Off-Policy Learning for Bipartite Output Regulation of Heterogeneous Multi-Agent Systems under Actuator Faults

The bipartite output regulation problem for a class of heterogeneous multi-agent systems with actuator bias faults is studied in this paper, where the dynamics of followers are completely unknown. To tackle such a problem, a kind of fixed-time distributed observers is proposed to learn the leader's dynamics and state. Then, a class of off-policy learning algorithms is developed based on the designed fixed-time observers to construct model-free fault-tolerant control schemes. It has been proved that the bipartite output regulation problem can be solved by the proposed fixed-time observer-based model-free fault-tolerant control protocols under some suitable conditions. Finally, the effectiveness of the presented theoretical results is demonstrated through performing numerical simulations.