Fullscreen
Loading...
 
Print

Symbrion job offers

TAO team is hiring a Post-Doc to work in Learning Algorithms for Swarm Robotics.

Context
The SYMBRION project is a European Integrated Project that aims to build self-assembling robot organisms from autonomous individual small robots.
Within Symbrion, the TAO team will investigate learning and optimization mechanisms in a multi-scale perspective, online (in situ and in simulation) and offline (from robotic logs).
Another related project is the ANR/JST SyDinMala, in cooperation with Prof. Einoshin Suzuki, Kyushu University.

Tasks
The task will be to design and implement Machine Learning algorithms under limited resources on the hybrid platform (software and hardware) provided by the project partners. The stress will be put on dealing with huge robotic logs and learning to construct efficient controller modules.
A first task will consist in applying some Learning and/or Data Mining techniques to multiple robot logs, while the robots are driven either by a human being or by a manually written controller. This should allow to automatically generate controller modules — off-line.
A second task will then be to design a high-level controller to perform the action selection among different modules, either hard-coded or automatically designed by the off-line procedure. This controller will be optimized either off-line using Evolutionary Computation, or on-line using some algorithms inspired from the Multi-Armed Bandit techniques.

The ultimate goal of the project is to design on-line protocols to allow open-ended evolution at both scales of high-level controller and generation of new low-level modules.

Expertise
Knowledge in Machine Learning and Statistics is of course mandatory, as well as excellent programming skills (C++ and Java). Previous experience in Robotics and/or Evolutionary Computation is welcome.

Practical information
The TAO team is part of INRIA CRI Orsay Île-de-France and its offices are in LRI at Université Paris-Sud. French public salaries include medical coverage.


Contributors to this page: evomarc and sebag .
Page last modified on Tuesday 25 of November, 2008 23:22:21 CET by evomarc.