Multi-unmanned vehicle collaborative path planning method based on deep reinforcement learning

This study aims to tackle the collaborative path planning issue in multi-unmanned vehicle systems using deep reinforcement learning. We’ve devised an efficient path planning framework by first establishing kinematic and mathematical models for differential-drive unmanned vehicles and collaborative obstacle avoidance scenarios. Then, we addressed the challenges of slow training, low sampling efficiency, and poor adaptability of reinforcement learning in complex dynamic scenarios. For cooperative obstacle avoidance and pursuit, we suggested an improved twin delayed deep deterministic policy gradient (AE-TD3) algorithm. By introducing random noise to pursuing unmanned vehicle actions, exploration in unknown environments is improved, leading to efficient and stable collaborative obstacle avoidance and pursuit. Our method is validated by simulation results, which show faster convergence and a 16.7% reduction in pursuit time when compared to the twin delayed deep deterministic policy gradient (TD3) algorithm.