MoCo-AIS: A Contrastive Learning Framework for Similarity Computation of Vessel Trajectories

2026-06-16

AI Summary

Existing trajectory similarity methods suffer from high computational costs, while learning-based approaches often rely heavily on labeled data and exhibit limited generalization due to the absence of a unified self-supervised evaluation framework. This work proposes the first unified self-supervised contrastive learning framework tailored for vessel trajectory similarity learning. Built upon the Momentum Contrast (MoCo) mechanism, the framework constructs positive and negative sample pairs to enable deep neural networks to learn effective trajectory embeddings. Evaluated on large-scale real-world AIS datasets, the method facilitates fair multi-model comparisons without requiring labels from traditional distance metrics. It significantly outperforms existing baselines across diverse maritime navigation scenarios and establishes a reproducible benchmark platform for trajectory representation evaluation.

Abstract

Trajectory similarity is a fundamental task in analyzing mobility patterns, essential for applications such as route pattern extraction, mobility prediction, and anomaly detection. Traditional distance-based measures for computing similarity incur high computational cost, driving the adoption of lightweight learning-based approaches. Supervised methods rely on extensive labels derived from traditional distance measures and often reproduce these metrics, which limits generalization. While self-supervised learning addresses this issue through contrastive learning, it lacks a unified framework, making it difficult to compare deep learning (DL) models for consistent trajectory representation. Accordingly, this paper presents MoCo-AIS, a unified framework for learning vessel trajectory embeddings based on the Momentum Contrast (MoCo) paradigm, which formulates similarity learning through positive and negative trajectory pairs. Within this framework, we evaluate a diverse set of leading DL models on large-scale, real-world vessel-tracking AIS datasets that capture diverse navigation behaviors and operating conditions. Results demonstrate that our framework significantly improves similarity learning over existing baselines, while providing a benchmarking platform for evaluating trajectory representation models.

Problem

• trajectory similarity
• contrastive learning
• self-supervised learning
• vessel trajectories
• deep learning

Innovation

• contrastive learning
• trajectory similarity
• MoCo
• self-supervised learning
• AIS data