Parte II: Investigación
2. Trabajo de campo-Intervención
2.2. El taller de arteterapia
As introduced previously, TrackMobile developed the first AVL system which used cellular
base stations as reference transmitters for radiolocation in 1989. In order to avoid any changes in the network infrastructure19 required for timing measurements, this vehicle alarm system
implemented a positioning method based on RSS measurements on the downlink signals from the base stations to the mobile terminal mounted inside the vehicle. Upon intrusion, the latter
sends RSS measurements relative to at least four base stations to a central computer which
18 Only a handful of commercial location-based services were available at the time, most of them using basic cell
identification for positioning [ABI05].
19
49 converts them to ranges and applies a circular lateration algorithm. The latter is a simple
geometrical method which uses the intersection of at least four circles to determine the target position. The range estimation is based on the simple inverse-square law formula [She91].
Without taking into account the large errors in the RSS measurements due to multipath propagation, the positioning outcome was deemed very inaccurate [Hil00].
In 1994, Frederick Leblanc adopted the concept of storing RSS measurements taken at different distances and directions relative to the base station in a database in order to model
the coverage of the BTS as a ‚scaled contour shape‛ *Leb96+. In order to determine the location of the mobile terminal, LeBlanc proposed matching real-time RSS measurements
relative to base stations with their corresponding contours and finding their intersection. The concept of linking ground-truth location with RSS measurements and storing the information
in a database dates back to 1969, when Figel et Al presented an Automatic Vehicle Location system based on signal attenuation [Fig69]. This concept would later become known as
location fingerprinting and become extensively used in both cellular and Wi-fi positioning.
Typically, location fingerprinting methods are based on a two-phase approach.
- A calibration phase is conducted offline to collect all the data necessary to form the so-
called fingerprints and store them in a database. The fingerprints include at least RSS measurements taken next to ground-truth and may contain other signal characteristics
such as the ID of the reference beacons.
- A location estimation phase applies some sort of correlation algorithm to associate real-
time RSS measurements and other signal characteristics to the fingerprints in the database in order to determine the target location.
50 After the E-911 mandate was issued, Oliver Hilsenrath and Mati Wax from U.S
Wireless Corp. developed a network-based location fingerprinting system called RadioCamera. While other time-based techniques aim at identifying and removing multipath
signals before applying the positioning process, RadioCamera uses signal characteristics including multipath patterns to determine the target position [Hil00]. The location estimation
process of RadioCamera is based on the correlation of real-time signal characteristics with signal signatures or fingerprints stored in the database during the calibration phase of the
system. A single base station is required for location estimation but needs to have an antenna array installed. The fingerprints are collected by the BTS which receives the signal
transmitted from the mobile phone inside a vehicle while driving the roads within the coverage area of the BTS. This would later become known as war-driving, although using the
downlink signals instead of the uplink signals. In 1998, U.S Wireless tested their system in the Baltimore area demonstrating that it meets and exceeds the FCC accuracy requirements for E-
911 Phase 2. However, this location fingerprinting method was not standardised mainly due
to the need for installing expensive antenna arrays, which is the same problem the AOA technique faced despite the fact that it was one of the earliest of all cellular positioning
techniques.
In around the same time in Japan, LBS vendor Locus Corp deployed the so-called
personal locator systems for PHS cellular networks. The method they adopted was known as Enhanced Signal Strength due to the fact that it is more robust than standard RSS based
lateration methods in the sense that they cope with multipath propagation effects. The database of fingerprints is collected using a propagation prediction tool which uses terrain
51 stations. Enhanced Signal Strength worked both indoors and outdoors achieving mean
accuracies of 40-50 metres [Kos00].
In 2001, Laitinen et Al introduced the Database Correlation Method (DCM), a location
fingerprinting method which estimates the position of a GSM mobile phone using RSS measurements on the downlink signals [Lai01]. The fingerprint database was constructed
using real RSS measurements in a dense urban environment (Helsinki, Finland). A correlation algorithm is used to extract the best matching fingerprint, whose position is taken as the
estimate, based on the nearest neighbour data association method. The latter consists of comparing the real time measurements against the calibrated measurements using the
Euclidean Norm as a metric to find the closest fingerprint in signal strength space20. The accuracy achieved by DCM as mentioned in [Lai01] is shown in table 2.2.
Table 2.2 DCM accuracy
Environment Rms 67% Rms 90%
Urban 44 m 90 m
Suburban 74 m 190 m
Given the signal strength measurement taken by the mobile terminal and one of the fingerprints belonging to the ith BTS , the Euclidean distance function is defined as:
√(
)
(2.5)
20
A WLAN localisation system known as RADAR adopted a similar correlation algorithm to find the best matching fingerprint [Rad00].
52 The accuracy of the DCM method depends on the resolution of the fingerprint database and
performs better in indoor and dense urban areas [Lai01]. The resolution refers to the spread of fingerprint samples across the trial area which depends on the sampling frequency of the
fingerprinting process. In order to achieve the high accuracies Laitinen relied on an extensive calibration phase which collected thousands of fingerprints in a relatively small areas.
DCM caught the attention of other researchers in the cellular positioning community who usually use the acronym DCM to refer to location fingerprinting applied to cellular
positioning as opposed to WLAN fingerprinting for instance.