Maritime traffic congestion identification and ship trajectory prediction using temporal graph convolutional...

Area of Science:

• Maritime logistics and safety
• Artificial intelligence in transportation
• Spatiotemporal data analysis

Background:

• Global maritime trade growth necessitates efficient traffic management.
• Current methods struggle with complex spatial and temporal dynamics.
• Accurate ship trajectory prediction and congestion identification are crucial.

Purpose of the Study:

• To develop a comprehensive framework for ship trajectory prediction and maritime traffic congestion identification.
• To enhance maritime safety and operational efficiency through proactive warnings.
• To leverage Automatic Identification System (AIS) data for improved traffic management.

Main Methods:

• Integration of spatiotemporal analysis with deep learning.
• Development of a Temporal Graph Convolutional Network (T-GCN) model combining Graph Convolutional Networks (GCN) and Gated Recurrent Units (GRU).
• Introduction of a congestion measurement indicator based on the Speed Performance Index (SPI).

Main Results:

• The T-GCN model effectively captures spatial dependencies and temporal dynamics of maritime traffic.
• Accurate ship trajectory prediction was achieved.
• The SPI-based indicator successfully quantifies and identifies maritime route congestion.
• Proactive congestion warnings were enabled.

Conclusions:

• The proposed framework significantly improves ship trajectory prediction accuracy.
• The method enables effective identification and warning of maritime traffic congestion.
• This approach contributes to enhanced maritime safety and operational efficiency.