水下动基座初始对准中的鲁棒自适应UKF方法  

Robust adaptive UKF approach for underwater moving base initial alignment

作　　者：朱兵 许江宁[1] 吴苗[1] 卞鸿巍[1] 李京书[1] Zhu Bing;Xu Jiangning;Wu Miao;Bian Hongwei;Li Jingshu(College of Electrical Engineering, Naval University of Engineering, Wuhan 430033, Chin)

机构地区：[1]海军工程大学电气工程学院

出　　处：《仪器仪表学报》2018年第2期73-80,共8页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(41574069);国家重大科学仪器开发专项(2011YQ12004502);国家重点研发计划(2016YFB0501700,2016YFB0501701);湖北省自然科学基金(2017CFB590)项目资助

摘　　要：捷联式惯导水下动基座初始对准中,当量测噪声具有不确定、非高斯的统计特性时,对量测噪声作高斯分布假设和对量测噪声阵R作常值处理会造成无迹卡尔曼滤波（UKF）精度和鲁棒性变差。针对此问题,提出了一种基于投影统计（PS）算法的鲁棒自适应UKF（ARUKF）方法。方法首先利用PS算法确定存储新息的权值,对异常的新息进行重加权;而后利用MyersTapley方法自适应地估计R阵;最后利用Huber方法中的权函数对估计出来的R阵进行修正。基于江试试验的实测数据,利用UKF、鲁棒UKF（RUKF）和ARUKF进行非高斯量测噪声条件下的动基座初始对准实验,结果表明：当观测量受到野值的污染时,ARUKF不仅具备RUKF的鲁棒性,而且能够准确地估计出观测量的噪声协方差阵R,比UKF和RUKF具有更高的初始对准性能。In the SINS underwater moving base initial alignment,when the measurement noise has uncertain and non-Gaussian statistic characteristics,assuming the measurement noise following Gauss distribution and processing the measurement noise covariance R as a constant matrix will make the precision and robustness of unscented Kalman filter（ UKF） poor. Aiming at this problem,this paper proposes an adaptive robust UKF（ ARUKF） method based on Projection Statistics（ PS） algorithm. Firstly,this method determines the weights of the stored measurement innovations,the abnormal innovation is re-weighted with PS algorithm,and then the Myers-Tapley method is used to adaptively estimate the measurement noise covariance R. Finally,the weighting function in the Huber method is used to modify the estimated measurement noise covariance R. The underwater moving base initial alignment experiment under non-Gaussian measurement noise condition was carried out using the UKF,robust UKF（ RUKF） and ARUKF algorithms,respectively based on the measured data in the river test. The experiment results demonstrate that the ARUKF method not only has the robustness of RUKF method,but also can accurately estimate the measurement noise covariance R,and has higher initial alignment performance compared with UKF and RUKF methods when the measurement is contaminated by outliers.

关 键 词：初始对准 无迹卡尔曼滤波 投影统计 鲁棒 自适应 

分 类 号：U666[交通运输工程—船舶及航道工程] TH89[交通运输工程—船舶与海洋工程]