Model-free Optimal Containment of Heterogeneous Multiagent Systems

This paper discussed optimal containment problem of heterogeneous multiagent systems using reinforcement learning technique, in which the leaders’ controls are nonzero. Distributed adaptive observers are presented to estimate the leaders’ dynamics and construct the state trajectories insider the convex hull of the leaders. To achieve the system’s optimality, we define a discounted cost function for each follower. We develop a reinforcement learning algorithm to address the ARE. Then we give a simulation to verify the effectiveness of the algorithm.