Adaptive Formation Framework with Obstacle Avoidance Based on Reinforcement Learning

Multi-robot formation control is a control technology that enables robots to maintain a specific formation and adapt to environmental constraints during task execution. In practical application, it is important to maintain structure and adaptively avoid obstacles. Therefore, a multi-robot adaptive formation obstacle avoidance framework is proposed in this paper. There are two pieces to this framework. An obstacle avoidance navigation algorithm built on reinforcement learning makes up the first section. Additionally, we optimize the system’s reward function using the artificial potential field method (APF). In the second part, we design a new formation control strategy based on the leader-follower model and propose the virtual obstacle avoidance point (VOAP) to solve the overlapping problem between virtual target points and obstacles in the leader-follower model. The results of the simulations demonstrate that the proposed formation obstacle avoidance framework can make unmanned ground vehicles (UGV) carry out formation keeping and adaptive obstacle avoidance independently.