一种新的抗差自适应Unscented粒子滤波  

Proposing a Robust Adaptive Unscented Particle Filtering(RAUPF) Algorithm

作　　者：薛丽[1] 高社生[1] 王建超[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《西北工业大学学报》2011年第3期470-475,共6页Journal of Northwestern Polytechnical University

基　　金：航空科学基金(20080818004);陕西省自然科学基金(SJ08F04)资助

摘　　要：针对粒子滤波存在的重要性密度函数难以选取和可能出现粒子退化的问题。提出了一种新的抗差自适应Unscented粒子滤波算法。该算法不但能利用等价权函数和自适应因子合理的分配信息,提高滤波精度,而且具有Unscented粒子滤波的优点,更好的适用于非线性、非高斯系统模型的计算。仿真结果表明,文中提出的抗差自适应Unscented粒子滤波算法,滤波性能明显优于扩展卡尔曼滤波和粒子滤波算法,并且能提高组合导航系统的定位精度。Particle filtering causes degeneration and has difficulties in selecting the importance density function. Aiming at this problem, we propose a RAUPF algorithm, which is better than the existing particle filtering algo-rithms, through adding the concept of robust adaptive filtering to the unscented particle filtering （UPF）. Sections 1 and 2 of the full paper explain our proposal. The core of section 2 consists of： （ 1 ） in selecting the importance density function, we combine the robust estimation with the UPF algorithm by utilizing the equivalent weight function and the adaptive factor as shown by eq. （19） ; （2） we reasonably distribute the effective model information, thus suppressing the influence of abnormal interference on the state model and the observation model. With two numeri- cal examples, section 3 verifies our RAUPF algorithm and applies it to a SINS/SAR integrated navigation system. The simulation results, given in Figs. 1 through 5, and their analysis show preliminarily that ： （ ! ） our RAUPF algorithm outperforms the conventional extended Kalman filtering and the particle filtering in terms of filtering precision, thus being more suitable for the filtering calculation of a nonlinear and non - Gaussian model ; （2） it can improve the location precision of the SINS/SAR integrated navigation system.

关 键 词：Unscented粒子滤波 抗差估计 等价权 自适应因子 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]