An Improved Multiple-Outlier-Robust Kalman Filter for GNSS/INS Tightly Coupled Positioning in Urban Canyons

In urban canyon environments, the integration of Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) faces significant challenges due to multipath effects, signal occlusions, and non-Gaussian noise, which degrade positioning accuracy and reliability. This study proposes an improved multiple-outlier-robust Kalman filter (MORKF) algorithm to enhance the accuracy and robustness of tightly coupled GNSS/INS positioning. The proposed algorithm integrates a bifactor robust estimation framework that leverages residuals to dynamically adjust the measurement noise covariance matrix, effectively mitigating gross errors and non-Gaussian noise. Additionally, an adaptive degree-of-freedom (DOF) estimation mechanism is introduced to dynamically tune the filters response to varying noise characteristics, reducing reliance on empirical parameter tuning. Real-world experiments in medium and high complexity urban canyons demonstrate the proposed algorithms superior performance, achieving 3D positioning error of 1.447 m and 4.638 m, respectively, with improvements over classical Kalman filter (KF), robust Kalman filter (RKF), and MORKF of up to 83.2%, 14.2%, and 75.3% for Dataset 1, and 50.5%, 16.1%, and 41.8% for Dataset 2. Compared with existing MORKF variants, the proposed algorithm also ensures robust trajectory estimation and a higher ambiguity fixing rate, enhancing its potential for high precision positioning in challenging urban environments.