Optimal Bipartite Consensus Control for Unknown Coopetition Multi-agent Systems with Time-delay via Reinforcement Learning Method

Abstract

In this article, a data-driven optimal bipartite consensus control (OBCC) scheme is proposed for unknown heterogeneous multi-agent systems (MASs) with time-delay via reinforcement learning (RL) algorithm. A directed signed graph is established to construct MASs with cooperative and competitive relationships, and model reduction method is developed to transform MASs with time-delay into a delay-free MASs. Then, based on Bellman’s optimal principle, a policy iteration method is utilized to design OBCC strategy. Further, based on Q-function, a model-free Q-function policy iteration algorithm is proposed to solve the OBCC problem for unknown MASs. And, only using input-output states of MASs to tackle the OBCC solution via RL algorithm, and it is implemented by actor-critic neural networks (NNs). Finally, simulation results are given to validate the feasibility and efficiency of the proposed algorithm.