The Beginners Guide To Playing Guitar Chords

Introduction

- Motivation

Crowdsourcing forDevice Diversity

- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup

- Varying number ofAPs

- Varying noise

- Varying number ofdevices

ExperimentalValidation

- Measurement Setup

Conclusions

- Concluding Remarks

Crowdsourced Indoor Localization for DiverseDevices through Radiomap Fusion

Christos Laoudias∗, Demetrios Zeinalipour-Yazti† andChristos Panayiotou∗

∗KIOS Research Center for Intelligent Systems and Networks, University of Cyprus†Department of Computer Science, University of Cyprus

ΕΝΤΥΠΟ ΥΠΟΒΟΛΗΣ ΕΝΔΙΑΜΕΣΗΣ ΕΚΘΕΣΗΣ ΠΡΟΟΔΟΥ ΕΡΕΥΝΗΤΙΚΟΥ ΕΡΓΟΥ ΤΗΣ ΔΕΣΜΗΣ

2009-2010

INTERIM PROGRESS REPORT FORM FOR RESEARCH PROJECTS FUNDED BY DESMI 2009-

2010

ΤΙΤΛΟΣ ΕΡΓΟΥ

PROJECT TITLE Fault tolerant detection, localization and tracking of multiple events using Wireless Sensor Networks (WSN)

ΑΝΑΔΟΧΟΣ ΦΟΡΕΑΣ

HOST ORGANISATION University of Cyprus

ΣΥΝΤΟΝΙΣΤΗΣ ΕΡΓΟΥ

PROJECT COORDINATOR Christos Panayiotou

1

Supported by the Cyprus Research Promotion Foundation under Grant TΠE/OPIZO/0609(BE)/06

International Conference on Indoor Positioning and Indoor Navigation, Montbeliard - Belfort, France 28 October 2013

Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Outline

Introduction

Crowdsourcing for Device Diversity

Simulation Results

Experimental Validation

Conclusions


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Motivation of our work

I Traditional RSS radiomap construction

I Laborious: Collectors need to visit several locationsI Time consuming: A large volume of data is requiredI Short-lived: Radiomap becomes obsolete with timeI Expensive: Cost can be prohibitive when the task is

undertaken by trained professionals

I Crowdsourcing comes to the rescue

I Volunteers are collecting location dependent RSSsamples, which they later contribute to the system

I Crowdsourced systems (e.g., Active Campus, Place Lab,Redpin, WiFiSLAM, Mole, Elekspot, FreeLoc)

I Or maybe not?

I Filtering incorrect contributions (aka polluted data)I Handling non-uniform fingerprint distributionI Managing the increasing radiomap sizeI Copying with heterogeneous mobile devices


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions









I Or maybe not?



Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions









I Or maybe not?



Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Crowdsourcing with RSS Fingerprints

Offline (training) phase

I Reference locations {L : ì = (xi , yi ), i = 1, . . . , l}, n APs

I Device D(m) visits {L(m) : ì = (xi , yi ), i = 1, . . . , l (m)},where m = 1, . . . ,M, L(m) ⊆ L and L =

⋃Mm=1 L(m)

I Reference fingerprint r(m)i = [r

(m)i1 , . . . , r

(m)in ]T collected at ì

is used to create the device-specific radiomap R(m) ∈ Z−l (m)×n

I Crowdsourced radiomap R ∈ Z−l×n

rij =1

Mi

Mi∑m=1

r(m)ij , 1 ≤ Mi ≤ M (1)

Online (localization) phase

I Use R and the new fingerprint s = [s1, . . . , sn]T measured atthe unknown location ` by the user carried device D(m′)

I (s) = arg minì d2i , d2

i =∑n

j=1

(rij − sj

)2


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Localization with DIFF Fingerprints

Radio propagation model

RSS [dBm] = A− 10γ log10 d + X , X ∼ N (0, σ2) (2)

DIFF approach1

I Takes the difference between all pairwise AP combinations

I The new fingerprints contain(n2

)= n(n−1)

2 RSS differences

I Crowdsourced radiomap R contains ri = [ri12, . . . , ri(n−1)n]T

where rijk = rij − rik , 1 ≤ j < k ≤ n

I s = [s12, . . . , s(n−1)n]T where sjk = sj − sk , 1 ≤ j < k ≤ n


i =∑n

k=2

∑k−1j=1

(rijk − sjk

)2

I Higher dimensionality leads to increased computations

1F. Dong, et al., A calibration-free localization solution for handling signal strength variance,

in MELT, 2009.


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Localization with SSD Fingerprints

SSD approach2

I Subtracts the RSS value of an anchor AP from the otherRSS values in the original fingerprint

I The new fingerprints contain n − 1 independent RSSdifferences

I Crowdsourced radiomap R contains ri = [ri1, . . . , ri(n−1)]T

where rij = rij − rik , j = 1, . . . , n, j 6= k

I s = [s1, . . . , sn−1]T where sj = sj − sk , j = 1, . . . , n, j 6= k


i =∑n

j=1j 6=k

(rij − sj

)2

I Lower dimensionality leads to higher localization errors

2A. Mahtab Hossain, et al., SSD: a robust RF location fingerprint addressing mobile devices’

heterogeneity, in IEEE Transactions on Mobile Computing, 2013.


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Simulation Setup

AP1

AP2

AP3

AP4

AP5

AP6

AP7

AP8

AP9

AP10

AP11

AP12

AP13

AP14

AP15

AP16

I Radiomap R(1) contains RSS values r(1)ij generated by the

propagation model of (2) with A = −22.7 dBm, γ = 3.3

I Radiomap R(m) contains RSS values such that

r(m)ij = α1mr

(1)ij + β1m, m = 2, . . . ,M,

I All M devices contribute their radiomaps R(m) to get thecrowdsourced RSS radiomap R according to (1), R and R

I User carries D(1) and may reside at any location

I Probability of correct location estimation Pc = Nc

Ns


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Varying number of APs

4 6 8 10 12 14 160

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of access points [n]

Pro

babi

lity

of c

orre

ct lo

catio

n es

timat

ion

[Pc]

RSSDIFFSSDDS

Pc for localizing device D(1) with M = 2 devices and σ = 3 dBm

I DIFF is better than SSD and performs equally well with DS

I RSS usually performs poorly, e.g., Pc = 0.45 for n = 6 APs

I For a large number of APs, e.g., n > 11, RSS looks fine

I For RSS, there are peaks in the Pc curve at n ∈ {4, 8, 12, 16}because the APs are evenly distributed around the area


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Varying noise standard deviation

2 4 6 8 10 12 140

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Standard deviation of noise [σ]

Pro

babi

lity

of c

orre

ct lo

catio

n es

timat

ion

[Pc]

RSSDIFFSSDDS

Pc for localizing device D(1) with M = 2 devices and n = 6 APs

I Under low noise conditions (σ = 1, 2 dBm), the performanceof SSD is similar with DIFF

I When σ ≥ 3 dBm, Pc is decreased by 5%–10% for SSD

I DIFF attains the same level of performance with DS

I For RSS, Pc < 0.5 even under low noise conditionsInternational Conference on Indoor Positioning and Indoor Navigation, Montbeliard - Belfort, France 28 October 2013

Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Varying number of crowdsourcing devices

1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of devices [M]

Pro

babi

lity

of c

orre

ct lo

catio

n es

timat

ion

[Pc]

RSSDIFFSSDDS

Pc for localizing device D(1) with σ = 3 dBm and n = 6 APs

I Pc decays linearly for the RSS approach

I DIFF and SSD approaches are extremely robust and theirperformance is not affected as more devices contribute data

I DIFF performs better than SSD and is very close to DSInternational Conference on Indoor Positioning and Indoor Navigation, Montbeliard - Belfort, France 28 October 2013

Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Measurement Setup

I Experimental data collected at the KIOS Research Center3

I RSS samples collected with 5 devices (HP iPAQ PDA,Asus eeePC laptop, HTC Flyer Android tablet, HTCDesire and Samsung Nexus S Android smartphones)

I 2100 location-tagged fingerprints for each devicecollected at 105 reference locations

I 960 location-tagged fingerprints for each devicecollected at 96 test locations

I Performance evaluation

I Used the reference data to build device-specificradiomaps and crowdsourced radiomaps with differentdevice combinations

I Used the test data to evaluate various crowdsourcingapproaches in terms of the localization error

I RSS, DIFF and SSD approaches for crowdsourcingcompared with DS (device-specific) RSS radiomap

3The KIOS dataset is available to download at http://goo.gl/u7IoG


http://goo.gl/u7IoG

Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Two-device crowdsourced radiomaps

RSS DIFF SSD DS0

1

2

3

4

5

6

7

8

Loca

lizat

ion

Err

or [m

]

RSS DIFF SSD DS0

1

2

3

4

5

6

7

8

9

Loca

lizat

ion

Err

or [m

]

Localization of the iPAQ (left) and Desire (right) devices

I Two contributing devices (iPAQ, Nexus) that fully cover thelocalization area for crowdsourcing the radiomap

I Differential approaches reduce error that is comparable to DS

I For iPAQ, the median error is 3.4 m for RSS against 2 m forDIFF and SSD (75th percentile drops from 5.2 m to 3 m)

I DIFF approach filters out high errors more effectively


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions


Crowdsourcing with multiple devices

1 2 3 4 50

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Number of Crowdsourcing Devices

Med

ian

Loca

lizat

ion

Err

or [m

]

RSSDIFFSSDDS

1 2 3 4 50

0.5

1

1.5

2

2.5

3

3.5

Number of Crowdsourcing Devices

Med

ian

Loca

lizat

ion

Err

or [m

]

RSSDIFFSSDDS

Localization of the iPAQ with fully overlapping radiomaps (left) and theeeePC with non-overlapping radiomaps (right)

I RSS performs poorly, e.g., for 5 devices the median error is 4.3 mcompared to 1.8 m for DIFF and 2.3 m for SSD

I For DIFF and SSD the localization error does not varysignificantly as suggested by the simulations

I DIFF outperforms SSD for any number of devices


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions

- ConcludingRemarks

Concluding Remarks

I Notes

I Crowdsourcing stands as the only viable solution forbuilding the radiomap considering effort, time and cost

I Our community has not appreciated its potential (0papers in IPIN’10-11, 1 in IPIN’12 and 2 in IPIN’13)

I Our Contributions

I Evaluated DIFF and SSD methods for creating the RSSdifferences from the original RSS fingerprints

I Simulation and experimental findings indicate thatdifferential fingerprinting is a promising solution

I DIFF performs better than SSD at the expense ofhigher computational complexity

I Future Work

I Investigate other issues related to crowdsourcing, e.g.polluted data, non-uniform fingerprint distribution andthe fast growing radiomap size


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions

- ConcludingRemarks

Thank you for your attention

ContactChristos LaoudiasKIOS Research Center for Intelligent Systems and NetworksDepartment of Electrical & Computer EngineeringUniversity of CyprusEmail: [email protected]


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions

- ConcludingRemarks

Extra Slides


Introduction

- Motivation


- RSS Fingerprints

- DIFF Fingerprints

- SSD Fingerprints

Simulation Results

- Simulation Setup


- Varying noise



- Measurement Setup

Conclusions

- ConcludingRemarks

Linear relation between RSS values

−100 −80 −60 −40 −20−100

−90

−80

−70

−60

−50

−40

−30

−20

α=0.9155β=13.3612

Mea

n R

SS

from

HP

iPA

Q [d

Bm

]

Mean RSS from HTC Flyer [dBm]

RSS pairsLinear Fit

−100 −80 −60 −40 −20−100

−90

−80

−70

−60

−50

−40

−30

−20

α=0.9439β=1.321

Mea

n R

SS

from

Asu

s ee

ePC

[dB

m]

Mean RSS from HTC Desire [dBm]

RSS pairsLinear Fit

I Several studies report a linear relation between the RSSvalues measured by heterogeneous devices

I r(m2)ij = αm1m2 r

(m1)ij + βm1m2 , m1,m2 ∈ {1, . . . ,M}, where

(αm1m2 , βm1m2 ) are the coefficients between D(m1) and D(m2)

I Direct fusion of the different RSS radiomaps using (1) maydegrade the quality of the crowdsourced radiomap


Documents

The Beginners Guide To Playing Guitar Chords