首页 | 本学科首页   官方微博 | 高级检索  
     检索      

几种非线性滤波算法的性能比较与分析
引用本文:余春平,李广云,张冠宇.几种非线性滤波算法的性能比较与分析[J].海洋测绘,2008,28(6):43-45.
作者姓名:余春平  李广云  张冠宇
作者单位:解放军信息工程大学,测绘学院,河南,郑州,450052
摘    要:在非线性状态估计中,传统的扩展卡尔曼滤波通过线性化来实现高斯近似,由于截断误差的存在很难保证估计精度;而基本粒子滤波容易出现粒子退化,导致滤波发散。针对粒子滤波的两个基本假设:蒙特卡罗假设和重要采样假设,采用蒙特卡罗随机链的方法来提高粒子的多样性,并利用无味卡尔曼滤波来产生更高精度的替代分布,发展了无味粒子滤波。通过仿真实验证明,相比较扩展卡尔曼滤波和基本粒子滤波,改进后的无味粒子滤波算法性能更优越,对含有非线性非高斯的状态估计问题有更好的滤波效果。

关 键 词:粒子滤波  蒙特卡罗马尔可夫链  无味粒子滤波  状态估计

Compare Analysis of Several Nonlinear Filters'Performance
YU Chun-ping,LI Guang-yun,ZHANG Guan-yu.Compare Analysis of Several Nonlinear Filters'Performance[J].Hydrographic Surveying and Charting,2008,28(6):43-45.
Authors:YU Chun-ping  LI Guang-yun  ZHANG Guan-yu
Institution:( Institute of Surveying and Mapping, Information Engineering University, Zhengzhou, Henan ,450052)
Abstract:Traditional EKF uses linearization to realize Gauss approximation which cannot guarantee estimate precision because of truncation error.The generic particle filter suffers from the particle degeneration.In order to improve the particle filter,a Markov Chain Monte-Carlo move step is introduced and formed PF-MCMC.UPF is another improvement of PF,which result from using a UKF for proposal distribution generation within a PF framework.Through the experiment,it is validated that the improved filters have better effect than EKF on the estimate of the nonlinear state estimation.
Keywords:particle filter  Markov Chain Monte-Carlo  unscented particle filter  state estimation
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《海洋测绘》浏览原始摘要信息
点击此处可从《海洋测绘》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号