Event-triggered optimal consensus tracking control for multi-agent systems with unknown internal states and disturbances

Abstract

In this paper, we propose an event-triggered consensus tracking control for unknown nonlinear multi-agent systems (NMASs) with disturbances by using adaptive dynamic programming (ADP) technique. Considering the disturbances as control inputs, the optimal consensus tracking control problem can be transformed into multi-player zero-sum differential games, where the controllers are designed to minimize the cost functions under the worst-case disturbances. To recover the unknown internal states and control coefficient functions, neural network (NN)-based observers are established. Then, NN-based critics are applied to approximate the value functions and help calculate the optimal control policies and disturbance policies. In order to save the computation resource and reduce the transmission load, the designed observers and controllers are updated only when the designed events are triggered. Stability of the proposed method is demonstrated by Lyapunov analysis for both the continuous and jump dynamics. Meanwhile, we can obtain that the weight estimated errors of the observer NNs, the critic NNs and the local neighbor consensus tracking errors are uniformly ultimately bounded (UUB). Finally, a simulation example is given to illustrate the effectiveness of the proposed approach.