A Dual-Adaptive Sage–Husa Kalman Filter with Model Matching Check for GNSS/INS-Integrated Navigation

GNSS/INS integrated navigation combines the long-term stability of satellite positioning with the high-frequency continuity of inertial navigation, and has been widely adopted in vehicle navigation, unmanned systems, and intelligent transportation systems. However, in complex urban environments, GNSS measurement noise exhibits significant time-varying characteristics, while high-dynamic carrier motions may cause mismatches between the dynamic model and the actual motion state. These two factors jointly degrade the performance of conventional Sage–Husa (cSage–Husa) filters, resulting in distorted measurement noise estimation and over-adaptation. To address this issue, this paper proposes a dual-adaptive Sage–Husa filtering algorithm considering model matching. The proposed algorithm integrates a forgetting-factor-based global recursive estimation mechanism with an adaptive sliding window strategy. Furthermore, a model matching test and a measurement noise covariance upper-bound constraint are introduced to distinguish model mismatch errors from normal measurement noise variations, thereby improving filtering stability and noise estimation reliability. Experimental results demonstrate that, under initial high-dynamic motions and instantaneous GNSS interference conditions, the proposed algorithm reduces the up-channel RMS error by 8.5% compared with cSage–Husa. Under sustained GNSS degradation conditions, the north RMS error is reduced by 8.7% and 2.5% compared with EKF and cSage–Husa, respectively, while the up-channel RMS error is reduced by 15% compared with cSage–Husa and maintains comparable accuracy to EKF. The proposed algorithm effectively mitigates over-adaptation and convergence degradation caused by model mismatch, achieving a favorable balance between time-varying noise tracking capability and filtering stability for urban canyon and high-dynamic vehicle positioning.