LAUSR

Abstract In this paper, a neuro-dynamic programming (NDP)-based event-triggered control (ETC) method is proposed for unknown non-affine nonlinear systems with input constraints. A neural network-based identifier is established with measurable input and output data to learn the unknown system dynamics. Then, a critic neural network is employed to approximate the value function for solving the event-triggered Hamilton-Jacobi-Bellman equation. Furthermore, an NDP-based ETC scheme is developed, which samples the states and updates the control law when the triggering condition is violated. Compared with the traditional time-triggered control methods, the ETC method can reduce computational burden, communication cost and bandwidth. In addition, the stability of the closed-loop system and the weight error convergence of the critic neural network are provided based on the Lyapunov’s direct method. The intersamling time is proved to be bounded by a positive constant, which excludes the Zeno behavior. Finally, two case studies are provided to verify the effectiveness of the developed ETC method.