一种自适应模糊Actor-Critic学习

提出一种基于模糊RBF网络的自适应模糊Actor—Critic学习．采用一个模糊RBF神经网络同时逼近Actor的动作函数和Critic的值函数，解决状态空间泛化中易出现的“维数灾”问题．模糊RBF网络能够根据环境状态和被控对象特性的变化进行网络结构和参数的自适应学习，使得网络结构更加紧凑，整个模糊Actor—Critic学习具有泛化性能好、控制结构简单和学习效率高的特点．MountainCar的仿真结果验证了所提方法的有效性．

A Kind of Adaptive Fuzzy Actor-Critic Learning

An adaptive fuzzy Actor-Critic reinforcement learning based on a fuzzy radial basis function network is proposed,which can solve the 'curse of dimensionality' problem caused by state space generalization.A fuzzy RBF network is used to approximate both the action function of Actor and the value function of Critic simultaneously.The fuzzy RBF network is able to adjust its structure and parameters in an adaptive way with a self-organizing approach according to the change of environment state and the characteristics of the plant during the learning process,which ensures network size is economical.The proposed fuzzy Actor-Critic learning has advantages of perfect generalization ability,simple control structure,and high learning efficiency.Simulation experiment on Mountain Car control shows the validity of the proposed algorithm.