Novel TD3 Based AUV Path Tracking Control

This paper investigates the path tracking problem of an autonomous underwater vehicle (AUV) at constant depth. For traditional methods cannot effectively solve problems such as strong coupling of AUV, we propose a method based on Twin Delayed Deep Deterministic policy gradient (TD3) and S-plane algorithm, named as S-TD3. We use S-plane for guided exploration and pre-train the actor neural network with the obtained data. Then we use the ε — greedy method to select the training strategy during the training. Finally, we illustrate with simulations that our method fits the desired path well and has strong robustness.