Heterogeneous formation control of multiple UAVs with limited-input leader via reinforcement learning

NEUROCOMPUTING

卷:412

Article

Liu, Hao^1,2  Meng, Qingyao^1,2  Peng, Fachun³  Lewis, Frank L.⁴ 

[1] Beihang Univ, Sch Astronaut, Beijing 100191, Peoples R China

[2] Beihang Univ, Key Lab Spacecraft Design Optimizat & Dynam Simul, Minist Educ, Beijing 100191, Peoples R China

[3] Gen Design Dept SCAAT, Chengdu 610100, Peoples R China

[4] Univ Texas Arlington, Res Inst, Ft Worth, TX 76118 USA

关键词: Formation control;    Multi-agent systems;    Heterogeneous systems;    Reinforcement learning;    Unmanned aerial vehicles;   

DOI  :  10.1016/j.neucom.2020.06.040

来源: Elsevier

【 摘 要 】

In this brief, a distributed optimal control method via reinforcement learning is proposed to address the heterogeneous unmanned aerial vehicle (UAV) formation trajectory tracking problem. The UAV formation is composed of a virtual leader with limited nonzero input and several follower vehicles with different unknown dynamics. The proposed control law contains a distributed observer and a model-free off policy reinforcement learning (RL) protocol. The distributed optimal trajectory tracking problem is formulated for the heterogeneous formation system. A RL algorithm is designed to obtain the optimal control input online without any knowledge of the followers' dynamics. Simulation example illustrates the effectiveness of the proposed method. (C) 2020 Elsevier B.V. All rights reserved.

【 授权许可】

Free