Patient Location via Received Signal Strength (RSS) Analysis D. Albano, C. Comeau, J. Ianelli, S. Palastro

1
Patient Location via Received Signal Strength
(RSS) AnalysisD. Albano, C. Comeau, J. Ianelli,
S. Palastro

• Components
• Hardware
• Pre-existing 802.11b infrastructure with at least
  3 APs
• Software
• Rmapr, a radio map generator - generates a
  location-specific radio map
• Creates radio map of a location by recording RSS
  values of reference APs at many points in
  the area
• Generates a list with the form (x, y, RSS1, RSS2,
  RSS3, RSS4, RSS5)
• As map density increases, accuracy increases, but
  set-up time increases as well
• Radio map is stored in server
• Trakr, the main end-user program - compares
  real-time RSS values with radio map data to
  approximate location
• As user moves, the software reads the RSS values
  of nearby APs
• These RSS values are compared to the radio map
• The closest match from the radio map is loaded
  and the location data is read
• This data is interpreted by the software and
  updated in the GUI
• The Viterbi algorithm allows us to predict the
  path and location of the user from the observed
  changes in signal strength
• Makes use of a moving average estimation
• Predicts current/future path based on past
  movement

Abstract The goal of this project is to become
familiar with and succeed in implementing a
successful location-aware computing network. This
position-tracking network will be RSS (Received
Signal Strength) based and will be used to
determine the precise positions of moving objects
in a given space. Initially, the recreation of
a previously established system will be the main
priority of the project. The positioning network
will be implemented by using wireless access
points (AP's) and laptop computer. After the
signal strength is received and logged in the
wireless device, the position of the laptop will
be calculated using a triangulating algorithm.
A radio-map of the sample space will also be
constructed using both (x,y) coordinates and RSS
values from five of the AP's. After mapping the
sample space, the mobile device will also be able
to be located via comparison with the database of
the various radio-map coordinates. The software
will then be calibrated to achieve a certain
level of accuracy within a tolerable
range. After the initial goal is realized, the
technology will then be implemented onto smaller,
more mobile devices such as PDAs and the process
will be repeated. The system will then transmit
data to a server, and the server will be used to
pinpoint the locations of multiple wireless
devices in the sample space. If successful,
multiple GUIs (Graphical User Interfaces) will be
adapted to both the mobile devises and the server
in order to make the system as user-friendly as
possible. Time permitting, additional features
will be added to the system, and the transfer of
information between server and mobile nodes will
also be implemented to add longevity to the
network.
Design Approach We have chosen a design approach
that will expand upon what has already been
accomplished in the field of location-aware
computing. We first intend to replicate the
works of the Microsoft project. We are taking on
the same design approach as they did by using an
802.11b infrastructure using five access points
strategically located throughout the floor of an
office building. We are then going to construct
a radio map a database consisting of all the
received signal strengths and (x,y) coordinates
of the floor. We intend to then use this radio
map as one of the ways of location detection by
taking the five received signal strengths given
from the access points and looking for (x,y)
coordinate pair that matches with those strength
readings.   To elaborate, we intend to place
the access points in the orientation shown in
Figure 1 to the right, with the test bed being
the fifth floor of the Sennot Square building.
The test AP's are Linksys 55AG and the test
computer will be a Lenovo ThinkPad X60 Tablet PC
with an Atheros 802.11a/b/g/n Wireless LAN
Mini-PCI Express Adapter network interface card.
The software application is coded in C using
Wireless Researchers Application Programming
Interface (WRAPI) to extract the RSS values from
the NIC. We are using WRAPI because it is
non-hardware specific, and thus we have the
freedom to perform this on any mobile device with
an 802.11b compatible wireless NIC. Once this
experiment is carried out successfully, the
software used on the laptop will then be
translated over to a PALM mobile device. This
will allow for a more real system involving a
true mobile device that is in use in todays
society. If successful, we will expand on this
simple design and create a user-friendly GUI for
the applications on both the mobile device and
the server. We will also develop, compare, and
combine other methods of location detection in
order to make the application as accurate as
possible. The algorithm will be tuned to produce
the most accurate results possible, and
additional features to the system (including data
transfer) will be added if time permits the work.
Figure 1. Simplified overhead view of 5th floor
Sennot Sq. The five AP locations are shown in
each corner and the center of the floor.
Figure 2. Example window outputs of the Trakr
command-line program back-end. A dynamic,
graphical GUI front-end is in development for
Windows, Linux, and Palm OS.
Radio Map Example
12345678900101 (01,01) lt--(x,y) Coordinate pair
with AP1 -12, AP2 -34, AP3 -56, AP4 -78,
AP5 -90 dBm 23456789010102 (01,02)
lt--Coordinate pair with AP1 -23dBm,
etc. 34567890120103 //This is the radio map of
the 5th floor of Sennot square starting in the
northwest corner //with the positive x direction
being east and positive y direction north. We
took points every //step and our map is outlined
visually. Each point is represented by five RSS
values //of five access points. Each gives us a
RSS in -dBm (we have truncated -100 or less to
-99). //These represent the first ten digits of
the line. The last four digits are the x,y
//coordinate pairs that will be referenced via
the map to approximate location. Notes If the
last RSS string ending in zero (i.e. the AP5 has
an RSS of 90, etc.), the reference will not find
it, as the compare statement will consider it a
null character.
Acknowledgments Thanks to the generous gift of
Drs. Hal Wrigley and Linda Baker, the
Bioengineering Department, Dr. Taieb Znati for
his time, effort, and funding, the Computer
Sciences Department, Bob Hoffman for network
access and troubleshooting, and Anandha Gopalan
for Linux advice and code troubleshooting.