Impact of the Time Window Length on the Ship Trajectory Reconstruction Based on AIS Data Clustering

Abstract

AbstractAutomatic identification system allows ships to automatically exchange information about themselves, mainly to avoid collisions between them. Its terrestrial segment, in which the communication is synchronized, has a range of only 74 km (40 nautical miles). To broaden this range, a satellite segment of the system was introduced. However, due to the fact that one satellite operates on multiple internally synchronized areas, the communication within satellite’s field of view suffers from message collision issue when two or more messages are received by satellite at the same time. This paper presents the first step of the approach of reconstructing lost or damaged AIS messages due to collision. The computational experiment was to cluster data (one cluster = messages from one ship) so that they can be further analysed to find abnormal ones among them and correct them. The clustering was conducted in a streaming approach, using a specific time window to mimic the time-changing character of AIS data. Numerical results that measure the clustering quality have been collected and are presented in the paper. The computational experiment proves that clustering might indeed help reconstruct AIS data and shows the best time window length from a clustering point of view.KeywordsAISAIS data analysisClusteringTrajectory reconstruction