An efficient neural-network based approach to automatic ship docking
Shuai, Yonghui; Li, Guoyuan; Cheng, Xu; Skulstad, Robert; Xu, jinshan; Liu, Honghai; Zhang, Houxiang
Journal article, Peer reviewed
Accepted version
View/ Open
Date
2019Metadata
Show full item recordCollections
Original version
https://doi.org/10.1016/j.oceaneng.2019.106514Abstract
Automatic ship docking is one of the applications of autonomous ships. How to realize autonomous low-speed maneuver under environmental disturbances for docking is the fundamental problem at present. This paper presents an efficient approach based on artificial neural network (ANN) for automatic ship docking. The problem is formulated and well-modeled for simulating ship docking operation. A joystick implementation in simulation provides manual maneuvering and thus enables collection of sufficient and reliable data from successful maneuvers. To keep consistent with the manual control, an ANN with two parallel structure is proposed to control the ship’s thrust and rudder, respectively. Feature selection technique and genetic algorithm (GA) are utilized to optimize the structure and reduce the training cost. Numerical simulations under different environmental disturbances, including no wind, constant wind and dynamic wind are carried out. The results show the ship is able to reach the dock smoothly, which confirms the effectiveness of the proposed approach.