Online Learning Cooperative Control for Heterogeneous Multi-Agent Systems

The cooperative control research of the leaderfollower heterogeneous multi-agent system with mixed dynamics is studied. To acquire the control objective of the uniformly ultimately bounded consensus, the direct heuristic dynamic programming (DHDP) based online learning controller is proposed. The actor-critic architecture is implemented for the controller algorithm. The critic network is employed to approximate the index function. The action network is applied to minimize the index function and produce the control signal. This method can be employed in real time and no accurate system model is needed. Finally, the numerical experimentation is given to reflect the practicability and reliability of the online controller.