1 May 2026 in Measurement
5 August 2026 in Measurement Science and Technology
Global Navigation Satellite Systems (GNSSs) integrated with inertial navigation systems (INS) have been widely applied in intelligent driving scenarios. However, their positioning performance is easily degraded by complex environmental interference. Moreover, the conventional GNSS/INS Kalman filter relies strictly on the first-order Markov assumption and depends only on the most recent state for prediction, increasing sensitivity to accumulated inertial errors and abnormal GNSS observations and limiting the effective utilization of long-term temporal dependencies in navigation states. To mitigate these limitations, this paper proposes an Adaptive multiple-outlier-robust fractional-order extended Kalman filter (AMOFEKF). This method employs an adaptive multiple statistical similarity measure (MSSM) that selects the optimal cost function based on innovation statistics. This process ensures consistency in multi-dimensional vector assessment, thereby inhibiting measurement outliers. Meanwhile, a fractional-order extended Kalman filter model is introduced for state and parameter estimation of continuous-time nonlinear fractional-order systems under noisy conditions. Finally, a field vehicle navigation test was conducted to verify the effectiveness of the proposed model. The results demonstrate that the proposed model achieves improved positioning performance under GNSS outlier conditions. Under the GNSS pseudorange differential positioning mode, the three-dimensional root mean square error (RMSE) of the integrated system is 0.90 m. Compared with the multiple-outlier-robust Kalman filter (MORKF), the proposed method further incorporates fractional-order state modeling, resulting in a 16.07% reduction in 3D RMSE. Compared with the robust fractional-order extended Kalman filter (RFEKF), the introduction of MSSM-based multi-outlier detection further reduces the 3D RMSE by 4.08%. Furthermore, compared with the fixed-parameter MOFEKF, the adaptive AMOFEKF achieves a 10.53% performance improvement under continuous 2 m gross errors, demonstrating the effectiveness of dynamic parameter tuning.