AIS-Based Vessel Trajectory Prediction Using Memory-Augmented Neural Networks

2026-06-06 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Transportation & Mobility · Depth: Expert, extended

Summary

An empirical investigation demonstrates that memory-augmented neural networks (MANNs) substantially enhance AIS-based vessel trajectory prediction. Researchers at KAIST adapted MANTRA, a seminal MANN, by incorporating Speed Over Ground (SOG) and Course Over Ground (COG) into its input features. Experiments using Automatic Identification System (AIS) data from the Gulf of Mexico (130,126 records, 369 vessels) and the New York Bight (22,741 records, 55 vessels) showed significant performance gains. MANTRA reduced Average Displacement Error (ADE) by up to 46.4% and Final Displacement Error (FDE) by up to 54.7% on the Gulf of Mexico dataset, and up to 33.3% (ADE) and 27.7% (FDE) on the New York Bight dataset, consistently outperforming six deep learning baselines across multiple prediction horizons.

Key takeaway

For maritime operations analysts or AI scientists developing navigation systems, this research indicates that integrating memory-augmented neural networks can significantly improve vessel trajectory prediction accuracy. You should consider adopting models like MANTRA, especially when dealing with diverse vessel behaviors, and ensure your input features include Speed Over Ground and Course Over Ground for optimal performance in collision avoidance and route optimization.

Key insights

Memory-augmented neural networks effectively predict vessel trajectories by storing and retrieving instance-level movement patterns.

Principles

• External memory improves trajectory prediction accuracy.
• SOG and COG are critical vessel motion features.
• Memory-based models capture diverse movement patterns.

Method

MANTRA uses an autoencoder to create past/future trajectory encodings. A memory controller selectively stores novel encoding pairs. Prediction involves retrieving top-K similar future encodings based on cosine similarity with the current past encoding.

In practice

• Integrate SOG/COG as input features for vessel models.
• Consider memory-augmented architectures for complex patterns.
• Apply cubic spline interpolation for missing AIS data.

Topics

• Vessel Trajectory Prediction
• Memory-Augmented Neural Networks
• Automatic Identification System
• MANTRA Model
• Maritime Navigation
• Deep Learning Baselines

Code references

• Marchetz/MANTRA-CVPR20

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Robotics Engineer

Related on AIssential

• AIS-Based Vessel Trajectory Prediction Using Memory-Augmented Neural Networks — Memory-augmented neural networks significantly improve vessel trajectory prediction using AIS data.
• Online Goal Recognition using Path Signature and Dynamic Time Warping — Path signatures offer a compact, expressive way to encode trajectories for online goal recognition.
• Forecasting Downstream Performance of LLMs With Proxy Metrics — Proxy metrics from expert trajectories reliably forecast LLM downstream performance, outperforming traditional loss and compute baselines.
• Titans: Learning to Memorize at Test Time — Training a dedicated neural memory module at inference time improves long-context sequence modeling on novel data.
• News - NSWCPD Applies AI and Machine Learning to Predictive Machinery Health - DVIDS — AI/ML-driven vibration analysis can predict mechanical failures, enhancing condition-based maintenance for naval platforms.
• Structural Grid Descriptors Predict Within-Task Solver Success on ARC-AGI — Structural grid descriptors at 50% trajectory completion reliably predict ARC-AGI solver success, enabling significant compute reductions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.