Ship Trajectory Prediction Based on Bidirectional Long Short-Term Memory

Abstract

In recent years, with the development of global trade, the shipping industry has become increasingly busy, and maritime traffic accidents have also increased. It is very important to ensure maritime traffic safety. Ship trajectory prediction is an important means for ships to avoid collisions and other accidents when sailing at sea. Bi-LSTM, a deep learning model, exhibits remarkable memory and control of information flow, as well as the capacity to manage two-way context information. It can be used in ship trajectory prediction to improve the accuracy of prediction results. In this paper, three models of RNN, LSTM and Bi-LSTM are constructed, and they are trained and tested by ship AIS data. The Bi-LSTM model's prediction results were the most precise of the three models, and its predictions were closest to the actual trajectory, as evidenced by the ultimate experimental results.