EM-CDKF算法及其SINS初始对准应用

发布时间：2018-02-04 11:25

本文关键词： 捷联惯导系统初始对准中心差分卡尔曼滤波极大似然准则极大期望算法　出处：《华中科技大学学报(自然科学版)》2014年10期 　论文类型：期刊论文

【摘要】：针对非线性捷联惯导系统噪声先验统计信息未知问题,基于中心差分卡尔曼滤波基本算法,采用极大似然准则构造极大期望最速下降梯度算法展开系统未知噪声统计特性在线估计计算研究,构建一类捷联惯导系统初始对准极大期望自适应中心差分最优滤波算法.该算法利用极大似然准则构造系统噪声统计特性对数似然函数,采用极大期望最速下降梯度法把系统噪声统计特性估计转化为对数似然函数期望最大值计算,获得系统过程噪声和观测噪声在线递推估计的自适应极大期望中心差分卡尔曼算法.经过大方位失准角捷联惯导系统初始对准仿真实验,与中心差分卡尔曼滤波基本算法相比,自适应极大期望中心差分卡尔曼算法能够有效解决基本算法在系统噪声先验知识未知情形下的滤波精度下降甚至发散问题,并且能够实现系统噪声统计特性的在线递推估计.
[Abstract]:Aiming at the unknown prior statistical information of noise in nonlinear strapdown inertial navigation system, a central differential Kalman filter algorithm is proposed. The maximum expectation maximum descent gradient algorithm is constructed by using the maximum likelihood criterion to study the online estimation of the unknown noise statistical characteristics of the system. A class of initial alignment maximum expectation adaptive center difference optimal filtering algorithm for strapdown inertial navigation system is constructed, which uses maximum likelihood criterion to construct logarithmic likelihood function of system noise statistics. The maximum expectation maximum descent gradient method is used to convert the estimation of the statistical characteristics of the system noise into the expected maximum value of the logarithmic likelihood function. The adaptive maximum expectation center differential Kalman algorithm for on-line recursive estimation of system process noise and observation noise is obtained. The simulation experiment of the initial alignment of the large azimuth misaligned sins system is carried out. Compared with the basic algorithm of central differential Kalman filter. Adaptive maximum expectation center differential Kalman algorithm can effectively solve the filtering accuracy decline or even divergence of the basic algorithm in the case of unknown prior knowledge of system noise. And it can realize the system noise statistical characteristics of the online recursive estimation.
【作者单位】：郑州轻工业学院电气信息工程学院;郑州轻工业学院软件学院;哈尔滨工程大学自动化学院;
【基金】：国家自然科学基金资助项目(U1204603) 郑州轻工业学院博士基金资助项目(2011BSJJ00048)
【分类号】：TN966
【正文快照】： 为解决非线性系统中扩展卡尔曼滤波算法估计误差较大以致引起滤波发散问题,近年来基于超球面确定性采样点的无迹变换类卡尔曼滤波(unscented tansformantion Kalman filter,UKF)方法在处理非线性系统最优估计问题中获得了较多的应用实践并取得大量研究成果[1].相比于扩展卡尔

【相似文献】