摘要
在动态环境下导航定位受到很多因素影响,且异常噪声会严重影响导航滤波结果。通过研究Bayes滤波的抗差方法,在导航观测方程中采用高斯混合模型,对混合模型不确定度参数采用层次模型实时估计。借助指示变量进行模型变换,削弱导航过程中异常噪声的影响。并采用Rao-Blcakwellized粒子采样方法,求取复杂、非标准形式的状态后验分布。最后通过卫星导航以及组合导航实验算例,分析验证了基于高斯混合模型的Bayes滤波在动态导航定位中的抗差性能。
Under the dynamic environment, navigation and positioning may be affected by many factors. Abnormal noise could seriously decrease the accuracy of navigation filtering results. In this paper the Bayesian filter methods were studied. And we used the Gaussian mixture models in the navigation observation equation. Then we estimated the uncertainty of parameter in mixture models with the way of hierarchical model. With the help of indicator variables, this algorithm could complete model transformation, and weaken the influence of abnormal noise in the process of navigation. Rao-Blcakwellized particle sampling method was adopted to calculate the complex and nonstandard forms of state sposterior distribution. Finally, through the satellite navigation and integrated navigation experiments, the performance of robust Bayesian filter based on Gaussian mixture model in dynamic navigation and positioning was provedby analysis. The work in this paper could have reference value in the robust filter research as well as its applications.
出处
《测绘科学技术学报》
CSCD
北大核心
2016年第1期27-32,共6页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(41274016
40974010)
地理信息工程国家重点实验室重点基金项目(SKLGIE2014-Z-2-1)
信息工程大学地理空间信息学院创优基金(XS201504)
关键词
动态导航定位
高斯混合模型
Bayes滤波
层次模型
粒子采样
后验概率
kinematic navigation and positioning
Gaussian mixture model
Bayes filter
hierarchical model
particle sampling
posterior distribution
