Kinematic motion models based vessel state estimation to support advanced ship predictors

Abstract

Advanced ship predictors can generally be considered as a vital part of the decision-making process of autonomous ships in the future, where the information on vessel maneuvering behavior can be used as the source of information to estimate current vessel motions and predict future behavior precisely. As a result, the navigation safety of autonomous vessels can be improved. In this paper, vessel maneuvering behavior consists of continuous-time system states of two kinematic motion models—the Curvilinear Motion Model (CMM) and Constant Turn Rate & Acceleration (CTRA) Model. Two state estimation algorithms—the Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) are implemented on these two models with certain modifications so that they can be compatible with discrete-time measurements. Four scenarios, created by combining different models and algorithms, are implemented using simulated ship maneuvering data from a bridge simulator. These scenarios are then verified through the proposed stability and consistency tests. The simulation results show that the EKF tends to be unstable combined with the CMM. The estimates from the other three scenarios can generally be considered more stable and consistent, unless sudden actions or variations in vessel heading occurred during the simulation. The CTRA is also proven to be more robust compared to the CMM. As a result, a suitable combination of mathematical models and estimation filters can be considered to support advanced ship predictors in future ship navigation.