Cooperative output regulation of heterogeneous directed multi-agent systems:a fully distributed model-free reinforcement learning framework  

作　　者：Xiongtao SHI Yanjie LI Chenglong DU Huiping LI Chaoyang CHEN Weihua GUI 

机构地区：[1]Guangdong Key Laboratory of Intelligent Morphing Mechanisms and Adaptive Robotics,School of Mechanical Engineering and Automation,Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China [2]School of Automation,Central South University,Changsha 410083,China [3]School of Marine Science and Technology,Northwestern Polytechnical University,Xi’an 710072,China [4]School of Information and Electrical Engineering,Hunan University of Science and Technology,Xiangtan 411201,China

出　　处：《Science China(Information Sciences)》2025年第2期166-181,共16页中国科学(信息科学)(英文版)

基　　金：supported by National Natural Science Foundation of China(Grant Nos.62303492,61977019,62222306);Shenzhen Basic Research Program(Grant Nos.JCYJ20220818102415033,JSGG20201103093802006,KJZD2023092311-4222045);Natural Science Foundation of Hunan Province(Grant No.2023JJ40765);Natural Science Foundation of Changsha(Grant No.kq2208287);Science and Technology Innovation Program of Hunan Province(Grant No.2022WZ1001);China Postdoctoral Innovation Talents Support Program(Grant No.BX20230430)。

摘　　要：In this paper,the cooperative output regulation(COR)problem of a class of unknown heterogeneous multi-agent systems(MASs)with directed graphs is studied via a model-free reinforcement learning(RL)based fully distributed eventtriggered control(ETC)strategy.First,we consider the scenario that the exosystem is accessible globally to all agents,an internal model-based augmented algebraic Riccati equation(AARE)is constructed,and its solution is learned by the proposed model-free RL algorithm via online input-output data.Further,for the scenario that the exosystem is accessible only to its adjacent followers,the distributed observers are designed for each agent to get the state of the exosystem,and an internal modelbased fully distributed adaptive ETC protocol is then synthesized to construct the corresponding AARE,and the feedback gain matrix is learned in a model-free fashion.The model-free RL-based control protocol proposed in this paper can not only remove the prior knowledge of agents'dynamics,but also release the dependence on global information by the adaptive event-triggered mechanism(ETM)and the new graph-based Lyapunov function.Finally,simulation results are illustrated to show the feasibility and effectiveness of the proposed control scheme.

关 键 词：model-free reinforcement learning unknown heterogeneous multi-agent systems fully distributed event-triggered control directedgraph 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] TP181[自动化与计算机技术—控制科学与工程]