Research on Global Ship Trajectories and Parameter Prediction by Grid-LSTM-GMM

This paper proposes the Grid-LSTM-GMM ship trajectory prediction model for the global trajectory prediction problem. First, voyage segmentation and data preprocessing are performed on existing ship trajectory data. A hierarchical clustering method based on two-dimensional grid-based route feature analysis is employed to construct a ship trajectory database. Secondly, a trajectory matching method based on LSTM-GMM is proposed to generate a trajectory classifier. Finally, a GMR ship trajectory data prediction model assisted by the trajectory archive is proposed. The model's validity was verified by designing comparative and ablation experiments. Experimental results demonstrate that the Grid-LSTM-GMM model achieves significant improvements in prediction performance compared to traditional models, and the proposed model enhancements effectively boost prediction accuracy.