基于混合核函数ν-SVR的船舶操纵运动非参数建模方法

PDF下载HTML

Title:

Nonparametric modeling method of ship maneuvering motion based on the ν-SVR with mixed kernel function

作者:

姜立超, 尚晓兵, 王伟, 张智, 李嘉祺

哈尔滨工程大学 智能科学与工程学院, 黑龙江 哈尔滨 150001

Author(s):

JIANG Lichao, SHANG Xiaobing, WANG Wei, ZHANG Zhi, LI Jiaqi

College of Intelligent Systems Science and Engineering, Harbin Engineering University, Harbin 150001, China

关键词:

非参数模型; 混合核函数; 系统辨识; ν-SVR; 船舶操纵运动; 遗传算法; SIMMAN2008; KVLCC2

Keywords:

nonparametric modeling; mixed kernel function; system identification; ν-SVR; ship maneuvering motion; genetic algorithm; SIMMAN2008; KVLCC2

分类号:

TP18; U661.33

DOI:

10.11992/tis.202310001

摘要:

非参数建模方法已广泛应用于船舶操纵运动建模。本文提出一种基于混合核函数（mixed kernel function, MK）的ν-支持向量回归（ν-support vector regression, ν-SVR）非参数建模方法，并采用遗传算法（genetic algorithm, GA）优化超参数，即遗传算法优化的混合核函数ν-支持向量回归（genetic algorithm-mixed kernel-ν-support vector regression, GA-MK-ν-SVR）。为了提高ν-SVR的性能，提出一种混合核函数，该核函数结合径向基和多项式核函数来同时捕获全局和局部性能。基于遗传算法，模型的超参数实现了优化。采用SIMMAN2008提供的KVLCC2油轮自由自航测试数据对该建模方法的性能进行了评估，并与多种现有操纵模型的预测结果进行了对比研究。试验结果表明，所提出的GA-MK-ν-SVR模型对船舶操纵运动具有良好的预测精度和较强的泛化能力。

Abstract:

Nonparametric modeling methods are widely used in modeling ship maneuvering motions. This study introduces a v-support vector regression (v-SVR) nonparametric modeling method based on a mixed kernel function (MK), optimized through a genetic algorithm (GA) called GA-MK-v-SVR. The MK aims to improve the performance of v-SVR by combining radial basis functions with polynomial kernel functions, thus capturing both global and local characteristics. A GA is employed to fine-tune the hyper-parameters. The performance of GA-MK-v-SVR was evaluated by using KVLCC2 free-running test data provided by SIMMAN 2008, and the results were compared with several maneuvering models. Experimental results show that the proposed GA-MK-v-SVR model delivers impressive prediction accuracy and robust generalization capabilities for ship maneuvering motion modeling.