In response to the challenge of diminished accuracy in integrated navigation for smartphones due to satellite signal interference in complex environments, this article proposes a robust extended Kalman filter enhanced by fault recovery. Firstly, this method uses equivalent weight factors to adjust observation weights in real-time, effectively reducing the impact of gross errors on combination navigation accuracy. Considering the low redundancy of observations in smartphone-based loosely-coupled navigation, this algorithm divides the detection range into three segments for fault-free, bias, and anomalies. In the absence of faults, no further processing is undertaken. When a deviation occurs, the observation value is reduced in weight. For anomalies, the predicted innovation is used to repair the fault amplitude and correct the observation value. Practical experimental results show that when a satellite experiences a single-epoch fault, the robust filter method can effectively improve the positioning accuracy of the smartphone PDR/ GNSS combination navigation. The maximum error in the north direction is reduced from 7. 27 m to 3. 20 m, and the maximum position error in the east direction is reduced from 24. 01 m to 6. 60 m. In the case of multiple-epoch consecutive faults in GNSS positions, the proposed fault recovery-enhanced robust filter method shows an average reduction of over 50% in positioning error compared to the classical robust filter method.