Online Learning based GA-BP Neural Network to Predict Ship Trajectory

Ship’s tracking prediction is an important part of ship navigation. By using neural network to predict ship’s track can avoid building complex mathematical model of ship motion. BP neural network is easy to fall into the trap of local error minimization, and it’s global search ability of genetic algorithm can overcome the shortcomings of BP neural network which easy to fall into local optimum. Experiment is using genetic algorithm to optimize the BP neural network on the number of hidden layer nodes, weights and thresholds. Finally, by training the neural network online learning method to real-time prediction of ship trajectory simulation experiment, through the contrast experiment, prove that GA-BP neural network online learning model has high precision by using the measured data and to verify the practicability and reliability of the GA-BP prediction model. This experimental result is helpful to the research of ship automatic control and comprehensive maneuvering.