A novel clustering method for ship trajectory prediction

Accurate ship trajectory prediction is crucial for ensuring maritime shipping safety. Current trajectory prediction methods primarily focus on analyzing the spatial interactions between ships, often overlooking the critical role of potential destinations. Furthermore, existing methods predominantly focus on straight trajectory prediction, resulting in suboptimal performance when predicting curved trajectories, which are prevalent in real-world scenarios. To address these issues, we propose a novel method based on bidirectional LSTM and destination/source clustering, referred to as GRU-Bi-LSTM-GRU (GBG). Our approach effectively captures various ship movement patterns by clustering destination/source clustering. Then, a bidirectional LSTM classification network is used to determine the ship’s destination. Finally, the trajectory output network generates a trajectory towards the predicted destination. We evaluated the GBG model on three public datasets, and the experimental results demonstrate that it surpasses existing methods in both multi-curved and multi-straight trajectory datasets, achieving higher prediction accuracy.