Maritime Trajectory Forecasting via CNN-SOFTS-Based Coupled Spatio-Temporal Features.

Area of Science:

• Maritime Logistics and Navigation
• Artificial Intelligence and Machine Learning
• Data Science and Analytics

Background:

• Accurate maritime trajectory forecasting is essential for navigation safety and traffic management.
• Existing methods struggle to effectively integrate spatial and temporal features from Automatic Identification System (AIS) data, especially in complex scenarios.
• Challenges include capturing intricate ship motion patterns and spatial structures.

Purpose of the Study:

• To develop a high-fidelity maritime trajectory forecasting framework by explicitly coupling spatial and temporal features.
• To enhance the prediction accuracy and computational efficiency of ship trajectory forecasting models.
• To improve inland river navigation safety and waterborne traffic monitoring capabilities.

Main Methods:

• A Convolutional Neural Network (CNN)-based spatial encoder was used to abstract spatial density distributions and learn global spatial structures.
• The Series-cOre Fused Time Series (SOFTS) model incorporated angular velocity, acceleration, and angular acceleration to capture temporal dependencies in ship motion states.
• A fusion regression module concatenated spatial and temporal features for trajectory prediction.

Main Results:

• The proposed framework achieved a Mean Squared Error (MSE) of 0.020 and a Mean Absolute Error (MAE) of 0.060 on the validation dataset.
• The method outperformed several advanced time series forecasting models in prediction accuracy and computational efficiency.
• Incorporating dynamic features (angular velocity, acceleration, angular acceleration) reduced MSE by 10.22% and MAE by 9.49%.

Conclusions:

• The developed framework effectively integrates spatial and temporal features for superior maritime trajectory forecasting.
• The inclusion of dynamic motion features significantly enhances prediction performance.
• The method holds strong potential for intelligent navigation systems, improving safety and traffic management.