Vessel Voyage Trajectory Extrapolation: Comparing the Performance of Kalman Filters

Abstract

Effective management of port terminal operations and logistics requires efficient allocation of resources for arriving ships. Predicting vessel arrival times is crucial for optimizing the allocation of resources and ensuring smooth operations. To this end, the Automatic Identification System (AIS) has emerged as a valuable source of data for vessel tracking and voyage-related information retrieval. In this study, we investigate the performance of two popular filtering algorithms, Discrete Kalman Filter (DKF) and Unscented Kalman Filter (UKF), in extrapolating the short-term (2-minute) trajectory of vessels using a Constant Velocity (CV) model. This can be useful in providing missing information needed by a vessel arrival time prediction model. Our experimental results show that the UKF and DKF perform similarly in vessel trajectory extrapolation, suggesting that the additional computational cost of sigma point sampling and propagation in the UKF may not be necessary for this application. This finding has implications for the development of vessel arrival time prediction models that rely on vessel trajectory information.