On-Call Antisubmarine Path Planning for AUVs Based on an Artificial Potential Field-Enhanced MADDPG Algorithm

Abstract：Objectives To enhance the efficiency and stability of cooperative AUV detection in complex underwater environments, a cooperative search and detection path planning model for AUVs based on an improved MADDPG algorithm with the Artificial Potential Field (APF) method is proposed. Methods To address the issues of local optima in APF-based path planning and poor convergence due to blind exploration in the early stages of MADDPG, the proposed APF-MADDPG algorithm integrates the gravitational field of the APF to guide the initial movement of the AUVs. A Monte Carlo method is used to simulate multiple possible target trajectories and statistically analyze the distribution of target positions at different times to predict the dispersion pattern of dynamic underwater targets. Additionally, the detection probability at various sonar ranges and the cumulative detection probability (CDP) formula are used as path evaluation metrics. The proposed algorithm is employed to simulate cooperative detection scenarios involving two and three AUVs. Results Simulation results show that the APF-MADDPG algorithm improves the cumulative detection probability (CDP) by 7 percentage points, reaching 80.93% in the two-AUV cooperative detection scenario, and by 0.6 percentage points, reaching 92.67%, in the three-AUV scenario, compared to the original MADDPG. Conclusions The APF-MADDPG algorithm effectively enhances the detection efficiency and stability in AUV cooperative search and detection tasks, providing valuable insights for improving the cooperative operational capabilities of AUVs in underwater environments. Future research could further explore the performance of other deep reinforcement learning algorithms in similar search and detection scenarios to further improve the efficiency and cooperative capabilities of multi-AUV systems.