Adaptive leaderless consensus of MIMO multi-agent systems with unknown dynamics and nonlinear dynamic uncertainties

Abstract

In this paper, the bipartite consensus problem of a class of unknown uncertain multi-agent systems (MASs) under a signed graph is investigated via a filter-based model-free reinforcement learning (RL) based fully distributed control scheme. First, considering the completely unknown dynamics of the agents, a novel filter-based model-free RL algorithm is proposed to learn the stabilizable feedback gain matrix via the online input–output data. Then, fully distributed algorithm with adaptive control gains is designed such that all agents reach bipartite consensus in the presence of dynamic uncertainties. Finally, a numerical simulation is illustrated to demonstrate the correctness and effectiveness of the proposed control scheme.