Adaptive Reinforcement Learning Tracking Control for Second-Order Multi-Agent Systems

Abstract

In this paper, the adaptive reinforcement learning tracking control problem is studied for second-order pure-feedback multi-agent systems (MASs). The pure-feedback MASs are transformed into strict-feedback form by using the mean value theorem. The reinforcement learning approach is applied to handle the unknown functions and system control performance index. Moreover, the error terms are introduced to the controller, which can improve the robust of the control scheme. The theoretical analysis indicates that all the signals and tracking errors in close-loop system are semi-global uniformly ultimately bounded (SGUUB), and the numerical simulation are conducted to verify the superiority of this scheme.