Validation of the Ornstein-Uhlenbeck route propagation model in the Mediterranean Sea

Abstract

Traffic route analysis and prediction are both crucial to maritime security. The ability to predict a vessel position in the future is essential to provide information on upcoming events. However, accurate prediction along a route is a challenging task in the maritime domain, due to the complex nature and variability of traffic patterns. Based on the popular Ornstein-Uhlenbeck stochastic mean-reverting processes, a novel method has been recently presented that enables the accurate prediction of future positions of a vessel under the hypothesis that it is following an established traffic pattern in the area. We present a large-scale extensive validation of the Ornstein-Uhlenbeck methodology applied to target predictions along routes in the maritime domain for several classes of vessel. This validation was done using a real-world dataset recorded in the Mediterranean Sea by a network of Automatic Identification System (AIS) receivers.