Group Formation Tracking of Heterogeneous Multiagent Systems Using Reinforcement Learning

This article introduces a novel distributed control protocol to address the problem of output group formation tracking for heterogeneous multiagent systems. In contrast to the existing group formation control protocols that rely on the system matrices, our approach leverages input-state data to design the optimal control gains in the framework of off-policy reinforcement learning. Specifically, an event-triggered distributed consensus estimator is proposed to estimate the leaders' system matrices and the convex hulls spanned by the leaders while ensuring the exclusion of Zeno behavior. Based on the proposed estimator, we establish an approximate optimal distributed control protocol for each follower to achieve output group formation tracking, which can be solved by the designed data-driven policy iteration algorithm. Finally, a numerical example is provided to show the efficacy of the proposed control approach.