ASCENT

1
Center for Embedded Networked Sensing
Omni-directional WiFi Localization
Rohun Bansal, Jessica Chen, Kevin Chen, Nathan
Schloss, Deborah Estrin, Cameron Ketcham, Rakhee
Patel, Martin Lukac OWL - http//owl.t4so.com/
Introduction Determining location through use
of sensors
Sensors on Mobile Phones
OWL Localization System

• Modern smart phones (such as the Android G1) have
  a variety of sensors
• GPS sensors can generally acquire fairly accurate
  location data.
• WiFi devices can scan for access points as well
  as create ad-hoc networks.
• Mobile phones are becoming increasingly more
  widespread
• As the number of phones in use increases,
  applications involving networks of phones become
  progressively more practical to implement.

• An Android application for the G1 which takes
  advantage of both GPS and WiFi sensors in a
  network of phones to estimate location
• The system is generally functional in many
  different situations.
• The application works to minimize battery usage
  and time to location fix while trying to improve
  accuracy.

• Relies more heavily on the phone itself for
  location fix

Problem Description Current localization
methods have disadvantages
WPS WiFi Positioning System
GPS Global Positioning System

• Requires line-of-sight view to satellites in
  order to function
• Although GPS is very accurate outdoors under an
  unobstructed sky, it works poorly or not at all
  in environments lacking such a view (notably
  indoors and near large buildings).
• Long time to first fix
• Because each GPS satellite only broadcasts
  ephemeris data every 30 seconds, the time for a
  receiver to get a fix is the time it takes to
  acquire each GPS satellites signal plus up to 30
  seconds for ephemeris information.

• Requires an internet connection
• A list of visible access points must be sent in
  an HTTP request to a central server a location
  estimate will be sent in the server's response.
• Accuracy is largely inconsistent
• Because the system uses a pre-compiled database
  of access points, the accuracy of location
  estimates depends completely on the coverage of
  these access points in the area, resulting in
  rather rough estimates.

Proposed Solution A networked localization
system that uses both GPS and WiFi
OWL Localization System

• Devices that can acquire an accurate GPS fix
  become anchors
• An anchor sets up its own ad-hoc networks, with
  its IMEI number (used as a unique identifier) and
  GPS location encoded into the network SSID.
• Devices that cannot acquire an accurate GPS fix
  become receivers
• A receiver periodically scans for available
  wireless networks set up by anchors.
• Distance to each anchor network is estimated
  using a simplified logarithmic correlation to
  signal strength (in dBm).
• Once there are at least 3 visible anchor
  networks, a receiver can estimate its position
  using multilateration.

X
80 m
30 m
Z
Y
50 m
Inside Building
A

• Screenshots of the OWL application
• initial check of GPS accuracy (left)
• receiver mode (right)

A can determine its location through
multilateration if it knows the positions of X,
Y, and Z, and its respective distances to them.
Comparison with GPS and WPS
Results
Further Work

• Average accuracy of receiver mode is almost that
  of GPS
• 29.16 of receiver calculations were 11m or
  better.
• 21.09 of GPS locations were 11m or better.
• Location fix time was lower than both GPS and WPS
• Battery usage was less than WPS and similar to GPS

• Fix issues with ad-hoc networks
• There were numerous problems with the anchor
  networks, including limited range, inconsistent
  signal strengths, and slow update rates.
• This seems to be a mainly hardware-specific
  issue.
• More testing
• The system needs testing in more controlled
  environments.

UCLA UCR Caltech USC UC Merced