Infrastructure for Context Driven Pervasive Computing Applications

1
Almost two thirds (63) of Americans say it is
annoying to hear ringing cell phones or cell
phone chatter in public places. Ringing of cell
phones is a big complaint and many times a cause
of embarrassment in classrooms and meeting rooms
How about making the environment smart so that it
takes care of silencing the cell phones even if
the users forget! Will save some embarrassment
for sure, isnt it?
2
Infrastructure for Context Driven Pervasive
Computing Applications

• Presented by
• Vishakha Gupta
• Advisor Prof. Peter Steenkiste
• Reader Prof. Raj Rajkumar

3
Agenda

• Goals
• Scenarios
• Requirements
• Concept
• Thesis Statement
• Motivation
• Related Work
• Architecture
• Evaluation
• History Information
• Conclusion
• Limitations
• Future Work

4
Goals

• In usual applications and services user has to
  initiate some action in order to use the results
• Smart environment for a better user experience
• Focus on area based user tracking in in-building
  environments
• Less attention has been paid to the fundamental
  and challenging problem of providing capability
  to an application in defining physical areas
• Determine with high probability when the user is
  in the area of interest to the application

5
Scenarios

• When a person enters a secure building, devices
  monitor his motion and warn him if he goes in a
  prohibited area, through a warning message passed
  to his cell phone
• In an auditorium, where a show or some program is
  going to begin, everyones cell phone is expected
  to be turned off or to be in silent mode
• When people are seated in the airplane, as the
  plane is about to take off, the cell phones and
  other electronic devices (if possible) could
  receive an interrupt indicating they should turn
  themselves off.
• At places such as a meeting hall, a classroom, a
  hospital, cell phone rings may cause disturbance.
  At the same time, people may want to attend their
  calls. So the devices should be signaled to
  change over to the silent mode

6
Requirements

• A method to define the regions of interest to the
  application
• An infrastructure to enable area based tracking
  for the client devices in an establishment
• Need for downloading of software on a handheld or
  including the APIs necessary for communication in
  the handheld by the device manufacturers
• Authenticity of the code getting downloaded and
  the source of the messages
• No effect on the normal communication or device
  use
• Interoperability and other challenges associated
  with a distributed system
• Generic APIs to help accommodate new applications
  that are developed

7
Concept
8
Thesis Statement

• Present an infrastructure for context-driven
  applications enabling them to specify area-based
  user tracking requirements
• Ability to determine with high probability when
  the user is in the area of interest to the
  application

9
Motivation

• Minimal change in the infrastructure.
• Easy testing and deployment
• No requirement to formulate a coordinate system
• Flexibility to an application in defining the
  attributes of an area as needed
• No restriction in terms of the shape of an area
  to be defined by the application

10
Related Work

• Area aware computing - relatively new concept
• Research work that comes really close to the
  concept presented in this thesis
• Area based triggers by Hermann et. Al.
• Fingerprinting using access points as in PlaceLab
  project by Intel

11
Related Work 2

• Other related projects
• RADAR project for in-building user tracking by
  Microsoft Research
• AURA hybrid space model by CMU
• CRICKET project by MIT

12
Terminology

• Zone A physical area defined by an application
  for its use. For example, the area where cell
  phones must be turned to silent mode like in an
  auditorium
• Region Any area in a building which could be a
  zone
• Signature Tuple consisting of ltAccess Point
  MAC, RSSI_min, RSSI_max, Weightgt used by the
  application to define zones in WiFi signal space
• Client Signature - Tuple consisting of ltAccess
  Point MAC, SSID, RSSI heardgt read by a client
  device from its wireless interface
• Region Definition or Rule A tuple of the form
  ltRegion ID, Signaturegt used in defining zones

13
Terminology 2

• Location A tuple of the form ltX, Y, Height,
  Descriptiongt defined keeping in mind a
  requirement of having actual Euclidean
  coordinates if needed anytime by the system.
  Currently we use the description member to make
  the location more meaningful wherever it is used.
  
• Percentage Match Its the probability with
  which the client signature matches a zone
  definition using the algorithms described later
• Message Its a string used as an attribute for
  the zone, defined by the application if it wants
  to convey something to the user when he is mapped
  to a certain zone
• Action The action that the application expects
  users to perform when they are in a certain zone

14
Architecture
15
Architecture Offline Zone Definition

• Repository
• Stores access point, zone and rule information
• Repository Manager
• Interface for application to provide requisite
  information

16
Architecture Offline Zone Definition 2
17
Architecture Server

• Listener Listens for client requests consisting
  of client signature
• Rule Manager
• Uses rule and zone information from the
  Repository
• Uses algorithm to find a matching zone for the
  current client signature
• Sender Sends the zone information to client
  with message and action

18
Architecture Client

• Wireless Device Reader reads wireless card
  information and forms client signature
• Sender sends client signature to the server for
  zone matching
• Listener receives zone information from server
• UI Component interacts with the user if
  required by the application

19
Algorithms

• begin procedure find-match
• Let A be the best heard access point by the
  client
• Let L denote the client signature
• Let M denote the list of zones with signatures
  consisting of A
• loop for all the zones Z in M
• store best zone(s) found
• end loop
• return best heard zone(s)
• end procedure

find-percent-match(Z, L)
20
Algorithms Exact Match

• begin procedure find-percent-match
• count number of access points in signature S
  used in definition of Z
• loop for all access points B used in S
• if (B exists in L and RSSI heard for it is
  within RSSI_MINA and RSSI_MAXA)
• increment match
• end loop
• percent match /count 100
• return percent
• end procedure

21
Algorithms Deviated Match

• begin procedure find-percent-match
• count number of access points in signature S
  used in definition of Z
• drop_per_deviation 0.2
• loop for all access points B used in S
• if (B exists in L and RSSI heard for it is
  within RSSI_MINA and RSSI_MAXA)
• increment match
• else
• dMatch (1 (deviation in RSSI from
  MIN or MAX) drop_per_deviation)
• if dMatch gt 0 then match dMatch
• end loop
• percent match /count 100
• return percent
• end procedure

22
Algorithms Weighted Match

• begin procedure find-percent-match
• count number of access points in signature S
  used in definition of Z
• loop for all access points B used in S
• if (B exists in L and RSSI heard for it is
  within RSSI_MINA and RSSI_MAXA)
• match WeightA
• end loop
• percent match /count 100
• return percent
• end procedure

23
Algorithms Weighted Deviation Match

• begin procedure find-percent-match
• loop for all access points B used in S
• if (B exists in L and RSSI heard for it is
  within RSSI_MINA and RSSI_MAXA)
• match WeightA
• else
• dMatch (1 (deviation in RSSI)
  drop_per_deviation)
• if dMatch gt 0 then match (dMatch
  WeightA)
• else if (dMatch lt 0) then
• dMatch exp (-WeightA
  deviation in RSSI / 100) match dMatch
• end loop
• percent match /count 100
• return percent
• end procedure

24
Evaluation Define Zones
25
Evaluation RSSI Measurement
26
Evaluation Entering Information

• Room 105 (Area 44.6 sq.m) can be defined as
• lt00022D04683B,-80,-70,1.5gt
  lt00601D23C5B5,-80,-70,1.0gt
  lt00022D51A90A,-80,-70,1.5gt
• Reception (Area 38.64 sq.m) can be defined as
• lt00022D04683B,-80,-70,1.0gt
  lt00601D23C5B5,-80,-70,1.25gt
  lt00022D51A90A,-65,-55,2.0gt
• Room 127 (Area 8.18 sq.m) can be defined as
• lt00022D04683B,-60,-50,2.0gt
  lt00601D23C5B5,-75,-65,1.0gt
  lt00022D51A90A,-80,-70,1.0gt

27
Evaluation Comparison of Algorithms
28
Evaluation Comparison of Algorithms 2
29
Evaluation Comparison of Algorithms 3
30
Evaluation - Conclusion

• Deviated Match algorithm better than remaining
  with number of access points(N) gt 3
• The number of access points determined by the
  zone under consideration
• The Weighted Deviation Match algorithm shows more
  consistency in accuracy
• Reduces spurious results acquired due to the
  exponential degradation
• Example on next slide

31
Evaluation Conclusion 2
Client signature at 1 lt00022D51A90A, -76gt
lt00022D04683B, -43gt lt00141B5A2290,
-89gt Zone definition for Room 127
lt00022D04683B,-60,-50,2.0gt
lt00601D23C5B5,-75,-65,1.0gt
lt00022D51A90A,-80,-70,1.0gt Zone definition
for Room 105 lt00022D04683B,-80,-70,1.5gt
lt00601D23C5B5,-80,-70,1.0gt
lt00022D51A90A,-80,-70,1.5gt

• At one point, Room 105 matched at Position 1 by
  Exact Match, Deviation Match and Weighted Match
• The Weighted Deviation Match algorithm showed
  Room 127

32
Evaluation - Limitations

• Only defined zone is Reception
• But Rooms 101, 103 and Lobby also show a match
  for Reception using any of the four algorithms
  presented

33
Evaluation Limitations 2

• Only defined zone is Room 127
• But the shaded portion on the floor plan also
  shows a match for Room 127 using any of the four
  algorithms presented

34
History Information

• Use past area information to conclude strongly
  about the present position
• Use bluetooth devices to conclude strongly about
  the present position
• New term Space - An area that is not defined by
  the application but which could be of consequence
  in defining zones
• E.g. A corridor in the building which might be
  leading to a zone
• Clients modified to report any nearby bluetooth
  devices as well as the previous area matched with
  a timestamp

35
History Using Space Information

• Consider two applications
• One has defined Zone Y
• The other has defined Zone Z
• At time t, the system knows that a device is in
  Space X (maybe by using one of the signature
  matching algorithms mentioned earlier)
• At time t 1, the client signature says Space X
  with timestamp t
• Higher probability that user must be in Zone Y
  for application 1 while in Zone Z for application
  2
• Example follows

36
History Using Space Information 2
37
History - Bluetooth
38
History New Algorithm

• begin procedure find-percent-match
• Let bDevice represent a bluetooth device
• if timestamp of client signature and server do
  not differ by MARGIN
• if bDevice heard by client and bDevice
  identifies Z
• return complete match
• if Space S heard by client before this iteration
  and S leads to Z
• perform signature based match
• percent (weight1 space match percent
  weight2 signature match percent) / 2
• return percent
• perform match as in previous signature based
  cases
• return percent match
• end procedure

39
Conclusions

• Solution for area aware computing using existing
  infrastructure
• Implemented and analyzed four algorithms for WiFi
  signal space matching of zones
• Weighted Deviation Match algorithm works best in
  general
• Improvement in identifying zones using history
  information

40
Limitations

• Zone definitions in WiFi signal space
• Configuration of access points
• Performance at different times of day
• Varying signal strength
• Bluetooth
• Chances of having Bluetooth devices installed

41
Future Work

• Experimental evaluation of history information
• Study the scalability of the system by
  introducing multiple clients
• Account for network usage and computation
  requirements on the client
• Implement an end to end system involving
• Download and verification of software on multiple
  devices
• Study the variation pattern in RSSI to define a
  model
• Define a variable signature depending on time of
  day etc. constituting an intelligent zone
  definition

42
(No Transcript)
43
Thank You
44
Implementation - Server
45
Implementation - Client