Robot Reinforcement Learning Methods Based on XCSG

Abstract

This paper proposed a robot reinforcement learning method based on learning classifier system. A Learning Classifier System is a accuracy-based machine learning system with gradient descent that combines reinforcement learning and rule discovery system. The genetic algorithm and the covering operator act as innovation discovery components which are responsible for discovering new better reinforcement learning rules. The reinforcement learning component is responsible for adjusting the fitness of rules in the system according to some reward obtained from the environment. The advantage of this approach is its accuracy-based representation, which can easily reduce learning space, improve online learning ability and robot robustness.