Black-Box Modeling of Ship Maneuvering Motion Based on Gaussian Process Regression with Wavelet Threshold Denoising

Abstract

A system identification method based on wavelet threshold denoising and Gaussian progress regression (WT-GPR) is proposed for identifying the black-box model of ship maneuvering motion. Wavelet threshold (WT) denoising is applied in data pre-processing to filter out the noise from the collected data of ship maneuvering motion. Gaussian process regression (GPR) is used for black-box modeling of ship maneuvering motion by using the denoised data, which improves the prediction accuracy and robustness of the identified black-box model. The proposed method is validated by utilizing the real measured data of 20°/20° zigzag tests with an unmanned surface vessel (USV) and those of 20°/20° and 10°/10° zigzag tests with the KVLCC ship model provided by SIMMAN2008 Workshop. Using the identified black-box models, zigzag tests and turning circle maneuvers of the USV, and zigzag tests of the KVLCC ship model are predicted. The prediction results are compared with the real measured data and the results of the basic model identified by GPR. It is shown that the black-box models identified by the proposed method have higher prediction accuracy and better generalization ability, and the proposed method is more suitable for modeling of ship maneuvering motion using the data with noise.

Keywords: ship maneuvering, system identification, black-box modeling, wavelet threshold denoising, Gaussian process regression