Intelligent Urban Positioning Using Smartphone-Based GNSS and Pedestrian Network

Sidewalk-level positions are required for a growing number of pedestrian applications. However, in urban canyons, buildings along both sides of the street severely obstruct global navigation satellite system (GNSS) signals, and the lack of redundant fault-free measurements leads to the poor accuracy in the cross-street direction, posing challenges in determining the side of the street solely based on GNSS positions. While 3-D building models have been utilized to improve position accuracy, particularly in the cross-street direction, techniques relying on these models face issues, such as position ambiguity, high computational load, and low accuracy in the along-street direction. In this study, we aim to develop a novel intelligent urban positioning system using smartphone sensors and pedestrian network. An algorithm is proposed to determine the side of the street by analyzing which half of the sky most of the line-of-sight (LOS) signals are observed. The additional virtual measurement derived from the sidewalk is combined with real measurements to solve GNSS position. It can achieve sidewalk-level positioning since the redundancy in the cross-street direction is significantly improved. The proposed system offers several advantages including elimination of the need for LOS/NLOS signal identification for each satellite and elimination of the need for 3-D building models. Extensive data sets were utilized to train the classification model and evaluate the system's performance. The results demonstrate a correct identification rate of better than 96% using single epoch GNSS observations. More importantly, the proposed positioning system achieves the accuracy of better than 5 m in urban canyons.