Deep Reinforcement Learning Based Multiple Omnidirectional Mobile Robots Control
Haoyu Tian, Yanqi Lu, Ouyang Zhang, Guanghui Sun, Chengwei Wu, Weiran Yao · 2021 China Automation Congress (CAC) · 2021
Three-wheeled omnidirectional mobile robot (TWODMR) is a new type of holonomic mobile robot, which performs better moving flexibility than nonholonomic mobile robot. In this paper, we focus on the multi-robot formation control problem of three-wheeled omnidirectional mobile robots. The deep deterministic policy gradient (DDPG) algorithm is applied to carry out the formation motion planning and controlling task and the artificial potential field (APF) algorithm is used as a comparison. According to the experiment results, the DDPG algorithm solves the multi-robot formation control problem. The DDPG algorithm optimizes its control strategy and the trained DDPG controller achieves more accuracy and adaptability than the APF controller.