Robust Genetic Network Programming using SARSA Learning for autonomous robots

Abstract

Adaptive and robust control has attracted increasing attention in the field of artificial intelligence. Adaptive controller makes use of some adaptation mechanisms which are designed to learn explicitly the unknown parameters of the system and the uncertain situation under control. An evolutionary algorithm called “Genetic Network Programming, GNP” has been already proposed to control intelligence systems. In conventional GNP, when they have a certain problems such as the hardware or software of the system, the nodes and connections of GNP may not work well. In this paper, GNP with SARSA Learning is applied to construct the robust GNP for autonomous robots.