Installation guide

19
Wi-Fi Location-Based Services—Design and Deployment Considerations
OL-11612-01
Cisco Location-Based Services Architecture
pre-packaged models enable calibration-less deployment in common office environments, which is a
significant advantage over approaches such as location patterning, especially in cases where easy and
rapid deployment is the primary concern.
In addition to the use of pre-packaged propagation models, RF Fingerprinting offers the ability to
develop a customized propagation model that enhances the default path-loss models based on an on-site
calibration phase. This process allows for the overall attenuation characteristics of the actual
environment to be taken into consideration during the calculation of both 2.4 GHz and 5 GHz path loss
exponents. For each calibration grid location, the physical location coordinates of the calibration client
(provided by the calibration operator) are recorded along with the client RSSI from three or more
LWAPP-enabled access points. This is performed until 150 location-to-access point measurements are
recorded per band from 50 distinct locations in the target environment.
The data accumulated during the calibration phase is statistically processed and groomed, then used to
build an RF propagation model where the path loss exponent, shadow fading, and PL
1meter
values are
calculated from the sample calibration data so as to better reflect specific propagation anomalies (such
as attenuation) that are present in the environment. This process consists of several computational cycles
where the previously-mentioned parameters are calculated for each band. The minimum mean square
error (MMSE) estimation technique is used to obtain the initial values for the parameters, as shown in
Figure 7, where the path loss exponent is represented by the slope of the applicable MMSE line of best
fit (that is, either default or corrected fit). However, note that in the RF Fingerprinting approach, the
selection of a path loss model does not end with MMSE. Rather, MMSE is used only as the starting point
for the selection of finalized parameters for each band, with the ultimate goal being the optimization of
the path loss model as it pertains to location accuracy instead of merely obtaining the best MMSE fit to
the calibration data.
Figure 7 MMSE Estimation
To locate a mobile client during the operational phase of RF Fingerprinting, RSS lateration is performed
using either a default RF model or a customized model created during the calibration phase. This process
yields the location(s) where the highest likelihood of client residence exists. Additional information
gleaned from statistical analysis of the distribution of calibration data is then used to further improve
location accuracy and precision over that of pure RSS lateration approaches.