传统组合导航中的实用Kalman滤波技术评述

在随机线性系统建模准确的情况下,Kalman滤波是线性最小方差无偏估计。针对传统惯导/卫导组合导航的实际应用,难以精确建模,给出了常用的建模方法、状态量选取原则、离散化方法及滤波快速计算方法。讨论了平方根滤波、自适应滤波、联邦滤波和非线性滤波等技术的适用场合,并给出了使用建议。针对前人研究可观测度中未考虑随机系统噪声的缺陷,提出了更加合理的以初始状态均方误差阵为参考的可观测度定义和分析方法。提出了均方误差阵边界限制方法,可有效抑制滤波器的过度收敛和滤波发散。该讨论可为工程技术人员提供一些有实用价值的参考。

Review on Practical Kalman Filtering Techniques in Traditional Integrated Navigation System

Kalman filtering is a minimum-variance unbiased estimator in the case of accurate modeling for stochastic linear systems. For practical application in the traditional INS/GNSS integrated navigation, it is always difficult to obtain an ideal precise model. In this paper, commonly used modeling methods, selection principle for state variables, discretization method and fast calcu-lation methods are presented. The applications of square root filtering, adaptive filtering, federa-ted filtering and nonlinear filtering are discussed, and then some useful suggestions are given. In view of the defects of previous studies on observability analysis without considering the noise effect of stochastic system, a more reasonable observability definition and analysis method with respect to the initial state mean square error matrix is proposed. The use of boundary limiting method for mean square error matrix is also proposed, which can effectively suppress filter excessive convergence and avoid filter divergence. It is hoped that the discussion can provide useful references for engineers and technicians.