L-VTP: Long-Term Vessel Trajectory Prediction Based on Multi-Source Data Analysis.

Area of Science:

• Marine Internet of Things (IoT)
• Mobile Delay Tolerant Networks (MDTN)
• Vessel Trajectory Prediction

Background:

• Ocean Mobile Delay Tolerant Networks (MDTN) are rapidly developing, making long-term trajectory prediction crucial.
• Existing on-land trajectory prediction methods are unsuitable for ocean vessels due to unique mobility patterns and environmental influences.
• A lack of map topology support and factors like fish moratoriums and sunshine duration complicate ocean vessel trajectory prediction.

Purpose of the Study:

• To propose a novel long-term, fine-grained trajectory prediction algorithm for ocean vessels, named L-VTP.
• To address the limitations of existing methods in the context of ocean vessel mobility.
• To incorporate privacy considerations into the trajectory prediction model.

Main Methods:

• Utilized multiple sailing-related parameters and K-order multivariate Markov Chains to build state-transition matrices.
• Implemented quantitative uncertainty analysis to manage trajectory sparsity and state missing problems.
• Incorporated distinct day and night mobility models and a privacy quantitative model for trade-offs.

Main Results:

• L-VTP achieves fine-grained long-term trajectory prediction for ocean vessels with privacy considerations.
• Demonstrated an average error of less than 500 m for 4.5-hour predictions using two years of real-world data.
• Extended prediction capability to 10 hours with an average error of 2.16 km, well within communication ranges.

Conclusions:

• L-VTP effectively overcomes the challenges of ocean vessel trajectory prediction, including data sparsity and environmental factors.
• The proposed method offers a practical solution for long-term, accurate, and privacy-aware trajectory forecasting in marine IoT.
• The algorithm's performance validates its suitability for applications in ocean Mobile Delay Tolerant Networks.