1. Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Yu-Chung Cheng (UCSD, Intel Research) Yatin Chawathe (Intel Research) Anthony LaMarca (Intel Research) John Krumm (Microsoft Research) Presented by Harsh Poddar Slides adopted from presentation by Cheng Yu, UCSD

2. slide2 Motivation Location-aware mobile applications – Maps – Location-enhanced content – Social applications – Emergency services (E911) PeopleNet Micro-Blog

3. slide3 What is needed? Location Systems

4. slide4 Location System Characteristics High Coverage Low calibration overhead Low cost

5. slide5 GPS High coverage Accurate

6. slide6 GPS Does not work indoors Any place with obstructions Not very prevalent

7. slide7

8. slide8 Riding the Wi-Fi wave Wi-Fi is everywhere now – No new infrastructure – Low cost – APs broadcast beacons Manhattan (Courtesy of Wigle.net)

9. slide9 Intuition “War-Drivers” already build AP Maps – Calibrated using GPS Position using Wi-Fi – Indoor Wi-Fi positioning gives 2-3m accuracy – But requires high calibration overhead: 10+ hours per building What if we use war-driving maps for positioning?

10. slide10 Methodology Training phase – Collect AP beacons by “war driving” with Wi-Fi card + GPS – Each scan records A GPS coordinate List of Access Points – Covers one neighborhood in 1 hr (~1 km 2 ) – Build radio map from AP traces Positioning phase – Use radio map to position the user – Compare the estimated position w/ GPS

11. slide11 Downtown vs. Urban Residential vs. Suburban Downtown (Seattle) Urban Residential (Ravenna) Suburban (Kirkland)

12. slide12 Evaluation Choice of algorithms – Naïve, Fingerprint, Particle Filter Environmental Factors – AP density: do more APs help? – #APs/scan? – AP churn: does AP turnover hurt? – GPS noise: what if GPS is inaccurate? Datasets – Scanning rate?

13. slide13 3 Different Algorithms Centroid Fingerprinting Particle Filters

14. slide14 Centroid Radio Map contains one location for each AP Places user at center of all heard APs – Average of the positions of each heard AP from radio map

15. slide15 Fingerprinting User hears APs with some signal strength signature Match closest 4 signatures in radio map using Euclidian distance formula Radio Map contains raw war-driving data RADAR: compare using absolute signal strengths [Bahl]

16. slide16 Alternative: Relative strength RANK: compare relative ranking of signal strengths [Krumm 03] Match signatures using the Spearman rank-order correlation coefficient

17. slide17 Particle Filter Probabilistic approximation algorithm for Bayes filter Radio Map: Tables containing response rates and signal strengths for every 5m increment in distance for each AP For each Wi-fi scan, lookup table and approximate position using probabilities

18. slide18 Baseline Results Algorithms matter less (except rank) AP density (horizontal/vertical) matters

19. slide19 Effect of APs per scan More APs/scan  lower median error Rank does not work with 1 AP/scan

20. slide20 Effects of AP Turnovers Minimal effect on accuracy even with 60% AP turnover

21. slide21 Effects of GPS noise Particle filter & Centroid are insensitive to GPS noise

22. slide22 Scanning density 1 scan per 10 meters is good == 25 mph driving speed at 1 scan/sec More war-drives do not help

23. slide23 Summary Wi-Fi-based location with low calibration overhead – 1 city neighborhood in 1 hour Positioning accuracy depends mostly on AP density – Urban 13~20m, Suburban ~40m – Dense AP records get better accuracy – In urban area, simple (Centroid) algo. yields same accuracy as other complex ones AP turnovers & low training data density do not degrade accuracy significantly – Low calibration overhead Noise in GPS only affects fingerprint algorithms

24. slide24 Criticism Some APs might not respond to probe packets sent out in active scanning mode; or even send out beacon packets signaling their presence All the algorithms except Rank use absolute signal strengths observed for location estimation Measured signal strengths may vary from chipset to chipset Poor performance of Rank How about using cell phones (GSM) as a location service

25. slide25 Questions? Comments? Concerns?