基于集合卡尔曼滤波的改进粒子滤波算法_杜航原.pdfVIP

33 7 V ol . 33 No. 7 2011 7 Sy stems Engineer ing and Electr onics July 2011 : 1001506X( 20 11 杜航原, 郝燕玲, 赵玉新 ( 哈尔滨工程大学自动化学院, 黑龙江 哈尔滨 150001 : 提出一种基于集合卡尔曼滤波的粒子滤波改进方法该方法利 集合卡尔曼滤波的最大后验概率 估计产生粒子滤波每一时刻各粒子的建议分布函数, 使建议分布函数融入最新观测信息的同时, 更加符合状态的 真实后验概率分布该方法在对粒子滤波的建议分布进行估计时使 采样方法近似非线性分布, 且采样点数灵 活可变, 使计算精度和算法效率得到提高仿真结果表明, 提出的集合卡尔曼粒子滤波的估计性能明显优于标准 粒子滤波扩展卡尔曼粒子滤波和无迹粒子滤波 : 非线性系统; 粒子滤波; 集合卡尔曼滤波; 建议分布函数 : T N 953 : A DOI: 10. 3969/ j. issn. 100 1506X. 20 11. 07. 42 Improved particle filter based on ensemble Kalman filter DU H angyuan , H A O Y anling , Z H A O Y ux in ( College of A utomation, H arbinEngineering Univers ity , H arbin150001, China) Abstract: A n impr ov ed part icle f ilter based on t he ensemble Kalman filt er ( EnKF is descr ibed. T he EnKF is used to propagate the part icles pr opo sal dist ribution of part icle f ilter in every t ime step. Because the EnKF can at tain a m ax im um posterior estim at ion of t he nonlinear sy stem, and t he proposal distr ibut ion f unct ion integr ates the late st observ at ion into sy stem st ate tr ansition dens ity , so th e pr opo sal distr ibution can appr ox i m ate the true post er ior distr ibut ion m or e accur ately . T his new algor it hm uses sample ensem ble s t o approx im ate t he nonliner pr oposal dist ribution, and the num ber of ensembles r equir ed in the EnKF is flex ible. In this w ay, the accur acy and efficiency of the algorithm is impr oved. The simulation results r eflect that the novel particle filter is super ior to the standard part icle filter , the extended Kalman particle filter and the unscented particle filter. Keywords: nonlinear syst