Predicting the Trajectories of Vessels Using Machine Learning
Xinglong Liu, Wei He, Jinguang Xie, Xiumin Chu
Abstract
Trajectory predicting plays an important role in port and shipping management. The Automatic Identification System (AIS) is conducive to ship navigation safety for it enables the navigation information exchanges between vessels and shore authorities. However, the AIS data get lost occasionally for the communication reliability reasons. Hence, it is necessary to predict the trajectory of vessels to enhance shipping management. In this study, an approach for predicting trajectories of vessels by utilizing historical AIS data was proposed. Firstly, an interpolation method was employed to regularize the AIS data. Accordingly, a historical AIS database composed of thousands of trajectories are created. Secondly, a method for querying similar trajectories from the historical AIS database was designed. With the similar trajectories, the lost track points can be predicted by a regression model named Least Squares Support Vector Machine (LSSVM). The minimum sum of mean square error (MSE) was treated as the optimization objective, and the Particle Swarm Optimization (PSO) was utilized to train the parameters of the regression model. In the end, the experimental results indicated that the proposed method is capable of accurately predicting the position of vessels.