An Effective GNSS Fault Detection and Exclusion Algorithm for Tightly Coupled GNSS/INS/Vision Integration via Factor Graph Optimization

• Published in: 2022 International Conference on Automation, Robotics and Computer Engineering (ICARCE) pp. 1 - 5
• Main Authors: Jiang, Haitao, Li, Tuan, Shi, Chuang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.12.2022
• Subjects: Factor graph optimization; Faults detection and exclusion; GNSS/INS/Vision integration; Positioning accuracy
• DOI: 10.1109/ICARCE55724.2022.10046595

Pseudorange measurements from GNSS (Global Navigation Satellite System) receivers are seriously affected by multipath in urban environments, which greatly degrades the positioning accuracy and reliability of GNSS/Inertial Navigation System (INS)/Vision integrated system. Fault Detection and Exclusion (FDE) module is essential to improve the robustness and positioning performance of the system. Recently, GNSS/INS/Vision integration via factor graph optimization (FGO) has attracted extensive attention due to its high accuracy and robustness. As measurements from multiple epochs can be used under FGO framework, it is expected that the detection capability of faulty pseudorange measurements can be improved significantly. Meanwhile, the inclusion of visual measurements could contribute to the capability of FDE of faulty GNSS measurements. In this contribution, we present a parallel GNSS FDE method via FGO, and it calculate the test statistics of each satellite based on the residuals of GNSS measurements in a sliding window. The public GVINS-dataset "urban" were used to evaluate the performance of the parallel GNSS FDE scheme in urban canyons. Experimental results show that compared with the GNSS/INS integration, the 2D positioning accuracy in terms of Root Mean Square Error of the parallel GNSS FDE scheme used for GNSS/INS/Vision integration is improved by 33.5% in urban complex environment. Additionally, compared with the sliding window-based FDE method, for GNSS/INS integration and GNSS/INS/Vision integration, the 2D positioning accuracy is increased by 12.1% and 11.7% respectively.