Dynamic Event-Triggered Optimized Backstepping Containment Control of a Class of Stochastic Nonlinear MASs Using Reinforcement Learning

This article focuses on the fully distributed dynamic event-triggered optimized backstepping (OB) containment control for a class of stochastic nonlinear high-order strict-feedback multiagent systems (MASs) based on reinforcement learning (RL) algorithm. First, an OB control scheme is proposed using the critic-actor structure. In this scheme, neural networks are applied to approximate the value function and the controllers are derived from the gradients of simple positive functions. This approach simplifies the algorithm and effectively addresses two fundamental requirements commonly encountered in RL-based optimal control: known dynamics and persistent excitation. Furthermore, the static and dynamic event-triggered mechanisms are designed in order to reduce the control actions in the highest layer for nonlinear high-order stochastic MASs. Then, the distributed event-triggered control protocols based on RL are proposed for the strict-feedback nonlinear MASs with stochastic disturbances. Finally, simulation results are shown to demonstrate the effectiveness of the theoretical analysis.