Relative Accuracy of GNSS/INS Integration Based on Factor Graph Optimization

The accuracy and robustness of factor graph optimization (FGO)-based Global Navigation Satellite System (GNSS) / inertial navigation system (INS) integrated navigation systems have been widely explored recently. Although factor graph methods can achieve precise state estimation in real-time, as measured by metrics like root mean square (RMS), there has been little focus on the relative changes in error series and the various error components across different time scales. However, in precision GNSS/INS applications like mobile surveying and railway irregularity detection, the relative accuracy, especially the short-term accuracy is of great concern. In this contribution, we analyze and evaluate the relative accuracy of the FGO-based GNSS/INS integrated navigation system for the first time. Allan variance, an effective metric for analyzing the relative accuracy in different temporal scales, is applied in this research. The enhancement in the relative accuracy provided by the FGO with respect to the extended Kalman filtering (EKF) is verified and compared using data from field tests. The examinations of the different sliding window sizes and different grade Micro-Electro-Mechanical System Inertial Measurement Units (MEMS-IMU) were also conducted. The results show an obvious improvement in the relative accuracy in comparison with EKF-based GNSS/INS integration, particularly in short-term relative accuracy, with gains of 23.6% in 3D position and 16.5% in 3D velocity. The proposed work can offer valuable insights into optimizing the design and sensor selection of GNSS/INS integrated systems across various precise engineering applications.