DOI: 10.18178/wcse.2019.06.139
Research of Ship Autopilot Rudder Based on Deep Belief Network
Abstract— In order to improve the control precision of the existing ship autopilot and improve the adaptive
capability of the autopilot, an autopilot control algorithm based on the deep confidence network (DBN) is
proposed. First of all, using the contrast divergence algorithm and the data recorded in the examination
system of the Shanghai Maritime University, the constrained Boltzmann machines (RBMs) that make up
each DBN are pre-trained in turn, and the results are used as the depth nerve Network weight of the initial
value. On this basis, the back propagation algorithm is used to fine-tune the multi-layer depth structure. The
simulation results show that the simulated sailing error between this method and the master captain is only
5.2%.
Index Terms— Autopilot Rudder, Deep Neural Networks, CD Algorithm, RBM, BP, Training
Li Shaowei
School of Mathematics and Computer Science, Jianghan University, CHINA
Wang Shengzheng
Merchant Marine College, Shanghai Maritime University, CHINA
Cite: Li Shaowei, Wang Shengzheng, "Research of Ship Autopilot Rudder Based on Deep Belief Network," Proceedings of 2019 the 9th International Workshop on Computer Science and Engineering, pp. 928-933, Hong Kong, 15-17 June, 2019.