基于自适应因子图优化的导航系统信息融合方法  

Information fusion method of navigation system based on adaptive factor graph optimization

作　　者：戴海发 卞鸿巍[2] 杜红松 姚曜 王涵 DAI Haifa;BIAN Hongwei;DU Hongsong;YAO Yao;WANG Han(91977 Troops of the Chinese People's Liberation Army,Beijing 100036,China;School of Electrical Engineering,Naval Engineering University,Wuhan 430033,China)

机构地区：[1]中国人民解放军91977部队,北京100036 [2]海军工程大学电气工程学院,武汉430033

出　　处：《中国惯性技术学报》2023年第1期45-52,60,共9页Journal of Chinese Inertial Technology

基　　金：国家自然科学基金项目(41506220,41876222)。

摘　　要：针对传统因子图优化方法(FGO)中量测协方差矩阵不准确带来状态估计精度下降问题,提出了一种基于滑动窗的自适应因子图优化方法(SWAFGO),将滑动窗技术应用到量测协方差矩阵的估计,利用滑动窗内的残差序列估计协方差矩阵。与传统的使用单点残差的自适应因子图方法(SRAFGO)相比,所提出方法不需要误差模型的先验知识,而且充分利用了历史残差数据,因此性能上有较大的提升。仿真实验和KITTI数据实验结果表明,在无先验知识条件下,所提出方法能够较准确地估计出量测协方差矩阵,且相较于SRAFGO方法位置估计精度提升12%。To solve the problem that the inaccuracy of the measurement covariance matrix in the traditional factor graph optimization method(FGO)will reduce the accuracy of state estimation,an adaptive factor graph optimization method based on the sliding window(SWAFGO)is proposed.The sliding window technology is applied to the estimation of measurement covariance matrix and the residual sequence in sliding window is used to estimate the covariance matrix.Compared with the traditional adaptive factor graph optimization method using single residual(SRAFGO),the proposed method does not need prior knowledge of error model,and makes full use of historical residual data,so the performance is greatly improved.Simulation experiments and KITTI data experiments show that the proposed method can accurately estimate the measurement covariance matrix without prior knowledge,and the position estimation accuracy is 12%higher than that of SRAFGO method.

关 键 词：惯性导航系统 组合导航 因子图 自适应 滑动窗 

分 类 号：U666.1[交通运输工程—船舶及航道工程]