几种非线性滤波算法的比较研究

Comparative Study on Some Nonlinear Filtering Algorithms

针对组合导航等非线性系统,扩展卡尔曼滤波算法(EKF)在初值不准确时存在滤波发散的现象,故提出U-卡尔曼滤波(UKF);粒子滤波算法(PF)适合于强非线性、非高斯噪声系统,但同时存在退化现象,故提出2种改进算法。前人的工作多集中在单一算法的研究,而在此是将上述各种算法应用到同一典型非线性系统,通过应用Matlab进行仿真实验得出具体滤波效果数据,综合对比分析了各算法的优缺点,得出一些有用的结论,为组合导航系统中非线性滤波算法的选择提供了参考。

For the nonlinear systems such as integrated navigation systems, since the extended Kalman filtering （EKF） has a dispersing phenomenon when the initial state value is inaccurate, the unscented Kalman filtering （UKF） is proposed, and although particle filtering （PF） is suitable for any nonlinear non Gaussian systems, it has a degeneracy phenomenon, then two kinds of improved filtering algorithms are put forward. Scientific researchers focused on single filtering before. The filte ring algorithms mentioned above are adopted in a same typical model of nonlinear system in this paper. The detailed data of the filtering algorithms were obtained by emulational experiments with Matlab. Some useful conclusios were acquired after the contrast and analysis of their advantages and disadvantages. A reference is offered in choosinga suitable nonlincarfiltering algorithm for integrated navigation systems.