Researcher in adaptive systems

Topics: autonomous flying animats, neural networks, evolutionary algorithms

My research activities consist in the design of autonomous flying animats. During my PhD, I worked on the automatic design of neuro-controllers by means of evolutionary algorithms. The test platforms were a 10m wide lenticular blimp and a simulated helicopter. Evolved neural networks were able to exploit aerodynamical features of the aicraft without the need to specify them neither in the fitness function nor in any bias in the neural network structure. Neural networks thus exploited the ability of the lenticular blimp to generate lift when facing the wind to efficiently control altitude.

My activities are now devoted to all aspects necessary to the development of an autonomous flying animat. I am then interested in the use of vision to implement a reactive behaviour in a flying animat and in the study of flapping wings animats.

This work is part of Robur Project, which aims at developping a completly autonomous flapping wings animat. This will require to adapt technics developed in our lab to the context of flying animats (cartography, action selection system), as well as the integration to a real platform.

I have developed the SFERES framework devoted to experiments involving evolution and simulation in collaboration with Samuel Landau. It is available now as an open source software under the CeCILL licence.

Whereabouts

AnimatLab
Tel : +33 1 44 27 87 45
Email : stephane.doncieux@lip6.fr