In [10], Fujita et.al. proposed a location estimation method
and system for subway stations using constrained conditions.
In this scheme, specification of a subway station from many
candidates of stations is carried out using Proximity first. Since
each station is far from another station and a set of BSSID
observed on each station is different from another station,
specification of the station is easy. Next a floor of the specified
station is identified from some floors in the station using
Proximity. Last, specification of an area is implemented. They
use Bayesian inference for area specifications.