A New Reinforcement Learning Method for Fuzzy Logic Controllers

摘要

A new reinforcement method for learning fuzzy logiccontrollers is proposed.The reinforcement learningscheme is composed of two fuzzy logic rule bases:oneacts as an adaptive critic,the other server as a control-ler.The proposed method is tested on the cart-pole sys-tem.Simulation results show that the method has betterlearning performance than Anderson’s neural network-based method.