Fault-tolerant control based on reinforcement learning and sliding event-triggered mechanism for a class of unknown discrete-time systems

Abstract

This paper presents an adaptive fault-tolerant control (FTC) system based on reinforcement learning using an even-triggered mechanism. The even-triggered mechanism is established through a justifiable sliding surface and triggered function, without the need for any fault detection or observer. The learning laws are derived to ensure the convergence of internal signals and tracking error, and an actor–critic architecture is designed accordingly. To validate the proposed scheme, an experimental system is constructed and tested using five typical actuator faults. The results indicate a positive closed-loop performance and a reduction of approximately 25% in data transmission for both cases with and without faults.