Prediction of Ship Trajectories in confined waters Based on N-gram Model

Addressing the issue of low accuracy in current ship trajectory prediction methods within complex maritime areas, this paper explores an improved N-gram model-based approach for ship trajectory prediction. The method utilizes historical grid-based trajectory data of ships, segments the grid-based trajectory sequences into sub-sequences of different lengths to create a grid-based trajectory database. Through the use of fuzzy queries, an N-gram model is constructed without the need for a corpus. Future ship motion direction and position are predicted iteratively. Experimental results demonstrate that, in the short-term ship trajectory prediction context, when input parameters involve 10 grids and predict 50 grid points, the method based on the improved N-gram model exhibits higher accuracy compared to the current mainstream LSTM and GRU models. The predictive accuracy is improved by 37.5% and 42.8%, respectively. This research holds significant academic and practical relevance for in-depth exploration of ship navigation in complex maritime environments.