Nonparametric modeling of ship maneuvering motion in waves based on Gaussian process regression

Abstract

A robust and easy-to-operate method based on Gaussian process regression (GPR) is proposed for nonparametric modeling of ship maneuvering in waves. The wave parameters that are difficult to measure are not needed when identifying the dynamic model. In order to extract the wave effect from the ship motion data, the black-box model is modified relative to that used for calm water case. Taking the KCS container ship and the S-175 container ship as study objects, the nonparametric models based on GPR are identified by utilizing the experimental data of turning circle maneuver of KCS model and zigzag maneuvers of S-175 model in regular waves. Using the identified nonparametric models, turning circle maneuver of KCS model and zigzag maneuvers of S-175 model in regular waves are predicted, and the prediction results are compared with the model test data to verify the proposed method of nonparametric modeling and to evaluate the performances of the identified nonparametric models. It shows that the trajectory of turning circle maneuver and the heading angle of zigzag maneuvers are well predicted, indicating that the nonparametric model based on GPR with the modified black-box model can predict the maneuvering motion in regular waves accurately.