Event-Based Data-Driven Consensus Control for Multiagent Systems with Unknown Dynamics

TL;DR: This paper proposes an event-triggered data-driven consensus control approach for multi-agent systems with unknown dynamics, utilizing a critic network and actor to optimize regulation and prevent Zeno behaviors through uniform ultimate boundedness of local neighborhood error.

Abstract: This paper addresses the problem of achieving tracking consensus control for multi-agent systems (MASs) with unknown dynamics. First, an augmented neighborhood error system model is designed, and the optimal consensus protocol of MASs is converted to the optimal regulation for augmented system with a performance function. A control policy based on event-triggered is proposed to address the optimal regulation of the augmented system. Then, a data-driven control approach is proposed to implement the event-triggered policy, including a critic network and an actor. The actor is utilized to generate control inputs and the critic network estimates the performance value. Through the selection of suitable event-triggered thresholds and learning rates, the local neighborhood error is proved to be uniformly ultimately bounded, effectively preventing the occurrence of Zeno behaviors. Finally, two simulation examples are presented and analyzed to validate the efficacy of the proposed control model.