Three-Dimensional Path-Following Control of an Autonomous Underwater Vehicle Based on Deep Reinforcement Learning

In this article, a deep reinforcement learning based three-dimensional path following control approach is proposed for an underactuated autonomous underwater vehicle (AUV). To be specific, kinematic control laws are employed by using the three-dimensional line-of-sight guidance and dynamic control laws are employed by using the twin delayed deep deterministic policy gradient algorithm (TD3), contributing to the surge velocity, pitch angle and heading angle control of an underactuated AUV. In order to solve the chattering of controllers, the action filter and the punishment function are built respectively, which can make control signals stable. Simulations are carried out to evaluate the performance of the proposed control approach. And results show that the AUV can complete the control mission successfully.