Resilient Pseudorange Error Prediction and Correction for GNSS Positioning in Urban Areas

Abstract

Positioning, navigation and timing (PNT) is essential for internet of things (IoT) communications and location-based services. Although GNSS can provide accurate PNT in open areas, obtaining reliable PNT is still a considerable technical challenge in complex urban environments. This is because the GNSS signals are more likely to be affected by multipath interference and non-line of sight (NLOS) reception issues arising from the obstructions and reflections in built environments. These introduce range measurement errors that degrade the GNSS positioning accuracy. This paper proposes two resilient pseudorange error prediction and correction strategies to improve the GNSS positioning accuracy in urban environments. In particular, considering the carrier-to-noise density (C/N0), satellite elevation angle and local positional information, the random forest based pseudorange error prediction and correction models are constructed in two variations, including: 1) the point-based correction (PBC), and 2) the grid-based correction (GBC). The final improved positioning solution is then calculated by using the least square method (LSM) of the corrected pseudoranges. Kinematic test results in urban environments show that both variations of the proposed model can improve the positioning accuracy by 42.9% and 40.8% in horizontal, and by 60.1% and 63.3% in 3D, respectively, compared to the positioning results obtained by traditional method without pseudorange error corrections. The improvements are 41.1% and 38.9% in horizontal, and 45.7% and 50.0% in 3D, respectively, compared with traditional elevation angle weighting method.