Urban GNSS Risk Assessment Using High-Definition 3D City Models: Validating the “Building-First” Hypothesis and the Necessity of Infrastructure Integration

Rito Yamasaki

Peer Reviewed

Abstract:	In dense urban environments, Global Navigation Satellite Systems (GNSS) suffer from signal degradation due to NonLine-of-Sight (NLoS) reception and multipath interference. While 3D city models are increasingly used to predict these risks, standard mapping approaches typically prioritize building geometries, often overlooking civil infrastructure. This paper tests the “Building-First” hypothesis—that a minimal 3D map consisting only of buildings is sufficient for navigation safety—and demonstrates its critical limitations. Using high-definition 3D data from Project PLATEAU (Tokyo), we propose a rapid vulnerability mapping framework that integrates overhead infrastructure to correct these safety blind spots. We converted open 3D data for Shibuya into a 5-meter resolution predicted risk map using a two-phase approach: Phase 1 computed a “Risk Horizon” based on building height, and Phase 2 applied a “Hybrid Override Logic” utilizing a binary overhead infrastructure flag. A rigorous field experiment across 45 stratified sites using raw GNSS logs from commodity smartphones revealed that the building-only model achieved a moderate AUC of 0.68. While effective in deep canyons, it critically failed to detect overhead hazards; specifically, it ranked a high-risk underpass (56.74 m error) as safe (Rank 25). The proposed Hybrid Override Logic rectified this failure, reclassifying the site to High Risk (Rank 2) and improving the overall AUC to 0.89 (95% CI: 0.78–0.97). The method achieved discrimination performance comparable to the traditional HDOP metric (AUC 0.84), while effectively avoiding the false-safety signals typical of HDOP in multipath environments. This study confirms that for safety-critical navigation, the 3D mapping of overhead infrastructure is as vital as that of buildings.
Published in:	Proceedings of the ION 2026 Pacific PNT Meeting April 13 - 16, 2026 Hilton Waikiki Beach Honolulu, Hawaii
Pages:	120 - 128
Cite this article:	Yamasaki, Rito, "Urban GNSS Risk Assessment Using High-Definition 3D City Models: Validating the “Building-First” Hypothesis and the Necessity of Infrastructure Integration," Proceedings of the ION 2026 Pacific PNT Meeting, Honolulu, Hawaii, April 2026, pp. 120-128. https://doi.org/10.33012/2026.20587
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In