Context-enhanced vessel prediction based on Ornstein-Uhlenbeck processes using historical AIS traffic patterns: Real-world experimental results

Traffic route analysis and prediction are both essential for maritime security. Specifically, the prediction of a vessel position is useful to provide alerts about upcoming events (e.g., opportunities and threats). However, accurate prediction along a route in the maritime domain is a challenging task, due to the complex and dynamic nature of traffic patterns. This paper presents a vessel prediction method, based on the popular Ornstein-Uhlenbeck stochastic processes, whose parameters are estimated from historical patterns of life. The historical traffic routes are obtained by pre-processing Automatic Identification System (AIS) data via the CMRE tool called Traffic Route Extraction for Anomaly Detection (TREAD). These recurrent routes allow prediction of the position of a vessel that is following one of these routes, surprisingly, by several hours. The method is validated using a case study related to the second data campaign of the EC FP7 Project New Service Capabilities for Integrated and Advanced Maritime Surveillance (NEREIDS)1. We demonstrate that the prediction accuracy is well represented by the Ornstein-Uhlenbeck model.