捷联惯性基组合导航中，量测值存在野值的情况难以避免，会导致无迹卡尔曼滤波(UKF)的估计精度下降。针对该问题，本文提出一种基于一类向量机（SVM）的鲁棒UKF算法（SVM-UKF）。首先使用一类支持向量机训练滑动窗，来辨别滤波中的新息是否为异常，对于正常新息不予处理，对于异常的新息采用指数加权的方法进行估计，使用新的估计值替换野值，并进行了船载实验，对含有野值的SINS/GPS系统使用SVM-UKF与常规UKF，RUKF滤波进行组合导航实验。实验结果表明，在量测值有野值污染的情况下，SVM-UKF具有较高的鲁棒性，对比于UKF，RUKF具有更高的估计精度。
In strapdown inertial base integrated navigation, it is difficult to avoid the existence of outliers, which will lead to the estimation accuracy of Unscented Kalman filter (UKF) decline. To solve this problem, a robust UKF algorithm (svm-ukf) based on a class of vector machines (SVM) is proposed. First, a kind of support vector machine is used to train the sliding window to identify whether the innovation in the filtering is abnormal. Normal innovation is not handled. For abnormal innovation, the method of exponential weighting is used to estimate, and the new estimated value is used to replace the outliers. In the experimental part, the Yangtze River shipborne experiment is carried out. For SINS / GPS system with outliers, svm-ukf and conventional UKF and RUKF are used to filter Line integrated navigation experiment. The experimental results show that svm-ukf has higher robustness in the case of outlier pollution, and RUKF has higher estimation accuracy compared with UKF.
2021,43(6): 141-145 收稿日期：2019-11-20
DOI：10.3404/j.issn.1672-7649.2021.06.027
分类号：TN967.2
基金项目：国家自然科学基金资助项目(61703419，61374206)
作者简介：张梦得(1992-)，男，博士研究生，从事惯性技术及应用研究
参考文献：
[1] B. ALANDRY, L. LATORRE, F. MAILLY, et al. A fully integrated inertial measurement unit: application to attitude and heading determination[J], IEEE Sensors Journal, 2011, 11(11): 2852−2860.
[2] 王泽元, 许昭霞. 基于UT变换改进多模型滤波的水下航行器导航方法[J]. 舰船科学技术, 2014(10): 73–77
WANG Z Y, XU Z X. Navigation method of underwater vehicle based on ut transformation and improved multi model filtering[J]. Ship Science and Technology, 2014(10): 73–77
[3] K. LI, L. CHANG, B. HU. Unscented attitude estimator based on dual attitude representations[J], IEEE Transactions on Instrumentation and Measurement, 2015, 64(12): 3564−3576.
[4] WU Y X, PAN X F. Velocity/position integration formula, PartI: Application to in-flight coarse alignment[J]. IEEE Transactions on Aerospace and Electronic Systems, 2013, 49(2): 1006–1023
[5] Di WANG, Xiao-su XU, Yong-yun ZHU. A novel hybrid of a fading filter and an extreme learning machine for GPS/INS during GPS outages[J]. Sensors, 2018(11): 3863
[6] 任民魁, 蒋国萍, 姜宇琛. 基于修正Rodrigues参数的鲁棒滤波算法在组合导航中的应用[J]. 电光与控制, 2019, 26(3): 39–43
REN Min-kui, JIANG Guo-ping, JIANG Yu-chen. Application of robust filtering algorithm based on modified Rodrigues parameters in integrated navigation[J]. Electro Optic and Control, 2019, 26(3): 39–43
[7] GU D Q, EL-SHEIMYN, HASSANT, et al. Coarse alignment for marine SINS using gravity in the inertial frame as a reference[J]. IEEE PLANS, CA, 2008: 961–965
[8] H. REN, AND P. KAZANZIDES. Investigation of attitude tracking using an integrated inertial and magnetic navigation system for hand-held surgical instruments[J], IEEE/ASME Transactions on Mechatronics, 2012, 17(2): 210−217.
[9] Rui XU, Meng-yu DING, Ya QI. Performance analysis of GNSS/INS loosely coupled integration systems under spoofing attacks[J], Sensors, 2018, 18(12), 4108.
[10] 陈建华, 朱海, 王超, 等. 水下SINS/DVL紧组合导航算法[J]. 海军工程大学学报, 2017, 29(2): 108–112
CHEN Jian-hua, ZHU Hai, WANG Chao, et al. An underwater SINS/DVL tightly integrated navigation algorithm[J]. Journal of Naval University of Engineering, 2017, 29(2): 108–112
[11] CRASSIDIS J L, MARKLEYF. Unscented filtering for spacecraft attitude estimation[J]. Journal of Guidance, Control, and Dynamics, 2003, 26(4): 536–542
[12] 任东彦, 孙明太, 周利辉. 基于鲁棒Kalman滤波的猎扫雷作战过程建模及效果评估[J]. 舰船科学技术, 2014, 36 (4): 143–146
REN Y D, SUN M T, ZHOU L H. Modeling and effect evaluation of mine hunting process based on robust Kalman filtering[J]. Ship science and technology, 2014, 36 (4): 143–146
[13] 季超, 杨晓东. 基于H∞的INS/GPS组合导航系统初始对准研究[J]. 舰船科学技术(5): 87−89.
JI C, YANG X D. Initial alignment of INS / GPS integrated navigation system based on H∞[J]. Ship Science and Technology (5): 87−89.
[14] 高敬东, 李开龙, 常路宾. 基于Huber的改进鲁棒滤波算法[J]. 系统仿真学报, 2014(8): 1769–1774
GAO J D, LI K L, CHANG L B. An improved robust filtering algorithm based on Huber[J]. Journal of System Simulation, 2014(8): 1769–1774
[15] 谭兴龙, 王坚, 韩厚增. 支持向量回归辅助的GPS/INS组合导航抗差自适应算法[J]. 测绘学报. 2014(43). 6: 590−597.
TAN Xing-long, WANG Jian, HAN Zeng-hou. Robust adaptive algorithm for GPS/INS integrated navigation aided by support vector regression[J]. Journal of Surveying and Mapping. 2014(43). 6: 590−597.
[16] DUAN L, XIE M, BAI T, et al. Anew support vector data description method for machinery fault diagnosis with unbalanced datasets[J]. Expert Systems with Applications, 2016, 64: 239–246
