未知测量噪声方差的平方根HCKF算法记作SHCKF-UN算法.DOCVIP

未知测量噪声方差的平方根高阶容积Kalman滤波 管冰蕾1，汤显峰2，葛泉波3 (1. 宁波工程学院 电子与信息工程学院, 浙江 宁波 315016；2. 浙江大学 教育学院, 杭州 310028；3. 杭州电子科技大学 自动化学院, 杭州 310018) 摘 要: 针对非线性系统中测量噪声统计特性未知导致滤波精度不高或发散的问题，提出了一种自适应平方根高阶容积卡尔曼滤波（high-degree cubature Kalman filter, HCKF）算法.首先,采用QR分解、Cholesky因子更新和高效最小二乘法等矩阵分解技术设计了一种平方根HCKF算法（SHCKF）,提高了滤波算法的数值稳定性,减小了状态估计误差.其次,引入Sage-Husa估计器在线估计未知测量噪声的方差,进一步提高了SHCKF的估计精度,扩大了算法的应用范围.最后,几个计算机仿真实验表明,与标准的HCKF算法相比,新算法具有更好的估计精度,尤其是在测量噪声统计特性不确定的场景. 关键词: 非线性系统; 高阶容积卡尔曼滤波; 自适应滤波; 矩阵分解; 噪声统计估计器 中图分类号: TN911　　　 Square-root high-degree cubature Kalman filter with unknown measurement noise covariance GUAN Bing-lei1, TANG Xian-feng2, GE Quan-bo3 (1. Academy of Electrics and Information, Ningbo University of Technology, Ningbo, Zhejiang, 315016, China；2. College of Education, Zhejiang University, Hangzhou 310028, China；3. School of Automation, Hangzhou Dianzi University, Hangzhou, 310018, China) Abstract: Aiming at low filtering precision and divergence caused by unknown measurement noise statistics in nonlinear system, an adaptive square-root high-degree cubature Kalman filter (HCKF) is proposed. First of all, matrix factorization techniques including QR decomposition, Cholesky factor updating and efficient least squares were introduced to design a square-root HCKF. Therefore, the filter numerical stability is effectively improved and the state estimation error is reduced. Secondly, Sage-Husa estimator was used to online estimate the unknown measurement noise covariance, which further improves the state estimation accuracy and enlarges its application scope. Finally, the simulations show that the proposed methods can provide better performance in estimation accuracy than the standard HCKF, especially in the case of uncertain measurement noise statistics. Key words: nonlinear system; high-degree cubature Kalman filter; adaptive filtering; matrix factorization; noise statistic estimator 引言 在诸多领域中，系统状态的动态估计问题一直是人们关注的焦点。线性高斯系统的状态估计问题，一般采用Kalman滤波方法[1]。但在处理实际问题时，比如目标跟踪、导航定位以及视频监控等，系统