Robust Helmert Variance Component Estimation for FGO-Based Multi-GNSS/INS Tightly Coupled Integration to Enhance Vehicle Navigation in Urban Environments

The integration of multi-global navigation satellite system (GNSS) and inertial navigation system (INS) has been proved as a more effective strategy for positioning in GNSS-challenged environments. Recent research suggests that factor graph optimization (FGO)-based GNSS/INS integration outperforms the extended Kalman filtering (EKF)-based one in terms of accuracy and robustness. However, the adaptive and refined stochastic modeling for FGO-based multi-GNSS/INS tightly coupled integration in urban environments remains challenging and unexplored. To fill this gap, we propose a robust Helmert variance component estimation (HVCE) for FGO-based multi-GNSS/INS tightly coupled integration to enhance vehicle navigation in urban environments. To further enhance the robustness of the HVCE algorithm, an outlier detection and exclusion algorithm was adopted to ensure the quality of multi-GNSS observations. Field vehicle-borne tests in urban environments were conducted to validate the proposed method. The results show that the robust HVCE algorithm provides refined stochastic models, which significantly improve positioning accuracy. Compared to the FGO-Only method, the accuracy improves by 46.4%, 38.3%, and 26.6% in the north, east, and vertical directions, respectively. The detailed analysis in terms of the robust algorithm and HVCE to the multi-GNSS/INS integration further validates the effectiveness of the proposed method in complex urban environments.