Unified Multimodal Vessel Trajectory Prediction with Explainable Navigation Intention

2025-11-18

AI Summary

Existing short-term multimodal trajectory prediction methods for vessels exhibit poor adaptability and weak interpretability in complex maritime environments. Method: This paper proposes a unified multimodal prediction framework that jointly models sustainable and transient navigation intentions. It introduces explicit, interpretable navigation intention modeling into vessel trajectory prediction for the first time—constructing a history-driven persistent intention tree and a conditional variational autoencoder (CVAE)-based transient intention model, fused via non-local attention to ensure cross-modal global consistency. Contribution/Results: Evaluated on real-world AIS datasets, the method significantly reduces average displacement error (ADE) and final displacement error (FDE), demonstrates strong cross-scenario generalization, and enhances semantic interpretability by outputting structured navigation intentions—thereby bridging the gap between low-level trajectory predictions and high-level navigational semantics.

Technology Category

• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Vessel trajectory prediction is fundamental to intelligent maritime systems. Within this domain, short-term prediction of rapid behavioral changes in complex maritime environments has established multimodal trajectory prediction (MTP) as a promising research area. However, existing vessel MTP methods suffer from limited scenario applicability and insufficient explainability. To address these challenges, we propose a unified MTP framework incorporating explainable navigation intentions, which we classify into sustained and transient categories. Our method constructs sustained intention trees from historical trajectories and models dynamic transient intentions using a Conditional Variational Autoencoder (CVAE), while using a non-local attention mechanism to maintain global scenario consistency. Experiments on real Automatic Identification System (AIS) datasets demonstrates our method's broad applicability across diverse scenarios, achieving significant improvements in both ADE and FDE. Furthermore, our method improves explainability by explicitly revealing the navigational intentions underlying each predicted trajectory.

Problem

• Predict multimodal vessel trajectories in complex maritime environments
• Address limited scenario applicability in existing prediction methods
• Enhance explainability by revealing navigational intentions behind predictions

Innovation

• Classifies navigation intentions into sustained and transient categories
• Uses CVAE to model dynamic transient intentions
• Employs non-local attention for global scenario consistency