Research Article Crowdsource Based Indoor Localization by...

Research ArticleCrowdsource Based Indoor Localization byUncalibrated Heterogeneous Wi-Fi Devices

Wooseong Kim1 Sungwon Yang2 Mario Gerla2 and Eun-Kyu Lee3

1Department of Computer Engineering Gachon University 1342 Seongnam-ro Gyeonggi 461701 Republic of Korea2Department of Computer Science University of California Los Angeles Los Angeles CA 90095 USA3Department of Information and Telecommunication Engineering Incheon National University 119 Academy-roYeonsu-gu Incheon 406772 Republic of Korea

Correspondence should be addressed to Eun-Kyu Lee ekleecsuclaedu

Received 15 March 2016 Accepted 4 May 2016

Academic Editor Sergio F Ochoa

Copyright copy 2016 Wooseong Kim et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Many indoor localization techniques that rely on received signals fromWi-Fi access points have been explored in the last decadeRecently crowdsourced Wi-Fi fingerprint attracts much attention which leads to a self-organized localization system avoidingpainful survey efforts However this participatory approach introduces new challenges with no previously proposed techniquessuch as heterogeneous devices short measurement time and multiple values for a single position This paper proposes an efficientlocalization method combating the three major technical issues in the crowdsourcing based systems We evaluate our indoorpositioning method using 5 places with different radio environment and 8 different mobile phones The experimental results showthat the proposed approach provides consistent localization accuracy and outperforms existing localization algorithms

1 Introduction

Navigation ability of smart phones and tablets at indoorenvironment becomes a challenge as recent mobile LocationBased Services (LBSs) [1ndash3] require more accurate and seam-less positioning at both indoor and outdoor environmentOutdoor localization had been provided reasonably by globalsatellite positioning systems using Global Positioning System(GPS) GLObal NAvigation Satellite System (GLONASS)Galileo and so forthOn contrary the indoor localization stillcontinues to a challenge even though it has been exploitedduring last several years Several techniques for the indoorlocalization have been proposed [4ndash6] In a survey of indoorlocalization techniques [7] many techniques are comparedsuch as GPS RFID WLAN and Bluetooth in terms of accu-racy complexity and scalability Some of previous works usesa specially designed hardware like a beacon installed on wallsor ceilings such as RFID ultrasound and infrared technolo-gies [8ndash10] Other systems use a combination of differentsensors to increase accuracy [11 12] Those which use thehardware of short range communication technologies are

costly to deploy on a large area even though those systemsprovide fine-grained accuracy in indoor localization

In contrast Wi-Fi based positioning systems are verycommon and easily achieved as shown that Internet mapservices of Google Apple Microsoft and so forth alreadyuse received signals frommany deployed access points (APs)for localization Recently Wi-Fi is becoming more popularand ubiquitous with countless Wi-Fi-enabled smart devicesAccordingly indoor localization using theWi-Fi APs receivesmuch attention frommobile computing research area whichhas the advantage of avoiding the cost of specialized hardwaredeployment

Typically Wi-Fi based localization estimates the loca-tion based on observed and stored ldquofingerprintsrdquo which arecomposed of a MAC address and corresponding ReceivedSignal Strength (RSS) value from a Wi-Fi AP [10 13] Whileprevious researches mainly dealt with localization algorithmsto improve accuracy based on the fingerprint database recentstudies focus on building the database with less effortldquoCrowdsourcerdquo Based Indoor Localization (CIL) [14ndash19] byautonomous users is one of cost efficient approaches which

Hindawi Publishing CorporationMobile Information SystemsVolume 2016 Article ID 4916563 18 pageshttpdxdoiorg10115520164916563

2 Mobile Information Systems

reduces the map building and maintenance expense com-pared to an expert surveyor based approach The CIL isdriven by normal mobile users who participate in the Wi-Fi AP survey as a contributor a user or both Also thismultiple-surveyor-multiple-usermodel has the advantages offast radio map building and quick update

Figure 1 shows an example of the CIL which consists oftwo phases ldquotraining phaserdquo (or ldquooffline phaserdquo) and ldquolocal-ization phaserdquo (or ldquoonline phaserdquo) In the training phasefingerprint data of locations have been captured to establisha radio map After then mobile phones inquire their locationwith measured fingerprint data to a remote server that holdsthe radio map in the localization phase In the figure the fiveusers (ie A B C D and E) are untrained normal peoplewho are carrying different types of mobile phones Users AD and E are surveyors for Room 101 Room 102 and Room103 whomeasure RSS ofWi-Fi APs and upload the measureddata with corresponding AP information to a local DB serverThe uploaded data is processed on the server and convertedto fingerprint dataWhen user B inquires the current locationinformation the server returns to Room 101 if user Brsquos RSSmeasurement matches one from the user A

However the CIL also introduces a new set of challengesFirst the CIL has to extract accurate fingerprint values fromshort measurement time for Received Signal Strength (RSS)because the mobile users as voluntary surveyors probablyprovide short-term RSS measurements Second the CILhas to support calibration-free indoor localization acrossdifferent devices Typically users as surveyors carry varioustypes of devices in terms of hardware and software for exam-ple Wi-Fi chipsets antenna pattern and operating systemsThe RSS measurement results in different values across theheterogeneous devices even at exactly the same positions

This crowdsourcing approach has been considered inmany recent researches [14ndash19] However many of themaddress those challenges in the CIL In [20] about survey-free localization several researches [21ndash23] dealt with het-erogeneous devices for CIL using difference of RSS samplesfrom APs instead of absolute RSS values to combat measure-ment variance from heterogeneous devices Previously weproposed ldquoFreelocrdquo [24] that uses a list of APs ordered by RSSfor the fingerprint instead of the RSS itself Thus the Freelocconsistently results in robust and reliable outcomes resolvingthose challenges In this paper we compare the Freeloc withexisting practical indoor localization algorithms for the CILin terms of accuracy In addition we investigate how todecide a gap between sample RSSs for difference localizationenvironment and conductmeasurement study for the Freelocin variance of wireless channel because of interference fromdense Wi-Fi APs deep fading from many pedestrians andpath loss from short distance

The remainder of this paper is organized as followswe introduce background knowledge with related works onindoor localization schemes in Section 2 Section 3 intro-duces existing localization algorithms for CIL Section 4describes our system and algorithm together with the threemajor challenges covered in the CIL We show the perfor-mance evaluation and comparison with existing localiza-tion algorithms from Sections 5 to 7 In Section 5 Freeloc

AP 1

AP 4

Room103

Room102

Room101

User E

User D

User A

User CUser B

AP 5

AP 6

AP 7

Remotelocalization sever Fingerprint

database

I am near ldquo103rdquo

I am in ldquo101rdquo


AP 3

AP 2

Where am I

Where am I

Figure 1 Crowdsource Based Indoor Localization (CIL) approachwhich collects RSS values of Wi-Fi APs measured by multiple usersusers AndashE at different positions of Rooms 101 102 and 103 togetherwith corresponding AP information which is finally converted tofingerprint data of each location

performance is compared according to various localizationenvironmentWe compare Freeloc with other CIL algorithmsin Section 6 In Section 7 Freeloc performance is investigatedwith varying delta valuesWe conclude the paper in Section 8

2 Background

21 Indoor Localization Techniques Many indoor localiza-tion techniques have been proposed over the past decadeSome indoor localization techniques require special hard-ware to determine the devices location like active badge [9]and active bat [25] that are attached on ceilings and walls ofbuildings and transmits out infrared and ultrasound pulsesfor proximity estimationThe cricket and cricket compass [826] used a combination of RF and ultrasound technologiesTechniques using active RFID were also proposed [27 28]

Recently Wi-Fi based indoor localization receives muchattention asWi-Fi APs become used for outdoor localizationLike the outdoor localization in cellular networks triangu-lation using time difference of signal arrival from neighbourAPs can be considered [13 29] This time of arrival (TOA)based approach could suffer from timely varying signal pathat complex indoor environment Instead Received SignalStrength (RSS) can be used RADAR [10] adopts this RF signalintensity first for the purpose of indoor localization Nowmany researchers concentrate on techniques that use Wi-Fi

Mobile Information Systems 3

RSS data Studies like [30] analyzed the properties of ReceivedSignal Strength values reported by wireless network interfacecards References [5 31] show that the Wi-Fi fingerprint (iea basic service set identification (BSSID) and its RSS value)methods can fulfil accuracy required for many indoor LBSapplications

22 Crowdsourced Indoor Localization Survey-free localiza-tion [16ndash19] orCrowdsource Based Indoor Localization (CIL)[14 15] has rigorously been explored in order to reduce effortand cost involved in expert surveyor based system as smarthand-held devices are deployed widely The CIL is a self-organized participatory system where normal mobile userscontribute measurement of Wi-Fi fingerprints at variouslocations for localization Reference [15] carried out experi-ments in indoor environments and have discussed encourag-ing results Teller et al in [16] have designed and deployed anorganic location system and achieved position accuracy thatis comparable to the accuracy achieved by a survey drivensystem A crowdsourcing based approach by Ledlie et almodels the world as a tree of hierarchical namespaces andprovides an algorithm that explicitly accounts for temporalvariations in signal space [19] Recently accelerator assisteddead-reckoning techniques are adopted for enhancing accu-racy of crowdsourced indoor localization [11 12]

A main issue with the CIL is the heterogeneity ofhardware devices usually a variety of mobile phones tocollect Wi-Fi fingerprints during the training phase Thisleads to variation in the values of observed signal strengthmeasurements due to the different chipsets present on differ-ent devices Park et al explore this issue in [17] and comparevarious methods used to mitigate this problem The mainparameters that are used are signal strength values and accesspoint detectionTheKernel function based estimation whichpredicts user location using a naive Bayes classifier has beenproposed to obtain better results than linear transformationbased approaches [32] Tsui et al proposed an unsupervisedlearning method that automatically tries to solve hardwarevariance problem in Wi-Fi localization [33] Wu et al pro-posed site survey-free indoor localization with off-the-shelfdevices which needs complex conversion from virtual tophysical space based on measurement [21] References [22ndash24] show that robust fingerprints can be derived using signalstrength difference (SSD) instead of absolute RSS values inthe heterogeneous mobile devices for measurement

3 Localization Algorithm for CIL

Although several localization algorithms were proposed forCIL most of them use log-distance or similarity of RSSs[20] We introduce several practical approaches that are lesscomplicated and easy to apply for the fingerprint basedindoor localization to compare performance with Freeloc

First Euclidean distance of RSSs of 119896-nearest neighbours(119896-NN) is calculated to find a location with minimum valueof summation of the distance values Here we considerunweighted 119896-NN in which distance values are equally addedat the summation Second similarity based on the Tanimotocoefficient is considered Third we evaluate the119873-gram that

compares BSSID subsequences between two fingerprints tocalculate probability of appearance of the same contiguoussubsequence With these four algorithms we apply the sameexperiment environment that is used for ldquoFreelocrdquo evaluationfor the comparison

(i) k-nearest-neighbour (119896-NN) is a popular localizationmethod to analyze RSS pattern from 119896 neighbour APsusing some recognition techniques For instance itcan simply calculate Euclidean distance of two finger-printsrsquo RSS values that are measured and stored in aremote server Like this 119896-NN estimates a locationwith the distance values of most frequently appeared119896 neighbour APs

arg min119891119871119894isin119865

radicsum

119894=1

11989610038171003817100381710038171003817119891119894 minus 119891119871 119894

10038171003817100381710038171003817 (1)

where 119871 119894 is locations 1198710 1198711 119871119899 and 119891119894 is ameasured fingerprint and 119891119871 119894 is a stored fingerprintof location 119894 That is to say 119896-NN selects a locationthat has minimum sum of distance of most strong 119896fingerprints between the measured and stored values

(ii) Two fingerprint similarities can be evaluated by Tan-imoto coefficient which is widely used for estimatingsimilarity ofWi-Fi fingerprints Reference [34] showsthat place detection can be achieved by similarity ofradio environment using Tanimoto coefficient As inbelow equation location 119871 119894 with maximum Tanimotocoefficient can be chosen

arg max119891119871119894isin119865

119879 (119891119894 119891119871 119894) =

119891119894 sdot 119891119871 119894

100381710038171003817100381711989111989410038171003817100381710038172+10038171003817100381710038171003817119891119871 119894

10038171003817100381710038171003817

2minus 119891119894 sdot 119891119871 119894

(2)

(iii) 119873-gram algorithm is a different approach comparedto above two approaches which calculates likelihoodof two subsequences of the measured and stored fin-gerprints [35] 119873-gram only concerns strength orderof the fingerprints while 119896-NN and Tanimoto con-siders signal strength itself The order of fingerprintstrength does not change probably even if the signalstrength of the fingerprints changes dynamically Forthis 119873-gram sorts the fingerprints in descendingorder of the RSS and compares subsequences ofthe fingerprints For instance subsequence 119873-gram(119891119904) = 119891119904 119891119904+1 119891119904+119899minus1 is a fingerprint sequencefrom a 119904th fingerprint to a 119894 + (119899 minus 1)th fingerprintwith size of 119899 Here we can find a location 119871 119894 thatmaximizes a following equation

arg max119891119871119894isin119865

prod

119904

119875 (119873-gram (119891119904) | 119871 119894) 119875 (119871 119894) (3)

That is to say the algorithm selects a location withmaximum probability that a set of the subsequences119891119904 appears in the location 119871 119894 119873-gram motivates usto improve it for crowdsourced localization in whichmeasured fingerprints are varying with locations anddevices as described in a following subsection


4 Freeloc Calibration-FreeIndoor Localization

In this section we briefly describe our localization algorithmFreeloc with a usage scenario in Figure 1 Freeloc solves fol-lowing three critical challenges in fingerprint derivationbased on the RSSs captured by untrained mobile users

(i) RSS measurement for short duration it is well knownthat multipath fading in an indoor environmentcauses RSS to fluctuate over time even if the receiveris absolutely fixed [36] One simple method to reduceRSS variance is to record the RSS data for a longtime As the number of RSS samples increases it iseasier to identify one single fingerprint value whichensures construction of a more robust and accurateradio map However it is almost impossible to havenormal users collect RSS data for such a long time inthe crowdsourcing based model We believe that theRSS survey time at each location will not exceed oneminute assuming all the surveyors are usual mobileusers Therefore a technique that extracts accuratefingerprint data from short-time measurement isnecessary

(ii) Device heterogeneity in CIL system it is inevitablethat diverse devices get involved in establishing theradio map database Since there is no expert surveyorwho uses specially designed hardware it is highlylikely that the RSS data gathered from each uservaries even though it is collected at the exact sameposition Different chipsets and antenna designs ofthe Wi-Fi devices cause varying RSS records pereach device and make it difficult to calibrate themThis device heterogeneity is another key problem forcrowdsourced localization

(iii) Multiple measurements for one location anotherproblem is that more than one user can upload onersquosown fingerprint data with the same location labelbut it is obtained at different measurement pointsThis results in multiple fingerprint data sets for onelocation which leads to inefficientmemory usage andmore time to estimate the current location Noneof the previous studies take this into account eventhough this is also a principal cause of localizationaccuracy degradation

41 Fingerprint Value Extraction This section presents how aRSS value of each fingerprint can be extracted at a particularlocation with variation in response rate and RSS values Incontrast RSS can bemeasured for a long time (eg frommorethan one hour to a month) to get average value in the expertsurveyor system [37] CIL has short measurement durationto extract robust fingerprint We propose a practical andresilient RSS abstraction method for the fingerprint based onthe experimental results we conducted

(i) AP response rate in our experiments we observedstrong correlation between AP response rate and RSSLedlie et al [19] thought that the correlation between

the response rate and RSS is rather weak which ledto using the response rate as fingerprint informationwith RSS However our results show that APs withRSS values of greater than minus70 dBm provide over90 of response rate and APs with RSS betweenminus70 dBm and minus85 dBm provide over 50 of responserate APs with RSS of less than minus90 dBm present verypoor response rateTherefore we decide to give lowerweights to weak RSS values which reflects the lowerAP response rate naturally for fingerprint

(ii) RSS variance RSS of a particular AP is varying overtime due to shadowing andmultipath fading and RSSfluctuation could be higher in indoor environmentrather than outdoor which degrades the localizationaccuracy For CIL that requires short measurementduration probabilistic methods based on Gaussiandistribution were proposed in several studies pro-posed [16 19 32] to obtain the average value whichare however feasible only in ideal environmentInstead we propose a simple yet effective methodbased on our experiment results that show the mostfrequently capturedRSS in short duration is very closeto the most RSS in the long-duration measurementcase In other words most frequent RSSs can beobtained regardless of the duration of measurementtime The proposed method extracts fingerprint RSSvalues by ignoring RSS records that are far awayfrom themost frequent RSS and giving the maximumweight to the RSSs with peak value as shown in (4)which provides tolerance to the RSS variation overtime119881fp is the fingerprint value for anAP andRSSpeakis the RSS value of highest frequency during themeasurementThewidth of the range is set by120596119871119879 and120596119877119879

119881fp

=sum120596119871119879119899=1 (RSSpeak minus 119899) + RSSpeak + sum

120596119877119879119898=1 (RSSpeak + 119898)

120596119871119879 + 120596119877119879 + 1

(4)

42 Online Localization Algorithm

421 Relative RSS Comparison Most of previous localizationtechniques estimate a location with an absolute RSS valuebased on the radio map built during the offline phase How-ever CIL where surveyors carry heterogeneous hardwaredevices can have different fingerprint data sets even for thesame location because RSS values measured by those devicesare different from each other

In this section we explain a novel localization techniqueFreeloc instead of modifying existing techniques for theheterogeneous device environment that is aligned with ourtarget indoor usage scenarios Thanks to the proliferation ofWi-Fi technology some office or university buildings havemore than 50 APs installed for personal purpose or foroffloading of wireless cellular networks RSS values of thoseAPs are widely distributed (eg minus40 dBm to minus100 dBm)which is good to abstract relative RSS strength among APs


Input fingerprint data FpunknownOutput estimated location Loc(1) score larr 0 scoreMAX larr 0

(2) bssidMAP larr

(3)(4) for each possible FpLoc119909 in the radio map do(5) for each AP119894 in Fpunknown do(6) if AP119894 is found in FpLoc119909 at119898th AP then(7) bssidMAP larr 119871(AP119895120575(119898)) vector where AP119894 = AP119898(8) for each BSSID in AP119895120575(119894) do(9) if BSSID is found in bssidMAP then(10) score larr score + 1(11) end if(12) end for(13) end if(14) end for(15) if score gt scoreMAX then(16) scoreMAX larr score(17) Loc larr Loc119909(18) end if(19) end for(20) return Loc

Algorithm 1 Estimate the location

Table 1 RSS measurement results from each user

AP User A User B User C User DAP1 minus50 minus65 minus65 minus92AP2 minus55 minus68 minus51 minus80AP3 minus67 minus74 minus53 minus67AP4 minus72 minus52 minus66 minus75AP5 minus88 minus91 minus82 minus58AP6 minus90 minus87 minus85 minus54

In contrast to several approaches using RSS difference ofAPs for fingerprints Freeloc generates fingerprints using anAP order instead of RSS value itself as shown in

Fp (Loc119909) = Loc119909 [AP119894 119871 (AP119895120575(119894))]

119894 119895 isin APs of Loc119909 119894 = 119895

RSS (AP119894) minus RSS (AP119895) ge 120575

(5)

Each fingerprint data element Fp of location Loc119909 has alocation label Loc119909 (eg room number) where measurementis taken and a detected APrsquos BSSID (AP119894) that has the 119894thstrongest 119881fp 119871(AP119895120575) is an order of other detected APsrsquoBSSIDs of which RSS is weaker than RSS of the AP119894 bymore than 120575 value Our proposed method adopts the delta(120575) as a marginal value for varying RSS which is differentfrom 119873-gram [35] that considers only similarity betweentwo continuous subsequences of the ordered fingerprint eventhough both schemes focus on relationship between detectedAPs in terms of RSS rather than RSS value itself Thedelta value keeps the relationship consistent under wireless

Table 2 Fingerprint data of each location

(a)

Room 101AP119894 AP119895120575(119894)AP1lowast AP3 AP4 AP5 AP6AP2lowast AP3 AP4 AP5 AP6AP3 AP5 AP6AP4 AP5 AP6AP5

AP6

(b)

Room 102AP119894 AP119895120575(119894)AP6lowast AP1 AP2 AP3 AP4AP5lowast AP1 AP2 AP4AP3 AP1 AP2AP4 AP1AP2 AP1AP1

channel fading and enables ourmethod to dealwith heteroge-neous devices without calibration efforts and alsomultiple Fpelements for the same location from many different devices

For example suppose that usersA andDare the surveyorsand users B and C want to know their positions in Figure 1and the measurement of user AndashD is presented in Table 1the fingerprints of Rooms 101 and 102 can be defined inTable 2 For instance AP1 with minus50 dBm RSS value has 4


(a) 5th floor of Gachon IT building

37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K

(b) 3rd floor of UCLA engineering building

(c) Resident house

FE

I A BG

H C

D

(d) UCLA NRL laboratory

(e) Indoor Coex shopping mall in Seoul

Figure 2 Floorplans for evaluation

other detected APs nearby (AP3ndashAP6 dBm) that show lowersignal strength by more than 10 dBm (ie the 120575 value)

From now on we describe details of Freeloc localizationalgorithm in Algorithm 1 In the aforementioned scenariousers B and C request their positions to a localization serverwith own fingerprint values in Table 2 and then the finger-print values are compared with the stored fingerprint map

that is created by users A and D First AP4 as the strongestAP of user B is searched in Room 101 map AP4 in the maphas other APs AP5 and AP6 that are weaker by the deltaSince AP5 and AP6 are commonly found in both the user Brsquosfingerprint and the location map Room 101 earns two pointsLike this our algorithm finally acquires 8 points for Room101 but only 1 for the 102 For the user C the scores are 9 and


S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)

(a) Desire HDD DSVSVDVDSVS

0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)

(c) Vega Iron 2D DNVNVDVDNVN

0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3

Figure 3 Impact of device heterogeneity experiment at 5th floor of Gachon IT building

D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2

Figure 4 Distance errors of localization at a resident house

2 for Room 101 and 102 respectively Therefore we concludethat both users B and C are in Room 101

In order to achieve fast retrieval (line 4 in Algorithm 1)we utilize an ldquoimportance flagrdquo a one-bit flag in fingerprintdata Table 2 The flag is set if the AP119894 ranks high with strongRSS When the localization server receives such locationrequest it selects fingerprint data with flagged AP119894 in themapwhich matches the flagged AP119894 in the request data

For example in the same aforementioned scenario theflagged AP119894s are AP1 and AP2 for Room 101 and AP6and AP5 for Room 102 if we set the criterion for the flag as

the top two APs The flagged AP119894s are marked with asteriskin Table 2 When users B and C request their positions atleast one flagged APlowast119894 in their fingerprint data is found in thefingerprint data of Room 101 (ie AP1 for user B and AP2for user C) while no matching flagged AP is found in thefingerprint data of Room 102 Therefore only Room 101 isconsidered for the possible location in this exampleHoweverwe can be bothered from high rank APlowast119894 s if adjacent locationshave high probability of receiving strong RSS from the sameAPs If no matches are found we regard the position wherethe measurement is taken as an unlabeled location


Table 3 Measured fingerprint data for localization

(a)

User BAP119894 AP119895120575(119894)AP4lowast AP1 AP2 AP3 AP5 AP6AP1lowast AP5 AP6AP2 AP5 AP6AP3 AP5 AP6AP6

AP5

(b)

User CAP119894 AP119895120575(119894)AP2lowast AP1 AP4 AP5 AP6AP3lowast AP1 AP4 AP5 AP6AP1 AP5 AP6AP4 AP5 AP6AP5

AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)

Figure 5 Localization accuracy comparison between laboratoryand third floor in UCLA building [24]

422 Fingerprint Update In CIL mobile users as surveyorsupload continuously different fingerprints for the same loca-tion with their heterogeneous devices which increases thenumber of RSS data sets for the same location label in themap database and time to compare all sets for localization

Freeloc can provide a simple update procedure thatbuilds a unique fingerprint without any calibration whichjust mergesmultiple fingerprints into a single fingerprint Forexample the new fingerprint data will be established as inTable 4 if B uploads its RSS measurements with the samelocation label as shown in Table 3 and they merge with anexisting Room 101 fingerprint in Table 2 created by user AThe proposed AP119894 119871(AP119895120575(119894)) fingerprint data structure andthe 120575 value increase the similarity amongmultiple fingerprintdata although they aremeasured at slightly different locationswith different devices

Table 4 Merged fingerprint data for Room 101

Room 101 user A + BAP119894 AP119895120575(119894)AP1 AP3 AP4 AP5 AP6AP2 AP3 AP4 AP5 AP6AP3 AP5 AP6AP4 AP1 AP2 AP3 AP5 AP6AP5

AP6

Table 5 Devices used for data collection

Device MFR Wi-Fi chipset AndroidBionic (B) Motorola TI WL1285C 234Galaxy S2 (G) Samsung BCM4330 234Galaxy S3 (S) Samsung M2322007 43Galaxy Note 4 (N) Samsung BCM4358 51Desire HD (D) HTC BCM4329 412Vega Iron 2 (V) Pantech 442Nexus One (NO) HTC BCM4329EKUBG 233Nexus S (NS) Samsung BCM4329GKUBG 236

5 Performance Comparison atDifferent Environment

From this section we evaluate the Freeloc with 8multivendorsmart phones in Table 5 which have differentWi-Fi hardwarechipsets and mobile OSs For evaluation we varied the num-ber and type of devices which participate in the fingerprintmap for each user deviceThe same device was never used forboth generating fingerprinting map and localization

51 Building Floor First 5th floor of the IT building atGachon University and 3rd floor of the UCLA engineeringbuilding as officeschool building environment where longcorridors and surrounded meeting rooms and laboratoriesexist as shown in Figures 2(a) and 2(b) The width of thecorridor is approximately 2m and some APs are visible withline-of-sightTheir floor plan is similar to each other butWi-Fi AP density is different Gachon IT building has more than300 detected APs while the UCLA building captures about50 APs We measured RSSs at 36 different positions at thethird floor of the engineering buildings in UCLA campus and35 positions at the IT building of Gachon University Smalldots in figure indicate the locations where RSSmeasurementswere taken The data collected at these points were usedfor experiments The adjacent points are approximately 6mapart

In the experiment Freeloc localization algorithm isdemonstrated for feasibility of CIL with heterogeneous mul-tiple devices at corridor based office buildings Localizationresults from the heterogeneous devices at the Gachon Uni-versity are shown in Figure 3 First three bars are the caseswhen only one device was used for the fingerprinting Thenext three bars are the cases when the fingerprint maps weregenerated from two different devices and these fingerprint


3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250

(a) Score at laboratory at positions 1ndash3

1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25

(b) Score at 3rd floor at positions 1ndash3

Figure 6 Localization scores at measurement locations of UCLA laboratory and 3rd floor

maps were merged to form one fingerprint map The last bardenotes the case when fingerprint maps have been generatedfrom three different devices and merged

Figure 3 shows that the accuracy did not degrademostly when more than one fingerprint data from differentdevices are merged Particularly HTC Desire HD (D) resultsin higher distance error to any fingerprint created by otherphones The distance error has been observed exceeding 4mwith the Desire HD But the distance error decreases withadditional fingerprints combined by other heterogeneousdevices to less than 1m Herein we can conjecture that moreadditional measurements from other heterogeneous devicescontribute to reduce the distance error Furthermore thedistance error will decrease considerably as the homogeneityof devices increase in the crowdsourced system

Experiment results in Gachon and UCLA buildings [24]show that Freeloc can provide consistent performance amongheterogeneous devices According to our observation inFigure 5 [24] error distance in Gachon IT building is lessthan UCLA engineering building about 1 versus 3m whichimplies that the number of APs and degree of their distribu-tion can affect localization performance in the Freeloc

52 Residential Area Next localization experiment for smallresidential area had been conducted at UCLA laboratoryand a residential house which have many different types offurniture and only a few residential people stay Howeverthe UCLA laboratory is partitioned by thin plywood forrooms and the partitions do not reach to ceiling while theresidential house has rooms separated by thick concretewallsfor example a living room a kitchen and a restroom In theresidential house and UCLA laboratory 29 and 34 positionsare measured respectively Each of the measurement pointsis approximately 15m apart

Figure 4 depicts distance errors of a resident house withthree heterogeneous devices (D S and V) Cross-deviceerror exceeds 15m that is a distance between measurementpositions in case of localization with a single fingerprintHowever the distance error becomes less than the 15mwith two fingerprints given by heterogeneous devices In allcases of the figure the accuracy did not degrade when morethan one fingerprint data from different devices are mergedThis reaffirms that the Freeloc can be a feasible solution forthe CIL

The localization performance at the resident house showsbetter than UCLA laboratory in Figure 5 even though theirsizes are similar We can conclude that the residential housecan obtain unique and stable fingerprints rather than theUCLA laboratory since the rooms of the resident house areseparated by the concrete walls and channel variation causedby residential people is smaller (eg more than 10 students inthe UCLA laboratory and 2 residents in the house)

Figure 5 shows that the average distance error at theUCLA laboratory compared to the third floor of UCLAengineering building The distance error at the UCLA labo-ratory is higher the third floor where the measurement pointinterval is longer than the laboratory (15 versus 6m)

The reason for this observation of degraded accuracy inthe laboratory can be difficulty to have a unique fingerprintcompared to the third floor that hasmore unique fingerprintsformed with thick concrete walls of the building due to pathloss while the laboratory ismade of thin plywoodmaterial forrooms inside

Figure 6 shows score distribution derived fromAlgorithm 1 at the measurement position from 1 to 3where localization error occurs at positions 2 and 3 of thelaboratory and position 3 of the third floor As can be seenin the figure the scores of the laboratory are very similar to


NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions

Figure 7 Analysis of localization performance at small region Distance errors of Bionic user device with a reduced number of APs andmeasurement positions

each other among measurement locations compared to thethird floor which leads to wrong prediction with higherprobability than the third-floor case

According to our analysis on the laboratory fingerprintswe observed that the similarity of the fingerprints is caused bysimilar order of the fingerprintsThat is to say fingerprints inthe laboratory have almost the same sequence of AP BSSIDsdue to plywood walls except several of the strongest APsthat should be highlighted To see impact of measurementdistance and number of APs for a fingerprint twofold isrevisited as inputs for fingerprints as shown in Figure 7

Since the short measurement distance at the small arealike a laboratory may increase localization error probabil-ity compared to the corridor based office buildings we exper-imented our localization algorithm with smaller number ofmeasurement spots With many APs in similar order a long

list of measured APs increases the score value and make thefingerprint undistinguishableThus we tested with a reducednumber of APs for example the strongest 15 20 and 25 APsfrom more than 40 captured APs in the laboratory

In Figure 7(a) 20 APs are most preferable for the finger-print of the laboratory The distance error decreases to 24and 25m from 28 and 32m in NexusS (NS) and Galaxy(G) respectively This error decreases more if the number ofmeasured sites decreases by a half The distance error of theNexusS and Galaxy decreases to 139 and 228m as shown inFigure 7(b) The distance error decreases more with smallermeasurement sites in Figure 7(c) Table 6 shows averagedistance error with a number of APs for the fingerprintsWithout resizing a list of captured APs the average distanceerror was not decreased only by reducing the number ofmeasurement sites (little bit increase in case of All) On


IT buildingLocations

Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n

Figure 8 RSS variation at Gachon IT building and Coex ∙ repre-sents the exceptional cases that are out of 95 confidence interval

Table 6 Average distance error with a reduced number of APs

APs 34 sites 17 sites 8 sites15 258 25 26120 25 201 14525 231 198 16All 277 301 308

contrary the average distance error decreases from 25 to145mwith 20 APs if themeasurement sites are reduced from34 to 8 sites Such reduction of AP sequence can also decreaselocation estimation time that is proportional to the numberof fingerprint data to be compared in addition to give ahighlight to the strong APs for a fingerprint

Localization with fine granularity is still challenge espe-cially in small open space where similar fingerprints arecaptured Accordingly measurement position distance or anumber of APs used for fingerprints should be adjustedaccording to the localization environment especially for thesmall area However we argue that Freeloc is still feasible forhome localization as one of key technologies for Internet ofThings (IoT)

53 Indoor Shopping Center Wi-Fi channels at indoor envi-ronment are varying with moving objects especially peoplewhich leads to variance in measured fingerprints We canconjecture that the fingerprint abstraction based on (4) candeal with this problem However largely distributed RSSs canstill mislead the localization with inconsistent fingerprints

In order to demonstrate this problem we experimentedour localization method at the crowd indoor shopping mallthe Coex one of the biggest indoor shopping centers in southKorea which holds large convention rooms for exhibitionsand conferences many restaurants and stores (Figure 2(e))

The Wi-Fi fingerprint data were collected at 40 positionsfor different Wi-Fi channel environment and each positionis about 6m apart the same as in the school buildings Wemeasured RSSs from more than 200 APs installed by Coexstores and Internet service providers

We used four smart phones Desire HD (D) Vega Iron(V) Galaxy S3 (S) and Note 4 (N) and measured RSSs witheach phone at different time to experience different channelsTo investigate degree of fading effect from pedestrians RSSvariation of two different places Gachon IT building andCoex mall is compared in Figure 8 The RSS variation in theCoexmall is higher than the IT building probably due tomorepedestrians RSS difference is almost about 10 dB in someof APs However the RSS distribution of strong APs is verysimilar in both locations

Figure 9 shows error distance of pairwise experiments ofthose heterogeneous devices at the Coex mall The distanceerror is bigger than the Gachon IT andUCLA building due tovarying channel status even though more APs were capturedfor fingerprints Also measurement by Desire HD leads toincrease distance error However the error distance decreaseswith fingerprints gathered by other three heterogeneousdevices up to less thanmeasurement distance 6mAccordingto this experiment results localization performance for CIL isaffected more by device characteristics that is heterogeneityrather than fading channel status because RSS value abstrac-tion based on multiple measurement samples (eg more orless 10 samples) reduces the fading effect Thus Freeloc canbe robust even in the crowd indoor environment that suffersfrom severe channel fading due to pedestrians

6 Performance Comparison ofLocalization Algorithms

First localization performance is compared when using aBionic device and three fingerprints built by others Figure 10shows CDF of error distance at the laboratory for eachdifferent fingerprint Error distance is slightly different fromthe fingerprints but comparable except 119896-NN Maximumerror distance is however very different from the localizationtechniques Tanimoto approach shows biggest error about 12meters among the four localization schemes since similaritybetween RSS fingerprints is undistinguishable to each otherat the narrow laboratory environment Figure 11 shows CDFof error distance at the third floor 119896-NN and Tanimotoapproaches show considerable error compared to the119873-gramand Freeloc 119873-gram and Freeloc show only about 5-metererror at more than 80 fingerprints error distance while 119896-NN and Tanimoto have more than 30-meter distance errorAccordingly algorithms using relative information for thefingerprint outperform ones using the absolute vale of theRSS especially for third-floor environment Comparing theFreeloc and119873-gram the Freeloc shows better accuracy sinceFreeloc candetect uniqueBSSID subsequence effectively evenunder varying radio condition

Figure 12 shows total average error distance of fourlocalization schemes using 4 user devices used in aboveexperiment at the laboratory and third floor ldquoSinglerdquo indi-cates an average distance error of single pairwise evaluations


V VSNSNVNVSNS012345678

Erro

r dist

ance

(m)

(a) Desire HDV VSDVSDSVDSD

02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(c) Vega Iron 2V NDVNDVDVNDN

0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3

Figure 9 Localization error at a crowd indoor place with many pedestrians

At the laboratory most algorithms show competitive resultsthat error distance is below 5metersHowever signal strengthbased localization techniques result in serious outcomes atthe third floor At the laboratory RSSs of many fingerprintscan help in elaborating localization accuracy but can misleadin wide area like third floor Considering typical room sizeit could be challenge in localization performance Freelocachieves competitive localization performance in both areasAverage error distance of localization based on heteroge-neous multiple devices (ie 2 or 3 devices) indicates barswith the label ldquomultiplerdquo in Figure 12 When using multipleheterogeneous devices for building a radio map 119873-gramand Freeloc that use relative information obtain notable gaincompared to the single device surveying On the contrarydistance error of 119896-NN and Tanimoto gets even worsethan single device case In 119873-gram and Freeloc mergedfingerprint data do not affect localization results because theyonly concern relative information of them As some omittedfingerprints from a single device are filled by others appro-priately the accuracy is improved However in 119896-NN andTanimoto the same fingerprint is not merged properly andnewly added values from other devices canmislead outcomesbecause heterogeneous surveying devices without calibrationcan get different RSS values for the same location Errorratio that is a success rate of localization against trials at allmeasurement points is shown in Figure 12(c) Error ratio atthe laboratory is higher than at 3rd floor because distancebetween measurement points is too small to detect exactlocation

Figures 13 and 14 illustrate distribution of error distance atthe laboratory and third floor respectively Error distribution

indicates stability of the localization algorithms Although 119896-NN shows lower error ratio at the laboratory in Figure 12(c)distance error is widely spread (max error distance is almost11meters) compared to the Freeloc At the third floor distanceerror of the Freeloc is also distributed within very shortrange than others including the 119873-gram of which errorratio is similar in Figure 12(c) Consequently Freeloc canestimate a location with relatively low error variance ratherthan other localization schemes In case of using multipledevices for fingerprint surveying improvement compared tothe single device case is not very notable The heterogeneousdevicemade fingerprints contribute to reducing error ratio inFigure 12(c) but not to reducing distribution of error distancemuch

7 Performance with Varying Delta Value

In this section we describe how to derive the delta value forFreeloc fingerprints which was not handled in our previouswork [24] Practically two delta value acquisition methodscan be considered for Freeloc implementation one is aninstant derivation that occurs occasionally with several storedfingerprints and the other is an accumulated evaluation thatoccurs every time a new measurement is received

71 Delta Derivation with Instant Fingerprints To find out theoptimal 120575 value for subsequent localization the distance errorof pairwise devices is calculated with 120575 values varied from 1 to15 for all devices From this we can find the optimal empirical120575 value which minimizes average distance error Here theerror distance of the pairwise devices is calculated for the


GalaxyNexusOneNexus

4 6 8 10 122Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(b) Tanimoto similarity

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc

Figure 10 Localization of a Bionic device at laboratory

cases where the predicted location of the user was differentfrom the real location For instance if fingerprints measuredat point 1 in Figure 2(b) are regarded as fingerprints of point 2by our localization algorithm then error distance is an actualdistance between those two measurement points

For example the instant average of each distance errorfrom pairwise evaluation of four phones at the UCLAlaboratory decreases until 120575 = 12 as shown in Figure 15(a)

Accordingly optimal delta value can be set around 12 dBmas noted in the average line in Figure 15 Practically thelocalization server holds RSSs reported by several survey-ors for the 120575 evaluation time and derives the optimal 120575internally This instant derivation can be performed peri-odically or occasionally to adapt to changed environmentsuch as newly installed Wi-Fi APs and new type mobiledevices


GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc

Figure 11 Localization of a Bionic device at 3rd floor

72 Delta Derivation with Accumulated Fingerprints Withaccumulated fingerprints the localization server derives anew optimal 120575with a newly reported RSS from a phone basedon the current fingerprint For example an initial fingerprintfrom the Bionic (B) is compared by the Galaxy (G) RSS that isreported later With two a first 120575 value is derived After thenNexusS (NS) andNexusOne (NO) can update 120575 sequentiallyWhenever the fingerprint is reported it is used to derive120575 combined to a single fingerprint per each location Here

the localization server does not need to maintain raw RSSvalues of the reported fingerprints to derive 120575 The serveronly updates the current 120575 value based on fingerprints newlyreported by surveyors In Figure 15(b) for example first theBG pair has an optimal 120575 in 14 dBm and 120575 is updated to12 dBm by the NS Finally 120575 is set as 13 dBm by the NO

Figure 16 shows localization errors of the HTC DesireHD (D) with varying 120575 values at Gachon IT building Amerged fingerprint of two different devices does not show


k-NN Tanimoto N-gram FreelocLocalization

SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)

(a) Localization comparison at laboratory


SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)

(b) Localization comparison at 3rd floor

LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)

Tanimoto N-gram Freelock-NNLocalization

(c) Location detection error ratio

Figure 12 Comparison of different localization schemes

better performance than a single device fingerprint in someof 120575 values However the most robust fingerprint has beenformed from three different devices even though distanceerror is varying according to the 120575 value This experimentjustifies that ourmethod can achieve impressive performanceamong heterogeneous devices in crowdsourced environmentWe believe that error will decrease considerably as thehomogeneity of devices increase in the crowdsourced system

In this study considering general performance in termsof total distance error 10 dBm 120575 value is used for the UCLAengineering building and the Coex mall and 2 dBm is used

for the Gachon IT building and the resident house in the restof the evaluations

8 Conclusion

FreeLoc is a novel calibration-free indoor localizationscheme that uses existingWi-Fi infrastructureThe proposedradio map building and localization techniques are based onthe overall relationship among RSS by APs Our techniquesprovide robust localization accuracy in a crowdsourcedenvironment in which device heterogeneity and multiple


k-NN FreelocN-gramTanimotoLocalization

0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)

(a) Fingerprint of a single device


0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)

(b) Fingerprint of multiple devices

Figure 13 Error distributions at laboratory


0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)


Figure 14 Error distributions at 3rd floor

untrained surveyors mainly cause performance degradationThe main contribution of our work is a novel approach tofingerprint data management and a localization algorithmthat are capable of handling diverse users and their deviceswithout complicated calibrations or transformations Exper-iments using heterogeneous devices in two sites that havedifferent environments have confirmed that our novel schemeis reliable and feasible and achieve better performance than

localization algorithms using RSS absolute values We willthen expand the scale of our experiments to cover the entireuniversity and perform long-term usability testing

Competing Interests

The authors declare that there are no competing interestsregarding the publication of this paper


BionicGalaxyNexusOne

NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)

(b) Accumulated average

Figure 15 120575 derivation using instant or accumulated fingerprints at UCLA laboratory

1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)

Figure 16 Localization performance of a Desire HD fingerprint at5th floor of the Gachon IT building with varying delta value

Acknowledgments

This research was supported by the Gachon UniversityResearch Fund GCU-2015-0044 of 2015

References

[1] S Wang J Min and B K Yi ldquoLocation based services formobiles technologies and standardsrdquo Electronics pp 78ndash922008

[2] DMohapatra and S B Suma ldquoSurvey of location basedwirelessservicesrdquo inProceedings of the 7th IEEE International Conferenceon Personal Wireless Communications (ICPWC rsquo05) pp 358ndash362 January 2005

[3] M Mohammadi E Molaei and A Naserasadi ldquoA survey onlocation based services and positioning techniquesrdquo Interna-tional Journal of Computer Applications vol 24 no 5 pp 1ndash52011

[4] Y Gu A Lo and I Niemegeers ldquoA survey of indoor positioningsystems for wireless personal networksrdquo IEEE CommunicationsSurveys and Tutorials vol 11 no 1 pp 13ndash32 2009

[5] V Honkavirta T Perala S Ali-Loytty and R Piche ldquoA com-parative survey of WLAN location fingerprinting methodsrdquoin Proceedings of the 6th Workshop on Positioning Navigationand Communication (WPNC rsquo09) pp 243ndash251 Hannover Ger-many March 2009

[6] D Zhang F Xia Z Yang L Yao and W Zhao ldquoLocalizationtechnologies for indoor human trackingrdquo in Proceedings of the5th International Conference on Future Information Technology(FutureTech rsquo10) May 2010

[7] H Liu H Darabi P Banerjee and J Liu ldquoSurvey of wirelessindoor positioning techniques and systemsrdquo IEEE Transactionson Systems Man and Cybernetics Part C Applications andReviews vol 37 no 6 pp 1067ndash1080 2007

[8] N B Priyantha A Chakraborty and H Balakrishnan ldquoCricketlocation-support systemrdquo in Proceedings of the 6th AnnualInternational Conference on Mobile Computing and Networking(MOBICOM rsquo00) pp 32ndash43 August 2000

[9] RWant AHopper V Falcao and J Gibbons ldquoThe active badgelocation systemrdquo ACM Transactions on Information Systemsvol 10 no 1 pp 91ndash102 1992

[10] P Bahl and V N Padmanabhan ldquoRADAR an in-building RF-based user location and tracking systemrdquo in Proceedings ofthe 19th Annual Joint Conference of the IEEE Computer andCommunications Societies (INFOCOM rsquo00) pp 775ndash784March2000


[11] H Wang S Sen A Elgohary M Farid M Youssef and R RChoudhury ldquoNo need to war-drive unsupervised indoor local-izationrdquo inProceedings of theACM10th International Conferenceon Mobile Systems Applications (MobiSys rsquo12) pp 197ndash210Ambleside UK June 2012

[12] G Shen Z Chen P Zhang T Moscibroda and Y ZhangldquoWalkiemarkie indoor pathway mapping made easyrdquo in Pro-ceedings of the 10th USENIX Conference on Networked SystemsDesign and Implementation pp 85ndash98 USENIX Association2013

[13] A LaMarca Y Chawathe S Consolvo et al ldquoPlace Lab devicepositioning using radio beacons in the wildrdquo in PervasiveComputing Third International Conference PERVASIVE 2005Munich Germany May 8ndash13 2005 Proceedings vol 3468 ofLecture Notes in Computer Science pp 116ndash133 Springer BerlinHeidelberg Berlin Heidelberg 2005

[14] P Bolliger ldquoRedpin-adaptive zero-configuration indoor local-ization through user collaborationrdquo in Proceedings of the 1stACM International Workshop onMobile Entity Localization andTracking in GPS-Less Environments (MELT rsquo08) pp 55ndash60ACM San Francisco Calif USA 2008

[15] M Lee H Yang D Han and C Yu ldquoCrowdsourced radiomapfor roomlevel place recognition in urban environmentrdquo in Pro-ceedings of the 8th IEEE International Conference on PervasiveComputing and Communications Workshops (PERCOM Work-shops rsquo10) pp 648ndash653 IEEEMannheim Germany April 2010

[16] S Teller J Battat B Charrow et al ldquoOrganic indoor locationdiscoveryrdquo Tech Rep MIT 2008

[17] J-G Park D Curtis S Teller and J Ledlie ldquoImplications ofdevice diversity for organic localizationrdquo in Proceedings of the30th IEEE International Conference on Computer Communica-tions (INFOCOM rsquo11) pp 3182ndash3190 Shanghai China April2011

[18] J-G Park B Charrow D Curtis et al ldquoGrowing an organicindoor location systemrdquo in Proceedings of the 8th Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo10) San Francisco Calif USA June 2010

[19] J Ledlie J geun Park D Curtis A Cavalcante and L CamaraldquoMole a scalable user-generatedwifi positioning enginrdquo inPro-ceedings of the International Conference on Indoor Positioningand Indoor Navigation (IPIN rsquo11) pp 1ndash10 IEEE GuimaraesPortugal 2011

[20] A K M Hossain and W-S Soh ldquoA survey of calibration-freeindoor positioning systemsrdquo Computer Communications vol66 pp 1ndash13 2015

[21] CWu Z Yang Y Liu andW Xi ldquoWILL wireless indoor local-ization without site surveyrdquo IEEE Transactions on Parallel andDistributed Systems vol 24 no 4 pp 839ndash848 2013

[22] A K M M Hossain Y Jin W-S Soh and H N Van ldquoSSD arobust RF location fingerprint addressing mobile devicesrsquo het-erogeneityrdquo IEEE Transactions onMobile Computing vol 12 no1 pp 65ndash77 2013

[23] F Dong Y Chen J Liu Q Ning and S Piao ldquoA calibrationfreelocalization solution for handling signal strength variancerdquo inMobile Entity Localization and Tracking in GPS-Less Environ-nments Second International Workshop MELT 2009 OrlandoFL USA September 30 2009 Proceedings pp 79ndash90 SpringerBerlin Germany 2009

[24] S Yang P Dessai M Verma and M Gerla ldquoFreeLoc cali-bration-free crowdsourced indoor localizationrdquo in Proceedingsof the 32nd IEEE Conference on Computer Communications(INFOCOM rsquo13) pp 2481ndash2489 Turin Italy April 2013

[25] A Harter A Hopper P Steggles A Ward and PWebster ldquoTheanatomy of a context-aware applicationrdquo in Proceedings ofthe 5th Annual ACMIEEE International Conference on MobileComputing and Networking (MobiCom rsquo99) pp 59ndash68 ACM1999

[26] N B Priyantha A K Miu H Balakrishnan and S Teller ldquoThecricket compass for context-aware mobile applicationsrdquo in Pro-ceedings of the International Conference on Mobile Computingand Networking (MobiCom rsquo01) Rome Italy July 2001

[27] L Ni Y Liu Y C Lau and A Patil ldquoLandmarc indoor locationsensing using active rfidrdquo in Proceedings of the 1st IEEE Inter-national Conference on Pervasive Computing and Communica-tions (PerCom rsquo03) 101109PERCOM20031192765 pp 407ndash415 Fort Worth Tex USA March 2003

[28] G-Y Jin X-Y Lu and M-S Park ldquoAn indoor localizationmechanism using active RFID tagrdquo in Proceedings of the IEEEInternational Conference on Sensor Networks Ubiquitous andTrustworthy Computing vol 1 pp 40ndash43 IEEE TaichungTaiwan June 2006

[29] M Kanaan and K Pahlavan ldquoA comparison of wireless geolo-cation algorithms in the indoor environmentrdquo in Proceedingsof the in Wireless Communications and Networking Conference(WCNC rsquo04) vol 1 pp 177ndash182 IEEE 2004

[30] K Kaemarungsi and P Krishnamurthy ldquoProperties of indoorreceived signal strength for WLAN location fingerprintingrdquo inProceedings of the 1st Annual International Conference onMobileand Ubiquitous Systems Networking and Services (MOBIQUI-TOUS rsquo04) pp 14ndash23 August 2004

[31] S-H Fang T-N Lin and K-C Lee ldquoA novel algorithm formultipath fingerprinting in indoorWLANenvironmentsrdquo IEEETransactions onWireless Communications vol 7 no 9 pp 3579ndash3588 2008

[32] A Haeberlen E Flannery A M Ladd A Rudys D S Wallachand L E Kavraki ldquoPractical robust localization over large-scale80211 wireless networksrdquo in Proceedings of the 10th AnnualInternational Conference on Mobile Computing and Networking(MobiCom rsquo04) pp 70ndash84 Philadelphia Pa USA October2004

[33] AW Tsui Y-H Chuang andH-H Chu ldquoUnsupervised learn-ing for solving RSS hardware variance problem inWiFi localiza-tionrdquoMobile Networks and Applications vol 14 no 5 pp 677ndash691 2009

[34] D H Kim Y Kim D Estrin and M B Srivastava ldquoSensLocsensing everyday places and paths using less energyrdquo in Pro-ceedings of the 8th ACM International Conference on EmbeddedNetworked Sensor Systems (SenSys rsquo10) pp 43ndash56 ACM ZurichSwitzerland November 2010

[35] Y Jiang X Pan K Li et al ldquoARIEL automatic Wi-Fi basedroom fingerprinting for indoor localizationrdquo in Proceedingsof the 14th International Conference on Ubiquitous Comput-ing (UbiComp rsquo12) pp 441ndash450 ACM Pittsburgh Pa USASeptember 2012

[36] K Pahlavan and P Krishnamurthy Principles of Wireless Net-works A Unified Approach Prentice Hall New York NY USA1st edition 2001

[37] A Smailagic J Small and D P Siewiorek ldquoDetermining userlocation for context aware computing through the use of awireless LAN infrastructurerdquo ACMMobile Networks and Appli-cations vol 6 pp 1ndash8 2000

Submit your manuscripts athttpwwwhindawicom

Computer Games Technology

International Journal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Distributed Sensor Networks


Advances in

FuzzySystems

Hindawi Publishing Corporationhttpwwwhindawicom

Volume 2014


ReconfigurableComputing

Hindawi Publishing Corporation httpwwwhindawicom Volume 2014


Applied Computational Intelligence and Soft Computing

thinspAdvancesthinspinthinsp

Artificial Intelligence

HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014

Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Electrical and Computer Engineering

Journal of

Journal of

Computer Networks and Communications


Hindawi Publishing Corporation

httpwwwhindawicom Volume 2014

Advances in

Multimedia


Biomedical Imaging


ArtificialNeural Systems

Advances in


RoboticsJournal of



Computational Intelligence and Neuroscience

Industrial EngineeringJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Human-ComputerInteraction

Advances in

Computer EngineeringAdvances in



reduces the map building and maintenance expense com-pared to an expert surveyor based approach The CIL isdriven by normal mobile users who participate in the Wi-Fi AP survey as a contributor a user or both Also thismultiple-surveyor-multiple-usermodel has the advantages offast radio map building and quick update

Figure 1 shows an example of the CIL which consists oftwo phases ldquotraining phaserdquo (or ldquooffline phaserdquo) and ldquolocal-ization phaserdquo (or ldquoonline phaserdquo) In the training phasefingerprint data of locations have been captured to establisha radio map After then mobile phones inquire their locationwith measured fingerprint data to a remote server that holdsthe radio map in the localization phase In the figure the fiveusers (ie A B C D and E) are untrained normal peoplewho are carrying different types of mobile phones Users AD and E are surveyors for Room 101 Room 102 and Room103 whomeasure RSS ofWi-Fi APs and upload the measureddata with corresponding AP information to a local DB serverThe uploaded data is processed on the server and convertedto fingerprint dataWhen user B inquires the current locationinformation the server returns to Room 101 if user Brsquos RSSmeasurement matches one from the user A

However the CIL also introduces a new set of challengesFirst the CIL has to extract accurate fingerprint values fromshort measurement time for Received Signal Strength (RSS)because the mobile users as voluntary surveyors probablyprovide short-term RSS measurements Second the CILhas to support calibration-free indoor localization acrossdifferent devices Typically users as surveyors carry varioustypes of devices in terms of hardware and software for exam-ple Wi-Fi chipsets antenna pattern and operating systemsThe RSS measurement results in different values across theheterogeneous devices even at exactly the same positions

This crowdsourcing approach has been considered inmany recent researches [14ndash19] However many of themaddress those challenges in the CIL In [20] about survey-free localization several researches [21ndash23] dealt with het-erogeneous devices for CIL using difference of RSS samplesfrom APs instead of absolute RSS values to combat measure-ment variance from heterogeneous devices Previously weproposed ldquoFreelocrdquo [24] that uses a list of APs ordered by RSSfor the fingerprint instead of the RSS itself Thus the Freelocconsistently results in robust and reliable outcomes resolvingthose challenges In this paper we compare the Freeloc withexisting practical indoor localization algorithms for the CILin terms of accuracy In addition we investigate how todecide a gap between sample RSSs for difference localizationenvironment and conductmeasurement study for the Freelocin variance of wireless channel because of interference fromdense Wi-Fi APs deep fading from many pedestrians andpath loss from short distance

The remainder of this paper is organized as followswe introduce background knowledge with related works onindoor localization schemes in Section 2 Section 3 intro-duces existing localization algorithms for CIL Section 4describes our system and algorithm together with the threemajor challenges covered in the CIL We show the perfor-mance evaluation and comparison with existing localiza-tion algorithms from Sections 5 to 7 In Section 5 Freeloc

AP 1

AP 4

Room103

Room102

Room101

User E

User D

User A

User CUser B

AP 5

AP 6

AP 7

Remotelocalization sever Fingerprint

database


I am in ldquo101rdquo


AP 3

AP 2

Where am I

Where am I

Figure 1 Crowdsource Based Indoor Localization (CIL) approachwhich collects RSS values of Wi-Fi APs measured by multiple usersusers AndashE at different positions of Rooms 101 102 and 103 togetherwith corresponding AP information which is finally converted tofingerprint data of each location

performance is compared according to various localizationenvironmentWe compare Freeloc with other CIL algorithmsin Section 6 In Section 7 Freeloc performance is investigatedwith varying delta valuesWe conclude the paper in Section 8

2 Background

21 Indoor Localization Techniques Many indoor localiza-tion techniques have been proposed over the past decadeSome indoor localization techniques require special hard-ware to determine the devices location like active badge [9]and active bat [25] that are attached on ceilings and walls ofbuildings and transmits out infrared and ultrasound pulsesfor proximity estimationThe cricket and cricket compass [826] used a combination of RF and ultrasound technologiesTechniques using active RFID were also proposed [27 28]

Recently Wi-Fi based indoor localization receives muchattention asWi-Fi APs become used for outdoor localizationLike the outdoor localization in cellular networks triangu-lation using time difference of signal arrival from neighbourAPs can be considered [13 29] This time of arrival (TOA)based approach could suffer from timely varying signal pathat complex indoor environment Instead Received SignalStrength (RSS) can be used RADAR [10] adopts this RF signalintensity first for the purpose of indoor localization Nowmany researchers concentrate on techniques that use Wi-Fi










arg min119891119871119894isin119865

radicsum

119894=1

11989610038171003817100381710038171003817119891119894 minus 119891119871 119894

10038171003817100381710038171003817 (1)



arg max119891119871119894isin119865

119879 (119891119894 119891119871 119894) =

119891119894 sdot 119891119871 119894

100381710038171003817100381711989111989410038171003817100381710038172+10038171003817100381710038171003817119891119871 119894

10038171003817100381710038171003817

2minus 119891119894 sdot 119891119871 119894

(2)


arg max119891119871119894isin119865

prod

119904

119875 (119873-gram (119891119904) | 119871 119894) 119875 (119871 119894) (3)












119881fp


120596119877119879119898=1 (RSSpeak + 119898)

120596119871119879 + 120596119877119879 + 1

(4)






(2) bssidMAP larr






Fp (Loc119909) = Loc119909 [AP119894 119871 (AP119895120575(119894))]

119894 119895 isin APs of Loc119909 119894 = 119895


(5)



(a)


AP6

(b)






37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K


(c) Resident house

FE

I A BG

H C

D








S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in












arg min119891119871119894isin119865

radicsum

119894=1

11989610038171003817100381710038171003817119891119894 minus 119891119871 119894

10038171003817100381710038171003817 (1)



arg max119891119871119894isin119865

119879 (119891119894 119891119871 119894) =

119891119894 sdot 119891119871 119894

100381710038171003817100381711989111989410038171003817100381710038172+10038171003817100381710038171003817119891119871 119894

10038171003817100381710038171003817

2minus 119891119894 sdot 119891119871 119894

(2)


arg max119891119871119894isin119865

prod

119904

119875 (119873-gram (119891119904) | 119871 119894) 119875 (119871 119894) (3)












119881fp


120596119877119879119898=1 (RSSpeak + 119898)

120596119871119879 + 120596119877119879 + 1

(4)






(2) bssidMAP larr






Fp (Loc119909) = Loc119909 [AP119894 119871 (AP119895120575(119894))]

119894 119895 isin APs of Loc119909 119894 = 119895


(5)



(a)


AP6

(b)






37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K


(c) Resident house

FE

I A BG

H C

D








S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in













119881fp


120596119877119879119898=1 (RSSpeak + 119898)

120596119871119879 + 120596119877119879 + 1

(4)






(2) bssidMAP larr






Fp (Loc119909) = Loc119909 [AP119894 119871 (AP119895120575(119894))]

119894 119895 isin APs of Loc119909 119894 = 119895


(5)



(a)


AP6

(b)






37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K


(c) Resident house

FE

I A BG

H C

D








S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





(2) bssidMAP larr






Fp (Loc119909) = Loc119909 [AP119894 119871 (AP119895120575(119894))]

119894 119895 isin APs of Loc119909 119894 = 119895


(5)



(a)


AP6

(b)






37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K


(c) Resident house

FE

I A BG

H C

D








S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





37703771

3759

3809

3815 3811

3820

3828

3803

3803

3803

3803

3532

3514

3506 3507

3256A

A3285

3413

3417

3713

37313732

H

G G

GL LH

A

A

A

BB

C

C

C

D

D

D

F

F

E

E

E

F

3714

3704

3440 3436 3428 3420 3292 3286 3276 3256 S 3256 N

3400

C

A

A 34093277

3273

A

B

B

C

3564

3562

3680

3760

3750

J I

J

K K


(c) Resident house

FE

I A BG

H C

D








S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




S SNVNVSVSNVN0

05

1

15Er

ror d

istan

ce (m

)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

1

2

3

4

5

Erro

r dist

ance

(m)


0

1

2

3

4

5

Erro

r dist

ance

(m)

(d) Galaxy S3


D DVV

Erro

r dist

ance

(m)

0

05

1

15

2

25

3

(a) Galaxy S3

Erro

r dist

ance

(m)

S SVV0

05

1

15

2

(b) Desire HD

Erro

r dist

ance

(m)

D DSS2

21

22

23

24

25

(c) Vega Iron 2








(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





(a)


AP5

(b)


AP6

B NSNOG

Laboratory3rd floor

005

115

225

335

Erro

r dist

ance

(m)






AP6








3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 331

020406080

100120140160180200

0

50

100

150

200

250

0

50

100

150

200

250


1 353331292725232119171513119753 1 353331292725232119171513119753 1 353331292725232119171513119753

0

50

100

150

200

250

020406080

100120140160180200

0

5

10

15

20

25













NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




NSGNO

20 25 All150

05

1

15

2

25

3

35Er

ror d

istan

ce (m

)

(a) 34 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(b) 17 positions

NSGNO

20 25 All150

05

1

15

2

25

3

35

4

Erro

r dist

ance

(m)

(c) 8 positions









Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Coex mall20

25

30

35

40

45

50

55

60

65

70RS

S va

riatio

n di

strib

utio

n
















Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





Erro

r dist

ance

(m)


02468

10121416

Erro

r dist

ance

(m)

(b) Galaxy Note 4

D DSNSNDNDSNS0

2

4

6

8

10

12

Erro

r dist

ance

(m)


0

2

4

6

8

10

12

Erro

r dist

ance

(m)

(d) Galaxy S3









GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




GalaxyNexusOneNexus


0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus


0

02

04

06

08

1

CDF


GalaxyNexusOneNexus


0

02

04

06

08

1

CDF

(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc






GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




GalaxyNexusOneNexus

10 20 30 40 50 600Error distance (m)

0

02

04

06

08

1CD

F

(a) 119896-NN

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF

10 20 30 40 50 600Error distance (m)


GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(c) 119873-gram

GalaxyNexusOneNexus

0

02

04

06

08

1

CDF


(d) Freeloc







SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





SingleMultiple

0

1

2

3

4

5

6Er

ror d

istan

ce (m

)



SingleMultiple

0

5

10

15

20

25

30

35

40

45

Erro

r dist

ance

(m)


LaboratoryFloor 3

0

02

04

06

08

1

12

Erro

r rat

io (

)







8 Conclusion




0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





0

2

4

6

8

10

12

14Er

ror d

istan

ce (m

)



0

2

4

6

8

10

12

14

Erro

r dist

ance

(m)




0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)



0

10

20

30

40

50

60

70

Erro

r dist

ance

(m)





Competing Interests




NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





NexusSAverage

26

28

3

32

34

36

38

4

42

44Er

ror d

istan

ce (m

)

2 4 6 8 10 12 14 160Delta value (dBm)

(a) Instant average

BGBGNSBGNSNO

16

18

2

22

24

26

28

3

32

34

36

38

Erro

r dist

ance

(m)

2 4 6 8 10 12 14 160Delta value (dBm)



1 2 3 4 5 6 7 8

SSN

SNVAverage

002040608

1121416

Erro

r dist

ance

(m)


Acknowledgments


References














































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






































Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in



Research Article Crowdsource Based Indoor Localization by...

Documents

Transcript of Research Article Crowdsource Based Indoor Localization by...