Factor Graph-Based Tightly Coupled PPP-B2b/INS for Real-Time Precise Positioning

Real-time precise point positioning (PPP) based on BDS-3 PPP-B2b corrections often suffers from limited robustness in complex environments. To address these challenges, this letter proposes a factor-graph-based tightly coupled PPP-B2b/inertial navigation system (INS) integration scheme. Unlike traditional filtering approaches, the proposed framework fuses asynchronous GNSS observations corrected by PPP-B2b with pre-integrated inertial measurements within a sliding-window optimization architecture. To explicitly mitigate non-Gaussian noise and measurement anomalies prevalent in urban canyons, a robust estimation strategy is implemented. This strategy incorporates Huber loss functions for outlier resistance and, critically, employs a time-varying, correction-age-dependent measurement covariance model to characterize the reliability degradation caused by correction latency. This formulation enables multiple relinearizations and effective outlier mitigation while maintaining bounded computational cost. Experimental results utilizing a dense urban dataset demonstrate that the proposed method reduces the ENU RMS position errors from 0.797/0.954/1.616 m to 0.490/0.671/0.797 m and the 3D RMS from 2.04 m to 1.15 m, representing a 43.6% improvement over the extended Kalman filter (EKF)-based solution. The factor-graph approach exhibits significantly smoother trajectory behavior in challenging environments, confirming that the proposed robust framework offers superior accuracy and reliability with the high computational efficiency required for real-time applications.