A Multi-Modal Knowledge-Enhanced Framework for Vessel Trajectory Prediction

TL;DR

This work tackles vessel trajectory prediction under irregular AIS sampling and complex motion. It introduces MAKER, a multi-modal knowledge-enhanced framework that fuses large language model guidance with knowledge-based self-paced learning to capture contextual information and progressive trajectory complexity. The LKT module combines a masked sequence encoder, an LLM-guided sequence model, and a cross-modal transfer to integrate trajectory data with textual prompts, while the KSL module leverages kinematic knowledge to guide a curriculum that gradually incorporates more challenging patterns. Experiments on two real vessel datasets show MAKER outperforms seven state-of-the-art methods by 12.08%–17.86% in MAE, demonstrating improved robustness to irregular sampling and better generalization to diverse trajectory patterns.

Abstract

Accurate vessel trajectory prediction facilitates improved navigational safety, routing, and environmental protection. However, existing prediction methods are challenged by the irregular sampling time intervals of the vessel tracking data from the global AIS system and the complexity of vessel movement. These aspects render model learning and generalization difficult. To address these challenges and improve vessel trajectory prediction, we propose the multi-modal knowledge-enhanced framework (MAKER) for vessel trajectory prediction. To contend better with the irregular sampling time intervals, MAKER features a Large language model-guided Knowledge Transfer (LKT) module that leverages pre-trained language models to transfer trajectory-specific contextual knowledge effectively. To enhance the ability to learn complex trajectory patterns, MAKER incorporates a Knowledge-based Self-paced Learning (KSL) module. This module employs kinematic knowledge to progressively integrate complex patterns during training, allowing for adaptive learning and enhanced generalization. Experimental results on two vessel trajectory datasets show that MAKER can improve the prediction accuracy of state-of-the-art methods by 12.08%-17.86%.

Figures (5)

• Figure 1: Challenges to vessel trajectory prediction.
• Figure 2: Multi-modal knowledge-enhanced framework for vessel trajectory prediction.
• Figure 3: The textual prompt, with underlined spaces indicating where sample-specific data is inserted.
• Figure 4: Multi-modal knowledge transfer component.
• Figure 5: Knowledge-based self-paced learning (KSL) module