Research on Cargo Ship Trajectory Based on Time Window Smoothing Filter Algorithm and Transformer Model

Abstract

With the flourishing development of maritime transportation, the safety of cargo ship navigation cannot be overlooked. To ensure the safety of cargo ship navigation and improve the prediction accuracy of cargo ship trajectories, the Transformer model is utilized to predict cargo ship navigation trajectories, enabling the anticipation of spatial and temporal location information of cargo ships in the near future. This facilitates proactive operations and early warnings, thereby ensuring the safety of cargo ship navigation. Previously, predictions of cargo ship navigation trajectories were mostly based on statistical and traditional machine learning methods, which exhibited poor model adaptability and high constraints, making them unsuitable for the current complex and thriving maritime traffic patterns. In the modern era, the rapidly developing Automatic Identification System (AIS) enables researchers to obtain vast amounts of cargo ship trajectory data, providing favorable conditions for mining data characteristics of cargo ships. Therefore, this paper, based on AIS data, combines the time window smoothing filter algorithm and employs the Transformer model to predict, research, and analyze cargo ship navigation trajectories. Furthermore, an early warning mechanism for cargo ships is established to alert to deviations from the planned route, thereby ensuring the safety of cargo ship navigation.