自适应SVD-UKF算法及在组合导航的应用  

Adaptive SVD-UKF algorithm and application to integrated navigation

作　　者：高社生[1] 王建超[1] 焦雅林[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《中国惯性技术学报》2010年第6期737-741,765,共6页Journal of Chinese Inertial Technology

基　　金：航空科学基金(20080818004);陕西省自然科学基金项目(N9YU0001)

摘　　要：提出一种新的自适应奇异值分解Unscented卡尔曼滤波(UKF)算法。该算法利用自适应因子平衡动力学模型信息与观测信息的权比,控制动力学模型误差对导航参数解的影响。用奇异值分解阵(SVD)的迭代计算代替协方差矩阵的迭代变换,提高了协方差矩阵的数值稳定性。将新算法应用于组合导航系统进行计算仿真,结果证明,新算法具有良好的鲁棒性,能有效改善滤波性能,提高组合导航系统的精度。This paper puts forward a new unscented Kalman filtering（UKF） algorithm.To control the effects of dynamic model errors on the solution of navigation parameters,the ratio of dynamic model information and observation data are made into equilibrium by an adaptive factor.Singular value decomposition（SVD） of matrix is adopted to substitute for the iteration of covariance matrix and improve the numerical stability of the covariance matrix.The simulation is made which applies the proposed algorithm into an integrated navigation system and compares it with adaptive Kalman filtering algorithm and adaptive UKF algorithm respectively.The results show that the new algorithm has higher robustness,and can effectively improve the filter performance and position accuracy of the integrated system.

关 键 词：自适应滤波 奇异值分解 UNSCENTED卡尔曼滤波 组合导航 

分 类 号：U666.1[交通运输工程—船舶及航道工程]