AUV Path Following Control using Deep Reinforcement Learning Under the Influence of Ocean Currents

Path following control of Autonomous underwater Vehicles (AUVs) is considered an important and challenging task, especially when AUVs are designed as underactuated systems with saturated inputs, and the ocean currents and disturbances in the environments are considered. Based on this premise, this work proposes a novel path following control approach based on deep reinforcement learning (DRL). Different from the existing DRL-based control methods, where only one single controller is used to control the motion of AUV, our proposed control approach consists of two subsystems, for controlling the velocity and steering of AUV, independently. Specifically, the deep deterministic policy gradient (DDPG) algorithm is adopted to achieve satisfactory control in the surge velocity control subsystem, and the soft actor-critic (SAC) algorithm is employed in the steering control subsystem to achieve high-accuracy path following control. Besides, we derive an enhanced Line-of-Sight (LOS) guidance method to make the AUV generate a drift angle to compensate for the sideslip induced by sway velocity of AUV and ocean currents. The numerical simulation results validate the advantages of the proposed DDPG-SAC-based control approach in great generalization ability, improved anti-interference capability and high robustness.