Enhancement of Smartphone Positioning for Urban Canyons: A Consensus-Based Recovery Vector Algorithm with GNSS Visibility Maps

Jungyo Park, Jeonghyeon Yun, Byungwoon Park

Peer Reviewed

Abstract:	The accuracy of Global Navigation Satellite System (GNSS) positioning is a critical factor for ubiquitous Location-Based Services (LBS). However, this accuracy is severely compromised in urban environments due to significant multipath and NonLine-of-Sight (NLOS) errors caused by high-rise buildings. This challenge is further compounded on smartphones, which suffer from measurement degradation due to inherent hardware limitations, such as linearly polarized antennas. Consequently, improving smartphone positioning in urban canyons remains a fundamental obstacle for advanced LBS applications. To mitigate these effects, 3D-mapping-aided (3DMA) techniques have been proposed; however, they often face computational bottlenecks due to the large uncertainty of the initial position. Our previous work addressed this by introducing the 'Recovery Vector' concept to decouple error estimation from iterative validation. While effective, the original method relied on a single pivot satellite, which can introduce bias if contaminated by multipath, especially in dynamic vehicle environments. This study enhances the algorithm for the smartphone environment with three primary contributions. First, we propose a robust 'Consensus-based Recovery Vector' technique that leverages multiple high-elevation satellites to filter out unreliable pivot candidates, ensuring unbiased multipath estimation. Second, we improve probabilistic map selection by integrating dual-frequency (L1/L5) measurements, applying independent probability models to account for their distinct signal characteristics. Third, we introduce a precise velocity estimation method using a Doppler residual-based weighting scheme, which provides reliable prediction updates for a continuous Kalman filter. The proposed framework was validated using dynamic data from Teheran-ro, a deep urban canyon in Seoul. The experimental results demonstrate significant performance improvements. While the conventional DGNSS showed a horizontal RMS error of 30.00 m, the proposed Consensus-based method achieved a horizontal RMS of 4.72 m and a 95th percentile error of 7.71 m, outperforming the Single Pivot method (RMS 6.13 m). Furthermore, the velocity estimation yielded a horizontal RMS error of 0.40 m/s, confirming the robustness of the proposed system for continuous urban navigation.
Published in:	Proceedings of the 2026 International Technical Meeting of The Institute of Navigation January 26 - 29, 2026 Hyatt Regency Orange County Anaheim, California
Pages:	540 - 549
Cite this article:	Park, Jungyo, Yun, Jeonghyeon, Park, Byungwoon, "Enhancement of Smartphone Positioning for Urban Canyons: A Consensus-Based Recovery Vector Algorithm with GNSS Visibility Maps," Proceedings of the 2026 International Technical Meeting of The Institute of Navigation, Anaheim, California, January 2026, pp. 540-549. https://doi.org/10.33012/2026.20562
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In