Obstacle avoidance path planning for AUVs in a three-dimensional unknown environment based on the C-APF-TD3 algorithm

To enhance the obstacle avoidance path planning ability of AUV in three-dimensional unknown underwater environments with obstacle constraints, an improved algorithm combining Constrained Artificial Potential Field (C-APF) and the Twin Delayed Deep Deterministic Policy Gradient algorithm (TD3) is proposed (C-APF-TD3). First, the kinematic constraints of AUV are incorporated into the APF algorithm, allowing C-APF to generate an approximate path for the AUV. Then, the AUV obstacle avoidance path planning problem is formulated as a Markov Decision Process (MDP), designing state space, action space, and reward functions. The TD3 training is guided by the approximate path planned using C-APF, ultimately resulting in a policy model for AUV path planning. Finally, various simulation experiments are established and designed for different obstacle scenarios. The experimental results show that the C-APF-TD3 algorithm produces more optimized trajectories compared to the C-APF algorithm. Furthermore, compared to the C-APF-DDPG algorithm, it achieves a policy model with efficient control performance with higher convergence efficiency and average returns, enhancing the adaptability and robustness of AUV obstacle avoidance path planning in three-dimensional unknown environments.