My Projects
This is a list of my projects over the last years, though I try to be up-to-date I will never fully be, so see this just to get you interested. You can get more information on separate websites or by contacting me!
Enhancing Biomorphic Agility
Through Variable Stiffness
The first project is nicknamed STIFF and is lead by the German Aerospace Center (DLR), more precisely their section working on biorobotics.. STIFF is a research project on enhancing biomorphic agility of robot arms and hands through variable stiffness & elasticity and is funded by the 7th framework programme of the European Union (grant agreement No: 231576). IDSIA is responsible for learning high-level task-specific controllers based on reinforcement signals for the flexible variable-impedance robot arm developed by DLR, and for extracting cost functions from human behaviours in collaboration with UEDIN (also sometimes referred to as inverse reinforcement learning). I am working here together with Tobias (evolutionary search for cost functions), Jan (Neurocontrollers) and recently with Simon.
Intrinsically Motivated Agents
The main projecte here is named IM-CLeVeR and the partners include CNR (in Rome), Universities of Ulster and Aberstwyth (UK). This project focusses on the iCub robot and how it can learn to interact with its environment. The project is funded by the EU Framework Programme and is officially called Intrinsically Motivated Cumulative Learning Versatile Robots.
It is quite a big project here at IDSIA and a lot of work is done to make the iCub perform right now. The main parts I am working on currently are to get the software setup and start on the interface between the Machine Learning (Reinforcement Learning) group and robotics. Right now I am helping to make the iCub see things, that is helping on the computer vision part of the project.
Evolving Docking
This project aims to evolve neurocontrollers that can be used to control spacecraft. Special focus is put on the problem of automatic rendezvous and docking (AR&D). The main goal is to find answers to the following questions: 'Can we design a neuro-controller that can adapt to varying conditions? What is the tradeoff between the optimality of such neuro-controllers and their adaptibility and robustness?'
As a first step towards a fully reactive neuro-controller (based on state feedback), an artificial neural network is trained offlne by a fitness based genetic algorithm to fulfill a docking task. Using numerical algorithms this returns the optimal control history as a function of time, u(t), an open loop control strategy that is bound to fail in an uncertain and disturbed environment. The vision, and basic motivation for this research, is to find a state-feedback, that is u(x), providing more adaptiveness and robustness to the controller.
Fractionated Spacecraft Simulation
The ACT in collaboration with DG-PS started the investigation of possible future and advanced scenarios and concepts in the broad area of space and security (e.g. disaster monitoring, dual use technology, space debris, NEOs, etc.) with a look at DARPA's F6 project because of its interest framework, which includes possible dual use, its aim for asset security, as well as, re-configurability. Apart from that the project also provides a new methodology for space systems design (see the proposed value-centric design method) with the ability of extending/reconfiguring space architectures over time by adding and/or dropping assets. more >
We aim to simulate dynamic resource sharing between multiple small spacecraft, taking a first step from theoretically described resource sharing to its application in orbit to ensure that each spacecraft module has the needed resources for operation. For this we use a Multi-Agent System (MAS) to model the spacecrafts as agents with a given resource requirement and the ability to transfer (point-to-point) resources between the agents. < less
Multi Robot Cooperation
During my Master's I investigated Multi Robot Formations for Area Coverage in Space Applications, on which I also wrote my thesis.
It presents two algorithmic implementations of multi-robot formation control for the area coverage problem. It uses a space exploration scenario, with a marsupial robot society, for tasks like mapping, habitat construction, etc. The solutions are though applicable to a wider range of applications, since area coverage is one of the canonical problems in multi-robot application. more >
An overview of multi-robot systems in space applications, both currently in use and planned for the near future, is given (and was recently cited by NASA/JPL working group for Mars Exploration here) and then shows two algorithms and their implementation in C++: (i) a vector force based implementation and (ii) a machine learning approach. The second is based on an organizational-learning oriented classifier system (OCS) introduced by Takadama an evolution of HollandŐs learning classifier system (LCS). To ease the development, testing and evaluation of the control algorithms a simulator, named SMRTCTRL, was implemented during a 3 months research stay at the University of Tokyo.
The vector-based control approach was tested using a multi-robot society of LEGO Mindstorms Robots and a comparison of the two algorithm was done with the help of the simulator. < less
CanSat
A CanSat is a “scale satellite” integrated within a soft drink can, that is launched in our case onboard a weather balloon (or an amateur rocket) to perform a fully autonomous measurement of temperature, pressure and so forth. During my Master's we worked on the EISBAR project (Educational Investigation into Satellite Building and Atmospheric Recording). more >
EISBAR designed, implemented and tested this small module to be launched. In order to meet the aim of measuring various specific values, the project was split into many sub-components so that it could be tackled by a group with differing specializations. The module was to be able to record pressure, temperature, position and radiation levels. The data from this CanSat is transmitted to a ground-station via radio signals (this will only work within a certain distance from the ground-station) and stored on-board. Afterwards it is analyzed to learn more about the radiation levels in the ascent/descent path. < less
Stereo-Vision Depth Perception in Humanoid Robots
This project aimed to give an the experimental evaluation of a dense disparity estimation algorithm, focusing on the most relevant aspects for humanoid robots: real-time functionality and ability to deal with calibration errors. The method and its real time implementation (using C++ and open source libraries) are tested in a variety of cases to show its performance and the quality of its output as a function of the design parameters. An early version of the iCub hardware (mounted on-top a torso with one robotic arm) together with the iCub abstractions for the hardware (YARP) were used.
Visual Programming
Mainly focusing on the programming language and IDE Processing, which evolved out of MIT in 2001. We tried to evaluate what it can be used for, featuring a short introduction of the IDEm the main libraries, as well as, a lot of community added libraries. Processing is used in various areas including Visualization, Hardware, Video, Animation, Installations and so forth. We also tried to look into some alternatives and offsprings of Processing like Arduino/Wiring as well as similar IDEs for other programming languages (e.g. Python).
SpaceMaster Robotics Team
The SpaceMaster Robotics Team is a student team consisting of members formerly studying in the Joint European Master in Space Science & Technology (SpaceMaster) programme. It is currently spread all over Europe. And we do cool things :) More about the following projects can be found on the SMRT homepage.
SMRT Smurfs
reel.SMRT
xgravler (HALE)
FSR Team Europe
HRI 2008 - UniRHOCU
Other Projects
Apart from the above projects I was also involved in the following...
Skriptenforum.net
Sozialpsychologie e-Learning
For more information feel free to contact me!