Trajectory prediction for ocean vessels base on K-order multivariate markov chain

Abstract

Trajectory prediction is a key problem in MDTN (Mobile Delay Tolerant Network). Because Vessel’s moving pattern is in free space and easily influenced by the fish moratorium, tide, weather, etc., it brings new challenges in free-space vessel trajectory prediction. In addition, the trajectory characteristics of a vessel are different from that on land, causing traditional trajectory prediction method can’t be directly used in ocean domain. To solve the problem above, we propose a novel trajectory prediction algorithm for ocean vessel called TPOV. We utilize k-order multivariate Markov Chain and multiple sailing related parameters to build state-transition matrixes. Through simulations and experiments on two-year trajectory data of two thousand vessels, we provide quantitative analysis of the proposed strategy. The results show that TPOV has high precision prediction with a minor error.