eyesofworld/AIS_Dataset

AIS Data-Driven Maritime Monitoring Based on Transformer

Paper Overview

This paper provides a comprehensive review of AIS data-driven maritime monitoring based on Transformer models, focusing on key techniques such as vessel trajectory prediction, vessel behavior detection, and vessel behavior prediction. With the increasing demands for safety, efficiency, and sustainability in global shipping, the application of AIS data plays a crucial role in enhancing maritime monitoring. With its advantage in handling long sequence data and complex temporal dependencies, the Transformer model has proven to be an effective tool for processing AIS data.

Paper Overview

If you find this work useful, please cite our paper:

• Paper Title: AIS Data-Driven Maritime Monitoring Based on Transformer: A Comprehensive Review
• Authors: Zhiye Xie, Enmei Tu, Xianping Fu, Guoliang Yuan, Yi Han
• arXiv: arXiv Link
• Published: International Joint Conference on Neural Networks (IJCNN), 2025

@inproceedings{xie2025ais,
    title={AIS Data-Driven Maritime Monitoring Based on Transformer},
    author={Xie, Zhiye and Tu, Enmei and Fu, Xianping and Yuan, Guoliang and Han, Yi},
    booktitle={2025 International Joint Conference on Neural Networks (IJCNN)},
    year={2025},
    url={https://arxiv.org/abs/2505.07374}
}

Research Focus

• Vessel Trajectory Prediction: Explores Transformer-based trajectory prediction methods, including generative and classification-based approaches.
• Vessel Behavior Detection: Evaluates Transformer-based methods for detecting vessel behaviors, focusing on anomaly detection and risk management.
• Vessel Behavior Prediction: Analyzes techniques for predicting future vessel states and navigational behaviors.

Datasets

This study utilizes multiple publicly available AIS datasets from the reviewed papers, which have been filtered, cleaned, and analyzed statistically. The datasets cover various types of vessels and provide critical information about vessel locations, speeds, and headings.

• The dataset includes AIS messages from 19,185 vessels, totaling approximately 640 million records.

Repository Structure

This repository contains the following:

• Datasets: The cleaned and filtered AIS dataset.
• Paper: Links to the full paper.

Acknowledgments

Thanks to Marine Cadastre Hub for providing the raw AIS data, and thank you all for your interest in this work!

---

license: apache-2.0