Identification of Unmanned Surface Vehicle Maneuverability Parameters Based on Improved WLSSVM

Abstract

A cosine processing method was designed to achieve high-precision identification modeling of unmanned surface vehicle(USV) maneuvering motion and address the issue that some parameters will be inaccurately identified when the second-order nonlinear response model of USVs is identified by least square support vector machine(LSSVM). On this premise, a weighted LSSVM(WLSSVM) algorithm that could optimize the weight was proposed. The algorithm was based on the idea of data weighting and used the adaptive particle swarm optimization technique with a mutation approach. Based on simulation data and actual ship data, the identification results indicate that the model after cosine reconstruction effectively handles the problem of inaccurate parameter identification. At the same time, the WLSSVM with optimized weights has better prediction accuracy for parameter identification modeling. The research findings can serve as a reference for high-precision parameter identification modeling of USV maneuvering motion.

Keywords

• unmanned surface vehicle
• motion parameter identification
• least squares support vector machine
• adaptive particle swarm