满秩分解最小二乘法船舶航向模型辨识  

Ship heading model identification based on full rank decomposition least square method

作　　者：包政凯 朱齐丹 刘永超 BAO Zhengkai;ZHU Qidan;LIU Yongchao(College of Intelligent Systems Science and Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150001

出　　处：《智能系统学报》2022年第1期137-143,共7页CAAI Transactions on Intelligent Systems

基　　金：绿色智能内河船舶创新专项(MC-202002-C01);国家自然科学基金项目(52171299)。

摘　　要：为了解决标准遗忘因子最小二乘法在线辨识船舶航向模型参数漂移和发散问题,考虑到船舶在实际航行中存在海洋环境扰动和数据欠激励的情况,提出并验证了一种基于满秩分解的递推最小二乘法。用实船数据进行船舶航向模型参数辨识,将辨识结果与标准遗忘因子最小二乘算法、多新息最小二乘法、最小二乘支持向量机的辨识结果进行对比,验证了满秩分解有效降低了在线辨识过程中扰动导致的参数漂移并成功抑制了参数的发散,提升了遗忘因子最小二乘法的辨识精度,减小了最小二乘法对持续数据激励的依赖。In order to solve the problem of parameter drift and divergence in the on-line identification of the ship heading model by the forgetting factor least squares method,considering the marine environment disturbance and data under-excitation in actual navigation,we propose a forgetting factor recursive least square algorithm based on full rank decomposition,which uses the ship navigation data to identify the ship heading model parameters,and compare the identification results with the identification results of the standard forgetting factor least squares algorithm,multi-innovation least square algorithm,least square support vector algorithm.The comparison result shows that the full rank decomposition method can effectively reduce the parameter drift caused by the disturbance in the online identification process and successfully suppress divergence of the parameters,improve the accuracy of the forgetting factor least square algorithm and reduce the dependence of the least square method on continuous data excitation.

关 键 词：遗忘因子最小二乘法 数据欠激励 船舶航向模型 满秩分解 参数辨识 海洋环境扰动 参数辨识收敛性 实船航行数据 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] U661.3[自动化与计算机技术—控制科学与工程]