Reporting Experiments on Integration of Learning Algorithms and Reactive Behaviour-Oriented Control Systems on a Real Mobile Robot

Abstract

This article reports on experiments that integrate reinforcement connectionist learning and self-organising neural networks with reactive behaviour-oriented mechanisms. The application concerns navigation towards a goal location with which a light source is associated. The experiments are executed on a real mobile robot. We compare the results of the experiments and look at the capability of the mechanisms to control a real mobile robot. We also propose changes to the navigation mechanisms that can improve their performance