Gaussian Sum Filtering Based on Uniformly Random Design with Application to Terrain Navigation
For nonlinear and non-Gaussian models,sequential updating of the filtering and predictive densities is not as straightforward as in the linear Gaussian model. In this paper, densities are approximated as finite mixture models as is done in the Gaussian sum filter (GSF). A novel GSF (UGSF) for filtering nonlinear non-Gaussian dynamic system is proposed which updates the means and covariances of the mixands using Uniformly Random Design (URD)method. To keep the number of the mixands constant, a method of weighted Expectation-Maximization (WEM)algorithm is used. The simulation of a univariate nonlinear model is presented here to exhibit the good performance of the novel filter compared with other two filters. The novel filter is also applied in 2-D terrain navigation. The simulation results also illustrate the performance outgoes GMSPPF and UKF.
Nonlinear Filtering Gaussian Sum Filtering UGSF URD Terrain Navigation
Dong Liu Zhiying Yao
Staff room 303 Xian Research Inst.of High-Tech Hongqing Town, Xian P.R.China
国际会议
The International Colloquium on Onformation Fusion 2007(2007年国际信息融合研讨会)
西安
英文
103-109
2007-08-22（万方平台首次上网日期，不代表论文的发表时间）