Data-Driven Model-Free Event-Triggered Distributed Adaptive Observer for Multiagent Systems Using Experience Replay

This paper proposes a model-free data-driven event-triggered adaptive observer (ETAO) for linear leader-follower multi-agent systems (MASs). The ETAO aims to estimate the leader's state for every agent without any information of the system dynamics. Initially, a data-driven experience-replay method is applied to identify the leader's dynamic matrix, utilizing only the leader's past state information. Furthermore, an innovative event-triggered mechanism is presented, effectively reducing the excessive communication transmissions among agents, and demonstrating Zeno behavior does not occur. Dynamic matrix recognition and the event-triggered observer are codesigned simultaneously, guaranteeing system stability. Finally, a simulation example validates the effectiveness of the ETAO.