Bootstrapping location on the World Wide Web

1
Bootstrapping location on the World Wide Web
Bill Schilit, Intel Research

• Anthony LaMarca, Anand Balachandran, Yatin
  Chawathe, Sunny Consolvo, Ian SmithIntel
  Research Seattle
  
• Gaetano Borriello, Jeff Hightower, David
  McDonald, Eithon Cadag, Jason Tabert, Ed
  LazowskaUniversity of WashingtonBill
  GriswoldUC San Diego
• Jason HongUC Berkeley

2
When dinosaurs walked the earth
3
PARCTab System (c. 1993)
For physical world computing, whats
different?what infrastructure is needed?
Roy Want, Bill Schilit, Norman Adams, John Ellis,
Karin Peterson, David Goldberg, Mark Weiser
Marvin Theimer
4
Becoming Physicalintimately connected with the
world around

• Information
• Getting information that applies
• Control
• Interacting with devices
• Making the environment responsive
• Communication
• Connecting with the right people at the right time

5
Context-Aware Computing
Location-based File System
Proximate Selection Widget

• Room level precision is good enough for a class
  of useful applications
  
• Lack of ubiquity was a big problem
• Privacy was a big concern

6
Mobisaic Browser (c. 1994)

• Mobile location-enhanced Mosaic
• Dynamic URLshttp//places/(Location).html
• Directions to closest suitable bus stop
• Active Documents
• Scripting inside comments subscribes to dynamic
  environment variables
  
• Reload page when location changes

Geoff Voelker Brian Bershad
7
10 years later location and computing have yet
to converge
Automotive GPS Chartplotter Fishfinder Aviatio
n
GPS Radio GPS handheld E911 phone
Mapquest Yahoo yellow pages Target store locator
Washington DOT Traffic Traffic Cams CompUSA in
-store pick-upBusview
8
Place Lab Research Challenge

• Challenge at MOBICOM WMASH workshop Sept 03
• Take location-aware out of the laboratory into
  the real world
  
• A multi-organization research collaboration for
• Designing, building, and distributing an open
  positioning test-bed
  
• Bootstrapping user and developer communities
  across many campus place labs
  
• Our approach

still fresh!
value to user
cost of use
9
What do we want to enable

• Widely available location-enhanced computing
• Notebooks, PDAs, and smart phones
• Build a positioning system that is
• Wide-scale, indoor outdoor
• Can be used everywhere
• Privacy-observant, low barrier to entry
• Can be used by everyone

10
Enabling virtually free device positioning

• Exploit wide-scale WiFi deployment
• Urban areas have dense WiFi coverage
• WiFi base stations beacon unique IDs
• Position WiFi devices using a map of
  base-station-IDs to location
  

11
BEACON VISUALIZATION
12
UbiComp 2003 Demo

• Client-side conference companion application
• Hotspots images, commentary, web content
• Available to conference attendees
• Demonstration and an invitation

13
Why not use GPS?

• GPS works for trucks, hikers, and cruise
  missiles
  
• GPS doesnt work indoors
• GPS works poorly in urban canyons
• This is where many people spend the majority of
  their time
  
• More computers with WiFi than computers with GPS
• But Jeff Hightowers Location Stack Technology
  will be used for sensor fusion of GPS and radio
  beacons

14
Virtually Free WiFi positioning You get what
you pay for

• Signal strength field of a single access pointin
  Lawrence Kansas

15
Components of the test-bed
Location-enhanced web services (online and OCC)
Seattle Wireless Club
Intel
Location-enhanced requests
Service data
Usage logs
Data snapshots
University of Washington
User device
RF beacon mapping service
16
What are the hard problems?

• Building the world wide database of AP positions
• Data collection
• Data munging - AP position estimation algorithms
• 2. Providing a privacy model acceptable to users
• Who computes my location?
• How do I control use of my location information?

17
Whats our approach to construct a global
database?

• RF beacon map is user-contributed
• Leverage war drivers and WiFi clubs
• Also collect AP logs from volunteers
• users without GPS devices
• beacon_at_home (sensor logs not CPU cycles)
• Privacy by segmenting logs by AP and time-shift

open
closed
WEP
18
Managing the user-contributed data

• Solution Distributed database with reputation
  management
  
• Distribution for trust - organizations want
  control over their own data
  
• Distribution for integrity making it harder to
  spoof a geo region
  

Seattle Wireless Club
Intel
Anonymous trace data reputations
Traces
Multi-server estimate
client
University of Washington
19
Mapping RF beacons in the wild

• Problem Signal strength alone is not sufficient
  to accurately map base station locations
  
• Solution Add contextual information to improve
  accuracy
  
• Apply techniques developed at IRS/UW for
  correcting GPS data and integrating multiple
  sensor readings over time
  
• Mine GIS maps for structural information e.g.,
  roads, building
  
• Learn simple models of AP radio propagation

t4 -56 db
t3 -40 db
t1 -95 db
t2 -46 db
Don Patterson, Dieter Fox, Dirk Haenel
20
Mapping RF beacons in the wild (cont)

• War drivers will not map all base stations
• Leverage users RF traces to complete the map
• Users can contribute a log of base stations they
  see
  
• (t1,b1) , (t2,b2) , (tn, bn)
• Refine base station locations using these logs
• Segmenting local processingfor privacy

21
Building a critical mass of data

• Beacon_at_home
• Make compact sniffers based on personal server
  platform
  
• PS WiFi Platform for users to carry
• PS WiFi GPS Wardriving package
• Deploy sniffers to gather data

22
How are we addressing privacy concerns?

• The computation trust problem is mostly solved by
  passive sensing RF beacons client-side
  computation
  
• Manage web access privacy by giving users
  visibility and control of how location is
  divulged
  
• Create privacy mechanisms to set level of
  disclosure
  
• Users should be able to reveal, hide, blur, even
  lie about location
  
• Create privacy policies to allow users to express
  preferences (UI)
  
• Trust Yahoo completely, give any of Maxs friends
  my city

23
How we are going about it

• Run project as open, collaborative effort
• Engage top researchers in the field
• Produce open-sourced artifacts
• Ensure interoperability, encourage
  experimentation via modular architecture
  
• Provide first Beacon Location database
• Share AP sniffer hardware to fill the database
• Offer a courseware for web programming classes
• Mapping WiFi on campuses
• Build a developer community for our new services
• Fostering application development

24
Campus Place Labsengage campus community
Research Faculty
Web, Ubicomp, Dist Sys, Networking Courses (CSE
590GB Location Aware Computing)
Students developers and users
25
Summary
Place Lab is a grassroots effort to create a
global indoor/outdoor positioning system test-bed
with a low barrier to participation

• Opportunities for research collaborations
• Open data through distributed sensing
• Hotspot P2P databases
• Geo-coding web content
• Privacy models and interfaces
• Location aware services

http//placelab.org