3D Building Model-Assisted Snapshot GNSS Positioning Method

Abstract

Global Navigation Satellite Systems (GNSS) have proven to be a viable and reliable solution in interference-free environments and in presence of Line-of-Sight (LOS) signals only. However, in urban canyons, multipath signals directly affect the pseudorange measurements resulting in degraded positioning performance of traditional GNSS receivers. Moreover, traditional GNSS receivers cannot distinguish between non-LOS (NLOS) and LOS signals, resulting in even worse performance if the receiver tracks NLOS-only signal. Hence, NLOS and multipath signals remains a dominant source of error in satellite-based navigation. Most of the existing research has focused on identifying and rejecting NLOS measurements. However, little research has used NLOS signals constructively. In this regard, this research uses snapshots of GNSS data in order to estimate position, utilizing all NLOS signals constructively with the help of a 3D Building Model (3DBM). Using a 3DBM and a ray-tracing algorithm, the number of reception paths and the corresponding path delays of reflected signals is predicted across a grid of candidate positions. These predictions are then used to compute least-squares fit to the GNSS receiver’s correlator outputs and the position with smallest residuals is selected as the position estimate. This approach is termed Signal Delay Matching (SDM) and yields a solution that is nearly unaffected by traditional GNSS error sources, and has capability of providing a position solution using a single satellite only. The use of snapshots of data mean the receiver need not perform tracking operations, thus making it easier to implement and power efficient.

The feasibility and performance of the algorithm was tested using data collected in downtown Calgary, Canada, where buildings reach heights of over 200 m. Contrary to traditional approaches, results for the proposed method show that positioning error decreases as sky-visibility decreases. For sky-visibility below 20%, the median error was found to be just over 3 m. Compared to two pseudorange-based receivers, the proposed method yields RMS errors improvements of 22% to 48% in the horizontal plane.