Past Projects



2019 | 2018 | 2017 | 2016 | 2015 | 2014 | 2013 | 2012 | 2011 | 2010 | 2009 | 2008 | 2007 | 2006 | 2005 | 2004 | 2003 | 2002 | 


2018


Evolution of controllers for soft modular robots with proprioception

Juraj Korcek (IN)

The goal of this project is to build a neural network controller for a modular tensegrity robot with proprioception. The network will utilize inputs from a sensing system consisting of cable stretch sensors, which can be considered a simple form of proprioception. This should allow the robot to intelligently interact with the environment and navigate rough terrains more efficiently than with the current open loop system. The neural network will be designed and trained using evolutionary algorithms. Evolution of constant weights will be compared to evolution of Hebbian learning rules.

Type: Master project
Period: 05.09.2018 - 30.03.2019
Section(s): IN ME MT MX
Type of work: 20% theory, 70% software, 10% testing
Requirements: C++, python
Subject(s): tensegrity robots, sensory feedback, proprioception, neural network, evolution of learning rules, adaptive learning, soft robotics
Responsible(s): Davide Zappetti, Fabian Maximilian Schilling
Report: Click here
 

Building a Robotic Flapping Wing to Study Bird Flight

Adrien Paolini (ME)

At the Laboratory of Intelligent Systems, we develop drones with bio-inspired morphing air frames to achieve increased flight agility. To better understand bird’s wing morphology, wing motion and aerodynamics, we have teamed up with the Animal Flight Lab (AFL) at Lund University (SWE), and are looking for a highly motivated master student who likes to think outside the box. We aim to design and build a novel, artificial bird wing with flapping capabilities, which can later be used to perform extensive wind tunnel tests. The main phases of this master’s project can be distinguished as follows: Analysis of the anatomy of bird’s wings and its bio-mechanics | Study of literature regarding existing flapping wing systems | Design of a mechanism considering the requirements by AFL | Manufacturing of the wing and its mechanisms (mechanics and control) | Fitting and assembly in Lund (1-2 trips to Sweden expected).

Type: Master project
Period: 19.11.2018 - 16.03.2019
Section(s): ME MT MX
Type of work: 30% theory, 60% hardware, 10%+testing
Requirements: CAD+software
Subject(s): Mechanics, Aerodynamics, Biology, Control
Responsible(s): Enrico Ajanic, Prof. Christoffer Johansson
Report: Click here

A path planning algorithms for delivery drones flying in urban environments

Julien Di Tria (Microengineering)

At the Laboratory of Intelligent Systems (LIS), we are developing a human-friendly drone delivery system called Dronistics. The system is composed of the safe-foldable drone called PackDrone and a software to control and monitor drones in real-time.

The first goal of this project is to extend the previous work of a semester project that aimed at computing the best path for the drone at a specified altitude (2D). The improved algorithm should be able to compute a path from point A to point B in a 3-dimensional space. The computed path should be able to avoid buildings, elevated lands, forbidden areas and other drones whose position and flight path are already known to the algorithm.

The second part of the project focuses on dynamic path planning. The algorithm should take into consideration unexpected obstacles such as buildings, construction cranes, trees, and other drones (not registered in a database) and re-plan the new optimum collision free path. Detection of obstacles should be based on existing on the market solutions for obstacles avoidance.

Finally, the system should be rigorously tested, and all the unintended situations should be handled. Tests should be initially done in simulation and validated in a real environment with a physical drone and obstacles. Further, the developed software should be well documented, and its performance should be analyzed. Moreover, the algorithms for path planning and obstacle avoidance should be implemented in the web application that navigates and monitors Dronistics’ drones.

Type: Master project
Period: 15.10.2018 - 15.02.2019
Section(s): IN MT SC
Type of work: 10% theory, 60% implementation, 30% tests
Requirements: Control theory, C++, obstacles avoidance, path planning
Subject(s): Flying Robot, transportation of packages,
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here
URL: Click here
 

Automated Propeller Geometry Measurement from 3D Scanning Data

Pauline Maury-Laribière (MT)

This project is part of LIS’s collaboration with Maxon Motor which aims to bring about widespread adoption of professional drones for commercial applications by providing the market with high quality, efficient and reliable propulsion system components. The trend in the drone market towards high-end professional use requires designers to adopt more rigorous methods for performance analysis and sizing of propulsion systems. A major challenge lies in accurately estimating the aerodynamics of the propellers. Since it is often impractical to carry out wind tunnel tests, particularly with large propellers, the Blade Element Momentum Theory is used to model the prop’s aerodynamics. However, this method requires detailed knowledge of the propeller geometry which manufacturers do not typically publish. The goal of this project is to create a software tool which takes as inputs STL or ASC files which contain 3D coordinates of points on the propeller surface and outputs the following parameters: hub radius, tip radius, blade twist, chord length and aerofoil shape at discrete points along the radius. Subsequently the aerofoil shape will be used to determine the lift and drag characteristics at each radial position. If this task is successfully completed the scope of the project may be extended to include developing processes to further automate the 3D scanning of the propeller using examples of tools available on the market.

Type: Semester project
Period: 18.09.2018 - 02.02.2019
Section(s): EL IN MA ME MT PH
Type of work: 30% theory, 70% software
Requirements: Coding experience in Python or MATLAB
Subject(s): Aerodynamics, UAV Propulsion Systems, 3D Scanning, Automation
Responsible(s): Sebastian John Steffen, Anand Bhaskaran
Report: Click here

Bandwidth efficient object recognition for drone swarms

Marco Zoveralli (IN)

The project aims at developing a bandwidth efficient distributed object detection system which can be flown on a drone swarm. The system exploits the different points of view of the drones in the swarm to improve object recognition, while keeping the amount of data that is transmitted to the ground station as low as possible. In this way a more efficient use of the limited wireless resources can be achieved. The projects will involve the use of both off-the-shelf neural network algorithms and WiFi communication protocols. The first part of the project will focus on the setup of the communication network between the communication modules to be mounted on the drones and the ground station. The second part of the project will focus on the setup of the image capture/object recognition system and some basic onboard image processing. The core part of the project will consist in the implementation and optimization of the detection and communication protocol, which will build upon the modules developed so far.This part will include the evaluation of the system performance in terms of both bandwidth efficiency and detection accuracy.

Type: Semester project
Period: 18.09.2018 - 31.01.2019
Section(s): EL IN MT
Type of work: 50% software, 30% hardware, 20% testing
Requirements: Good knowledge of WiFi communication protocols (hands-on experience desirable), programming skills (shell scripting, Python, C/C++), familiarity with computer vision
Subject(s): Distributed sensing, drones communications, computer vision, swarm robotics
Responsible(s): Giuseppe Cocco, Fabian Maximilian Schilling
Report: Click here
 

Development and testing of a device to give haptic guidance based on the Hanger Reflex when flying a drone with upper body movements

Maxim Pavliv (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To do so, we created a wearable interface, called the FlyJacket, with which the user can control a drone with intuitive body movements. The control of the drone can even be enhanced by giving a haptic guidance to the user using tactile feedback. The goal of this project is to integrate a simple, not cumbersome and lightweight haptic guidance device into the FlyJacket based on the Hanger Reflex (https://www.youtube.com/watch?v=on22yoI40TI&t=7s, publication: Development of a Head Rotation Interface by Using Hanger Reflex, Sato et al.). During this project, the student will have to familiarized him/herself with the Hanger Reflex principle. Then, he/she will have to develop the hardware and the electronics of a lightweight and portable device to trigger the Hanger Reflex by applying force on the torso. This device will be integrated into the FlyJacket., Finally, the student will have to test its effectiveness by doing extensive test with a significant number of human subjects. This will be done by comparing performances of a flight task between subjects with the haptic guidance and a control group that doesn't have this feedback. This project require a highly motivated, independent and creative student. Indeed, a reliable device need to be implemented early in the project in order to have the time to perform tests on human subjects. We are expecting the student to come with his/her own ideas and always be one step ahead of the current state of the project.

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): ME MT
Type of work: Mechanical+workshop+experience +Solidworks +motivation
Requirements: 10%+theory +50%+hardware +40%+test+on+subjects
Subject(s): Mechanics +Haptics +Wearable +Actuator +Human+Machine+interaction
Responsible(s): Carine Rognon, Vivek Ramachandran
Report: Click here

Give the sense of touch to soft modular robots

Siqi Zheng (MA)

Soft modular robots are versatile systems that can be assembled into different task-specific morphologies to safely locomote and manipulate beside or cooperatively with humans or in un-constructed environments. Soft robots, in fact, can freely deform along any direction and comply with any unexpected or excessive external force. However, their soft bodies have unlimited degrees of freedom and it is very challenging to control their motions and the forces they apply to the environment. Indeed, a distributed sensory feedback to detect global and local deformations should be included in the practical design of soft robots to improve their controllability and include detections of loads from the environment. The objective of the semester project is to design and manufacture sensorized soft modules able to detect deformations. At first, the student will implement a technology available in our lab to sensorize the modules. Secondly, he will make a set-up for detecting deformations of the modules under different conditions and will characterize them. Finally, a demo application of a multi-module structure will be manufactured and characterized to show the potential of the results.

Type: Semester project
Period: 15.09.2018 - 30.01.2019
Section(s): MA ME MT
Type of work: 10% theory, 80% hardware, 10% software
Requirements: CAD (Inventor, SolidWorks or similar), good understanding of structures and materials, Arduino/matlab
Subject(s): soft robotics, bio-inspired robotics, stretchable sensors
Responsible(s): Davide Zappetti, Matteo Macchini
Report: Click here

Development of a bio-inspired robotic manipulator

Roc Arandes (MT)

Vertebrates have manipulation capabilities - in terms of dexterity and versatility - still unmatched by robotic applications. For this reason at laboratory of intelligent systems (LIS) we are focusing on extract bio-inspired principles to improve next generation of robots. In this proposed semester project, we seek to increase robot manipulation capabilities by developing a bio-inspired robotic modular manipulator. At first, the student will design and fabricate the manipulator based on an available design concept. He will use lightweight materials and 3D printing technologies. Secondly, a mechanism for grasping will be designed and manufactured. Finally, the performances of the actuated manipulator will be assessed in terms of workspace area, actuation forces and robustness.

Type: Semester project
Period: 04.06.2018 - 30.01.2019
Section(s): MA ME MT
Type of work: 20% theory 80% hardware
Requirements: CAD (Inventor, SolidWorks or similar), good understanding of mechanisms and materials
Subject(s): soft robotics, bio-inspired robotics, manipulation
Responsible(s): Davide Zappetti, Enrico Ajanic
Report: Click here

Development and Visualization of Neural Network in Robogen

Camilla Carta (CH)

RoboGen™ (www.robogen.org) is an open source platform for the co-evolution of robot bodies and brains (neural network). It features an evolution engine and a physics simulation engine. The goal of this project is to further develop the neural network framework in RoboGen to evolve more capable controllers for different robot morphologies. The first part of this project is to understand the working of RoboGen software and the theory behind recurrent neural networks. This will be followed by the development of an improved network architecture that will also include sensory-feedback modulated oscillator models to improve the robot’s capabilities. The second part of the project is to extend the RoboGen platform to introduce visualization of the neural controller of the robot. For this part, various open-source neural network visualization tools will be suggested to the student, who could also develop his own solution. The third part of this project is to further develop recurrent neural networks by adding various other state-of-the-art neuron models. Finally, the developed software should be tested, validated and theoretical software analysis (such as o-analysis) should be performed. This project will be done in collaboration between two labs: Biorobotics Laboratory (BioRob), and Laboratory of Intelligent Systems from Ecole Polytechnique de Lausanne.

Type: Semester project
Period: 15.09.2018 - 30.01.2019
Section(s): MT
Type of work: 20% Theory, 80% software
Requirements:
Subject(s): Evolutionary Algorithm, Neural Networks, Modular robots
Responsible(s): Anand Bhaskaran, Davide Zappetti, Shravan Tata Ramalingasetty, Jonathan Arreguit, Fabian Schilling
Report: Click here
URL: Click here

Control a Robot by means of an Adaptive Body Motion Decoder

Loic Niederhauser (CH)

Learning a teleoperation task normally requires a high cognitive effort as well as extensive training. In the frame of the Symbiotic Drone project, we are designing a new Human-Robot Interface (HRI) based on wearable technology in order to ease the process of learning a new interface as well as the dynamics of the controlled machine. This interface needs to adapt to one’s control style in order to grant responsiveness and error correction. In this preliminary stage, the student will design a variable structure with gain-adaptive Single-Input Single-Output (SISO) interface and test in VR for a drone teleoperation task. Subsequently, this system will be adapted through machine learning to the control of a robot by means of body motion. Students interested in neural networks, HRI as well as VR programming are encouraged to apply.

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): IN MT
Type of work: 30% theory, 70% experimental
Requirements: Python or Matlab, C# basics, elementary knowledge of control systems
Subject(s): Error feedback, teleoperation, adaptive control
Responsible(s): Matteo Macchini, Sebastian John Steffen
Report: Click here

Development and test of a human motion capture system

Victor Camille Baptiste Faraut (CH)

Wearable sensors are a relatively young technology which has been object of great interest for several industrial, clinical and research applications in the past years. One of the main limitations in this field is the lack of a complete and versatile environment for the interface of such devices. In the LIS we are planning to move a first step towards a unified framework for wearable technology, to allow the user to handle in a simple and rapid way the interface with wearable systems of different nature from multiple suppliers. Currently, a hardware infrastructure has been proposed and designed consisting in modular parts with plug-and-play functionalities: the next step is to implement a real application. For this project, you will test the board (a BeagleBone Green Wireless) in different configurations and estimate time requirements for several communication and processing tasks, including acquisition from different sensors, signal processing, multiplexing. Later, you will use the same infrastructure to develop an acquisition system from a set of IMUs, to track a the motion of a human, and validate this system in camparison with a commercial IMU-based motion capture system and a IR camera-based one. Students interested in embedded software and processing and wearable technology are encouraged to apply.

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): EL MT PH
Type of work: 30% theory, 20% hardware, 50% experimental
Requirements: embedded software, real-time systems
Subject(s): real-time capabilities assessment, hardware
Responsible(s): Matteo Macchini, Olexandr Gudozhnik
Report: Click here

Haptic display for a wearable Human-Robot Interface

Hugo Kohli (CH)

Haptic is becoming a prominent area in the field of human-robot interfaces. Force feedback and vibrating haptic devices are capable of displaying the presence of obstacles and other physical objects and forces, as well as providing info on robot status and dynamics. In the LIS we are developing a wearable interface finalised to the control of a flying robot. For this application, we are going to adopt a shared control architecture to support human teleoperation in case of difficulties due to the operator cognitive state and challenging environmental conditions. Shared control can be a misleading tool in that the robot’s behaviour does not always reflect user’s inputs. The goal of this project is to design, prototype and test a haptic display which will be installed on the user’s body to inform them about the current shared control state. We want to investigate amplitude and direction resolution for the device consisting in an array of vibrotactile motors. A good plus would be the integration of such device in the current wearable system. Students interested in haptics, wearable hardware design and validation are encouraged to apply.

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): EL IN ME MT
Type of work: 30% theory, 50% hardware, 20% testing
Requirements: Hardware interfacing, embedded software
Subject(s): Haptic feedback, wearable technology
Responsible(s): Matteo Macchini, Carine Rognon
Report: Click here

Wearable Technology : Integration of hardware and UI for a new framework

Cyrill Florin Lippuner (MT)

Wearable sensors have become very popular in many applications such as medical, entertainment, security, and commercial fields. In the LIS, our goal is to develop a unified framework to allow the user to easily design and prototype experiments and application exploiting this kind of technology. So far, several efforts have been done in this sense on the software, hardware and interfacing sides. Now, we want to integrate the currently available advancements in a first integrated version of the framework. - Integration of the current hardware and user interfaces in a single GUI for system prototyping. - Implementation of a new UI for hardware diagnostics and control Students interested in wearable technology implementation and software development are encouraged to apply.

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): EL IN MT PH
Type of work: 20% theory, 20% hardware, 60% software
Requirements: python, basics of embedded software coding
Subject(s): Hardware and Software integration, UI design
Responsible(s): Matteo Macchini, Olexandr Gudozhnik
Report: Click here

Implementation of a novel connection strategy for soft modular robots

Ludovic Sébastien Pierre Coullery (MT)

Soft modular robots are versatile systems that can be assembled into different task-specific morphologies to safely locomote and manipulate beside or cooperatively with humans or in un-constructed environments. Soft robots, in fact, can freely deform along any direction and comply with any unexpected or excessive external force. Consequently, it is very challenging to design a connection strategy for such modules that should also exchange power and information. Indeed, a novel connector able to mechanically connect the deformable modules and, at the same time, allow exchange of energy and information should be included in the practical design of such soft robot. The objective of the semester project is to design and manufacture such new connector based on a novel connection strategy. At first, the student will familiarize with the current soft modular robot and the connection strategy developed in our lab. Secondly, he will design and manufacture a new connector and will characterize it in terms of mechanical robustness, electrical conductivity and quality of the data transmission. Finally, a demo application of a multi-module structure will be manufactured and characterized to show potential applications.

Type: Semester project
Period: 15.09.2018 - 30.01.2019
Section(s): EL MA ME MT MX
Type of work: 10% theory, 70% hardware, 20% testing
Requirements: good understanding of electrical/mechanical engineering, CAD technologies a plus
Subject(s): soft modular robotics, power and data transmission, bio-inspired robotics
Responsible(s): Davide Zappetti, Anand Bhaskaran, Olexandr Gudozhnik
Report: Click here

Towards solar panels cleaning through drones

Michael Perret (MT)

Dirt on the surface of solar panels (SPs) represents a significant factor in drops of their efficiency [1, 2]. Cleaning SPs can be expensive, hazardous and often tricky. This project aims at providing a first step towards the solution of this problem. In the following, some details and phases of this project are given. BLOCK 1: INTEGRATION OF THE DRONE ARCHITECTURE (SW/HW) In this block, the student must quickly familiarize with the literature related to the general topic of the project. After that, the drone architecture has to be built. It consists of assembling drone hardware (e.g., Pixhawk flight controller [5], motors, ESCs, power distribution board, propellers and frame) with an ODROID XU4 [6] and an onboard camera (e.g., OpenMV Cam M7 [7]). It will be fundamental the understanding of the PX4 software [9, 10]. Once the drone is assembled it will be the turn of understanding how to create an object in the Motion Capture System (MoCap) environment (DroneDome). Subsequently, the student should familiarize with the ROS concepts. ROS [8] is a fundamental tool for many robotics projects and it will be used, in this project, to link the flight controller to the MoCap and the laptop/pc used by the student. The drone should be able to autonomously go from a point A to a point B in the DroneDome. BLOCK 2: DETECT AND LAND ON A SOLAR PANEL In this block, the student should understand the basic concepts of visual servoing. These will be used to detect a solar-panel-like structure placed in the DroneDome. This task will be achieved by placing fiducial markers (e.g., Apriltags [3, 4]) on the SP and, without relying on the MoCap, landing at the center of the SP. The drone should be able to autonomously go from a point A to a point B in the DroneDome, find a SP by detecting the visual fiducials and land on it. BLOCK 3: INTELLIGENT DETECTION AND DOWNWASH CLEANING This block can be split into two main parts: • instead of directly landing on the SP, the drone will plan and execute an intelligent trajectory which, thanks to the downwash effect, will give an initial cleaning of debris • some basic edge-detection technique [11] will be investigated and applied to detect a SP without the aid of fiducial markers. The presented approach represents a first fundamental step towards the solution of the complex task of cleaning solar panels. Note also that this project might also be a good starting point for future master thesis projects. ADDITIONAL INFORMATION The student is expected to document his/her work throughout the semester project carefully. Moreover, as usual, he/she will have to do a mid-term and final presentation of his/her work. A final report also has to be provided. REFERENCES [1] M. R. Maghami, H. Hizam, C. Gomes, M. A. Radzi, M. I. Rezadad, and S. Hajighorbani, “Power loss due to soiling on solar panel: A review ” Renew. Sustain. Energy Rev., vol. 59, pp. 1307–1316, 2016. [2] M. H. Bergin, C. Ghoroi, D. Dixit, J. J. Schauer, and D. T. Shindell, “Large Reductions in Solar Energy Production Due to Dust and Particulate Air Pollution ” Environ. Sci. Technol. Lett., vol. 4, no. 8, pp. 339–344, 2017. [3] E. Olson, “AprilTag: A robust and flexible visual fiducial system ” Proc. - IEEE Int. Conf. Robot. Autom., pp. 3400–3407, 2011. [4] J. Wang and E. Olson, “AprilTag 2: Efficient and robust fiducial detection ” Proc. IEEE/RSJ Int. Conf. Intell. Robot. Syst., pp. 2–7. 2016. [5] Pixracer - https://docs.px4.io/en/flight_controller/pixracer.html [6] ODROID XU4 - https://www.hardkernel.com/main/products/prdt_info.php?g_code=G143452239825 [7] OPENMV Cam M7 - https://openmv.io/products/openmv-cam-m7 [8] ROS - http://wiki.ros.org/ROS/Introduction [9] PX4 user guide - https://docs.px4.io/en/ [10] PX4 developer guide - https://dev.px4.io/en/ [11] Canny edge detection OpenCV - https://docs.opencv.org/3.1.0/da/d22/tutorial_py_canny.html

Type: Semester project
Period: 18.09.2018 - 30.01.2019
Section(s): MT
Type of work: 5%+state+of+the+art +40%+hardware +40%+software +15%+misc
Requirements: Basic+aerial+vehicle+knowledge+and+programming
Subject(s): Control +hardware +drone +MoCap
Responsible(s): Fabrizio Schiano, Fabian Maximilian Schilling, Anand Bhaskaran
Report: Click here

Development of the chassis and steering system for a long delivery drone

Damien Roger Richard Coulon (ME)

The Lake Victoria Challenge (LVC) has been created to address these problems and to promote drone transportation technologies in Africa. This semester project is made to develop a part of the flying platform which will take part in the Lake Victoria Challenge and will focus on the development of the chassis and steering system. Due to this environment with weak infrastructure (low runway quality and size), the chassis needs to be designed to allow take-off and landing on a very short and potentially rough runway all while keeping the payload safe. This chassis needs to connect the wing, the motor block, the wheels and the cargo bay of the drone. To assure the shortest possible take-off and landing against the wind and for precise good delivery, the drone must be maneuverable on the ground. A steering system has to be implemented to this end. As for the chassis, the steering system has to support rough runways and imprecisions at landing. Since the weight is an important factor, all the requirements should be met with minimum weight. The drone will be used for transportation in a region of Africa with many big water areas (e.g., lakes, rivers) that need to be overflown and a wet season. Thus, the steering system has to be waterproof (IPX7) to withstand emergency water landing and heavy rain. In parallel to these tasks, the risks before, during and after flight will be evaluated and gathered in the form of a risk assessment that will be integrated to the mandatory and necessary risk assessment documentation of the global project.

Type: Semester project
Period: 28.09.2018 - 30.01.2019
Section(s): ME
Type of work: 10% theory, 30% software, 40% hardware, 20% test
Requirements: CAD software
Subject(s): Mechanics, Conception, Structure
Responsible(s): Przemyslaw Kornatowski, Fabrizio Schiano
Report: Click here

Cargo bay, waterproof protection design and wind sensors for a UAV

Paul Megevand (ME)

The Lake Victoria Challenge (LVC) has been created to address these problems and to promote drone transportation technologies in Africa. This semester project is made to develop a part of the flying platform which will take part in the LVC and will focus on a way to transport packages, protect the drone against water and a way to measure wind speed. Firstly, as the main objective of the competition is to transport several 250 g packages defined by the LVC rules from A to B, the drone will have to integrate a cargo bay in which cargo pick-up will have to require a minimal intervention at the destination location. Due to the African environment with weak infrastructure (low runway quality and size), the cargo bay will also have to be designed in a way to absorb hard landings, bumps and drops. As the drone will be used for transportation in a region of Africa with many big water areas (lakes, rivers…) that need to be overflown and a wet season, a water crash must be anticipated. To do so, critical electronics (e.g. autopilot) will have to be protected against water and a specific design will have to be implemented on the drone so it will become waterproof (IPX7: immersion in 1m depth). In the event of a water crash landing, electronics have also to be protected against short-circuits. Thus, a system that mechanically unplugs the battery from the drone will have to be designed and integrated on the drone. However, to avoid any battery disconnection during flight, a special attention to the false positive case will have to be given. Finally, as the autopilot will have to know the precise wind speed during flight and the wind speed and direction during taxi before take-off, a wind measurement system will have to be designed and embedded on the drone. In parallel to these tasks, the risks before, during and after flight will be evaluated and gathered in the form of a risk assessment that will be integrated to the mandatory and necessary risk assessment documentation of the global project.

Type: Semester project
Period: 28.09.2018 - 30.01.2019
Section(s): ME
Type of work: 10% theory, 30% software, 40% hardware, 20% test
Requirements: CAD software
Subject(s): Mechanics, Conception, Structure
Responsible(s): Przemyslaw Kornatowski, Fabrizio Schiano
Report: Click here

Integration of flight telemetry, precise landing and ground navigation of a UAV

Louis Munier (MT)

The Lake Victoria Challenge (LVC) has been created to address these problems and to promote drone transportation technologies in Africa. This semester project is made to develop a part of the flying platform which will take part in the LVC and will focus the flight telemetry and analysis of the data, the landing, and autonomous ground navigation. A C2/3 telemetry link is a criterion required by the challenge and needed during all flights. For this purpose, existing hardware will be used to send all the sensor data to the ground base. All communication needs to be encrypted between our drone and the ground control station as required by the challenge. Regarding this, state-of-the-art encryption methods will be studied to find the best solution and adapt it to our application. An automatic diagnostic procedure will be implemented on the ground control station using the logs provided by our platform. The landing will take place in a 10x20 meters area, so it must be precise and short. It will be assured by using existing autopilot hardware and adapting it to have a smooth and short landing. Different existing landing approaches, including experimental ones such as deep stall landing, will be investigated and tested to shorten the necessary runway to a minimum. Since the cargo pick-up criteria are specified to require minimal human intervention, automatic land navigation from the runway to the pick-up location will be designed. The navigation to the take-off will be done with the same algorithm. In parallel to these tasks, the risks before, during and after flight will be evaluated and gathered in the form of a risk assessment that will be integrated to the mandatory and necessary risk assessment documentation of the global project.

Type: Semester project
Period: 28.09.2018 - 30.01.2019
Section(s):
Type of work: 30% theory/state of the art, 50% software, 20% flying test
Requirements: Basic aerial vehicle and avionics knowledge and programming
Subject(s): Control, autonomous landing and ground navigation
Responsible(s): Fabrizio Schiano, Przemyslaw Kornatowski
Report: Click here
URL: Click here

Flight control and telemetry integration for short take-off of a UAV

Joël Zbinden (ME)

The Lake Victoria Challenge (LVC) has been created to address these problems and to promote drone transportation technologies in Africa. This semester project is made to develop a part of a flying platform which could take part in the LVC and will focus on the implementation and programming of the flight controller and telemetry hardware. The LVC demands an autonomous and easy-to-use system. Therefore, the selected platform will need to be equipped with a compatible flight control unit which will contain all the necessary sensors for speed measurement, consumption control and reliable positioning in space (e.g. GPS receiver). These components will be integrated with the provided chassis. A suitable autopilot will be selected among the ones present on the market. Then, it will be programmed, calibrated and tuned to satisfy our needs. Since the runway will have limited dimensions (e. g. 10m x 20m), different autonomous take-off sequences will be investigated and tested to assure stable and effective take-off. For the necessary and mandatory observability of the status of the drone and recoverability in case of a crash landing on beyond visual line of sight (BVLOS) flights, telemetry hardware with enough coverage and range will need to be embedded on the drone. In parallel to these tasks, the risks before, during and after flight will be evaluated and gathered in the form of a risk assessment that will be integrated to the mandatory and necessary risk assessment documentation of the global project.

Type: Semester project
Period: 28.09.2018 - 16.01.2019
Section(s):
Type of work: 10% theory and state of the art, 40% software, 20% hardware, 30% test
Requirements: Basic aerial vehicle and avionics knowledge and programming
Subject(s): Control, autonomous flight, aerial robotics
Responsible(s): Fabrizio Schiano, Przemyslaw Kornatowski
Report: Click here

Comparative Study of Coaxial and “Flat” Multicopter Propeller Configurations

Marjorie Marie Joséphine Lasson (MT)

In the future drones will become a ubiquitous feature of urban airspaces, carrying out deliveries, aerial inspections and perhaps even transporting people. If they are to coexist with us near our homes and places of work we will require drones to become safer, quieter and more efficient. The quadcopter is currently the best known and most common rotor configuration but it has a fatal flaw: losing a single rotor makes the craft uncontrollable under most conventional control algorithms. Hence, we are interested in studying systems with six or eight propellers that could continue performing -albeit with a limited flight envelope-, in the event of rotor damage. The project will compare flat hexa and octo-copters to an “X8” configuration with 4 arms, each holding two coaxial propellers. The project has two primary goals. The first is to experimentally investigate various coaxial configurations including, but not necessarily limited to different combinations of blade diameters and pitches, different numbers of blades per propeller and using different rotational speeds for the top and bottom props. Some of the performance metrics to be considered include the specific thrust in terms of force per unit of power, heating of the motors and ESCs, and the level of noise produced by the system. The second goal is then to use these same metrics to compare the better performing coaxial configuration(s) to flat 6 and 8 rotor systems that achieve the same level of thrust. The key tasks for this project include the design and manufacturing of a rig for attaching the propellers to our thrust stand, carrying out the experiments on various propeller configurations, and processing and analysis of the data.

Type: Semester project
Period: 17.09.2018 - 14.01.2019
Section(s): GR ME MT
Type of work: 60% Testing, 25% Hardware, 10% Software, 5% Theory
Requirements: Basic Aerodynamics, Coding for automating experiments and data acquisition
Subject(s): Aerodynamics, UAV Propulsion Systems, Experimental Methods
Responsible(s): Sebastian John Steffen, Przemyslaw Kornatowski
Report: Click here

Development and study of ailerons vs. folding wings

Didier Negretto (ME)

We aim to improve the roll effectiveness of a highly maneuverable feathered drone with folding wings capability. To operate beyond the convectional flight envelope which includes flying at extended angles of attack and low velocities, sufficient roll control is required to allow effective maneuvering in cluttered environment. This can be achieved through the methodical application of properly sized ailerons and folding wings. The goal of the project is to design and develop ailerons for the feathered drone and subsequently, carry out a comparative analysis of roll effectiveness between the ailerons and the folding wings using the wind tunnel.

Type: Semester project
Period: 18.09.2018 - 11.01.2019
Section(s): Robotics Microengineering Mechanical Engineering
Type of work: Theory 10% hardware 60% testing 30%
Requirements:
Subject(s): Aerodynamics, Flight mechanics
Responsible(s): Enrico Ajanic, Mir Feroskhan
Report: Click here
URL: Click here

Design and development of morphing tail

Rayan Mouwafak (ME)

The pursuit of equipping drones with advanced flight capabilities such as aggressive maneuvering and perching calls for a more bio-inspired design and function of the tail. The aim of this project would be to develop an aircraft tail with artificial feathers and characterize its aerodynamics in the wind tunnel. By improving the current tail and feather design, the student is tasked to build small tail prototypes using 3D-printers and our laser cutter and subsequently compare their aerodynamic characteristics. To find out more about the project, please contact the supervisor stated below.

Type: Semester project
Period: 18.09.2018 - 11.01.2019
Section(s): Robotics Microengineering Mechanical Engineering
Type of work: Theory 10% hardware 60% testing 30%
Requirements:
Subject(s): Aerodynamics, Flight mechanics
Responsible(s): Enrico Ajanic, Mir Feroskhan
Report: Click here
URL: Click here

Brake thruster for precision landing

Hugo Bordog (ME)

As multi-rotor platforms offer low aerodynamic efficiency and limited range, the concept of equipping fixed wing drones with precision landing capabilities is an intriguing one. Although vertical take-off and landing (VTOL) techniques such as the tilt-rotor and tilt-wing methods have been researched widely, most mechanisms tend to be complex with high energy requirements and involve highly nonlinear dynamics. Therefore, the project aims to develop dedicated high impulse thruster that can be deployed on the aircraft’s fuselage to bleed off the remaining velocity during a high angle of attack landing. To find out more about the project, please contact the supervisor stated below.

Type: Semester project
Period: 18.09.2018 - 11.01.2019
Section(s): Robotics Microengineering Mechanical Engineering
Type of work: Theory 10% hardware 60% testing 30%
Requirements:
Subject(s): Mechanics
Responsible(s): Enrico Ajanic, Mir Feroskhan
Report: Click here

Omnidirectional vision for drones

Robin Wütschert (IN)

Omnidirectional imaging is a hot research topic and has important applications to aerial robots equipped with multiple cameras. In addition to numerous use cases in cinematography and film-making, omnidirectional vision can enhance drone autonomy by providing 360-degree visual sensing for tasks such as collision avoidance. The present project aims at implementing and comparing efficient stitching algorithms for omnidirectional vision, possibly enhancing existing solutions. Since the algorithm is to be run on the drone onboard computer and potentially used for real-time applications such as collision avoidance, it should have low complexity and must be able to run at an acceptable speed. The first part of the project will involve the selection and implementation of suitable stitching algorithms. Existing algorithms may be enhanced by taking the a priori information such as the camera configuration into account. The second part of the project will involve testing of the proposed algorithms on a physical drone platform.

Type: Semester project
Period: 18.09.2018 - 31.12.2018
Section(s): EL IN MT
Type of work: 40% software, 30% hardware, 20% testing, 10% theory
Requirements: Programming skills (shell scripting, Python, C/C++), image processing
Subject(s): Omnidirectional vision, computer vision, multi-camera, aerial robotics, image processing
Responsible(s): Fabian Maximilian Schilling, Giuseppe Cocco
Report: Click here

Control interface for drone swarms

Mahmoud Zgolli (MT)

At the Laboratory of Intelligent Systems, we are developing algorithms for controlling drone swarms. However, controlling several agents at the same time is not trivial and the need for an easy-to-use interface for swarm control arises. The goal of the project is to develop an intuitive interface for controlling a swarm of drones using only a single control mechanism (via gesture recognition, leap motion, remote controller, joystick, etc.). Existing flocking algorithms already allow the agents to avoid collisions with each other and to stay cohesive as a group. The control interface you develop on top should allow a user to steer the swarm as a single cohesive unit and it should be as intuitive as controlling a single stabilized drone. The project will involve the selection of a control mechanism, the design, and implementation of a control algorithm that fulfills the above requirements, followed by extensive testing thereof in simulation. Optionally, the algorithm can be validated in a state-of-the art motion tracking hall with real drones.

Type: Semester project
Period: 18.09.2018 - 31.12.2018
Section(s): EL IN MT
Type of work: 60% software, 20% testing, 20% theory
Requirements: Programming skills (Python, C++, Matlab), previous knowledge of ROS and the PX4 autopilot is a plus.
Subject(s): Swarm robotics, multi-agent control, human-robot interaction
Responsible(s): Fabian Maximilian Schilling, Enrica Soria
Report: Click here

Swarming algorithms for quadcopters

Yoann Yves Lapijover (MT)

The state of the art of swarming algorithms is rich. Many of these algorithms are inspired by animal behaviours such as ants, bees and birds and they are designed to visually resemble to them. The most evident example is the Reynolds flocking model [1], developed by the homonymous author to simulate flocks of birds in computer games. DEVELOPMENT AND TESTING OF AN OPTIMAL DECENTRALIZED SWARMING ALGORITHM When considering drone applications, Olfati-Saber model well suits the needs to keep the robots separated by a known constant inter-agent distance and make them navigate in a preferential and common direction [2]. Furthermore, in realistic scenarios taking into account optimal control helps improving the performances of the flock [3]. The outcome of this first work-package should be the implementation of an optimal swarming algorithm. SWARM DEPLOYMENT IN SIMULATION By integrating an existing simulator (ideally Gazebo [4], [5]), you will deploy your swarm in a search and rescue mission, where the dense group will disperse to explore an area and then become compact again when the goal is reached. The outcome of this second section should be a simulated demo of the mission. ROBUSTNESS OF THE SWARM IN STRESSFUL CONDITIONS The performances of a flock are highly sensible to the failure of some individuals, to the physical constraints of the sensors (for instance limited field of view) and to real world perturbations (like wind). The aim of the last section of the project is to test the robustness of the algorithm under deliberately stressful conditions, like the limitation of the agent’s visual capabilities (cameras with limited field of view) or the arrival of external intruders. The outcome of this work-package is the implementation of one scenario at your choice among the ones cited before and the analysis of the flock performance. You will develop the code in Matlab, Python and/or ROS/Gazebo. Good programming skills are required. References: [1] C. W. Reynolds, “Flocks, Herds, and Schools: A Distributed Behavioral Model ” p. 21, 1987. [2] R. Olfati-Saber, “Flocking for Multi-Agent Dynamic Systems: Algorithms and Theory ” IEEE Transactions on Automatic Control, vol. 51, no. 3, pp. 401–420, Mar. 2006. [3] O. Saif, “Reactive navigation of a fleet of drones in interaction ” 2016. [4] http://gazebosim.org/ [5] http://www.ros.org/

Type: Semester project
Period: 18.09.2018 - 21.12.2018
Section(s): IN MT SC
Type of work: 20% theory, 60% programming, 20% testing
Requirements: Programming skills (Python, MATLAB, C++), design of experiments, familiarity with autopilots like PX4 and simulator like Gazebo would be a plus.
Subject(s): Swarm robotics, multi-agent control
Responsible(s): Enrica Soria, Fabian Maximilian Schilling
Report: Click here

Development of a dynamics simulator for swarms of quadcopters

Victor Pierre Guy Delafontaine (MT)

At the LIS we are currently working on swarming algorithms for quadcopters. In this regard, reproducing the dynamics of a drone in simulation with fidelity is essential to make multiple drones fly together in reality. Drones simulators are effective tools to test the behaviour of a platform without taking the risk of damaging real hardware and being capable of running hundreds of experiments at a click of a mouse. In the framework of this project, you will gain hands-on experience in several key aspects of quadcopter control and navigation, as well as a good knowledge of decentralized drone flocking. ENHANCEMENT OF A QUADCOPTER SIMULATOR By integrating an existing simulator developed in Matlab, you will enhance the low level control strategy, ideally robust control (to take perturbations into account [1]), and implement an optimal path following algorithm. For this, you will make use of the already implemented physical and graphical packages. The outcome of this work-package should be a simulated demo of the quadcopter following the input commands in position and velocity. GENERALIZATION TO MULTIPLE QUADCOPTERS The generalization of the simulator to multiple drones will make it possible to fly a small fleet of drones at the same time. Multiple instances of the same quadcopter should be generated to run in parallel. The paradigm of object oriented programming (in Matlab) may be helpful to this aim. The outcome of this work-package should be a software allowing to simulate the dynamics of a variable number of drones. DECENTRALIZED SWARMING ALGORITHM To simulate a decentralized swarm of drones, an algorithm of flocking has to be integrated to the previous work. A decentralized flocking algorithm that suits the needs to keep the robots separated by a known constant distance and to make them move in a preferential direction is Olfati-Saber [2] model. Building up on the Olfati Saber algorithm, LQR controllers can improve the performance of the swarm [3]. The outcome of this section should be a simulated demo of a flock navigating according to a decentralized algorithm. The implementation of a simple interface could help to set the parameters of the flock. The code will be developed in Matlab/Simulink. Previous knowledge on control and estimation theory are required. References [1] C. MASSÉ, O. GOUGEON, D. NGUYEN, and D. SAUSSIÉ, “Modeling and Control of a Quadcopter Flying in a Wind Field: A Comparison Between LQR and Structured ??Control Techniques ” in 2018 International Conference on Unmanned Aircraft Systems (ICUAS), 2018, pp. 1408–1417. [2] R. Olfati-Saber, “Flocking for Multi-Agent Dynamic Systems: Algorithms and Theory ” IEEE Transactions on Automatic Control, vol. 51, no. 3, pp. 401–420, Mar. 2006. [3], O. Saif, “Reactive navigation of a fleet of drones in interaction ” 2016.

Type: Semester project
Period: 18.09.2018 - 21.12.2018
Section(s): EL MT SC
Type of work: 30% theory, 50% software, 20% testing
Requirements: Programming skills (MATLAB, Simulink, Python), control theory, familiarity with quadcopters would be a plus.
Subject(s): Quadcopter control, navigation, guidance
Responsible(s): Enrica Soria, Julien Lecoeur, Schiano Fabrizio
Report: Click here

Middleware for Transportation Drone: Communication with the drone

Anton Hào-Chen Lu (MA)

At the Laboratory of Intelligent Systems (LIS) we are developing drones for a last-cm delivery. These delivery drones are fully autonomous and are controlled in the real-time and monitored by the Dronistics’ web-application framework. The main goal of this project is to design a communication software between the autopilot of the drone (PX4) and Web-API. The first step of this project is to analyse various existing frameworks (such as Drone/Core, Dronekit, FastRTPS, etc.) that enables reliable communication with the autopilot. The selected framework for the communication software should be implemented and validated with the simulated version of the autopilot. Additionally, the software should be capable of capturing in the real-time drone’s parameters (such as battery status, system status, GPS location). Secondly, the theoretical analysis of the implemented software (using Big O notation) should be performed and efficiency parameters such as computation time and memory usage should be optimised. Additionally, developed software should be integrated into two types of companion computers: Odroid and Snapdragon (both with Linux OS), and the performance metrics should be analysed. Thirdly, the communication software developed should be capable of reprogramming the drone autopilot for different morphologies and should also be capable of modifying the gain parameters Finally, the monitoring and controlling endpoints of the communication software should be well documented and every endpoint should be tested independently. Thus the final goal of this project is to develop an integration-ready software that can be combined with the Web-API to control and monitor the drone. The diagram on Figure 1 explains the concept of the described above software framework.

Type: Semester project
Period: 19.02.2018 - 30.06.2018
Section(s): MT
Type of work: 20% theory, 50%+software, 30%+hardware
Requirements:
Subject(s): Basic understanding of programming and software architecture
Responsible(s): Anand Bhaskaran, Przemyslaw Kornatowski
Report: Click here

Middleware software design for transportation drones

Elisabet Arvidsson (MA)

At the Laboratory of Intelligent Systems (LIS) at École Polytechnique Fédérale de Lausanne (EPFL) we are developing drones for last-cm delivery. These delivery drones are fully autonomous. They are controlled and monitored in the real-time by the Dronistics’, web-application framework. The main goal of this project is to design a Web-API that runs on the companion computer of the drone. The Web-API should accept high-level messages from the external application and perform the corresponding operations on the autopilot. Additionally, the Web API should follow publish-subscribe architecture. Thus the WebAPI should notify the subscribed applications on specific changes in drone parameters (such as a battery, system failure, GPS change, etc. ) The first step of this project is to analyse various existing protocols and open-source REST (Representational State Transfer) compliant Web-API framework for the implementation. During this analysis, comparative study on reliability, security and bandwidth requirement should be performed. Then, the chosen Web-API framework should be implemented. However, the student can use pre-defined values for the drone-parameters. Secondly, a simple web-based application should be developed for a remote server that can reliably control and monitor the drone parameters to validate the WebAPI developed in the first step. Then, the developed Web-API should be ported to different companion computers (Raspberry Pi Zero, Odroid, etc.) and bottleneck analysis should be performed. Finally, the software should be integrated with the real-time drone parameters. Thus, the web application (developed in the step 2) along with the Web-API should be capable of monitoring and controlling the simulated drone reliably. The diagram on Figure 1 explains the concept of the described above software framework.

Type: Semester project
Period: 19.02.2018 - 30.06.2018
Section(s): IN MT SC
Type of work: 80%software, 20%testing
Requirements: Programming
Subject(s): Software Architecture IoT
Responsible(s): Anand Bhaskaran, Przemyslaw Kornatowski
Report: Click here
URL: Click here

Improvements and new manufacturing methods of the PackDrone drone

Nathanaël Yvon Ferraroli-Tissot-Daguette (MT)

At the LIS, we are developing a new type of a safe and foldable delivery drone called the “PackDrone”. The features of the drone can be find here: https://dronistics.epfl.ch/EPFL/ Existing cage: The cage of the drone is manufactured with strong and lightweight pultruded carbon-fibre tubes connected by flexible 3D printed joints that allow folding of the drone. However, pultruded carbon fibre tubes have limited energy absorption. Thus they break during strong collisions with obstacles. Moreover, the broken carbon parts are dangerous for people. Flexible joints are 3D printed which limits the possibility to produce a bigger number of them in a short time. Additionally, tubes and joints are assembled and bonded manually which is a time-consuming process. The main goal of this project is to find the best available materials and new techniques of manufacturing, to create a stronger light-weight cage in a shorter amount of time. Secondly, the selection of materials must be validated with various lab-experiments. In the end, the prototype of the drone should be built with proposed manufacturing method, and its performance should be measured. The first task of the project is to find materials that could absorb more energy and could be safe after breaking. This part also requires identifying a new method of manufacturing of curved carbon rods to be used in the cage. The second task is to find new methods of manufacturing of flexible joints, e.g. injection moulding or overmoulding. For this task redesign of the joint might be required. Thirdly, the new method of assembling of the cage should be developed to reduce the manufacturing time of the cage. Additionally, the proposed new method should also enable easy replacement of the damaged parts. Finally, the new CAD design of the drone should be done considering the new materials and manufacturing methods. Any improvements/modifications to the other parts of the drone (quadcopter) should be made if required. In the end, the prototype of the drone should be built and tested. If time allows a smaller version of the cage could be designed and built.

Type: Semester project
Period: 19.02.2018 - 30.06.2018
Section(s): MT MX
Type of work: 20% theory, 10% software, 70% hardware
Requirements: CAD design, carbon manufacturing, moulding methods
Subject(s): Flying Robot - drone, transportation of packages, manufacturing, mechanical design
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here
URL: Click here

Adaptive Estimation of Multirotor Geometry and Inertia

Louis-Dominique Renaud (MT)

At the LIS, we are developing a new type of a safe cargo delivery drone. The drone is inside a cage that can be folded to be safe while flying close to people, and to protect the package. However, transportation of heavy packages might be dangerous if the drone cannot adapt to packages with unknown mass. Furthermore, the drone should be able to adapt to misalignment of rotors that may result from an improper unfolding of the drone.

Indeed, drone autopilots need to have knowledge of the geometry and inertia of a multirotor in order to transform roll pitch and yaw commands into motor commands. This is usually performed using a mixer matrix that is generated according to pre-defined geometries. Using a mixer matrix that does not correspond to the actual geometry of the multirotor may lead to controller instability and crash.

The goal of this project is to allow the drone to estimate its geometry and inertia while flying. The first task of this project will consist in a review of the existing methods, followed by implementation in the PX4 autopilot. The second task is to validate the algorithm. Flight tests will be performed on a multirotor with modified geometry (for example by displacing one of the motors) and inertia (for example by adding a mass). The student will characterize the accuracy of the estimation, and the improvements in controller stability.

Type: Semester project
Period: 09.02.2018 - 30.06.2018
Section(s): IN MA MT PH SC
Type of work: 10% theory, 40% software, 10% hardware, 40% experiments
Requirements: Good programming level in C++, data analysis in python. Basics in linear algebra
Subject(s): drone, parameter estimation, mixing
Responsible(s): Julien Lecoeur, Przemyslaw Kornatowski
Report: Click here

A swapping and recharging battery ground station for a delivery drone

Hannes Kaspar Rovina (MT)

At the LIS, we are developing drones for fast transportation of lightweight packages (up to 2 kg). Although the drones can quickly deliver goods at a cheaper cost, the range of delivery is still limited. Inefficient batteries are one of the main reason for this limitation. This project aims at eliminating these traditional limitations by the development of an automated battery-swapping mechanism, which could be used by drones during long-range deliveries. The objective of this project is to develop a ground station for the “PackDrone”. The first goal is to design a mechanism that enables quick battery replacement. The second goal is to implement all required sensors that help in precise landing and orientation of the drone during landing. Finally, the prototype of the ground station should be designed, build and validated with the real drone. The first task of this project is to redesign holder enclosing batteries of the drone. The design should ensure the ease of battery swapping (it may require finding the different shape of batteries). Precise orientation of the drone on the ground station is an important constraint for automatic battery replacement mechanism. Thus the second task is to design a mechanical system that helps to precisely position the drone while landing in the ground station. Additionally, hardware integration of existing supporting system for precise landing with the ground station is required, e.g. IR beacon, vision-based landing or RTK GPS. Thirdly, a mechanism to automatically swap batteries should be designed. The ground station should also have the capabilities to charge used batteries for further use. The ground station should be weatherproof and have integrated weather sensors such as anemometer and thermometer. Finally, the ground station should be extensively tested at the different weather and lighting conditions and its operational reliability should be validated.

Type: Semester project
Period: 19.02.2018 - 30.06.2018
Section(s): MT
Type of work: 20% theory, 20% software, 60% hardware
Requirements: CAD design, manufacturing
Subject(s): Flying Robot - drone, transportation of packages, ground station
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here
URL: Click here

Incremental Non-linear Dynamic Inversion (INDI) control

Lombardo Julien Maxime (MT)

At the LIS, we are developing a new type of a safe cargo delivery drone. The drone has shielded propellers to be safe while flying close to people. However, transportation of heavy packages might be very dangerous while a drone is flying during strong winds. Especially, strong wind gust can destabilise the platform, cause fall from the sky thus injure people and damage property.

The goal of this project is to adapt existing control algorithm for quadcopters to improve resistance to wind changes while flying with packages of different size and weight delivered by a drone developed at the LIS.

The first task of this project is the implementation of the Incremental non-linear dynamic inversion controller (INDI) (existing implementation in Paparazzi) to PX4 autopilot to keep the platform stable during sudden changes of wind.

The second task is to validate the controller resistance to the wind and characterise adaptation to varying inertia of the drone due to changes of the payload. Flight tests with Bebop 2 (Parrot’s drone) and/or the PackDrone (delivery drone developed in the LIS) should be done.

Type: Semester project
Period: 19.02.2018 - 30.06.2018
Section(s): IN MT SC
Type of work: 10% theory, 60% implementation, 30% tests
Requirements: Control theory, C++
Subject(s): Flying Robot, transportation of packages
Responsible(s): Julien Lecoeur, Przemyslaw Kornatowski
Report: Click here

Improvements in safety of a protective cage for a delivery drone

Jan Niklas Benzing (MT)

At the LIS, we are developing a new type of a safe delivery drone. The drone is designed to transport lightweight items such as medicaments, dressings for wounds, small equipment over short distances. This drone has an all-around protective cage, which separates propellers from the environment thereby protecting the drone and people. Especially, while delivering goods directly to peoples hands. However, the low density structure of the cage allows fingers to reach the inside of the cage easily. Thus fingers can be injured by the fast rotating propellers.

The main goal of this project is to redesign and build a protective cage for an existing in the LIS delivery drone. The cage should have a structure dense enough to separate fingers of an adult person from the propellers. In the same time, the structure should minimally disturb the airflow which allows the drone to fly.

The first task of this project is to find lightweight materials, fabrication techniques, and mechanical structures that could be used to improve safety of the existing design of the cage. The selection of the best structure should be based on drag calculations verified by experiments (test bench with precise force sensor will be provided).

The second task is to design and build the cage using selected materials and techniques of manufacturing. Implementation of the existing quadcopter in the cage will be required. In the end, the performance of the developed drone should be validated in the controlled environment using motion capture system.

Type: Semester project
Period: 01.03.2018 - 22.06.2018
Section(s): MT
Type of work: 20% theory, 10% software, 70% hardware
Requirements: CAD design, carbon manufacturing, moulding methods,
Subject(s): Flying Robot - drone, transportation of packages, manufacturing, mechanical design
Responsible(s): Przemyslaw Kornatowski, Mir Feroskhan
Report: Click here

Cellular Controlled Drone

Florian Kaufmann (CH)

The goal of this project is to prototype a drone that allow for its control logic to be hosted on a distributed cloud accessible via cellular connectivity. Furthermore, all communication needs will have to be implemented via cellular connectivity. Beyond the prototyping, the student will be asked to explore and measure how factors such as latency, range and throughput affect the ability to control and steer drones. Of particular interest is the placement of control function (on the drone or in the network), and the possibilities offered by centralizing parts or all of the control logic.

Type: Master project
Period: 19.02.2018 - 19.06.2018
Section(s): MT
Type of work: 30% theory, 40% software, 30% hardware
Requirements:
Subject(s): flight control, state estimation, networking
Responsible(s): Julien Lecoeur, Alexandru Rusu (alexandru.rusu@swisscom.com)
Report: Click here

Flying in vision denied cluttered environments using contact-based navigation.

Valentin Kindschi (MT)

At the LIS, we are developing a new type of a safe and foldable delivery drone called the “PackDrone”. This drone has an all-around protective cage, which separates propellers from the environment thus protecting people and the drone. The unique cage design of the “PackDrone” also ensures the safe flight in the cluttered environments (while flying close to obstacles). Additionally, the cage is equipped with a mechanical solution to withstand collisions, thereby allowing it to physically interact with its environment by flying along flat surfaces such as walls, and ceilings.

The present version of the “PackDrone“ is autonomously stabilised but has not got yet any haptic sensors or high-level controller that tells it where to go after the collision. To navigate autonomously in cluttered vision denied environments (fog, smog or smoke), we aim at using the haptic-based interactions with the environment and apply some reactive control to the contacts that occur during flight. The goal of this project is to design and test haptic-based sensors for the drone.

The first task of the project is to design and implement haptic-based sensors within the protective cage. The sensor should be a part of the mechanism that allows to withstand collisions, thus redesign of the mechanism is required.

The second task is a validation of the implemented sensor with the device to withstand collisions. The validation should be performed by acquiring and analysing the sensor data using Arduino electronic development board. Tests will be performed by usage of motion capture system.

Type: Semester project
Period: 14.02.2018 - 14.06.2018
Section(s): MT
Type of work: 10% theory, 60% implementation, 30% tests
Requirements: microcontrollers e.g. Arduino, C/C++,
Subject(s): Flying Robots, transportation of packages, sensors
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here
URL: Click here

Wearable Technology: Prototyping of a Modular Hardware Front-End

Gabriel Neamtu (IN)

Wearable sensors are a relatively young technology which has been object of great interest for several industrial, clinical and research applications in the past years. One of the main limitation in this field is the lack of a complete and versatile environment for the interface of such devices. In the LIS we are planning to move a first step towards a unified framework for wearable technology, to allow the user to handle in a simple and rapid way the interface with wearable systems of different nature from multiple suppliers. The main goals of this project can be summarised as follows: • Research of the connectivity requirements for a set of wearable devices. • Choice of a hardware platform (microcontroller) for the development of the system. The choice should be justified by research on the market based on the needed specifications. • Design and prototyping of the front-end (analog/digital transmission, power supply, communications) • Implementation of hardware recognition and data streaming to a PC through a selected protocol. • Development of a wireless interface to communicate with a computer (bonus) • PCB design (bonus) • Design of different modules (bonus) The student will mainly work on hardware design, driver development and embedded software. Applicants are encouraged to contact directly the reference person for the project.

Type: Semester project
Period: 21.02.2018 - 08.06.2018
Section(s): EL MT
Type of work: 10%+theory +60%+hardware +30%+embedded+software
Requirements:
Subject(s): hardware design, hardware, interfacing, signal acquisition, communications
Responsible(s): Matteo Macchini, Anand Bhaskaran
Report: Click here

Precision landing for a cargo drone

Kunal Shrivastava (CH)

Description: Multicopters are recently used to deliver different types of goods in dense environments such as cities, due to their manoeuvrability and possibility for vertical take-off and landing. However, landing among high buildings may cause reflections or even loss of the GPS signal causing imprecision in the localisation of the drone, thus unprecise landing. In this project, we aim at using a vision based technique to land precisely in a designated place by a drone sender or recipient.

Specifically, the first goal of this project will be to localise and remember a take-off position of a drone for precise landing in the same place after delivery. The second task is to find a landing spot indicated on Google maps by a recipient. Thirdly, the algorithm should verify the position of obstacles on the ground such as people, cars, trees etc. and not land on them. Then, the landing performance will be validated on the prototype of the safe foldable delivery drone developed at EPFL.

Goal: The goal of this project is to design and implement a vision based solution for precise landing in cluttered outdoor environments.

This project will be done in collaboration between two labs: Robotic and Perception Group from the University of Zurich, and Laboratory of Intelligent Systems from Ecole Polytechnique de Lausanne.

Contact Details: Davide Scaramuzza (sdavide@ifi.uzh.ch), Przemyslaw Kornatowski (przemyslaw.kornatowski@epfl.ch)

Type: Master project
Period: 19.02.2018 - 01.06.2018
Section(s): IN MT
Type of work: 25% théorie, 60% software, 15% hardware
Requirements:
Subject(s): drones, vision based navigation
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran, Jeff Delmerico
Report: Click here
URL: Click here

2017


Development of a simulator for soft modular robots

Jean Marc Bejjani (MT)

Soft modular robots are versatile systems that can be assembled into different task-specific morphologies. Such robots are expected to safely locomote and manipulate beside or cooperatively with humans or in un-constructed environments. Soft robots, in fact, can freely deform along any direction and comply with any unexpected or excessive external force. On the other hand, it is very challenging to simulate the kinematics and dynamics of their soft deformable bodies. At LIS we are investigating a new approach to develop soft modular robots that can lead to a simplified representation of their body deformations. Hence, such simulations can be performed with classical physics simulators to predict soft modular robot dynamics in real time. The objective of this semester project is to utilize a classical physics simulator (BULLET engine) to perform real time dynamic simulations of soft modular robots. At first, the student will familiarize with the selected simulator and with the prototype of soft modular robot that we developed at LIS. Secondly, the student will modify the classical simulator to simulate the soft modular robot. Thirdly, the student will implement a specific control algorithm to control the robot in the simulated environment and extract its behavior data. Finally, a comparative study between the performance of simulated and real prototype will be performed.

Type: Semester project
Period: 15.02.2018 - 30.06.2018
Section(s): IN
Type of work: 20% theory 80% software
Requirements: C++, MATLAB, knowledge in kinematics and dynamics
Subject(s): Dynamic simulations, soft modular robots
Responsible(s): Davide Zappetti, Anand Bhaskaran
Report: Click here
 

Development of a controller for soft modular robots based on self-sensing

Luigi Campanaro (ME)

Soft modular robots are versatile systems that can be assembled into different task-specific morphologies to safely locomote and manipulate beside or cooperatively with humans or in un-constructed environments. Soft robots in fact can freely deform along any direction and comply with any unexpected or excessive external force. However, their soft bodies have unlimited degrees of freedom and it is very challenging to control their motions and the forces they apply to the environment. lndeed a distributed sensory feedback to detect local deformations should be included in the practical design of soft robots to improve their controllability. The objective of the thesis project is to develop a controller for actuated soft modules able to self-sense their deformations. At first the student will analyze state of art control strategy of soft robots. Secondly, he will implement state estimation control in one single actuated soft module in simulation and later in a real available prototype. Finally a possible control strategy for different assembled soft modules will be proposed and investigated.

Type: Master project
Period: 01.10.2017 - 31.03.2018
Section(s): ME
Type of work: 20% theory, 40% software, 40% hardware
Requirements:
Subject(s): soft robotics, modular robotics, control engineering
Responsible(s): Davide Zappetti, Jun Shintake
Report: Click here
 

A path planning algorithm for drones flying in urban environments

Julien Di Tria (Microengineering)

At the Laboratory of Intelligent Systems (LIS), we are developing a drone-based point-to-point logistics solution to transport lightweight packages (up to 2 kg). We have a JAVA-based software (Dronistics) that can control and monitor drones in real-time.

The objective of this project is to develop an algorithm that could compute the best path from point A to point B in a 3-dimentional space. This computed path should be capable of avoiding all the buildings, elevated lands and other obstacles. 3D model of the environment could be obtained from existing databases (such as Google earth). The developed algorithm should calculate the path in terms of GPS-waypoints and also be capable of providing various details of the path (such as the flight distance). The computed path should be drone-friendly, which means the path should be away from the obstacles (such as buildings and trees). The algorithm should be tested at different environments and should be capable of visualizing path in a 3D map.

Finally, the theoretical analysis of the algorithm (using Big O notation) should be performed and efficiency parameters such as computation time and memory usage should be optimized. The algorithm should be compatible with Dronistics software which is developed at LIS.

Type: Semester project
Period: 19.09.2017 - 03.02.2018
Section(s): IN MT
Type of work: 80% software programing, 20% testing
Requirements: Java programming, Data Structures, 3D geometry
Subject(s): Flying Robots, obstacle avoidance, path planning
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here

Rendering the sensation of flying: design, manufacturing and testing on subjects

Quentin De Longraye (IN)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To boost immersion, we aim to develop portable feedback interface to transmit haptic feedback from the drone to the user. To do so, we created a wearable and portable interface called the FlyJacket. With it, users can control a drone with intuitive body movements. An important factor to improve this control and also to deepen the embodiment is to render force that the drone is experiencing when flying. The goal of this project is to integer a simple, not cumbersome and lightweight device into the FlyJacket to render the haptic feedback and test its effectiveness by doing extensive test with a significant number of human subjects. The project work flow is the following: the student will have to do a literature research on the state of the art of wearable and portable devices that give somatic feedback. Then, the best solution will be chosen and a prototype will be designed, manufactured and implemented in the FlyJacket. The second half of this project is to test the efficiency of the device on a significant number of human subjects. This will be done by comparing performances between subjects with the haptic feedback and a control group that doesn't have this feedback.

Type: Semester project
Period: 19.09.2017 - 03.02.2018
Section(s): IN
Type of work: 20% literature review, 30% hardware, 50% experiment
Requirements: Motivation! 😉
Subject(s): haptic feedback, human drone interaction, actuator, wearable
Responsible(s): Carine Rognon, Seunghee Jeong
Report: Click here

A software framework for adaptive structure of multirotor transportation drone

Karim Zeid (MT)

At the LIS, we are developing modular-drones for fast transportation of lightweight packages (up to 2 kg) with medicaments, dressings for wounds, small equipment etc. As the packages have different weight and dimension, the modular-drone (of LIS) should adapt its body according to the package. These drones will be controlled in real-time using Dronistics (a JAVA-based software).

The objective of this project is to develop a software, which could firstly define a geometry of the modular drone depending on the size and weight of the parcel. Secondly, the software should adapt gains for proposed geometry. Thirdly, the developed software should be well tested and should be integrated with our existing software Dronistics.

The first task is to write a JAVA-based software that could define number of motors, and its position for given parcel. Additionally, the software should be capable of visualizing the proposed configuration.

The second task is to find the gain for the proposed configuration by interpolating the look-up table (non-model based methods such as gain scheduling).

The third task is to experimentally tune gains of different configurations of the multicopters in order to fill in the look-up tables required for the second task.

The last but not least, extensive tests should be done to validate the algorithm. The flight controller will be based on the PX4 and PixHawk autopilot (software and electronic control board) which communicates with the software Dronistics.

Type: Semester project
Period: 19.09.2017 - 03.02.2018
Section(s): IN MT
Type of work: 20% theory, 60% software, 30% hardware
Requirements: Java programming, programming PID controllers,
Subject(s): Flying Robot - drone, transportation of packages, adaptive morphology
Responsible(s): Przemyslaw Kornatowski, Anand Bhaskaran
Report: Click here

Development of variable stiffness dielectric elastomer actuator

Egor Piskarev (MA)

Dielectric elastomer actuator (DEA) is a class of electroactive polymer that exhibits high compliance, large actuation stroke, and fast response speed. DEAs are composed of two compliant electrodes sandwiching an elastomer membrane. Applying high voltage induces electric charges on the electrodes that squeeze the membrane in thickness direction and cause expand in area as actuation. The idea of this project is to use low-melting point alloy (LMPA) with DEA. LMPA changes its phase between solid and liquid as a function of temperature. Thanks to this behavior, DEA with LMPA can have variable stiffness of the structure to withstand large external force or to keep an actuated shape without consuming energy. During the project, the student will design a proof of concept of this variable stiffness dielectric elastomer actuator (VSDEA). Subsequently he will establish fabrication process of the device, and characterize it in terms of actuation stroke, force, and rigidity change. Finally, the student will develop a simple 2-fingered gripper using VSDEA to demonstrate the effect of stiffness change.

Type: Semester project
Period: 20.09.2017 - 01.02.2018
Section(s): MA
Type of work: 25+%+theory +75+%+hardware
Requirements:
Subject(s): Smart+materials
Responsible(s): Jun Shintake, Seunghee Jeong
Report: Click here
 

Smart Variable Stiffness Fibers for Soft Robotics

Ece Ozelci (ME)

Variable stiffness components (i.e components that change their stiffness and deformability under a specific stimulus) can be used in soft robotics structures to increase deformability and adaptability of the system while minimizing actuation components. In this context, we want to develop a variable stiffness component by using low melting point materials (alloys, polymers) encapsulated in soft materials (elastomers, textiles): the stiffness change is here due to the phase change (solid to liquid and vice versa) of the low melting point material, which is triggered by thermal energy.  In order to achieve better controllability, heating and cooling systems will be designed and integrated in the variable stiffness component. For example, microfluidic channels can be fabricated around a variable stiffness component and can be used as cooling systems. To fabricate them, 2D, 3D printing, laser machining, material patterning and heterogeneous material assembling technology can be used. The component and systems should be designed with consideration on a targeting application, and developed. The component performance (e.g heating and cooling times and control schemes) should be tested and optimized. We most welcome email contact for exploring ideas and discussing project details.

Type: Semester project
Period: 20.09.2017 - 31.01.2018
Section(s): MA ME MT MX
Type of work: 30% theory and design, 50% fabrication, 20% evaluation
Requirements: Heat transfer, Dielectrics, Electromechanical devices
Subject(s): Variable stiffness, Tensegrity robot
Responsible(s): Seunghee Jeong , Davide Zappetti, Jun Shintake
Report: Click here
 

Simulation model of upper body movement with kinetic assistance

Alaa Bakr Maghrabi (MT)

As part of a collaborative study with the BioRobotics Laboratory (BioRob), one of our research goals is to facilitate intuitive teleoperation of mobile robots. The approach that we have adopted involves mapping the degrees of freedom of the human torso to those of an aerial robot, and providing kinetic feedback based on this configuration, to assist the user in controlling the drone’s trajectory. To validate the user’s response to varying force feedback, we have developed a Simulink model of the torso and the upper body as a simple inverted pendulum. While this model serves as a good rudimentary platform for our analysis, we would like to develop a new model in OpenSim, a software platform that is built specifically for biomechanical systems analyses. The goal of this project is to build a human model and carry out a comparative study of the forward dynamics predicted by the two simulated systems with those measured during subject tests. Students should have some experience with Matlab and Simulink. Experience with OpenSim would be advantageous but is not required. Students will be co-advised with Amy Wu from BioRob.

Type: Semester project
Period: 20.09.2017 - 31.01.2018
Section(s): ME MT SV
Type of work: 20% theory, 10% hardware, 70% software
Requirements: Matlab and Simulink
Subject(s): Dynamics Model, Programming, Robotics, Simulator
Responsible(s): Vivek Ramachandran, Amy Wu
Report: Click here

Model and Design of a Soft, Wearable Sleeve

Pol Michel Alain Banzet (ME)

In recent years, there has been a burgeoning interest in the field of wearable robotics for rehabilitation and health monitoring. We are invested in broadening the scope of the current research to encompass applications like telerobotics, that would enable users to interact with robots in an intuitive manner, when provided with passive haptic feedback., The feedback provided through a system of compliant modular clutches, which conform to the human body, are meant to constrain human joint motions. The objective of the project is to design, model and fabricate an upper extremity sleeve integrated with these clutches. This project will entail a study of the surface interactions between various fabric substrates and the upper limb. Drawing on established design principles for orthotics, the student will present ergonomic solutions for the strategic placement of the clutches on the sleeve.

Type: Semester project
Period: 20.09.2017 - 31.01.2018
Section(s): MA ME MT MX SV
Type of work: 30% theory; 50% modelling; 20% design
Requirements: Matlab, Simulink/OpenSim, Solidworks
Subject(s): Biomechanics, Mechanics of Materials. Design
Responsible(s): Vivek Ramachandran, Carine Rognon, Amy Wu
Report: Click here
 

Soft active joint for human assistance

Jules, Mertens (CH)

Worldwide an increased demand for assistance and rehabilitation technologies is driven by the advent of an aging society. Assistance of limbs joints movements is one of the most important for a user. However, it is complex to design and functionalize a lightweight and flexible solution able to guide and allow mobility of the user at the same time. We seek to develop a wearable, compliant and lightweight tensegrity active joint for human limbs assistance. At first, the student will investigate state of the art upper limb wearable assistive robotic devices. Secondly, he/she will design and fabricate a wearable actuated tensegrity joint based on an available passive structure involving the usage of 3D printing technologies. Finally, the performances of the actuated soft wearable module will be assessed in terms of weight, compliance, and actuation forces.

Type: Semester project
Period: 28.09.2017 - 28.01.2018
Section(s): MT
Type of work: 30% theory, 70% hardware
Requirements: CAD (Inventor, SolidWorks or similar), good understanding of mechanisms
Subject(s): soft robotics, wearable robots
Responsible(s): Davide Zappetti, Jun Shintake, Rognon Carine
Report: Click here
 

A Software for Motion Tracking and Acquisition of Biometric Signals from Wearable Systems

Hugo Elysée Arthur Meyer (MT)

Wearable sensors are a relatively young technology which has been object of great interest for several industrial, clinical and research applications in the past years. One of the main limitation in this field is the lack of a complete and versatile environment for the interface of such devices. In the LIS we are planning to move a first step towards a unified framework for wearable technology, to allow the user to handle in a simple and rapid way the interface with wearable systems of different nature from multiple suppliers. The main goals of this project can be summarised as it follows: • Development of drivers for a set of different devices (IMUs and EMGs) • Design and development of a basic Graphic User Interface for the interaction with the hardware and the signal readout • Real time data storage and graphic output consisting in plots and 3D visualization of motion tracking and biometric signals. The interdisciplinary nature of the tasks would allow the student to learn how to tackle issues in several fields of engineering such as programming, hardware interfacing, signal processing, 3D graphics. Applicants are encouraged to contact directly the reference person for the project.

Type: Semester project
Period: 19.09.2017 - 12.01.2018
Section(s): IN MT
Type of work: 10%+theory +20%+hardware +70%+software
Requirements: python+or+java
Subject(s): wearable+devices +hardware+interfacing +signal+acquisition+and+processing +graphics
Responsible(s): Matteo Macchini, Anand Bhaskaran
Report: Click here

Modelling the Neuromuscular system and the Kinematics of the Human Body

Johann Hêches (MT)

Wearable sensors have become very popular in many applications such as medical, entertainment, security, and commercial fields. In the LIS, our goal is to develop a unified framework to allow the user to easily design and prototype experiments and application exploiting this kind of technology. In order to do this, and accurate model of the human body needs to be prepared and implemented as a 3D avatar for the tool. The main phases of the project can be summarised as it follows: • Study and production of a preliminary report (3-5p.) of the state of the art of wearable systems: sensors and actuators, technology, placement on the body, signal acquisition and processing. • Analysis of biomechanical aspects for a kinematic model of the human body. • Analysis of the human neuromuscular structure for the accurate placement of sensors and actuators. • Implementation of the acquired knowledge in the production of a 3D virtual model of the human body, on which a set of sensors (visualised as simple markers) can be applied. The model needs to be scalable and adjustable to different body structures. • As an additional task, the student can propose a tool or a procedure to automatically adapt the appearance of the avatar to different persons. Applicants are encouraged to contact directly the reference person for the project.

Type: Semester project
Period: 19.09.2017 - 12.01.2018
Section(s): IN MT SV
Type of work: 40% theory, 30% modelling, 30% software
Requirements: python or java, preferably
Subject(s): wearable devices, modelling, biometric signals, graphics
Responsible(s): Matteo Macchini, Seunghee Jeong
Report: Click here

Development of a biotensegrity mobile robot

Timothée Peter (MT)

Biotensegrity could be a novel design principle for bio-inspired mobile robots. The aim of this project is to proof this concept through development of a bio-inspired mobile robot using tensegrity structure. The robot will consist of a tensegrity structure driven by a servo motor. The student is expected to develop a prototype through literature review, designing, fabrication, and characterization. Please contact the responsible person for more detail.

Type: Semester project
Period: 17.09.2017 - 22.12.2017
Section(s): MA ME MT MX
Type of work: 25 % theory, 75 % hardware
Requirements: Solidworks, 3D printing
Subject(s): Bio-inspired robots, tensegrity
Responsible(s): Jun Shintake, Davide Zappetti, Yusuke Ikemoto
Report: Click here
 

Development of a smart materials based actuation for soft modular robots

Sébastien Rosat (ME)

Modular robots that can change morphology through re-assembly are a robust and versatile solution in situations where multiple tasks are required and the operating environment is not well defined or even unknown. Soft modules can be implemented in such systems to display physical compliance providing safer and better interaction of these robots with the environment. Moreover, the body softness provides a very high robustness absorbing high mechanical shocks. We seek to develop a soft actuator technology based on smart materials to improve such robots in terms of robustness and compliance. At first, the student will investigate state of the art smart materials based actuators. Secondly, he will design and fabricate an actuation mechanism, and implement it in an available prototype of soft modular robot developed in our lab. Finally, the performances of the actuated soft modular robot will be assessed in terms of deformability, robustness and force.

Type: Semester project
Period: 20.09.2017 - 20.12.2017
Section(s): MA ME MT MX
Type of work: 30% theory 70% hardware
Requirements: smart and novel materials, good understanding of mechanisms, CAD (Inventor, SolidWorks or similar)
Subject(s): smart materials, soft robots, modular robots
Responsible(s): Davide Zappetti, Jun Shintake
Report: Click here

Indoor speed estimation for a contact tolerant drone

Simon Pyroth (MT)

Flyability redefines UAV boundaries and brings drones indoors, in complex and confined spaces, and in contact with people. Our goal is to shape the future of the drone market through innovation and commercialization of novel products changing how people work, play and communicate. Flyability is an investor-backed tech-focused company which has received multiple awards and has a worldwide recognition. Flyability’s first product, Elios, is designed for industrial inspection professionals who can now get access for the first time to complex places in seconds without risk. The goal of this master project is to work on potential future technologies for the next Flyability drone. The primary focus will be researching and prototyping sensing solutions which would allow the drone to estimate it’s current speed in unstructured industrial environments. Some of the tasks will include: - Studying the state of the art of sensors which could be used for position or speed estimation. - Testing and evaluating various sensors and evaluation kits, such as optical flow and inertial sensors. - Creating prototypes of possible solutions, including electrical design and software. - Performing tests to assess and compare the performance of various solutions in different environmental conditions. - Choosing the most appropriate solution and providing guidelines for the implementation of this solution on the next Flyability product. As this master project is focusing on the development of future technologies for future products, the student should be conscious that the scope of this internship can be redefined from time to time to fit better-emerging ideas.

Type: Master project
Period: 25.01.2017 - 25.08.2017
Section(s): MT
Type of work: 30% theory, 30% software, 40% hardware
Requirements:
Subject(s): state estimation, sensor fusion
Responsible(s): Julien Lecoeur, Ludovic Daler and Alexandre Pabouctsidis
Report: Click here

Design and manufacturing of a slack-enabling tendon drive for soft upper-limb exosuit

Alexandre Moscardini (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To boost immersion, we aim to develop portable feedback interface to transmit haptic feedback from the drone to the user. A way of doing it is using cable driven actuation. One drawback of this actuation is that the tension should always be maintained in the tendon to prevent derailment from the spool. The goal of this project is to develop and test a slack-enabling tendon drive. As this device will be integrated in a wearable device, it has to be small, lightweight, passive and the friction should be minimized., During this project, the student will propose different solutions and a prototype will be designed, manufactured, and extensively tested (optimization of the size and weight, system reactivity, friction …).

Type: Semester project
Period: 20.02.2017 - 20.06.2017
Section(s): ME MT
Type of work: 20% theory, 50% hardware, 30% experiment
Requirements: SolidWorks is an advantage
Subject(s): Mechanics, Wearables, cable actuation
Responsible(s): Carine Rognon, Stefano Mintchev
Report: Click here

Design, manufacturing and testing of a bidirectional glove

Hugo Viard (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive interfaces to interact with distal robots. To improve the control, we aim to develop portable bidirectional interface that can command distal robots and receive haptic feedback from them. The goal of this project is to develop and control a smart bidirectional glove. This glove will sense finger movements using liquid metal sensors, and will give back pressure sensation (such as in case of gripping an object) using fishing line actuators. During this semester project, the student will have to get familiarized with human proprioception, study and develop both the sensing and the actuation parts. He will have to integrate both device in a fully functional glove. The issue of the high temperature for the actuation will have to be solved. The student will also have to make the control (reading sensors and actuations) and to do extensive tests on the interface performances.

Type: Semester project
Period: 20.02.2017 - 20.06.2017
Section(s): MT
Type of work: 20% theory, 50% hardware, 30% experiment
Requirements: Electronic is an advantage
Subject(s): Soft robotic, haptic feedback, sensor, wearable
Responsible(s): Carine Rognon, Alice Tonazzini
Report: Click here
 

Variable stiffness t-shirt for wearable human-robot interfaces

Louai Bensaid (MT)

At the Laboratory of Intelligent Systems (LIS) we develop variable stiffness fabrics to be used in wearable human-robot interfaces. The goal of this project is to develop a prototype of a self-adaptable t-shirt capable to vary its stiffness and deformability: in the soft state, the t-shirt should softly conform to the morphology of the user, like a normal elastic t-shirt. Then, once adopted the user morphology, it should become stiff and not deformable in order to rigidly maintain the adopted shape and block any further movement, like a carapace. Such a smart t-shirt has to be lightweight and portable. During this project, the student will study a specific variable stiffness strategy (i.e. layer jamming) and will design, manufacture and test a functional prototype.

Type: Semester project
Period: 20.02.2017 - 02.06.2017
Section(s): MA MT MX
Type of work: 20%theorie, 30% experiments, 50% hardware
Requirements:
Subject(s): Wearable robotics, soft robotics, variable stiffness
Responsible(s): Alice Tonazzini, Carine Rognon
Report: Click here
 

Fabrication and characterization of low cost ultra stretchable sensors

Egor Piskarev (MA)

We have developed a low cost ultra stretchable sensor. The next step is to characterize the sensor performance in order to apply this technology to robotic applications. The student will work on: 1) Fabrication of the sensor. 2) Characterization of the sensor in terms of strain, cycle, and repeatability. 3) (optional) If the project goes quickly, there will be chance to prototype a robotic application.

Type: Semester project
Period: 20.02.2017 - 02.06.2017
Section(s): MA
Type of work: 25% theory, 75 % hardware
Requirements:
Subject(s): Soft robotics, stretchable electronics
Responsible(s): Jun Shintake
Report: Click here
 

2016


Highlight of geo-referenced information during flight

Lucas Monnin (IN)

At the Laboratory of Intelligent Systems (LIS), we aim to bring the human closer to the drone. By developing a bidirectional interaction we obtain a simpler interaction and allow more people to fly. In our solution, the human is flying a fixed-wing drone using his upper body gestures and at the same time is having a visual and a haptic feedback. The sensors to monitor the gestures and the actuators are embedded in a 'flight jacket' the user is wearing. The goal of this project is to develop relevant tools to be able to use the fly jacket in search and rescue scenarios. In simulation, a quadcopter will first be designed and used with the fly jacket. Then, highlights of geo-referenced information such as interest/hazard spots triggered by an external device will be implemented. This will then be tested with a real quadcopter using augmented reality. Simulation and reality tests will conclude the success of this project.

Type: Semester project
Period: 20.02.2017 - 17.06.2017
Section(s): IN MT SC
Type of work: 100% software
Requirements: Programming, aerodynamics, natural interface
Subject(s): Aerial Robots, Simulation
Responsible(s): Alexandre Cherpillod, Grégoire Hilaire Marie
Report: Click here

Symbiotic User Interface between the fly jacket and a commercial drone

Cyril Stuber (MT)

Drones are becoming ubiquitous in our daily lives and are challenging us to develop new avenues of interactions. Existing interfaces such as joysticks and remote controllers require training and constant cognitive effort and provide a limited degree of awareness of the robots’ state and its environment. At the laboratory of Intelligent Systems (LIS) we are developing a new immersive and intuitive interface that allows novice and experienced users to fly drones. The setup uses a “fly jacket” that tracks user hand gestures and provides visual feedback from a camera mounted on the drone. The goal of this project is to interface the fly jacket with a commercial drone. A first objective of the thesis is to develop the required hardware and software to interface the fly jacket and the drone. A second objective is to ensure a safe interaction with the drone. This will encompass automatic take-off and landing, geo-fencing, stable flight guarantee, etc. A third objective is to tune the level of aggressiveness of the drone during flight. An integrated system and demonstration is expected by the end of this project.

Type: Master project
Period: 20.02.2017 - 17.06.2017
Section(s): EL MT
Type of work: 10% theory, 70% software, 10% hardware
Requirements: Programming, aerodynamics, natural interface
Subject(s): Aerial Robots, Natural User Interaction
Responsible(s): Alexandre Cherpillod, Grégoire Hilaire Marie
Report: Click here

Development of a soft, jumping robot

Grégoire Besson (MT)

Locomotion on rough terrain is a difficult task for robots of small in size, such as those used for inspection or search and rescue scenarios. To solve this problem, several small animals in nature use jumping as their main means of locomotion. We seek to develop a novel type of soft jumping robot composed of a deformable body., At first, the student will analyze state of art solutions of jumping robots. Secondly he/she will develop and manufacture an actuation system for an available prototype of a deformable and lightweight, smart structure developed in our lab. Thirdly, a simple controller will be developed and tested to perform jumps and angled jumps. Finally, the prototype will be tested to assess the jumping capability, the speed and jumping angle precision.

Type: Semester project
Period: 15.02.2017 - 15.06.2017
Section(s): MA ME MT
Type of work: 20% theory 30% research 50% hardware
Requirements: Solidworks or similar, microcontroller programming, good understanding of mechanisms
Subject(s): Jumping robots, soft robots
Responsible(s): Davide Zappetti, Jun Shintake
Report: Click here

Design of a Disposable Robot for Exploration in Search and Rescue Scenarios

Jérémie Willemin (MT)

The robots currently recruited for search and rescue missions are very complex and expensive machines, hence fragile and difficult to deploy on the field. A different approach consists in deploying a swarm of minimalistic and inexpensive robots capable to rapidly explore the environment in order to stream vital information to the rescuers. For example, the robots can penetrate the disaster zone and go beneath the rubble where their sound, heat, motion and CO2 sensors can be used to detect signs of life. Using as a starting point an available prototype of disposable flying robot, the goal of this project is twofold: to improve the mechanical design of the robot aiming at ensuring locomotion in uneven environments; and to implement a control strategy for random locomotion. At first the student will analyze state of the art solutions in the field of swarm robotics for search and rescue missions. Secondly the student will improve the mechanical design of the available prototype focusing on weight reduction, ease of manufacturing and crash resilience. Thirdly, a simple flight controller will be developed and tested. Finally, the student will perform tests to assess the performances (weight, resilience, operation time) of the prototype.

Type: Semester project
Period: 20.02.2017 - 02.06.2017
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements: Solidworks or similar, programming, aerodynamics
Subject(s): Swarm robotics, Aerial robotics
Responsible(s): Stefano Mintchev, Zappetti Davide
Report: Click here

Design and Manufacturing of Foldable Wings Based on Layer Jamming

Guglielmo Milan (ME)

Aerial robots provide valuable support in several high-risk scenarios thanks to their capability to quickly fly to locations dangerous or even inaccessible to humans. In order to fully benefit from these features, aerial robots should be easy to transport and rapid to deploy. One solution to solve this transportation issue is to reduce the size of the drones. The shortcoming is a reduction in payload capabilities and in flight time. Therefore, foldable structures, which can be packaged for transportation and quickly deployed for operation, are a promising solution to enable full benefits from the potential of flying platforms. The goal of this project is to develop foldable wings for small sized flying robots (wingspan 0.5-1 m) based on layer jamming. Layer jamming is instrumental in stiffening the wing in the deployed configuration ensuring load bearing capabilities. First, an analysis of the state of the art on jamming technologies and origami wing will be done. Then, based on this analysis a design will be proposed and a prototype of the foldable wing will be manufactured. Finally, static tests and flight tests will be performed to assess the design and suggest improvements. The goal of the test is to assess the deployment (folding and unfolding) time, the volume reduction and the wing weight and stiffness.

Type: Semester project
Period: 20.02.2017 - 02.06.2017
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements: Solidworks or similar
Subject(s): Aerial Robots, Foldable mechanism
Responsible(s): Stefano Mintchev, Vivek Ramachandran
Report: Click here

Foldable wings for VTOL UAVs to facilitate storage and transportation

Cameron Dowd (CH)

Since Africa is growing too fast to build out its road network, some goods will have to be supplied from the sky. We believe Africa will be the first continent to build out unmanned air cargo at massive scale.

That is why Laboratory of Intelligent Systems collaborate with the "Red/Blue Line" flying robot consortium (http://afrotech.epfl.ch/page-115280-en.html) which is a spin-off of Afrotech based at EPFL (http://afrotech.epfl.ch). Consortium aims to set up the first flight-cargo robot route in Africa by 2016. It will be about 80 kilometres long and will connect several towns and villages. The first use case will be to fly units of blood from a centralized blood bank to peripheral health clinics. By bringing blood to severely anaemic young children and mothers, as well as to trauma patients, the route will prove that autonomous flight can save lives.,

The goal of this project is to design foldable wings to facilitate storage and transportation of VTOL platforms used for transportation by Red Line.

The project will involve two important parts:

Identification of different solutions of mechanical structures taking into account scalability – different sizes of the platforms for different payloads.

The second part will address the investigation of materials and manufacturing technologies that could be used for different sizes of VTOL platforms.

This project will involve dimensioning, CAD design and simulations of foldable wings. Proposed foldable mechanism should be manufactured and tested in order to assess the design and suggest improvements.

Type: Master project
Period: 21.09.2016 - 15.02.2017
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements: CAD software, mechanical knowledge
Subject(s): Flying Robots, Foldable structures, lightweight materials, fixed wing platforms
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev
Report: Click here
URL: Click here

Development of a novel type of bio-inspired underwater robot

Thibaut Paschal (MT)

We seek to develop a novel type of bio-inspired underwater robot capable of multiple swimming modes. The robot will consist of silicone based soft body actuated by servo motors powered either tethered or untethered condition. The student is expected to develop a prototype through literature review, designing, fabrication, and characterization. Please contact the responsible persons for more detail.

Type: Semester project
Period: 20.09.2016 - 18.01.2017
Section(s): ME MT MX
Type of work: 25 % theory, 75 % hardware
Requirements: Solidworks
Subject(s): Soft robotics, underwater robots, bio-inspired robots
Responsible(s): Jun Shintake, Stefano Mintchev
Report: Click here

Soft actuators based on a new material

Egor Piskarev (MA)

We recently developed a new material for soft robotics. Our next step is to create a soft actuator out of this material. In this project organized between LIS and RRL, the student is expected to (1) characterize the mechanical property of the material using a mechanical tester, (2) fabricate a pneumatic soft bending actuator by reference to methods available in literature, and (3) characterize the actuator in terms of bending angle and force using an existing setup. The set up may be needed to modify according to the specification of the actuator. For more details, please contact the responsible person.

Type: Semester project
Period: 20.09.2016 - 15.01.2017
Section(s): EL MA ME MT MX
Type of work: 80+%+hardware +20+%+theory
Requirements: Solidworks
Subject(s): Soft+robotics +pneumatic+actuator
Responsible(s): Jun Shintake, Harshal Sonar
Report: Click here
 

Wind speed measurement and wind audio feedback for aerial robots

Tristan Besson (MT)

At the Laboratory of Intelligent Systems (LIS), we aim to bring the human closer to the drone. By designing a bidirectional interface between a human and a drone, we allow more people to use the interface in a simpler way. At LIS, we also aim to develop aerial robots capable of monitoring their environment for different tasks such as search and rescue, environmental monitoring and inspection. The goal of this project is to design a solution to measure the wind speed and to use the same hardware to record the sound of the wind and to transmit it to the ground user. Indeed both of these functions require engine noise removal. This project will be based on a previous semester project which only adressed the speed measurement. A hardware setup will be built and signal processing will be used. The whole system will be testd on a real flying platorm by the end of the project.

Type: Semester project
Period: 20.09.2016 - 13.01.2017
Section(s): EL MT
Type of work: 20% theory, 40% hardware, 40% software
Requirements:
Subject(s): Human Drone Inteface, Flying Robotics
Responsible(s): Alexandre Cherpillod, Grégoire Hilaire Marie , Meysam Basiri
Report: Click here

Gimbal camera with on-screen display

Cyrill Baumann (MT)

At the Laboratory of Intelligent Systems (LIS), we aim to bring the human closer to the drone. By designing a bidirectional interface between a human and a drone, we allow more people to use the interface in a simpler way., The visual feedback is an important component to have immersion. This has to mimick head orientation and be able to display other desired information. The goal of this project is to design a 3-axis gimbal camera and to add specific on-screen information to the video displayed in the Oculus Rift DK2. Based on maxon motors, a given controller board and a tiny camera, the mechanical design of the gimbal will have to be designed. Then, the image received on a computer will be added custom display information such as the level of battery, etc. The setup will be tested in real experiment at the end of the project.

Type: Semester project
Period: 20.09.2016 - 13.01.2017
Section(s): EL ME MT
Type of work: 50% hardware, 50% software
Requirements:
Subject(s): Human Drone Inteface, Visual feedback
Responsible(s): Alexandre Cherpillod, Przemyslaw Kornatowski
Report: Click here

Implementing a grammar for RoboGen body evolution

Carlos Malanche+Flores (MA)

RoboGen™ is an open source platform for the co-evolution of robot bodies and brains. It features an evolution engine, and a physics simulation engine. The goal of the project is to implement new encoding methods for RoboGen robot body descriptions which makes morphology mutations more natural and efficient in terms of exploration of the search space.

Type: Semester project
Period: 20.09.2016 - 13.01.2017
Section(s): EL IN ME MT MX PH SC
Type of work: 40%+Theory+60%+Software
Requirements: C/C+++Programming +Artificial+Evolution
Subject(s): Artificial+Evolution, Genetic Algorithms, Programming
Responsible(s): Alice Concordel, Basil Huber, Josh Auerbach
Report: Click here
URL: Click here

Creating more flexible parts in the RoboGen simulator

Gaël Gorret (MT)

RoboGen™ is an open source platform for the co-evolution of robot bodies and brains. It features an evolution engine, and a physics simulation engine. Learn more at www.robogen.org The goal of the project is to restructure how parts are defined in the simulator, in order to make it easier to create new parts with more flexible parameters. The simulator is coded with C, .

Type: Semester project
Period: 20.09.2016 - 13.01.2017
Section(s): IN SC
Type of work: 20%+Theory +80%+Software
Requirements: C/C+++Programming +Simulation
Subject(s): Programming +Simulation +Physics-based+simulation
Responsible(s): Alice Concordel, Basil Huber, Joshua Auerbach
Report: Click here
URL: Click here

Automated Flight Test of MAV

Dalmir Hasic (IN)

At the Laboratory of Intelligent Systems we are developing an autopilot software for drones. Since the development of this software is continuous, frequent testing is crucial to assure functionality. In this project, the student will develop an automatic testing procedure for the autopilot software. The testing software should perform flight tests (in simulation and reality) to assure the autopilot is fully functional and measure its performances, such as needed time for a mission and trajectory error. The software should then output a comprehensive report about the test results. The student will determine the testing procedure and the performance metrics. He/she will then implement the automatic testing and define the format of the resulting report. The testing should then be included in the development workflow (github).

Type: Semester project
Period: 19.09.2016 - 13.01.2017
Section(s): IN MT
Type of work: 20% theory;, 80% software;
Requirements: Programming (C++)
Subject(s): Drones, Autopilot, Testing
Responsible(s): Basil Huber, Julien Lecoeur
Report: Click here
 

Creating a Ground Control Station Software for Drones

Arthur Pierre Philippe Hervé Gay (MT)

Ground control software is a key part of the operation of Drones. It allows to monitor the current state of the Drone, plot and log telemetry data, and also send commands to the drone. Current ground control station are generally large pieces of software, which are difficult to set-up and modify for custom needs. The student will improve and extend a ground control software that was developed by a student during the spring semester. The focus of the project is to adapt the software to the needs of the laboratory while being compatible with various open source software.

Type: Semester project
Period: 19.09.2016 - 13.01.2017
Section(s): IN MT
Type of work: 80%+programming +20%+experiments
Requirements: programming:+python ++javascript +html +css
Subject(s): Human-Drone+interface
Responsible(s): Basil Huber, Julien Lecoeur
Report: Click here
URL: Click here
 

Design and manufacturing of an active arm support to receive haptic feedback from a drone

Matthieu Boubat (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To boost immersion, we aim to develop portable feedback interface to transmit haptic feedback from the drone to the user. The goal of this project is to develop and manufacture an active arm support. This device as two functions: to support that arm to prevent fatigue and to give kinetic feedback for the abduction movement. In addition, it has to be lightweight, wearable, user-friendly, and given forces tunable., This project deals with human proprioception, mechanics, and electronics. Based on the knowledge acquired during a study of the human arm proprioception, the student will propose different solutions and a prototype will be designed, manufactured, and tested.

Type: Semester project
Period: 20.09.2016 - 10.01.2017
Section(s): ME MT
Type of work: 20% theory, 50% hardware, 30% experiment
Requirements: SolidWorks is an advantage
Subject(s): Haptic feedback, human robot interaction, exoskeleton
Responsible(s): Carine Rognon, Stefano Mintchev
Report: Click here

Development of a novel adaptive drone

Maimun Al-Tayar (MT)

We seek to develop a novel type of quadcopter capable of carrying objects of different shapes and sizes by adapting its stretchable, soft silicone body. In the robot, the object is held by contraction force generated from the stretched body. In other words, the object provides rigidity to the robot to fly stable. The student is expected to develop a prototype through literature review, designing, fabrication, and characterization. Please contact the responsible persons for more detail. Responsible(s): Przemyslaw Kornatowski, Jun Shintake, Stefano Mintchev

Type: Semester project
Period: 10.09.2016 - 10.01.2017
Section(s): ME MT MX
Type of work: 25 % theory, 75 % hardware
Requirements: Solidworks
Subject(s): Soft robotics, flying robots
Responsible(s): Przemyslaw Kornatowski, Jun Shintake
Report: Click here

Design, manufacturing and control of a kinetic feedback device for the torso

Valérie Céline Springmann (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To boost immersion, we aim to develop portable feedback interface to transmit haptic feedback from the drone to the user. The goal of this project is to develop and control a device to give kinetic feedback at the torso level in bending motion (front, back, right and left). The device has to be lightweight, user-friendly, and given forces tunable., This project deals with human proprioception, mechanics, and electronics. Based on the knowledge acquired during a study of the human torso proprioception, the student will propose different solutions and a prototype will be designed, manufactured, and tested.

Type: Semester project
Period: 20.09.2016 - 10.01.2017
Section(s): ME MT
Type of work: 20% theory, 50% hardware, 30% experiment
Requirements: SolidWorks is an advantage
Subject(s): Haptic feedback, human-robot interaction
Responsible(s): Carine Rognon, Stefano Mintchev
Report: Click here

Voice assistance for delivery drones

Nicolas Winteler (MT)

At the LIS we are developing a drone for fast transportation of different packages to individual users. We would like to implement within a drone a system of voice communication/information in order to help user to interact with a drone during procedures of: landing, removing a package from a drone and taking off.

The first task of this project is to implement on a drone system that will allow users for voice communication in a real time. This feature will help user to interact with the drone in case of any problems during the last phase of the delivery. This task will require to interface an appropriate microphone and a lightweight powerful speaker with a GSM module to send and receive signals on a long distances.

Implementation of GSM module will require to adapt/write algorithms for voice communication. Second task is to develop an algorithm that will filter the sound of propellers and allow users for voice communication during the landing or take off procedure. This task will require to use additional electronic platform for computation and interface it with electronics from the task one.

The last but not least assignment will be to implement voice instructions for the recipient of the package for different steps of the delivery.

Type: Semester project
Period: 10.09.2016 - 10.01.2017
Section(s): EL MT
Type of work: 30%+theory +50%+software +20%+hardware
Requirements: Programming +signal+processing +PCB+design+
Subject(s): Flying+Robot ++transportation +sound+filtration +voice+communication
Responsible(s): Przemyslaw Kornatowski, Grégoire Hilaire Marie
Report: Click here

Variable stiffness wearable cast

Lucie Gabrielle Sarah Houel (MT)

Orthopedic casts have to be breathable, lightweight, fully conformable to the patient anatomy and stiff enough to support the injured body part. The goal of this project is to develop a wearable orthopedic cast made of variable stiffness fiber, that can be easily worn and removed like a sleeve; this cast will allow an easy removal (definitive / temporary for diagnosis or for adapting to a less swollen injured area) and therefore it will be re-used (by the same patient or others) .Technological solutions in the fields of variable stiffness have be studied and the requirements for a safe wearability have to be taken into account; a prototype has to be designed, fabricated and tested.

Type: Semester project
Period: 15.09.2016 - 31.12.2016
Section(s): ME MT MX
Type of work: 30% theory, 50% hardware and manufacturing, 20% experiments
Requirements:
Subject(s): Soft and wearable devices, Materials, Mechanics
Responsible(s): Alice Tonazzini, Stefano Mintchev
 

Robotic fabric for next generation human-robot interfaces

Lea Bole-Feysot (MT)

At the Laboratory of Intelligent Systems (LIS) we aim to develop multifunctional tissues that integrate sensing and actuation. Those tissues enable a wide range of applications, from rehabilitation exosuits to wearable human-robot interfaces. The goal of this project is to develop a proof-of-concept fabric capable to conform to a curved surface (e.g. human arm) and sense its deformation; the fabric has to be actuated, lightweight and breathable. Technological solutions in the fields of wearable sensors, fiber-like actuators and textile manufacturing have be studied; once selected the ones suitable for the targeted application, a prototype has to be designed, manufactured, and tested.

Type: Semester project
Period: 15.09.2016 - 31.12.2016
Section(s): ME MT MX
Type of work: 30%theory +50%hardware and manufacturing +20% experiments
Requirements:
Subject(s): Soft and wearable devices, Materials, Mechanics
Responsible(s): Alice Tonazzini, Carine Rognon
 

Development of a Quadcopter with Adaptive Morphology

Benjamin Vincent Camile Bonnal (MT)

Morphology plays an important role in behavioral and locomotion strategies of living and artificial systems. There is biological evidence that adaptive morphological changes can not only extend dynamic performances by reducing trade-offs during locomotion, but also provide new functionalities. The goal of the project is to develop a quadcopter that exploits adaptive morphology: i) for folding for ease of storage and transportation, ii) for achieving different modes of flight that can enable critical and opposing requirements such as maneuverability and efficiency. Based on available technologies for morphing that have been developed at LIS, the student will: i) design a prototype, ii) manufacture a prototype, iii) assess the results.

Type: Semester project
Period: 20.09.2016 - 18.12.2016
Section(s): ME MT MX
Type of work: 20% theory, 40% mechanical designs, 30% hardware and manufacturing, 10% experiments
Requirements: CAD (Solidworks preferred), Basic electronics and programming
Subject(s): Flying robotics, , Multi-Modal Locomotion
Responsible(s): Stefano Mintchev

Design and Control of a Foldable Quadcopter

Frédérick Matthieu  Gusset (MT)

Aerial robots provide a valuable support in several high-risk scenarios thanks to their capability to rapidly fly in locations dangerous or even inaccessible to humans. In order to fully benefit from these features, aerial robots should be easy to transport and rapid to deploy. With this aim, at LIS we have developed a novel pocket sized quadrotor with foldable arms. The quadrotor can be packaged for transportation by folding its arms around the main frame. Before flight, the quadrotor’s arms self-deploy in 0.3 second. Using as a starting point the available prototype and control algorithms, the goal of this project is twofold: to add mechanical protections that improve collision resilience; and to implement stabilization algorithms to hold altitude and to recover stability after being launched in the air by hand. At first the student will analyze state of the art solutions in the field of origami robotics, foldable structures and variable stiffness materials. Secondly the student will design and manufacture a prototype of the foldable arms/frame. Finally, the controller will be developed and tested.

Type: Semester project
Period: 20.09.2016 - 16.12.2016
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements: CAD (better if SolidWorks), C
Subject(s): Origami robotics, Foldable mechanisms
Responsible(s): Stefano Mintchev, Grégoire Hilaire Marie
URL: Click here

Design a software framework for transportation drones

Anand Bhaskaran (MT)

At the Laboratory of Intelligent Systems (LIS) at École Polytechnique Fédérale de Lausanne (EPFL) we are developing a drone for point to point delivery. We are targeting drones for long range (more than 10 km) delivery to our houses of lightweight packages (up to 2 kg).

The main goal of this project is to design a software framework to monitor transportation of parcels by a flying platform. The software should allow drone to fly autonomously to desired location and verify person responsible to receive package.

First step of the project is to implement WebServer and WebApp for a smartphone (or use an existing smartphone application) to control point to point navigation. Afterwards, implementation of Database Server to maintain the logs and run-time information has to be done. To constantly monitor the status and the position of the drone over long distances GSM interface has to be implemented. User should also have the possibility to change parameters of the mission and allow the sender to decide what to do in case of unforeseen situations. Additionally, this framework should support the management of multiple drones.

Secondly, the algorithm should be written to compute on the server the best path to fly from point A to point B avoiding forbidden areas (airports, stadiums, crowded places etc.). Information about those areas will be downloaded from the existing Geofence Database Server. Computed way points have to be send to the drone through GSM Cellular Network.

The third feature of the software framework, should be recognition and verification of the recipient of the package (verification should be made by the application on the recipient's smart phone). Recipient should acknowledge receipt of the package using the application. In case of negative authorization appropriate action should be performed by a drone. For example: land on the spot pointed by the sender or fly directly back home.

Finally the developed software framework by the student should be tested using one of the flying platforms designed at LIS for transportation of goods. The software framework should be capable to interface with the MAVRIC autopilot (electronic control board and software) developed at LIS. Diagram on the Figure 1 explains concept of the described above software framework.

In terms of research, the project entails:

• A review and comparison of point to point navigation approaches, • Status and positioning monitoring, and a modeling of possible deviating situations • A tradeoff analysis regarding local and remote (cloud) computations • An investigation into the parameterization of the drone behaviour from the perspective of the definer of a mission.

Type: Master project
Period: 01.03.2016 - 31.08.2016
Section(s): MT
Type of work:
Requirements:
Subject(s): programming
Responsible(s): Przemyslaw Kornatowski, Grégoire Hilaire Marie
 

Bio-inspired collision avoidance in cluttered environment

Darius Constantin Merk (PH)

Insects are known to rely heavily on vision during flight, more specifically on optic flow which is the apparent motion on their retina as they move through a scene. A clever use of optic flow cues can explain most of the behaviors exhibited by flying insects such as body stabilization, position and speed control, height control, visual odometry and landing. In a recent publication [Bertand2015], a method was proposed to explain how insects modulate the amplitude of saccades in order to avoid collisions in cluttered environments. Visual motion is measured all around a spherical eye and used to build a proximity map of the environment. On this basis, a motion direction is computed for the next saccade. The goal of this project is to implement this method on a real flying robot. The project will be validated by autonomous flights in cluttered artificial environments with controlled geometry and texture, as well as in natural environments. Depending on results and time, the student will propose improvements to the method and implement them.

Type: Master project
Period: 01.02.2016 - 08.08.2016
Section(s): MT
Type of work: 50% software, 50% experiments
Requirements: Good programming level
Subject(s): Optic flow, autonomous flying robot, collision avoidance
Responsible(s): Julien Lecoeur, Basil Huber

Miniature vision sensor for flying robot

William Ponsot (MT)

The goal of this project is to develop a miniature smart vision sensor for drones, composed of a conventional camera and a microcontroller. Image processing algorithms such as optic flow extraction and obstacle detection will run on the microcontroller and extract useful information of the drone's motion and the environment. The extracted data will then be sent to the drone's autopiloy. Several of these smart sensors will be mounted on a drone and used for autonomous navigation tasks, such as collision avoidance and target following. The project consists of designing the PCB with the camera, the microcontroller and other components (power regulator, clock, etc.). In the second phase, the student will implement the image acquisition on the microcontroller. The acquisition should be coded in C/C, . Sample code for the acquisition will be provided. If time allows, the student could implement basic image processing functions on the microcontroller and test it on a flying robot.

Type: Semester project
Period: 08.01.2016 - 08.08.2016
Section(s): EL IN MT
Type of work: 30% software, 70% hardware
Requirements: PCB design, embedded C/C++ programming
Subject(s): Electronic design, digital camera, smart sensor
Responsible(s): Julien Lecoeur, Basil Huber

Onboard simulation of the dynamics of a fixed wing UAV

Nicolas Jacquemin (MT)

The MAVRIC autopilots developed at LIS are capable of simulating the dynamics of a quadcopter on the onboard microcontroller. When in simulation mode, the values given by embedded inertial sensors, barometer, sonar and gps are not used. Instead, a dynamic model is updated, and simulated sensor values are computed accordingly. This is very useful for development because all the rest of the hardware is used, including radio link, so a flying UAV can communicate with a UAV in simulation mode on the ground. Furthermore this can be used for fault detection by comparing simulated sensors and real sensors. However, because the simulation runs on the microcontroller of the autopilot, it has to be computationally efficient. The simulation is currently limited to quadcopters. The goal of this project is to implement the simulation of the flight dynamics of a fixed wing UAV on MAVRIC autopilots. A good trade-off between accuracy and computational complexity will have to be found. The resulting simulation will be first tested on the ground, then compared to the actual dynamics of an existing fixed wing UAV.

Type: Semester project
Period: 05.01.2016 - 06.08.2016
Section(s): EL IN MT
Type of work: 40% theory, 40% software, 20% experiments
Requirements: C/C++ programming, basics of flight mechanics
Subject(s): Flight mechanics, embedded programming, fixed wing UAV
Responsible(s): Julien Lecoeur, Grégoire Hilaire Marie

Optic flow based estimation of angle of attack on a VTOL UAV

Brice Platerrier (MT)

At LIS, we developed the Ywing : a VTOL (vertical take off and landing) drone capable of flying at any speed between its cruise speed and hover by varying its angle of attack. The angle of attack (angle between wing chord and airflow) cannot always be approximated by pitch angle (angle between wing chord and horizon) because the drone is not always flying at constant altitude. The goal of this project is to use optic flow measurements from a wide field of view camera embedded on the drone to estimate the angle of attack. Indeed, in translation flight, all optic flow vectors point to the current direction of flight, which can be used to estimate angle of attack. The estimation of angle of attack will be implemented on the onboard controller using optic flow vectors computed on the camera, and validated in flight. If time allows, this estimate will be used to control the altitude of the Ywing.

Type: Semester project
Period: 06.01.2016 - 06.08.2016
Section(s): IN MT
Type of work: 20% theory, 50% software, 30% experiments
Requirements: C/C++ programming, 3D geometry
Subject(s): vtol uav, optic flow, great circles, angle of attack
Responsible(s): Julien Lecoeur

Interface of Mavric autopilots with motion tracking system

Tracchia Tommaso (MT)

Motion tracking systems are more and more popular in flying robotics. By detecting the position of IR reflective markers on a set of high speed cameras, commercially available motion tracking software are capable of tracking and streaming the 3D position and orientation of a rigid object. This provides solid ground truth for experiments with autonomous flying robots. The goal of this project is to interface a motion tracking system with the MAVRIC drones developed as LIS. The first task will be to forward tracking data to the drone via wireless communication. This data will be used onboard to close the control loop and provide safety features such as wall avoidance and fallback controller in case of emergency. The main challenge to tackle will be the latency introduced by the tracking software, the packet forwarding software, and radio link. If time allows, the student will implement additional features such as automatic identification of each drone in a swarm.

Type: Semester project
Period: 06.01.2016 - 06.08.2016
Section(s): IN MT
Type of work: 60% software, 30% hardware, 10% theory
Requirements: C/C++ programming, basic networking
Subject(s): Motion tracking system, flying robots
Responsible(s): Julien Lecoeur, Alexandre Cherpillod

A lightweight and modular Ground Control Station

Stéphane Ballmer (MT)

Ground control software is a key part of the operation of UAVs, it allows to monitor the current state of the UAV, plot and log telemetry data, and also send commands to the drone. Current ground control station are generally large pieces of software, which are difficult to set-up and modify for custom needs. The goal of this project is to build a lightweight alternative to current ground control station software. The focus of the development will be on the ease of use, modularity and extensibility. A promising approach is to use a server-client architecture with the interface displayed in the web-browser, the student can take inspiration and reuse open source libraries such as Mavue, MavProxy, Dronekit or Mavelous.

Type: Semester project
Period: 06.01.2016 - 18.07.2016
Section(s): MT
Type of work: 80% software, 20% experiments
Requirements: Good programming level
Subject(s): Human-Drone interface
Responsible(s): Julien Lecoeur, Basil Huber

Bringing RoboGen Online To the Public

Guillaume Leclerc (IN)

At the Laboratory of Intelligent Systems we are actively involved in the development and, maintenance of RoboGen: an open-source software, hardware platform for co-evolving brains and bodies of 3D-printable robots. Recent work has ported the software portion of RoboGen to run completely inside of a web browser while also allowing the computationally expensive fitness evaluations to be distributed to other peers as well as cloud servers., This creates a powerful combination where anyone, anywhere in the world can visit the website and run meaningful experiments in Evolutionary Robotics without the need for complicated installation procedures or having their own access to HPC resources. While the functioning of this system has been succesfully demonstrated, the development process has not yet been completed to the point of releasing the system to the public., This project will complete the final steps to, bring RoboGen online to the public: (a) fix the outstanding bugs and usability issues to make the software robust, (b) add some small additional functionality to make the software more useful for experimentation, and (c) perform user tests with the Student of Micro 515 (Artificial Evolution) who will use the software for TPs and their class mini projects.

Type: Semester project
Period: 22.02.2016 - 30.06.2016
Section(s): IN
Type of work: 60% software, 40% testing
Requirements:
Subject(s): Software, Evolutionary Robotics, WebAL
Responsible(s): Josh Auerbach, Giovanni Iacca
URL: Click here
 

Flywheel device to apply torque feedback for hand orientation

Aloïs Aebischer (MT)

At the Laboratory of Intelligent Systems (LIS) we are investigating the development of more intuitive and immersive controllers for drones. To boost immersion, we aim to develop portable feedback devices to transmit haptic feedback from the drone to the user. The goal of this project is to develop a small and lightweight device, based on the flywheel concept, to give perceptible torque to the user at the hand level. Along this project, the human torque perception and flywheel mechanisms will be studied. Based on this knowledge, a prototype will be designed, manufactured, and tested. Depending on the advancement of the project, an optimization of the torque vs weight will be performed.

Type: Semester project
Period: 22.02.2016 - 03.06.2016
Section(s): ME MT
Type of work: 20% theory, 50% hardware, 30% experiments
Requirements: SolidWorks is an advantage
Subject(s): Haptic feedback, Human-robot interaction
Responsible(s): Carine Rognon, Stefano Mintchev

2015


Development of a new multi-modal robot for hovering and terrestrial locomotion

Florian Reinhard (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and take off. These abilities bring this new flying robot closer to the capabilities of animals, such as bats, that are much more adaptive to their environment than current flying robots. The goal of this project is to conceive a new multi-modal robot capable of hovering, terrestrial locomotion and transition between the two. Based on state-of the art solutions and already available prototypes at LIS, different design solutions will be identified and compared. Secondly, a functional prototype of the robot will be designed and manufactured. Finally, a preliminary assessment of the performance will be performed.

Type: Semester project
Period: 22.02.2016 - 03.06.2016
Section(s): ME MT
Type of work: 20% theory, 40% mechanical designs, 30% hardware and manufacturing, 10% experiments
Requirements: CAD (Solidworks preferred)
Subject(s): Flying robotics, Multi-Modal Locomotion
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
URL: Click here

Drone's repulsive behavior from virtual fence

Cyril Stuber (MT)

At the Laboratory of Intelligent Systems (LIS), we aim to bring the human closer to the drone. By designing a natural interface, emcompassing feedbacks, between a user and a drone, we decrease the cognitive effort dedicated to the interface. This allows a wider diversity of people to use it, and allows people to focus more on the mission than on the interface. The goal of this project is to implement a repulsive behavior for a quadcopter while approaching a virtual geo-referenced fence. As the repulsion actions will be in the future fed back to the user through the natural interface, care will be taken to minimize the dis-comfort created by these actions. The repulsive behavior will be based on an artificial force field. We will first optimize this field to reduce at most the user's discomfort. This will be tested at first in simulation and then in a real experiment.

Type: Semester project
Period: 22.02.2016 - 03.06.2016
Section(s): IN MT PH
Type of work: 30% Theory, 50% Software, 20% Experiment
Requirements: Good knowledge of C, basics of control theory
Subject(s): GPS navigation, Artificial force field
Responsible(s): Alexandre Cherpillod, Grégoire Hilaire Marie

Novel ways to interact with contact-tolerant drones

Arnaud Janvier (MT)

Flyability is an EPFL spin-off dedicated to bringing to the market the first collision-tolerant and safe drone. The company aims at making a significant change in the use of drones by allowing them to enter spaces where no robot can currently penetrate and allow them to fly in contact with humans without risk. The main goal of this master project is to study physical interactions between humans and drones. The drone developed at Flyability can be safely manipulated while flying. This capability enables interactions between the drone and its environment and thus, opens new possibilities. The first task of this project will be to understand how a user can control the drone by other means than a conventional remote control. Then, sensors will be selected and integrated on the drone in order to enable different types of interactions with its environment and with humans. Finally, the last task will be to program the drone in order to control it by the use of these interactions during flight.

Type: Master project
Period: 14.09.2015 - 31.03.2016
Section(s): MT
Type of work:
Requirements:
Subject(s): sensory, mechanical design, programming in C
Responsible(s): Ludovic Daler, Alexandre Cherpillod
 

Visual Sensor Feedback for RoboGen

José Gallardo (MT)

At the Laboratory of Intelligent Systems we are actively involved in the development and maintenance of the open-source RoboGen platform for co-evolving brains and bodies of 3D-printable robots. This platform has been used successfully for class mini-projects in our Master's level class: MICRO-551, Bio-Inspired Artificial Intelligence and is currently being leveraged as research platform for the European project INSIGHT. As part of the INSIGHT project we have been investigating learning techniques that require relatively high dimensional sensor feedback such as that provided by a camera. In order to facilitate these efforts this project will investigate integrating visual input into the RoboGen robots. The main tasks of the project will be as follows: 1. Researching different cameras to find one that is reasonably low cost, the proper size, and could ideally interface with our current Arduino microcontroller (likely through the use of a dedicated microcontroller for image processing / data compression). 2. Once a camera is chosen, actually get it to feed into the robot controller: this will involve some electronics work and a lot of micro controller programming. 3. Program a model of the camera into our simulator. This will involve C, programming to capture the rendered scene from the perspective of the camera.

Type: Semester project
Period: 14.09.2015 - 31.01.2016
Section(s): IN ME MT
Type of work: 20% Literature Review, 40% Hardware, 40% Software
Requirements: Electronics, Programming (C/C++), Embedded Systems
Subject(s): Robotics, Microcontrollers, Simulation
Responsible(s): Josh Auerbach, Basil Huber
URL: Click here

Real-time monitoring and feedback of robot behaviors.

Maxime Esparbès (MT)

At the Laboratory of Intelligent Systems we are actively involved in the development and maintenance of the open-source RoboGen platform for co-evolving brains and bodies of 3D-printable robots. This platform has been used successfully for class mini-projects in our Master's level class: MICRO-551, Bio-Inspired Artificial Intelligence.     Currently RoboGen operates by evolving a robot morphology and controller for a given task entirely in simulation and then transferring the evolved solution to reality by 3D printing the morphology and loading the controller onto the embedded micro controller board. Once the robot has been fabricated and the controller loaded, the robot is free to behave in the real world. Due to the fidelity of the simulator, this process often works well, but other times the real robot fails to replicate the behaviors seen in simulation -- a phenomenon known as the "reality gap."     In order to overcome the reality gap we would like to be able to bring the hardware into the optimization loop by performing fitness evaluations on the real hardware and/or allowing the controller to continue adapting once it has been loaded onto the robot. For both of these approaches we need a robust method of monitoring the performance of real robots and sharing this performance (potentially in real time) with the optimization procedure and the micro controller.     The aim of this project will be to implement these abilities: (a) real-time monitoring of robot behaviors: tracking the position/orientation of robots over time. (b) communicating this tracked information to the evolutionary algorithm. (c) communicating a performance metric and/or new controller parameters to a robot's micro controller wirelessly. (d) Use the above capabilities to optimize a robot controller with some trials on the real robot. (For simplicity, this project will assume that just the control strategy of a fixed morphology robot is being optimized.)

Type: Semester project
Period: 14.09.2015 - 31.01.2016
Section(s): EL IN MT
Type of work: 40%+hardware+40%+software+20%+theory
Requirements: Previous+experience+with+microcontrollers+(arduino)+and+C+++programming
Subject(s): Robotics +Tracking +Communication
Responsible(s): Josh Auerbach, Grégoire Hilaire Marie
URL: Click here

Web Browser Based Distributed Robot Evolution

Guillaume Leclerc (IN)

At the Laboratory of Intelligent Systems we are actively involved in the development and maintenance of the open-source RoboGen platform for co-evolving brains and bodies of 3D-printable robots. This platform has been used successfully for class mini-projects in our Master's level class: MICRO-551, Bio-Inspired Artificial Intelligence.

Recently we have ported the entire RoboGen software stack to run natively inside a web browser., This drastically lowers the barrier to entry to using the software, since it just requires a user to visit a web page rather than downloading and installing the software and all of its dependencies.

In order to make full use of the potential of RoboGen on the web it would be extremely useful to be able to distribute fitness evaluation to servers on the cloud and/or to other users., This project will involve implementing this functionality and exploring strategies for increasing user participation.

Type: Semester project
Period: 01.09.2015 - 31.01.2016
Section(s): IN
Type of work: 20% Theory, 80% Software
Requirements:
Subject(s): Evolutionary Robotics, Distributed Computation
Responsible(s): Josh Auerbach, Giovanni Iacca
URL: Click here

A safe self-stabilizing landing system for a multicopters transporting cargo

Elena-Sorina Lupu (MT)

At the LIS we are developing a flying platform for fast transporting lightweight packages of various sizes with medicaments, dressings for wounds, small equipment, books, etc.

One of the reasons of the largest number of the accidents that occurs with multicopters is caused by failure of the propulsion system of the unit. This might happen as a result of uncharged battery, after collision with obstacle or strong gust of the wind which destabilizes platform. Most of those situations does not damage the aircraft. The biggest destruction occurs during the impact with the ground. Problem increases when drone has to transport payload which also shouldn’t be damaged.

In order to reduce damage of delivered items as much as it is possible, transported cargo will be placed on top of the multicopter and energy absorption elements to protect robot and payload will be installed below the copter. That is why the main goal of this project is to investigate possible solutions to self-stabilize platform in the way that could safely land vertically after partial or even full malfunction of the propulsion system.

First task of this project is to investigate different possible hardware solutions combined with control to stabilize platform during the fall of the aircraft, so it could “land on its legs”.

Second task is to use control algorithms to stabilize platform enough for safe vertical landing using only propulsion system but when one or more motors will stop working (dependent from the platform).

Both tasks will require to work and do tests of proposed solution on real platform. Required mechanical elements have to be designed and installed on the platform. Appropriate software has to be written in the C language. The flight controller will be based on a MAVERICK autopilot which will be placed on quadcopter LEQUAD.

Type: Semester project
Period: 11.09.2015 - 16.01.2016
Section(s): ME MT MX
Type of work: 30% theory, 40% software, 30% hardware
Requirements: Programming microcontrollers in C language, knowledge of CAD software
Subject(s): Flying Robot, transportation of packages, safety landing
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev
 

Design of an adaptive structure for multirotors to transport different sizes packages

Marc André Leroy (MT)

At the LIS we are developing a flying platform for fast transporting lightweight packages of various sizes with medicaments, dressings for wounds, small equipment etc.

The main goal is to design a mechanism, which will allow to adapt the size of different multi-copters accordingly to the size of the carried packages.

First task of this project is to investigate standard dimensions and types of the existing packages in post offices, logistic companies in different countries etc. that could be used to transport different types of goods.

Second task is to design mechanical structure of the robot that could adapt its shape to carry different parcels e.g. available post office boxes or pizza boxes.

It has to be taken into consideration that robot will fly among people and that is why its fast rotating propellers should be protected. Prototype has to be build and test it.

Type: Semester project
Period: 11.09.2015 - 16.01.2016
Section(s): ME MT MX
Type of work: 30%+theory +15%+software +55%+hardware
Requirements: Good+skills+at+CAD+software
Subject(s): Flying Robot, transportation of packages
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev

Prepare mass and power model for an adaptive structure of multirotors

Antoine Tardy (MT)

At the LIS we are developing a flying platform for fast transporting lightweight packages of various sizes with medicaments, dressings for wounds, small equipment etc.

The main goal is to prepare and verify mass and power model for different multi-copters used to transport different sizes of the packages.

First task of this project is to adapt and extend already available existing mass and power models, used to design hovering platforms, in order to choose components of the robot for a different payload and flight duration.

Second task is to build simple quadcopter based on existing mainframe of quadcopter LEQUAD using selected components for given payload and distance and test its capabilities.

The last but not least assignment will be to programme PID controller that could adapt to the different geometries of the robot as function of the payload size. The flight controller will be based on a MAVERICK autopilot (electronic control board and software) developed at LIS.

Type: Semester project
Period: 11.09.2015 - 16.01.2016
Section(s): ME MT MX
Type of work: 30% theory, 30% software, 40% hardware
Requirements: Programming PID controllers
Subject(s): Flying Robot, transportation of packages
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev

Portable localization system for indoor aerial robots.

Xaver Bandi (ME)

Localization is one of the key challenges that needs to be considered to design truly autonomous robots. Knowledge about the three-dimensional position of an aerial robot is essential for allowing it to navigate autonomously to different points in space. Since GPS cannot be used indoors, existing solutions for localization of small indoor aerial robots have been limited to large, expensive and impractical motion tracking systems. The goal of this project is to develop a low cost, small size and portable localization system that can localize small aerial robots simply based on the sound that is emitted from their engines. For this project, the student needs to understand/adapt an audio-based localization module (an embedded system having an AVR32 microcontroller) to implement his design and algorithm. This project is novel and has a high chance of publication at the end.

Type: Master project
Period: 01.09.2015 - 15.01.2016
Section(s): EL IN ME MT MX
Type of work: 25% theory, 35% software, 15% hardware, 20% experiments
Requirements: Experience with micro-controllers and programming. Basic knowledge in signal processing and robot localization problem.
Subject(s): Robot localization, Microcontrollers, Programming, Signal processing,
Responsible(s): Meysam Basiri, Grégoire Hilaire Marie

Bearing-only collision-avoidance for teams of aerial robots

Riço Caldas Bruno Oscar (CH)

At the Laboratory of Intelligent Systems we are developing teams of autonomous aerial robots to accomplish different tasks in a collaborative manner. Robots within an aerial team need to detect their team mates and avoid collision with them. One important information a robot could obtain is the relative direction of its local team mates. This information can be measured by the robots independently with on-board sensors such as cameras or microphone arrays. The goal of this project is to design a mid air collision avoidance strategy for teams of aerial robots based only on the relative bearing information between the robots. For this, the student need to study previous work on multi-robot collision avoidance techniques, propose bearing-only collision avoidance methods, investigate and compare methods in simulation.

Type: Semester project
Period: 01.09.2015 - 15.01.2016
Section(s): EL IN MA ME MT MX
Type of work: Theory 50%, Software 30%, Hardware 20%
Requirements: Mathematics, basic control theory, programming
Subject(s): Mathematics, control theory
Responsible(s): Meysam Basiri, Basil Huber

Wind speed measurement system for aerial robots

Charly Blanc (MT)

At the laboratory of intelligent systems, we are developing small autonomous aerial robots that could be used outdoors and for different applications, such as search and rescue missions, environmental monitoring, aerial surveillance and mapping. To design truly autonomous robots, the robots need to obtain different information about their environment. The aim of this project is to design a wind sensor, that is suitable for small aerial robots and in particular quad-rotors, for measuring the wind speed from these airborne robots. For this project, the student will start by studying the state of the art on wind measurement techniques that are suitable for aerial robots. The student is then required to study the feasibility of using microphone sensors on the robot for measuring the wind speed. This project will involve real world experiments, signal processing and implementation of algorithms on a micro-controller for real-time sensing.,

Type: Semester project
Period: 01.09.2015 - 15.01.2016
Section(s): EL IN MA ME MT MX
Type of work: Theory 30%, Hardware 20%, Software 30%, Experiments 20%
Requirements: Signal Processing, Micro-controller Programming
Subject(s): Signal/Audio processing, Micro-controller Programming
Responsible(s): Meysam Basiri, Grégoire Hilaire Marie

Improving GPS accuracy for our Drones

Steven Junod (MT)

At LIS laboratory we are working on flying robots. In many projects, we are working-on, we need a good position estimation, both for safety reason and for path planning. Nowadays, GPS are cheap and available in very small package. However, their accuracy is limited between, or - 5m horizontally and even more vertically. The precision may also vary due to weather conditions. In a lot of different applications, such as collision avoidance tests, or outdoor flights in cluttered environments, it would be necessary to have a better position estimation. Using differential GPS (DGPS), we could get this better position estimation. It consists in using 2 GPS units. One at a fixed well known position, and a second one on the robot. Since the first one is fixed, it can deduce at any time the GPS position estimation error and forward it to the robot in order to improve its position estimation in real time. The idea of that project is to first get familiarized with DGPS method, and implement the differential GPS method to improve the position estimation. In a second step implement it in our drone framework and characterize its improvement.

Type: Semester project
Period: 15.09.2015 - 12.01.2016
Section(s): IN MT SC
Type of work: 20%+theory +60%software+and+20%outdoor+tests
Requirements: computer+science
Subject(s): Flying+robot +GPS
Responsible(s): Grégoire Hilaire Marie , Nicolas Dousse
 

Autopilot for a fixed wing platform

Simon Pyroth (MT)

At LIS laboratory we are working on flying robots. We are using different platform for different projects. We are aiming at using the same autopilot for all the platforms to ease interaction and speed-up outdoor experiments of new robot or algorithm. In this project, you will have to adapt the code of our own quad-copter to port it for a fixed wing platform. For this you will have to first understand the code architecture of our quad copter. Then you will have to adapt it to the need of a flying wing (non holomic platform) Finally you will have to make it as autonomous as possible, by adding, for example, some take off and landing strategies.

Type: Semester project
Period: 15.09.2015 - 12.01.2016
Section(s): IN MT
Type of work: 20% theory, 60%software and 20%outdoor tests
Requirements: C / C++ and some basics about flying robot
Subject(s): autopilot, control, flying robot
Responsible(s): Grégoire Hilaire Marie , Nicolas Dousse
 

Development of a multi-modal wing for flapping flight and terrestrial locomotion

Fabio Zuliani (MT)

At the LIS we are developing a novel robot capable of aerial and terrestrial locomotion. This new platform will be able to hover, fly, and crawl on the ground using its wings. These abilities bring this new flying robot closer to several animals that are capable to transition between multiple substrates, like the vampire bat Desmodus Rotundus. Like this animal, recruiting the wings for locomotion in both environment and adapting their morphology is a key capability to enable multi-modal locomotion. The goal of this project is to conceive a wing with adaptive morphology for flapping flight and terrestrial locomotion. Based on state-of the art solutions and already available prototype at LIS and BioRob, different design solutions will be identified and compared. Secondly, a simple prototype of wing will be designed and manufactured. The wing will be capable to adapt its morphology during the transition from air to water. A preliminary evaluation of the lift and crawling performances of the multi-modal wing will be performed. The project will be performed in collaboration between the Laboratory of Intelligent Systems (LIS) and the Biorobotics Laboratory (BioRob)

Type: Semester project
Period: 14.09.2015 - 18.12.2015
Section(s): ME MT
Type of work: 20% theory, 20% research, 40% hardware, 20% experiments
Requirements:
Subject(s): Flying robotics, Legged robotics, Multi-Modal Locomotion
Responsible(s): Stefano Mintchev, Peter Eckert
URL: Click here
 

A novel bioinspired strategy for the development of collision tolerant drones

Sébastien Douglas De Rivaz (MT)

At the LIS we are developing drones for search and rescue operations. These drones must be easy to transport and capable to fly in cluttered environments. Therefore the two main challenges are foldability and crash resilience. Recently we have developed a foldable drone capable of autonomous deployment in 0.3 seconds. However, the current design is very fragile against collisions. Starting from the existing prototype, the goal of this project is to exploit foldability to develop drones capable to withstand collisions. Insects use this strategy in order to mitigate the wear of their wings when flying in very compact environment. At first, the new strategy will be modelled and advantages/disadvantages compared to current solutions (e.g. protective carbon fiber cages or foam protections) will be evaluated. Secondly, a prototype of a rugged foldable drone will be developed and characterized.

Type: Semester project
Period: 14.09.2015 - 18.12.2015
Section(s): ME MT MX
Type of work: 20% theory, 20% research, 40% hardware, 20% experiments
Requirements:
Subject(s): Flying Robotics, Bioinspired Robotics, Crash Resilience
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
URL: Click here

Improving Aquatic Capabilities of a Multi-Modal Flying Robot.

Théo-Tim John Denisart (MA)

At the LIS we are developing a novel flying robot, which has the ability to move easily on multiple environments. This new platform is able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of animals, such as bats or birds, which can easily transition between multiple environments. The goal of this project is to improve aquatic locomotion abilities on a flying and diving robot, which has already been designed in a previous project. We want to implement wings with adaptive morphology to facilitate the transition from air to water and vice-versa and to reduce drag underwater. For this project, we will first investigate different technologies for the development of foldable / morphing wings and then we will modify the concepts to fit the requirements of hybrid aerial/aquatic locomotion. The main challenge in this project is thus to conceive, design and develop a flying robot with adaptive morphology capable to maximize locomotion effectiveness (i.e. maneuverability, efficiency) in multiple environments. We will develop morphing wings and implement them on a working prototype for characterization.

Type: Semester project
Period: 14.09.2015 - 18.12.2015
Section(s): MX
Type of work: 20% theory, 20% research, 40% hardware, 20% experiments
Requirements:
Subject(s): Flying robotics, Underwater robotics, Multi-Modal Locomotion
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
 

Variable Stiffness Thread For Flying Robots

Paolo Bertero (ME)

Materials that can drastically change their stiffness are of great interest for engineering applications. This is especially true in robotics, where the ability to dynamically change stiffness allows the development of robots with adaptive morphology. When the material is soft the robot can adapt its morphology, when the material became rigid again the morphology is freezed. We develop variable stiffness structures is by fabricating threads (variable stiffness thread VST) composed of low-melting-point-metals (LMPMs) embedded in silicone tubes. The experiences a large change in stiffness when the LMPM transition from solid to liquid. The challenge of this project is to improve the specific stiffness of the VST in order to use it as frame for flying robots. This can be accomplished by embedding fibers into the VST. During the project the improved VST will be characterized and a frame for a quadcopter will be investigated as a proof of concept.

Type: Semester project
Period: 14.09.2015 - 18.12.2015
Section(s): ME MT MX
Type of work: 20% theory, 20% research, 30% hardware, 30% experiments
Requirements:
Subject(s): Variable Stiffness Materials, Soft Robotics, Flying Robotics
Responsible(s): Stefano Mintchev, Alice Tonazzini
 

Active Learning with Deep Networks

Martin Gammelsaeter (IN)

In many prediction tasks it is easy to collect input data, but expensive to collect targets (e.g. class labels or predictions). For example, when building forward models for robotics, it is possible to generate a large (possibly infinite) number of potential actions, but learning the results of these actions requires actually performing them on the robot and observing the results.      This cost of obtaining targets data is a major hindrance to constructing effective models. Moreover, the data points that can be sampled from a given [hidden] target distribution will contain differing amounts of information for learning. Active learning is the branch of machine learning that seeks to tackle the problem of data scarcity by intelligently choosing unlabelled data points for labelling and training. One effective approach to active learning is called "Query by Committee" (Seung et al, 1992) where a committee of models is queried to find the most informative data point to query: the point that maximizes disagreement among committee members.      In recent years deep neural networks have shown state of the art performance for many prediction tasks. REMAINDER OF DESCRIPTION HIDDEN WHILE WORK IN PROGRESS     

Type: Master project
Period: 19.02.2015 - 31.07.2015
Section(s): IN
Type of work: 50%+theory +50%+software+(++some+hardware+robotic+tests+if+time+allows)
Requirements: Neural+Networks +Backprop
Subject(s): Machine+Learning +Active+Learning +Neural+Networks +Robotics
Responsible(s): Josh Auerbach, Giovanni Iacca

Comparison of Design Strategies for Multi-Modal Locomotion

Darius Constantin Merk (PH)

At the LIS we are developing a new generation of robots with enhanced versatility thanks to Multi-Modal Locomotion capabilities. These robots are inspired by animals that can easily transition between different environments. For example, inspired by the vampire bat “Desmodus rotundus”, we have successfully developed a flying robot with the ability to crawl on the ground. Animals evolved different strategies for multi-modal locomotion. Some species recruit a different locomotor system for each environment, while others exploit a single one adapted for multiple environments. Furthermore, several animal species adapt their morphology during the transition from one environment to the other. Similarly to animals, these strategies have been implemented in multi-modal robots. However, the best design strategy for multi-modal locomotion and associated trade-offs are still unknown. The first goal of the project is to develop a simplified framework that allows to compare different design strategies for multi-modal robots. The second goal of the project is to apply the model to case studies of multi-modal locomotion.

Type: Semester project
Period: 16.02.2015 - 30.06.2015
Section(s): PH
Type of work: 30% theory, 30% research, 40% Modeling
Requirements: Dynamic modeling
Subject(s): Adaptive morphology, biological actuators, animal locomotion
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
 

Design and Manufacturing of Insect Inspired Foldable Wings

Louis Dufour (ME)

Aerial robots provide valuable support in several high-risk scenarios thanks to their capability to quickly fly to locations dangerous or even inaccessible to humans. In order to fully benefit from these features, aerial robots should be easy to transport and rapid to deploy. One solution to solve this transportation issue is to reduce the size of the drones. The shortcoming is a reduction in payload capabilities and in flight time. Therefore, foldable structures, which can be packaged for transportation and quickly deployed for operation, are a promising solution to enable full benefits from the potential of flying platforms. The goal of this project is to investigate insect inspired solutions for the development of foldable wings for small sized flying robots. First, an analysis of the state of the art of both artificial and bio-inspired solutions will be done. Then, based on this analysis a design and a manufacturing process will be proposed and a prototype of the foldable wing will be implemented. Finally, static tests and flight tests will be performed in order to assess the design and suggest improvements.

Type: Semester project
Period: 16.02.2015 - 30.06.2015
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements: Solidworks or similar
Subject(s): Mechanical design, Foldable mechanism, Bioinspired Robotics
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
 

2014


Quality Diversity of Evolvable Robot Morphologies

Pierre Quinton (IN)

At the Laboratory of Intelligent Systems we are actively involved in the development and maintenance of the open-source RoboGen platform for co-evolving brains and bodies of 3D-printable robots. This platform has been used successfully for class mini-projects in our Master's level class: MICRO-551, Bio-Inspired Artificial Intelligence.

While we have had success in evolving robot morphologies with RoboGen, we still see a noticeable lack of diversity of the robot morphologies that have been evolved. Ongoing work at LIS is investigating how changes to the morphological building blocks of RoboGen may aid in evolving a more diverse and interesting set of robot morphologies, but it is also possible to modify the way that artificial evolution functions in order to promote such diversity. Specifically, recent work within the field of Evolutionary Computation has stressed the potential for Evolutionary Algorithms to function not just as optimizers looking for a single fit solution, but as accumulators of a diverse set of well performing solutions. This new focus has been dubbed Quality Diversity (QD) by Pugh et al. In this project we will apply recently introduced QD algorithms towards the ultimate goal of evolving a diverse set of well performing physical robots. The project will involve implementing these algorithms into RoboGen and extensive experimentation to find appropriate dimensions on which to encourage diversity.

Type: Semester project
Period: 15.09.2015 - 31.01.2016
Section(s): EL IN MT SC SV
Type of work: 40%+Theory +60%+Software
Requirements: Proficiency with C++ Programming and knowledge of Evolutionary Computation and Neural Networks
Subject(s): Evolutionary Robotics, Nueral Networks
Responsible(s): Josh Auerbach, Giovanni Iacca
URL: Click here

Enabling a Flying Robot With Aquatic Locomotion Abilities

Louis Moreau-Gentien (MA)

At the LIS we are developing a novel flying robot, which has the ability to move easily on multiple environments. This new platform is able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of animals, such as bats or birds, which can easily transition between multiple environments. The goal of this project is to investigate and enable aquatic locomotion abilities on a flying robot, which has already been designed in a previous project. We want to use the existing actuators of this robot to also move below the water surface. For this project, we will first investigate how we can reuse these actuators to move and steer the robot underwater and then we will modify the design accordingly and make the robot waterproof. The main challenge in this project is thus to make a flying robot waterproof and capable to use the same propulsion system both in the air and in the water. Different techniques used for underwater robot will be studied and adapted to fit the need of our robot. We will build a first prototype capable of flying forward and of moving underwater.

Type: Semester project
Period: 16.02.2015 - 30.06.2015
Section(s): ME MT MX
Type of work: 20% theory, 20% research, 40% hardware, 20% experiments
Requirements: Solidworks
Subject(s): Flying robotics, Underwater robotics, Multi-Modal Locomotion
Responsible(s): Stefano Mintchev, Przemyslaw Kornatowski
URL: Click here

Kin recognition and memory effects in the evolution of altruism

Matvey Khokhlov (IN)

An open question in social evolution is whether kin recognition and memory play a role in the evolution of altruism. Kin recognition, meaning the capability of an individual to acknowledge another individual as kin, might trigger specific behaviour aiming at helping closely related individuals. On the other hand, memory or previous interactions between individuals might lead to cooperative or spiteful behaviour, depending on what was the outcome of previous interactions. The purpose of this project is to test these two hypotheses via in silico evolutionary experiments. A structured population of simulated robots controlled by neural networks will be evolved in a public good production scenario, and the individual level of altruism will be measured. The measured level of altruism when kin recognition and memory are active will be compared statistically with respect to baseline configurations respectively without kin recognition and memory, so to draw conclusions about the two hypotheses.

Type: Semester project
Period: 01.02.2015 - 30.06.2015
Section(s): EL IN MA MT SC SV
Type of work: 30% theory, 70% software
Requirements: Java, Python
Subject(s): evolutionary computation, multi-agent systems
Responsible(s): Giovanni Iacca, Josh Auerbach
URL: Click here

Caste differentiation and the joint evolution of altruism and dispersal

Pedro Amorim (IN)

One of the open questions in social evolution is the reciprocal interaction between altruism and dispersal. Altruism occurs when an individual suffers from a fitness cost in exchange of a fitness benefit for its neighbors (which are typically related individuals). Dispersal occurs when an individual leaves its native colony, for example because of an increased local competition due to lower resource availability. A possible factor affecting the joint evolution of dispersal and altruism is caste differentiation. For instance, in an ant colony foragers might be more prone to disperse than soldiers or mating individuals. On the other hand, foragers typically collect resources and share them with the rest of their colony. In this project, the effect of castes will be investigated in evolutionary robotics experiments, where a structured population of up to 100 robots will be evolved. The robots will be assigned different castes and properties, e.g. the maximum energy they can spend during a generation. The influence of these factors will be analyzed and the resulting levels of altruism and dispersal will be compared against a baseline population with no caste differentiation.

Type: Semester project
Period: 01.02.2015 - 30.06.2015
Section(s): EL IN MA MT SC SV
Type of work: 30% theory, 70% software
Requirements: Java, Python
Subject(s): evolutionary computation, multi-agent systems
Responsible(s): Giovanni Iacca, Josh Auerbach
URL: Click here

Improving the Usability of RoboGen with WebGL

Guillaume Leclerc (IN)

At the Laboratory of Intelligent Systems we are actively involved in the development and maintenance of the open-source RoboGen platform for co-evolving brains and bodies of 3D-printable robots. This platform has been used successfully for class mini-projects in our Master's level class: MICRO-551, Bio-Inspired Artificial Intelligence.

Recent work (http://jaredmmoore.com/WebGL_Visualizer/visualizer.html) has demonstrated the potential for running complex 3D visualizations in the browser using WebGL. This technique allows interactive visualizations of 3D physics simulations without installing any software on the client machine. It could be especially powerful when used to visualize the results of evolutionary search running on a remote cluster.

In this project the student will extend Jared's visualizer for RoboGen and create the necessary communication infrastructure so that RoboGen can send its data to the visualization engine., Ideally this communication can take place in real time, so that a user may view a robot in real time being simulated on a remote server.

Type: Semester project
Period: 01.02.2015 - 15.06.2015
Section(s): EL IN MT SC SV
Type of work: 100% Software
Requirements: Familiarity with C++, and Javascript., Previous experience with WebGL is a plus.
Subject(s): 3D Rendering, Web APIs, Evolutionary Robotics
Responsible(s): Josh Auerbach, Giovanni Iacca
URL: Click here

Design of a methodology for comparison of collision avoidance strategies

Gaëtan Burri (MT)

The myCopter project tends to enable the technologies for Personal Aerial Vehicle. This project proposes an innovative way to overcome the financial and environmental cost of current road transportation by using the 3rd dimension for personal transportation such as commuting. Here at LIS, we focus on collision avoidance strategies in dense environments. With the large number of user in the sky, the risk of collision is growing accordingly and ways to ensure safety should be addressed. Here at LIS, we developed and built quadrotors allowing outdoor multi-MAV tests. The state-of-the-art in collision avoidance is very large, each proposing its own scenario to validate its approach. There are no way to compare existing collision avoidance strategies. The aim of this project is to design a methodology to test different collision avoidance strategies on flying quadrotors in a GPS environment. Ce projet peut être réalisé en français.

Type: Master project
Period: 15.09.2014 - 13.02.2015
Section(s): EL ME MT
Type of work: 80% software, 20% hardware
Requirements: C programming
Subject(s): C programming, collision avoidance
Responsible(s): Nicolas Dousse , Julien Lecoeur

Flight-Initiating Jump in Bats

Charlotte Evéquoz (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly and walk on the ground. These abilities bring this new flying robot closer to the capabilities of animals, such as bats, that are much more adaptive to their environment than current flying robots.

The goal of this project is to investigate how to recruit wings for dynamic terrestrial gaits (running, jumping). The vampire bat Desmodus Rotundus takes off from the ground with a dynamic jump (flight-initiating jump). At first we will analyze the available biological data of the vampire bat during the jump. Secondly we will develop a mechanical model of the jump in order to provide useful information to mimic this capability in a robotic platform. Depending on the advancement of the project, we will build a preliminary demonstrator.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% software/simulation
Requirements: Dynamic modelling (Matlab or Mathematica), simulation, CAD
Subject(s): Biomechanical modelling, jumping robot
Responsible(s): Stefano Mintchev, Ludovic Daler
URL: Click here

Collision avoidance with imperfect sensors

Maryon Grandjean (MT)

The myCopter project tends to enable the technologies for Personal Aerial Vehicle. This project proposes an innovative way to overcome the financial and environmental cost of current road transportation by using the 3rd dimension for personal transportation such as commuting. Here at LIS, we focus on collision avoidance strategies in dense environments. With the large number of user in the sky, the risk of collision is growing accordingly and ways to ensure safety should be addressed. Here at LIS, we developed a real-time simulator allowing to simulate large number of flying agents and test different collision avoidance strategies. Usually collision avoidance strategies assume perfect senors with no delays and no performance decrease with distance. Therefore, the aim of this project is to evaluate in simulation the loss of performance when such unrealistic assumptions are removed. The student will model different realistic senors in the existing simulation framework and will evaluate the loss of performance regarding collision avoidance. Ce projet peut être réalisé en français.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL IN ME MT
Type of work: 80% software, 20% theory
Requirements: ADA programming is an advantage
Subject(s): Collision Avoidance, Simulations
Responsible(s): Nicolas Dousse , Maja Varga

Coordination among PAVs

Clement Kunz (ME)

The myCopter project tends to enable the technologies for Personal Aerial Vehicle. This project proposes an innovative way to overcome the financial and environmental cost of current road transportation by using the 3rd dimension for personal transportation such as commuting. Here at LIS, we focus on collision avoidance strategies in dense environments. With the large number of user in the sky, the risk of collision is growing accordingly and ways to ensure safety should be addressed. Here at LIS, we developed a real-time simulator allowing to simulate large number of flying agents and test different collision avoidance strategies. The aim of this project is to implement coordination strategies to develop global traffic strategies such as no flying zones, flocking, dealing with non cooperative flying vehicles. This work will be carried in simulation. Ce projet peut être réalisé en français.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL IN ME MT
Type of work: 10% theory, 90% software
Requirements:
Subject(s): Collision Avoidance
Responsible(s): Nicolas Dousse , Julien Lecoeur

Implementation of a Controller for Transitioning between Flight and Hover

David Wuthier (ME)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to implement a flight controller for a robot that can fly forward, hover, and move on the ground. The specificity of this robot is that it controls the flight by turning the tips of its wings, this allows to use the same actuators for the different modes of locomotion. This project will involve a theoretical study of the dynamics of such a platform and the implementation of a controller for the forward flight and the hover. At first the robot will be remote controlled and the controller will improve the stabilization during forward flight, autonomously stabilize the platform during hover, and manage the transition between this two types of flight configuration. Depending on the advancement of the project, a GPS-based 3D vector field controller will be used to perform trajectories in 3D which will be composed of hover and forward flight phases (this vector field controller has already been successfully tested on a quadrotor).

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work: 30% theory, 20% research, 50% software
Requirements:
Subject(s): Flying Robot, Flight Control
Responsible(s): Ludovic Daler, Julien Lecoeur
 

Task allocation in swarms

Loris Sandro Aiulfi (IN)

Swarms of robots have a huge potential in distributed sensing. They establish communication network over large area to collect local information about the environment. This local information should be shared among robots and robots should be able to allocate tasks in a distributed way. This means that robots should be able to reach decisions about information they are sharing. The aim of this student project is to investigate existing algorithms in a distributed task allocation, perform literature survey and compare algorithms. Selected algorithms should be implemented and tested in C, /Matlab.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL IN ME MT PH
Type of work: 50% theory, 50%software
Requirements:
Subject(s): Robotics, Computer science, communication systems
Responsible(s): Maja Varga, Nicolas Dousse
 

Design of a Swarm Interface

Thibault Jean Roger Alcouffe (MT)

Here at LIS, we developed a quadrotor and build a few copy. The goal is to perform multiple robot experiments. Setting up experiments with multiple robots is a time-consuming task. The goal of this project is, on one hand, to develop a strategy to high level control multiple robots in order to set rapidly experiments in the future and to control a swarm in a dynamic way, and, on the other hand, to implement this strategy with the help of an existing open-source communication software. At the end, the goal is to demonstrate the strategy with the real platforms. Ce projet peut être effectué en français.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL IN ME MT PH
Type of work: 80%+Software +20%+Theory
Requirements: Programming+language
Subject(s): Human Machine interface, Multi robot experiments
Responsible(s): Nicolas Dousse , Julien Lecoeur

Development of soft jellyfish type underwater robot

Bendicht Grossniklaus (MT)

As a collaboration work between the Laboratory of Intelligent Systems (LIS) and the Microsystems for Space Technologies Laboratory (LMTS), we developed a soft actuator using dielectric elastomer actuators (DEAs), also known as artificial muscle, capable of bending actuation. Now we are interested in developing novel robotic devices based on this new actuator. The purpose of this project is to develop a soft jellyfish type underwater robot based on the actuator. The student will work on design and fabrication of the robot using DEA technology and its fabrication process. After developing the device, it will be characterized on performance such as the swimming speed. Experiments will be carried out in tethered or untethered configuration.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL ME MT MX
Type of work: 20% Theory, 50% Hardware, 30% Experiments
Requirements: Experience of SolidWorks is an advantage
Subject(s): Robotics, Electronics, Materials, Mechanics
Responsible(s): Jun Shintake, Stefano Mintchev

Design of Foldable Spherical Cage for GimBall robot using Tensegrity Structures

Kevin Owen (MT)

At the LIS we are developing a novel flying platform called the GimBall which has the ability to not only fly indoors, but to physically interact with the environment. The GimBall is able to resist collisions with obstacles and still continue flying after a crash without falling to the ground. The robot is able to achieve this goal thanks to a gimbal system implemented in the structure which decouples the outer cage from the inner frame. This means that after a collision the cage can rotate freely around the robot while the inner frame remains stable.

The goal of this project is to design foldable spherical cage for this platform using the concept of Tensegrity Structures. Tensegrity is a structural principle based on the use of isolated components in compression inside a net of continuous tension, in such a way that the compressed members (usually bars or struts) do not touch each other and the prestressed tensioned members (usually cables or tendons) delineate the system spatially. Spherical cage has to be designed in a way that would be totally self-erected from a low-volume disk shape by tensioning cables. Such a capability will be useful for transportation of the robot in the back-pack to the not easily procurable places.

This project will involve identification of different solutions and their comparison based on performance index (weight, rigidity). Prototypes of selected solutions will be manufactured and tested.

Type: Semester project
Period: 17.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work:
Requirements:
Subject(s): 30% theory, 30% research, 40% hardware
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev
URL: Click here

Tiny cameras for flying robots

Raphaël Johannes Charles Mottet (MT)

There is an increasing interest in flying robotics to diminish the size of future drones in order to fly safely in very constrained environments. However, drones have to carry out navigation tasks in these conditions with more limited resources in terms of sensing, computation, or control. Recently developed artificial compound eyes might be a very suitable solution as smart vision sensors for low-energy and computation power flying control. The aim of this project is to develop a fully functional smart mini-sensor that can be easily integrated with flying robots or other platforms. The student will use an existing miniature camera. S/He will design the necessary architecture in terms of PCB and electronic parts. In addition, the student will implement existing algorithms for data processing in the controlling units. The student is encouraged to develop a demo at the end of the project to illustrate the functionality of the device.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL MT
Type of work: 20% design; 30% hardware; 20% experiments; 20% programming
Requirements: microcontroller programming; hardware electronics
Subject(s): vision sensors; optic flow; mobile robots
Responsible(s): Géraud L'Eplattenier, Ramòn Pericet Camara

Design of Foldable Spherical Cage for GimBall robot using kirigami technic

Johann Edmond Bigler (MT)

At the LIS we are developing a novel flying platform called the GimBall which has the ability to not only fly indoors, but to physically interact with the environment. The GimBall is able to resist collisions with obstacles and still continue flying after a crash without falling to the ground. The robot is able to achieve this goal thanks to a gimbal system implemented in the structure which decouples the outer cage from the inner frame containing the propulsion and control units. This means that after a collision the cage can rotate freely around the robot while the inner frame remains stable.,

The aim of this project is to design a foldable spherical cage in order to reduce the volume of the Gimball for transportation purposes. The design will be inspired by the concepts of the origami and of the kirigami, which is a variation of the origami technique that includes cutting of the paper.,

This project will involve a preliminary investigation and comparison of possible solutions for the development of a spherical foldable cage using origami/kiragami techniques. Afterwards, the best solution will be dimensioned, designed (CAD software) and prototyped for characterization.,

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work: 30%+theory +30%+research +40%+hardware
Requirements:
Subject(s): Flying+Robot +Foldable+structures +kirigami +origami
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev
URL: Click here
 

Design of a contact-sensitive flying robot using force sensors

Florent-Valéry Coen (MT)

At the LIS we are developing a novel flying platform called the GimBall which has the ability to not only fly indoors, but to physically interact with the environment. The GimBall is able to resist collisions with obstacles and still continue flying after a crash without falling to the ground. The robot is able to achieve this goal thanks to a gimbal system implemented in the structure which decouples the outer cage from the inner frame. This means that after a collision the cage can rotate freely around the robot while the inner frame remains stable. Additionally thanks to its spherical shape robot can roll on walls and ceiling to find its way or floor to save energy.

The goal of this project is to work on the contact sensing ability of the Gimball, so that the robot is able to detect when an external force is applied anywhere on its external structure. The first task is to evaluate the intensity of the collision on the GimBall. The second task of the project is the choice of force sensors for contact detection, and how to integrate them to the existing outside structure of the platform in a way that they could detect collisions and their directions. Afterward, the sensing capabilities will be experimentally assessed on a real platform or using a test bed.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL MT
Type of work: 50%+hardware +20%+testing +30%+electronics
Requirements:
Subject(s): sensing +flying +robotics
Responsible(s): Przemyslaw Kornatowski, Stefano Mintchev
URL: Click here
 

Multi-modal Locomotion in Aquatic Birds

Darius Constantin Merk (PH)

At the LIS we are developing multi-modal robots that have the ability to move over different substrates. The common guillemot, Uriaa Alge, is a diving bird that exploits deployed wings for flight and partially folded wings for swim. The transition from air to water is extremely challenging due to changes in the physical property of the environment: water has 800 time greater density and 15 times greater viscosity compared to air. Therefore, the locomotion in such different environments imposes conflicting evolution pressure on the animal. However, the common guillemot is perfectly evolved to live in-between the two media. This locomotion versatility is even more intriguing considering that the muscles are actuators optimized to work in a very limited range of force (stress) VS contraction velocity (strain). However, beside the different physical properties and the potentially different locomotion dynamics during flight and swim, the common guillemot always recruits the pectoralis muscles for each locomotion modes.

The project aims to investigate the hypothetical synergy between wings adaptive morphology and muscles. The first step is to develop a first order dynamic model of both locomotion modes in order to understand the synergies between wings' adaptive morphology and muscles. Based on simulation results, the second step is to abstract bioinspired design principles to implement more advanced multi-modal robots.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): PH
Type of work: 30% theory +30% research +40% Modelling
Requirements: System dynamics, fluid structure interaction, biomechanics
Subject(s): Adaptive morphology, biological actuators, animal locomotion
Responsible(s): Stefano Mintchev, Ludovic Daler
URL: Click here
 

Implementation of a Controller for Transitioning between Ground and Hover

Yannick Poffet (ME)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to implement a flight controller for a robot that can fly forward, hover, and move on the ground. The specificity of this robot is that it controls the flight by turning the tips of its wings, this allows to use the same actuators for the different modes of locomotion. Furthermore, the robot uses these same actuators to upright on the ground in order to transition between ground locomotion and hover. This project will involve a theoretical study of the dynamics of such a platform and the implementation of a controller for the hover. The robot will be remote controlled and the controller will be used to take-off from the ground and to autonomously stabilize the platform during hover. Depending on the advancement of the project, the take-off/hover controller will be tested on different terrains to see if the robot can take-off from uneven surfaces.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work: 30% theory, 20% research, 50% software
Requirements:
Subject(s): Flying Robot, Flight Control
Responsible(s): Ludovic Daler, Julien Lecoeur
 

Online Evolution of Neural Features

Martin Gammelsaeter (IN)

In a recent publication: Auerbach, Fernando and Floreano (2014), we introduced the idea of using competing "neuronal replicators" to aid in solving an online learning task. Here, the units of replication are individual hidden units of an artificial neural network (ANN) that is actively learning from a stream of data, similar to a robot acting in the world. In that publication, we showed the utility of neuronal replication on a toy problem with some simplifications (binary inputs, binary hidden unit activations, uniformly distributed inputs), but we would like to extend this method to more complex, real-world problems where those simplifications are relaxed. This project will involve extending the existing system to perform on more complex problems. This will involve first implementing an indirect encoding for the feature weights so that larger dimensional problems are evolvable. The next step will involve experimentation with parameters, fitness functions, diversity mechanisms, etc. to discover a setup that can (hopefully) outperform other online learning frameworks.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): EL IN MT SC SV
Type of work: 40% Theory, 60% Software
Requirements: Programming, Evolutionary Computation, Neural Networks
Subject(s): Evolutionary Computation, Nueral Networks, Machine Learning
Responsible(s): Josh Auerbach, Giovanni Iacca
URL: Click here
 

Handling equality constraints by viability evolution

Amr Arbani (MT)

Viability Evolution is an abstraction of artificial evolution which operates by eliminating candidate solutions that do not satisfy viability criteria. In a constrained optimization problem, these criteria are naturally defined as boundaries on the values of objectives and constraints of the problem at hand. One of the key concepts of Viability Evolution is that by adapting these boundaries, it is possible to drive the search towards desired feasible regions of the solution space. An open question in evolutionary computation is how to handle properly equality constraints, since these constraints reduce the size of the feasible space to a zero-volume region. An unexplored feature of the viability abstraction is the possibility to add or remove viability criteria dynamically. This aspect may be exploited, for instance, to temporarily relieve one or more equality constraints, thus allowing solutions to overcome local minima which are due to the shape of the constraints themselves. The purpose of this project is to understand if advantages can be gained in the solution of equality-constrained optimization problems by means of viability evolution. A simple viability evolutionary algorithm that can add and remove dynamically equality constraints will be implemented and tested on a benchmark fitness landscape characterized by one or multiple constraints of this kind. A comparison with baseline results will be performed to validate the proposed approach.

Type: Semester project
Period: 29.09.2014 - 31.12.2014
Section(s): EL IN MA MT SC SV
Type of work: 30% theory, 70% software
Requirements: Any programming language, preferably Matlab
Subject(s): evolutionary computation, stochastic optimization
Responsible(s): Giovanni Iacca, Andrea Maesani
URL: Click here

Low-level collision-free navigation with tiny vision sensor

Abdelhak Amine Bensalah (MT)

In LIS, we are interested in developing tiny components to be implemented in future microrobots, such as microflyers. These tiny components would save weight and size with the cost of much lower capabilities. Thus, new strategies have to be developed for the robots to still perform complex tasks with lower resources. In this project, we aim at implementing low-level collision-free navigation to a wheeled robot assisted only by a tiny very-low-resolution camera. The camera possess only three pixels and is aimed at extracting an optic flow vector. The student will optimize the existing algorithms for real-time data processing to get the optic flow, according to the robot navigation characteristics. The student is expected to design the mechanical attachment of the camera to an e-puck robot as well as the communication with the central processor. He will aim at implementing control commands to perform tasks like corridor or wall following and avoidance. A final demo, consisting of the autonomous low-level collision-free navigation of the robot, would validate the implementation.

Type: Semester project
Period: 15.09.2014 - 19.12.2014
Section(s): EL MT
Type of work: 40% software (control and communication); 20% hardware; 30% experiments; 10% data processing
Requirements: microcontroller programming; some mobile robotics knowledge
Subject(s): bio-inspired robotics; mobile robotics; vision sensing
Responsible(s): Géraud L'Eplattenier, Ramòn Pericet Camara

Onboard simulation of a VTOL UAV

David Leydier (IN)

At the laboratory of intelligent systems, we develop flying robots using inspiration from biology. We built a prototype inspired from the way insects fly. This prototype is able to fly at any speed between hovering and fast forward flight, it is controllable and capable of aggressive maneuvers at any flight speed.

The goal of this project is to implement a simulation mode on an existing autopilot. The goal is to enable the fine tuning of the flight controller while the robot is on the ground. In simulation mode, the autopilot will emulate sensors values according to the dynamics of the platform.

The student will have to parametrize the geometry of the VTOL platform and model its dynamics accordingly. Then, the simulator will be implemented and optimized to be executable on the on-board micro-controller.

Ce projet peut être réalisé en français.

Type: Semester project
Period: 07.02.2014 - 15.07.2014
Section(s): EL IN MT PH SC
Type of work: 20% Theory, 50% Software, 30% Hardware
Requirements:
Subject(s): Simulation, Embedded, flight dynamics
Responsible(s): Julien Lecoeur, Nicolas Dousse

Active rolling for acrobatic autonomous flight

Nikita Filippov (EL)

At the laboratory of intelligent systems, we are developing a flying robot that is able to fly straight as well as upside-down and at any intermediate roll orientation. Thus, by actively rotating the robot during flight, its sensors can cover the environment at 360°, providing more information to avoid collisions with limited additional weight. The goal of this project is to implement a flight controller for active rolling flight.The student will have to adapt the code of an existing autopilot and enable the control of the UAV while it is rolling at constant rate. In the end of the project, the student will demonstrate human-piloted flights during which the robot constantly rolls and the human pilot controls it as if it was not rolling.

Type: Semester project
Period: 09.02.2014 - 15.07.2014
Section(s): CH EL IN MA MT PH
Type of work: 20%Theory, 40% Software, 40% Experiments
Requirements:
Subject(s): Flight control, Aerobatics, Embedded programming
Responsible(s): Julien Lecoeur, Grégoire Hilaire Marie

Development of soft bio-inspired jellyfish type underwater robot Enseignant(s)

William Conus (MT)

As a collaboration work between the Laboratory of Intelligent Systems (LIS) and the Microsystems for Space Technologies Laboratory (LMTS), we developed a soft actuator using Dielectric Elastomer Actuators (DEA) capable of bending actuation. Now we are interested in developing novel robotic devices based on this new actuator. In recent years, several smart materials such as Shape Memory Alloys and Electroactive Polymers have been applied to bio-inspired underwater robots because their simple direct drive mechanism imitates biomechanical behavior, which is often difficult by conventional motors and gears. However, despite their promising properties, DEA have been used only in a few cases. The purpose of this project is to develop a bio-inspired jellyfish type underwater robot. The student will work on design and fabrication of the robot using DEA technology and its fabrication process. After developing the device, it will be characterized on performance such as the swimming speed. Experiments will be carried out in tethered configuration.

Type: Semester project
Period: 17.02.2014 - 30.05.2014
Section(s): EL ME MT MX
Type of work: 10% Theory, 60% Hardware, 30% Experiments
Requirements: Experience of SolidWorks is an advantage
Subject(s): Robotics, Electronics, Materials, Mechanics
Responsible(s): Jun Shintake, Bryan Schubert

Ground Controller for a Flying and Walking Robot with Adaptive Morphology

Ismaël Marc Zeâf (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to enable a robot, which can fly forward like a plane, hover vertically like an helicopter and roll on the ground using its wings, with autonomous capabilities. The current version of the robot also has deployable wings which allow him to change its morphology in order to be more efficient while moving on the ground. Furthermore, this mechanism can be used during walking to increase the steps of the robot. A first controller has been implemented to control the robot on the ground. The goal of this project is to improve this controller in order to control as well this deployable mechanism. Then we will use this controller to investigate different walking gaits in order to see which are more efficient. The final goal of this project will be to show that the robot can autonomously and efficiently navigate on rough terrains.

Type: Semester project
Period: 17.02.2014 - 28.05.2014
Section(s): ME MT MX
Type of work: 20% theory, 40% software, 20% hardware, 20% experiments
Requirements:
Subject(s): Flying Robot, Walking Robot, Control
Responsible(s): Ludovic Daler, Grégoire Hilaire Marie

Design and Manufacturing of a Foldable Micro Flying Robot

Dimitri Mazourenko (MT)

At the LIS we are interested in the design of miniature flying robots for their amazing flight maneuverability. The goal of this project is to design and manufacture a miniature foldable flying robot that could fit easily in a pocket. The design will be based on the one of the "Walkera QR Ladybird Ultra Micro Quadcopter" or similar. Once deployed the robot should have the same flight abilities as the Ladybird (same span, maneuverability, and impact robustness) and in the folded configuration its volume must be reduced to a minimum. As a first step, the foldable mechanism will be actuated by hand and in a second step, active autonomous deployment will be investigated. This project will involve the design and dimensioning of this foldable micro flying robot, CAD design, and manufacturing of one or several working prototypes.

Type: Semester project
Period: 17.02.2014 - 28.05.2014
Section(s): ME MT MX
Type of work: 30% theory, 30% research, 40% hardware
Requirements:
Subject(s): flying micro robot, foldable mechanism
Responsible(s): Ludovic Daler, Bryan Schubert

2013


Enabling a Walking, Flying and Hovering Robot with Aquatic Locomotion Abilities

Eric Ansgar Unnervik (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of animals, such as bats, that are much more adaptive to their environment than current flying robots.

The goal of this project is to enable a walking, flying, and hovering robot with aquatic locomotion abilities. A robot which has already been designed in a previous project can walk on the ground, fly forward, and hover. We now want to use the existing actuators of this robot to also move on the water surface. For this project, we will first investigate how we can reuse these actuators to move and steer the robot on water and then we will modify the design accordingly and make the robot waterproof. The main challenge in this project is thus to make a flying robot waterproof. Different techniques used for underwater robot will be studied and adapted to fit the need of our robot. Depending on the advancement of the project we will build a prototype capable of flying forward and of moving on water.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements:
Subject(s): Flying Robot, Waterproof Robot, Mechanical Design
Responsible(s): Ludovic Daler, Stefano Mintchev
URL: Click here

An Evolved Controller for a Flying and Walking Robot with Adaptive Morphology

Noémie Laure Gwendoline Jaquier (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to enable a robot, which can fly forward like a plane and also walk on the ground using its wings, with autonomous capabilities. The current version of the robot also has deployable wings which allow him to change its morphology in order to be more efficient while moving on the ground. Furthermore, this deployable mechanism can be used during walking to increase the steps of the robot. A first controller has been implemented to control the robot on the ground. The goal of this project is to improve this controller in order to control as well this deployable mechanism during walking. To do so, the robot will be simulated in an existing physics based simulator. Then, the synchronization of the motors used for the deployable mechanism and for the walking will be evolved in the simulator in order to find an optimal gait. The final goal of this project will be to show that the robot can evolve different gaits depending on the terrain. Finally, depending on the advancement of the project, these different gaits will be tested on the real robot.

Type: Semester project
Period: 15.09.2014 - 16.01.2015
Section(s): IN ME MT
Type of work: 30% theory, 50% software, 10% hardware, 10% experiments
Requirements:
Subject(s): Flying/Walking Robot, Evolutionary Controller
Responsible(s): Ludovic Daler, Josh Auerbach
URL: Click here

Audio-based Control of Micro Air Vehicles

Walid Amanhoud (MT)

We are interested in the design of an on-board audio-based system for our Micro Air Vehicles (MAVs) that allows them to perceive acoustic targets in the environment and furthermore navigate relative to these targets. Some potential applications of such a system are:
• Formation control:, Allowing a group of flying robots to follow and/or maintain predefined formations relative to a sound emitting leader robot.
• Target Pursuing: To pursue an acoustic target on the ground, such as a person in need of help who is blowing into a safety whistle or having a personal alarm.
• Interaction with a human operator:, Allowing a human operator to use acoustic signalling to control the motion of robots (eg: commanding it to land in a desired spot)
The goal of this project is to develop an on-board audio-based motion control system for a micro air vehicle capable of navigating the robot relative to a desired sound source. For this project, the student needs to adapt the current version of a sound source localization module (an embedded system having an AVR32 micro controller) and to interface this module with the robot’s on-board computer to plan and send appropriate navigation commands to the robot.

Type: Semester project
Period: 15.02.2014 - 15.06.2014
Section(s): EL IN ME MT MX
Type of work: 20% theory, 50% software, 10% hardware, 20% experiments
Requirements: Applicants must be proficient in C programming and have some experience with microcontrollers and electronics.
Subject(s): Microcontroller programming, Signal processing, Control
Responsible(s): Meysam Basiri, Grégoire Hilaire Marie

Active environment scanning for optic-flow based reactive flight control

Anwar Quraishi (MT)

At the laboratory of intelligent systems, we develop flying robots using inspiration from biology. The goal of this project is to implement an obstacle avoidance strategy inspired from insects on a flying robot. Optic flow, which is the apparent visual motion generated as an observer moves through the world, can be used to estimate the distance to obstacles and to react before a collision. The novelty of this project is that the robot will use a single optic flow sensor with limited field of view. We built a flying robot that is able to fly straight as well as upside-down and at any intermediate roll orientation. Thus, by actively rotating the robot during flight, its optic flow sensor will cover the environment at 360°, providing enough information to avoid collisions with limited additional weight. The project involves the interfacing of an optic flow sensor with an existing autopilot. The main challenge is the implementation of an algorithm that, as the sensor rotates, updates a map of optic flow vectors generated by the surrounding environment. The speed of rotation, frequency of optic flow extraction, as well as field of view discretization will be optimized.

Type: Semester project
Period: 22.01.2014 - 10.06.2014
Section(s): EL IN MT
Type of work: 50% Software, 40% Experiments, 10% Electronics
Requirements: good knowledge of C, basics of control theory
Subject(s): Obstacle avoidance, Optic flow, Active scanning
Responsible(s): Julien Lecoeur, Grégoire Hilaire Marie

Flight control of a small UAV capable of smooth transition between hovering and forward flight

Nicolas Vaucher (MT)

At the laboratory of intelligent systems, we develop flying robots using inspiration from biology. We built a prototype inspired from the way insects fly. This prototype is able to fly at any speed between hovering and fast forward flight, it is controllable and capable of aggressive maneuvers at any flight speed. The goal of this project is to implement an adaptive control architecture on an existing autopilot. The goal is to enable the fine tuning of the flight controller for any angle of incidence. The student will start by modeling the basic behaviour of the prototype in flight. Then, he will implement one of the state of the art adaptive control techniques. Finally, flight tests will be performed in order to assess the efficiency of the control.

Type: Semester project
Period: 22.01.2014 - 10.06.2014
Section(s): EL IN MT
Type of work: 40%Theory, 30% Software, 30% Implementation
Requirements: flight control, good knowledge of C
Subject(s): Adaptive control, flight control
Responsible(s): Julien Lecoeur, Ludovic Daler

Autonomous GPS navigation on a fixed wing UAV

Lukas Hostettler (MT)

At the laboratory of intelligent systems, we developped an autopilot that performs the stabilisation of a quadrotor robot. It is also capable of GPS navigation, which consists of guiding the UAV along predefined GPS waypoints. Fixed wing UAVs, contrary to multicopter, are not able to hover and need a minimum speed to keep on flying. This imposes additionnal constraints for the GPS navigation because the UAV cannot stay on the last GPS waypoint. The goal of this project is to use the existing autopilot and adapt it to fixed wing platforms. The student will start by adapting the low level control to the new platform. Then, he will implement a GPS navigation technique that is suitable for fixed wing design, such as vector field path following. Test flights will be performed to assess the reliability of the implemented algorithm.

Type: Semester project
Period: 23.01.2014 - 06.06.2014
Section(s): EL IN MT
Type of work: 20%Theory, 40% Software, 40%+Experiments
Requirements: good knowledge of C, basics of control theory
Subject(s): GPS waypoint navigation, flight control
Responsible(s): Ludovic Daler, Julien Lecoeur

Exploration of application based on artificial muscle

Geoffroy Le Pivain (MT)

As a collaboration work between the Laboratory of Intelligent Systems (LIS) and the Microsystems for Space Technologies Laboratory (LMTS), we developed a foldable actuator using Dielectric Elastomer Actuators (DEA) capable of 1-DOF antagonistic actuation. Now we are interested in developing novel robotic devices based on this new actuator. Adding foldability to robotic devices enhances their usability by making them easier to transport. An example application of this robotic technology is a manipulator used in a remote place such as space. For space missions the folding capability is necessary to carry the robot by rocket, where the volume capacity is limited. An example is EPFL’s CleanSpace One where a set of deployable arms are used. The purpose of this project is to develop a foldable robotic arm composed of several DEAs connected in series. The student will work on design and fabrication of the robot using DEA technology and its fabrication process. After developing the device, the student will perform characterization of the actuator performance.

Type: Semester project
Period: 17.02.2014 - 30.05.2014
Section(s): EL ME MT MX
Type of work: 10% Theory, 60% Hardware, 30% Experiments
Requirements: Experience of SolidWorks is an advantage
Subject(s): Robotics, Electronics, Materials, Mechanics
Responsible(s): Jun Shintake, Bryan Schubert

Morphology Optimization of a Multi-Modal Walking, Flying and Hovering Robot

Quentin Vichard (ME)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, walk on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to model a multi-modal robot capable of walking, flying, and hovering in order to optimize its morphology. Many versions of this robot have been designed in previous projects but different geometric parameters can still be optimized in order to improve the efficiency of the robot in the different modes of locomotion. The first task is to establish a mass and power model of the robot based on two existing models, one developed for a flying wing platform and an other one for a hovering quad-rotor. This model can then be used to optimize the geometrical parameters of the robot and its scale.

Ce projet peut être réalisé en français.

Type: Semester project
Period: 17.02.2014 - 28.05.2014
Section(s): EL IN ME MT MX SC SV
Type of work: 30% theory, 30% research, 40% software
Requirements:
Subject(s): Multi-Modal Robot, Morphology Optimization, Mass and Power Model
Responsible(s): Ludovic Daler, Julien Lecoeur
URL: Click here

3D Variable Stiffness Microstructures

Bertrand Buisson (ME)

Materials that can drastically change their stiffness are of great interest for engineering applications. This is especially true in robotics, where the ability to dynamically change stiffness can enable a robot to become soft to squeeze through small openings but become rigid when supporting loads.

One approach to creating variable stiffness is by fabricating microstructures composed of low-melting-point-metals (LMPMs) embedded in a soft membrane. By transitioning the LMPM from solid to liquid, the overall device experiences a large change in stiffness. The soft membrane prevents loss of liquid LMPM and defines the shape of the structure.

The challenge of this project is to design and fabricate PCM microstructures that maximize stiffness changes while limiting power consumption and transition times. This can be accomplished by creating low-density, 3-dimensional microstructures that have high strength, fast heating/cooling and low mass. The creation of this novel material will require the creative use of advanced fabrication processes available in the LIS and CMi.

Type: Semester project
Period: 17.09.2013 - 31.01.2014
Section(s): CH EL ME MT MX PH
Type of work: 30% design, 50% fabrication, 20% testing
Requirements: Microfabrication experience is a plus.
Subject(s): Materials, Electronics, Microfabrication
Responsible(s): Bryan Schubert, Jürg Markus Germann

Evolution of insect walking.

Thibault Asselborn (MT)

How have insects evolved to walk adaptively in complex environments? Answers to this question will greatly advance the development of robust terrain navigation in miniature bio-inspired robots.

For this project the student will use genetic algorithms to artificially evolve a "computational fly” that emulates insect walking in a 3D world simulator. They will test resulting gaits in a physical hexapod robot.

We will examine the resulting control strategies to provide a view on biological solutions to this problem as well as a means to build robotic controllers that achieve the resilience of insect walking behavior.

This project will be done in the Laboratory of Intelligent Systems (EPFL) and in collaboration with the Benton Lab (UNIL). Therefore this is an extremely unique project right at the interface between engineering, computer science, & neurobiology.

Type: Semester project
Period: 04.09.2013 - 31.01.2014
Section(s): IN MA MT PH SC SV
Type of work: 60% software 30% research 10% theory
Requirements: C++/Python
Subject(s): Genetic Algorithms, Neural Networks, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Andrea Maesani
URL: Click here

Design and Manufacturing of Protective Structure for a Hovering Robot

Pierre Jacques François Lourdais (ME)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to fly indoors and to physically interact with its environment. The AirBurr is able to resist collisions with obstacles and go back in flight after a crash.

The current prototype has a cage that protects it from big obstacles such as wall, but not from small obstacles that can easily reach the inside of the structure. The goal of this project is to design and manufacture additional protections for the existing cage of the AirBurr which should protect it much better against small objects or wires that could get into the structure and damage it. For example special mesh will also help to protect people from inadvertently getting into contact with fast-rotating propellers in the core of the robot.

The first task of this project is to investigate what kind of lightweight materials, fabrication techniques, and mechanical structures could be used to protect the robot and then test it (test bench with precise force sensor will be provided) to check if proposed structures don’t disturb the airflow of the robot which allows hovering.

The second task of the project is to manufacture and implement chosen structure on the real robot and test its performance in the real scenarios.

The project will involve an analysis of disturbance of the airflow, selection and characterization of materials, CAD design of 3D-printed connection elements and the construction of one or several prototype protective structures.

Type: Semester project
Period: 17.09.2013 - 24.01.2014
Section(s): EL MT
Type of work: 20% theory, 30% research, 50% hardware
Requirements: Basics of aerodynamics
Subject(s): Hovering Robot, Mechanical Design
Responsible(s): Przemyslaw Kornatowski, Adrien Briod
URL: Click here

Information consensus in swarms of flying robots

Francesca Sorba (MT)

Swarms of flying robots have enormous potential in distributed sensing., They are usually deployed in outdoor environments, they establish communication network over large area to collect local information about the environment. This local information should be shared among robots and robots should be able to allocate tasks in a distributed way. This means that robots should be able to reach a consensus and decisions about information they are sharing. The aim of this student project is to investigate existing algorithms in a distributed decision-making and consensus and propose the methodology to solve following problems in ad-hoc aerial networks: What is the topology of the communication network? Which agent has the highest energy level? Which agent is the closest to the predefined target? How will agents synchronize their observations in mutual observation map? For each consensus and decision-making strategy, student should measure time it takes to reach the decision and the network load in the process of reaching the decision and consensus. The algorithms should be implemented and tested in simulation, using Argos simulator and NS3 network simulator. Algorithms should be implemented on the flying platforms and tested in outdoor environment for different network topologies.

Type: Semester project
Period: 17.09.2013 - 24.01.2014
Section(s): EL IN ME MT PH
Type of work: 30% theory, 50% software, 20% hardware
Requirements:
Subject(s): Robotics, Wireless networks, Physics
Responsible(s): Maja Varga, Iliana Spartali

Active mechanism for flexible vision sensor

Raphael Valceschini (MT)

The development of novel flexible vision sensors has opened up a new avenue of more resilient and compact components to be implemented in soft robots, revealing many advantages with respect to traditional stiff single-aperture cameras. Such flexible vision sensors could exploit their physical characteristics to tune their functionality in line with the robot specific needs. Thus, new solutions have to be engineered in order to implement this capabilities onboard a robot. The goal of the project is to design a mechanism that makes active bending of a flexible camera possible. The student will make use of a recently developed flexible compound vision sensor in LIS. The student will look into mechanical methods to actively bend the flexible sensor and change configurations. The final prototype will be integrated with mobile e-puck robot.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL ME MT
Type of work: Research 10%; design 30%; fabrication 40%; testing 20%
Requirements: imagination; experience in microfabrication and electronics is an advantage; basics in solid works and matlab are a plus
Subject(s): flexible sensors; soft robotics; soft actuation; mobile robotics
Responsible(s): Ramòn Pericet Camara, Jürg Markus Germann
URL: Click here

Design of Deployable Wings for a Rolling, Flying and Hovering Robot

Jean-Charles Gasche (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design and manufacture deployable wings for a robot that can fly forward like a plane, hover vertically like an helicopter and roll on the ground using its wings. A previous version of the robot can hover, fly forward, and roll on the ground; this prototype has a fixed wingspan which makes it not very maneuverable on the ground. Another version of the robot has deployable wings which make it more efficient on the ground and allows it to go through narrow openings smaller than its wingspan. This project will include a new design of the robot to merge these two platforms into one prototype which will be capable of three modes of locomotion and will have a deployable mechanism that allows to reduce the wingspan of the robot. This deployable mechanism must be as lightweight as possible, must be quick and must not impair the hover and flight capabilities of the platform. Furthermore, it should enhance the hovering and rolling capabilities of the robot, in terms of stability in the air and maneuverability on the ground. This project will involve the design and dimensioning of the wings, CAD design, and manufacturing of one working prototype.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements:
Subject(s): Flying Robot, Hovering Robot, Deployable Wings
Responsible(s): Ludovic Daler, Julien Lecoeur

A Flying Robot Capable of Autonomous Aggressive Maneuvers

Eric Sen Nguyen Van (ME)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to do autonomous aggressive maneuvers with a flying wing platform. The robot that will be used in this project can fly forward, and roll on the ground by using its wings. However, the robot uses different actuators for rotating its wings to roll on the ground and for the control of the flight. The goal of this project will then be to control the flight by turning the wings of the robot (when the wings are used to control the flight they are called "wingerons"), this will allow to remove the flaps on, the wings, to do more aggressive maneuvers with the platform, and to use the same actuators for both modes of locomotion. This project will involve a theoretical study of the aerodynamics of such a platform, and the design, dimensioning, and manufacturing of a test prototype. The main challenge of this project will be to design a controller that allows to realize autonomous aggressive maneuvers with the platform. In this project, we will also test different geometries and sizes of wingerons in flight; the on-board IMU will be used to detect a stall of the wingerons, and these results will be compared to the theory.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL IN ME MT MX SV
Type of work: 30% theory, 30% research, 20% hardware, 20% software
Requirements:
Subject(s): Flying Robot, Aggressive Flight Maneuvers, Flight Control
Responsible(s): Ludovic Daler, Julien Lecoeur

Improving the Robustness of a Rolling, Flying and Hovering Robot

Pablo Klemm (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to improve the robustness and reduce the weight of a robot, which can fly forward like a plane, hover vertically like an helicopter and roll on the ground using its wings. The current version of the robot is entirely made of 3D printed parts which makes it easy to build and test, but is also too heavy and not very robust to crashes. Therefore, the goal of this project is to investigate what kind of materials, manufacturing techniques, and mechanical structures could be used to improve the robustness and reduce the weight of the robot. This project will involve the design and dimensioning of the robot, CAD design, and manufacturing of one working prototype.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL ME MT MX SV
Type of work: 20% theory, 30% research, 50% hardware
Requirements:
Subject(s): Flying Robot, Rolling Robot, Mechanical Design
Responsible(s): Ludovic Daler, Przemyslaw Kornatowski

Assessment of bio-inspired flight control strategies for UAV flight in forest

Audrey Boulard (MT)

Despite their tiny brains, flying insects are able to navigate safely using panoramic vision as main sensory input. By measuring image motion on their retina (optic flow) across a wide field of view, they are able to control their speed, altitude and to stay away from obstacles. However, when flying in complex environments such as a forest, little is known about how motion cues are spatially integrated into reactive flight control commands. The goal of this project is to test hypothesis on how flying insects use vision to fly in complex environments. The student will use an existing flight simulator and replicate recent experiments performed with free-flying bumblebees. Several flight control strategies for collision-free navigation will be formulated and tested on this simulator. Also, in order to assess the relevance of the results for robotic implementations, an flight arena similar to the one used with bumblebees will be built and used to test the selected strategies on a quadrotor robot.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL IN MT PH SC
Type of work: 20%Theory, 40% Software, 40% Experiments
Requirements: Programming, Data analysis, Control theory
Subject(s): Flight control, Optic flow
Responsible(s): Julien Lecoeur, Ramòn Pericet Camara

Comparison of Collision Avoidance Strategies

Mathias Heyraud (MT)

The myCopter project tends to enable the technologies for Personal Aerial Vehicle. This project proposes an innovative way to overcome the financial and environmental cost of current road transportation by using the 3rd dimension for personal transportation such as commuting. Here at LIS, we focus on collision avoidance strategies in dense environments. With the large number of user in the sky, the risk of collision is growing accordingly and ways to ensure safety should be addressed. We developed a real-time simulator with realistic dynamics on which we can test and compare different collision avoidance strategies. This semester project aims to implement and compare simple collision avoidance strategies in the simulator such as Reynolds Flocking. Ce projet peut être réalisé en français.

Type: Semester project
Period: 16.09.2013 - 20.12.2013
Section(s): EL IN MA ME MT PH
Type of work: Theory 20% Software 80%
Requirements: Programming knowledge, ADA would be a plus
Subject(s): Programming
Responsible(s): Nicolas Dousse , Felix Schill
URL: Click here

A Flying Robot Capable of Dynamic Running

Beat Geissmann (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design a mechanism that allows a flying wing platform, which can also move on the ground by using its wings, to achieve "dynamic running". Previous work was done on the design of a mechanism that allows a flying robot to walk on the ground with its wings, and on the design of deployable wings. The goal of this project is thus to combine these two mechanisms to achieve greater amplitude of steps and possibly "dynamic running". First a study of the system will be done using an existing physics-based simulator, general principles will be extracted from that study and will be used to design and manufacture a mock-up prototype. Depending on the advancement of the project, the mechanism could be implemented in a real flying platform.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): ME MT SV
Type of work: 20% theory, 20% research, 30% software, 30% hardware
Requirements:
Subject(s): Flying Robot, Walking Robot, Mechanical Design
Responsible(s): Ludovic Daler, Julien Lecoeur

Enabling a Rolling, Flying and Hovering Robot with Autonomous Capabilities

Yann Roth (MT)

At the LIS we are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to enable a robot, which can fly forward like a plane, hover vertically like an helicopter and roll on the ground using its wings, with autonomous capabilities. The current version of the robot can hover, fly forward, and roll on the ground, but it is still completely remote controlled. Therefore, this project consist in investigating what kind of sensors could be used to enable the robot with autonomous behaviors, specifically for the ground locomotion mode. A first electronic board was designed to control the motors used for the ground locomotion, and this project will possibly include a second design of this custom electronic board for the robot to add new sensors. In this project, we will also investigate different gaits to see which are more efficient. The final goal of this project will be to show that the robot can autonomously navigate on rough terrains.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL ME MT MX SV
Type of work: 20% theory, 40% research, 40% hardware
Requirements:
Subject(s): Flying Robot, Rolling Robot, Sensors, Electronic Design
Responsible(s): Ludovic Daler, Grégoire Hilaire Marie

Implementation of low-power motion detection in microcontrollers

Marcel Starein (MT)

In our lab, we have a strong interest micro air vehicles (MAVs), and how these vehicles can navigate based on vision similarly to insects. In order to further reduce the size of such flying robots, new navigation control methods to be used with lower resources have to be investigated. A good strategy is to detect motion of the environment to carry out various tasks like obstacle avoidance or flight stabilization. Developing new methods to extract motion with very low-power microprocessors would be very advantageous to be implemented onboard a very small flying robot. In this project, the goal is to investigate, develop, characterize and validate a new method of motion detection using a very small, very low resolution vision sensor. The student will develop such method optimizing available algorithms in literature, or based on his own ideas. He will be asked to implement such method in a low-power microcontroller and will carry out experiments to characterize and validate the motion extraction capabilities of the method.

Type: Semester project
Period: 17.09.2013 - 20.12.2013
Section(s): EL MT PH SC
Type of work: research 20%; software 50%; experiments 30%
Requirements: microcontroller programming or C/C++ knowledge
Subject(s): bio-insipired robotics; image processing; flying robotics; vision-based navigation
Responsible(s): Ramòn Pericet Camara, Adrien Briod
URL: Click here

Magnetic connection mechanism for soft modular robots

Jonathan Zuercher (MT)

At the LIS we are developing soft modular robots to form artificial mult-cellular systems that can change their morphology to suit task or to adapt to the environment. This makes these robots potentially more robust and flexible compared to traditional fixed-morphology systems, especially in unknown or difficult environments.

One of the major challenges in the design of soft modular robots is the availability of a soft reversible connection mechanism. Therefore, the goal of this semesterproject is to investigate a novel clamping/gripping technology based on magnets and pneumatics. Magnets enable reliable attachment with high forces, however require a mechanism for detachment. Recent progress in soft, pneumatics show that this technology is fast, powerful and reliable, and thus is a promising actuator for a detachment mechnanism.

This project will tackle the challenge of design, dimensioning and fabrication of this mechanism. More specifically, the student is expected to investigate how to embed magnets in soft polymer, to propose and evaluate different designs for the mechanism and to dimension the mechanism to minimize weight and size while ensuring high connection strength and reliable detachment. Easy integration of the mechanism into the soft modules is another important requirement. At the end of the project the final design of the mechanism will be characterized while integrated in current prototype modules.

Type: Semester project
Period: 18.02.2013 - 01.06.2013
Section(s): MT
Type of work: 30% theory, 70% hardware
Requirements:
Subject(s): Soft robotics, soft materials, smart acutators
Responsible(s): Jürg Markus Germann, Bryan Schubert
 

Sensor fusion and ad hoc communication networking

Philippe Paccaud (MT)

The myCopter project tends to enable the technologies for Personal Aerial Vehicle. An innovative way to overcome the financial and environmental cost of current road transportation. At LIS, we focus on collision avoidance strategies as well as testing on small-scaled platforms. To validate the collision avoidance strategies, a light-weight quadrotor was developed at LIS. The aim of this project is twofold: first, to implement sensor fusion techniques such as Kalman filtering to do position control and navigation on a quadrotor. GPS information and an IMU (inertial measurement unit) will be fused. And second, to program data broadcasting between a swarm of quadrotors (~10 platforms) using a XBee communication module. If time permits, a simple collision avoidance strategy will be implemented.

Type: Semester project
Period: 18.02.2013 - 01.06.2013
Section(s): EL IN ME MT
Type of work: Hardware 30%, Software 70%
Requirements: Programming skills
Subject(s): Flying robots, data fusion, communication
Responsible(s): Nicolas Dousse , Felix Schill
URL: Click here

Pneumatic Soft Robotic Actuator with Controllable stiffness

Simon Houis (MT)

Pneumatic, elastomer actuators are an interesting technology for soft robots because they are completely elastic and highly deformable. They are made by carefully designing a hollow rubber structure that, when inflated, expands to a particular shape. However, one limitation of these actuators is that producing multiple shapes requires the introduction of hard, heavy valves or the use of multiple pumps. One approach to avoid this problem is to produce a single soft actuator capable of multiple shapes by using a material with controllable stiffness. Then, by varying the stiffness of certain regions of the device, it would be possible to control the overall shape of the actuator. More specifically, we propose creating a soft pneumatic actuator with embedded low-melting-point alloy structures that can be used to control the stiffness of local regions of the actuator. This project will focus on the creative design of the composite actuator, including design of the overall soft pneumatic structure and the dimensioning and placement of the controllable stiffness elements. Also, the student will need to develop a clever manufacturing method, using 3D printing and polymer molding, in order to produce a working prototype capable of attaining many shapes with only a single pump.

Type: Semester project
Period: 18.02.2013 - 31.05.2013
Section(s): EL ME MT MX
Type of work: 40% Design, 60% Fabrication
Requirements:
Subject(s): mechanics, thermodynamics, electronics
Responsible(s): Bryan Schubert, Jürg Markus Germann

Real-time Ad-hoc networking for robotic swarms

Raffael Hochreutener (EL)

Robots operating collectively in swarms require reliable, decentralised communication with low, predictable latency and fast update rates. Within the MyCopter project, we require reliable communication between flying robots for mid-air collision avoidance. Current communication systems are generally designed for sporadic, point-to-point traffic, while in robotic swarms the mode of communication is continuous dissemination of information. We previously developed a distributed time-division channel access (TDMA) protocol that addresses these issues. The goal of this semester project is to implement and evaluate this TDMA protocol on readily available commodity radio modules. Experiments shall be conducted to measure real-world performance in static and dynamic networks, and compare the results with previous work in simulations. If time permits, further work can be carried out on the integration of routing algorithms with the scheduling method.

Type: Semester project
Period: 18.02.2013 - 31.05.2013
Section(s): EL IN MT
Type of work: 30% theory, 50% embedded software, 20% hardware
Requirements: embedded systems, good programming skills, graph theory, electronics (RF)
Subject(s): wireless communication, real-time embedded systems robotics
Responsible(s): Felix Schill, Nicolas Dousse
URL: Click here

2012


Egomotion estimation of a microflyer assisted by a compound camera

Florian Gerlich (MT)

Flying insects possess compound eyes that assist them to perform a variety of navigation tasks with limited resources. Novel compound cameras with comparable size are taking this inspiration to be used onboard of microflyers. The aim of this project is to implement egomotion estimation on a flying microrobot assisted by a smart compound camera. The student will deal with the design and implementation of the compound camera on the flying robot. A customized mechanical attachment of the camera to the robot will be designed and implemented. A new available method for egomotion estimation based on inertial and visual sensor fusion will be implemented by the student onboard the microcontroller of the camera. Special attention will be put on the development of low-computation optic flow extraction with the camera prototype. The student will make use of simulation tools and prototype platforms for implementation and optimization. Flying experiments to validate the implementation will be realized assisted by a VICON system. The student will be encouraged to validate the viability of the implementation with a flying demo.

Type: Master project
Period: 18.02.2013 - 21.06.2013
Section(s): EL IN ME MT
Type of work: 40% software; 20% hardware; experiments 40%
Requirements: microcontroller programming (C or C++), some knowledge on control
Subject(s): bio-inspired robotics; flying robotics; vision-based navigation
Responsible(s): Ramòn Pericet Camara, Adrien Briod
URL: Click here

Collision-free navigation of a microflyer assisted by a compound camera

Steven Briquez (MT)

Flying insects possess compound eyes that assist them to perform very efficient collision-free navigation with limited resources. Novel compound cameras with comparable size are taking this inspiration to be used onboard of microflyers. The aim of this project is to implement collision avoidance on a flying microrobot assisted by a smart compound camera. The student will deal with the design and implementation of the microrobot control directly assisted by the smart compound camera. Available or new methods for collision avoidance will be implemented by the student onboard the microcontroller of the camera. This requires special work on optimization of data processing in terms of computational load. The student will make use of simulation tools as well as camera and flyer prototypes for implementation and optimization of optic flow extraction and control methods. The final demonstration will show the collision-free, (semi-)autonomous flight of the microflyer in a highly-cluttered environment.

Type: Master project
Period: 18.02.2013 - 21.06.2013
Section(s): MT
Type of work: 40% software; 20% hardware; experiments 40%
Requirements: microcontroller programming (C or C++), some knowledge on control
Subject(s): bio-inspired robotics; flying robotics; vision-based navigation
Responsible(s): Ramòn Pericet Camara, Julien Lecoeur
URL: Click here

Modular airfoil design for low Reynolds number environments

Leon Duplay (MT)

The development of miniature aerial vehicles (MAVs) is currently one of the major challenges in robotics research. These MAVs are characterized by small physical dimensions, low flight speeds, reduced payloads, and reduced aerodynamic performance due to operating in a low Reynolds number environments (below 105). On the other hand, MAVs hold several advantages over larger flying robots thanks to favorable scaling: better structural strengths, reduced stall speed, and better impact tolerance. In particular, MAV flight performance is strongly influenced by wing geometry. To date, very little research has been done on the impact of wing design and in particular, no research has systematically explored the effects of the wing cross-section or airfoil on flight performance. This Master thesis will provide an exploration of the impact on flight performance of different airfoil shapes for small flying robots. The first phase of the project consists of designing, building and characterizing a modular wing system that can change airfoil shape. The prototype will be controllable by software and be able to perform automated testing. In the second phase, the prototype will be used in conjunction with the Harvard Microrobotics Lab wind tunnel setup, to perform fully automated testing and to measure the impact of the different airfoil dimensions and characteristics, allowing for the discussion of principles for the design and fabrication of airfoils for future MAV research.

Type: Master project
Period: 18.09.2012 - 19.04.2013
Section(s): MT
Type of work:
Requirements:
Subject(s): Airfoil optimization
Responsible(s): Ludovic Daler, Robert Wood
 

Sensor fusion for mini UAV

Christophe Barraud (MT)

We developed a mini-UAV that is equipped with a set of positioning (GPS), inertial, magnetic and pressure sensors to implement autonomous navigation. The Kalman filter currently implemented can only estimate the attitude (roll and pitch angle). The goal of this project is to design, implement and characterise one or more new Kalman-based state estimation filter(s) that can provide an estimate for additional variables, including at least position, velocities, heading, course and wind speed/direction.

The project includes: 1) Design of an experimental setup for in-flight data logging and ground truth values acquisition (based on a photogrammetric process). 2) Offline design, tuning, optimisation and characterisation of the state estimation filter(s). 3) Implementation of the optimised filter(s) on the mini-UAV’s autopilot and characterisation of the final result.

This project will take place in a start-up company active in the development and sales of mini-UAV and will be run in close collaboration with the LIS.

Type: Master project
Period: 09.09.2012 - 01.04.2013
Section(s): EL IN ME MT
Type of work: 30% theory, 30% testing, 40% coding
Requirements: embedded programming, basics of sensor fusion
Subject(s): sensor fusion, sensor characterization, flight tests, embedded systems
Responsible(s): Adrien Briod, Antoine Beyeler

Optic-flow based obstacle avoidance on a flying robot

Michael Spring (MT)

The AirBurr platform is a flying robot that can collide with obstacles without breaking. Unlike conventional platforms that use heavy sensors to model the environment and avoid all obstacles, the AirBurr can thus navigate with simpler and lighter sensors that provide incomplete information about the environment, as collisions aren’t critical anymore. However, even though some collisions with obstacles are acceptable, they should be avoided most of the time for efficient navigation, which is the subject of this Master project.

Recent work showed that optic-flow and inertial sensors could be used for ego-motion estimation and keep the speed of the robot stable as long as the robot moves according to a certain pattern. The goal of this Master project is to take these constraints into account while implementing an obstacle avoidance behavior using the optic-flow measurements. Even though the robot's position is unknown, which prevents building a map of the environment, the algorithm will be able to use a short-term odometry that is available from the speed estimation algorithm. A final demonstration should show the robot flying completely autonomously in a cluttered environment, trying to stay away from the obstacles around it even though it might collide into those that it couldn't detect.

The platform is already built and the low-level stabilization controllers and optic-flow extraction are already implemented. The work will thus mostly focus on the algorithms for obstacle detection from optic-flow and the controller for avoidance. The algorithm will be programmed in C and should run in real time on a microcontroller.

Type: Master project
Period: 15.10.2012 - 15.03.2013
Section(s): EL ME MT
Type of work: 25% theory, 25% simulation, 25% coding, 25% flight tests
Requirements: embedded programming, robotics (if possible: students should have followed the 'mobile robots' course)
Subject(s): obstacle avoidance, optic-flow, minimal sensing
Responsible(s): Adrien Briod, Jean-Christophe Zufferey
URL: Click here

Wing Shape Optimization for a Flying and Hovering Robot

Charles Lambelet (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design a robot that can both fly horizontally like a plane and hover vertically like an helicopter. This project will include the design of fixed wings for this flying robot, in order to optimize its flight capabilities. This includes the characterization of the wing in terms of size, aspect ratio, and chord length and also the manufacturing of one or several wings. The wings will then be integrated into a flying version of the robot and tests will be performed to characterize the flight performances in both modes of locomotion.

Type: Semester project
Period: 18.09.2012 - 02.02.2013
Section(s): ME MT MX
Type of work: 30% theory, 30% research, 40% hardware
Requirements: Basic knowledge in aerodynamics
Subject(s): Flying robot, Aerodynamics
Responsible(s): Ludovic Daler, Felix Schill

Design of Robust Deployable Wings Using Tensegrity Structures

Johann Groll (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design robust compliant deployable wings for this platform using the concept of Tensegrity Structures. Tensegrity is a structural principle based on the use of isolated components in compression inside a net of continuous tension, in such a way that the compressed members (usually bars or struts) do not touch each other and the prestressed tensioned members (usually cables or tendons) delineate the system spatially. This project will involve the design and dimensioning of suitable deployable wings, CAD design, and manufacturing of one or several working prototypes.

Type: Semester project
Period: 18.09.2012 - 02.02.2013
Section(s): ME MT MX
Type of work: 30% theory, 30% research, 40% hardware
Requirements:
Subject(s): Flying Robot, Deployable Wings, Tensegrity Structures
Responsible(s): Ludovic Daler, Felix Schill

Wing-Actuated Ground Locomotion of a Flying and Rolling Robot

Patrizia Bernadette Hählen (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design deployable wings for this platform in order to enhance the ground capabilities of the robot. Previous work was done on an actuation mechanism to deploy the wings and on uprighting of the platform using the wings. Future work consist on improving the ground locomotion of the robot. We will study the concept used by passive-dynamic walking robot and apply it to our robot to improve the efficiency and the speed of the wing-actuated ground rolling. Furthermore, the robot should be able to not only use its wings to roll on the ground but also to steer. This project will involve the design and dimensioning of suitable deployable wings, CAD design, and manufacturing of one or several working prototypes.

Type: Semester project
Period: 18.09.2012 - 02.02.2013
Section(s): EL ME MT MX
Type of work: 30% theory, 30% research, 40% hardware
Requirements:
Subject(s): Flying Robot, Deployable Wings, Passive-dynamic Rolling
Responsible(s): Ludovic Daler, Adrien Briod

Design and Manufacturing of Soft Inflatable Wings for a Flying and Rolling Robot

Matteo Nessi (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design soft inflatable wings for this robot, based on embedded pneumatic networks (EPNs). This networks of channels can inflate like balloons for actuation. We will investigate how to design soft compliant joints that can provide actuation and that would improve the robustness of the system. Furthermore, we will investigate the concept of jamming for controlling the stiffness of the joints. Jamming is a physical process by which granular materials become rigid with increasing density. This project will involve the design and dimensioning of suitable inflatable wings, CAD designs, and manufacturing of one or several working prototypes.

Type: Semester project
Period: 18.09.2012 - 02.02.2013
Section(s): ME MT MX SV
Type of work: 30% theory, 30% research, 40% hardware
Requirements:
Subject(s): Flying Robot, Soft Inflatable Wings
Responsible(s): Ludovic Daler, Jürg Markus Germann

3D modelling of hexapod walking.

Alban Le Vallois (MT)

How have insects evolved to walk adaptively in complex environments? Answers to this question will greatly advance the development of robust terrain navigation in miniature bio-inspired robots.

For this project the student will use a "computational fly” that emulates insect walking in a 3D world simulator and bio-inspiration to test models of how hexapods attain stable gaits in 3D terrain.

This project will be done in the Laboratory of Intelligent Systems (EPFL) and in collaboration with the Benton Lab (UNIL). Therefore this is an extremely unique project right at the interface between engineering, computer science, & neurobiology.

Type: Semester project
Period: 11.09.2012 - 01.02.2013
Section(s): IN MA MT PH SC SV
Type of work: 60% software 30% research, 10% theory
Requirements: C/C++
Subject(s): Genetic Algorithms, Neural Networks, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Andrea Maesani
URL: Click here

Modeling and control of a flying robot

Lucas Turrian (MT)

VTOL (vertical take-off and landing) flying robots are often unstable systems, that need a performant active stabilization mechanism. The goal of this project is to realize a simple model of the AirBurr platform (lis.epfl.ch/airburr), and improve the existing flight controller. This task involves:
1) the characterization of motor inputs versus applied force and torque. For this task, a test-bed with a contra-rotating motor and the flaps will be mounted on a force sensor (mechanical setup will be provided) and the goal will be to do several measurements in order to model the forces (such as the force created by the flaps in function of their area and distance to the propellers, or the force torque created by the propellers).
2) the design of a simple simulator using the force model obtained from point 1)
3) the development of control algorithms and realization of tests in simulation before applying the best algorithm to the real platform.

Type: Semester project
Period: 09.09.2012 - 01.02.2013
Section(s): EL ME MT
Type of work: 30% theory, 40% experiments, 20% software, 10% hardware
Requirements:
Subject(s): control theory
Responsible(s): Adrien Briod, Przemyslaw Kornatowski
URL: Click here

Improvement and optimization of a sensor fusion algorithm for ego-motion estimation

Alexandre Pabouctsidis (MT)

An important problem in flying robotics is the ego-motion estimation of a mobile device: the estimation of the velocity of a robot using only embedded sensors (no GPS or external cameras). A method to estimate ego-motion thanks to inertial sensors and optic-flow sensors has been developed and showed promising results.

The goal of this project is to improve the existing algorithm by using together a unimodal state estimation method (like a Kalman filter) with a multimodal state estimation method (like a particle filter), in order to account for the strong non-linearities and to remain computationally reasonable for an embedded implementation. The new method will be tested and optimized in Matlab on real flight experiments data, and compared to the ground truth recorded thanks to a tracking system. Depending on the results, an algorithm to estimate position from integrated velocity will be developped. Finally, if time allows, the method will be implemented in C on the embedded STM32 processor for real-time operation on the flying robot itself.

Type: Semester project
Period: 09.09.2012 - 01.02.2013
Section(s): EL ME MT
Type of work: 40% theory, 30% implementation, 30% tests
Requirements:
Subject(s): sensor fusion, ekf, particle filter
Responsible(s): Adrien Briod, Jean-Christophe Zufferey
 

Extracting principles of touch evoked leg kinematics in Drosophila.

Martin Savary (MT)

Insects exhibit robust terrestrial locomotion while relying on a relatively small and simple controller, the nervous system. This makes it an ideal source of bio-inspiration for robust legged walking robots. However, owing to the tiny size and rapid kinematics of most insect behavior we cannot model this behavior precisely without advanced computer vision approaches.

The main objective of this project is to use an algorithm for extracting the leg positions of freely walking flies in high-speed, high-resolution movies to perform a quantitative analysis of the means by which flies use dynamic gaits to achieve optimal avoidance of other flies during collective behavior. This new knowledge will provide inspiration for efficient and robust robotic locomotion as well as a substrate for further studies into neural control of gait in animals.

The project will primarily be supervised at the Laboratory of Intelligent Systems (EPFL) and in collaboration with the Biomedical Imaging Group (EPFL), and the Benton Lab (UNIL). Therefore this is an extremely unique project right at the interface between engineering, computer science, & neurobiology.

Type: Semester project
Period: 18.09.2012 - 01.02.2013
Section(s): IN MA MT PH SC SV
Type of work: 60% software, 20% research, 20% theory
Requirements: C/C++
Subject(s): Image processing, Computer Vision, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Cédric Vonesch

Design and control of a contact-sensitive flying robot using whiskers and accelerometers

Marco Pagnamenta (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr is able to resist collisions with obstacles and go back in flight after a crash. It is for now autonomously stabilized but hasn't yet any high-level controller that tells it where to go. In order to navigate autonomously indoors, we aim at using the interactions with the environment and apply some reactive control to the contacts that occur during flight.

The goal of this project is to work on the contact sensing ability of the robot and the subsequent high-level autonomous control. The first task of the project is to interface whisker sensors to the onboard microcontroller, test them, and assess their usability on the flying robot. The second task of the project is to use whisker sensors and/or accelerometers (depending on the results of the first task) in order to program a simple autonomous behavior, showing the robot change direction after a collision in the air.

Type: Semester project
Period: 09.09.2012 - 01.02.2013
Section(s): EL MT
Type of work: 30%+programming +30%+testing +20%+theory +20%+hardware
Requirements:
Subject(s): sensing +flying+robotics
Responsible(s): Adrien Briod, Przemyslaw Kornatowski
URL: Click here

Radar based navigation and collision avoidance

Marc Schönenberger (MT)

In the scope of the myCopter project, we are enabling new technologies for collision avoidance in aerial vehicle. Radar modules are well developed for heavy and large platforms. The emergence of centimeter scale sensors allows their implementation on Micro aerial vehicles. The student is asked to implement a strategy to use radar modules for indoor navigation and collision avoidance on a quadrotor. The student will first try to fully control the robot in an indoor environment, then implement strategies in order to avoid incoming obstacles or even fly in the room while avoiding obstacles. If these objectives are satisfied, the transition for outdoor obstacle avoidance (using GPS navigation) can be assessed. Ce projet peut être fait en français.

Type: Semester project
Period: 18.09.2012 - 18.01.2013
Section(s): MT
Type of work: 90% software 10% hardware
Requirements: C language
Subject(s): Aerial robotics
Responsible(s): Nicolas Dousse , Felix Schill
URL: Click here

Pendul'air

Alexandre Cherpillod (MT)

The student will develop a new flying platform based on the AirBurr that will be able to fly not only indoor but also outdoor and whose weight should not exceed 1kg.

This platform will be used as a demonstrator for the myCopter project and is thus required to have a central cavity to simulate the presence of crew. It will have to perform vertical takeoff and landing (VTOL), to hover and to follow GPS navigation points.

The idea is to merge existing sensors such as the autopilot from SenseFly platform and the micro-controller software from AirBurr.

Ce projet peut être fait en français.

Type: Semester project
Period: 18.09.2012 - 18.01.2013
Section(s): EL IN ME MT
Type of work: 70% design & construction, 30% software
Requirements: C, C++
Subject(s): Aerial robotics
Responsible(s): Nicolas Dousse , Felix Schill

Autonomous Micro-Aerial Vehicle Navigation Using a Custom Optic Flow Sensor Ring

Raphael Cherney (MT)

The RoboBees project[1] is an effort to build a swarm of flapping-wing micro-aerial vehicles to collectively perform tasks such as crop pollination, disaster search, and target tracking. Each RoboBee is projected to weigh half a gram and be about 3 cm in length. Correspondingly, each RoboBee is extremely resource-scarce. However, the swarm is expected to be very large with hundreds of RoboBees. Given such a swarm, one of the main challenges in using it to perform the tasks listed above is coordination. To this end, the RoboBees project continues to research various ways of coordination to overcome the limitation of individual RoboBees and efficiently execute applications using the swarm[2]. Due to weight and energy limitations, it is hard to instrument micro-aerial vehicles with a variety of sensors. Since the RoboBees are currently under development, we use micro-helicopters as proxies for them. The objective of this project is to use a custom ring with eight optic flow sensors to perform ego motion estimation as well as indoor navigation on a micro-helicopter with most of the computation on-board. This will be used as the basis for ongoing research in studying distributed techniques for executing the target applications. [1] The RoboBees Project, http://robobees.seas.harvard.edu [2] Karthik Dantu, Bryan Kate, Jason Waterman, Peter Bailis, Matt Welsh, "Programming Micro-Aerial Swarms with Karma", In SenSys '11: Proceedings of the 9th International Conference on Embedded Networked Sensor Systems, Seattle, Washington, Nov. 1-4, 2011.

Type: Master project
Period: 18.09.2012 - 18.01.2013
Section(s): MT
Type of work:
Requirements:
Subject(s): Robotics
Responsible(s): Maja Varga, Karthik Dantu
 

Variable stiffness actuator mechanism based on low melting point metals for soft robots

Quentin Cabrol (MT)

At the LIS we are developing soft modular robots to form artificial mult-cellular systems. These systems can change their morphology by changing the softness of their module shell to suit task or to adapt to the environment. This makes them potentially more robust and flexible compared to traditional self-reconfigurable robots and fixed-morphology robotic systems.

In order to change the softness of the shell of a module, one needs to carefully design a mechanism featuring a variable stiffness actuator. A promising solution for this functionality is the use of low melting point metals., When embedded as tracks in a soft polymer, this actuator enables large stiffness changes going from rigid to a completely liquid state.

The goal of this semesterproject is to develop a variable stiffness actuation mechanism based on these low-melting point metals. The main challenge in this project is the smart design of the metal tracks/layers to be embedded in a polymer while optimizing for large stiffness change and small power consumption. Work will include fabrication of polymer structures featuring complex metal tracks, evaluation and testing of different track/layer compositions and the integration of the mechanism into soft modular robots.

Type: Semester project
Period: 18.09.2012 - 13.01.2013
Section(s): CH EL ME MT MX PH
Type of work:
Requirements: Some expierence with electronics and microfabrication techniques is advantageous
Subject(s): Electronics, Physics, Microfabrication
Responsible(s): Jürg Markus Germann, Bryan Schubert

Active flexible sensor for adaptive robot navigation

Benoit Seguin (IN)

The development of novel flexible vision sensors has opened a new avenue of more compliant and compact components to be implemented on robotic platforms. Such sensors reveal many advantages with respect to traditional stiff single-aperture cameras. In example, flexible vision sensors could exploit their physical characteristics to adapt to different underlying surfaces or to tune their functionality in line with the robot specific needs in a number of situations. For instance, a flexible camera could gain field of view by just curving its surface to a lower radius of curvature. However, new solutions have to be engineered in order to implement this capabilities onboard a real robot. The goal of the project is to investigate the possibility of performing collision avoidance task using the flexible vision sensor developed in the LIS laboratory. The student is expected to (a) look into algorithms to perceive motion suitable for the flexible vision sensors (b) implement the chosen one onboard the sensor’s microcontroller and perform tests under various curvature to assess its functionality, (c) evaluate various methods to actively bend the flexible sensor and implement the chosen one, (c) interface the sensor to the e-puck robot and prepare final demonstration.

Type: Semester project
Period: 18.09.2012 - 04.01.2013
Section(s): IN
Type of work: 40% motion extraction theory, 40% software developement, 20% hardware developement
Requirements: programming in C is an advantage
Subject(s): Motion Extraction, C Programming, Flexible Camera, Image Processing
Responsible(s): Michal Dobrzynski, Ramòn Pericet Camara
URL: Click here
 

Development of novel collision avoidance methods for microflyers using compound cameras

Jean-Luc Liardon (MT)

Flying insects possess compound eyes that assist them to perform very efficient collision-free navigation with limited resources. Novel compound cameras with comparable size are taking this inspiration to be used onboard of microflyers. However, these novel small and light sensors possess limited computational power and optimization of data processing has to be carried out while keeping efficient flight control. The aim of this project is to develop novel adaptive strategies for collision avoidance with constrained resources. First, the student will report the state of the art of optic flow extraction methods susceptible of being optimized. Further, the student will adapt such methods to the available sensor's characteristics. Then, those methods will be tested for various navigation scenarios making use of an available simulation tool. The conclusions of this project will include an evaluation of those methods for real implementation on physical robots.

Type: Semester project
Period: 18.09.2012 - 21.12.2012
Section(s): EL IN MT PH SC
Type of work: 30% theory, 40% software, 30% simulations
Requirements: C, C++, Matlab
Subject(s): bio-inspired engineering, vision processing, flying robotics
Responsible(s): Ramòn Pericet Camara, Michal Dobrzynski
URL: Click here

Implementation of collision-free navigation on a robot using a compound camera

Thibault Priquel (MT)

Flying insects possess compound eyes that assist them to perform very efficient collision-free navigation with limited resources. Novel compound cameras with comparable size are taking this inspiration to be used onboard of microflyers. The aim of this project is to implement collision avoidance on a wheeled robot assisted by a smart compound camera. The student will deal with the design of the mechanical assembly and the communication interface of the sensor on an e-puck platform. Available or new methods for collision avoidance will be implemented by the student onboard the microcontroller of the camera. The final demonstration will show the collision-free fully autonomous navigation of the e-puck in a highly-cluttered environment.

Type: Semester project
Period: 18.09.2012 - 21.12.2012
Section(s): EL MT
Type of work: hardware 10%; programming 40%; research 30%; control 20%
Requirements: C or C++; microcontroller programming
Subject(s): bio-inspired engineering; mobile robotics; vision processing
Responsible(s): Ramòn Pericet Camara, Michal Dobrzynski
URL: Click here

The synergy between neural path integration and landmark-based mapping

Florentin Marty (MT)

The navigational capabilities of rats have called much attention to the scientific community. The rats' spatial processing mechanisms have been widely studied in the recent years and, by the discovery of place cells in its hippocampus, they are believed to have a cognitive map of the environment. A relevant part of the place cell mechanism is the network of grid cells, found upstream in the medial entorhinal cortex. Those cells are believed to be the main mechanism of the path integration.Neuromorphic models of the grid cells were implemented in simulation and real robots. Through Hebbian learning allothetic sensory information was used to associate landmarks with position information. Whenever a landmark was revisited, the learnt knowledge was used to form additional inputs for grid cells. As a result, the system was able to correct and reset the grid cells' position coding and to stabilize the trajectory. Afterwards, the synergistic interaction between path integration in the grid cells and landmark mapping in the place cells was studied. We were able to show that landmarks and path integration are closely linked in the neural processing of spatial information and that solely landmark navigation leads to problems in neural orientation, because ambiguous sensory inputs are likely to cause error. Figure 2 illustrates that both, insufficient and excessive landmark density result in weak navigation performance of the simulated robot. Finally, new experiments for rats were proposed, in which cue ambiguity plays a key element.

Type: Master project
Period: 23.04.2012 - 10.09.2012
Section(s): MT
Type of work:
Requirements:
Subject(s): Robotics
Responsible(s): Maja Varga, Cesar Renno Costa
 

Towards an AR.Drone 2.0 Based Prototype For Autonomous 3D Mapping

Axel Murguet (MT)

Rattaché à l'équipe Image vous aurez en charge la mise en oeuvre d'algorithmes de construction de cartes et navigation sur un prototype de Drone. Le projet consistera à évaluer les stratégies pour cartographier une pièce avec un drone équipé d'une centrale inertielle, d'une caméra verticale et d'une Kinect. Les missions du stage sont: , . Prendre en main les algorithmes Parrot , . Intégrer la Kinect sur un prototype , . Participer à l'élaboration et l'évaluation d'algorithmes de cartographie. En collaboration avec l'équipe vous définirez des critères pertinents pour évaluer les performances des solutions. Les outils de travail sont Matlab et C/Cpp, un système de motion tracking et divers prototypes de Drones. La bonne compréhension algorithmique des techniques employées est nécéssaire.

Type: Master project
Period: 07.02.2012 - 06.09.2012
Section(s): EL MT
Type of work: 30% theory, 30% software, 10% hardware, 30% experiments
Requirements:
Subject(s): embedded programming, sensor fusion
Responsible(s): Adrien Briod, Gaspard Florenz

Wind resistance and VTOL capabilities for small flying-wings (in industry)

Loic Zimmermann (MT)

The candidate will first assess various possibilities of increasing wind resistance of small flying wings thanks to airframe optimization considering wing geometry, wing loading and thruster specifications. The impact of the proposed optimisations on airspeed, energy consumption, flight endurance, landing and takeoff capabilities will be systematically analysed by means of theory backed by in flight measurements. In a second part, adding VTOL capabilities using cheap strategies will be theoretically compared. This comparison will constitute the starting point of a design process and the production of at least one flying prototype as a proof-of-concept. The impact on cost, exploitation and handling capabilities will be considered and discussed.

Type: Master project
Period: 20.02.2012 - 20.08.2012
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler (senseFly)
 

Indirect encodings for soft-multicellular robots

Jérémie Despraz (PH)

Since the seminal work of Sims on virtual creatures, different systems for the evolution of morphology and control of modular robots have been proposed. However, the aim of generating robots that could reach levels of complexity comparable to the ones observed in natural systems is far from being achieved.
To achieve this goal, many challenges must still be solved. It is clear that to design the structures of such multi-cellular robots, automatic design methods are needed that could possibly replicate the incredible diversity level produced by nature in an artificial system. Various generative encodings have been proposed in the past, including grammar-encoding and methods that simulate natural morphogenesis.
In this project, the student will investigate existing indirect encodings for multi-cellular systems and test them on morphology matching problems. In the first part of the project, the student is expected to review existing encodings for the automatic design of multi-cellular structures. Then, the student will select the most promising encodings and will perform a series of experiments to evaluate their capabilities on morphology matching benchmarks. The outcome of this project should possibly be the selection of an existing encoding or the synthesis of a new methodology for the evolution of multi-cellular structures that will fulfill the assigned requirements. If time allows the capability of the chosen encoding will be demonstrated on an existing soft cell simulator.

Type: Semester project
Period: 21.02.2012 - 21.06.2012
Section(s): IN MT PH
Type of work:
Requirements:
Subject(s): Artificial Evolution, Modular Robotics, Soft Robotics
Responsible(s): Andrea Maesani, Jürg Markus Germann
 

3D Physics-Based Soft Multi-Cellular Simulator

Loïc Perruchoud (PH)

The tremendous technological advance we are currently experiencing will eventually lead to the feasibility of soft multi-cellular robots, which could potentially display many of the characteristics that can be observed in natural organisms. At the LIS, we are currently investigating different aspects of such multi-cellular artificial systems, both at the level of hardware and in software simulations.
In this project, the student is expected to port an existing 2D physics-based soft-multi cellular robot simulator into 3D. The current 2D simulator supports features like soft cell membranes, active/passive membrane adhesion, active deformation of the membranes and sensing capabilities (detection of external stimuli).
The student is expected to perform an evaluation study of existing 3D physics engines. Then, in the first phase of the project the student will select the best existing technology and implement the mechanisms already present in the 2D version of the simulator. Subsequently, the student will characterize the scalability of the simulator and demonstrate its capabilities in a few test scenarios.

Type: Semester project
Period: 21.02.2012 - 21.06.2012
Section(s): IN MT PH
Type of work:
Requirements:
Subject(s): Physics Based Simulations, Soft Robotics
Responsible(s): Andrea Maesani, Jürg Markus Germann
 

Viability Evolution for Prediction of Protein Structures

Adrien Béraud (MT)

Viability Evolution is a novel evolutionary meta-heuristic developed at LIS, that drives evolution by “shaping” the environment where the candidate solutions live and reproduce. The environment shaping is obtained by the dynamic modification of a number of constraints during the optimization process, which determines at each time step the solutions that will survive and those that will be eliminated.
In this project, the student is expected to apply Viability Evolution to a real-world problem, namely the prediction of protein assemblies conformations. In order to achieve a specific function, proteins often assemble in complexes having a unique shape of minimal energy. While determining experimentally the structure of a single protein at atomistic resolution is usually easy, determining the structure of a large assembly can be challenging. In this context, predicting the structure of a protein complex on the base of the structure of its individual components and low resolution experimental data acting as geometric constraints could be of great benefit. Furthermore, the energy landscape associated to proteins conformations is large and extremely rough, that makes traditional optimization methods unsuitable.
In the first part of the project, the student will integrate the existing optimization framework available at LIS with the specific problem implementation available at LBM (Laboratory for Biomolecular Modeling). Then, the student will rigorously evaluate the performance of the existing Viability Evolution algorithm against a traditional Evolutionary Algorithm. In the second part of the project, the student is expected to test different “environment shaping” strategies to improve the performance of Viability Evolution.

Type: Semester project
Period: 20.02.2012 - 20.06.2012
Section(s): MT PH
Type of work:
Requirements:
Subject(s): Artificial Evolution, Protein Structure Prediction
Responsible(s): Andrea Maesani, Matteo Thomas Degiacomi
 

Adaptation of ego-motion algorithm for embedded systems

Géraud L'Eplattenier (EL)

The goal of this project is the adaptation of an existing ego-motion estimation algorithm for an embedded artificial compound vision system. This novel sensor has a spherical shape containing 272 pixels and contains a gyroscope, an accelerometer and two microcontrollers within the spherical structure. Currently, the ego-motion algorithm is working on a demonstration platform with 6 optic flow sensors, a 3-axis gyroscope, a 3-axis accelerometer and two microcontrollers. In this project, we replace the six high quality optic flow sensors with compound vision sensor that provides low-quality but high density optic flow signals. This will require the existing ego-motion algorithm to be adapted to account for the special type of optical information produced as well as the limited computational power available. The experiments will be performed using an existing simulation tool that can provide virtual environments required for the characterization of the ego-motion algorithm. Specifically, the tool allows to simulate the noisy optic-flow and inertial sensors values obtained by the Spherical CURVACE moving along a predefined path in the virtual environment.

Type: Semester project
Period: 20.02.2012 - 01.06.2012
Section(s): EL
Type of work:
Requirements:
Subject(s): bio-inspired vision;vision-based navigation; vision processing
Responsible(s): Ramòn Pericet Camara, François Tièche (HE - ARC)
 

2011


Design and control of a contact-sensitive flying robot using force sensors

Arnaud Garnier (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr is able to resist collisions with obstacles and go back in flight after a crash. It is for now autonomously stabilized but hasn't yet any high-level controller that tells it where to go. In order to navigate autonomously indoors, we aim at using the interactions with the environment and apply some reactive control to the contacts that occur during flight.

The goal of this project is to work on the contact sensing ability of the robot, so that the robot is able to detect when an external force (in the order of 0.1N) is applied anywhere on its external structure. The first task of the project is the choice of force sensors for contact measurement, and how to integrate them to the existing outside structure of the platform (support from our mechanical engineer will be provided for part production and construction). Then the interface with the embedded electronics will be developed in C, so as to provide useful commands to the control system. If time allows, autonomous control can be tackled to explore a room or find openings in the wall.

Type: Semester project
Period: 09.09.2012 - 01.02.2013
Section(s): EL ME MT
Type of work: 20% theory, 40% hardware, 20% electronics, 20% programming
Requirements: mechanical design (CAD software), sensors, electronics, embedded programming
Subject(s): force sensing, control, robotics
Responsible(s): Adrien Briod, Przemyslaw Kornatowski
URL: Click here

Electroadhesion at a microscale

Constantinos Stergiopulos (MT)

At the LIS we are developing soft modular robots to form artificial mult-cellular systems that can change their morphology to suit task or to adapt to the environment. This makes these robots potentially more robust and flexible compared to traditional fixed-morphology systems, especially in unknown or difficult environments. One of the biggest challenges in the design of soft modular robots is the availability of a soft reversible connection mechanism. Therefore, the goal of this project is to investigate a novel clamping/gripping technology called "electroadhesion" at a microscale. This adhesion technology is electrically controllable and induces electrostatic charges on a substrate using a power supply connected to compliant pads situated on the robot. Electroadhesion enables high-clamping forces on a wide variety of substrate. Electrostatic forces occur between the substrate material and electroadhesive pads. These pads are comprised of conductive electrodes that are deposited on the surface of a polymer. When alternate positive and negative charges are induced on adjacent electrodes, the electric fields set up opposite charges on the substrate and thus cause electrostatic adhesion. This project will involve the fabrication and testing of electroadhesive pads at the microscale and the design of supply electronics. A major challenge will be the dimensioning and integration of pads for the use of this technology in a soft modular system.

Type: Semester project
Period: 18.09.2012 - 13.01.2013
Section(s): EL ME MT
Type of work: 20% theory, 60% hardware, 20% testing
Requirements:
Subject(s): Soft robotics, Flexible electronics
Responsible(s): Jürg Markus Germann, Bryan Schubert

Benchmarking and performance assessment of network inference methods

Urtzi Alfaro (SC)

The effective reverse engineering of Gene Regulatory Networks (GRN) is one of the great challenges of systems biology and is expected to have substantial impact on the pharmaceutical and biotech industries in the next decades. A gene network is formed by regulatory genes, which code for proteins that enhance or inhibit the expression of other regulatory and/or non-regulatory genes, thereby forming a complex web of interactions. The goal of reverse engineering is to automatically identify such a network from experimental data.

The goal of this project is the partial implementation of two reverse engineering algorithms. The first one is based on AGE, a Analog Genetic Encoding developed at LIS. The second is based on CMA-ES which has been shown multiple times to be an efficient evolutionary strategy. The main objective of this project is to compare quantitatively the performance of the AGE-based and CMA-based gene network inference methods.

Our long-term goal is the development of an open-source reverse engineering library for the bio-computing community. Thus, this project is very demanding with respect to programming and requires strong interest in development of extensible and reusable object-oriented software.

Type: Semester project
Period: 20.02.2012 - 01.07.2012
Section(s): IN MA MT PH SV
Type of work: 50% software, 50% research
Requirements: Java / Matlab
Subject(s): systems biology, gene networks, reverse engineering, benchmark, DREAM challenges
Responsible(s): Thomas Schaffter, Trevis Alleyne

Target detection using flying robots

Leon Duplay (MT)

This semester project aims at implementing an image recognition algorithm on flying robot platform for detecting targets on the ground for future usage in a search-and-rescue mission., The first task of the candidate will be to study different algorithms proposed in literature and provide state of the art review of the current algorithms used in the field of image recognition. Based on the literature survey, candidate should define the approach how to determine probabilities of finding the target while flying over a predefined area., The main task of the project is to implement and test the approach on the embedded computer of our flying robot platform. The embedded computer, Overo from Gumstix, is running Linux and is equipped with a camera. Chosen algorithm can be directly programmed on Gumstix computer in C++ using OpenCV library. The detection should be done in real-time during the flight and results of the target detection algorithm should be sent to the other parts of the control system for further evaluation. The task also includes outdoor flight-testing and the analysis of the flight data. The performance of the algorithm should be evaluated for different flight altitudes and lightning conditions and robustness of the algorithm should be determined in presence of turbulences.

Type: Semester project
Period: 20.02.2012 - 22.06.2012
Section(s): EL IN MT
Type of work: 20% research, 40% software, 40% experiments
Requirements: C/C++
Subject(s): Aerial Robotics, Image Processing
Responsible(s): Maja Varga, Jean-Christophe Zufferey

Design and Manufacturing of an Actuation Mechanism for a Deployable Wing

Mathieu Bonny (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to explore different solutions for the actuation of a deployable wing. First, it will be necessary to establish the state of the art of the different technologies that could be used. Then, one or several methods will be tested, and the most appropriate solution, will be implemented into a complete deployable wing. Also, depending on the progress made, this wing will be integrated into a flying robot.

Type: Semester project
Period: 20.02.2012 - 22.06.2012
Section(s): EL ME MT MX
Type of work: 20% theory, 30% research, 50% hardware
Requirements:
Subject(s): Smart materials, Actuation, Deployable wing
Responsible(s): Ludovic Daler, Bryan Schubert

Wing-Aided Uprighting of a Flying and Rolling Robot

Frank Bonnet (MT)

We are developing a novel flying platform which has the ability to both move easily through the air and on the ground. This new platform will be able to hover, fly, roll on the ground, and upright itself thanks to its deployable wings. These abilities bring this new flying robot closer to the capabilities of birds that are much more adaptive to their environment than current flying robots.

The goal of this project is to design an active uprighting mechanism for this platform. The robot should be able to use its wings to upright itself in order to get back to a takeoff-ready position. This mechanism will allow the robot to transition between the ground locomotion mode to the hover locomotion mode. Using the deployable wings of the robot for uprighting is a very efficient solutions since it does not require an additional actuator for this task. This project will involve the design and dimensioning of this uprighting mechanism, CAD design and manufacturing of one or several working prototypes. The mechanism will then be integrated into a flying version of the robot, depending on the advancement of the project.

Type: Semester project
Period: 20.02.2012 - 22.06.2012
Section(s): EL ME MT MX
Type of work: 20% theory, 20% research, 60% hardware,
Requirements:
Subject(s): Flying robot, Deployable wings, Uprighting mechanism
Responsible(s): Ludovic Daler, Adrien Briod

Environment shaping strategies for Viability Evolution

Adrian Tudor Panescu (IN)

Evolutionary computation (EC) is the field of science that aims to develop problem-solving tools by modeling the evolutionary process in nature. Viability Evolution is a novel evolutionary meta-heuristic developed at LIS, that drives evolution by "shaping" the environment where the candidate solutions live and reproduce. The higher performances of Viability Evolution with respect to classical Evolutionary Computation methods are mainly due to the higher level of diversity maintained during the search process.
In the current implementation, the environment is described by a series of requirements (viability constraints) that determine the possibility of an individual in the population to survive and reproduce. Individuals that do not comply with the viability constraints are eliminated from the population. Moreover, constraints are only made tighter as evolutionary time proceeds. The absence of a mechanism that could relax (under specific conditions) the viability constraints might eventually hamper the ability of the algorithm of escaping local optima.
In this project the student is expected to implement a new environment shaping policy for Viability Evolution that supports constraints relaxation. The student will perform experiments on a dynamic optimization benchmark as well as static optimization problems. Finally, the student will analyze the results of the experiments.

Type: Semester project
Period: 20.02.2012 - 22.06.2012
Section(s): EL IN MA ME PH SC SV
Type of work: 25% theory, 25% research, 50% software
Requirements: knowledge of C/C++ is beneficial
Subject(s): Artificial Evolution, Genetic Algorithms, Optimization
Responsible(s): Andrea Maesani, Pradeep Ruben Fernando

Miniature electro-permanent magnets for modular soft robots

Mohamed Raad (MT)

Modular robots require connections that are strong when engaged, act over a distance, self align, require little or no holding power, and are easy to release. Permanent magnets satisfy most of these conditions, but they require a high force to separate. Traditional electromagnets allow easy release, but they require constant electrical power to maintain engagement. A hybrid of these two technologies, called Electro-Permanent Magnets (EPMs), has been developed that satisfies all the above mentioned criteria. These devices use a small current pulse to switch between a permanent magnetic state and an off state. However, this technology currently exists only on the millimeter scale.

The challenge of this project is to generate functioning devices on the micrometer scale. This will allow the scaling of robotic modules, thereby increasing their range of possible applications. In this project, the student will explore the underlying principles of EPMs in order to develop designs and fabrication methods that allow their miniaturization. Successful prototypes will be characterized and integrated with soft robotic modules.

More specific, the project involves the following work packages:
(i) Fabrication and characterization of different types (e.g. Neodym and AlNiCo) of micrometer-sized magnets (by using magnetic powder)
(ii) Development of a fabrication method to assemble the fabricated magnets and wrap coils around the magnets (iii) Characterization of developed EPMs and its fabrication method

Type: Semester project
Period: 20.02.2012 - 20.06.2012
Section(s): EL ME MT MX
Type of work: 10% design, 60% fabrication/hardware, 30% characterization
Requirements: Creativity and desire to fabricate novel structures.
Subject(s): Electromagnetics, circuits, materials
Responsible(s): Jürg Markus Germann, Bryan Schubert

Multi-objective Optimization using Viability Evolution

Marina Mircheska (IN)

Evolutionary Algorithms are well suited to perform multi-objective optimization as they maintain a population of solutions that can be used to simultaneously explore a bigger subset of the solution space (compared to Simulated Annealing). But, simple weighted sum approaches to evolutionary multi-objective optimization (EMO) perform poorly due to lack of diversity in terms of all the objectives. When augmented with special diversity preserving techniques, evolutionary multi-objective optimization approaches have shown good performance but still need additional mechanisms to deal with optimization of more than 3 objectives. At LIS, a novel Evolutionary Computation method called Viability Evolution (ViE) has been developed that utilizes “environment shaping” and “eliminations” to drive evolution rather than depend on an absolute (or partial) ranking of individuals. Hence, the ViE algorithm will not suffer from the problems faced by traditional EMO algorithms, and initial experiments show promising results on multi-objective benchmarks when compared against the state-of-the-art EMO techniques such as Non-dominated Sorting Genetic Algorithm (NSGA-2). The goal of this project is to modify the existing ViE algorithm to produce a well-distributed Pareto front in multi- and many-objective optimization problems. Additionally, the scalability of the developed ViE algorithm (in terms of the number of objectives) should be analyzed.

Type: Semester project
Period: 20.02.2012 - 20.06.2012
Section(s): EL IN MA ME MT SC SV
Type of work: 20% theory 30% research 50% software
Requirements:
Subject(s): Multi-objective optimization, Evolutionary Algorithms
Responsible(s): Pradeep Ruben Fernando, Andrea Maesani

Build a Computational Fly.

Raphael Cherney (MT)

To understand the behavior of complex biological systems it is often useful to build a physically accurate simulation. Robotics has a history of using such computational tools and these can also be exploited to reverse engineer biological behaving systems. In return, lessons learned from these simulations may then be directly applied to the generation of advanced artificial intelligent systems.

We are studying the behavior of the hexapod insect, Drosophila melanogaster. Owing to hundreds of millions of years of insect evolution, its sensing and actuation mechanisms serve as useful guides for the development of sophisticated behaving robots.

For this project we will develop a 3D simulation environment for a morphologically and kinematically accurate computer generated fly. With this tool, we will test the means of achieving fly-like locomotion using dimension-reducing control strategies. Results from these experiments may suggest bio-inspired strategies for robotic locomotion.

Type: Semester project
Period: 20.02.2012 - 08.06.2012
Section(s): IN MA MT PH SC SV
Type of work: 70% software, 20% research, 10% theory
Requirements: C++
Subject(s): Robotics, 3D Simulation, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Andrea Maesani
URL: Click here

High resolution, high-speed localization of Drosophila touch responses.

Lukas Frisch (MT)

Identification of the mechanisms that underlie collective behavior requires a high-resolution understanding of the behavior of individuals. Therefore, for this project the student will use a high-speed, high-resolution behavioral imaging apparatus to collect videos of freely moving flies interacting with one another. Subsequently, using state-of-the-art tracking algorithms, they will compile these interaction responses into a "touch map" that identifies the walking response of flies to touch. This map will include the precise location of each touch, providing crucial information regarding the location of neurons responsible for conveying this information.

This project will be done at both the Laboratory of Intelligent Systems (EPFL) and in the Benton Lab (UNIL).

Type: Semester project
Period: 12.09.2011 - 01.02.2012
Section(s): MT
Type of work: 40% software 40% experiments 20% theory
Requirements: Matlab
Subject(s): Machine Vision, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Adrien Briod
URL: Click here

Generate in silico benchmarks for developmental network inference methods

Jonathan Rohrbach (IN)

One of the most challenging goals in systems biology is the development of computational tools for the reverse engineering of gene regulatory networks, that is, to infer the network of gene interactions from quantitative experimental data. Knowing how genes interact which each other is a primordial information required to design novel drugs that are aimed to target very specific genes.

Numerous methods have been developed for inference of gene regulatory networks from expression data, however, their strengths and weaknesses remain poorly understood. Accurate and systematic evaluation of these methods is hampered by the difficulty of constructing adequate benchmarks and the lack of tools for a differentiated analysis of network predictions on such benchmarks.

GeneNetWeaver (GNW) is a Java tool we developed for in silico benchmark generation and performance profiling of network reverse engineering algorithms. GNW is used by a world-wide community to generate in silico ("virtual") models of gene regulatory networks that can be simulated to reproduce many biological experiments (knockouts, knockdowns, multifactorial perturbation, etc.).

The goal of this project is to extend the generation of benchmark from currently single-celled to multi-cellular tissues by incorporating our detailed gene network model into a 2D cellular tissue model implementing spatial diffusion of proteins involved in the system.

Type: Semester project
Period: 20.09.2011 - 01.02.2012
Section(s): IN MT SC SV
Type of work: 50% software, 50% research
Requirements: Java / Matlab
Subject(s): systems biology, gene networks, reverse engineering, benchmark, DREAM challenges
Responsible(s): Thomas Schaffter, Pavan Ramdya

Extracting individual behaviors during collective encounters.

Winnie Wing Yee Tse (MT)

The actions of individuals form the basic building blocks of collective or swarm behavior. We aim to extract these individual responses from videos capturing the behavior of groups of fruit flies in an attempt to understand the emergence of collective olfactory decision making in these animals. In this project, the student will the develop computational tools necessary to automatically extract the responses of flies to physical interactions with one another. This data will be processed and analyzed to identify if a somatotopic or body-location dependent set of rules applies to these interactions. Finally these tools will be applied to understanding if such somatotopic rules hold true in the presence of an aversive odorant. This project will be done in the Laboratory of Intelligent Systems (EPFL) using data acquired the Benton Lab (UNIL).

Type: Semester project
Period: 13.09.2011 - 01.02.2012
Section(s): MT
Type of work: 80% software, 20% theory
Requirements: Matlab
Subject(s): Animal behavior, computer vision
Responsible(s): Pavan Ramdya, Thomas Schaffter
URL: Click here
 

Signal to noise ratio enhancement for an audio based relative positioning system

Christophe Barraud (MT)

At the LIS we are working on an audio based relative positioning system for a group of outdoor flying robots that can provide every robot with information about the position of its neighboring robots. Similar to how animal’s hearing system works, the idea here is to mount a set of microphones on the robots and to estimate the direction of the sound emitted from other robots by computing the time delay among signals of different microphone pairs., However, microphone sensors are very sensitive to wind and platform vibrations that highly affects the localization performance when the robots are in the air. Also the self engine noise of the robots has a high influence on the localization performance. The goal of this semester project is to investigate different microphone mounting strategies for reducing the effect of noises and hence enhancing the signal to noise ratio of the system. More specifically, the student is required to design, develop, test and compare different wind protections, anti vibration mounts and microphone mounting positions. The limitations of the flying platform in terms of weight and aero dynamics must be taken into account during the design.

Type: Semester project
Period: 20.09.2011 - 20.01.2012
Section(s): EL IN MA ME MT
Type of work: 30 % theory, 40% design, 30% experiments
Requirements:
Subject(s): Microphone, Wind protection, anti-vibration mounts, Signal to noise ratio
Responsible(s): Meysam Basiri, Felix Schill

Active acoustic signaling motor driver

Maxime Brülhart (MT)

Teams of flying robots can accomplish aerial coverage tasks more robustly and more efficiently compared to a single flying robot. Tasks such as security patrols or searching for victims inside a disaster area can benefit from several autonomous robots operating in parallel. However a challenge faced in the design of MAV swarms is to achieve motion coordination and collision avoidance. At the LIS we are working on developing an audio based relative positioning system which can help the robots to gain information about the location of other robots and hence address these challenges. The goal of this semester project is to design a motor driver that is capable of emitting acoustic signals using the motor. These signals can be perceived and used by other robots in the group to estimate the position of the sound emitting robot., More specifically, in this project it is required to modify an available robot motor controller in a way to be able to control both the motor speed and the pure tone sound of the motor. Furthermore experiments should be performed to identify the limitations of sound generation and the detection range for different sound patterns.

Type: Semester project
Period: 20.09.2011 - 20.01.2012
Section(s): EL IN MA ME MT PH
Type of work: 10 % theory, 50% software, 20% hardware, 20% experiments
Requirements: Programming, microcontrollers
Subject(s): motor driver, acoustic signaling
Responsible(s): Meysam Basiri, Felix Schill

Real time sound source localization

Michael Spring (MT)

At the LIS we are working on swarms of autonomous flying robots that can work together towards achieving efficient and robust aerial coverage tasks. One requirement for achieving this is that each robot should be equipped with a relative positioning system allowing it to obtain some information about the relative location of other robots. However, designing a relative positioning system for our flying robots is challenging since the robots can only carry a maximum payload of 150 grams. Inspired from nature, where animals use sound to communicate and localize each other, we are interested in developing an audio based relative positioning system which satisfies the constraints of our robots. The goal of this semester project is to develop an embedded system for real time direction estimation of a sound emitting robot., More specifically, this project involves the interfacing of four microphone sensors with a microcontroller and designing and implementing a real time sound source localization algorithm for estimating the direction of a sound source. The limited computational power and memory of the microcontroller have to be taken into account during the design.

Type: Semester project
Period: 20.09.2011 - 20.01.2012
Section(s): EL IN MA ME MT PH
Type of work: 20% theory, 60% software, 5% hardware, 15% experiments
Requirements: good programming skills, microcontrollers
Subject(s): microcontrollers, microphones, sound source localization
Responsible(s): Meysam Basiri, Felix Schill

Development and Evaluation of a small RADAR system for detecting unmanned aircraft

David Morisod (MT)

Mid-air collision avoidance is an important topic in aviation in general. With the increasing number of UAVs (unmanned aerial vehicles) it becomes a challenge for aerial robotics as well. We are investigating a range of sensor modalities to sense and avoid other aircraft in the vicinity, including very small 25GHz doppler radar modules. This project is to develop a small, lightweight radar system based on a 25 GHz FMCW radar frontend. The analog radar frontend is supplied as a module. In this project the student will develop, implement and evaluate radar algorithms on a microcontroller-based signal processing board to detect e.g. doppler velocity, range and approximate direction of nearby unmanned aerial vehicles (UAVs). The goal of this project is to evaluate the performance and capabilities of very small and light radar systems for detection of small aircraft. Experiments will be carried out to identify performance criteria (e.g. maximum range, field of view, etc.).

Type: Semester project
Period: 20.09.2011 - 20.01.2012
Section(s): EL IN MA ME PH
Type of work: 20% theory, 50% hardware/construction, 30% experiments
Requirements: background in radio technology desirable
Subject(s): RADAR, sensing, UAV
Responsible(s): Felix Schill, Meysam Basiri
URL: Click here

Antenna modification for a sub-miniature doppler RADAR

Alexander von Mach (MT)

Mid-air collision avoidance is an important topic in aviation in general. With the increasing number of UAVs (unmanned aerial vehicles) it becomes a challenge for aerial robotics as well. We are investigating a range of sensor modalities to sense and avoid other aircraft in the vicinity, including very small 25GHz doppler radar modules. There are very small and lightweight commercial RADAR modules available that weigh only a few grams. However, due to the small size of the antenna, the beam width is quite wide (40-80 degrees), limiting the detection range. The goal of this project is to design an antenna upgrade to be retrofitted to our RADAR module, that is lightweight, small enough to be integrated into our UAVs without adding too much aerodynamic drag, that will focus the beam to 10-20 degrees and increase the detection range. After evaluating different topologies (waveguide, Yagi, dish, etc.), prototypes have to be built and tested. During the design phase, multiphysics simulation software is available. For the experimental validation, a test rig may have to be constructed to achieve repeatable, reliable measurements of beam pattern and increase in signal strength.

Type: Semester project
Period: 20.09.2011 - 20.01.2012
Section(s): EL MA ME MT PH
Type of work: 20% theory, 20% simulation, 30% hardware/construction, 30% experiments
Requirements: background in radio technology desirable
Subject(s): RADAR, sensing, UAV
Responsible(s): Felix Schill, Meysam Basiri
URL: Click here

Increasing the evolvability of the Analog Genetic Encoding (AGE)

Dimitri Weideli (MT)

The Analog Genetic Encoding (AGE) representation has been successfully applied to a number of evolutionary robotics experiments, demonstrating to be a powerful representation for the evolution of neural networks and their topologies. Unfortunately, one drawback of AGE is its difficulty of evolving large networks. AGE generally evolves fully-connected networks, hampering the evolvability of the encoding.

In this project, the student is expected to implement in AGE a novel approach to overcome this problem. The student will implement one mechanism (more if time allows) to limit the interactions within AGE genes, which could potentially lead to an increased evolvability of AGE. The student will have to systematically analyze the tested approaches on a series of network matching benchmarks and assess their performances with respect to the standard AGE implementation. The evolutionary framework with the standard AGE representation is already available in the lab. The student is expected to implement the code needed to run the chosen benchmark experiments, collect and analyze the relevant statistics.

Type: Semester project
Period: 20.09.2011 - 14.01.2012
Section(s): IN MA MT PH SC SV
Type of work: 70% software, 30% theory
Requirements: familiarity with C/C++ would be an advantage
Subject(s): artificial evolution, age
Responsible(s): Andrea Maesani, Trevis Alleyne

Physics-based Simulator for Soft Robotics

Manuel Stöckli (ME)

We are currently developing mechanically soft and highly deformable modular robots. In order to test different aspects of these robots, we aim at creating a physics-based simulation platform.

The goal of this project will be to set up a simulation platform based on an existing 2D physics engine (Box2D) for soft modular robotics. After a familiarization phase with the software environment, the student is expected to implement a framework that allows for the simulation of a variable number of soft modules having customizable softness, size, number of connection points, internal connection control, simulation accuracy, and external environmental flow. A major challenge will be to enable connection and disconnection between soft objects based on the physical modelling of different attachment mechanisms. At the end of the project the simulation environment should be demonstrated in a few possible scenarios (self-reconfiguration, self-assembly, etc).

Type: Semester project
Period: 20.09.2011 - 13.01.2012
Section(s): IN MT PH
Type of work: 30%+theory +70%+software
Requirements: C++
Subject(s): Simulation, Soft robotics
Responsible(s): Jürg Markus Germann, Andrea Maesani

Soft spherical mirrors manufacturing process

Ionut Halasz (MT)

Spherical mirrors are popular tools, which allow, for e.g., enlarging the field of view of a camera or, redirecting the light-ray emitted by a laser. They are usually built of a glass or a metal surface covered by a reflective coating layer. The first goal of this project is to validate a new technique of soft spherical mirrors manufacturing, developed in the LIS laborathory. The technique bases on curing a PDMS polymer in a mould with a spherical opening and coating it with a micro-layer of gold. This new technique promises several advantages over the traditional mirror manufacturing process as quick mirror development, adjustable shape and low cost. The second goal of this project is to extend the functionality of a light-based shape sensor developed in the LIS laboratory by using the new type of flexible mirrors. In this project student is expected to: a. derive the optimal parameters of the flexible mirrors manufacturing process b. develop an analitical model of the process, which allows prediction over the process parameters c. apply the soft mirrors to the light-based shape sensor d. characterize the performance of the shape sensor with the mirror applied.

Type: Master project
Period: 20.09.2011 - 23.12.2011
Section(s): CH EL IN MA ME MT PH
Type of work: 60% experimental, 20% designing, 20% analytical
Requirements: no prior knowladge is required
Subject(s): spherical mirrors, polymers processing
Responsible(s): Michal Dobrzynski, Ramòn Pericet Camara

Bio-inspired arrangement of pixels on spherical artificial compound eye

Florian Gerlich (MT)

The compound eye of insects consists of a highly packed arrangement of individual sensing units called ommatidia. The arrangement of such ommatidia depends strongly on the eye region. Moreover, the available literature demonstrates that evolution has optimized such arrangement to provide efficient visual information for various functions, such as egomotion estimation or flight stabilization. In LIS, we are working on the design and fabrication of compound cameras inspired by insect eyes to be applied as vision sensors for flying robots. With the available technology, it is not possible to achieve the same pixel configuration as the one of ommatidia in insect eyes. Thus, alternative solutions must be found. The aim of the project is to evaluate various configurations of pixels on a camera with spherical geometry and demonstrate the most efficient solutions for egomotion estimation. The student is expected to set up the available simulator with various pixel arrangements according to the specifications of the available prototype designs. He is also expected to investigate egomotion estimation methods available in the literature that potentially fit the characteristics of the sensor and implement them in the simulator. Finally, he is expected to evaluate the different pixel arrangements of the sensor with respect to ground-truth values.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): MT
Type of work: Simulations 40%, programming 30%, theory 30%
Requirements: C++ or C
Subject(s): Bio-inspired engineering; computer vision; vision-based navigation; optical flow
Responsible(s): Ramòn Pericet Camara, Michal Dobrzynski
URL: Click here

Co-evolution of communication

Nicolas Uffer (MT)

The evolution of communication is usually only studied among members of the same group. Only very little is known about how the evolutionary dynamics between different species or groups influence the efficiency, composition and development of signals. In this project the student is expected to further develop an existing minimalistic simulation to investigate how co-evolving groups that compete for common resources influence each other in the evolutionary development of their signaling systems. More specifically, systematical investigations should be conducted by varying the parameters that control resource abundance and the way co-evolving species can interact. Interaction can be defined on two levels: The level of competition (e.g. competition for common resources) and the level of inter-specific communication. We expect that the level of communication overlap between species will influence the level of cooperation within and among species. Also, higher competition should lead to a decrease in interspecific cooperation and an increase in intraspecific cooperation. However, both are interdependent especially if there is a communication overlap between the species that may allow eavesdropping. Preliminary studies have suggested that this leads to a tragedy of the commons and low levels of cooperation on both levels. The student is expected to verify and quantify this hypothesis by running systematic simulation experiments and thorough statistical analysis.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): IN MT PH SC SV
Type of work: 20% theory, 80% software
Requirements: prior programming knowledge is an advantage
Subject(s): evolution, social behavior, communication
Responsible(s): Steffen Wischmann, Pawel Lichocki
URL: Click here

Efficient communication in groups of robot

David Mansolino (MT)

One of the key innovations during the course of evolution of life on earth has been the emergence of efficient communication systems. Conducting experimental evolution on social traits such as communication is unfortunately complicated by the extreme difficulty to assess individual fitness within groups and select individuals accordingly from one generation to the next. Therefore, individual-based models or simulated robots are used to study the link between inter-individual interactions and behavioral effects and to conduct unbiased analysis of the factors driving the evolution of social behavior. The degree of realism provided by such systems greatly exceeds current analytical and game-theoretic models and allows experiments that cannot be readily performed with real organisms. However, it is often unclear how well the evolved behavior in simulation also constitutes an efficient behavior in real embodied systems, such as robots. In this project we will consider communication behaviors that have been previously evolved and studied in physics based simulations. The student is expected to transfer a set of those behaviors onto a group of real EPuck robots. Careful analysis have to be conducted to identify crucial differences between simulation and reality (e.g., which aspects of the real environment and robot influence the effectiveness of different communication strategies). Different sensors and actuators have to be investigated for the potential to realize communication in groups of real robots. The system should be designed that an easy transfer of controllers developed in simulation to the real hardware is possible. It should also allow a quantitative comparison of group performance in both systems. Furthermore, it should be scalable to more than two EPuck robots that can interact autonomously in a common environment.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): EL ME MT
Type of work: 10% theorie, 40% software, 50% hardware
Requirements:
Subject(s): communication, social behavior
Responsible(s): Steffen Wischmann, Pawel Lichocki
URL: Click here

Task allocation algorithms for multiple robots

Paul Koch (MT)

The field of multi-agent systems is concerned with societies of autonomous agents (both artificial and natural) that interact to efficiently achieve their goals. This project focuses on a common problem of task allocation, where multiple robots distribute themselves to different tasks. The student will implement a task allocation algorithm proposed in LIS. The main idea is to use a threshold-like mechanism with a given target distribution of agents to tasks, which the system should display. The student will implement two versions of the algorithm: deterministic and probabilistic, for a real hardware (i.e., e-puck robots). This will require implementing also few simple robotic behaviors (e.g., line following, obstacle avoidance). A working demo is expected at the end of the project. If time permits the algorithm should be extended to handle task precedence constrains.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): EL IN MT
Type of work: 10% theorie, 50% software, 40
Requirements: programming skills (C), prior knowledge about e-puck is desirable
Subject(s): task allocation, decentralized algorithms
Responsible(s): Pawel Lichocki, Steffen Wischmann
 

Aerodynamical wing surface optimization using hierarchical structures for flying micro robots

Gilles Roulet (MT)

The Harvard Microrobotics Laborory has developed a series of biologically inspired flying micro robots. One of the major challenges in the development of these robots is to keep the energy consumption during flight at a minimum. A promising approach is to adapt gliding flight as part of their locomotion strategy. However, to date very little research has addressed gliding flight and its implications on the design of flying micro robots. In the animal kingdom, many gliding insects and mammals have hierarchical structures on their wings. Results from biomechanics research in Biology suggests that these structures can increase the lift to drag ratio of the wings and delay stall at high angles of attack, both which are desirable effects for flying micro robots. This Master Thesis will explore the manufacturing and testing of hierarchical structures on the wings of a micro scale gliding robot. The thesis will include the following work packages: 1) Summary of different wing surface structures found in butterflies, gliding mammals and fish. 2) Exploration of possible fabrication methods to create similar hierarchical structures using state of the art microfabrication tools. 3) Exploration of the effects of wing surface properties and architectures using a commercial Computational Fluid Dynamics (CFD) software. 4) Fabrication of a chosen set of structures on a given wing and its systematic comparison in the wind tunnel. 5) Characterization and discussion of the aerodynamical boundary layer effects based on the measurements. 6) Reporting and presentation of the results

Type: Master project
Period: 12.02.2011 - 12.09.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jürg Markus Germann, Mirko Kovac
 

Wing shape optimization for flying micro robots

Daniel Vogt (MT)

The Harvard Microrobotics Laborory has developed a series of biologically inspired flying micro robots. One of the major challenges in the development of these robots is to keep the energy consumption during flight at a minimum. A promising approach is to adapt gliding flight as part of their locomotion strategy. However, to date very little research has addressed gliding flight and its implications on the design of flying micro robots. In particular, no research has systematically explored the effects of wing morphology which is needed to perform efficient gliding flight in the Reynolds Number regime between 1000 and 10000. This Master Thesis will provide a first systematic exploration of the gliding performance of different wing shapes for micro robots. As a starting point, the project will focus on wing shapes found in proficient gliders in the animal kingdom and will test systematic variations of their wing shapes. The thesis will include the following work packages: 1) Extraction of wing shape of a first set of gliding butterflies (25 shapes) and variation of the forewing orientation. 2) Implementation of these shapes in a commercial Computational Fluid Dynamics (CFD) software. 3) Aerodynamical measurements of those shapes in CFD 4) Aerodynamical measurements of the same shapes in our low Re number wind tunnel 5) Validation of the CFD measurements and adaption of the CFD parameters to the wind tunnel experiments 6) Systematic optimization of the wing shape in CFD and validation of the most promising set of wing shapes in the wind tunnel. 7) Reporting and presentation of the results Added work can address the testing of these shapes on a flapping wing test bed in order to explore trade-offs between wing shapes optimized for gliding and flapping flight.

Type: Master project
Period: 12.02.2011 - 12.09.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jürg Markus Germann, Mirko Kovac
 

Next generation avionics for miniature UAV

Tilman Schneider (MT)

The goal of the project is to develop the next generation of a small UAV’s avionics. We include in avionics the autopilot and its sensors as well as its user interface (which, by nature, sits remotely in the ground station). On the embedded side, the goal is to develop a new autopilot based on modern microcontrollers and sensors, including the low-level software support. On the ground station side, the goal is to develop a novel, map-based interface for the UAV. Embedded development: - Study the suitability of Microchip microcontrollers for the purpose (in particular PIC32). - Design the general architecture of the autopilot based on provided constraints (sensor selection, etc.). - Develop a prototype for testing (using commercial demo-boards and sensor boards when available and/or designing custom PCB for sub-system to be tested). - Implement low-level software interface of the selected sensors and devices (in C). - Design an initial version of the autopilot in the final form factor (if time allows). Ground station development: - Develop a full-featured map widget (in Qt) - Implement a set of widgets (to be used as map overlays) to reflect the state of the autopilot and to interact with it.

Type: Master project
Period: 21.02.2011 - 12.09.2011
Section(s): MT
Type of work: 60% embedded system development, 40% ground control software
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
 

Smart actuators for soft flying robots

Adrià Manuel Pérez (EL)

At the LIS we are developing soft flying robots which will feature amongst other the ability to actively deform.

The goal of this masterproject is to find, build and integrate a suitable smart actuator for soft flying robots that matches several criterias such as to not impair the modules softness while being lightweight, small sized etc.
The student is expected to conduct a thorough conceptional analysis of novel smart actuators like e.g. electroactive polymers, electroactive paper or pneumatic actuators. Based on this analysis and evaluation one chosen technology will be investigated, designed and tested. Towards the end of the project the actuator will be dimensioned adequately and finally integrated into a soft flying robot. If time allows the new integrated mechanism will be tested in different situations.

Type: Master project
Period: 21.02.2011 - 15.07.2011
Section(s): EL
Type of work: 30% theory, 50% hardware, 20% testing
Requirements:
Subject(s): Artificial muscles, soft robotics
Responsible(s): Jürg Markus Germann, Ramòn Pericet Camara
 

Wing shape optimization for flying micro robots

Daniel Vogt (MT)

The Harvard Microrobotics Laborory has developed a series of biologically inspired flying micro robots. One of the major challenges in the development of these robots is to keep the energy consumption during flight at a minimum. A promising approach is to adapt gliding flight as part of their locomotion strategy. However, to date very little research has addressed gliding flight and its implications on the design of flying micro robots. In particular, no research has systematically explored the effects of wing morphology which is needed to perform efficient gliding flight in the Reynolds Number regime between 1000 and 10000. This Master Thesis will provide a first systematic exploration of the gliding performance of different wing shapes for micro robots. As a starting point, the project will focus on wing shapes found in proficient gliders in the animal kingdom and will test systematic variations of their wing shapes. The thesis will include the following work packages:

1) Extraction of wing shape of a first set of gliding butterflies (25 shapes) and variation of the forewing orientation.
2) Implementation of these shapes in a commercial Computational Fluid Dynamics (CFD) software.
3) Aerodynamical measurements of those shapes in CFD
4) Aerodynamical measurements of the same shapes in our low Re number wind tunnel
5) Validation of the CFD measurements and adaption of the CFD parameters to the wind tunnel experiments
6) Systematic optimization of the wing shape in CFD and validation of the most promising set of wing shapes in the wind tunnel.
7) Reporting and presentation of the results Added work can address the testing of these shapes on a flapping wing test bed in order to explore trade-offs between wing shapes optimized for gliding and flapping flight.

Type: Master project
Period: 21.02.2011 - 01.07.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Flying robotics
Responsible(s): Jürg Markus Germann, Mirko Kovac
 

Aerodynamical wing surface optimization using hierarchical structures for flying micro robots

Gilles Roulet (MT)

The Harvard Microrobotics Laborory has developed a series of biologically inspired flying micro robots. One of the major challenges in the development of these robots is to keep the energy consumption during flight at a minimum. A promising approach is to adapt gliding flight as part of their locomotion strategy. However, to date very little research has addressed gliding flight and its implications on the design of flying micro robots. In the animal kingdom, many gliding insects and mammals have hierarchical structures on their wings. Results from biomechanics research in Biology suggests that these structures can increase the lift to drag ratio of the wings and delay stall at high angles of attack, both which are desirable effects for flying micro robots. This Master Thesis will explore the manufacturing and testing of hierarchical structures on the wings of a micro scale gliding robot. The thesis will include the following work packages:

1) Summary of different wing surface structures found in butterflies, gliding mammals and fish.
2) Exploration of possible fabrication methods to create similar hierarchical structures using state of the art microfabrication tools.
3) Exploration of the effects of wing surface properties and architectures using a commercial Computational Fluid Dynamics (CFD) software.
4) Fabrication of a chosen set of structures on a given wing and its systematic comparison in the wind tunnel.
5) Characterization and discussion of the aerodynamical boundary layer effects based on the measurements.
6) Reporting and presentation of the results

Type: Master project
Period: 21.02.2011 - 01.07.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jürg Markus Germann, Mirko Kovac
 

Precision control of a flying robot

Ryan Ammoury (MT)

The Swinglet developed by Sensefly is a fixed-wing flying platform that can autonomously operate in the air. Applications include exploration of remote environment, aerial photography, etc. Many of these tasks require position control for navigation. To achieve this, GPS is used to follow trajectories thanks to a vector-field based algorithm. However, for some applications where a precise position control is required (accuracy under 3 meters along all 3 axes), the current method has shown some limits. The goal of this semester project is to characterize and improve the trajectory control for the Swinglet platform. In a first step, a method to establish a ground-truth (reliable measurement of position) has to be developed for the characterization. Several methods to improve the position control will then be explored, such as improving the trajectory following algorithm, augmenting the GPS with other sensors or improving the model for wind compensation, etc. The final demonstration should show the platform realize a precise position control by flying through a 6-meters-wide arch, which corresponds to the scenario of the IMAV 2011 competition.

Type: Semester project
Period: 21.02.2011 - 30.06.2011
Section(s): ME MT
Type of work: 30% theory, 40% experiments, 30% software
Requirements:
Subject(s): aerial robotics, control&systems
Responsible(s): Adrien Briod, Antoine Beyeler

Evolution of referential communication

Fabian Santi (SV)

Communication, a ubiquitous behavior in the animal kingdom, can be defined as a process wherein the behavior of one individual influences the future behavior of another individual. Many animals, such as honeybees, rely on referential communication--the transition from private to social information. Private information is acquired via direct interaction of an individual with the environment, as opposed to social information that is acquired through the actions (e.g. waggle dance), body structures (e.g. shapes or colors) or products (e.g. pheromones) of other individuals. Often, referential communication relies on multicomponent signals (i.e., signal comprising more than one informational component). In this project the student is expected to built upon and extend a model of referential communication that has been developed in the LIS lab. In particular, experimental evolution experiments have to be conducted with this model to investigate the following questions: Does evolution converge toward distinct communication strategies? If there is variation in how individuals use referential communication, what are the reason for this variation and what are the consequences with respect to fitness? If referential communication is based on multicomponent signals, how can we distinguish individual components? And how do individual components develop during the evolutionary process? Of particular interest is the transition from private to social information, i.e., how much information is provided by the sender and how much of this information is acquired by the receiver?

Type: Semester project
Period: 21.02.2011 - 16.06.2011
Section(s): IN MT PH SC SV
Type of work: 30% theory, 70% software
Requirements: some prior programming knowledge is an advantage
Subject(s): evolution, communication
Responsible(s): Steffen Wischmann, Pawel Lichocki
URL: Click here

Measuring the atomic units of fly olfactory behavior

Nicolas Beuchat (SV)

What are the underlying algorithms of seemingly complex animal behaviors? Answers to this question will greatly advance the development of robotic systems capable of producing complex animal-like behaviors. One means of identifying the structure of an unknown system is to probe it with controlled stimuli. The output or behavior of the system can then serve as a dataset for exploring potential explanatory models.

For this project the student will use state-of-the-art machine vision algorithms and data acquired with automated, custom-built behavioral odor delivery chambers to computationally dissect the actions of freely moving flies in response to odors. Together we will explore this data to explain the behavioral strategies of odor avoidance and attraction as well as the degree of individual differences in behavior or "personalities". These analyses will provide a starting point for developing robotic controllers that mimic real biological olfactory behaviors.

This project will be done in the Laboratory of Intelligent Systems (EPFL) and in collaboration with the Benton Lab (UNIL). Therefore this is an extremely unique project right at the interface between engineering, computer science, & neurobiology.

Type: Semester project
Period: 21.02.2011 - 10.06.2011
Section(s): IN MA MT PH SC SV
Type of work: 50% software 30% research, 20% theory
Requirements: familiarity with Matlab and/or C++
Subject(s): Machine Vision, Animal Behavior, Neuroscience
Responsible(s): Pavan Ramdya, Thomas Schaffter
URL: Click here

2010


Integrating opti-flow-based control on a mini-UAV

Ryan Ammoury (MT)

Recent progresses in our lab enabled demonstration of fully autonomous low-altitude flight and collision avoidance with a 400-gram flying wing using a set of optic-flow sensors. This project aims at developing an integrated version of the control strategy and vision system that is practical for use on commercial mini-UAVs. To this end, the candidate will carry out a quick comparison of the most recent optic-flow sensors before theoretically comparing the minimal requirements (number and orientation of viewing directions, dynamic range, etc.) to achieve collision avoidance, terrain following and altitude control for precise landing. For each of these cases, several control strategies and integration schemes will be devised and theoretically compared. The most promising ones will then be implemented and systematically tested in flight, which will require some interfacing and programming efforts. Finally, a concept will be devised for integrating the optic-flow sensors into the existing airframe while limiting their impact on mass budget, aerodynamic drag and handling requirements.

Type: Master project
Period: 20.09.2011 - 31.03.2012
Section(s): MT
Type of work: 50% hardware, 20% software, 30% tests
Requirements:
Subject(s): Visual system, aerial robotics
Responsible(s): Jean-Christophe Zufferey, Beyeler Antoine
URL: Click here

2D Deformation Mechanism for Soft Robots based on SMA

Sélim Gawad (MT)

Modular robots can change morphology to adapt to changing tasks and environments by rearranging the connectivity of their own modules. In contrast to fixedmorphology robots, modular robots have the potential to be more flexible, robust and cheap. However, many challenges prevent the full realization of these potentials. Over the last two decades several sophisticated module designs have been proposed, most them featuring hard building blocks with rigid connection mechanism. Although this design guarantees controllability and stability, it minimizes flexibility. One solution to overcome the issue of rigidity in large numbered modular systems is to use modules that could become mechanically soft when desired. Hence, at the LIS we are investigating soft modular robots which will feature amongst other the ability to actively deform. The goal of this semesterproject is to develop a novel deformation mechanism based on Shape Memory Alloy coils. This project focuses in a first phase on the fabrication of SMA coils (i.e. most importantly the design of a winding setup), in a second phase on the conceptual design and the dimensioning of the integration of the SMA's into soft 2D modules. At the end of the project the implemented solution featuring the SMA coils will be tested to assess its performance as deformation mechanism.

Type: Semester project
Period: 20.09.2011 - 13.01.2012
Section(s): CH EL ME MT MX PH
Type of work: 40% electronics, 40% fabrication, 20% testing
Requirements: Creativity and research attitude
Subject(s): Electronics, Physics, Fabrication
Responsible(s): Jürg Markus Germann, Michal Dobrzynski

Design and fabrication of an actuation system for an active deformable spherical soft robot

Gabriel Safar (MT)

Modular robots can change morphology to adapt to changing tasks and environments by rearranging the connectivity of their own modules. In contrast to fixed morphology robots, modular robots have the potential to be more flexible, robust and cheap. However, many challenges prevent the full realization of these potentials. Over the last two decades several sophisticated module designs have been proposed, most them featuring hard building blocks with rigid connection mechanism. Although this design guarantees controllability and stability, it minimizes flexibility. One solution to overcome the issue of rigidity in large numbered modular systems is to use modules that could become mechanically soft when desired. Hence, at the LIS we are investigating soft modular robots. When in a soft state controllability of the morphology of the single module as well as of the robot becomes highly challenging. Towards orchestrating the rich number of degrees of freedom of one module, we aim at developing an active deformation mechanism of our spherical modules. The goal of this project is to develop a novel deformation mechanism for spherical soft modules. The main requirements at this mechanism are to be as soft as possible (to keep the flexibility of the modules), to enable high global deformation of the membrane (e.g. deformation from a spherical shape into an ellipsoid), and to be small and lightweight. This project is divided into several workpackages: (i) conceptual design of the deformation mechanism, (ii) review and choice of a "soft" actuator technology, (iii) design and dimensioning of the actuator, (iv) fabrication of the deformation mechanism, (v) testing of the performance of the mechanism when integrated in soft modules.

Type: Master project
Period: 20.09.2011 - 13.01.2012
Section(s): CH EL MA ME MT MX PH
Type of work: 30% theory, 70% hardware
Requirements: Creativity and research attitude
Subject(s): Electronics, Physics, Fabrication
Responsible(s): Jürg Markus Germann, Ramòn Pericet Camara

Measuring multi-robots performance in task allocation

Chen Xiang (IN)

The field of multi-agent systems is concerned with societies of autonomous agents (both artificial and natural) that interact to efficiently achieve their goals. In this work, the student will implement simple behaviors for the e-puck robot allowing it to perform basic sub-tasks (e.g. navigate to base, bring an item, avoid obstacles and other robots). The goal of this project is to measure quantitatively in simulation the performance of a team of robots trying to collaboratively solve a given problem (e.g. search and forage for items) with respect to the number of robots in the team. If time permits, the evaluation of the results with robotic hardware should be performed. In the future, the results of this experiment will allow to devise realistic models of scalability of multi-robot systems.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): IN MA MT
Type of work: 60% software, 40% experiment
Requirements: C
Subject(s): multi-robots, scalability, behavioral-based programming
Responsible(s): Pawel Lichocki, Steffen Wischmann

The role of chance for the evolution of communication

Jérôme Waeber (MT)

Evolution usually doesn't follow a straight line and the outcome often depends on noise in mutation, selection or environmental conditions. The aim of this project is to investigate the role of chance, that is, contingencies in evolutionary history, on the evolution of communication and social behavior. For this purpose, the student is expected to develop a minimalistic simulation to test the divergent character of evolution and its consequences. More specifically, the student should build upon a previously developed simulation that investigated the evolution of referential communication. In the last project we found that individuals can use two fundamentally different ways of communicating the location of a displaced food location, either via their movement behavior or via direct signaling channels. The student is expected to extend this simulation so that individuals have genes that determine which of the two communication strategies to use. Replicated evolution experiments need to be performed, the information content of signaling strategy have to be quantified as well as their evolutionary trajectory. Eventually, we want to disentangle the effect that stochastic events during evolution have from the fitness benefits of a specific order of beneficial mutations.

Type: Semester project
Period: 20.09.2011 - 23.12.2011
Section(s): IN MT PH SC SV
Type of work: 20% theory, 80% software
Requirements: some prior programming knowledge is an advantage
Subject(s): evolution social behavior communication
Responsible(s): Steffen Wischmann, Andrea Maesani
URL: Click here

Altitude estimation for an indoor flying robot

Thibault Dupont (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr will be able to resist collisions with obstacles, and to carry out autonomous tasks with a minimal sensor suite.

The goal of this project is to work on the altitude estimation of this robot using only two extremely lightweight passive sensors: a 3-axis accelerometer and a barometric pressure sensor. None of these sensors provide a good enough altitude estimate by themselves. It is expected that a fusion of the two will provide a reliable altitude or vertical speed estimate. The work will consist in implementing such an algorithm in simulation first (using data logs from real flights) and to code it in C on the embedded micro-controller (provided). If time allows, an altitude controller using the sensor data will be implemented.

Type: Semester project
Period: 21.02.2011 - 30.06.2011
Section(s): MT
Type of work: 25% theory, 25% simulation, 25% programming, 25% testing
Requirements:
Subject(s): sensor fusion, microcontroller programming
Responsible(s): Adrien Briod, Adam Klaptocz
URL: Click here

Tactile feedback system for blind people

Seifeddine Mejri (MT)

The first part of the project is to understand how people perceive small amplitude, high frequency vibrations punctually applied to the skin of the head. This knowledge will be used in a second part of the project to design a vibration-based feedback system for blind or visually impaired people. The feedback system should be able to transmit at least two types of information (for e.g., about the distance to an object and its position in an unknown environment), and will later on be integrated with a collision alert device developed in our lab.

In this project student is expected to: design an experimental protocol and perform a set of tests on a group of users, answering the questions: how many motors should be used, what is their optimal placement and how to modulate the vibration signal to provide the feedback in a clear and understandable way. The obtained results should be statistically reliable. Based on the results, a feedback system should be designed, built and tested.

Type: Master project
Period: 01.02.2011 - 30.06.2011
Section(s): EL IN MA ME MT SC SV
Type of work: 50% experimental work, 25% theory, 15% software, 10% hardware
Requirements: no prior knowladge required
Subject(s): sensory substitution, vibration perception
Responsible(s): Michal Dobrzynski, Steffen Wischmann

Long-range mini-drone for atmospheric sampling

Cédric Schwab (MT)

This project will be carried out in collaboration with the EPFL spin-off senseFly and MeteoSuisse. The goal is to further develop the embedded electronics of an existing mini-drone for atmospheric sampling. Within this semester project, the candidate will mainly focus on enabling long-range, bidirectional communication between the ground control station (GCS) and the remote mini-drone. Solutions (radio modems) to extend the range up to 10-30 km will be analyzed. The most promising module fitting the limited available payload will be selected, interfaced to the existing autopilot, integrated in the mini-drone and fully characterized. Optionally, the candidate will prototype a release mechanism to allow pulling the mini-drone underneath an helium-filled balloon and releasing it at relatively high altitude. Test flights will be conducted with MeteoSuisse in Payerne.

Type: Semester project
Period: 21.02.2011 - 20.06.2011
Section(s): EL MT
Type of work: 30% hardware, 40% software, 30% flight testing
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler

Evolutionary dynamics in mutualistic networks

Joël Rey (MT)

Mutualism refers to the interactions between different species that have beneficial fitness effects on both partners. This is a widespread phenomena in biology. Examples are the symbiosis between plant roots and fungi, interactions between ants and aphids or cleaner fish and their clients. Measuring the exact fitness benefits in biological individuals is difficult and, thus, explaining the the evolution of mutualistic networks becomes challenging especially when more than one species is involved. The aim of this project is to develop a general, yet rather simple, individual-based model of multi-species mutualism that allows to investigate the co-evolutionary dynamics in mutualistic networks. This model should be implemented in a two dimensional grid-world simulation that allows easy manipulation of resource distribution and possible species interactions via signaling. More specifically, systematical investigations should be conducted by varying the parameters that control resource abundance and the number of co-evolving species. An interface between the simulation and an existing evolutionary algorithm library (ECJ, TEEM, or similiar) has to be implemented to conduct experiments investigating the likelihood of the evolution of mutualistic inter-species relationships and their evolutionary stability.

Type: Semester project
Period: 21.02.2011 - 16.06.2011
Section(s): IN MT PH SC SV
Type of work: 30% theory, 70% software
Requirements: some prior programming knowledge is an advantage
Subject(s): evolution, mutualism
Responsible(s): Steffen Wischmann, Andrea Maesani
URL: Click here

Co-evolution of communication

Thibaut Watrin (IN)

The evolution of communication is usually only studied among members of the same group. Only very little is known about how the evolutionary dynamics between different species or groups that influence the efficiency, composition and development of signals. For this purpose, the student is expected to extend a simulation model based on the Epuck robot to investigate how co-evolving groups that compete for common resources influence each other in the evolutionary development their signaling systems. For this project we will consider two groups of robots in a common foraging area. Based on previous experiments that did not involve co-evolution, the student is expected to implement an algorithm that allows to co-evolve two groups of robots that have the ability to communicate via light signals. This algorithm should be designed in a way that simulations can be run in a distributed fashion on a high performance computer cluster. Systematic experiments have to be conducted to identify appropriate selection mechanisms. Further the student is expected to analyze the evolutionary development of signaling in both groups of robots and its relation to fitness development. In particular, we are interested in the influence that signaling behavior in one group has on the signaling behavior of the other group during evolutionary development.

Type: Semester project
Period: 21.02.2011 - 16.06.2011
Section(s): IN MT PH SC SV
Type of work: 20% theory, 80% software
Requirements: some prior programming knowledge is an advantage
Subject(s): evolution, social behavior, communication
Responsible(s): Steffen Wischmann, Pawel Lichocki
URL: Click here

Perspectives d'utilisation d'un mini-drone dans l'agriculture (projet externe à Changins)

Yannick Bergem (MT)

Les photos aériennes de cultures et de pâturages sont extrêmement intéressantes à analyser. Elle donne tout d'abord une vision globale des champs. Dans le spectre visible elles permettent d'observer l'homogénéité d'une culture ou l'avancement de la forêt dans un pâturage. Avec des images infrarouges, il est possible de quantifier la densité d'azote dans le sol qui est un facteur déterminant dans la croissance des cultures. Dans la recherche en agriculture ces images peuvent également servir pour étudier différentes parcelles de cultures tests. Il existe actuellement plusieurs outils à disposition pour obtenir des vues aériennes, dont le satellite ou les prises d'avion. Les images satellites dépendent de la couverture nuageuse du site à observer et ont une résolution assez faible. Les photos prises depuis des avions sont moins dépendantes de la météo et ont une bien meilleure résolution. Cependant le vol en avion reste cher. Les drones de par leur légèreté et leur flexibilité d'utilisation offrent un outil particulièrement adapté. Ils permettent de prendre des images hebdomadaires des champs à faible coût. Dans ce projet nous allons explorer les différentes applications possibles que peuvent nous apporter un drone. Une analyse de la précision nécessaire pour les différentes applications serait faite. Nous allons également essayer de reconnaitre de manière automatique les types de cultures à partir des photos.

Type: Master project
Period: 20.09.2010 - 25.04.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Dr. Raphaël Charles
 

Embedded control for a crawling robotic insect

Christian Heimlich (MT)

We have recently developed a prototype ambulatory microrobot based upon biological principles derived from cockroach locomotion., This device weighs less than one gram and can achieve locomotion speeds greater than four body lengths per second., Current work on the mechanics and design is focused on decreasing the size, optimizing stride motion, and integrating various attachment devices. However, the device is currently tethered for control and power., This project will involve the development and integration of power and control circuitry to bring the robot to full autonomy., There are four primary components of this circuit: 1) high voltage boost conversion: we have existing designs but have not yet integrated them into the body of the robot with small-scale components. 2) simple microcontroller for programmable gait control. 3) power source/circuitry: we have appropriate batteries and the capabilities to physically modify them to desirable form factors. This will need to be determined along with appropriate power conditioning circuitry. 4) interface: we need a way to program the controller to execute a variety of gaits. Time permitting, we may also want to explore one or two simple sensors to incorporate in the body of the robot. All circuit prototypes will be generated in-house, first with bench-top breadboards, then migrating down to custom circuit boards based upon flex circuits (which we can make in-house) and small discrete components or bare die components (whichever is commercially available). Expected final demonstration: embedded circuits to enable autonomous locomotion of a cockroach-scale robot

Type: Master project
Period: 21.09.2010 - 15.04.2011
Section(s): EL MT
Type of work:
Requirements:
Subject(s): microrobotics, microelectronics
Responsible(s): Adrien Briod, Robert Wood
URL: Click here

Design, Manufacturing and Implementation of Fiber Adhesives-Based Perching Mechanism for Flying Robots

Ludovic Daler (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr will be able to resist collisions with obstacles, detect the force of impact, attach to surfaces and to autonomously take off again after a collision. These types of interactions bring it closer to the capabilities of insects which are much more aware of their environment than current indoor flying robots.
The goal of this project is to design a fiber adhesives-based attachment mechanism for the AirBurr. This mechanism should be able to attach to different smooth surfaces, support the weight of the AirBurr, and detach for takeoff. This project will involve the design and dimensioning of a suitable attachment mechanism, CAD design and the manufacturing of one or several prototypes. The mechanism will then be integrated into a flying AirBurr platform.

Type: Master project
Period: 19.07.2010 - 30.03.2011
Section(s): MT
Type of work: 30% theory, 50% mechanical design, 20% testing
Requirements:
Subject(s): Flying Robots, Gripping, Mechanics
Responsible(s): Adam Klaptocz, Sitti Metin
URL: Click here

Second generation of OptiPilot frontend

Sandro Montanari (ME)

At the Laboratory of Intelligent Systems, we have developed a very successful control strategy for unmanned aircraft named optiPilot. This control strategy enables low-altitude flight, terrain following, collision avoidance, automatic take-off and landing, etc. It uses a series of divergent optic-flow detectors to sense proximity all around the aircraft longitudinal axis. This project is about designing a new visual system including 12 new-generation motion detectors. This includes choosing the materials, designing the underlying PCB as well as the connecting parts. Some code will then be implemented in order to interface the 12 sensors to the existing autopilot in an optimized way. Finally, the entire visual system will be characterized by means of well-thought experiments.

Type: Semester project
Period: 21.09.2010 - 03.02.2011
Section(s): MT
Type of work: 30% theory, 40% hardware, 30% software
Requirements: electronic design, embedded programming
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
 

Electroadhesion for flying robots

Michael Dommer (MT)

At the LIS we are developing flying robots which will feature the ability to attach to surfaces (like e.g. the AirBurr).

The goal of this semesterproject is to investigate a novel clamping technology called "electroadhesion". This adhesion technology is electrically controllable and induces electrostatic charges on a substrate using a power supply connected to compliant pads situated on the robot. Electroadhesion enables high-clamping forces on a wide variety of substrate.

Electrostatic forces occur between the substrate material and electroadhesive pads. These pads are comprised of conductive electrodes that are deposited on the surface of a polymer. When alternate positive and negative charges are induced on adjacent electrodes, the electric fields set up opposite charges on the substrate and thus cause electrostatic adhesion.

This project will involve the fabrication and testing of electroadhesive pads and the design of supply electronics. If time allows the clamping technology will be integrated into a flying robot.

Type: Semester project
Period: 21.09.2010 - 03.02.2011
Section(s): EL MA ME MT
Type of work: 40% electronics, 40% fabrication, 20% testing
Requirements:
Subject(s): Electroadhesion, Electronics, Polymers, Fabrication methods
Responsible(s): Jürg Markus Germann, Adam Klaptocz

Automated extraction of fly behavior

Adrian Cabrera (MT)

What is the structure and connectivity of brain networks that underlie the behavior of even a simple organism like the fly? Answers to this question will have considerable impact on the development of robotic and artificial systems capable of producing complex animal-like behaviors. One means of identifying the structure of an unknown system is to probe it with controlled stimuli. The output or behavior of the system can then serve as a dataset for exploring possible underlying network mechanisms through a reverse engineering of the system.

The focus of this project is to develop the tools necessary to begin this kind of systems analysis in fly olfaction (smell). The student will first design and build fly chambers as well as an automated odor delivery system. Using image analysis software for recording the behavior of individual flies as they respond to computer-controlled odor stimuli, the student will develop quantitative measures of this data in order to shed light on topics including the degree of variability and individuality observed in fly olfactory behavior.

This project will be done in the Laboratory of Intelligent Systems (EPFL) and in collaboration with the Benton Lab (UNIL). Therefore this unique project is right at the interface between engineering, computer sciences, and neurobiology.

Type: Semester project
Period: 20.09.2010 - 03.02.2011
Section(s): MT
Type of work: 10% theory, 60% hardware/electronics, 30% software
Requirements: Component design and fabrication, Electronics, C++/Matlab
Subject(s): Animal Behavior, Image Analysis
Responsible(s): Pavan Ramdya, Thomas Schaffter

Optic flow detection using a radial-polar image sensor

Christophe Paccolat (MT)

We recently demonstrated fully autonomous low-altitude flight control and collision avoidance with a small flying wing in natural environments using a series of 7 computer mouse sensors as optic flow detectors. This semester project aims at replicating this feat using a single imager with a custom-designed arrangement of pixels. The embedded processor that will be used to process the images is a Blackfin microcontroller. The main focus of the project will be on optic flow extraction. The optic flow extraction will be done using a gradient-based method, such as the Lucas-Kanade algorithm. This kind of methods and the different versions of the algorithms will first be studied in the literature. Then, the routines will be implemented in Matlab in order to be characterized and compared using different sequences of images taken in flight, some sequences with impending collisions and other without. The goal will be to study the false positives and false negatives collision detections. If the obtained results and the time permit it, the routines will then be optimized and implemented on the Blackfin microcontroller, for real-time testing and optic flow characterization.

Type: Semester project
Period: 21.09.2010 - 03.02.2011
Section(s): MT
Type of work: 15% literature review, 70% software development, 15% tests
Requirements: good programming skills
Subject(s): aerial robotics
Responsible(s): Yannick Gasser, Jean-Christophe Zufferey
URL: Click here

Optimal team composition in division of labor

Patrick Farnole (MT)

The field of multi-agent systems is concerned with societies of autonomous agents (both artificial and natural) that interact to efficiently achieve their goals. In this work, the student will evolve in silico teams of agents that are capable of displaying division of labor (i.e. agents specialize in a strict subset of all tasks). The main question is how diverse a team needs to be in order to efficiently divide a labor for a given problem. The goal of this project is to test quantitatively how varying the level of relatedness between the workers affect the speed of the evolution and the final performance of the evolved teams. From engineering perspective, addressing this issue could lead to new scalable optimization techniques suited for systems composed of large number of agents. From biological perspective, this may shed light on processes shaping the colony structure and provide some explanations for the variance of within colony relatedness that has been observed in nature.

Type: Semester project
Period: 20.09.2010 - 03.02.2011
Section(s): IN MA MT
Type of work: 25% theory, 50% software, 25% analysis
Requirements:
Subject(s): team optimization, division of labor
Responsible(s): Pawel Lichocki, Steffen Wischmann
 

User-friendly GUI for mini-drones

Tinasoa L. Ramahaly (MT)

At LIS, we developed a micro-UAV designed to be easy-to-operate and used by untrained users. The ground control station (GCS) software currently used to monitor and control the swinglet is a flexible monitoring software for research and development in robotics. While it served this purpose well, its very generic flexibility has become an obstacle to create a GCS software offering an experience as user-friendly as our micro-UAV’s hardware. The goal of the project is to redesign a GUI specifically designed for this micro-UAV, based on the existing communication architecture. The GUI should also be usable with touch-screen-based mobile devices. The project includes: - state of the art of the existing GCS software and analysis of the functionality of the micro-UAV - design of a GUI following suitable Human-Machine Interface principles and based on a sane implementation architecture - implementation of a prototype GUI with basic functionalites and architectural foundations suitable for future completion/expansion

Type: Semester project
Period: 29.09.2010 - 31.01.2011
Section(s): IN MT
Type of work:
Requirements:
Subject(s): Flying robots
Responsible(s): Antoine Beyeler, Jean-Christophe Zufferey
 

Flight control for image acquisition

Loic Zimmermann (MT)

The swinglet CAM is a micro-UAV weighing 500 grams and carrying a high-resolution camera that is fixed with respect to the airframe. Many applications, such as aerial photography for architecture or real-estate or inspection of non-horizontal surfaces, require images taken from a specific location and non-vertical orientation. The goal of the project is to implement strategies to that allow the acquisition of arbitrary oblique images with the swinglet CAM. The project includes: 1) characterisation of the swinglet CAM’s dynamics and IMU performance; 2) characterisation of the image acquisition timing; 3) (if needed) design and implementation of a shutter timing readout; 4) implementation of one (or more) strategy for arbitrary oblique image acquisition; 5) characterisation of the strategy.

Type: Semester project
Period: 20.09.2010 - 31.01.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial Robotics
Responsible(s): Antoine Beyeler, Jean-Christophe Zufferey
 

Self calibration of OptiPilot frontend

Raphaël Zaugg (MT)

At the Laboratory of Intelligent Systems, we have developed a very successful control strategy for unmanned aircraft named optiPilot. This control strategy enables low-altitude flight, terrain following, collision avoidance, automatic take-off and landing, etc. It uses a series of divergent optic-flow detectors to sense proximity all around the aircraft longitudinal axis. This project entails the setup of a visualization and data-recording software for a new OptiPilot frontend comprising 12 motion sensors. In a second step, the new generation motion chips should be systematically characterized and compared with the last generation sensors. Finally, a viewing-direction calibration routine should be implemented to allow self-calibration of the system based on gyroscopic data.

Type: Semester project
Period: 20.09.2010 - 31.01.2011
Section(s): MT
Type of work:
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
 

Sensor-laden roll cage for an indoor flying robot

Florentin Marty (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr will be able to resist collisions with obstacles, detect the force of impact, attach to surfaces and to autonomously take off again after a collision. These types of interactions bring it closer to the capabilities of insects which are much more aware of their environment than current indoor flying robots.

The goal of this project is to re-design the current protection cage of the AirBurr to increase its robustness to collisions and ease of construction, as well as to integrate sensors directly into this cage. Such a cage should be able to not only survive a collision with an obstacle, but to be able to detect this obstacle, its position, texture and force of contact. The project will involve an analysis of impact forces, selection and characterization of materials, CAD design of cage joints and the construction of one or several prototype roll cages. This project will focus on mechanical design, but will also include sensor selection and, if time allows, signal processing of integrated sensors.

Type: Semester project
Period: 13.09.2010 - 31.01.2011
Section(s): EL ME MT MX
Type of work: 20% theory, 30% materials, 50% mechanics
Requirements:
Subject(s): Soft Materials, Mechanics, Flying Robots
Responsible(s): Adam Klaptocz, Adrien Briod

Gripping-based attachment mechanism for an indoor flying robot

Steven Roelofsen (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr will be able to resist collisions with obstacles, detect the force of impact, attach to surfaces and to autonomously take off again after a collision. These types of interactions bring it closer to the capabilities of insects which are much more aware of their environment than current indoor flying robots.

The goal of this project is to design a gripping, spike-based attachment mechanism for the AirBurr. This mechanism should be able to attach to different rough surfaces, support the weight of the AirBurr, and detach for takeoff. This project will involve the design and dimensioning of a suitable attachment mechanism, CAD design and the manufacturing of one or several prototypes. If time allows the mechanism will be integrated into a flying AirBurr platform.

Type: Semester project
Period: 01.08.2010 - 31.01.2011
Section(s): EL MA ME MT
Type of work: 50% mechanics, 20% design, 30% software
Requirements:
Subject(s): Flying Robots, Gripping, Mechanics
Responsible(s): Adam Klaptocz, Jürg Markus Germann

Contact sensing on a flying robot

Kevin Jamolli (MT)

At the LIS we are developing a novel flying platform called the AirBurr which has the ability to not only fly indoors, but to physically interact with its environment. The AirBurr will be able to resist collisions with obstacles, detect the type and force of impact, and navigate in the environment by using these interactions.
The goal of this project is to work on the contact sensing ability of the robot. Several solutions can be explored to sense contact or interaction forces through distributed sensors (for example short-range distance sensors, force sensors, ...). A solution appropriate for the AirBurr platform requirements (crash-resistant, lightweight, ease of integration in the current structure) will have to be selected and implemented in hardware. The sensors will be interfaced by a micro-controller (provided) programmed in C, and if time allows, some signal processing and even simple behaviors using the sensor data will be implemented.

Type: Master project
Period: 21.09.2010 - 21.01.2011
Section(s): EL ME MT
Type of work: 20% theory, 60% hardware, 20% sofware
Requirements:
Subject(s): force sensing, flying robots, signal processing
Responsible(s): Adrien Briod, Adam Klaptocz
URL: Click here

Passive connection mechanism for soft self-reconfigurable robots

Laurent Blanchet (MT)

At the LIS we are investigating soft self-reconfigurable flying robots., On the one hand, self-reconfigurable robots offer many potential advantages with respect to traditional robots thanks to their ability to adapt their morphology to a given task or environment. Self-reconfigurable robots are expected to show significant abilities like e.g. adaptability, fault tolerance, flexibility, versatility, low cost. On the other hand, soft flying robots benefit from several advantages compared their rigid counterparts in indoor environments. Using their softness the modules can conform to obstacles and therefore adapt to the environment requiring less sensorial and controlling computation without causing damage. Furthermore, soft flying robots can squeeze through openings smaller than their nominal size and are dexterous to reach confined spaces, thus have a more enhanced field of operation. A major challenge that has to be tackled is to enable modules to attach and detach when desired. The goal of this semester project is to explore different passive connection technologies (materials, substrates) for soft modules. The work will include a review on existing technologies (like e.g. surface tension, magnets) and materials, and subsequently the implementation and testing of the best suited solutions.

Type: Semester project
Period: 21.09.2010 - 13.01.2011
Section(s): MA ME MT MX
Type of work: 30%+theory +40%+hardware +30%+testing
Requirements:
Subject(s): Mechanics
Responsible(s): Jürg Markus Germann, Ramòn Pericet Camara
 

Alternative propulsion methods for soft flying robots

Martin Liniger (MT)

At the LIS we are investigating soft self-reconfigurable flying robots., On the one hand, self-reconfigurable robots offer many potential advantages with respect to traditional robots thanks to their ability to adapt their morphology to a given task or environment. Self-reconfigurable robots are expected to show significant abilities like e.g. adaptability, fault tolerance, flexibility, versatility, low cost. On the other hand, soft flying robots benefit from several advantages compared their rigid counterparts in indoor environments. Using their softness the modules can conform to obstacles and therefore adapt to the environment requiring less sensorial and controlling computation without causing damage. Furthermore, soft flying robots can squeeze through openings smaller than their nominal size and are dexterous to reach confined spaces, thus have a more enhanced field of operation.

However, the actuation of a soft flying robot remains a major challenge. The goal of this semester project is to explore the use of an alternative propulsion method for soft flying robots in the context of self-reconfigurable robots. Taking inspiration from nature, the idea is to exploit Brownian motion (or ambient motion) in the environment to actuate modules. This alleviates the need for actuators for locomotion of the individual module and thus greatly simplifies module design. Accordingly, a conceptual analysis of different actuation (and control) principles to propulse a soft flying module using environmental forces will be realised. Then the work will include the design and implementation of a prototype for the validation of the method.

Type: Semester project
Period: 21.09.2010 - 10.01.2011
Section(s): EL MA ME MT
Type of work: 50% mechatronics, 30% design, 20% software
Requirements:
Subject(s): Flying robots, Bio-inspired locomotion
Responsible(s): Jürg Markus Germann, Adam Klaptocz

Scalability and evolvability of modular neural controls systems

Maxime Petit (MT)

Several studies in the literature indicate that modular genetic representations of artificially evolved neural control systems outperform non-modular representations in tasks which require modular solutions. It has been hypothesized that the increased evolvability of the modular mappings is caused by their variational properties, which may facilitate solving problems with a modular structure. The goal of this master project is to investigate the properties of a modular implicit genetic representation based on analog genetic encoding (AGE). To this end, a robotic control task which requires a modular control system will be studied: the control of a six-legged robot. As the problem can be altered by manually decomposing the problem structure, the required modularity of the evolved solutions can systematically be varied.The student should implement a simulation of the task and perform a series of evolutionary experiments using an existing implementation of AGE both with and without a modular mapping. Then, the results have to be analyzed with respect to network performance, topology and variational properties and compared to the results obtained with other methods reported in the literature.

Type: Master project
Period: 08.03.2010 - 27.08.2010
Section(s): MT
Type of work:
Requirements:
Subject(s): AGE +Modularity
Responsible(s): Peter Duerr, Thomas Schaffter
 

Wind-optimised descent

Benjamin Fankhauser (MT)

This project concerns the optimisation of ground distance covered by an unmanned aircraft by adjusting its gliding speed (or angle of attack) based on the knowledge of the wind profile (wind speed vs. altitude). Smart navigation strategies will be designed to take advantage of the specificities of the wind profile in order to optimise the distance flown during the descent. The student will have to: 1) Design in-flight experiment to identify the longitudinal dynamics of a small unmanned aircraft (in particular, the lift and drag coefficients). 2) Implement in matlab the dynamic model identified in (1) and develop a 2D simulation setup to test various descent strategies. By 2D, we assume level flight as well as purely linear trajectories and wind. The simulation setup should allow to use real wind profile data obtainable from meteorological institutions. 3) Propose one or more descent strategies and test them using the simulation setup of (2) and optionally the real platform.

Type: Semester project
Period: 23.02.2010 - 26.06.2010
Section(s): MT
Type of work: 30% literature and theory, 50% programming, 20% test and analysis
Requirements: Matlab, basics in flight dynamics, C programming
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
 

The evolution of sympatric speciation

Yannick Koller (IN)

150 years after the publication of the "Origin of Species" by Charles Darwin, many open questions still remain regarding how species evolve. It is unknown, for example, whether it is possible for a species to divide into two in the absence of a physical barrier. This mechanism of speciation has been termed "sympatric speciation" and now constitutes one of the hot topics in evolutionary biology.

The goal of this project will be to extend an individual-based model proposing an explanation for the evolution of sympatric speciation by natural and sexual selection. The student is expected to study the literature on the topic, to understand the existing individual-based model and to test whether the model still holds if one breaks down some of its strong assumptions. In particular, the student will explore the influence of evolving sexual signals to the existing model.

Type: Master project
Period: 22.02.2010 - 01.06.2010
Section(s): IN
Type of work:
Requirements:
Subject(s): Evolutionary biology, modeling
Responsible(s): Steffen Wischmann, Sara Mitri
 

2009


Mini-drone for atmospheric sampling

Sascha Spiegl (MT)

The goal of this project is to equip an existing autonomous mini-drone with all necessary sensors and electronics to enable meteorological measurements. This involves high-precision temperature, humidity and wind measurements. More specifically, this project involves the interfacing of predefined temperature and humidity sensors with the onboard autopilot running a microcontroller from Microchip. Optionally, a long-range radio modem may be integrated as well. The existing wind estimation algorithm may also be characterized and improved. The implemented sensors and routines will be tested and characterized both on the ground and in flight.

Type: Semester project
Period: 20.09.2010 - 31.01.2011
Section(s): EL MT
Type of work: 30% hardware, 35% software, 30% testing
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler

Design of Advanced Crash-Proof Flying Robot

Ludovic Daler (MT)

At the LIS we are working on flying robots that can autonomously avoid obstacles. We cannot however avoid any and all crashes, and are thus working on systems that can re-deploy after a crash with an object. A first prototype system called the HoverMouse was designed to passively return to a takeoff-ready position after a crash, but still requires a short runway.

The goal of this project is to design a flying platform that can take off without the need for a runway after a crash. Designs from several previous semester projects can be used as inspiration for this new platform, which should weigh below 80g. The project may include the design of new control electronics (such as a dual-motor brushless controller), mechanical design of parts for 3D printing, and flight tests. An active takeoff mechanism (such as thrust vectoring) should be integrated into this design. A first fully functional and flying prototype should be ready by the end of the project.

Type: Semester project
Period: 15.02.2010 - 01.07.2010
Section(s): EL ME MT MX
Type of work: 20% theory, 60% construction, 20% electronics/programming
Requirements:
Subject(s): Mechanics, Electronics, Aerodynamics
Responsible(s): Adam Klaptocz, Adrien Briod

Automatic Adjustment of Autopilot Parameters

Markus Zimmermann (MT)

In the frame of a PhD project at LIS, we have developed a small, fixed-wing aerial robotic platform and an electronic flight control system (autopilot). Two different control strategies are used in current research projects, both relying on PID control loops. Currently, all controller parameters have to be set up manually during a sequence of flight experiments. A simulation of the aircraft is available, but not precise enough to calculate controller parameter values with high fidelity. Although not too difficult, the process of setting up controller parameters manually could be sped up by developing a semi- or fully automatic tuning mechanism. This is especially true if considering an extended flight envelope, and not only a single operating point at which the non-linear platform behavior can be approximated by linear functions, essential for PID controllers. This Semester Project will therefore look at the possibilities of automatic, fast parameter tuning for PID controllers on aerial platforms, without any restriction to a predefined platform hardware. The first step will be a thorough investigation of mechanisms already described in literature: A first set of approaches passes by a system identification process, based on underlying aircraft models or black-box models like neural networks. A second set of approaches passes by optimization metrics, like integral squared error (ISE) functions to evaluate and find error minima without any prior system identification. A comprehensive documentation of the literature research is required. A second step consists in the selection (with well-founded motivation) and implementation of a tuning approach. Since both controller strategies described above use PID controllers, the tuning approach can be tested on both and compared. A third step will evaluate the performance of the implemented parameter tuning (e.g. operating conditions, tuning speed, applicability to an aerial platform with different hardware but the same controller electronics and PID loops). At last, a complete documentation of the performed tasks, motivations and achievements concludes the Semester Project.

Type: Semester project
Period: 22.02.2010 - 30.06.2010
Section(s): EL IN MA ME MT PH
Type of work: 40% theory, 30% software, 30% experiments
Requirements: control theory, Matlab, micro-controller C-programming, interest in flying robots
Subject(s): automatic control, unmanned air vehicles
Responsible(s): Severin Leven, Jean-Christophe Zufferey
URL: Click here

Vision Tape – Hermetic Packaging

Diego Joss (ME)

Vision Tape (VT) is a flexible, light weight, low resolution array of photoreceptors developed in LIS laboratory, inspired by the insect’s eye. The objective of the sensor is to estimate the optic flow.

This semester project focuses on the design and fabrication of protection layer for a flexible VT available in our lab. The hermetic package should be built in such a way to not decrease the VT sensing abilities. Moreover we believe that with proper design, the optical parameters of VT could be even amplified.

The goal of the project is it to find a proper material, design concept and packaging method for VT encapsulation. In the project student is expected to: (a) design and manufacture testing boards, (b) specified potential materials and investigate their properties, (c) design the protection layer, (d) apply the obtained solution to the VT device

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): CH EL ME MT MX
Type of work: 10% theory, 30% design, 50% hardware, 10% experiments
Requirements: Creative thinking, Interest in the experiments
Subject(s): materials, vision, bio-inspired robotics
Responsible(s): Michal Dobrzynski, Mirko Kovac

Hybrid algorithm for inferring gene regulatory networks

Gilles Roulet (MT)

The effective reverse engineering of Gene Regulatory Networks (GRN) is one of the great challenges of systems biology and is expected to have substantial impact on the pharmaceutical and biotech industries in the next decades. A gene network is formed by regulatory genes, which code for proteins that enhance or inhibit the expression of other regulatory and/or non-regulatory genes, thereby forming a complex web of interactions. The goal of the reverse engineering is to automatically identify such networks from experimental data.

The goal of this project is the implementation of a hybrid reverse engineering algorithm. The first stage will consist in performing a fast analysis (e.g. Z-score analysis) on the available data to obtain a first network structure, that will then be used as starting point for an evolutionary reverse engineering algorithm developed at LIS to infer a dynamical model and eventually refine the network structure. An additional stage will be implemented to take advantage of a modular decomposition of the network inferred so far to perform additional, parallel reverse engineering (using dynamical models) of the different sub-modules extracted. The benchmarks of two international reverse engineering competitions (DREAM - Dialogue for Reverse Engineering Assessments and Methods, 3rd and 4th edition) will serve as main test cases to assess the performance of this novel method.

Our long-term goal is the development of an open-source reverse engineering software for the bio-computing community.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): IN MT
Type of work: 30% research, 70% software
Requirements: C/C++, Matlab
Subject(s): systems biology, gene networks, reverse engineering
Responsible(s): Thomas Schaffter, Peter Duerr
URL: Click here

State estimation for flying robots

Winnie Wing Yee Tse (MT)

To control efficiently a flying robot that has to achieve precise tasks (landing, precise navigation, etc.), it is very important to have a good estimation of the robot's orientation and position. Different algorithms exist to fuse the different sensor inputs in order to get the best estimation. At the Laboratory of Intelligent Systems, a Kalman filter running on a microcontroller has been developped to estimate the orientation of a flying robot (roll pitch yaw), using 3-axis MEMS rate gyros, 3-axis accelerometers, and 3-axis magnetometers. Work was done toward a 6dof Kalman filter for orientation and position estimation using in addition GPS and pressure sensors (for altitude and speed).

The goal of this project is to understand and finish the work toward a 6dof Kalman filter. Real flight tests with a Swinglet and the output of a commercial 6dof module will be used to optimize the filter's parameters, compare different filter configurations and characterize the results. Finally, the candidate will implement the best filter on Matlab simulating the on-board microcontroller (dsPIC33) environment, taking into account the limited processing power. The final demonstration should present the results of the on-board position estimation algorithm during a real flight, with the estimation hopefully giving a better precision than the GPS.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): ME MT
Type of work: 30% theory, 50% software, 20% experiments
Requirements:
Subject(s): autonomous flying robots, sensor fusion
Responsible(s): Adrien Briod, Antoine Beyeler

From Simulator to Reality: Swarm Search with a Sensor Network

Marco Parisi (MT)

As part of the Swarmanoid Project http://www.swarmanoid.org/, the LIS lab has developed a novel flying robot for indoor exploration: the Eye-bot. The Eye-bot, is designed to fly in a swarm of up to 20 coordinating robots in indoor environments. Previously we have developed a swarm search algorithm that deploys the eye-bots into unknown indoor environments using a strategy inspired by Wireless Sensor Networks (WSN). WSNs form a communication network expanding the local communication abilities of each robot and can perform entirely distributed processing of local sensory information. The network of robots facilitates simple navigation and exploration without requiring global information such as GPS or maps. We have so far verified the ability of our algorithm in simulation environments. This project involves taking the ideas from the simulated behaviours and implementing them on the the real Eye-bot hardware. Importantly, this project is concerned with the sensing, processing and communication of the robots within the WSN. It does not involve the low level control of flying robots. However, if time allows, experiments involving controlling flying Eye-bots are possible. This project requires the development of the WSN technologies including: a robust communication network allowing the propagation of simple messages across a network of robots; processing the robot's sensors (such as infra-red distance sensor) including noise filtering; developing tools to make debugging easier. The student has the opportunity to learn diverse real-world robot technologies.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): EL IN MT
Type of work: 20% Theory, 80% Software
Requirements: C/C++, linux useful, microcontrollers
Subject(s): embedded programming, communication protocol, robotics
Responsible(s): Timothy Stirling, James Roberts
URL: Click here

Optic-Flow-based Anti-Drift Control

Romain Besuchet (MT)

At the laboratory of intelligent systems (LIS), we are developing indoor flying robots with the ability to hover in place (such as quadrotors). Though stable, these platforms have a tendency to drift in position while in flight since the on-board inertial sensors (accelerometers, gyroscopes ...) do not provide an outside reference to their position. To get an absolute reference, many solutions exist, such as beacon-based positioning systems or complex laser systems. But nature tells us that it is possible to solve the problem with quite simple sensors, for example the fly only uses inertial sensors and vision (optic-flow) to navigate in cluttered environments. From this observation, we want to use simple optic-flow sensors coupled with inertial sensors to achieve non-drifting indoor navigation with our robots. Typically optical mouse sensors can be used, as they are low-cost, passive (low-consumption), lightweight, and do not require high processing power. The goal of this project is to install a certain number of these optical mouse sensors on an existing coaxial flying platform, and realize a control system that allows to hold a non-drifting position and orientation. The candidate will have to study the literature to get inspiration from the existing strategies to solve questions like how many sensors are needed, how to position the sensors on the robot, what is the most suited strategy for our platform, etc. The candidate will use existing electronics and will code the control algorithm in C on a dsPIC33. The final demonstration should show the robot flying autonomously indoors, holding a position with no or very few drift.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): EL ME MT
Type of work: 25% theory, 50% software, 25% experiments
Requirements:
Subject(s): autonomous flying robots, optic-flow, control
Responsible(s): Adrien Briod, Jean-Christophe Zufferey

Viability selection

Ken Larpin (MT)

Current evolutionary algorithms are based on simple selection operators. The performance of individuals is evaluated according to a predefined fitness function and then a selection operator is applied based on individual performance. Natural evolution however, also features a mechanism called elimination which is based on the viability of the individual. The goal of this project is to implement a viability based selection mechanism in matlab, and to study its properties using a set of standard benchmark problems for multi-objective optimization algorithms. The student should carry out a series of numerical experiments and the results should be analyzed and compared to the results obtained with traditional algorithms.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): IN MT
Type of work: 50% théorie, 50% software
Requirements: C++, matlab
Subject(s): Evolutionary Algorithms, Multi-objective Optimization
Responsible(s): Peter Duerr, Thomas Schaffter

Vision Tape to Vision Matrix

Veronica Andrade (EL)

Vision Tape is a flexible, light weight, low resolution array of photoreceptors developed in LIS laboratory able to perform simple 1D vision. The objective of the sensor is to estimate the optic flow.

The aim of this project is to transform the 1D Vision Tape into 2D Vision Matrix, not losing its main advantages as flexibility, low weight and robustness.

This semester project focuses on the: (a) Vision Tape electrical circuit design, (b) pixels reading algorithm. At the end of the project student is expected to have a working 2D Vision Matrix prototype.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): EL IN MT
Type of work: 60% hardware 30% software 10% experiments
Requirements:
Subject(s): electronics, microcontrollers, sensors
Responsible(s): Michal Dobrzynski, Adam Klaptocz

The evolution of referential communication

Rolando Rodas (MT)

A number of research projects have studied the emergence of referential communication in autonomous agents. However, it is still unclear what the minimal prerequisites required for such behaviour are. The goal of this project is to use analog genetic encoding (AGE) to automatically synthesize the control systems of two autonomous agents. The agents, controlled by evolved artificial neural networks, have to solve a cooperative task. With AGE, both the topology and the parameters of the evolved networks are subject to evolution. In order to investigate the evolution of network architectures, the student should carry out a series of numerical experiments. The results of these experiments should then be analyzed with respect to the evolution of network topology and behaviour of the agents. As a reference, the results of a series of experiments with a hand-designed network topology can be used.

Type: Semester project
Period: 22.02.2010 - 04.06.2010
Section(s): IN MT
Type of work: 50% théorie, 50% software
Requirements: C++, matlab
Subject(s): Evolutionary Algorithms, Artificial Neural Networks, Communication
Responsible(s): Peter Duerr, Kuniaki Noda

Crash Detection of a Flying Robot

Matthieu Bippus (MT)

At the LIS we are working on flying robots that can autonomously avoid obstacles. We cannot however avoid any and all crashes, and thus are interested in characterizing crashes to be able to autonomously react and recover from them.

The goal of this project is to devise a system to detect crashes (both small bumps and major crashes) and the airborne state of a flying robot (ie. if it is on the ground or in the air). The project will begin with an analysis of possible sensors such as accelerometers or contact potentiometers, as well as signals inherent to the platform, such as the motor power or current. A sensor circuit board will then be developed and tested. Finally the sensors will be evaluated in their capacity to detect collisions and their severity.

Type: Semester project
Period: 01.09.2009 - 30.03.2010
Section(s): EL IN MT
Type of work: 30% electronics, 30% programming, 40% signal processing
Requirements:
Subject(s): signal processing
Responsible(s): Adam Klaptocz, Adrien Briod
URL: Click here

Design of Advanced Crash-Proof Flying Robot

Christian Heimlich (MT)

At the LIS we are working on flying robots that can autonomously avoid obstacles. We cannot however avoid any and all crashes, and are thus working on systems that can re-deploy after a crash with an object. A first prototype system called the HoverMouse was designed to passively return to a takeoff-ready position after a crash, but still requires a short runway.

The goal of this project is to design a flying platform that can take off without the need for a runway after a crash. Designs from several previous semester projects can be used as inspiration for this new platform, which should weigh below 80g. The project may include the design of new control electronics (such as a dual-motor brushless controller), mechanical design of parts for 3D printing, and flight tests. An active takeoff mechanism (such as thrust vectoring) should be integrated into this design. A first fully functional and flying prototype should be ready by the end of the project.

Type: Semester project
Period: 01.09.2009 - 30.03.2010
Section(s): EL ME MT MX
Type of work: 20% theory, 60% construction, 20% electronics/programming
Requirements:
Subject(s): Mechanics, Electronics, Aerodynamics
Responsible(s): Adam Klaptocz, Mirko Kovac

Vision based take-off and landing for unmanned aircraft

Guillaume Monnard (MT)

At LIS we have developed a series of vision-based robots capable of flying outdoors without human intervention. Based on a vision system, these robots are able to autonomously avoid obstacles and safely fly in cluttered environments. The aim of this project is to add vision-aided take-off and landing. This project will involve developing and testing algorithms for target-oriented navigation. The algorithms must be able to visually detect a target and control the flight with respect to its location in order to stabilize the take-off phase or enhance precision in the landing phase. The vision processing and control stabilization should run in real time on an embedded system. During this project, the student will become familiar with vision processing on embedded Linux and understand the requirements of real time image processing in a highly dynamic environment. He will summarize the state of the art in video tracking and feature recognition algorithms before developing an algorithm able to track a target independent of its scale and orientation. The system will be tested in order to characterize it in both open-loop and close-loop.

Type: Master project
Period: 12.09.2009 - 13.03.2010
Section(s): MT
Type of work: 40% theory, 40% software, 20% characterization
Requirements:
Subject(s): Aerial robotics
Responsible(s): Yannick Weibel, Jean-Christophe Zufferey
 

Time-of-flight omni-directional distance scanner for flying robots

Klaus Doth (MT)

In order for a miniature aerial vehicle (MAV) to navigate and fly within a cluttered indoor environment it is necessary that the MAV is able to detect objects and avoid collisions. The goal of this project is to develop a new omni-directional scanner that uses time-of-flight sensing capable of measuring the distance to objects with sub centimeter precision. The output of the sensor would be a full omni-directional depth map of the surrounding environment. Existing scanners are either too heavy, or have very limited specifications regarding acquisition speed, resolution and range., The project involves: 1. designing the circuitry based on a time-to-digital converter 2. selecting the correct light weight optics 3. selecting the low power light source for 5m range 4. designing light-weight scanner mechanics 5. basic sensor characterization The prototype will be fabricated so that it can interface to the existing eye-bot hovering MAV.

Type: Master project
Period: 07.09.2009 - 07.03.2010
Section(s): MT
Type of work: 10% theory, 10% software, 30% experimentation, 50% hardware
Requirements: Mechanical design, electronics
Subject(s): Sensing, mechanics, electronics, MAVs
Responsible(s): James Roberts, Jean-Christophe Zufferey
URL: Click here

Optimisation of mouse-chip-based vision system

Michael Minnig (MT)

At the Laboratory of Intelligent Systems (LIS), we developed a smart control strategy named optiPilot enabling unmanned aircraft to fly at low altitude while avoiding collisions with the terrain and other obstacles. The current implementation uses a set of seven optical computer mouse sensors to sense optic flow in various divergent viewing directions. The goal of this project is to redesign this compound vision system in order to increase its sensitivity, reduce its weight while enhancing its robustness. The student will start by evaluating various optic-flow sensors available on the market in order to select the best suited chip. He will design the electronics, optics and packaging of the new vision system, which will then be constructed and tested in flight on an existing flying platform.

Type: Semester project
Period: 14.09.2009 - 15.02.2010
Section(s): MT
Type of work: 80% hardware, 20% software
Requirements:
Subject(s): Aerial robotics
Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
URL: Click here
 

Pheromone-based swarming for networks with multiple dynamic users

Pascal Bach (MT)

At the LIS we aim at designing a swarm of Micro Air Vehicles (SMAVs) capable of autonomously establishing emergency wireless networks (SMAVNETs) between multiple ground-users in a disaster area. Up until now, algorithms designed in simulation have been adapted to scenarios where only two users on ground could communicate via the SMAVNET. Instead, in this project the student will have to extend in simulation the pheromone-based algorithm (Hauert et al. 2008) to scenarios with multiple users on the ground or mobile users. The student should develop and compare several strategies to connect at least 5 users together and determine what these strategies requires in terms of resources (number of robots, deployment time). In addition, there should be a clear model to describe how the controllers should be parametrized for different environments. Finally, these strategies should be extended to 1-5 mobile users.

Type: Semester project
Period: 12.09.2009 - 15.02.2010
Section(s): EL IN MA MT PH SC SV
Type of work: 60% software, 20% theory, 20% experiments
Requirements: C++, linux an advantage
Subject(s): Swarming, Micro air vehicles, evolution, ant-based swarming
Responsible(s): Sabine Hauert, Severin Leven
URL: Click here

Chain-based swarming for networks with multiple dynamic users

Raphaël Maegli (MT)

At the LIS we aim at designing a swarm of Micro Air Vehicles (SMAVs) capable of autonomously establishing emergency wireless networks (SMAVNETs) between multiple ground-users in a disaster area. Up until now, algorithms designed in simulation have been adapted to scenarios where only two users on ground could communicate via the SMAVNET. Instead, in this project the student will have to extend in simulation the chain-based algorithm (Hauert et al. 2008) to more complex scenarios. The student should develop and compare several strategies to maintain a communication link between a mobile user and a base station. The best strategy should then be extended to situations were multiple users (at least 5) must be found in the environment. To make the controller practical for use in real-world applications, a clear model of the system should be developed and used to automatically parametrize the robot controllers. If time permits, other scenarios (such as 360° search) can be considered.

Type: Semester project
Period: 13.09.2009 - 01.02.2010
Section(s): IN
Type of work:
Requirements:
Subject(s): MAVs, swarming, communication networks
Responsible(s): Sabine Hauert, Severin Leven

Wing design for the self deploying microglider

Mohammadmehdi Shafizadeh Khoolenjani (ME)

Jumping and gliding is a promising locomotion method for small robots to move in rough terrain. However, very little is known about how to design optimal wings for small gliding systems and how to choose the shape and size that gives the best flight performance. This semester thesis consists of reviewing the state of the art on wing design in the MAV and hobbyist community, compare the different approaches and propose the most suitable one for the indoor flying robots developed at LIS. The final prototype of the project is to use the best options and fabricate the foldable wings for the final version of the self deployable Microglider.

Type: Semester project
Period: 14.09.2009 - 15.01.2010
Section(s): MT
Type of work:
Requirements:
Subject(s): hardware implementation
Responsible(s): Mirko Kovac, Michal Dobrzynski
 

Sensor and control set for the self deploying microglider

Manon Picard (MT), Mohammadmehdi Shafizadeh Khoolenjani ()

Jumping and gliding is one very promising way of deploying nodes of sensor networks. This semester project consists of developing a electronics set to sense, monitor and control a small jumping and gliding robot. Based on an existing PCB (with integrated dsPic and communication unit), the student will first propose sensors that can be integrated on this board (e.g. temperature/humidity sensors, microphones or cameras) and implement them on a novel PCB. In a next step, the student will extend an existing version of a monitoring software to be able to receive the sensor information on a laptop. The second part of the project includes the integration and packaging on the self deploying microglider to balance the weight distribution and allow it perform stable flight. Finally, the student and will demonstrate the working prototype integrated on the platform. This project also comprises the choice of teh wing size and shape based on aerodynamical theory and experiment in order to be able to balance the flying system.

Type: Semester project
Period: 14.09.2009 - 08.01.2010
Section(s): MT
Type of work:
Requirements:
Subject(s): PCB design, microelectronics, wing design
Responsible(s): Mirko Kovac, Adam Klaptocz
URL: Click here
 

Wing Folding Implementation for the Self Deploying Microglider

Wassim Hraiz (MT), Oriol Fauria Torrent ()

A new biomimetic platform for flying miniature robots capable of jumping and gliding is currently in development at LIS. This system will possess wings that are foldable and a jumping mechanism in order to propel it into the air, open the wings and glide The goal of this master thesis is to implement a hardware design for the wing folding mechanism in a Computer Aided Design (CAD) program and fabricate a working prototype. The work includes the following work packages:

-
  • Integrate the existing wing folding frame with the existing jumping mechanism in CAD.
  • -
  • Fabricate the wing folding frame and wings using the 3D printer (several iterations until the design is mature).
  • -
  • Fabricate the wing folding frame in PCB, assemble the wings and demonstrate the working remote controlled prototype.
  • -
  • Translate the previous two wing folding designs in SolidWorks and compare them to the new one.
  • -
  • Using an existing catapult and the high speed camera, determine the aerodynamical friction coefficient for the three given wing folding designs and the energy that is needed to propel it up to a given height.
  • -
  • Document and present the work.
  • Type: Master project
    Period: 14.09.2009 - 08.01.2010
    Section(s): MT
    Type of work: 30% theorie, 30% software, 40% hardware
    Requirements:
    Subject(s): hardware implementation
    Responsible(s): Mirko Kovac, Michal Dobrzynski
    URL: Click here
     

    A minimalist jumping and climbing robot

    Manuel Schlegel (MT)

    One of the specialties of the Nanorobotics Lab at Carnegie Mellon University in Pittsburgh is their gecko inspired wall climbing robots. On the other hand, the Laboratory of Intelligent Systems of the Ecole Polytechnique Fédérale de Lausanne has developed a minimalist jumping robot with selfrecovery capabilities. The aim of the Master project is the combination of the strength of both techniques in order to build a jumping robot which is able to jump towards an obstacle, attach to it and overcome the obstacle by climbing. As Manuel Schlegel already worked on the minimalist jumper of the LIS within a semester project and knows well its strengths and weaknesses, he will use this existing mechanism for the jumping part of the new hybrid locomotion device. A minimalist climbing part, which can be fixated to this existing jumping robot, as well as an adapted fixation system have to be developed within this master project. The existing jumping robot may have to be slightly modified in order to adapt its geometry to the climbing requirements and increase its payload capabilities. In a first step magnetic feet on a ferromagnetic surface could be considered to implement the climbing mechanism and its integration with the jumping robot. When this system works, the replacement of the magnetic feet by gecko-feet can be envisaged. Two points of most importance during the development of the climbing part are its weight and size so that the robot still achieves a jumping height of several times its own size. It should always be kept in mind that an additional locomotion capability - gliding - is envisaged to be added to the robot in a later project. The jumping robot with and spare parts are sent to the Nanorobotics Lab by the LIS. As the jumping robot, which represents the propulsion unit of this robot, is a robust functional system it, should be modified as few as possible.

    Type: Master project
    Period: 14.09.2009 - 08.01.2010
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): hardware implementation
    Responsible(s): Mirko Kovac, Matthew Woodward (CMU)
     

    Smart Materials for Drop Foot Soft Orthotics

    Philippe Bérard (MT)

    One class of soft active materials relies on the response of active fluids (e.g. electrorheological and magnetorheological fluids) to create stiffness changes and actuation. The method used to contain these fluids is crucial for designing the active response of the system. Techniques for fabricating and filling micro fluidic channels in soft materials have been established, but typically the thickness of the supporting structure is much larger than then the characteristic size of the channels. To create fluid filled material systems where the mechanical properties are dominated by the viscosity of the fluid, it is necessary to fabricate thin film materials which contain organized micro fluidic channels, are hyperporous, or can be swollen with an active fluid. Additionally, the electric and magnetic fields must be applied locally and the components required to create these fields must not significantly restrict the motion or dominate the stiffness of the system. This project is divided in three different steps and sub goals that lead to the overall goal ‐ the development of a smart material that has the capabilities of changing its proprieties in order to be employed in an ankle soft orthotic for drop foot patients. (These steps might be modified as each ongoing step depends strongly on the success of the previous steps).Step 1: Quantification Describe the design space of these systems (i.e. what stiffness changes can we expect? what fields do we need?) based on small test patterns. Step 2: Integration Develop fabrication techniques and chose appropriate materials for prototyping Step 3: Orthotic prototype A first ankle orthotic is developed under consideration of the specifications given by the project leaders of the Harvard Wyss Institute.

    Type: Master project
    Period: 14.09.2009 - 08.01.2010
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): hardware implementation
    Responsible(s): Mirko Kovac, Rob Wood (Harvard University)
     

    The Evolution of Mutualistic Cooperation

    Manuel Wüthrich (MT)

    Ecosystems are based on interactions between many species. Cooperative interactions between species, i.e., "mutualisms" are very common, yet difficult to explain, since they are susceptible to cheating individuals that reap the benefits of the interaction without paying the cost. The dynamics determining the stability of mutualisms and their immunity towards cheaters are not yet well understood.

    Using an evolutionary model of software agents, this project will explore the influence of two parameters on the dynamics of mutualistic interactions: (i) relatedness between individuals of each of the two species and (ii) the within-species distribution of the benefits gained from the mutualistic interactions. Theoretical work has generated concrete predictions with respect to these two factors, but these have been difficult to test using living systems. This project will thus represent the first test of the theory using an evolutionary system. The student is expected to study the literature on mutualism from a theoretical and empirical perspective, to build the software model and to set up and run the necessary experiments to explore the influence of the two above-mentioned factors on the stability of mutualistic interactions.

    Type: Semester project
    Period: 14.09.2009 - 18.12.2009
    Section(s): IN MT SV
    Type of work: 30% theory, 30% software, 40% data analysis
    Requirements:
    Subject(s): Evolutionary biology, agent-based simulation
    Responsible(s): Sara Mitri, Steffen Wischmann
     

    An evolutionary study of communication in bacteria

    Yannick Koller (IN)

    Evolutionary models represent a powerful tool to answer many questions in evolutionary biology. In this project, the evolution of communication between bacteria - generally referred to as "quroum sensing" - will be explored by evolving the behavior of simulated bacteria in an individual-based model.

    The study of quorum sensing in bacteria generally involves the use of mathematical models. However, testing the validity of these models using real bacteria can be problematic. Instead, in this project, we propose to use a model where we evolve simulated bacteria to test the validity of, and possibly modify, one such mathematical model.

    The student is expected to read some literature on quorum sensing and work both with the mathematical model as well as the simulation model to determine the effect of a number of parameters on the evolution of quorum sensing.

    Type: Semester project
    Period: 14.09.2009 - 18.12.2009
    Section(s): IN SC SV
    Type of work: 20% research, 50% software, 30% mathematical analysis, 30% data analysis
    Requirements: Programming C/C++, basic mathematics
    Subject(s): Biological modelling, evolutionary biology, evolutionary robotics
    Responsible(s): Sara Mitri, Steffen Wischmann
    URL: Click here

    Drift-Control and Positioning of a Quadrotor Using an Infra-red Relative Positioning Sensor

    Sebastien Mamin (MT)

    Hovering platforms suffer from unknown ergo-motions and drift due to sensor noise and platform dynamics. Conventional methods to control this drift depend on GPS receivers, computer vision, or the accurate positioning via an external system such as infra-red tracking cameras in the VICON system., GPS is not usable indoors due to weak signal strength and multipath reflections. Computer vision requires good illumination and complex real-time onboard vision processing. Systems such as the VICON system are prohibitively expensive and require to be pre-installed. The Laboratory of Intelligent Systems has developed a low cost solution that provides relative positioning to small flying robots. This project entails the complete characterisation of the sensor and improvement of the firmware. The improved sensor will then be used to control the drift of a quadrotor, and finally, to position the quadrotor within a small room.

    Type: Semester project
    Period: 18.02.2009 - 30.06.2009
    Section(s): EL IN MA MT
    Type of work: 20% Theorie, 10% Hardware, 50% Software, 20% Analysis
    Requirements:
    Subject(s): Programming, sensor characterisation, MAV control
    Responsible(s): Timothy Stirling, James Roberts

    Hole detection and positioning using a hovering MAV

    Iván García García (IN), Antonio Morales García ()

    In the indoor mission of the EMAV09 there are several key difficulties that need to be accomplished in order to complete the mission. One of the most challenging tasks is to have the ability to detect a window opening and then to fly through this opening. This task being the main goal of this project. Two students will be selected to systematically break down the problem and devise a solution. The goals being devised below: 1. Research the state of the art in hole detection methodology. 2. Initial sensor testing and selection. 3. Algorithm testing and selection. 4. System evaluation and characterisation. 5. Implementation on a hovering MAV platform.

    Type: Semester project
    Period: 16.02.2009 - 15.06.2009
    Section(s): EL MT
    Type of work: 30% theory, 40% software, 30% hardware
    Requirements: C++code, electronics
    Subject(s): MAV, electronics, algorithms
    Responsible(s): James Roberts, Timothy Stirling
    URL: Click here

    Active Take-off Mechanism for a Flying Robot

    Corentin Ryser (MT)

    At the LIS we are working on flying robots that can autonomously avoid obstacles., We cannot however avoid any and all crashes, and are thus working on systems that can re-deploy after a crash with an object., A first prototype system called the HoverMouse was designed to passively return to a takeoff-ready position after a crash, but still requires a short runway.

    The goal of this project is to design a flying platform that can take off without the need for a runway after a crash., The system may be passive, based on gravity, or active, based on some simple and light actuators that up-right the platform., The student is free to adapt the current platform to perform to these requirements by adapting its structure, or to design an add-on module that actively positions the platform vertically after a crash, ready for take-off.

    Type: Semester project
    Period: 15.02.2009 - 10.06.2009
    Section(s): EL IN MT
    Type of work: 20% theory, 60% construction, 20% electronics/programming
    Requirements:
    Subject(s): Mechanics
    Responsible(s): Adam Klaptocz, Mirko Kovac
    URL: Click here

    Hybrid Gene Regulatory Network Reverse Engineering Method

    Zeineb Benabdallah (IN)

    The effective reverse engineering of Gene Regulatory Networks (GRN) is one of the great challenges of systems biology and is expected to have substantial impact on the pharmaceutical and biotech industries in the next decades. A gene network is formed by regulatory genes, which code for proteins that enhance or inhibit the expression of other regulatory and/or non-regulatory genes, thereby forming a complex web of interactions. The goal of reverse engineering is to automatically identify such a network from experimental data.

    The winning method of DREAM3 is very effective in predicting network structures, but it does not infer a predictive model of the network. Thus, we can use the inferred network structure by the winning method as starting point for our method to infer a dynamical model and refine the network structure. We hope that this hybrid approach increases both the speed and the accuracy of the reconstruction as compared to directly inferring a dynamical model from scratch.

    In this project, the student is expected to: 1) assess the performance of the DREAM3-winning method and our own GRN reverse engineering algorithm on both DREAM2 and DREAM3 In-Silico Challenges, 2) augment the current implementation or our algorithm to confer it the capacity to integrate into its process a network structure previously inferred by the DREAM3-winning method and 3) conduct a comparative analysis on the performance of our algorithm used with and without this novel extension.

    Type: Semester project
    Period: 16.02.2009 - 29.05.2009
    Section(s): IN
    Type of work: 60% software, 40% research
    Requirements: C++
    Subject(s): Systems Biology, Gene Networks, Reverse Engineering
    Responsible(s): Thomas Schaffter, Daniel Marbach

    2008


    Long-range airborne video streaming

    Yannick Gasser (MT)

    At LIS, we developed an autonomous micro-UAV (unmanned aerial vehicle) capable of acquiring and processing videos. The goal of this project is to implement a wireless TCP/IP link to digitally stream this imagery to a ground station. Two technologies will be considered: 3G mobile telephony and 5.8GHz WiFi. The project includes: * interfacing both technologies on the existing Linux board and camera module; * setting up the software for digital video streaming using existing libraries; * thorough in-flight testing, characterisation (range, throughput, lag, etc.) and comparison of both approaches.

    Type: Master project
    Period: 12.09.2009 - 18.03.2010
    Section(s): EL IN MT SC
    Type of work: 20% hardware, 40% software, 40% tests
    Requirements: Linux and networking configuration
    Subject(s): GSM, GPRS, camera, GPS, wifi
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler

    State estimation for flying robots

    Kevin Jamolli (MT), Daniel Vogt ()

    This project is about developing and characterizing a state estimator using a Kalman filter for sensor fusion for a fixed-wing UAV. The filter should rely on an existing hardware (see http://lis.epfl.ch/smavs) which includes the following sensors: 3-axis MEMS rate gyros and 3-axis accelerometers, barometer and airspeed sensors, 3-axis magnetometers and a GPS. The full filter should be implemented in a cascaded configuration, and should provide an accurate estimation of the orientation and heading in a first stage, position and velocity in a second stage. It will be coded in C, and will have to run in real-time on the embedded auto-pilot. The limited computational power of the available microcontroller (dsPIC33) will have to be taken into account during the design. The obtained results will be compared to the output of a commercially available module providing the same information (Xsens MTi-G). The project can be carried out by two students, with the following task distribution : At first, both students will work on separate implementations, based on an existing filter. The first student will include the 3-axis magnetometer to the filter while the second student will include the GPS speed and heading. Both students will then compare the results of their different approaches. They will then work together on the second part of the filter, using the full position information from the GPS for position estimation.

    Type: Semester project
    Period: 14.09.2009 - 16.02.2010
    Section(s): IN MA ME MT
    Type of work: 30% theory, 50% software, 20% testing
    Requirements: C programming, microcontrollers
    Subject(s): sensors, state estimation, Kalman filter
    Responsible(s): Adrien Briod, Jean-Christophe Zufferey

    Evolving heterogeneous teams of robots

    Adrian Arfire (MT)

    The use of groups of robots can be highly beneficial in solving distributed tasks. This is particularly the case if the task is composed of multiple sub-tasks that need to be accomplished in parallel. Such tasks require heterogeneous robot teams. Moreover, evolutionary algorithms have proven to be useful in designing controllers for groups of robots. A number of algorithms have been proposed for evolving groups of heterogeneous robots. A common problem with many of these algorithms is that they are not able to adapt to changes in the environment that require a dynamic reallocation of individuals to the different sub-tasks. In this project, the student is required to develop an application-oriented robotic task, to design and implement an algorithm that is flexible to such environmental changes, and finally to conduct a systematic comparison of the performance of different algorithms on the same benchmark task.

    Type: Master project
    Period: 14.09.2009 - 18.12.2009
    Section(s): IN SC SV
    Type of work: 20% research, 50% software, 30% data analysis
    Requirements: Strong C/C++, interest in biology
    Subject(s): Biological modelling, evolutionary biology, evolutionary algorithm
    Responsible(s): Sara Mitri, Steffen Wischmann
    URL: Click here

    GSM-based UAV data link

    Nathanaël Mägli (MT)

    This project aims at developing the hardware and firmware to provide a small unmanned aerial vehicle (UAV) with long-range data communication capabilities. This will be achieved by relying on the mobile phone network (GSM). Therefore the project will start by a review of existing GSM module and continue with its interfacing to the existing embedded electronics. Some software modification will be required at the level of the ground station as well. Finally, a thorough characterization of the robustness and data throughput will be carried out both in static configurations and in flight.

    Type: Semester project
    Period: 16.02.2009 - 30.06.2009
    Section(s): EL IN MT SC
    Type of work: 40% hardware, 40% software, 20% tests
    Requirements: C programming, microcontrollers
    Subject(s): GSM, GPRS, camera, GPS
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler

    Flying air quality sensor

    Sylvain Candolfi (EL)

    At LIS, we developed a small autonomous airplane weighing less than 500g and capable of flying in various kinds of outdoor environments including cities and mountains, both and medium and low altitude. This airplane has an additional payload capability of approximately 100g. The goal of this project is to select and integrate onboard the airplane a sensor for air quality monitoring (e.g. ozone sensor). Once the system is operational, it should be deployed according to various scenarios in order to analyze the added value of such a flying sensor with respect to other more conventional systems (ground stations, full-scale airplanes, LIDARs, etc.). This applicative part of the project will be co-supervised by EPFL experts in air quality monitoring (EFLUM).

    Type: Semester project
    Period: 16.02.2009 - 30.06.2009
    Section(s): EL MT
    Type of work: 30% hardware, 30% software, 40% experiments
    Requirements:
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler

    Cost and the evolution of communication

    Arnaud Jutzeler (IN)

    It is difficult to conceive of any behavior in humans -- or animals -- that is cost-free. The same goes for communication, which can be costly because of the physiological investment in producing the signal, or because of eavesdropping from potential enemies.

    The aim of this project is to understand how the cost of a signal can influence its evolutionary trajectory. To explore this question, the student will develop a minimalistic individual-based model where individuals can communicate with each other. Communication will be evolved using a genetic algorithm, and the effects of different signaling costs observed, analyzed and interepreted.

    Type: Semester project
    Period: 16.02.2009 - 15.06.2009
    Section(s): IN SC SV
    Type of work: 20% research, 50% software, 30% data analysis
    Requirements: Programming C/C++
    Subject(s): Biological modelling
    Responsible(s): Steffen Wischmann, Sara Mitri
    URL: Click here

    Programming of Flying Robots through Wireless Communication Link

    Yannick Gasser (MT)

    At LIS we work on many flying robotic platforms that communicate with a ground station through a wireless link. Our smallest robots use a new ultra-light wireless chip, the Nordic nRF24L01, along with a custom-designed low-level communication protocol to achieve transfer speeds higher than Bluetooth or Zigbee. For the moment however the robots themselves are still programmed using a cable., In addition the communication protocol can only communicate between two devices.

    The goal of this project is thus two-fold:, to design a bootloader that can re-program a microcontroller (on-boar a robot) using only this wireless link, and to improve the protocol between wireless chips. The bootloader should be able to program a robot wirelessly using the existing nRF24 dongle, and will be tested on a robotic platform such as the MC3., As for the enhanced protocol, it should be able to allow communication between at least 3 devices without requiring a master., Optionally, a protocol able to deal with an undefined number of devices, creating ad-hoc connections when new devices arrive, will be developed.

    The student will learn about embedded programming (C and assembler), wireless protocols and how to deal with problems intrinsic to wireless communication, such as packet loss and handshaking.

    Type: Semester project
    Period: 15.02.2009 - 10.06.2009
    Section(s): EL IN MT
    Type of work: 30% theory, 60% software, 10% testing
    Requirements: C programming
    Subject(s): Wireless communication, embedded programming
    Responsible(s): Adam Klaptocz, James Roberts
    URL: Click here

    Biologically inspired soaring for increased endurance of, Unmanned Aerial Vehicles

    Saurabh Indra (IN)

    Birds of prey and migratory birds actively seek out and utilize the energy of atmospheric thermals to gain height in between powered phases. Such behavior could provide significantly increased endurance for unmanned aerial vehicles. This project will investigate this behavior and adapt an electric glider for this purpose. Preparation for the project: Some of the sensors and electronics to be used for the project will be developed at the Laboratory of Intelligent Systems, EPFL. In particular an autopilot previously developed here will be adapted. Steps of the project: 1. Conduct literature survey of the field: Study of avian behavior, modeling of atmospheric thermals. Identify appropriate control techniques for best utilization of thermals. 2. Modeling and simulation of the thermal model coupled with an aircraft model with autopilot. Verify behavior of control algorithm(s) in simulation. 3. Is it possible to utilize commercial off the shelf thermal sensors to fabricate a sensing system which would provide an edge over just accelerometer data? If possible construct this sensor. 4. Adapt an electric R/C glider for the purpose of this project. Add the autopilot and an embedded computer. Implement control algorithms and install required sensors. 5. Verify studied/developed control algorithms with the adapted test bed. Make a quantitative study of the improvement in endurance.

    Type: Master project
    Period: 15.09.2008 - 30.03.2009
    Section(s): IN
    Type of work: 40% theory and simulation, 30% implementation in hardware, 30% experiments
    Requirements: flight control, sensors, simulation (SIMULINK), interest in bio-inspiration
    Subject(s): flight control, flight and environment simulation, electronics, flight field tests
    Responsible(s): Severin Leven, Jean-Christophe Zufferey
     

    Designing and building a rolling legged tensegrity robot

    Simon Fivat (MT)

    Tensegrities are structures made out of struts and cables arranged such that the structure retains a 3D shape without the struts touching each other. They have an excellent strength to weight ratio and have been used frequently in architectural design. Others have used tensegrity principles to explain structural properties of biological cells, yet their potential for robotics applications remains largely unexplored. Their advantages over many other legged robots are their resilience to deformation, their ability to change shape and their high strength-to-weight ratio. At the Computational Synthesis Laboratory at Cornell (CCSL) some crawling tensegrity robots have already been built. A new tensegrity robot capable of rolling will be built in this project. Different possible shapes of the robot will be evaluated and the best one will be further analyzed. A real tensegrity robot based on that selected shape will then be designed and built. Finally, controllers for different gaits (crawling and rolling and eventually climbing) will be evolved and applied to the real robot.

    Type: Master project
    Period: 15.09.2008 - 13.03.2009
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): tansegrity structures
    Responsible(s): Mirko Kovac, Hod Lipson
    URL: Click here
     

    Design and Assessment of a Robot Curriculum based on the e-Puck Robot and Webots

    Nicolas Heiniger (IN)

    The goal of this Master project is to complete, test and disseminate a robot curriculum currently being developed by Cyberbotics., This curriculum is a document that contains exercises aimed at teaching robotics from the beginner to the expert level, to be used mostly by teachers of different levels but also by robotics clubs or hobbyists., Exercises are all implemented using the e-Puck educational robot and the Webots software.

    The beginning of the project will involve the completion of the Advanced section of the curriculum., Several new exercises should be designed and tested, based partially on exercises already developed at EPFL., A cognitive robotics benchmarks section should also be expanded with more benchmarks and completed., In parallel with the design of the curriculum, sections of the curriculum will be tested extensively through several in-class experiments with students at the grade- or high-school level, and then revised according to feedback received from the students and their teachers., The final revision (version 1.0) of the document will then be published on the web under the Common Creative license, both in the form of a clean PDF version and an interactive wiki version designed by the student.

    Type: Master project
    Period: 15.09.2008 - 01.03.2009
    Section(s): IN
    Type of work: 30% Theory, 50% Testing, 20% Software
    Requirements:
    Subject(s): low-level programming, education
    Responsible(s): Adam Klaptocz, Michel Olivier
    URL: Click here
     

    Radio Leashing of a Micro Air Vehicle (MAV)

    Cristina Vinasvinuales (IN)

    In the frame of the SMAVs project at LIS, we are currently developing a flying robotic platform based on an 80cm wingspan, flying-wing type airplane. It is designed to be less complex and less expensive compared to existing platforms on the market, but still provides all necessary flight stabilization, sensing and processing to be suited not only for the SMAVs project, but also for other projects in research and education. Recently, platform and flight control development has reached the point of being operational. In order to be interesting for a wide range of people, the platform must be easy to use. The goal of this master project is to take advantage of a wireless link (via ZigBee modem) between the plane and a ground station and indicators like RSSI (Received Signal Strength Indicator) to control the plane's flight in such a way that it stays in a safe range from the ground station. Additionally, a supervision of other flight parameters like altitude, speed, turn rate is planned. The project comprises the search and selection of flight control strategies (on the navigation level) based on wireless communication and sensor measurements, the adpatation of a graphical user interface and a thorough series of experiments to prove the chosen implementation.

    Type: Master project
    Period: 15.09.2008 - 31.01.2009
    Section(s): EL IN MA ME MT PH SC
    Type of work: 40% software, 30% theory, 30% experiments
    Requirements: C/C++ programming, RC piloting skills are an advantage
    Subject(s): Wireless communication, ZigBee, autonomous control, flight envelope supervision
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here

    Sensor Network Deployment for MAVS Indoors

    Vasili Massaras (MT)

    As part of the European funded Swarmanoid (http://www.swarmanoid.org/) project, the LIS lab is developing a unique flying robotics platform for indoor exploration: the Eye-bot. The Eye-bot, helicopter is designed to perform swarm search and exploration in groups of 20 coordinating robots in indoor environments consisting of doorways, corridors and rooms., In indoor environments GPS positioning is unreliable and flying robots have limited odometery sensing making classic robot localisation prohibitive. Instead, we have developed a novel localisation-free search method that deploys our flying robots into a distributed sensor network which can provide easy navigation. This project involves expanding upon this prior research developing a more robust deployment method, and building upon our gained knowledge. You are free to apply a variety of different techniques or develop your own novel algorithm. Work will be done in simulation, with the potential for limited testing on actual hardware.

    Type: Semester project
    Period: 11.09.2008 - 31.01.2009
    Section(s): IN MA MT PH
    Type of work: 25% theory, 65% software, 10% analysis
    Requirements: C++ Programming
    Subject(s): Sensor networks, multi-robotic search, flying robots, MAVs
    Responsible(s): Timothy Stirling, Sabine Hauert
    URL: Click here

    Data fusion of lidar and digital camera data

    Simon Rutishauser (MT)

    The goal of this project is to efficiently fuse the data from lidar, scans with panoramic digital camera images taken using a standard compact digital camera mounted on the K10 mobile robot, in order to augment the 3D depth map from the lidar scan with texture and color information.
    This project will involve computer vision theory, Cpp development (using the NASA Vision Workbench on Linux) and testing on data from K10 mobile robot test runs.
    This project will be carried out at NASA Ames under the supervision of Terry Fong and Matthew Deans.

    Type: Master project
    Period: 15.09.2008 - 31.01.2009
    Section(s): IN MT
    Type of work:
    Requirements:
    Subject(s): Mobile robot control
    Responsible(s): Jean-Christophe Zufferey, Terry Fong
     

    Micro-GPS for quadrotor waypoint navigation

    Nicolas Wicht (MT)

    Here at the laboratory of intelligent systems (LIS) a new and exciting type of hovering platform has been developed called the 'OB1-QuadroB'. This project would involve designing a ultra small GPS module that can be used on this testbed quadrotor platform. The goals are defined as: 1. Design a small printed circuit board to carry a tiny GPS receiver, GPS antenna and dsPIC microcontroller. 2. Write a simple waypoint controller to run on the microcontroller. 3. Interface the system with an existing 2D mapping program, that will monitor the trajectory. 4. Test the system on the quadrotor/flying wing.

    Type: Semester project
    Period: 15.09.2008 - 31.01.2009
    Section(s): EL IN MT
    Type of work: 30% theory, 30% software, 40% hardware
    Requirements: Basic electronics & programming skills
    Subject(s): Miniature Aerial Vehicles, GPS, Waypoint Navigation
    Responsible(s): Severin Leven, James Roberts
    URL: Click here

    Sensor Optimization for a Micro Air Vehicle (MAV)

    Adrian Arfire (MT)

    In the frame of the SMAVs project at LIS, we are currently developing a flying robotic platform based on an 80cm wingspan, flying-wing type airplane. It is designed to be less complex and less expensive compared to existing platforms on the market, but still provides all necessary flight stabilization, sensing and processing to be suited not only for the SMAVs project, but also for other projects in research and education. Recently, platform and flight control development has reached the point of being operational. Still, improvements can be brought to the sensing technologies currently in use: the 2 pressure sensors and 2 gyroscopes which constitute the core sensors of the platform are subject to drift, which can potentially be a problem for long flight times. The goal of this semester project is to find out ways to compensate for drift, be it with additional sensors, temperature compensation etc., to implement the selected solutions and to document the improvement.

    Type: Semester project
    Period: 15.09.2008 - 16.01.2009
    Section(s): EL IN MA ME MT PH
    Type of work: 20% hardware, 20% theory, 30% software, 30% experiments
    Requirements: advantageous are: sensors, electronics, flight control, motivation, RC piloting skills, C programming
    Subject(s): sensors, electronics, programing
    Responsible(s): Severin Leven, James Roberts
    URL: Click here

    Monocular-Visual Depth Estimation in Indoor Environments

    Lorenz Küchler (MT)

    As part of the European funded Swarmanoid (http://www.swarmanoid.org/) project, the LIS lab is developing a unique flying robotics platform for indoor exploration: the Eye-bot. The Eye-bot, helicopter is designed to fly in a swarm of up to 20 coordinating robots in indoor environments consisting of doorways, corridors and rooms - as it's name suggests, vision is a key sensory modality. One limitation of the Eye-bots is the inability to estimate the distances to far-away walls or estimate the area of rooms. Perspective is a powerful visual tool frequently used by humans to estimate 3D shape and distances from single still images without requiring complex and well-calibrated stereoscopic processing. This semester project entails the development of monocular (single image) vision algorithms that can estimate distances to walls in indoor rectilinear environments exploiting geometric simplicities afforded by the regular structure and orthogonal-planar constraints. Other useful information can also be extracted from such images such as camera orientation and alignment in rooms and corridors. This project starts with literature based research to understand the state of the art and theory in monocular image depth processing followed by a brain-storming period which will finalise a vision processing system suitable for the Eye-bots' hardware constraints. Development, implementation and analysis are equally important.

    Type: Semester project
    Period: 15.09.2008 - 15.01.2009
    Section(s): IN MT PH
    Type of work: 40% theory, 50% software, 10% analysis
    Requirements: Programming, reasonbale mathematics. Computer vision or image, processing and machine learning useful
    Subject(s): Computer Vision
    Responsible(s): Timothy Stirling, Yannick Weibel
    URL: Click here

    Spherical Jumping Rolling Robot

    Manuel Schlegel (MT)

    Locomotion in rough terrain is one of the main challenges for mobile robots. One promising strategy is to jump and thus be able to overcome relatively big obstacles. So far, we developed a novel 7gram jumping robot that is able to overcome obstacles of up to 1.4m heigth, which corresponds to more than 27times its own size, but, without the ability to upright itself after landing. This project consists of developing a spherical cage system for the jumping robot so to enable it to jump and roll. Such a combination of locomotion modes can be very effective and be applied for a new generation of robot swarms in rough terrain on earth and other planets. The project includes a systematic innovation/creativity phase and subsequent the implementation of the ideas and production of a functional prototype. The final prototype should be able to jump, roll and jump again, once in proper position. This may be ensured by a passive mechanism to ensure the right position prior to jumping or be determined by the robot using on board sensing.

    Type: Semester project
    Period: 15.09.2008 - 10.01.2009
    Section(s): EL IN MA ME MT MX
    Type of work: 10% théorie, 20% software, 70% hardware
    Requirements: Creativity, Ambition, Research Attitude
    Subject(s): hybrid locomotion
    Responsible(s): Mirko Kovac, Adam Klaptocz
    URL: Click here

    Development of an acoustic communication turret for the e-Puck robot

    Matthias Furler (MT)

    One research focus at LIS lies in the evolution of communication and cooperation in groups of robots. So far, signaling between robots relied on visual information. However, for more sophisticated forms of collective behavior and communication multiple communication modalities are required.

    The goal of this project is to develop an efficient acoustic communication system for the e-Puck robot, in the form of an add-on turret. The ultimate aim is to realize robust signaling among up to 20 interacting robots. The project will start with a review of acoustic transmission theory and an evaluation of the existing microphone turret to identify the parts that require redesign (speaker, microphones, sound processor, etc.). A new turret will then be designed and a prototype will be built that can produce at least four distinct sound signals and detect the direction and distance/intensity of sounds emitted by other turrets. The system then needs to be proven to function robustly with many simultaneously interacting robots.

    Optionally, cooperative behaviors based on acoustic communication can be implemented (such as collectively finding energy sources, etc.) to demonstrate the turret's functionality.

    Type: Semester project
    Period: 15.09.2008 - 31.12.2008
    Section(s): EL IN MT
    Type of work: 20% theory, 40% hardware, 40% software
    Requirements:
    Subject(s): Acoustics, electronics, circuit board design, signal processing
    Responsible(s): Adam Klaptocz, Steffen Wischmann
    URL: Click here

    Observation of Insect Collisions

    Adrien Briod (MT)

    Some flying insects are remarkably adept at navigating cluttered environments at high speeds. Elegant obstacle avoidance sensors and algorithms have been suggested as models for how insects operate in constrained environments., But obstacles are not always avoided: it is common to observe insects violently crashing into objects with little or no detriment to their flight performance or mechanical integrity. This project will involve staging experiments in which we induce flies to collide with objects in their environment while observing both body and wing kinematics with an off board high speed motion capture system. The end goal is to characterize the dynamics of collisions and subsequent recovery in an attempt to identify salient features (both in the system dynamics and the mechanics) necessary for rapid recovery of a flapping-wing MAV.

    Type: Master project
    Period: 15.09.2008 - 31.12.2008
    Section(s): ME MT SV
    Type of work: 50% experiments, 50% analysis
    Requirements:
    Subject(s): biology
    Responsible(s): Adam Klaptocz, Robert Wood
    URL: Click here

    Advanced omni-directional distance scanner using 3D image sensor

    Peter Oberhauser (MT)

    In order for a miniature aerial vehicle (MAV) to navigate and fly within a cluttered indoor environment it is necessary that the MAV is able to detect objects and avoid collisions. The MAV can then determine possible flight paths for exploration. A new 3D image sensor that has been developed by Dr Cristiano Niclass under the supervision of Dr Edoardo Charbon, is capable of measuring the distance to objects with sub centimeter precision. Currently this new technology is patterned in an array of single photon detectors able to 'see' objects much like a normal camera but in three dimensions. In this project a new scanning device will be developed to move the sensor about a full 360 degrees in order to obtain a full omni-directional depth map of the surrounding environment. The device will require close collaboration with the sensors developers as listed above, in order to 1. find a correct package for the sensor. 2. choose the correct light weight optics. 3. choose the correct low power modulated light source. 4. interface the sensor to a computer. A prototype will be fabricated that can interface to the existing eye-bot hovering MAV.

    Type: Master project
    Period: 15.09.2008 - 19.12.2008
    Section(s): MT
    Type of work: 10% theory, 10% software, 30% experimentation, 50% hardware
    Requirements: mechanical design, electronics
    Subject(s): 3D imaging, Mechanics, Electronics, MAVs
    Responsible(s): James Roberts, Timothy Stirling
    URL: Click here

    Micro-Quadrotor-Mach2

    Kaspar Leuenberger (MT), Philippe Bérard ()

    A prototype of the Micro-Quadrotor has been developed last semester by Simon Fivat and Lucas Oehen. This semester the project will continue, the goal will be to enable stable flight and to optimize both the avionics hardware and software., This project will require two students to: 1. achieved stable flight with the current system. 2. optimise the avionics so that the system uses minimal processors and fits on a single printed circuit board. 3. improve the interface for the extended sensor connection. 4. improve the radio control interface. If time permits the system will be implemented and tested with a RABIT/GPS sensor.

    Type: Semester project
    Period: 15.09.2008 - 19.12.2008
    Section(s): MT
    Type of work: 10% theory, 40% software, 50% hardware
    Requirements: electronics & C programming skills
    Subject(s): aerial robotics, avionics, miniaturisation
    Responsible(s): James Roberts, Adam Klaptocz
    URL: Click here
     

    Optic-flow-based tracking of moving objects from a moving viewpoint

    Yannick Weibel (MT)

    This project will involve developing and testing algorithms for moving-object detection by (a) computer simulation (b) tethering the vision system on the arm of a moving gantry in our lab, and, when the algorithms are satisfactory (c) testing on a vision system mounted on a model aircraft, with the task of detecting another model aircraft flying in the vicinity. Tracking and interception could be implemented later, if time permits and depending upon how well the project progresses.

    Type: Master project
    Period: 18.02.2008 - 11.07.2008
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): computer vision, aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Prof. Srinivasan
     

    Waypoint Navigation with an MAV

    Adrien Briod (MT)

    This semester project aims at implementing a waypoint navigation algorithm for a micro air vehicle (MAV) based on GPS position measurements First, the candidate will study different strategies proposed in literature, compare them, and make a short evaluation of the most interesting algorithms with a given flight simulation. The second and main task of the project is to implement a chosen waypoint algorithm on a real flying platform, which will be provided by the Lab. It already includes a basic electronic control system which directly permits the implementation of navigation algorithms. The second task also involves extensive flight testing and the analysis of recorded flight data. The goal of the obtained flight controller should finally be to show an example of the existing platform's usefulness for educational purposes and to be used in an autonomous MAV competition in 2008.

    Type: Semester project
    Period: 18.02.2008 - 30.06.2008
    Section(s): EL IN MA ME MT
    Type of work: 20% theory, 40% software, 40% experiments
    Requirements: C-programming, basic electronic knowledge, Interest in Flying Robotics
    Subject(s): Micro Air Vehicle Flight Controller Waypoint GPS Navigation
    Responsible(s): Severin Leven, Jean-Christophe Zufferey
    URL: Click here

    The effects of noise in evolutionary robotics

    Steve Berger (SV)

    Evolutionary algorithms are optimization procedures inspired by the principles of Darwinian evolution. One of the main challenges in applying such methods is that most real world optimization problems are subject to considerable amounts of noise. It has been suggested to increase the robustness of evolutionary algorithms to noise by using large population sizes or by evaluating individual solutions multiple times. However, these methods drastically increase the computational cost per generation and there is very little research dealing with this trade-off. The goal of this semester project is to systematically investigate the trade-off between population size and multiple evaluations for a difficult evolutionary robotics experiment. The experiment consists of a wheeled robot which has to find reward tokens in a double T-maze. The robot is controlled by a modulatory neural network whose parameters are encoded in the artificial genome. In a first step, the existing software simulation of the robot has to be integrated with the artificial evolution framework OOGA. In a second step, a set of numerical experiments should be carried out to investigate the influence of population size, number of evaluations per individual, and selection operator on algorithm performance. Finally, the obtained results should be analyzed statistically.

    Type: Semester project
    Period: 18.02.2008 - 15.06.2008
    Section(s): EL IN MT SV
    Type of work: 50%Théorie, 50% Software
    Requirements: C++, Matlab
    Subject(s): Evolutionary Computation, Noisy Fitness
    Responsible(s): Peter Duerr, Daniel Marbach
     

    Micro-Quadrotor

    Simon Fivat (MT), Lucas Oehen ()

    The current worlds smallest quadrotor, the MicroX4, measures only 160mm and weighs a mere 40g (designed by Pascal ZUNINO and Jean-Claude PESCE, see picture). Here at the laboratory of intelligent systems (LIS) we intend to develop a similar technology but also add a level of practicality by adding a new range and bearing sensor (RABIT) developed in the LIS. This project aims at developing a small and light weight micro quadrotor helicopter that has the expansion capability for the RABIT. Two students will work together to develop the propulsion system, body structure and integrate this with the avionics firmware and hardware. Example schematics and micro-controller control code will be supplied. First, a study of viable propulsion systems will be carried out, followed by experimental bench testing. The characteristics obtained in these experiments will give an incite on the available payload and endurance and provide an outline on the design criteria for the structural body. The avionics system will then need to be integrated with this design and the system will then be fabricated.

    Type: Semester project
    Period: 18.02.2008 - 15.06.2008
    Section(s): EL MA ME MT
    Type of work: 10% theory, 5% software, 30% experimentation, 55% hardware
    Requirements: mechanical design, electronics
    Subject(s): aerial robotics, avionics, miniaturisation
    Responsible(s): James Roberts, Adam Klaptocz
    URL: Click here

    Scalable reverse engineering of nonlinear gene networks in the presence of noise

    Andreas Weishaupt (EL)

    The effective reverse engineering of gene regulatory networks is one of the great challenges of systems biology. We have previously developed an evolutionary reverse engineering method that allows an accurate reconstruction of the gene network using biologically plausible, dynamical models. A problem decomposition approach is used to achieve scalability for large networks. An implementation of the method in C++ is already available. In this project, the student is expected to adapt this code and apply the method to a benchmark dataset of a fifty-gene in silico network, which has been published as an international reverse engineering competition for the second DREAM conference (New York, 2007). Furthermore, the student will develop, implement, and test different variations of the problem decomposition strategy and analyze the performance in the presence of noise.

    Type: Semester project
    Period: 18.02.2008 - 06.06.2008
    Section(s): EL
    Type of work: 30% théorie, 70% software
    Requirements: C++
    Subject(s): gene regulatory networks, reverse engineering
    Responsible(s): Daniel Marbach, Claudio Mattiussi
    URL: Click here
     

    Long Range Autonomous MAV Using Mobile Communication Network

    Lionel Brocard (MT)

    The purpose of this master project is to develop an MAV able to follow a predefined flight plan using GPS waypoint navigation and to send images of spots of interest on the ground through a mobile communication network, thus allowing the user to have visual insight into remote and maybe dangerous places. The first part of the project is the development of the carrier plane. Either a commercial or an in-house (LIS) developed autopilot will be fitted on the airframe in order to obtain a stable and safe aircraft capable of carrying autonomously a payload at a low altitude (100m) following a given flight plan (waypoints). During setup of the autopilot on the plane, a safety pilot will take over control in case of failure. Since this only works at short ranges, additional safety measures like a parachute system and its opening system will be evaluated. The second part of the project consists in adding an orientable camera to take pictures of spots of interest on the ground and to send them through a high-bitrate mobile communication network like UMTS or HSDPA. Due to inhomogeneous coverage of the several high-bitrate communication protocols in use in Switzerland, the system must be able to 1) dynamically switch between communication protocols and 2) adapt the size of the images or to save them on an embedded memory in order to send them later or to bring them back to the user in case of poor coverage.

    Type: Master project
    Period: 15.09.2007 - 10.02.2008
    Section(s): MT
    Type of work: 100% system development
    Requirements: Electronics, C/C++-Programming, Knowledge in Communication Technologies
    Subject(s): MAV, UMTS, HSDPA, Waypoint Navigation, Autonomous
    Responsible(s): Severin Leven, Jean-Christophe Zufferey

    2007


    In flight dynamics of the Self Deploying Microglider

    Cécile Grometto (IN)

    A novel palm sized biomimetic flying platform is currently in development at LIS. It will be very light weight, able to climb up walls or jump, open its wings and do subsequent goal directed gliding flight and attachment to walls. There are a number of technical and conceptual challenges that are encountered when trying to realize such a robotic system. One of the key issues is the ability recover in mid air so to prepare the gliding phase. The goal of this project is to explore different kinds of stabilization methods for gliding flight and implement a solution on the existing flying platform and characterize the in flight dynamics. The application of our new high speed imaging equipment and our 'holodeck' will be part of it. The goal of this project is to determine stability parameters and recovery height when dropped. The parameters to change are (i) the position of the center of gravity, (ii) wing dihedral and (iii) angle between the main wing and the tail. The work includes the fabrication of winged prototypes and the anaylsis of the flight path movies in ProAnalyst.

    Type: Semester project
    Period: 15.09.2008 - 10.01.2009
    Section(s): EL IN MA ME MT
    Type of work: 30% théorie, 20% software, 50% hardware
    Requirements: Creativity, Ambition, Research Attitude
    Subject(s): aerial robotics, bioinspired locomotion
    Responsible(s): Mirko Kovac, Severin Leven
    URL: Click here

    Design of a Hovering Microflyer

    Grégoire Boutinard Rouelle (MT)

    At LIS we have a vision of one day being able to reproduce the complex yet elegant flight behaviours that exist in nature, in particular the housefly. Current research has yielded a robotic microflyer capable of autonomous flight in a confined indoor environment at a weight of only 10g! Based on an airplane-type design, however, means that the platform must be in constant forward motion. The next step is to add the capability of hovering in one spot to this platform.

    This project consists of the initial design of a basic single- or multi-rotor hovering, with an emphasis on crash-proofness and self-recovery. An extensive state of the art will examine existing or possible platform configurations (such as single rotor, flapping wing, 2 contra-rotating rotors, etc), weighing the advantages of each solution. The first stage of the design will include tests of platform components, such as the motor and propeller or the actuators. A first prototype of the new platform will then be designed in CAD, and optionally will be built and demonstrated by remote control.

    Type: Semester project
    Period: 15.09.2008 - 31.12.2008
    Section(s): ME MT
    Type of work: 30% Theory, 70% mechanics
    Requirements:
    Subject(s): Aerial Robotics
    Responsible(s): Adam Klaptocz, Jean-Christophe Zufferey
    URL: Click here

    Design of a monitoring GUI software for flying robots

    Alexandre Habersaat (IN)

    The LIS is involved in several projects in the domain of flying robotics (Microflyers, Swarming MAVs, Swarmanoid, etc.). There is a need for a versatile, customisable and cross-platform GUI software to remotely monitor and control experiments, based on an existing communication architecture (Ishtar).

    The student will have to:
    - define specifications for the monitoring system;
    - study the provided communication architecture (Ishtar) and assess its features with respect to the specifications;
    - if needed, enhance Ishtar to include the required features;
    - implement the ground-based monitoring GUI software using the Qt framework;
    - demonstrate the system using a real flying robot and Xbee-based communication.

    Since the software is expected to be used for many current and future projects, we want to put emphasis on versatility, sound design and high-quality programming.

    Type: Master project
    Period: 18.02.2008 - 11.07.2008
    Section(s): IN
    Type of work: 80% software 20% experiment
    Requirements: C++ programming, software architecture, GUI programming
    Subject(s): Scientific software development
    Responsible(s): Antoine Beyeler, Severin Leven
    URL: Click here

    Hardware and control for near obstacle flight

    Laurent Coutard (MT)

    One big issue in the future of small UAVs (Unmanned Aerial Vehicles) is the ability to fly at very low altitude, close to and among obstacles. This project aims at identifying which sensors and control strategies can be efficiently used to autonomously fly very close to a horizontal surface (terrain following), to land safely or to avoid collisions with vertical objects. The testbed will be a flying wing of approximately 150 g, which can embed an additional 100 g of payload. The first step will consist in reviewing range finders that could be used for this task. In a second step, the candidate will integrate a selection of these sensors with the on-board electronics in order to perform open-loop tests and recordings. After analysis, the student will try and close the loop to demonstrate fully autonomous operation in the above-mentioned situations.

    Type: Semester project
    Period: 18.02.2008 - 30.06.2008
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Severin Leven
     

    Evolving neural networks for sleep / wake detection

    Jérémie Guignard (SV)

    Artificial neural networks can be used to classify sleep / wake patterns from physiological signals like respiration or heart activity. Until now, the topology of these networks has been hand-designed for the use with a supervised learning algorithm. In this semester project, analog genetic encoding (AGE), a new method for the evolution of topology and synaptic weights of neural networks, should be applied to the sleep detection of the Solar Impulse pilot. The students work will consist of implementing a fitness function for the application of AGE with physiological time series. Then, a series of numerical experiments should be done, followed by a thorough analysis of the results regarding network topology and performance. Finally the obtained networks can be compared to the hand-designed structures employed so far.

    Type: Semester project
    Period: 18.02.2008 - 15.06.2008
    Section(s): IN MT
    Type of work: 60% theory, 40% software
    Requirements: C++, neural networks
    Subject(s): AGE, neural networks, biomedical signal processing
    Responsible(s): Walter Karlen, Peter Duerr

    Optimization and scaling in jumping robots

    Michael Schwander (EL)

    Jumping is a very efficient way to locomote on ground. Different animals at different sizes adopt this principle to overcome large obstacles or just to help them getting into the air. In robotics however, jumping mechanisms at small scale and low weight do almost not exist and need further exploration. Recently, a 6g jumping robot has been developed at LIS that allows jumping as high as 27 times its own body size. This performance can however be optimized even further by adopting knowledge gained from biological systems to the geometric space of this existing jumping robot. The goal of this project is thus to explore what is known from biology regarding jumping performance and conduct a series of experiments using our jumping robot. The work can be roughly subdivided in the following work packages: - Identification of the relevant parameters on jumping performance from biological jumping models - Fabrication of different leg-lengths - Conduct of jumping experiments (monitoring using high speed imaging and the 6DOF ground force sensor) - Summary and discussion of the results

    Type: Semester project
    Period: 11.02.2008 - 15.06.2008
    Section(s): EL IN MA ME MT MX
    Type of work: 30% théorie, 10% software, 60% hardware
    Requirements: Creativity, Ambition, Research Attitude
    Subject(s): bioinspired locomotion, jumping robots
    Responsible(s): Mirko Kovac, Severin Leven
    URL: Click here

    Curriculum of exercices for e-puck and Webots

    Fabien Rohrer (IN)

    The Cyberbotics enterprise wants to create a new offer to its customers: to sell a new package composed of its main software, a mobile robot and a curriculum (tutorial). Project is constituted of two main parts: improve interactions between software and robot and create this curriculum. For the first part a lot of things are already done: E-Puck 3D-modelisation, E-Puck integration in the Webots environment, bluetooth connexion between both (remote control session), possibility to upload a cross-compiled program on E-Puck, etc. But in each of these steps there are some bugs or some features to ameliorate. On the other side there are also some new features to add like the possibility to create cross-compiled program by using directly Webots text editor. Second part objective is to create a kind of tutorial for customers. It must be addressed for a large target public i.e. with a very different skills and particularly in an education domain. The curriculum will take the following form: a succession of exercises containing theoretical parts. They begin with a graphical programing interface (BotStudio) and end up with advanced exercises (like image processing, artificial intelligence). The final part of the curriculum will include cognitive benchmarks.

    Type: Master project
    Period: 18.09.2007 - 30.03.2008
    Section(s): IN
    Type of work:
    Requirements:
    Subject(s): Robotics, simluation
    Responsible(s): Jean-Christophe Zufferey, Olivier Michel
     

    Avian tail for the Self Deploying Microglider

    Joël Israël (IN)

    A novel palm sized biomimetic flying platform is currently in development at LIS. It will be very light weight, able to climb up walls or jump, open its wings and do subsequent goal directed gliding flight and attachment to walls. There are a number of technical and conceptual challenges that are encountered when trying to realize such a robotic system. The proper control of the flight direction while ensuring stability is one of them. Nature is a potentially useful source of inspiration to tackle technical challenges. In the animal kingdom, e.g. many birds use their tail to control pitch, roll and drag at the same time. The goal of this semester project is to develop a novel bioinspired adaptive steering system for gliding microrobots for simultaneous control of pitch and roll. The work can be divided in the following work packages: - Overview of different steering mechanisms in robots and in animals - Concept and evaluation phase of different designs for a suitable mechanism for the Microglider - Implement one solution in a CAD program (ProE) - Fabricate the mechanism and demonstrate its behavior - Integrate the mechanism on a glider and balance the flight dynamics based on data from on board sensors (i.e. anemometer, gyroscope, accelerometer)

    Type: Semester project
    Period: 18.09.2007 - 15.02.2008
    Section(s): EL MA ME MT MX
    Type of work: 30% théorie, 20% software, 50% hardware
    Requirements: Creativity, Ambition, Research Attitude
    Subject(s): Aerial Robotics, Hybrid Locomotion
    Responsible(s): Mirko Kovac, Walter Karlen
    URL: Click here

    Embedded electronics for the next generation microflyers

    Adam Klaptocz (MT)

    The goal of this project is to integrate custom vision sensors on the existing 10-gram microflyer available in our laboratory in order to achieve autonomous flight in an experiment room. This master project is divided into the two following parts: 1) Further development and test of the new embedded electronic system (pevopic5) including a new set of chips: aVLSI-based artificial eye, 2-axes gyros, anemometer, and Bluetooth module. Special care should be taken to achieve a robust and easy to use system, both from the hardware and software perspective. 2) Development and test of an efficient control strategy allowing for autonomous flight in changing visual environments. Special emphasis will be made on the variation of spatial frequency and ambient light.

    Type: Master project
    Period: 18.09.2007 - 10.02.2008
    Section(s): MT
    Type of work: 50% hardware & firmware, 50% control and experiments
    Requirements:
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here
     

    MAV Flight Model Based on Sensor Data

    Vincent Bozzo (IN)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two or more ground emitters (base station and user stations) is currently being developed in our lab. The MAVs navigate using swarm intelligence relying only on information obtained through local communication with neighboring MAVs and internal sensory information. One of the main difficulties is that no positioning information is present in the system (i.e no GPS or other positioning sensors). The goal of this project will be to fine tune the simulated ODE-based flight model of the current MAV platform. Having an accurate flight model is an important step in making transfer of simulated controllers to reality feasible. A strategy for automatically determining the coefficients of the flight model based on in-flight test data should be implemented and thoroughly tested. A final effort will go towards demonstrating the accuracy of the simulated trajectories with respect to outdoor experiments conducted in the scope of the "transfer to reality" effort during which tests with 1 to 3 MAVS and a ground station will be done. Interested students can also build 1-2 MAV platforms.

    Type: Semester project
    Period: 18.09.2007 - 01.02.2008
    Section(s): IN
    Type of work: 20% theory, 50% software, 30% tests
    Requirements: linux and c++ an advantage
    Subject(s): MAV flight tests, transfer from simulation to reality, flight model
    Responsible(s): Sabine Hauert, Severin Leven
    URL: Click here

    Evolving communication in a group of e-puck robots

    Alban Laflaquière (MT)

    Recent work at the Laboratory of Intelligent Systems has explored the evolution of a group of communicating S-bots required to solve a foraging task by evolving neural controllers. Although the S-bot robot has been a satisfactory tool in exploring the evolution of communication, the robots are unreliable and require a high level of maintenance to run extensive experiments. We would therefore like to switch to using the e-puck robot for these experiments. The aim of this project is to transfer the neural controllers of a group of communicating e-pucks evolved in simulation onto the physical robots. This would consist of an initial comprehension of the behaviour of the robots evolved in simulation, followed by a transfer of that behaviour to the real robots. The student should systematically compare the performance of the algorithm on the physical robots to that of the robots evolved in simulation.

    Type: Semester project
    Period: 14.09.2007 - 01.02.2008
    Section(s): EL IN MT SC
    Type of work: 30% running simulations, 60% programming robot, 10% analysis
    Requirements: C/Cpp, preferable: experience with e-puck
    Subject(s): evolutionary robotics
    Responsible(s): Sara Mitri, Julien Hubert
    URL: Click here

    Porting Aerial Swarming to Reality

    Damien Rosat (IN)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two or more ground emitters (base station and user stations) is currently being developed in our lab. The MAVs navigate using swarm intelligence relying only on information obtained through local communication with neighboring MAVs and internal sensory information. One of the main difficulties is that no positioning information is present in the system (i.e no GPS or other positioning sensors). The goal of this project will be to automatically transfer simulated swarm controllers to the actual MAV linux boards and user stations. MAVs and ground users should be capable of broadcasting and receiving simple messages which are needed for the control of the swarm. The software interface between microcontroller and linux-processor should be set up. If time permits, a ground station consisting of a PDA and wireless device will be implemented. Final effort will go towards demonstrating successful transfer of swarm algorithms with out-door experiments in the scope of the "transfer to reality" effort during which tests with 1 to 3 MAVS and a ground station will be conducted. Interested students can also build 1-2 MAV platforms.

    Type: Semester project
    Period: 18.09.2007 - 01.02.2008
    Section(s): IN
    Type of work: 10% theory, 30% hardware, 30% software, 30% tests
    Requirements: linux, c++
    Subject(s): embedded linux, wireless communication, tranfer to reality
    Responsible(s): Sabine Hauert, Severin Leven
    URL: Click here

    Experimental Protocol and Monitoring for the SMAV Projec

    Alexandre Habersaat (IN)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two or more ground emitters (base station and user stations) is currently being developed in our lab. The MAVs navigate using swarm intelligence by relying only on information obtained through local communication with neighboring MAVs and internal sensory information. One of the main difficulties is that no positioning information is present in the system (i.e no GPS or positioning sensors). The goal of this project will be to implement the necessary tools to allow for outdoor experiments with multiple MAVs and one ground station. These are: a connector-board Microcontroller/Linux-Processor/Wireless device and the adaptation of the current flight monitoring software for use with multiple MAVs. Final effort will go towards demonstrating the tools with outdoor experiments in the scope of the "transfer to reality" effort during which tests with 1 to 3 MAVS and a ground station will be conducted. The experimental protocol discussed with the assistants should be taken into account. Interested students can also build 1-2 MAV platforms.

    Type: Semester project
    Period: 18.09.2007 - 01.02.2008
    Section(s): IN
    Type of work: 10% theory, 20% hardware, 40% software, 30% tests
    Requirements: c++
    Subject(s): MAV flight tests, transfer from simulation to reality, robot monitoring
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here

    Scalable Reverse Engineering of Nonlinear Gene Networks

    Thomas Schaffter (MT)

    The effective reverse engineering of gene regulatory networks is one of the great challenges of systems biology and is expected to have substantial impact on the pharmaceutical and biotech industries in the next decades. A gene network is formed by regulatory genes, which code for proteins that enhance or inhibit the expression of other regulatory and/or non-regulatory genes, thereby forming a complex web of interactions. The goal of reverse engineering is to automatically identify such a network from experimental data. In this project, we explore an evolutionary reverse engineering approach for gene networks of medium size. As benchmark we use a dataset of a fifty-gene in silico network, which has been published as an international reverse engineering competition for the second DREAM conference (Dialogue for Reverse Engineering Assessments and Methods). Our goal is to provide an accurate reconstruction of the gene network using biologically plausible dynamical models. With this kind of nonlinear models, traditional methods of system identification are generally not applicable. Hence, the reverse engineering algorithm must rely on stochastic global search methods and a "divide-and-conquer" strategy to achieve scalability.

    Type: Master project
    Period: 18.09.2007 - 18.01.2008
    Section(s): MT
    Type of work:
    Requirements:
    Subject(s): Systems Biology
    Responsible(s): Daniel Marbach, Claudio Mattiussi
     

    Omni-directional Distance Scanner for the Eye-bot

    Peter Oberhauser (MT)

    As part of the European funded, Swarmanoid (http://www.swarmanoid.org/), project, the LIS lab is developing a unique flying robotics platform for indoor exploration: the Eye-bot. As the Eye-bot will fly indoors it is a necessity that the immediate surrounding environment be scanned for obstacles so that the Eye-bot can maneuver without crashing. This project involves the design and testing of a 360 degree distance scanner based on the Sharp GP2Y3A002K0F distance sensor. The prototype must meet strict size, weight and power consumption requirements. To accompany the sensor a simple c program will be developed using Monitor to display the information in a 2D obstacle map for easy visualization.

    Type: Semester project
    Period: 18.09.2007 - 12.01.2008
    Section(s): EL ME MT
    Type of work: 20% theory, 25% software, 55% hardware
    Requirements: basic electronics, mechanics, c programming
    Subject(s): sensors, aerial robotics
    Responsible(s): James Roberts, Timothy Stirling
    URL: Click here

    Gecko inspired climbing mechanism

    Jerome Favre (MT), Andreas Elmer ()

    A novel palm sized biomimetic flying platform is currently in development at LIS. It will be very light weight, able to climb up walls or jump, open its wings and do subsequent goal directed gliding flight and attachment to walls. There are a number of technical and conceptual challenges that are encountered when trying to realize such a robotic system. One of the key issues is the ability to attach and detach to walls or obstacles when desired. The goal of this semester project is to explore different kinds of dry adhesive materials and propose a design of an attachment mechanism that is suitable for our new robot. The work can be divided in the following work packages: - Overview of different climbing mechanisms in robots and in animals - Concept and evaluation phase of different designs for a suitable mechanism for the Self Deploying Microglider - Implement one solution in a CAD program (ProE) - Fabricate and characterize the mechanism and demonstrate its behavior

    Type: Semester project
    Period: 18.09.2007 - 19.12.2007
    Section(s): EL MA ME MT MX
    Type of work: 30% théorie, 20% software, 50% hardware
    Requirements: Creativity, Ambition, Research Attitude
    Subject(s): aerial robotics, bioinspired locomotion
    Responsible(s): Mirko Kovac, James Roberts
    URL: Click here

    Solar Cells for Self Deploying Microglider

    Jean-Paul Fuchs (MT)

    A novel palm sized biomimetic flying platform is currently in development at LIS. It will be very light weight, able to climb up walls or jump, open its wings and do subsequent goal directed gliding flight and attachment to walls. The contribution of this semester project is to develop a solar cell system to recharge the robot and enable it to move autonomously and independently for long periods of time. The work can be divided in the following work packages: - Overview on existing small ground/aerial robots that use solar cells - Evaluation which solar cells can be used for the Self Deploying Microglider and which performance can be expected, summary of the main challenges. - Development of a PCB including the required electronics to charge (i) a 20mAh LiPo battery and (ii) a capacitor directly connected to the motor using the chosen solar cell technology. - Integration of the entire solar cell-electronics-battery system into an existing wing structure on the jumping robot, (including eventual modifications of the wing structure) - Characterization and demonstration of the working prototype.

    Type: Semester project
    Period: 18.09.2007 - 19.12.2007
    Section(s): EL IN MA ME MT MX
    Type of work: 20% théorie, 20% software, 60% hardware
    Requirements:
    Subject(s): aerial robotics, electronics
    Responsible(s): Mirko Kovac, Walter Karlen
    URL: Click here

    Emergent Communication and Coordination for Optimal Swarm Coverage

    Nicolas Delieutraz (MT)

    As part of the European funded, Swarmanoid (http://www.swarmanoid.org/), project, the LIS lab is developing a unique flying robotics platform for indoor exploration: the Eye-bot. Coverage of an indoor multi-room environment by multiple robots is a critical basis for swarm search that has attracted much attention recently. Coverage entails the systematic search of every location of an enclosed environment, preferably optimally without search duplication (multiple robots searching the same location), but while retaining robustness in case of robot failures. The advantages of an evolutionary approach is that individual behaviours can be co-evovled with social behaviours and novel communication strategies which improve coordination. This project entails the development of an algorithm that performs multi-robot coverage using an evolutionary approach. In particularly, studying the emergence of communication signals that improve performance. The evolved behaviours should be characterised under various conditions (environment size and complexity, team size, time limit, etc.). Work will be done in simulation.

    Type: Semester project
    Period: 16.09.2007 - 18.12.2007
    Section(s): EL IN MT SC
    Type of work: 30% theory, 60% software, 10% analysis
    Requirements: c/cpp programming
    Subject(s): evolutionary robotics, swarm robotics
    Responsible(s): Timothy Stirling, James Roberts
    URL: Click here

    Embedded Evolution on a Team of e-pucks

    Patrick Terreaux (IN)

    Artificial evolution is a powerful method to develop control systems for robot teams. Most current work uses computer simulations to evolve robot teams and subsequently transfers the resulting control systems to the robots. However, with this method inadequacies of the software simulation inevitably lead to a performance loss. The alternative is to evolve control systems in a real environment using real robots. This solves the so-called "reality gap" problem but requires a particular setup to monitor the behavior of the robots.

    The aim of this project is to implement the necessary architecture for embedded evolution using the e-puck robots and evolve a control system for a simple task involving a team of robots. More specifically, this project will accomplish the following points:

    * Provide the infrastructure needed for embedded evolution:
    - Adapt an existing camera based robot tracking system to the e-puck robots.
    - Establish infrared communication between tracking system and the e-puck.
    - Program or port a feed-forward neural net implementation to the e-puck.
    - Implement a system detecting the battery charge of the robots and replacing them automatically
    * Use embedded evolution to solve a simple task with a small e-puck team.

    Type: Master project
    Period: 19.03.2007 - 01.08.2007
    Section(s): IN MT
    Type of work: 70% software, 30% research
    Requirements: C/C++ programming, computer vision, neural networks
    Subject(s): robots, evolutionary robotics, neural networks, communication
    Responsible(s): Julien Hubert, Danesh Tarapore
    URL: Click here

    Omni-Directional Vision for Hovering MAV

    Adam Klaptocz (MT)

    As part of the European funded Swamanoid project, the LIS lab is developing a unique robotics system based on a contra-rotating, mechanically self-stabilizing hovering platform for search-and-rescue type applications - the "Eye-bots". This project pertains to the design and development of an omni-directional vision sensor(s) with 3D views, thus providing a visual overview of the environment from the unique vantage point afforded by the Eye-bots. The Eye-bots with their top-down vision will help coordinate the activities of other ground-based robots. The flying platform enforces tight design requirements including the absolute reduction of weight, power consumption, dimensions and processing requirements. Various types of vision sensor designs will be compared including: a single sensor with an omni-directional optical lens/mirror system, omni-directional articulated camera system and a multi-camera system with multiple view points. A prototype vision system, with elementary support drivers, will then be developed based on the best solution. The system will then be implemented and tested on the Eyebot platform, to review the overall performance of the sensor.

    Type: Semester project
    Period: 12.03.2007 - 29.07.2007
    Section(s): MT
    Type of work: 60% hardware, 25% theory, 15% software
    Requirements: Hardware Knowledge of Cameras Advantageous
    Subject(s): Camera and Vision Sensors
    Responsible(s): James Roberts, Timothy Stirling

    Data Monitoring System for a Micro Air Vehicle

    Laurent Hauser (MT)

    In the "Swarming MAVs" project at LIS, a fleet of autonomous Micro Air Vehicles (MAV) is developed which will operate as a wireless communication relay. At the current stage of the project, onboard data of an MAV is sent continuously during flight to a ground station (PC). This flight data (airspeed, altitude, attitude etc.) is transmitted via a radio link. For recording and online displaying purposes, a monitoring software is used which has been developed at LIS. The semester projects aimes at adding improvements and customizations to this software, namely: 1) a module to show in realtime the attitude (Euler Angles) of the airplane using an OpenGL window, 2) a module to play back recorded flight data after a flight (trajectory from GPS and attitude), if possible including a topographical map as ground, 3) adding parameter configuration buttons to the GUI, 4) creating a proper way of interpreting, channeling and logging incoming data streams (configurable in the Cpp-Code).

    Type: Semester project
    Period: 01.04.2007 - 15.07.2007
    Section(s): EL IN MA ME MT
    Type of work: 100% software
    Requirements: C++, OpenGL
    Subject(s): Sensor Data Monitoring Software Micro Air Vehicle
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here

    Active vision and online adaptation to different environments

    Marc Clapera (MT)

    In the human eye movements during facial and object recognitions, Noton and Stark (1971) has claimed that when a particular visual pattern is viewed, a particular sequence of eye movements is executed and furthermore that this sequence is important in accessing the visual memory for the pattern. Inspired by this work, we speculate that effective eye movements may allow an autonomous car to quickly ``identify'' an environment and to execute a control strategy suitable for the environment. That way, the car may successfully navigate in different environments. The aim of this project is to test the validity of this hypothesis. The student engaged in this project uses a 3D physics-based simulator for modeling the car and the environments. She or he can use the code developed already in our previous work. The student is required to 1) design several types of environment, 2) implement and test several types of recurrent neural controller (e.g., continuous time recurrent neural network), 3) perform a set of artificial evolution experiments and analyze their outcome and 4) analyze the scanning strategy and internal states of the best evolved neurocontroller. At the end of the project, the student is expected to demonstrate successful car navigation in two significantly different environments.

    Type: Semester project
    Period: 16.03.2007 - 09.07.2007
    Section(s): IN
    Type of work: 60% software, 40% experiments and analysis
    Requirements: good knowledge of Neural Networks and C/C++ programming
    Subject(s): computer simulation, neural networks, evolutionary robotics
    Responsible(s): Mototaka Suzuki, Julien Hubert
    URL: Click here

    Jump to Glide Transition

    Vitanov Vasko (MT)

    A new biomimetic platform for flying miniature robots capable of terrestrial and aerial locomotion is currently in development at LIS. This system will possess wings that are foldable and some sort of running or jumping mechanism in order to move on ground, open the wings and fly.

    The goal of this semester thesis is to accurately detect the highest point of the jump phase of a small robotic system. This achievement is of key importance to coordinate the transition from jumping to the wing unfolding and subsequent gliding phase. The work includes the following work packages:

    -
  • Evaluation of different sensors that could be used for this task.
  • -
  • Proposition of a justified choice of one sensor setup
  • -
  • Design and fabrication of a PCB populated by the chosen sensor setup, a Bluetooth communication link and a microcontroller
  • -
  • Development of a small catapult for repetitive launching of the PCB
  • -
  • Programming of the microcontroller in C to detect the highest point of the jump phase -
  • In flight experiments with high speed recording as reference
  • Type: Semester project
    Period: 16.03.2007 - 09.07.2007
    Section(s): EL IN ME MT
    Type of work: 40% hardware, 20% theory, 40% software
    Requirements:
    Subject(s): Aerial Robotics, Hybrid Locomotion
    Responsible(s): Mirko Kovac, Walter Karlen

    Evolving neural networks for foraging robots

    Yannick Weibel (MT)

    This project integrates itself in the study of evolution of cooperation and division of labor in the ANTS project. Artificial ants, implemented as micro-robots, are engaged in a foraging task, trying to deliver food items scattered throughout the arena to the nest. The foraging behavior of the Alice micro-robots used in these experiments is controlled by a feed forward neural network. Due to the limitations of the sensors and motors of the robot the neural network needed has to be robust. Also considering the memory constrains on the micro-robot, the number of neurons in the network must be low. The aim of this project is to understand the influence of different neural network topologies on the foraging performance of a colony of micro-robots. The detailed subtasks of the project are listed below 1. Reading on artificial evolution and neural networks. 2. Introduction to the Teem simulator used in the ANTS project and understanding the neural network code of this simulator. 3. Implementing different neural network topologies. 4. Performing evolutionary experiments with different neural network topologies and comparing the results. 5. Documentation including a project report and documentation of source code

    Type: Semester project
    Period: 12.03.2007 - 29.06.2007
    Section(s): MT
    Type of work: 60% coding in C++, 40% experiments
    Requirements:
    Subject(s): Collective Robotics, Artificial Evolution, Neural Networks
    Responsible(s): Danesh Tarapore, Markus Waibel
    URL: Click here

    2006


    Flying squid inspired passive wing folding mechanism

    Lydia Lostan (MT)

    A new biomimetic platform for flying miniature robots capable of terrestrial and aerial locomotion is currently in development at LIS. This system will possess wings that are foldable and some sort of running or jumping mechanism in order to move on ground, open the wings and fly. Such a combination of locomotion modes is widely used in the animal kingdom (e.g. gliding squirrels, gliding frogs, gliding snakes, gliding lizards, gliding ants, flying fish etc.), can lead to significant improvements of traveling distance per energy unit and enable a small robot to overcome large obstacles. The goal of this project is to implement a hardware design for a completely passive wing folding mechanism that is integrated with the existing jumping mechanism. The project can roughly be subdivided in the following work packages: - Definition of the requirements for the mechanism and its different functions - Systematical search and choice of the mechanical design - Implementation in CAD (ProE) - Fabrication of the components - Demonstration of the working prototype

    Type: Semester project
    Period: 11.02.2008 - 15.06.2008
    Section(s): EL ME MT MX
    Type of work: 30% hardware, 30% theory, 40% software
    Requirements:
    Subject(s): Aerial robotics
    Responsible(s): Mirko Kovac, James Roberts
    URL: Click here

    Interface robotique pour téléphones mobiles

    Sophie Mathis (MT)

    Le but de ce projet est d’implémenter une interface pour robots mobiles compatible avec une majorité de téléphones portables. Un couche logicielle en Java a déjà été développée dans notre laboratoire pour permettre la communication entre le microcontôleur du robot et un téléphone utilisant Bluetooth. Dans une première phase, il s'agira de passer en revue les possibilités des différentes classes de téléphones afin de réaliser un catalogue d'applications possibles. L'accent sera mis sur des applications éducatives ou récréatives. Dans une deuxième phase, une ou deux applications spécifiques seront implémentées. Leur fonctionnement sur différent téléphones mobiles sera testé et démontré.

    Type: Semester project
    Period: 19.09.2007 - 31.01.2008
    Section(s): IN MT SC
    Type of work: 90% software, 10% hardware
    Requirements: Java programming
    Subject(s): miniature robotics, Bluetooth communication
    Responsible(s): Jean-Christophe Zufferey, Julien Hubert

    Mouse optical sensors for flying robots

    Grégoire Salamin (MT)

    At the laboratory of intelligent systems (LIS), we are developing flying robots that rely on optic flow to navigate. Until now, the optic-flow detection was made using linear cameras and tiny microcontrollers (see for instance Klaptocz, 2005; Zufferey et. al. 2006). The goal of this project is to interface, test and characterize modern mouse optical sensors to see whether they could advantageously replace the modules used so far. The candidate will first review existing mouse sensors, chose one of these and build a demo board to interface it to a microcontroller, which will send data to a PC for monitoring and data-logging. A well-suited and reasonably lightweight optical system will be designed. Characterization experiments (which may involve in-flight tests either with an outdoor airplane or an indoor helicopter) must finally be implemented to conclude about the usability of the sensors for flying robots.

    Type: Semester project
    Period: 12.03.2007 - 28.07.2007
    Section(s): MT
    Type of work: 50% software, 50% hardware
    Requirements:
    Subject(s): Sensors, optic flow detection
    Responsible(s): Jean-Christophe Zufferey, James Roberts

    Interface USB pour télécommande

    Simon Wiedmer (MT)

    Le but de ce projet de semestre consiste en l’interfaçage d’une télécommande d’avion modèle réduit ultraléger à un ordinateur. Il s’agit de remplacer le microcontrôleur de la télécommande (modèle spécifique) par une version qui supporte le protocole USB et d’écrire le firmware nécessaire afin que la télécommande soit reconnue comme un joystick standard par n’importe quel ordinateur sous Windows, Mac ou Linux. Une demonstration de la fonctionnalité du système sera faite avec un simulateur de vol pour modèles réduits. Ce projet commence par une revue des microcontrôleurs (essentiellement Microchip) supportants USB et un compréhension du protocol implémentant un joystick standard via USB. Il s'agira ensuite de développer un nouveau PCB en tenant compte des contraintes physiques de la télécommande existante. Finalement, le firmware doit être implémenté et testé. Si le temps le permet, une option Bluetooth sera étudiée.

    Type: Semester project
    Period: 12.03.2007 - 28.07.2007
    Section(s): IN MT SC
    Type of work: 50% hardware, 50% software
    Requirements:
    Subject(s): microcontroller programming
    Responsible(s): Jean-Christophe Zufferey, Jean-Daniel Nicoud

    Jumping Mechanism for Flying Robots

    Martin Fuchs (MT)

    A new biomimetic platform for flying miniature robots capable of terrestrial and aerial locomotion is currently in development at LIS. This system will possess wings that are foldable and some sort of running or jumping mechanism in order to move on ground, open the wings and fly.

    The goal of this semester thesis is to explore small jumping systems and summarize the key ingredients for a successful jumping robot. Design solutions shall be examined and one jumping principle realized. The work includes the following work packages:

      -
    • Summary of different jumping mechanisms in robots and in animals
    • -
    • Given a designs for the jumping mechanism, model the forces acting on the system and propose a dimensioning based on the calculations
    • -
    • Implement one solution in a CAD program (ProE)
    • -
    • Fabricate the mechanism and demonstrate its behavior
    • Type: Semester project
      Period: 16.03.2007 - 09.07.2007
      Section(s): EL IN ME MX PH
      Type of work: 30% théorie, 30% software, 40% hardware
      Requirements:
      Subject(s): Aerial Robotics, Evolutionary Robotics, Jumping Robots
      Responsible(s): Mirko Kovac, James Roberts

    Wing Folding Mechanism for Flying Microrobots

    Grégory Savioz (MT)

    A new biomimetic platform for flying miniature robots capable of terrestrial and aerial locomotion is currently in development at LIS. This system will possess wings that are foldable and some sort of running or jumping mechanism in order to move on ground, open the wings and fly.

    The goal of this semester thesis is to implement a hardware design for the wing folding mechanism in a Computer Aided Design (CAD) program. The work includes the following work packages:

    -
  • Summarize the wing folding principles found in the animal kingdom.
  • -
  • Given a designs for the wing folding mechanism, model the forces acting on the system and propose a dimensioning based on the calculations
  • -
  • Implement one solution in a CAD program (ProE)
  • -
  • Fabricate the mechanism and demonstrate its behavior
  • Type: Semester project
    Period: 16.03.2007 - 09.07.2007
    Section(s): MA ME MT
    Type of work: 20% hardware, 20% theory, 60% software
    Requirements:
    Subject(s): aerial robotics
    Responsible(s): Mirko Kovac, James Roberts

    Evolving efficient swarm controllers: Investigation of optimal automatic design of heterogeneous robot teams

    Martijn Bosch (MT)

    Automated synthesis of robot swarm controllers is a promising method to create large distributed teams of agents, that surpass current systems in scalability, robustness and cost-effectiveness. Most current research in swarm robotics is limited to homogeneous agent teams, foregoing the potential benefits of efficient search and task specialisation. Our first investigations of the performance of heterogeneous robot teams uncovered two significant problems: (1) Comparatively low performance in heterogeneous colonies due to inefficient selection. (2) Comparatively slower convergence on optimal solutions in homogeneous colonies due to reduced diversity (n times fewer genomes are being tested in homogeneous colonies, where n is the team size; cmp. Bull and Holland 1997). This master project will develop and compare methods to automatically design controllers for groups of morphologically similar agents. It will investigate the impact of group composition, fitness evaluation methods and the influence of group/genepool size on the performance of evolved groups. Work can be based on an existing physics-based simulation of a multi-agent system.

    Type: Master project
    Period: 01.05.2006 - 01.05.2007
    Section(s): EL IN MT MX PH
    Type of work: 50% software, 50% research
    Requirements: good knowledge of C++
    Subject(s): Evolution of heterogeneous robot teams
    Responsible(s): Markus Waibel, Claudio Mattiussi
     

    Biologically Inspired Swarms of MAVs for Radio Communication Relay

    Laurent Winkler (MT)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two or more ground emitters (base station and user stations) is currently being developed in our lab. The MAV controllers rely on information obtained through local communication with neighboring MAVs and internal sensory information. Also no positioning information is present in the system (i.e no GPS or positioning sensors). In order to overcome these fairly restrictive constraints, inspiration can be taken from the deployment of Army Ant Raid Patterns. Such biological systems make use of stigmery (i.e communication through the environment), which is difficult to achieve with aerial systems. The Master student should start by an overview of aerial and ground robotic swarms that take inspiration from ant systems as well as an overview of "Army Ant Raid Pattern" mechanisms. Interesting algorithms should then be adapted to the generation and maintenance of communication networks established in 2D simulation using a swarm of MAVs. Three phases are considered, the deployment of the swarm of MAVs and the search for one or more user stations, the maintenance of the communication network when established and the retraction of the swarm to the base station. The main challenge will be to imagine an algorithm where no positioning information is needed. The student should determine the limitations, robustness and efficiency of such an approach when simple MAV dynamics and communication are taken into account. If time permits, simulations can be done in 3D.

    Type: Master project
    Period: 23.10.2006 - 28.03.2007
    Section(s): IN MA MT PH SC
    Type of work: 30% théorie, 70% software
    Requirements: C++Programming
    Subject(s): Micro Air Vehicles, Army Ant Raid Patterns, Stigmergy, Aerial swarms, Bio-inspired Algorithms
    Responsible(s): Sabine Hauert, Sara Mitri
    URL: Click here

    GPS Module for trajectory plotting and tracking of MAVs

    Hervé Péter-Contesse (MT)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two ground emitters (base station and user station) is currently being developed in our lab. Each swarm member will consist of a flying platform, a main processing unit and a set of various sensors and actuators. A radio link will be employed to transmit flight data to a ground station. The task of this project is to create a system that provides GPS data (position, speed, time) for an existing MAV electronics board, when the latter asks for this information. The long term goal is to track and analyze the MAV's flight by sending the GPS data to a ground station. For the design, a focus should be made on small form factor, light weight, small power consumption, accuracy and reliability. To start with, some GPS components are already available at the LIS. The student can nonetheless propose other components if this seems promising or even necessary. A special issue is the antenna, where active and passive antenna solutions should be evaluated, especially in flight. Also, it should be investigated whether a DGPS (differential GPS) solution is feasible for the project and an overview of the characteristics (performance, advantages and inconveniences) should be given. If possible, a DGPS solution should be implemented. Finally, the complete system will be tested on the ground and during flight. An evaluation of its performance and a complete description of its characteristics (e.g. GPS accuracy and cases with sensing problems) conclude the project.

    Type: Semester project
    Period: 23.10.2006 - 28.02.2007
    Section(s): EL IN ME MT
    Type of work: 70% software, 30% hardware
    Requirements: General knowledge of microcontrollers, sensors and programming
    Subject(s): Micro Air Vehicles, microcontrollers, sensors, interfaces
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here

    Altitude and speed control of MAVs

    Vitalis Hirschmann (MT)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two ground emitters (base station and user station) is currently being developed in our lab. Each swarm member will consist of a flying platform, a main processing unit and a set of various sensors and actuators. A radio link will be employed to transmit flight data to a ground station. The goal of this semester project is to develop a module, which measures the flight altitude and the airplanes speed relative to the surrounding air. As sensors, pressure sensors can be used, but the student can suggest other promising technologies, as far as the sensors are available on the market. Furthermore, the student should provide a mechanism for keeping airspeed and altitude constant as a function of imposed input values. The sensor module's output will be connected to an existing control electronics board. Within this module, control laws "altitude hold" and "speed hold" should be implemented in software to actuate the aerodynamic control surfaces. An interface between the sensor board and the existing controller board has obviously to be created, and appropriate software must be written. Finally, the created system's functionality and characteristics (e.g. behaviour for different temperatures and transduction equations) must be described completely. For doing so, the module has to be evaluated on ground and during flight, where the aircraft's direction will be controlled via other means.

    Type: Semester project
    Period: 23.10.2006 - 28.02.2007
    Section(s): MT
    Type of work: 30% theory, 30% software, 40% hardware
    Requirements: Electronics and microcontroller basics
    Subject(s): Electronics development, programming and testing
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here
     

    Turn control system for MAVs

    Pascal Gilbert (MT)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two ground emitters (base station and user station) is currently being developed in our lab. Each swarm member will consist of a flying platform, a main processing unit and a set of various sensors and actuators. A radio link will be employed to transmit flight data to a ground station. This semester project focuses on the development of a light, small and cheap flight control system which suits the project's needs, that is, a system for turn coordination. It takes as input a demanded value for the turn rate and, with supposedly constant altitude, flies the aircraft with this turn rate. The task comprises the development of a sensor board which will be connected to already existing flight controller electronics. Off-the-shelf sensors should be used, and sensors other than gyroscopes and accelerometers (e.g. magnetometers) should be explored in the first line. On the controller board, software has to be written to produce suitable commands for the aerodynamic control surfaces from the flight state values provided by the sensor board. Finally, the complete system must be tested on the ground and in flight. An evaluation of its performance and a complete description of its characteristics shall conclude the project.

    Type: Semester project
    Period: 23.10.2006 - 28.02.2007
    Section(s): MT
    Type of work: 30% theory, 30% software, 40% hardware
    Requirements: Electronics and Microcontroller basics
    Subject(s): Electronics development, programming and testing
    Responsible(s): Severin Leven, Sabine Hauert
    URL: Click here
     

    Modularity and Community Structure in Complex Networks

    Thomas Schaffter (MT)

    Recently there has been a lot of interest in complex networks. Examples are the internet, the worldwide web, biochemical networks, food webs, neural networks, social networks, etc. It has been found that many complex networks share surprising similarities with respect to structural properties of their architecture.
    A particularly interesting property of many networks is modularity, also called community structure. The above mentioned networks can be grouped into modules / communities. Nodes within a community are densely connected together and have much fewer connections that reach outside the community. Analyzing the community structure of a complex network is highly interesting - for example, one can detect functional modules in a biochemical network or communities in a social network.
    Identification of the community structure of a network amounts to finding an optimal partition of the network into subgroups. This is a very difficult problem as the complexity grows exponential with the network size. Many algorithms have been proposed in the literature, but none guarantees finding the optimal solution.

    Type: Semester project
    Period: 23.10.2006 - 25.02.2007
    Section(s): EL IN MA PH SC
    Type of work: 30% research, 70% software
    Requirements: C / C++
    Subject(s): complex networks, modularity
    Responsible(s): Daniel Marbach, Claudio Mattiussi

    Sensor Eyewear for fatigue detection

    Samuel Hauser (MT)

    A possibility to detect the fatigue of a person is to monitor the movement of his eyelids. Usually eye and eyelid movements are detected by complex tracking systems which are unsuitable for the use on the Solar Impulse plane. This semester project focuses on building a simple device which can detect reliably the eyelid closure. The system has to be robust against influences from ambient light and strong sun. Low power consumption and lightweight is a must. A miniature microcontroller card with Bluetooth connection is available for the signal processing. Some preprocessing of the sensor data should be implemented on the microcontroller. The student will explore the different reflection properties of the eye and the eyelid. An analysis of existing systems and patents will be useful. The student will also propose a simple algorithm for the sleep onset detection and eventually for fatigue prediction and implement it to the new developed device. Finally, the complete system will be tested. An evaluation of its performance and a complete description of its characteristics conclude the project.

    Type: Semester project
    Period: 23.10.2006 - 25.02.2007
    Section(s): MT
    Type of work: 20% Theory 50% Hardware 30% Software
    Requirements:
    Subject(s): PIC microcotroller, eye movement detection, IR sensor
    Responsible(s): Walter Karlen, Mirko Kovac

    Optic-flow-based navigation with an e-puck robot

    Pablo Tueta (MT)

    The goal of this project is to develop efficient optic-flow-based control strategies for the e-puck robot (www.e-puck.org) as a showcase of the technology the lab is developing for aerial robots. The candidate will chose different approaches (hand-design, learning, genetic algorithms, etc.) to produce efficient control strategies for an e-puck robot in different kinds of environments (e.g., wide rooms of different shapes, corridors, mazes). Special care will be made on adaptation to background light and metrics to evaluate and compare the obtained behaviors. At the end, a nice demo setup should be put together allowing for easy demonstrations even when operated by unskilled people.

    Type: Semester project
    Period: 23.10.2006 - 25.02.2007
    Section(s): MT
    Type of work: 60% software, 40% hardware
    Requirements:
    Subject(s): Mobile robotics
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
     

    Sunburn detector

    Pierre Jaquier (MT)

    Skin cancer is a severe health problem in developed countries. Especially people exposed to high UV radiation (mountaineers, construction workers…) have high risk to get a melanoma. Many UV radiation measurement devices exist, but they lack of intelligence. The student will develop an electronic sunburn warning device based on UV radiation measures. Tiny sensors mounted on a PIC microcontroller board will help to evaluate the risk for the human skin. The sensor system should process intelligently the information collected eg. automatically adapt to the wearers skin type, time of day, season etc. An analysis of existing systems and patents will be useful. The system has to be lightweight and low power consuming. The student has also to find the optimal location for the device. Finally, the complete system will be tested. An evaluation of its performance and a complete description of its characteristics conclude the project.

    Type: Semester project
    Period: 23.10.2006 - 25.02.2007
    Section(s): MT
    Type of work: 30% Theorie 50% Hardware 20% Software & Experimentation
    Requirements:
    Subject(s): Senors, UV detection, microcontrollers
    Responsible(s): Walter Karlen, Mirko Kovac

    Mice optical sensors for flying robots

    Matthias Tschudi (MT)

    At the laboratory of intelligent systems (LIS), we are developing flying robots that rely on optic flow to navigate (http://lis.epfl.ch/microflyers). Until now, the optic-flow detection was made using linear cameras and tiny microcontrollers (Klaptocz, 2005; Zufferey et. al. 2006). The goal of this project is to interface, test and characterize modern mice optical sensors to see whether they could advantageously replace the modules used so far. The candidate will first review existing mice sensors, chose one of those and build a demo board to interface it to a microcontroller, which will send data to a PC for monitoring and data-logging. A well-suited and reasonably lightweight optical system will be designed. Characterization experiments (which may involve in-flight tests) must finally be implemented to conclude about the usability of the sensors for flying robots.

    Type: Semester project
    Period: 23.10.2006 - 25.02.2007
    Section(s): MT
    Type of work: 60% software, 40% hardware
    Requirements:
    Subject(s): Sensors, optic flow detection
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
     

    In Silico Evolution of Biochemical Networks

    Fanny Riedo (MT)

    This project aims at studying the evolution of biochemical networks in silico (i.e. using artificial evolution and computer simulation). In the first weeks, we will choose a simple problem, for example evolution of chemotaxis (the ability of bacteria to swim up food gradients). Chemotaxis in bacteria is controlled by a simple biochemical network that is well studied. Using a minimal simulation of a 'virtual bacteria' and an appropriate evolutionary algorithm, we evolve networks that mediate efficient chemotaxis and compare these networks with the real one.

    Depending on the progress of the project, we can investigate interesting questions in evolution of biochemical networks. For example, chemotaxis of bacteria is extremely robust. By using different setups (e.g. a static or a changing environment of the bacteria), we can study under which conditions a robust network evolves. Other interesting questions are the evolution of modular networks or the impact of the genetic encoding on the network structure.

    Type: Master project
    Period: 23.10.2006 - 23.02.2007
    Section(s): MT
    Type of work: 35% research; 65% software
    Requirements: Evolutionary algorithms
    Subject(s): Artificial evolution
    Responsible(s): Daniel Marbach, Claudio Mattiussi
     

    Human Sleep detection with miniature movement sensor

    Iván Muñoz (IN)

    Actigraphy (recording of body parts acceleration) is often used for context awareness and long term sleep monitoring outside hospitals. We want to explore the possible use of accelerometers for the state detection of the pilot on the Solar Impulse plane. The aim of this project is to find good algorithms to detect sleep onset of the pilot with the help of accelerometer information. The task is to explore different known algorithms for sleep detection and modify them for the use with one of our own developed PIC microcontroller cards. Together with a well chosen MEMS accelerometer this card will then be used for real time experiments. Ideally the processing is done on the microcontroller. However, for complex calculation a Bluetooth connection to a master PC is also possible. The optimal location on the human body has also to be found by the student. After, the system has to be adapted for a use on planes, where external vibrations are more important than on the ground. Finally, the student evaluates the performance of the different configurations and gives a complete description of the characteristics of the system.

    Type: Semester project
    Period: 23.10.2006 - 20.02.2007
    Section(s): EL IN MT
    Type of work: 30% Theorie 20% Hardware 50% Software & Experimentation
    Requirements: C
    Subject(s): Accelerometry, PIC programming, Sleep onset detection
    Responsible(s): Walter Karlen, Mirko Kovac

    Controller design for the homing of swarms of MAVs

    Alexandre Habersaat (IN)

    A swarm of micro air vehicles (MAVs) capable of establishing a communication relay between two ground emitters (base station and user station) is currently being developed in our lab. Once the communication between the base and user station finished, a homing mechanism must ensure the safe return of the MAVs to the base station.
    This semester project aims at implementing in simulation a controller capable of retracting a previously deployed swarm of connected MAVs to a given base station. The MAV controllers rely on information obtained through local communication with neighboring MAVs and internal sensory information. To add to the challenge, no positioning information is present in the system (i.e no GPS or positioning sensors).
    Robot controllers will be either hand designed or evolved within the TEEM framework and 2D ENKI viewer. Interesting controllers should be studied in terms of efficiency and reliability. The initial requirements necessary for a successful retraction of the swarm can also be addressed.

    Type: Semester project
    Period: 23.10.2006 - 20.02.2007
    Section(s): IN MA MT PH SC
    Type of work: 30% théorie, 70%software
    Requirements: C++Programming
    Subject(s): Genetic Algorithms, Neural Networks, Micro Air Vehicles, Controller Design
    Responsible(s): Sabine Hauert, Julien Hubert
    URL: Click here

    Bluetooth and control software for a miniature robot

    Alex Vulliemin (MT)

    The goal of this project is to implement an efficient way to interact with an existing didactic miniature robot developed by DIDEL. The robot user - a student or a player - is in front of a Bluetooth-enabled PC or pocketPC. He must be able to steer the robot in real-time, display the status of the sensors, and download new programs into the robot processor. More specifically, this project consists of developing the required extension board with a small microcontroller and Bluetooth module that are already used on other robots in our lab, interface the new board with the underlying robot processor, and implement software for the robot as well as for the controlling PC and/or pocketPC. Beyond these technical developments, the candidate will prepare a few demonstrations showing some abilities of this miniature robot equipped with the new communication board and software.

    Type: Semester project
    Period: 13.03.2006 - 15.07.2006
    Section(s): EL IN MT SC
    Type of work: 10% literature reading, 40% hardware, 50% software
    Requirements: C-programming, hardware
    Subject(s): miniature robotics
    Responsible(s): Jean-Christophe Zufferey, Jean-Daniel Nicoud

    2005


    Modelization of wireless communication for swarms of MAVs

    Daniel Horacio Arze Pando (SC)

    We aim at designing a swarm of Micro Air Vehcles (SMAVs) capable of autonomously establishing emergency wireless networks (SMAVNETs) between multiple ground-users in a disaster area. The SMAVNET is required to be rapidly deployable in any environment following the paradigms of swarm robotics. MAVs are designed to be minimal (low-cost, light-weight, simple electronics) and do not use position sensors (cameras, GPS), which are dependent on the environment, expensive in terms of energy, cost, size and weight or unusable at large ranges. Rather than that, agents rely on local communication with immediate neighbors and proprioceptive sensors which provide heading, speed, altitude and angular velocities.

    This semester project will be concerned with the, characterization of the wireless hardware developed for the MAVs in terms of bandwidth, communication range, reconnection times and multi-hop routing protocols. Initial tests will be conducted on hardware setups with 2 to 10 wireless nodes on ground. Parameters derived from these tests will then be implemented in our 3D simulator, which will be used to asses more complex scenarios and possibly identify bottlenecks in the network.

    Type: Semester project
    Period: 01.02.2008 - 01.07.2008
    Section(s): EL IN MA MT PH SC
    Type of work: 33% théorie, 33% software, 33% testing
    Requirements: C++ programming
    Subject(s): Wireless Communication, Network Protocols. Modelization, Micro Air Vehicles
    Responsible(s): Sabine Hauert, Severin Leven
    URL: Click here

    Miniature multi-1D camera for indoor microflyers

    David Jeanbourquin (EL)

    This project is about designing and building an ultralight (about 1g) vision system to fit the requirements of our indoor microflyers (http://lis.epfl.ch/microflyers). It will be based on several 1D sensors placed on a 3D arrangement of printed circuit board in order to optimize the estimation of radial optic flow at 45° off the flying direction. The prototype must be interfaced to an existing microcontroller in order to be tested and characterized.

    Type: Semester project
    Period: 12.03.2007 - 28.07.2007
    Section(s): EL MT
    Type of work: 50% hardware, 20% software, 30% experiments
    Requirements:
    Subject(s): Miniature camera design
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here

    Evolution of Specialisation

    Sarah Marthe (IN)

    Social insects have long been a great source of inspiration to engineers. Their success in solving problems is illustrated by the fact that they outnumber us a million to one, and the total weight of ants matches that of the entire human race. One key aspect that has led to this success is their ability to cooperate and perform efficient division of labour - which are also a key challenges in designing multi-robot systems.
    Different social insects have found surprisingly different solutions to optimal task allocation. While some insect species have evolved specialists for different tasks (soldiers with large jaws, foragers that can carry 5 times their own body weight or living honey pots), others, like bees, have developed an age-dependant task distribution, with the youngest colony members working on brood care while the oldest work as soldiers. The reasons for this are largely unknown.
    A fundamental design choice for multi-robot groups or robot swarms is whether group members should be identical or whether to combine two or more different types of specialist robots.
    This research Master project will address challenges in collective robotics by drawing on parallels to the biology of social insects. To determine when specialisation pays and under what conditions specialists are outperformed by generalists, we will use an agent based model to simulate various conditions (task count, environmental fluctuations, group size, noise). Successful strategies for groups of agents will be selected using artificial evolution.

    Type: Master project
    Period: 21.08.2006 - 15.12.2006
    Section(s): EL IN MT PH
    Type of work: 50% software, 50% research
    Requirements:
    Subject(s): biological modelling
    Responsible(s): Markus Waibel, Danesh Tarapore
    URL: Click here

    Generation of Biologically Plausible GRN Inference Benchmarks

    Nadir Hadbi (IN)

    Reverse engineering and simulation of biochemical networks may contribute substantially to our biological knowledge in the post-genomic era. The fast progress in high-throughput experimental technology is boosting research in reverse engineering of biochemical networks, especially genetic regulatory networks (GRNs). The goal of this project is developing an open-source software tool for the automatic generation of appropriate benchmarks for GRN reverse engineering algorithms. Besides from reviewing the state-of-the-art in GRN simulation and reverse engineering, the student will also have to study biological literature to acquire the necessary basic knowledge of genomics. The focus of this project lies on the generation of biologically plausible artificial GRNs that can be used as benchmarks for reverse engineering algorithms. The student is expected to research methods for: (1) Meaningful sampling of known GRNs from model organisms to generate plausible topologies; (2) Generation of plausible parameters values for the GRN model; (3) The realistic simulation of experimental data from the generated models. From a practical point of view, the expected outcome of the project is a cross-platform C++ library for the automatic generation of GRN inference benchmarks. Emphasis is put on usability, flexibility and extensibility through adequate object-oriented design.

    Type: Master project
    Period: 03.04.2006 - 28.07.2006
    Section(s): IN SC
    Type of work: 70% software, 30% research
    Requirements: Object oriented software design, C++
    Subject(s): software architecture design, biocomputing
    Responsible(s): Daniel Marbach, Claudio Mattiussi

    Capteur de distance ultrason ultraléger

    Gavrilo Bosovic (MT)

    Le capteur de distance ultrason le plus léger actuellement sur le marché pèse un peu plus de 3g (cf. http://zuff.info/RangeFindersComp_E.html). L'objectif de ce projet est de développer, à partir de l'existant, un capteur 3 fois plus léger pour qu'il puisse être utilisé sur un robot volant de 10g (cf http://microflyers.zuff.info).

    Type: Semester project
    Period: 13.03.2006 - 15.07.2006
    Section(s): EL MT
    Type of work: 50% hardware, 30% software, 20% testing
    Requirements:
    Subject(s): Distance sensors
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here

    Miniature aVLSI camera for indoor microflyers

    Tobias Greuter (EL)

    This project is about designing and building an ultra-light camera (less than 1g) using a special vision chip, which has been specifically designed at INI-ETHZ to fit the requirements of our indoor microflyers. It is using the aVLSI technique and has adaptive photoreceptors, which can quickly adapt to changing background light. Such custom vision sensors are shipped without any packaging (raw dies) and must be bonded onto a printed circuit board (PCB), which is to be developed by the candidate. Optics will be design based on existing solutions. The prototype camera will be interfaced with a microcontroller in order to be fully tested and characterized.

    Type: Semester project
    Period: 13.03.2006 - 15.07.2006
    Section(s): EL MT
    Type of work: 50% hardware, 20% software, 30% experiments
    Requirements:
    Subject(s): Miniature camera design
    Responsible(s): Jean-Christophe Zufferey, André Guignard

    Visual Tracking of Indoor Flying Robots

    Julien Reuse (MT)

    At our lab we are working on bio-inspired, ultra-light indoor flying robots that use optic-flow to navigate. We just set up a new virtual-reality experimental room equipped with a ceiling, fish-eye camera for testing those flying robots under different visual conditions. The goal of this master project is to develop a software using the ceiling camera for tracking a flying robot while it is freely manoeuvring within the experimental room. The first step would be to retrieve the 2D trajectory, but further developments should lead to an estimate of all the 6 degrees of freedom (DoF). This project shall start with a review of existing software that could ease de development of the final system. In particular, the real-time 3D tracking and detection system from the CVLab-EPFL will be evaluated. The project will then proceed with the development of the tracking software while taking into consideration the constraints related to this particular camera (geometry, type of connection) as well as real-time image processing. A precise characterization of the system (precision for all 6 DoF, update rate, etc.) will be carried out at the end.

    Type: Master project
    Period: 23.10.2005 - 22.03.2006
    Section(s): IN MT
    Type of work: 20% theory, 50% software, 30% experiments
    Requirements: C++, image processing
    Subject(s): visual tracking
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here
     

    Simulation of Swarming MAVs for Communication Relay

    Sabine Hauert (IN)

    This master project is part of a long-term research project aiming at developing a fleet of simple Micro Air Vehicles (MAV) for use as RF communication relay in disaster areas. Each MAV should navigate autonomously based on swarm intelligence, using distributed control and onboard sensors. The primary goal of this aerial swarm will be to establish a robust and low-power communication link between two or more emitters located on ground. In this master project, the candidate will first review the state-of-the-art in swarm intelligence, collective aerial robotics, biological models of social behaviours of interest, and simulation frameworks for collective robotics. She/he will then propose at least two approaches to design distributed controllers capable of producing global behaviours required for the three following phases: (1) deployment and target finding, (2) network maintenance, (3) retraction. Those controllers will be first tested, analysed and compared in an existing minimalist 2D simulation framework. Then, depending on time left, the complexity of the simulation will be progressively increased by incorporation, e.g., modelization of wind, random crashes of MAVs, dynamic ground emitters, more realistic communication, more realistic sensors, 3D MAVs, etc.

    Type: Master project
    Period: 23.10.2005 - 22.03.2006
    Section(s): IN
    Type of work: 40% theory, 30% software, 30% analysis
    Requirements:
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Markus Waibel
     

    Implementation and test of a new algorithm for optical flow estimation

    Samuel Gmehlin (IN)

    When an object moves in the field of view of a grayscale camera it generates an optical flow, that is, a “flow” of grayscale values on the plane of the camera sensor. One of the classical problems in visual motion computation is the estimation of the optical flow from the sequence of frames captured by the camera. The goal of this project is to implement and test a new algorithm for optical flow estimation, and compare the results with those of state-of-the-art optical flow estimation algorithms.

    Type: Semester project
    Period: 07.10.2005 - 07.01.2006
    Section(s): EL IN MT
    Type of work: 30% theory, 40% software, 30% testing
    Requirements: C/C
    Subject(s): Machine Vision
    Responsible(s): Claudio Mattiussi, Jean-Christophe Zufferey

    Influence of parameter encoding in the artificial evolution of neural networks

    Fanny Riedo (MT)

    Artificial neural networks are information processing systems inspired by the architecture of biological neural systems. Evolutionary algorithms can be used to assign the values of the parameters that determine the behavior of an artificial neural network. This implies the genetic encoding of the parameters of the neural network. The goal of this project is the implementation and testing of a non-conventional genetic encoding of neural network parameters. The candidate will implement both the conventional and non-conventional genetic encoding and run a series of evolutionary experiments to compare the performance of the two kinds of encoding when applied to some neural network benchmark problems.

    Type: Semester project
    Period: 07.10.2005 - 07.01.2006
    Section(s): EL IN MT
    Type of work: 100% software
    Requirements: C/C++/Matlab programming
    Subject(s): Neural networks, Genetic algorithms
    Responsible(s): Claudio Mattiussi, Antoine Beyeler

    Intelligent vision module for Khepera robot

    Davis Daidie (MT)

    An existing 2D color vision module exists for the Khepera robot. The goal of this project is to update the module with a dsPIC (powerful new microcontrollers from Microchip) and adapt the existing code to drive the camera. Some vision algorithms will be implemented on the module, and made available for use by the Khepera robot.

    Type: Semester project
    Period: 07.03.2005 - 01.07.2005
    Section(s): EL IN MT
    Type of work: 60% electronic design 40% embedded software development
    Requirements: Digital Eletronic, C Programming
    Subject(s): Electronic, Vision
    Responsible(s): Antoine Beyeler, Francesco Mondada

    Optic-flow-based navigation for an ultra-light indoor aircraft

    Laurent Gillet (MT)

    Within the Bio-inspired Vision-based Flying Robots project, a 30-gram slow-flyer was developed to avoid collisions like a fly, using optic-flow. This semester project is about corridor following. The candidate will first assess some optic-flow-based control strategies on a small wheeled robot before implementing the best ones on an ultra-light airplane and demonstrating its ability to fly autonomously down a corridor.

    Type: Semester project
    Period: 07.03.2005 - 27.06.2005
    Section(s): ME MT PH
    Type of work: 50% software dev., 50% experiments
    Requirements: C programming
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here

    Automation of x/y robot with camera tracking system

    Frédéric Magnard (MT), Alexis Reust ()

    This semester project integrates inself in the study of evolution of cooperation and division of labour in the ANTS project. Artificial ants, implemented as microrobots, are engaged in a foraging task, trying to deliver food items scattered throughout the arena to the nest. An overhead tracking camera is used to determine the positions of all objects in the arena.
    The goal of the first part of this project is the design and construction of an automated x-y robot that can be used to reposition collected food items to random starting positions and integrate it with an existing online tracking system. The work on the final robot for the 100x100cm arena can be based on a prototype built for a smaller, 50x50cm arena.
    The second part consists of integrating this x-y robot with an existing tracking software and of implementing the software required to run evolutionary experiment up to the microrobot's maximum power autonomy (~ 20h).

    The final goal of this project is a fully automatic, robust experimental arena for evolutionary experiments with swarms of up to 20 Alice microrobots.

    Type: Semester project
    Period: 07.03.2005 - 27.06.2005
    Section(s): EL IN MA ME MT PH
    Type of work: 30% design, 60% hardware, 10% testing and 30% design, 60% software, 10% testing respectively
    Requirements: basic English communication skills
    Subject(s): robotics, mechanics, software
    Responsible(s): Markus Waibel, Danesh Tarapore, Gilles Caprari
    URL: Click here

    Optic-flow-based altitude control for an ultra-light indoor aircraft

    Laurent Bernau (MT)

    Within the Bio-inspired Vision-based Flying Robots project, a 30-gram slow-flyer was developed to avoid collisions like a fly, using optic-flow. This semester project is about altitude control. The candidate will first assess some optic-flow-based control strategies on a small wheeled robot before implementing the best ones on an ultra-light airplane and demonstrating its ability to automatically regulated its altitude.

    Type: Semester project
    Period: 07.03.2005 - 27.06.2005
    Section(s): MT
    Type of work: 50% software dev., 50% experiments
    Requirements: C programming
    Subject(s): Aerial robotics
    Responsible(s): Jean-Christophe Zufferey, Antoine Beyeler
    URL: Click here
     

    The influence of environmental pressure on competition

    Dominique Zosso (EL)

    All naturally evolving systems are faced with a tradeoff between allocating resources to development versus competition. Examples include a company's choice in focusing on manufacturing vs. advertising, a nation's expenditures for military spending vs. its investments for development or the shift of effort from growth and reproduction to sexual competition in all living organisms. Our hypothesis is that a main factor in determining a system's relative investment in competition is the amount of environmental pressure on its agents. Monopolistic market segments in economics, stable political environments as well as animals way up there in the food chain are expected have higher investment into development vs. competition. This project consists of building a minimalist agent based computer model to investigate the influence of environmental pressure on the amount of resource allocation on development vs. competition.

    Type: Semester project
    Period: 07.03.2005 - 27.06.2005
    Section(s): EL IN MA MT PH
    Type of work: 33% théorie, 33% software, 33% analysis
    Requirements: good coding skills
    Subject(s): system modelling
    Responsible(s): Markus Waibel, Danesh Tarapore

    Caméras miniatures pour robots volants ultra-légers

    Adam Klaptocz (MT)

    Dans le cadre du développement de nos avions d’intérieurs bio-inspirés, nous avons besoin de caméras de plus en plus petites et légères. Le but de ce projet consiste en la mise au points d’une caméra miniature pesant moins de 0,5g (l’actuelle pèse 1g). Pour la partie capteur, l’étudiant utilisera soit des chips à basse résolution disponibles dans le commerce (par exemple CMOS 1D), soit des capteurs avec prétraitement d’image intégré (aVLSI) développés sur mesure par l’INI. Le design du packaging ainsi que de l’optique représentera une partie importante du développement. Finalement, la caméra sera interfacée à un microcontrôleur (existant) afin de pouvoir être caractérisée soigneusement.

    Type: Semester project
    Period: 07.03.2005 - 27.06.2005
    Section(s): EL MT
    Type of work: 60% hardware, 20% programmation, 20% experimentation
    Requirements: optique, programmation C, électronique
    Subject(s): vision
    Responsible(s): Jean-Christophe Zufferey, André Guignard
    URL: Click here

    2004


    Active Vision and Independent Visual Feature Extraction

    Sylvain Quartier (IN)

    In neurobiology it has been found that the positions (on the scale of receptive field sizes) of contours change much faster than their orientation. In a simulated neural network this property, in combination with a recently proposed learning rule, allows learning translation-invariant receptive fields. This matches properties of complex neurons found in primary visual cortex (Hubel and Wiesel 1962), suggesting an unsupervised learning algorithm to provide a factorial code of independent visual features. It is still unexplored whether such invariant features can be mapped to the motor outputs successfully in mobile robot controller. The goal of this project is to implement the learning algorithm in the robot controller, to evolve the robot in simulation, and to analyze the behavior and resulting receptive fields.

    Type: Semester project
    Period: Winter 2004/2005
    Section(s): EL IN MT SC
    Type of work: 20% theory, 50% software, 30% simulation and analysis
    Requirements: C/C++, neural network, PCA (and ICA)
    Subject(s): active vision, online learning, receptive field, ICA
    Responsible(s): Mototaka Suzuki, Markus Waibel
    URL: Click here

    Active Vision with Omnidirectional Camera

    Antoine Béguin (EL)

    Omnidirectional camera has 360 degrees of field of view. Its application to Active Vision System has a lot of potentials. First, evolved artificial retina is potentially capable of selecting important features in such a broad field of view without mechanical camera control (pan/tilt). Additionally, active vision control can dispense a computationally expensive step, the “unwrapping process”. The goal of this project is to explore the usability of the omnidirectional camera for active vision system. First goal (hardware part) is to setup the connection between the camera and CPU. Then implement the active vision system on the basis of the software which has already been developed in our lab.

    Type: Semester project
    Period: Winter 2004/2005
    Section(s): EL IN MT SC
    Type of work: 50% hardware, 50% software
    Requirements: CCD (and CMOS) camera, C/C++
    Subject(s): omnidirectional camera, active vision
    Responsible(s): Mototaka Suzuki, Dario Floreano
    URL: Click here

    Télécommande Bluetooth

    Alexandre Faudot (EL)

    Le but de ce projet consiste en l'implémentation d'une télécommande de dirigeable (taille ~150cm, voir cette page) à partir d'un téléphone mobile avec Bluetooth. Le téléphone sera de type Nokia 3650 fonctionnant avec le système d'exploitation Symbian. Le dirigeable étant équipé d'une caméra, il faudra envisager une communication bi-directionnelle permettant aussi de transmettre et d'afficher le flux vidéo sur l'écran du téléphone en temps réel. Le dirigeable est déjà équipé d'un microcontrôleur 8-bit (PIC) ainsi que d'un module Bluetooth. Le projet inclue aussi l'amélioration du driver Bluetooth pour le PIC.

    Type: Semester project
    Period: 18.10.2004 - 04.02.2005
    Section(s): IN MT SC
    Type of work: 90% programmation, 10% hardware
    Requirements: programmation C/C++ pour systèmes embarqués
    Subject(s): Bluetooth, Symbian OS, PIC
    Responsible(s): Jean-Christophe Zufferey, Frédéric Pont

    Simulateur, évolution et émergence du language chez les s-bot simulés

    Vincent Porchet (IN)

    Le but est de créer un simulateur simple mais très rapide, gérant une arène circulaire en 2d. Les robots (s-bots) sont représentés par des cercles. Les capteurs sont simulés par ray-tracing en intersectant les cercles par des droites. Une expérience validera le simulateur. Elle consistera à tester l'émergence d'un proto-language entre les s-bots simulés destiné à optimiser la recherche de nourriture.

    Type: Semester project
    Period: Summer 2004
    Section(s): IN
    Type of work: 20% théorie, 40% software, 40% experiment
    Requirements: C++, base of evolution
    Subject(s): simulator, evolution, neural networks, genetic algorithms, C++
    Responsible(s): Stéphane Magnenat, Markus Waibel
     

    Labour distribution in colonies of artificial ants

    Julien Chassot (MT)

    Ant colonies have long fascinated because of their enourmous success and complex social structure. One key aspect is their ability to perform task distribution without any central planning. Due to the great number of parameters influencing the biological system, identifying key attributes in real ant colonies is very difficult. Computer models are a powerful way to generate new insights into the complex working of task distribution in biological systems and can at the same time provide methods for translating bio-inspired algorithms into applications for groups of autonomous robots.
    The goal of this project is the implementation of a simple threshold model of labour division into an existing simulator (Ant Farm) and the realisation of a series of evolutionary experiments. This semester project is part of an ongoing research cooperation with the department of ecology of the UNIL.

    A good knowledge of C++, interest in interdisciplinary work and the motivation to apply computer science to real world problems are required.

    Type: Semester project
    Period: Summer 2004
    Section(s): EL IN MT
    Type of work: 50% software, 50% experiments
    Requirements: good knowledge of C++
    Subject(s): Labour distribution in colonies of artificial ants
    Responsible(s): Markus Waibel
    URL: Click here

    Caractérisation de robot s-bot

    Christian Studer (MT)

    Le robot mobile miniature "s-bot" a été développé dans le cadre du projet "swarm-bot". La particularité de ce robot est d’être capable de s'accrocher physiquement à d'aut