Deep Reinforcement Learning-Based Cooperative Control for Multimobile Robots With Obstacle Avoidance

作者：Yu-Kai Fu, Yiyang Liu, Chao Deng · 发表于：IEEE Transactions on Industrial Electronics · 年份：2025 · DOI：10.1109/tie.2025.3581265 · 被引用次数：5 · 研究领域：Reinforcement Learning in Robotics、Traffic control and management

In this article, the formation and obstacle avoidance problem is investigated for multimobile robots (MMRs). Compared with the traditional formation and obstacle avoidance method, a novel data-driven hierarchical control framework is proposed, which can be divided into designing an obstacle avoidance reference trajectory algorithm, developing a trajectory smoothing generator, and designing distributed adaptive controllers. In particular, a deep reinforcement learning-based trajectory generation algorithm is proposed to generate a reference trajectory, which can achieve both obstacle avoidance and reach the prespecified target. Based on this trajectory, a smooth trajectory generator is designed to generate a smooth trajectory with up to third-order derivatives, which facilitates the design of an adaptive tracking controller using the backstepping method. Moreover, distributed adaptive controllers are designed for all MMRs to ensure that both the formation and obstacle avoidance objectives can be achieved. Finally, a numerical simulation and experiment are conducted to verify the effectiveness of the proposed hierarchical control method.