MIMO non-parametric modeling of ship maneuvering motion for marine simulator using adaptive moment estimation locally weighted learning

Abstract

This work explores an adaptive moment estimation locally weighted learning (AME-LWL) method to develop a novel high-precision non-parametric modeling technology for ship maneuvering motion and conducts full-scale tests. First, a non-parametric learning framework is used to avoid unmodeled dynamics and parameter drift; second, Tikhonov method and multi-innovation adaptive moment estimation (AME) algorithm are proposed for local model adaptive learning; third, considering rudder and propeller joint effect for hydrodynamic forces, a multiple-input multiple-output (MIMO) continuous-time model is established to realize ship motion dynamic simulation under interferences. This scheme leads to a high-precision continuous-time non-parametric model that can be easily implemented, robust, less time-consuming and insensitive to initial parameters. In order to verify the properties of AME-LWL non-parametric model, ‘Mariner’ vessel and ‘Delta Linda’ tug two ship types were taken as objects for simulation experiments of comprehensive maneuvering tests, rudder-propeller coupling tests and wind interferences test. Finally, simulation results verify the effectiveness of the proposed scheme.