Synchronizability-Based Distributed Learning Control for Multi-Agent Systems

In this brief, the distributed learning control problem is addressed for multi-agent systems with irregular dynamics by focusing on the transient synchronization objective. First, the synchronizability property is defined to reveal whether the multi-agent system has the ability to reach the transient synchronization objective, for which a synchronizability criterion is developed in a distributed manner. Then, given the synchronizability property, a distributed learning control protocol is further proposed through the full usage of the experience information from neighbor agents. Under the quasi-strong connectivity, the necessary and sufficient condition is exploited to accomplish the transient synchronization objective for the multi-agent system despite whether it is regular. In particular, the specific selection condition on the gain matrix is presented to provide a fully distributed design manner for the proposed distributed learning control protocol.