GNSS/IMU与里程计紧-松耦合的因子图融合定位方法

Factor Graph Fusion Localization Method with Tight and Loose Coupling of GNSS/IMU and Odometry

开展了不同深度的信息融合对微惯性导航(micro inertial navigation system,M-INS)与全球导航卫星系统(global navigation satellite system,GNSS)组成的导航系统精度的影响研究,以及室内外一体化定位研究。首先,基于因子图优化(factor graph optimization,FGO)算法分别建立M-INS/GNSS单点定位的松耦合系统和M-INS/伪距/多普勒速度的紧耦合系统,并通过实验验证紧耦合设计的优势。然后,将视觉惯性里程计(visual inertial odometry,VIO)因子以松耦合的方式接入紧耦合M-INS/GNSS系统的因子结构中,提出了基于FGO紧-松耦合的多源融合算法模型。最后,通过实验验证系统的定位精度和室内外一体化定位能力。实验结果表明,在无法观测到4颗卫星而导致无法获得单点定位解时,M-INS/GNSS紧耦合系统的定位精度可以达到M-INS/GNSS松耦合系统的2倍;基于FGO的M-INS/GNSS/VIO紧-松耦合系统能够无缝应对室内外环境之间的变化,在GNSS或VIO任意一种传感器突然失效的情况下保证持续可靠定位,东、北向的平均误差比基于滤波方法的M-INS/GNSS/VIO系统精度分别提升了53.98%、54.74%。

Objectives We investigate the impact of information fusion at different depths on the accuracy of navigation systems composed of micro inertial navigation systems(M-INS)and global navigation satellite systems(GNSS),as well as indoor and outdoor integrated positioning.Methods First,based on the factor graph optimization(FGO)algorithm,a loosely coupled system of M-INS/single-point GNSS and a tightly coupled system of M-INS/pseudorange/Doppler velocity are established respectively,and the advantages of the tightly coupled design are verified by experiments.Then,the visual inertial odometry(VIO)factor is inserted into the factor structure of the tightly coupled M-INS/GNSS system in a loosely coupled manner,and a tight-loosely coupled multi-source fusion algorithm model based on FGO is proposed.Finally,the positioning accuracy and indoor and outdoor integrated positioning capability of the system are verified by experiments.Results The experimental results show that:(1)When four satellites cannot be observed and the single-point positioning solution cannot be obtained,the positioning accuracy of the M-INS/GNSS tightly coupled system can reach twice that of the M-INS/GNSS loosely coupled system.(2)In the case of the sudden failure of either GNSS or VIO,the FGO-based M-INS/GNSS/VIO tight-loosely coupled system can continue to provide reliable positioning,and the average errors of the east and north directions are improved by 53.98%and 54.74%respectively compared with the M-INS/GNSS/VIO system based on the filtering method.Conclusions Compared with the loosely coupled M-INS/GNSS system,the M-INS/GNSS tightly coupled system effectively reduces the impact of GNSS signal instability on the system,and has higher accuracy and robustness.The proposed tightly-loosely coupled system can seamlessly respond to changes between indoor and outdoor environments.