Black-box modeling of ship maneuvering motion based on Gaussian process regression with wavelet threshold denoising

A system identification method based on Gaussian progress regression (GPR) combined with wavelet threshold denoising (WT) is proposed for identifying the black-box model of ship maneuvering motion. WT is applied for data preprocessing to filter out the noise in the collected data of ship motion; GPR is used for modeling with the denoised data. Two study objects are considered. One is the KVLCC1 tanker model for verifying the basic identification ability of the proposed method by utilizing the available data of zigzag tests. The other one is an unmanned surface vessel (USV) for further verifying the effectiveness and advantages of the proposed method by using the measured data of zigzag tests and turning test. For the KVLCC1, the results of the proposed method are compared with the experimental data, the results of the basic GPR model and the model identified by BPNN combined with WT. For the USV, the results of the proposed method are further compared with those of the model identified by GPR combined with lowpass filter and the model identified by BPNN combined with WT. It is shown that the proposed method is more suitable for modeling of ship maneuvering motion using limited and noisy data. •A modeling method based on Gaussian process regression (GPR) combined with wavelet threshold denoising is proposed.•The proposed method is implemented for black-box modeling of ship maneuvering with the measured data of free-running tests.•Case studies are carried out for the KVLCC tanker model and an unmanned surface vessel.•The effectiveness of the proposed method is shown by comparing its results with the test data and those of the basic GPR.•The superiority of the proposed method is demonstrated by comparing its results with those of other modeling methods.