Towards Behavior Control for Evolutionary Robot Based on RL with ENN

Abstract

This paper proposes a behavior-switching control strategy of an
evolutionary robotics based on Artificial Neural
Network (ANN) and Genetic Algorithms (GA). This method is able not only to construct the
reinforcement learning models for autonomous robots and evolutionary robot modules that
control behaviors and reinforcement learning environments, and but also to perform the
behavior-switching control and obstacle avoidance of an evolutionary robotics (ER) in
time-varying environments with static and moving obstacles by combining ANN and GA.

The experimental results on the
basic behaviors and behavior-switching control have demonstrated that our
method can perform the decision-making strategy and parameters set opimization of
FNN and GA by learning and can escape successfully from the trap of a local
minima and avoid \emph{"motion deadlock" status} of humanoid soccer robotics agents,
and reduce the oscillation of the planned trajectory between
the multiple obstacles by crossover and mutation. Some results of the proposed algorithm
have been successfully applied to our simulation humanoid robotics soccer team CIT3D
which won \emph{the 1st prize} of RoboCup Championship and ChinaOpen2010 (July 2010) and \emph{the $2^{nd}$ place}
of the official RoboCup World Championship on 5-11 July, 2011 in Istanbul, Turkey.

As compared with the conventional behavior network and the adaptive behavior method,
the genetic encoding complexity of our algorithm is simplified, and the network
performance and the {\em convergence rate $\rho$} have been greatly
improved.

DOI: http://dx.doi.org/10.11591/ijra.v1i1.259