Stackelberg Game-Based Finite-Horizon Optimal Fault-Tolerant Control of Heterogeneous MASs
研究了异构非线性离散时间多智能体系统的有限时域最优容错控制问题，基于Stackelberg博弈和图形博弈框架，将智能体系统与故障估计器分别视为领导者和跟随者，设计了分布式最优故障估计与容错控制方案，并通过自适应动态规划在线求解交互Stackelberg均衡。
This article investigates the problem of finite-horizon optimal fault-tolerant control for heterogeneous nonlinear discrete-time multiagent systems. A distributed optimal fault estimation and optimal fault-tolerant control scheme is presented based on the Stackelberg game and graphical game frameworks, in which agent systems and corresponding fault estimators are treated asthe “leader” and “follower” in the Stackelberg game, and the agent systems constitutes a graphical game. This is more general than the existing integration design of fault estimation as well as fault-tolerant control, as it better reveals the interaction between the agent system and its fault estimator. To achieve interactive Stackelberg equilibrium, an auxiliary controller for fault estimator is designed, which exhibits non cooperative properties with the fault-tolerant controller of the agent. Then, a finite-time distributed estimator for discrete-time case is developed for the first time to rapidly estimate the matrix and state of the leader agent. The coefficient matrices of the considered fault signals obey a hidden semi-Markov switching rule, which can encompass the existing actuator and sensor fault models. Further, two critic neural networks are established to obtain approximate solutions of interactive Stackelberg equilibrium online by employing the adaptive dynamic programming technology. Finally, simulation examples are provided to verify the efficacy and applicability of the proposed scheme.