Ships' Trajectories Prediction Using Recurrent Neural Networks Based on AIS Data

Abstract

The objective of this research is to develop a method for predicting the future behavior of ships and detecting anomalous behavior based on their past location coordinates and a set of context features. We use a Recurrent Neural Network model with inputs extracted from Automated Information System (AIS) data. This data includes ship coordinates, speed and course, and the ships call sign, size, and type. These features are appropriately encoded to amplify significant predictive structures within the data. The ability to automate the task of track prediction and the process of detecting anomalous ship behavior serves to increase maritime domain awareness and aid security analysts in deciding how to best allocate limited resources. Furthermore, these capabilities enable the investigation of potential threats, prevention of collisions, and planning for search-and rescue missions.

Document Details

• Document Type: Technical Report
• Publication Date: Sep 01, 2018
• Accession Number: AD1065429

Entities

People

• Shay P. Liraz

Organizations

• Naval Postgraduate School

Tags

Communities of Interest

• Ground and Sea Platforms

DTIC Thesaurus Topics

• Artificial Intelligence
• Artificial Intelligence Software
• Artificial Neural Networks
• Bayesian Networks
• Cognitive Science
• Computational Science
• Computer Languages
• Computer Programming
• Computer Science
• Data Processing
• Dimensionality Reduction
• Experimental Design
• Information Processing
• Information Science
• Information Systems
• Machine Learning
• Maritime Domain Awareness
• Network Science
• Neural Networks
• Pattern Recognition
• Probabilistic Models
• Recurrent Neural Networks
• Search And Rescue

Fields of Study

• Computer science

Readers

• Neural Network Machine Learning.
• Sensor Fusion and Tracking Systems.
• Theoretical Analysis.

Technology Areas

• AI & ML
• AI & ML - Neural Networks