The NearMe Wireless Proximity Server

1
The NearMe Wireless Proximity Server

• John Krumm and Ken Hinckley
• Microsoft Research
• Redmond, WA, USA

2
The Basic Idea
802.11 Wi-Fi access point
NearMe Proximity Server
3
Location vs. Proximity
x1 (x,y) location
x2 (x,y) location
d12 f(x1, x2)
s1 measured signals
s2 measured signals
4
Who Cares?
Schilit, Adams, Want, Context-Aware Computing
Applications, 1994.

• Which of my friends, colleagues, relatives, or
  enemies are nearby?
• Who are these strangers with me?
• What are the names of nearby printers?
• What conference rooms are nearby?
• How far away is the nearest receptionist desk?

Other elements of context velocity, mode of
transportation, task, goal, mood
5
Other Proximity Methods

• Location from
• Active badges
• Cameras
• Wi-Fi
• GPS
• Cell Phones
• Conference Devices
• SpotMe
• nTAGTM
• IntelliBadgeTM - NCSA
• Conference Assistant Georgia Tech
• Proxy Lady Viktoria Institute
• Digital Assistant ATR
• Synchronous Context
• Synchronous Gestures Microsoft
• Are You With Me? U. Washington
• SyncTap Sony
• Stitching Microsoft

SpotMe
nTAGTM

• Services
• Trepia
• iChat Apple
• Short Range Networking
• Infrared
• Bluetooth

6
NearMe Client
PocketPC 2003
Windows XP

• Requirements
• Windows XP
• WWW access
• Microsoft .NET Framework

7
NearMe Client Test Connections
8
NearMe Client Register

• Register with
• Name
• Email (optional)
• URL (optional)
• Expiration interval

9
NearMe Client Report Wi-Fi

• List of detectable Wi-Fi access points
• Access points used only as beacons
• Periodic reports for mobility

10
NearMe Client -- Query
Adjustable Look back time to filter outdated
reports
11
NearMe Client Nearby Things
12
NearMe Server
SQL Server
.NET Web Service
Clients
13
NearMe Server Data
Expiration deletes source of Wi-Fi data, but not
Wi-Fi data itself
14
Access Point Layout
1
2
A
F
3
15
Hops Distance Estimate
Extends proximity range beyond radio range
16
Access Point Travel Times
Minimum observed client travel time between
access points

• Recomputed every hour
• Could be computed from other wardriving data,
  e.g.
• Intel Research Place Lab
• NetStumbler
• To Do
• Path plan, e.g. A?C A?B B?C
• Cluster times, e.g. times for walk, bike, drive

17
Travel Times Distance Estimate
Observed lower bound on travel time
18
Distance Estimation
How to estimate distance between things?
19
Distance Functions
Wi-Fi Signature s (mac1, rssi1), (mac2,
rssi2), , (macn, rssin)
sa
sb
d f(sa,sb)

• Candidate Distance Features
• nn number of access points in common
• ssdn sum of squared differences of rssi in
  common
• nu number of access points not in common
• ?s Spearman rank coefficient of access points
  in common

20
Compare Ranked Signal Strengths
mac2 mac3 mac3
mac1 mac2 mac3
-1 lt ?s lt 1
Spearman rank ?s
RightSPOT, UbiComp 2003
21
Distance Functions
d f(sa,sb)
d f(nn, ssdn, nu, ?s)
Example d ann2 bssdn cnu3 d?s
4 features polynomials to 3rd degree 45
different test functions
22
Test Data

• Wi-Fi Cards
• Dell built-in laptop
• Microsoft Wireless USB
• ORiNOCO PC Card
• Actiontec USB
• Cisco Aironet PC Card
• Linksys USB

Office Building 1,441,739 pairs Cafeteria
572,027 pairs
23
Simple Distance Function
d -2.53nn 2.90?s - 22.31
rms error 14.04 meters ?s 0.39
24
Applications
Stitching
Look up URLs of nearby people
Send email to people nearby
25
What Makes NearMe Different?

• Works on existing Wi-Fi devices
• Minimal setup
• Out of the box for nearby people
• Capability grows with use
• Short range proximity
• Accounts for different Wi-Fi hardware
• Sorted by distance
• Long range proximity
• Hop distances via adjacency analysis
• Travel times via timestamp analysis

26
Demo