Adaptive consensus for uncertain multi-agent systems with stochastic measurement noises

Abstract

This paper addresses asymptotic consensus for multi-agent systems with the measurement of neighbor information interfered by stochastic noises. Notably, intractable parametric uncertainties are allowed in the agent dynamics, which, coupling with inherent nonlinearities, firmly ask for distributed adaptive compensation. To limit the influence scope of measurement noises on adaptive mechanism, the distributed protocol design is separated into two aspects: (i)Reference consensus based on relative information. Local dynamic generators, providing certain reference signals, are constructed by exploiting the relative information multiplying a subtle time-varying gain. With the measurement noises attenuated via the time-varying gain, all the generators are guaranteed to evolve for all time and reach consensus in the almost sure sense. (ii)Agent tracking based on individual estimate. Utilizing individual information, local adaptive estimates are proposed for the agent uncertainties while not involving the noise measurements. In this way, a distributed adaptive protocol is designed to make each agent state track the corresponding reference signal while preventing the inherent nonlinearities from incurring finite-time explosion, and furthermore, to enable all the agents to reach average consensus in the almost sure sense.