Improving Smartphone Pedestrian Positioning in Urban Canyons Through PDR-Constrained GNSS Shadow Matching

Smartphone global navigation satellite system (GNSS) positioning in building-obstructed environments is degraded by non-line-of-sight reception and multipath. Visibility-based shadow matching (SM) exploits three-dimensional building models to constrain cross-street position but has limited along-street observability. This study proposes a pedestrian dead reckoning (PDR)-constrained SM framework with two implementations: extended Kalman filter fusion (EKF-PDR/SM) using heading-orthogonal candidate search and anisotropic SM measurement uncertainty, and PDR-driven SM using candidate propagation, visibility scoring, and GNSS-proximity weighting. Evaluation used one smartphone and one pedestrian completing one traversal of each of two campus routes with different satellite-visibility conditions. Under common PDR prediction and process-noise settings, EKF-PDR/SM achieved cross-street/along-street root mean square errors (RMSEs) of 3.57/8.28 m on Route 1 and 2.46/7.83 m on Route 2, compared with 5.23/8.98 m and 4.29/12.38 m for EKF-PDR/GNSS, respectively. PDR-driven SM achieved the lowest directional RMSEs among the six compared methods: 2.92/7.47 m on Route 1 and 1.54/5.89 m on Route 2. All errors were evaluated relative to common model-derived reference trajectories. These controlled results support the use of PDR constraints to improve visibility-based SM under the tested conditions.