Multi-Unmanned Underwater Vehicles Cooperative Obstacle Avoidance Method Based on Artificial Potential Field Method and DQN

To address the large formation errors and long recovery times that arise when multiple unmanned underwater vehicles (UUVs) operate in unknown environments, we propose a collaborative obstacle-avoidance scheme that integrates the artificial potential field (APF) method with Deep Q-Networks (DQN). We first design a formation controller using a consensus algorithm and compute individual avoidance commands through an APF-based potential model. Building on this, we introduce a weight-based cooperative avoidance mechanism in which a DQN adaptively sets the weights according to the relative positions of obstacles. Experiments demonstrate that, during formation navigation around obstacles, the proposed approach yields smaller formation deviations and faster recovery than a conventional APF baseline, and it generalizes effectively to complex, cluttered environments.