基于RFR辅助的因子图组合导航算法  

An Integrated Navigation Algorithm of Factor Graph Based on RFR

作　　者：栾啸飞 张华强[1] 李孝旭 陈雨 LUAN Xiaofei;ZHANG Huaqiang;LI Xiaoxu;CHEN Yu(Shandong University of Technology,Zibo 255000,China;Beijing Institute of Space Launch Technology,Beijing 100000,China)

机构地区：[1]山东理工大学,山东淄博255000 [2]北京航天发射技术研究所,北京100000

出　　处：《电光与控制》2023年第6期69-73,共5页Electronics Optics & Control

基　　金：国家自然科学基金青年基金(61803035)。

摘　　要：针对在INS/GNSS组合导航系统中,由GNSS自身故障或外部环境遮挡造成的信号缺失问题,提出了一种利用随机森林回归辅助因子图的组合导航算法。首先,采用因子图算法对惯性导航系统、全球导航卫星系统进行建模,搭建了INS/GNSS组合导航的因子图模型。其次,引入随机森林理论搭建随机森林,并在GNSS信号有效时进行训练,模拟卫星导航失效时的GNSS信号输出值。最后设计了仿真实验,结果表明:改进的因子图算法相比联邦卡尔曼滤波算法在导航精度上有了10%~15%不等的提升,同时,所提出的随机森林回归辅助因子图算法在GNSS信号丢失的情况下仍能保持较高精度。In order to solve the problem of GNSS signal loss caused by GNSS faults or external environment occlusion in INS/GNSS integrated navigation system,an integrated navigation algorithm based on Random Forest Regression(RFR)factor graph is proposed.Firstly,the Inertial Navigation System(INS)and Global Navigation Satellite System(GNSS)are modeled by using factor graph method,and the factor graph model of INS/GNSS integrated navigation is built.Secondly,the random forest theory is introduced to build the random forest,and the training is carried out when the GNSS signal is effective,and the GNSS signal output when the satellite navigation fails is simulated.Finally,a simulation experiment is designed,and the results show that the improved factor graph algorithm improves the navigation accuracy by about 10%~15%compared with the federated Kalman filter algorithm.Meanwhile,the proposed RFR factor graph algorithm can still maintain high accuracy in the case of GNSS signal loss.

关 键 词：组合导航 因子图 随机森林回归 数据融合 

分 类 号：U666.1[交通运输工程—船舶及航道工程]