基于Takagi-Sugeno的再励学习模糊神经网络控制

提出一种模糊神经网络的自适应控制方案。针对连续空间的复杂学习任务,提出了一种竞争式Takagi-Sugeno模糊再励学习网络,该网络结构集成了Takagi-Sugeno模糊推理系统和基于动作的评价值函数的再励学习方法。相应地,提出了一种优化学习算法,其把竞争式Takagi-Sugeno模糊再励学习网络训练成为一种所谓的Takagi-Sugeno模糊变结构控制器。以一级倒立摆控制系统为例,仿真研究表明所提出的学习算法在性能上优于其它的再励学习算法。

FUZZY NEURAL NETWORKS CONTROL BASED ON TAKAGI- SUGENO AND FUZZY REINFORCEMENT LEARNING

This paper presents a new adaptive control based on fuzzy neural network. For solving complicated learning tasks of contin- uous domains, this paper proposes a competitive Takagi- Sugeno fuzzy reinforcement learning network. It is constructed by combining Takagi- Sugeno type fuzzy inference systems with action- value- based reinforcement learning methods. Accordingly, it proposes a opti- mal learning algorithm. It leads to so called Takagi- Sugeno fuzzy variable structure controllers. Simulation experiments on the signal inverted pendulum system demonstrate the superiority of this learning methods.