Candidate density as boundary-matching difficulty
GPS research on roads and sidewalks knows a similar problem. Weng et al. (2025) need a special sidewalk-matching method, because a sidewalk is only about 2 m wide, while GPS error between tall buildings reaches 18.4 m (RMS) when standing and up to 32 m across the street when walking. Quddus et al. (2007) report three related problems: point-to-line matching becomes unstable when roads are close together, one matching method gets worse on parallel streets, and at a Y-shaped junction both roads can be almost equally close.
Candidate density bias toward building is a similar problem at a larger scale. When many candidates of the same type are packed close together, a small GPS error has more chances to land on a wrong one.
The comparison has a limit. Those papers choose between places of the same kind, for example two roads. Here, the problem is a competition between types: streets lose to buildings. The two problems share one cause, candidates packed close together, but they work in different ways.
A subsection of the Discussion in the thesis. Part of Error analysis.
Linked from Candidate density bias toward building · Error analysis