A Deep Learning Model for Ship Trajectory Prediction Using Automatic Identification System (AIS) Data

The rapid growth of ship traffic leads to congestion, which causes maritime accidents. Accurate trajectory prediction can improve the efficiency navigation and safety. Previous studies have focused on developing a model using deep learning approach, such as long short-term memory (LSTM) network. However, convolutional neural network (CNN) has rarely been applied extract potential correlation among different variables (e.g., longitude, latitude, speed, course over ground, etc.). Therefore, this study proposes deep-learning-based (namely, CNN-LSTM-SE) that considers temporal characteristics. This integrates CNN module, an LSTM module squeeze-and-excitation (SE) module. is utilized data relationship speed ground), capture dependencies, SE introduced adaptively adjust importance channel features focus more significant ones. Comparison experiments two cargo ships at time interval 10 s show proposed CNN-LSTM-SE obtain best performance compared with other models evaluation indexes average root mean squared error (ARMSE), absolute percentage (AMAPE), Euclidean distance (AED), ground (AGD) Fréchet (FD).