Optimal Control for Fuzzy Markov Jump Singularly Perturbed Systems: A Hybrid Zero-Sum Game Iteration Approach
Jing Wang, Yaling Huang, Xiangpeng Xie, Huaicheng Yan, Hao Shen
Abstract
This article addresses a novel optimal fuzzy control scheme for a category of nonlinear Markov jump singularly perturbed systems with mixed zero-sum (MZS) game. First, a Takagi–Sugeno fuzzy structure is employed to describe the nonlinear plant. On this basis, two types of value functions for multiplayer MZS game are developed and a series of fuzzy controllers are designed. Subsequently, a hybrid iteration (HI) approximation learning framework is constructed for approaching the solutions of the fuzzy game coupled algebraic Riccati equations. The HI eliminates initial admissible control dependencies and has a desirable learning iteration rate, in contrast to the traditional policy iteration and value iteration. We present a novel model-based HI scheme and further develop it to a model-free version. The model-free HI scheme uses only online data collected along the system trajectories for controller design, and it eliminates the traditional model-dependent issues, which is more suitable for practical applications. Finally, the validity and flexibility of the proposed algorithm are verified by a simulation example.