| High-precision vessel track prediction has a very high practical value in both civil and military applications.The vessel track prediction can assist the ship to avoid the collision risk in civil field,it is also able to complete the target’s fixed-point forecast in the military field.Vessel is disturbed by complex marine environments when it navigates on the ocean,which including current,sea breeze,wave,etc.However,current vessel track prediction methods rarely consider these factors,which simply use machine learning methods or simplified kinematic model to estimate the future vessel track,and its prediction accuracy is low.Therefore,in order to improve the accuracy of the vessel track prediction,it is necessary to study an effective and high-precision vessel track prediction method based on the marine environment during the vessel sailing process.Firstly,this paper briefly introduced the purpose and significance of the vessel track prediction and considering the marine environment interference in the prediction process,analyzed the research status and progress of the marine environments disturb vessel motion and track prediction methods at home and abroad.This paper established the mathematical model of the vessel’s motion,it determined the marine environmental factors affecting the vessel’s motion mainly included ocean currents,sea breeze and ocean waves,analyzing the impact process and mathematical model of those factors,then the disturbance force and torque of the sea breeze were simulated.It also predicted vessel’s track based on uniform linear motion.Secondly,this paper analyzed the current mainstream model-free track prediction methods,comparing the performance of RBF and BP neural network on track prediction,it found that BP neural network is more suitable for ship motion model.Based on this conclusion,the interference of sea breeze and current was added to BP track prediction model for improving prediction accuracy.Then the genetic algorithm was used to optimize the BP prediction model parameters to avoid falling into local minimum values.Thirdly,in order to predict the position of vessel at multiple times in the future,this paper proposed the prediction method based on neural network combined model.It predicted the speed and heading of vessel at multiple times in the future with the Long Short-Term Memory(LSTM)model.Then the GA-BP neural network proposed could predict the positions of vessel at multiple times in the future based on the results of Long Short-Term Memory(LSTM)model.Finally,the simulation experiments were carried out based on the data of this subject.It dealt with the vessel navigation data and the marine environment data by interpolation method in order to satisfy the experimental requirements.After the simulation comparison,it found that the BP model considering the marine environment interference is 23% higher than the BP model without considering the marine environment interference on accuracy.The prediction accuracy of GA-BP is 13% higher than that of the traditional BP model.Finally,the simulation experiment of the neural network combined model showed that when the position error is within 500 meters,the track of vessel in the next 15 minutes can be predicted. |
