基于奇异值分解的球型平方根UKF滤波算法  

Spherical SRUKF based on singular value decomposition

作　　者：叶泽浩 周升响[1] 晏凯[1] 涂灏 YE Zehao;ZHOU Shengxiang;YAN Kai;TU Hao(Air Force Early Warning Academy,Wuhan 430019,China)

机构地区：[1]空军预警学院,武汉430019

出　　处：《空天预警研究学报》2022年第1期10-14,共5页JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH

摘　　要：为了解决传统平方根UKF(SRUKF)跟踪精度易受参数选择影响且易出现奇异矩阵而导致滤波失效问题,研究了一种基于奇异值分解的球型平方根UKF(SVDS-SRUKF)滤波算法.该算法首先改用球型无迹变换对权系数以及sigma点进行选取,其次引入奇异值分解方法,对协方差矩阵进行奇异值分解,并化成平方根的形式进行算法迭代.仿真结果表明,与基于奇异值分解的平方根UKF(SVD-SRUKF)算法和SRUKF算法相比,SVDS-SRUKF算法在减少计算量的同时提高了滤波精度、收敛速度,具有较好的可靠性以及适用性.Traditional square root unscented Kalman filter(SRUKF)is easily affected in the tracking accuracy by parameter selection,and is also prone to occurrence of singular matrix,which leads to filtering failure.In order to solve the problems,this paper researches a spherical SRUKF algorithm based on singular value decomposition(SVD).This algorithm first uses spherical unscented transformation to select the weight coefficients and simga points.And then,the SVD method is introduced to carry out SVD on the covariance matrix,and transform it into a square root form for algorithm iteration.Simulation results show that,compared with SVDbased SRUKF algorithm and SRUKF algorithm,SVDSSRUKF algorithm improves filtering accuracy and convergence speed while reducing computation.Thus it has better reliability and applicability.

关 键 词：平方根UKF 奇异矩阵 滤波失效 奇异值分解 

分 类 号：TN957[电子电信—信号与信息处理]