Accuracy Characterization for Metropolitan-scale Wi-Fi Localization Ying Wang, Xia Li Ying Wang, Xia Li.

Accuracy Characterization for Accuracy Characterization for Metropolitan-scale Wi-Fi LocalizationMetropolitan-scale Wi-Fi Localization

Ying Wang, Xia LiYing Wang, Xia Li

OutlineOutline

Introduction

Methodology

Results

Evaluation

Summary

Introduction-Introduction-What does the paper doWhat does the paper do

Outdoor Location mechanism based on Wi-Fi

Explore the question of how accurately a user's device can estimate its location using existing hardware and infrastructure and with minimal calibration overhead

slide3

Introduction-Introduction-Why We Need LocationWhy We Need Location

Context-aware applications are prevalent– Maps– Location-enhanced content– Social applications– Emergency services (E911)

A key enabler: location systems– Must have high coverage

Work wherever we take the devices

– Low calibration overhead Scale with the coverage

– Low cost Commodity devices

Introduction-Introduction-Why not just use GPS?Why not just use GPS?

High coverage and accuracy (<10m)

But, does not work indoors or in urban canyons

GPS devices are not nearly as prevalent as Wi-Fi

Introduction-Introduction-Why Wi-FiWhy Wi-Fi

Wi-Fi is everywhere now– No new infrastructure– Low cost– APs broadcast beacons– “War drivers” already build AP

maps Calibrated using GPS Constantly updated

Position using Wi-Fi– Indoor Wi-Fi positioning gives 2-

3m accuracy– But requires high calibration

overhead: 10+ hours per building

Manhattan (Courtesy of Wigle.net)

MethodologyMethodology

1. Training phase (war driving)

Position 1Position 2 Position 3

GPS

Wifi card

(x1, y1)(x3, y3)

(x2, y2)

• A GPS coordinate• List of Access Points

MethodologyMethodology

2. Positioning phase

(x1, y1)

(x3, y3)

(x2, y2)

Position 1

Position 2

Position 3

• Use radio map to position the user

(x’, y’)

A

B C

MethodologyMethodology

Problem: How to make position estimation?

(x’, y’)

(x3, y3)

Answer: By using positioning algorithms

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

1.Centroid Algorithm• Basic Centroid

• Weighted Centroid

2. Fingerprinting Algorithm• Radar Fingerprinting

• Ranking Fingerprinting

3. Particle Filters

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

1. Centroid Algorithm

Basic Centroid

AP1(x1,y1)

AP3(x3,y3)

AP2(x2,y2)

3'

3'

321

321

yyyy

xxxx

(x’, y’)

Estimated

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

1. Centroid Algorithm

Weighted CentroidAP1 (x1,y1)

AP3 (x3,y3)

AP2 (x2,y2)

3'

3'

332211

332211

ywywywy

xwxwxwx

ss1ss2

ss3

(x’, y’)

132

132

www

ssssss

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

2. Fingerprinting Algorithm

What is Fingerprinting?

(x1, y1)

ss

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

2. Fingerprinting Algorithm

Radar Fingerprinting

A

C

B ssA ssB

ssC

ss’A ss’B

ss’C 222 )'()'(' CCBBAA SSSSSSSSSSSS ）（

choose “4” nearest GPS coordinates

GPS coordinate

Access Points

New user

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

2. Fingerprinting Algorithm

Ranking Fingerprinting

All hardware will not give same signal strength Instead of comparing signal strength directly, this method compares the rank of signal strength

is spearman coefficient. Higher -> more similar rankings

     SS = (-20, -90, -40)    R = (1,3,2)

sr sr

222 )'()'(' CCBBAA SSSSSSSSSSSS ）（

Methodology-Methodology-Positioning AlgorithmPositioning Algorithm

3. Particle Filters

Key point of Particle Filter: Fusion

2211 estimationestimationfinal pwpwp

Sensor Model Motion Model

Note: The actual fusion calculation is more complicated, not this linear equation

ResultsResults-AP-AP DensityDensity

Downtown(Seattle)

Urban Residential(Ravenna)

Suburban(Kirkland)

Results-Results-TableTable

Median error in meters for all of algorithms across the three areas

Results-Results-HistogramHistogram

0

10

20

30

40

50

60

70

Downtown UrbanResidential

Suburban

Me

dia

n E

rro

r (m

ete

rs) Centroid (Basic)

Fingerprint (Radar)

Fingerprint (Rank)

Particle Filter

• Algorithms matter less (except rank)• AP density (horizontal/vertical) matters

EvaluationEvaluation

Choice of algorithms– Naïve, Fingerprint, Particle Filter

Environmental Factors– AP density: do more APs help?

– AP churn: does AP turnover hurt?

– GPS noise: what if GPS is inaccurate?

– Scanning rate?

Effect of APs per scanEffect of APs per scan

• More APs/scan lower median error• Rank does not work with 1 AP/scan

Effects of AP TurnoversEffects of AP Turnovers

0

20

40

60

80

100

0% 20% 40% 60% 80% 100%AP Turnovers

Med

ian

erro

r (m

eter

s)

centroid

particle filter

radar

rank

• Minimal effect on accuracy even with 60% AP turnover

Effects of GPS noiseEffects of GPS noise

• Particle filter & Centroid are insensitive to GPS noise

Scanning densityScanning density

• 1 scan per 10 meters is good == 25 mph driving speed at 1 scan/sec• More war-drives do not help

SummarySummary

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) 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

Q & AQ & A

Any Questions?

*The slides were edited based on the original ppt from Yu-Chung Cheng