A novel path following approach for autonomous ships based on fast marching method and deep reinforcement learning

Path following is one of the indispensable tools for autonomous ships, which ensures that autonomous ships are sufficiently capable of navigating in specified collision-free waters. This study proposes a novel path following approach for autonomous ships based on the fast marching (FM) method and deep reinforcement learning (DRL). The proposed approach is capable of controlling a ship to follow different paths and ensuring that the path tracking errors are always within a set range. With the help of the FM method, a grid-based path deviation map is specially produced to indicate the minimum distance between grid points and the path. Besides, a path deviation perceptron is specifically designed to simulate a range sensor for sensing the set path deviation boundaries based on the path deviation map. Afterwards, an agent is trained to control a ship following a circular path based on the DRL. Particularly, the approach is validated and evaluated through simulations. The obtained results show that the proposed method is always capable of maintaining high overall efficiency with the same strategy to follow different paths. Moreover, the ability of this approach exhibits a significant contribution to the development of autonomous ships.