PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Embedded Evolutionary Robotics: The (1+1)-Restart-Online Adaptation Algorithm
Jean-Marc Montanier and Nicolas Bredeche
In: IROS 2009 Workshop on Exploring new horizons in Evolutionary Design of Robots, Oct 2009, USA.

Abstract

This paper deals with online onboard behavior optimization for an autonomous mobile robot in the scope of the European FP7 Symbrion Project. The work presented here extends the (1+1)-online algorithm introduced in earlier publication. This algorithm is a variation of a famous Evolution Strategies adapted to autonomous robots. In this paper, we address a limitation of this algorithm regarding the ability to perform global search whenever a local optimum is reached. A new implementation of the algorithm, termed (1+1)-restart-online algorithm, is described and implemented within the Symbrion robotic Cortex M3 microcontroller. Results from the experiments show that the new algorithm is able to escape local optima and, as a consequence, converge faster and provides a richer set of relevant controllers.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
ID Code:6789
Deposited By:Nicolas Bredeche
Deposited On:08 March 2010