Output synchronization of unknown heterogeneous agents via distributed model reference adaptation

Abstract

This work presents a distributed model reference adaptive methodology for output synchronization of heterogeneous linear agents with unknown dynamics. We consider a setting in which the control input is communicated among neighbors, instead of observer variables. For those agents that can access the signals of the reference model, classical model reference adaptation laws lead to leader synchronization; for those agents that cannot access such signals, synchronization must be achieved by taking the neighboring agents as an alternative reference model. We show that these two groups of agents give rise to two types of matching conditions: the standard conditions to match the reference model, and new distributed matching conditions among neighboring agents. Since all matching gains are unknown, the gains are adapted online via Lyapunov-based estimation. Asymptotic synchronization is proven analytically, and numerical examples show the effectiveness of the approach.