Enhanced Online Lever-Arm Estimation for GNSS/INS Integration using Factor Graph Optimization

Accurate and rapid spatial calibration has been demonstrated to be a crucial technology for Global Navigation Satellite System (GNSS)/Inertial Navigation System (INS) integrated navigation systems. Inaccurate spatial transformation between the GNSS antenna and the center of the Inertial Measurement Unit (IMU), commonly called the lever arm, may introduce additional errors into the system. Existing research has endeavored to estimate the lever arm, primarily using the Kalman Filtering (KF) model. However, the observability of the lever arm is closely related to sufficient excitation from the angular motion of the vehicle. This may pose a challenge for land vehicles, particularly large carriers such as trains, which exhibit limited angular motion. Recent research suggests that GNSS/INS integration based on factor graph optimization (FGO) outperforms the KF-based approach due to its capability to account for multiple historical epochs and perform iterative linearization. This inspires us to propose the online estimation of the lever arm using FGO-based GNSS/INS integration. By leveraging these advancements in FGO, motion information from multiple epochs can be utilized to iteratively estimate the lever arm between GNSS and INS. Field tests on vehicles and trains were conducted to validate the proposed method. The results indicate that the lever arm estimation using the FGO-based method demonstrates higher accuracy, faster convergence, and better repeatability in comparison with the KF-based method. This approach holds promise for enhancing the accuracy of lever-arm estimation in vehicle-mounted GNSS/INS platforms across a variety of transportation carriers.