Predictor-Based Event-Triggered Optimized Control for Uncertain Nonlinear Systems With Time Delays: A Reinforcement Learning Approach

Abstract

This article investigates the adaptive optimized control issue for uncertain nonlinear systems with time delays, where output signal can only be available through periodic sampling. Based on the RBF neural network approximation methods, a fresh predictor-based continuous-discrete fuzzy state observer is presented to estimate the unmeasurable states. Specially, in the backstepping design process, we introduce the reinforcement learning algorithm with actor–critic architecture to achieve better optimal control performance. Moreover, an adaptive auxiliary system is presented to eliminate the effect of input delays. To reduce the sampling of output signals, a novel periodic event-triggered controller is presented. With Bellman–Gronwall inequality and Lyapunov stability theory, the semiglobally uniformly ultimately bounded of all signals is proved. Finally, two illustrative examples are included to demonstrate the validity of control framework.