A Novel CNN-LSTM Method for Ship Trajectory Prediction

With the rapid development of the shipping industry, the traffic volume increases rapidly, which may leads to traffic risks such as collisions, stranding, running on rocks. Therefore, accurate ship trajectory prediction is of great significance. In this paper, we construct a novel CNN-LSTM trajectory prediction model, using the trajectory data recorded by the Automatic Identification System (AIS) as the training and testing sets. In the data aspect, we propose a new data processing scheme to partition AIS latitude and longitude data into neatly arranged regions. In the model aspect, we introduce the residual block and attention mechanism to the basic CNN-LSTM model. Moreover, the Ranger optimizer is applied to the optimization process of the model's loss function. The practicability and effectiveness of our model in ship trajectory prediction are verified by model evaluation. This not only has important application value for ships to travel safely and efficiently in complex navigable environments, but also for maritime traffic management.