Reinforcement Learning-Based Adaptive Fault-Tolerant Optimal Control with Time Delays for Discrete-Time Multi-Agent Systems

This paper investigates adaptive fault-tolerant optimal control for discrete-time multi-agent systems with time-varying input delays and actuator faults using reinforcement learning. A controller is developed via a policy iteration method to solve the discrete-time Hamilton-Jacobi-Bellman equation. In addition, actor-critic neural networks enable real-time approximation of the control input and performance index. An M network addresses input delays, while a fault compensation scheme mitigates actuator faults. Convergence and stability are proven, and numerical simulations validate effectiveness.