Distributed Tracking Control of Completely Unknown Heterogeneous MASs Based on Reinforcement Learning

In this paper, a policy iteration control method is proposed to address the distributed tracking control problem of heterogeneous multi-agent systems where the system matrices of both the leader and the followers are unknown. The system matrix of the reference signal is first identified. Then, a distributed observer is developed to realize the asymptotic estimation of the leader state. The algebraic Riccati equation (ARE) is derived and the optimal solution of ARE is obtained based on reinforcement learning. Under the proposed distributed tracking control framework and cooperative control protocol, all followers can asymptotically synchronize the reference signal generated by the leader. Finally, the theoretical feasibility of the proposed method is verified by numerical simulation.