An unsupervised learning based simplification on ship motion model and its verification

In this paper, simplification method for ship model is proposed. Firstly, sensitivity index with a depressive strategy is introduced to characterize the significance of hydrodynamic coefficients for higher accuracy. Secondly, an unsupervised learning based simplification is proposed and can properly and effectively reduce the hydrodynamic coefficients. Thirdly, open-loop and closed-loop simulation are carried out to verify that simplification. Experiment result shows that clustering are effective in horizontal rotational movement and horizontal zigzag maneuver with the maximum and minimum errors of the motion parameters are 4.75% and 0.24% respectively within acceptable range.