Virtual WiFi Network Database Construction for Positioning Services

1
Virtual WiFi Network Database Construction for
Positioning Services
CFI 2009 2009. 06. 19. Jinyoung Han, Jeongkeun
Lee, Munyoung Lee, Kideok Cho, Taejoon Ha, Ted
Taekyoung Kwon, Yanghee Choi
2
Contents

• Introduction
• Related work
• VNDB Construction
• Experiment Results
• Conclusion

3
Introduction

• Context-aware/Location-based Services will be
  pervasive in future Internet
• Quality of services will be mainly determined by
  the location accuracy

4
Localization

• Network-based localization using NDB

Calculating my position
AP1
AP4
RSS1
ltNDBgt
RSS2
How to construct a NDB for localization?
RSS2
AP2
AP3
5
Network Database (NDB)

• Network-based localizations employ network
  database (NDB)
• NDB maintains the location information of
  reference points
• Real NDB (RNDB)
• Network providers maintain real locations of
  reference points
• Ex. in cellular networks
• Virtual NDB (VNDB)
• Estimated locations of reference points (or
  APs)
• Due to reference point installation of private
  users
• Accuracy is the key of VNDB
• Ex. In WiFi networks

6
War-driving

• Constructing VNDB through war-driving

AP1
RSS2
RSSk
RSS1
(37.123,127.123)
(37.223,127.123)
(37.441,127.111)
ltMaking AP map! - VNDBgt
7
Related work

• War-driving
• Signal collecting process by driving through the
  area of interest
• Many academia or industrial groups exploit the
  war-driving data for localization
• However, the accuracy of war-driving data is
  often disregarded
• M. Kim et al. highlighted the proper
  war-driving strategy for path-planning
• Measurement points might be biased toward
  war-driving paths
• However, some refining strategies for war-driving
  data are also needed

8
Contributions

• We propose novel filtering methods to construct a
  precise VNDB in WiFi Networks
• RSS filtering
• Grid-based filtering
• Experiments are conducted on a real testbed
• We compare the localization results using a RNDB
  and VNDBs

9
VNDB Construction

• Estimate each APs position
• AP position calculation
• received-signal-strengths (RSSs) for weight
• positions where the signals are captured
• AP map (i.e., VNDB) construction

10
Filtering inaccurate data (1/2)

• Often, the captured RSS data can be inaccurate
• Due to interference and shadowing
• To filter out the inaccurate data is needed
• RSS filtering
• Grid-based filtering
• RSS filtering
• Use only top 1 rank of
• strong RSS information
• Simple
• But, sampling bias problem

11
Filtering inaccurate data (2/2)

• Grid-based filtering
• Divide the area into number of grids
• Use the average of strong RSS data for each grid
• Exclude the grid from the calculation
• if the number of data in the grid lt 2

12
Testbed for Experiments

• Testbed installation
• 3 x 3 APs at distance of 200m in a parking place
• War-driving to collect 802.11 beacons from the 9
  APs
• NetStumbler v0.4 with GPS

13
VNDB Accuracy Experiments

• Estimation error for APs positions
• No filtering
• RSS filtering
• Grid-based filtering

14
Localization Experiments

• Localization test in specific points using an
  RNDB and VNDBs
• RNDB
• Real AP positions using GPS
• VNDBs with 12m / 64m / 137m / 178m error

15
Conclusion

• Propose filtering processes for accurate VNDB
• RSS filtering and Grid-based filtering
• Localization results with an accurate VNDB are
  comparable to those with an RNDB
• Future Work
• Explore a stochastic approach to construct a
  precise VNDB
• Carry out more experiments on urban areas

16
Questions Comments
17
Backup

• Classification of localization techniques
• GPS-based localization
• Accurate, but not available in indoor
• Need additional hardware costs
• Network-based localization
• Available both in indoor and outdoor
• Additional hardware is not needed
• Especially, WiFi network-based localization
  technique is viable
• Highly Accurate
• due to dense AP deployment in urban
• Inexpensive