ISIR - Institut des Systèmes Intelligents et de Robotique

Institut des Systèmes Intelligents
et de Robotique

Partenariats

UPMC

CNRS

Rechercher

A voir également

Profil

mouret Jean-Baptiste
Title : Associate Professor
Addres : Pyramide ISIR - 4 Place Jussieu 75005 Paris
Phone : 01 44 27 51 06
Email : mouret(at)isir.fr
Team : SIMA (Integrated, Mobile and Autonomous System)
Member of project(s) : ROBUR

Research Activities

My current researches focus on the use of evolutionary algorithms to synthesize complex structures useful in robotics (neural networks, control loops, whole robot, ...). Most of the approaches I work on rely on Pareto-based multiobjective evolutionnary algorithms (MOEA) and are currently applied to evolve neural networks to control robots or unmaned aerial vehicles (UAV).

In this context, I mainly work on the following questions:

  • What genotype can be defined to encode structures (graphs, neural networks, ...) and exploit modularity (see [2008ACLI931]).
  • How are linked modularity and selection gradients? (see Mouret and Doncieux, 2009)
  • How to apply a selective pressure of increasing difficulty to overcome the bootstrap problem ? (see [2008ACTI850 ])
  • How to maintain the diversity of a population ? (see Mouret and Doncieux, 2009)
  • How to employ on real robot the results on a evolutionary algorithm ? (see Koos et al, 2009)

 

The source code for all our experiments is available in the EvoRob database .

 I currently work on improving  these ideas and on importing the knowledge from computational neuro-science in an evolutionary process.

 

Some of the proposed evolutionary methods have been (or will be) employed in the Robur [2006ACTI361 ], which aims at creating a bird-like flapping wing robot. In this project, I contributed on the following topics:

  • Evolving the morphology and the kinematics of a flapping-wing robot, with a special emphasis on the comparison between our results and biological data ([2008ACTI937 ] et [2007ACLI732 ]);
  • Evolving neuro-controlers to build a closed-loop neuro-controllers for a flapping wing UAV [2006ACTI462 ].
  • Evolving neuro-controlers to control a soaring UAV (exploiting thermals and ridge soaring ) [2007ACTI734 ] ;
  • Control of a quadrotor [2007ACTI735].

The following video illustrates the evolution of a neuro-controler to secure an horizontal flight at constant speed despite perturbations: