Event-Triggered Integral Reinforcement Learning Control Based on Recursive Terminal Sliding Mode

Abstract

For a class of continuous-time non-linear systems with saturated input and unknown non-linear disturbance, a novel event-triggered integral reinforcement learning (IRL) control strategy based on recursive terminal sliding mode (RTSM) is proposed in this paper. Firstly, a novel performance index function is designed based on RTSM and a two-player zero-sum game, and the robust control problem with saturated input and unknown disturbance can be transformed into an optimal control problem. To avoid the requirement of drift dynamics, the IRL technique is introduced. Secondly, a critic neural network is used to approximate the optimal value function, which not only simplifies algorithm implementation structure, but also relaxes initial admissible control in the learning of neural network weights. Then, considering the event-triggered mechanism, the asymptotic stability of the closed-loop system and the uniformly ultimately boundedness of weight estimation errors are proved by utilizing the Lyapunov theory. Finally, simulation results illustrate the effectiveness of the proposed control method.