Institut des Systèmes Intelligents
et de Robotique

Partenariats

Sorbonne Universite

CNRS

INSERM

Tremplin CARNOT Interfaces

Labex SMART

Rechercher

A voir également

Short bio

doncieux Stéphane
Title : Professor
Address : 4 place Jussieu, CC 173, 75252 Paris cedex 05
Phone : +33 (0) 1 44 27 87 45
Email : doncieux(at)isir.upmc.fr
Group : AMAC (AMAC)

 

I am currently the head of the AMAC research team at ISIR and the scientific coordinator of the DREAM European project

My research activities follow the frame of the animat approach, whose goal is to design autonomous robots able to adapt their behavior to their environment. This approach, in which biology is a fundamental source of inspiration, is centered on the interactions of the robot with is environement and the ability to extract useful information from it to change or modify its behavior depending on the context (motor failure, for instance) in order to solve some tasks with as little human intervention as possible.

I am thus interested in learning in a wide acceptance and in particular in evolutionary robotics, i.e. on the use of algorithms inspired from natural selection (genetic algorithms or more generally evolutionary algorithms) applied to robotics problems.


Evolutionary robotics deals with the use of evolutionary algorithms to design robots together with their control system and to make them exhibit some behavior with desired features: walking, swimming, flying, avoiding obstacles, surviving as long as possible, etc. The main interesting aspect of these methods lies in the crucial role of the interactions between the robot and the environment. A specific exploration procedure allows thus to exploit real robot abilities as we don't need to do any prior simplification. Innovative and original solutions that are also simple and yet efficient, may thus be found. We have thus obtained, in the frame of the ROBUR project, neural networks exploiting the features of a lenticular blimp or simple neural networks able to exhibit a thermal exploitation behavior while using only few informations (pitch, roll and vertical speed).

Our work in this field are focused on the consequences on the evolutionary algorithms of focusing on robotics applications. More precisely, we are concerned with:

  • control system encoding that impacts on the search space and how to explore it,
  • pressure selection that may be adapted to robot behavior search,
  • links between simulation and reality.