Design of USV Navigation System Based on Lidar Image-Satellite Map Fusing

The real-time point-cloud Image from shipborne Lidar scanning in conjunction with satellite map is used to construct the environment water area with shore baseline, which provides USV (Unmanned Surface Vehicle) with indication of feasible area or path, supports the design of control system. The USV navigation system with DQN (Deep Q-Learning Network) is trained in a set water environment to tune its parameters and optimize its experience pool. The trained USV is able to autonomously navigate in unfamiliar waters and avoid obstacles. Tests are carried out and good results achieved:error of 3.05 m in open water area and 3.81 m in narrow water area with swift current.