IEEE Systems, Man and Cybernetics Magazine - October 2020 - 28

make convergence upon a position
Wi-Fi Mapping and
difficult when the models are utiModeling
If the path around the
lized for positioning.
The Wi-Fi mapping and modeling
When the area of interest has a
method utilizes three separate
region of interest is
closed, or near-closed path around
stages prior to presenting a probanot closed or nearly
it, models for the RSSI can be genble location utilizing an RSSI (see
closed, the models
erated. The models are based upon
Figure 3). The first stage depends
Delaunay triangulation. The vertiupon a 360° survey while logging
will have jagged
cal component of the populated is
both the ground truth in addition
points of inflection or
based off of a linear interpolation
to the down-selected sets of BSSID
of the measurements of the point
RSSI values. One industry where
points of focus.
map. Once the models are generatthis method has direct application
ed, they must be validated against
is in firefighting.
each other.
Structural firefighters often
Model validity is accomplished by finding the composmake a 360° survey of a property prior to making entry.
ite difference between the models. Any regions where the
This example will be utilized for illustrative purposes. Fire
models have near zero difference suggest singularities in
apparatus in the example are marked as E51, E53, and E59.
those physical locations. If the model is singularity free
Two of the apparatus are modeled as equipped with wireand near closed, the RSSI can be filtered to find regions
less APs. Although this example gives APs of a known oriwhere the current measurements, when fused with the
gin, it is not necessary to know where the Wi-Fi signals
positioning model, converge to form a region of possioriginated from. The experimental results leveraged inforble locations.
mation from Wi-Fi signals of unknown location.
A point map is generated based on the RSSI and the
GNSS position transformed to the tangent frame as a
Point Map Generation
ground truth. Errors in either the RSSI or a drift in ground
It becomes convenient to utilize the t-frame calibrated reltruth during this period of the model generation will casative to a single point (see Figure 4). This allows for units
cade to errors in the model around that point. These errors
to stream locations with minimal overhead within the
local area. The result is a set of xy coordinates relative to
an initial measurement point. The fusion of the spectral
and spatial measurements produce a map of points in 0 6 .
The first four dimensions of this map represent spectral
C

E59

R10

POV

E53

E51

A
Figure 4. Firefighters and first responders commonly

mark the scene of action with A, B, C, and D on the
edges of a scene sketch. Fire apparatus are often
denoted with an "E" for engine and an "R" for rescue
unit. The numeral code notes the individual unit
number. In keeping with this convention, the RSSI point
map is shown with measurement points depicted as
circles, the sides of the scene of action noted with A,
B, C, and D and the engines and rescue units noted
within the figure. POV: privately owned vehicle.
28

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE O ctober 2020

A
Figure 5. A triangulated point map with

measurement points depicted as circles. The model
itself cannot take into account the position of the
apparatus because that information, including the
general layout, is based off of visual data from the
first responder.

IEEE Systems, Man and Cybernetics Magazine - October 2020

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - October 2020