Determination of the optimal values of parameters in reinforcement learning for mobile robot navigation by a genetic algorithm

Reinforcement learning is useful in location estimation, obstacle avoidance and path planning for autonomous mobile robots. We have to determine various parameters in reinforcement learning without prior information. We propose to determine the optimal values of parameters with the help of a genetic algorithm. We have succeeded in accelerating the speed of learning by about 20-30% and in decreasing the number of actions needed to reach the goal by about a half. (C) 2004 Elsevier B.V. All rights reserved.