Robust GNSS shadow matching for smartphones in urban canyons

Abstract

GNSS is being widely used in different applications in navigation. However, GNSS positioning is greatly challenged by notorious multipath effects and non-line-of-sight (NLOS) receptions. The signal blockage and reflection by buildings cause these effects. In other words, the more urbanized the city is, the more challenge on the GNSS positioning. The conventional multipath mitigation approaches, such as the sophisticated design of GNSS receiver correlator, can efficiently mitigate the most of multipath effects. However, it has less capability against NLOS reception, potentially leading to several tens of positioning errors. Therefore, the 3D mapping aided (3DMA) GNSS positioning is introduced to exclude or even use the NLOS signal. Shadow matching is to make use of the similarity between building geometry and satellite visibility to improve the positioning performance. This paper introduces a machine learning intelligent classifier with features to distinguish LOS and NLOS. With the NLOS reception classification, the positioning accuracy of shadow matching can be increased. In addition, this paper develops several indicators to label the unreliable solution of shadow matching. These indicators are to examine the complexity of the surrounding environment, which is the key factor relating to the proposed shadow matching performance. Several designed experiments were done in Hong Kong to evaluate the proposed method. With the intelligent classifier, the average positioning accuracy is about 15m and 6m on 2D and the across-street direction, respectively. Simultaneously, the reliability evaluation rules can exclude unreliable epoch and improve the positioning results, especially on smartphone data.