researcharticle a wireless location system in lte networks

Research ArticleA Wireless Location System in LTE Networks

Qi Liu Rongyi Hu and Shan Liu

Network Technology Research Institute China Unicom Beijing 100084 China

Correspondence should be addressed to Shan Liu liushancandy163com

Received 29 September 2016 Revised 23 December 2016 Accepted 15 January 2017 Published 16 February 2017

Academic Editor Xiaohong Jiang

Copyright copy 2017 Qi Liu et al This is an open access article distributed under the Creative Commons Attribution License whichpermits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

Personal location technologies are becoming important with the rapid development of Mobile Internet services In traditionalcellular networks the key problems of user location technologies are high-precision synchronization among different base stationsinflexible processing resources and low accuracy positioning especially for indoor environment In this paper a new LTE locationsystem in Centralized Radio Access Network (C-RAN) is proposed which makes channel and location measurement moreavailable allocation of baseband processing resources more flexible and location service capability opening The location systemcontains more than two antenna clusters and each of them gets time-difference-of-arrival (TDOA) of sounding reference signals(SRSs) from different antennas Then based on data provided by location measurement units (LMUs) the location informationserver calculates TDOAs and derives the usersrsquo position Furthermore a new location algorithm is raised which can achievedistributed antennas collaboration and centralized location computing And an improved optimized algorithmwith the best TDOAselection is proposed Finally simulations are given out to verify the efficiency of the proposed algorithm in this LTE location system

1 Introduction

In recent years wireless location technologies especiallytechnologies for indoor environment have attracted atten-tion of many powerful companies including network oper-ators Although the Global Positioning System (GPS) couldprovide personal location services the 3rd Generation Part-ner Project (3GPP) has also proposeduser location projects inLTE networks Particularly terminals maybe cannot receiveGPS signals in indoor environment due to the weak receivedsignal and multipath reflection As LTE networks are beingdeployed around the world location technologies with LTEnetworks need to be studied further which could enhanceuser location services in all different environments [1ndash3]

There are mainly two challenges in traditional LTE loca-tion system [4 5] One is the limitation of indoor networksarchitecture causing the difficulties of distinguishing posi-tioning signals from DAS (distributed antenna system) andtime synchronization among different base stations And theother is the low positioning accuracy due to the locationalgorithm and the complex LOSNLOS communication envi-ronment [6 7]

According to the study of previous work [8ndash11] time-of-arrival (TOA) time-difference-of-arrival (TDOA) andhybrid location algorithms based on time-related measure-ments are the research hotspot at present Most locationsystems used to take TOATDOA techniques However fewlocation algorithms or schemes are proposed for LTE systemswith 20MHz bandwidth especially for complex scenariossuch as indoor positioning So new studies or algorithmsrelated to LTE systems are needed in order to improvepositioning performances of cellular networks especiallylocation accuracy and implementation complexity [12 13]

In this paper a wireless location system in LTE networksbased on C-RAN (Cloud Radio Access Networks) architec-ture is introduced [14ndash16] To be specific traditional cellularRadioAccessNetworks (RAN) usually consist ofmany stand-alone base stations It is a proposed architecture for futurecellular networks called C-RAN or Cloud-RAN which allowdistributed transmission of RF signals from remote radioheads (RRH) centralized processing of baseband signals inbaseband units (BBU) pool and real-time virtualization andcloud computing For the cellular location services thesefunctions can better support dynamic computing resource

HindawiMobile Information SystemsVolume 2017 Article ID 6160489 11 pageshttpsdoiorg10115520176160489

2 Mobile Information Systems

allocation and flexible selection of RF antennas which wouldguarantee both communication quality and positioning accu-racy Meanwhile high-precision synchronization among dif-ferent base stations for TDOA measurement is easily solvedby involving the LMU (location measurement unit) whichcan independently measure and calculate the TDOAs ofdifferent antennas By the way both stand-alone LMU andbase station with LMU function are supported in our system

In this paper we also propose a new method thatcalculates and analyzes the TDOAs of sounding referencesignals (SRSs) in uplink channel of C-RAN According tothe TDOAs of SRS and classical nonlinear least squares(NLLS) algorithm a new location algorithm based on theLTE location system is proposed It combines informationfrom different antenna clusters provided by LMUs whichdoes not require synchronization among BSs In addition wepropose an improved algorithm by optimizing the selectionof TDOAs in order to enhance the accuracy of the usersrsquolocation Both the experiment test and simulation resultsshow the performance gain of the proposed algorithms andverify the efficiency

The remainder of the article is organized as followsA wireless location system in LTE networks is given inSection 2 including the system model the performance ofSRS correlation for TDOA estimation and the proposedlocation algorithms In Section 3 the simulation and testresults are described Finally Section 4 concludes the paper

2 A Wireless Location System inLTE Networks

21 System Model Generally in order to ensure the locationperformance location systems need to be low-latency andhigh-precision especially for multiuser systems So the loca-tion procedure shall be completed in a short time (eg 10ms)in LTE networks including channel measurement large-scale computing and information feedback Meanwhilewith the development of positioning technology a softwareupdate of the related equipment comes naturally Traditionalcellular network architecture obviously cannot support thesefeatures However in C-RAN architecture operators canrapidly deploy or upgrade their networks and make full useof base station and antennas resources

In this section we introduce the user location mecha-nism in LTE system with the C-RAN architecture whereUplink-Time Difference of Arrival (U-TDOA) technology isapplicative Here location measurement of SRSs and channelmeasurement functions are achieved in LMUs RF antennasor RRHs are flexibly selected and managed and locationalgorithms and computing are centrally realized in BBU poolFigure 1 gives the LTE location system model in C-RANarchitecture

Firstly 119872 (119872 isin N) receiver units are used for collectingthe LTE signals from the target User Equipment (UE) orMobile Terminals (MT) In other words there are119872 antennaclusters deployed and 119873119898 (119873119898 isin N 119873119898 ge 2) antennas incluster-119898 (119898 = 1 2 119872) and the antenna gap is usuallyset as 1 to 10 meters for more accurate location measurement

By the way each location antenna cluster should contain atleast two antennas to insure the available measurement ofTOATDOA

Secondly there are 119872 wireless location measurementunits (LMUs) which connect the location servers in thissystem In cluster-119898 a LMU receives SRSs from each antennain this cluster extracts the time differences of SRSs fromdifferent antennas and calculates TDOAs It is worth men-tioning that there is no need for synchronization betweendifferent LMUs or base stations

Thirdly LMUs report location measurement data to thelocation information server including TDOAs and relatedantenna clustersrsquo information The server collects time dif-ferences from 119872 clusters and gets user location informationby the positioning algorithm In this architecture the timedelay of location information transmitting will be less thantraditional cellular architecture

Figure 2 describes functions and signal processing inthis location system The RF front ends receive the LTEsignals and transmit the signals to LMUs LMUs process thebaseband analog signals through Analog to Digital signalsconversion (AD) at first then exact single userrsquos SRS andcalculate TDOAs Finally location information server esti-mates the userrsquos position through the location algorithm andstores the information in databases of BBU pool

22 SRS Correlation Performances for TDOA EstimationTwo methods for TDOA estimation are widely used Oneis calculating the cross-correlation between two receivedsignals and getting TDOA [17] But it does not work well inmultipath environments especially for the indoor localization[18] Another is estimating TOA at first by cross-correlationbetween the received signal and the transmitted pseudoran-dom sequence And then the difference between the twoTOA estimations is calculated assuming that all receivers aresynchronized In this paper TDOA is detected and estimatedbased on the second method upon which the TDOA errormodel is built

SRS is a reference signal in LTE networks It is usuallyused to figure out the channel quality of uplink path MobileTerminals (MT) send SRS at the last symbol of a slot A MTcan transmit SRS signal every 2 subframes at the most andevery 32 frames (320 subframes) at the least A signalingparameter transmitted by eNodeB named as SRS-Config-Index tells UEs the periodicity of SRS transmission and theperiod can be 2 5 10 20 40 80 160 and 320 milliseconds(ms) [5] SRS is generated by Zadoff-Chu sequences whichare good candidates as their ideal correlation properties

In this section we evaluate SRS correlation performancein LTE networks by simulations In the simulations SRS isgenerated by a MT Then SRS and useful information dataare combined in a frame according to 3GPP LTE standardsThe eNodeB receives the uplink signals and maps frequencysignals to time domain obtaining the Single-Carrier Fre-quency Division Multiple Access (SC-FDMA) signals Thenthe received signals are correlated with the local SRS whichis the same with SRS generated by the MT So the correlationpeak is found The peak time is what we want to get for ourlocation system which is helpful for obtaining the accurate

Mobile Information Systems 3

LTE signalantennacluster-1

LTE signalantenna

cluster-M


LMULMULMU

Hub

Virtual BS1 Virtual BS2 Virtual BS3

BBU pool (location information server)

RRHRRH

RRH RRH

RRH

RRH RRHUE

RRHRRH UE

UE

UE

middot middot middot


Figure 1 LTE location system in C-RAN architecture

SRS correlationamp TDOA

calculation

digitalsignals

SRS parameters

Wireless positioningalgorithm

RF frontend

RF frontend

AD

AD

Multiuser baseband

analog signals

Multiuser baseband

Multiuserbaseband

digitalsignals

Multiuser baseband

analog signals

Single userrsquosTDOAs

amp antennainformation

Single userrsquos

Single userrsquos


calculation

Single userrsquos

TDOAs amp antenna

information

Location information

database SRS parameters

LMU-1

LMU-M

Location information serverAntennacluster-1

cluster-MAntenna

Figure 2 LTE location system function diagram

TDOA To be simplified the Additive White Gaussian Noise(AWGN) channel model is applied in the simulations Andthe default sampling rate is 3072MHz

In LTE networks SRSs from different users are trans-mitted by time-division mode or code-division mode Theircorrelation results are shown in Figure 3 In time-divisionmode as Figure 3(a) shows SRS from a MT is easy to distin-guish from othersrsquo which has a perfect performance of corre-lation peak Then we can easily get the TOAs and TDOAsof the different signals by detecting and analyzing theirSRSs

However as Figure 3(b) shows in the case of code-division mode there are more side peaks in correlationresults The time interval between each side peak is approxi-mately 410 ns It is greater than the accuracy of time advance(TA) which is a parameter in LTE networks So we can useTA to assist in finding the main peak time

In addition we can take advantage of oversampling beforethe SRS correlation uponwhich the errors of time differencescan be reduced Figure 4 shows SRS correlation performancecomparison between nonoversampling and oversampling Itis obvious that the correlation curve of oversampling is more


X 9333e minus 05Y 2698

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

86 8884 92 949 98 10 10296Time (s) times10

minus5

(a) Time-division mode

X 9333e minus 05Y 2684

X 9292e minus 05Y 2638

times10minus5

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

1029694 98 109288 98684Time (s)

(b) Code-division mode

Figure 3 SRS correlation performance in time-division and code-division mode

X 9333e minus 05Y 2698

X 9334e minus 05Y 1358

X 9333e minus 05Y 1356

times10minus5

9332 93339331 9335 93369334Time (s)

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n in

tens

ity

(a) Default sampling rate

X 9396e minus 05Y 1966e + 04

X 9396e minus 05Y 1981e + 04

X 9396e minus 05Y 2146e + 04

times104

times10minus5

0

05

1

15

2

25Th

e cor

relat

ion

inte

nsity

93945 93955 93965 93975 9398593935Time (s)

(b) Oversampling

Figure 4 SRS correlation performance comparison between nonoversampling and oversampling

smooth and this could increase the accuracy of peak timewith increased intensity of sample points So oversamplingcan be used for TDOA detection

23 Location Algorithm First we assume one cluster with119873 antennas in this location system and x represents aMTrsquos position to be estimated x minus x119899 is the true distancebetween antenna-119899 and theMTwhere x119899 represents antenna-119899rsquos position 119899 = 1 2 119873 Set antenna-1 as an anchor thenℎ119899(x) = xminusx119899minusxminusx1 represents the difference of distanceIt is defined that h(x) = [ℎ2(x) ℎ119899(x)]T where [sdot]T isthe transpose operation r represents measured difference ofdistance shown as (1) Here 120576 is measurement error andobeying normal distribution (0 1205901198992)

r = h (x) + 120576 (1)

h(x) is linearized about an initial reference position x0ignoring high-order terms by Taylor series expansion shownin (2) HereH0 is the Jacobian matrix of h(x) at x0

h (x) asymp h (x0) + H0 lowast (x minus x0) (2)

So (1) can be expressed as follows

r = y + h (x0) minus H0x0y = H0x + 120576 (3)

Here y is a linearized approximation of r Then x can becalculated by least squares (LS) estimation method shownin (4) Here R is the covariance matrix of 120576 and (sdot)minus1 is theinverse matrix operation

x = (HT0Rminus1H0)minus1HT

0Rminus1y (4)


In addition the algorithm also extends to the scenario of119872 clusters in the location system Assuming that each clustercan have different number of antennas (at least 2) we cancombine information getting from119872 clusters to calculate theMTrsquos position by (4) where R1015840 is the covariance matrix of 1205761015840

y1015840 =[[[[[[[

y1y2y119872

]]]]]]]

H10158400 =[[[[[[[

H01H02

H0119872

]]]]]]]

1205761015840 =

[[[[[[[

1205761

1205762

120576119872

]]]]]]]

(5)

As the closed form of the estimation position cannot bederived we apply an iterative method to minimize the erroraccording to Taylor series expansion as follows To be specificthe estimated position for the 119896th iteration should be derivedas

x (119896) = (H1015840T0 R1015840minus1H10158400)minus1H1015840T0 R1015840minus1y1015840 (6)

Algorithm 1 (the proposed algorithm)

Step 1 Initialize the position x0 and calculate y1015840H10158400 R1015840 with

(3) and (5)

Step 2 (iteration) (1) Estimate x(119896) with (5) and (6) where119896 isin N+ and x(0) = x0

(2) If x(119896) minus x(119896 minus 1) le 120590 stop else continue(3) Set x0 = x(119896) go to Step 1

Step 3 Return the final estimated position until a presetnumber of iterations are reached or until convergence

24 Improved LocationAlgorithm In LTE systems we cannotensure the links between signal sources and receivers (or thereceiving antenna clusters) are always in line of sight (LOS)due to the multipath Though all TDOAs getting from 119872clusters can be gathered together for calculation the badestimated TDOAs caused by NLOS or penetrationmaymakethe performance of the proposed Algorithm 1 worse

To deal with the bias caused by bad estimation weproposed an improved algorithm of TDOA selection whichis mainly based on kNN (119896-NearestNeighbor) algorithm anonparametricmethod used for classification and regression

First by kNN classification each estimated position calcu-lated by randomly divided TDOAs is classified by a majorityvote of its neighbors with the estimation being assigned tothe class most common among its 119896 nearest neighbors (119896 isa positive integer typically small) Particularly if 119896 = 1 it issimply assigned to the class of that single nearest neighborThen by kNN regression (which is simplified as calculatingthe average of the values of 119896 nearest neighbors) the bestestimated positions and their reliable TDOAs with propertyvalue are selected and the worst ones are got rid of

The detailed procedure of the improved algorithm ofTDOA selection is as follows Firstly a coarse selection isused Set an upper limit of TDOA (119879upper) according tothe maximum distance between receiving antennas in onecluster Then for all 1198730 TDOAs get rid of the obviously badones if

TDOA (119894) gt 119879upper 119894 = 1 2 1198730 (7)

Secondly the remaining TDOAs whose number isassumed as 1198731 will be selected in an accurate way Here tosimplify we denote TDOAs as 119879(119894) 119894 = 1 2 1198731 Choose119872 (3 le 119872 le 1198731) of the TDOAs randomly

1198601 1198602 1198601198621198721198731

sub 119879 (1) 119879 (2) 119879 (1198731) (8)

where 119860 119894 ≜ 119879(1199091198941) 119879(1199091198942) 119879(119909119894119872) and 1199091198941 1199091198942 119909119894119872 sub 1 2 1198731 119894 = 1 2 1198621198721198731 And calculate the predicted positions with (5) and (6)

in Algorithm 1 for all 1198621198721198731 cases expressed as pos(119894) (119894 =1 2 1198621198721198731)

Then according to the distribution of positions the rangeof all positions is divided into continuous 119869 (119869 ge 4)equivalent segmented blocks The cumulative probability ofeach segment 119895 (119895 = 1 2 119869) is expressed as

119875 (119895) = sum1198621198721198731119894=1 119901119895 (119894)1198621198721198731

119895 = 1 2 119869 (9)

where

119901119895 (119894) =

1 dis (119894) isin segment (119895)0 dis (119894) notin segment (119895)

119895 = 1 2 119869 119894 = 1 2 1198621198721198731 (10)

Therefore according to the cumulative probability dis-tribution we can get the densest segment expressedas segment(119895max) Furthermore the related positions insegment(119895max) are obtained whose number is added up as 119871Here1198621198721198731119869 lt 119871 le 1198621198721198731 Then themedian position of 119871 pointscan be easily obtained called point 119860 (pos119860) The distancebetween the positions and point 119860 can be expressed as

dis (119894) = 1003817100381710038171003817pos (119894) minus pos1198601003817100381710038171003817 119894 = 1 119871 (11)

By sorting all these distances the former 119870 (119870 le 119871)values can be achieved which are considered as the nearest


neighbor points and most likely estimated positions of targetlocation Here 119870 depends on the size of sample data

Finally according to the 119870 nearest neighbor points therelated TDOAs for pos1015840(119896) (119896 = 1 2 119870) are easy toachieve which are marked as 119860119895119896 = 119879(1199091198951) 119879(1199091198952) 119879(119909119895119872) where 119860119895119896 isin 1198601 1198602 119860119862119872

1198731

119895 isin 1 2 1198621198721198731 Then we can calculate the times of selected TDOAs asthe weight of TDOA(119894)

119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =

1 if 119894 = 119909119898 (119895) 119879 (119894) isin 119860 (119896)0 if 119894 = 119909119898 (119895) or 119879 (119894) notin 119860 (119896)

119898 = 1 sdot sdot sdot119872 119896 = 1 sdot sdot sdot 119870 119895 = 1 2 1198621198721198731 119894 = 1 sdot sdot sdot 1198731(13)

By sorting the weights of these TDOAs we can select thetop 1198732 (119872 le 1198732 le 1198731) TDOAs which are the optimizedselection of TDOAs Based on these TDOAs and Algorithm 1above we can recalculate the optimized estimated position

The steps of the improved Algorithm 2 are as follows

Algorithm 2 (the improved algorithm)

Step 1 (coarse selection) Set an upper limit of TDOA and getrid of the bad TDOAs with (7)

Step 2 (accurate selection) (1) Calculate the positions with(5) (6) and (8) by random 119872 TDOAs for all 119862119872119873 cases

(2) Calculate the cumulative probability of continuous 119869segmented blocks with (9) and get the densest segment 119895maxand the related 119871 positions

(3) Select the 119870 nearest neighbor points with (11) afterresorting the distances

(4)Weight the related TDOAswith (12) and select top1198732TDOAs

Step 3 Calculate the final optimized position with theselected TDOAs using Algorithm 1

3 Simulation and Test Results

31 Simulation Results In this section we mainly evaluatethe performance of proposed location algorithms in the LTElocation system through several simulations

Some common simulation settings are as follows SRS isconfigured in time-division mode among different users Tobe simplified the Additive White Gaussian Noise (AWGN)channel model is applied in the simulations The defaultsampling rate is 3072MHz and 10 times oversampling isused in this location system We assume that the simulationscenario is a room whose size is 20m times 20m times 3m and119888 (119888 = 1 2 5) antenna clusters are deployed in it Eachcluster contains 119898 (119898 = 2 3 4) antennas

Two-cluster algorithmSingle-cluster algorithm

01

015

02

025

03

035

Estim

ate e

rror

(m)

2 3 4 5 6 7 8 9 101Measurement error covariance of time difference

among different antennas (1205902)

Figure 5 The performance of LTE location system between singlecluster and 2 clusters

In the first simulation group users are randomly dis-tributed Firstly performance comparisons between singlecluster (with 4 antennas) and 2 clusters (one with 4 antennasand the other with 2 antennas) are given in Figure 5 Thelocation accuracy of 2 clusters is better than that of singlecluster Its estimating error is improved about 25

Furthermore Figure 6 shows the cumulative distributionfunction (CDF) of estimation error with different antennaclusters (clust119873) each of which has 4 antennas Owing to thefact that simulation scenario (a room 20m times 20m times 3m) islimited the performance of 1 cluster is not quite good butseems to be nearly perfect when there are above 2 antennaclusters deployed It is easy to find that the estimation errorgoes smaller with antenna clusters number increasing andthe estimation error is controlled within 11 meters whenthere are more than 2 clusters Particularly for 5 clusters theestimation error can be even all controlled below 05 metersIn some degree the results show that the more antennaclusters can bring better performance

In the second simulation group we assumed that 2 clus-ters are used and the userMTmoves in a certain area amongthe clusters The location performances are compared whenthe user is in different positions As shown in Figure 7 as theuser moves from the edge of the area to the center in a linethe location accuracy is obviously improved For examplewhen the user moves at the center areas among the antennaclusters like (3 1) or (35 1) the estimating error is lower than01 meters rather better than that of the edge areas especiallyat (0 1) and (45 1)

In the last simulation group assuming the MT randomlydistributed we give out the performance by the cumulativedistribution function (CDF) of estimation error for differentclusters (clust119873 = 3 4 5) And the comparison betweenthe proposed algorithms that is Algorithm 1 based on the


0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5

Figure 6 The performance of LTE location system with differentantenna clusters where clust119873 = 1 2 5

005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)



Figure 7 The performance for user in different positions in LTElocation system

weighted least square method and the improved Algorithm 2based on TDOA selection is also analyzed Here we assumedthe good TDOAs (1198732) are mostly accounted for 75 of all(1198730)

As shown in Figure 8 it is easy to find that the estimationerror goes smaller with clusters number increasing for bothlocation algorithms Meanwhile for all three cases with

AlgorithmAlgorithmAlgorithm


16 18 208 1 12 140402 060Estimation error (m)

0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5

Figure 8 The performance comparison of the proposed two algo-rithms where clust119873 = 3 4 5 and each has 4 antennas

different clusters the cumulative probability of estimationerror with Algorithm 2 has a better performance especiallywhen there are 3 antenna clusters At the expense of increasedtime complexity Algorithm 2 can obviously improve theaccuracy of location which proves that the optimization ofTDOA selection is effective and feasible

32 Test Results In order to testify the performance of theproposed system and algorithms we have done some experi-ment in a conference room The size of the room is 122m lowast17m lowast 38m and 2 antenna clusters are deployed in it eachcluster has 4 antennas The experiment system is composedas we describe in Section 2

LMUs are used to process the data received by antennaclusters and the composition and features of LMU arepresented in detail as Figure 9 Each unit has 4 receiving chan-nels and the signal in each channel is expected to be sampledapproximately at the same time Due to the fact that a smalltiming error may cause a large mistake during estimation forTDOA localization it is imperative to precisely synchronizeeach channel Through location measurement module loca-tion data can be collected and sent to the location sever foranalyzing and the estimated position is finally calculated bythe proposed algorithms In a word this design can releasethe demand of synchronization between base stations andsave the cost of network side

Firstly we try to locate fixed targets in the conferenceroom by using the deployed system Figure 10 shows thepositioning results of 6 fixed targets Each picture presents theestimated position distribution of 50 test results where theblue star is the target (as the center of a blue circle with radius


4 channels4 ADCs

FPGA

ARM

VC0clockbuffer

Ports(Ethernetserial)

Figure 9 The composition of location measurement unit

10 155Width of lab (m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



Figure 10 Multiple measurement results

of 3 meters) and the red points are the 50 estimations Thefinal position results are shown in Figure 11 by aggregatingthe multiple measurement results

From the statistic analysis about the positioning results inTable 1 as we can see the possibility of measurement error

less than 3m is about 80 And the maximum estimationerror is 3m the minimum is 047m and the average is 13mDue to the complexity of real environment the measurementerror is larger than the simulation results but still can satisfythe indoor positioning precision


5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)


15105Width of lab (m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


Figure 11 Final estimation of position

Table 1 Statistic analysis of measurement results

Number lt2m () lt3m () lt4m () Measurementerror (m)

1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130

Then we testify the UErsquos path tracing performanceFigure 12 presents the path tracking result of the UE movingat walking speed The red line denotes the true path and theblue line is the estimated trajectory It can be seen that theestimated trajectory is very close to the true path and themeasurement error is below 1m Comparing with the fixedmeasurement the system of dynamic positioning is morestable andmore preciseThe possible explanation of the resultis that UErsquos movement makes the error furtherly comply withGaussian distribution which can increase the precision byaveraging multiple measurements

In addition both the simulation and test results show thatthe LTE location system and algorithm have good perform-ance In order to deal with the complexity of real scene weimprove our algorithm by selecting TDOA and get the finalposition result by aggregating multiple position points Thismethod can reduce the influence of multipath

But our test environment is still relatively simple andhollowness In a more complex environment especially withserious multipath like indoor case our algorithms may notwork well Other improved algorithms or multipath miti-gation technologies should be taken into consideration Forexample in cellular communication systems the hybridTDOA and the Angle Of Arrival (AOA) location algorithmcan reach higher accuracy than traditional TDOA methoddoes

And terminal-side hybrid locations withWiFi Bluetoothor other RF location systems are not precluded to improvethe poor precision of wireless location in cellular network [1920] Besides the assistant positioning technologies of termi-nal-side equipment like inertial navigation and better filteringalgorithms of tracking will be good for location estimationand prediction [21]

4 Conclusions

This paper introduces a new user location system in LTE net-works with C-RAN architecture In this system uplink SRS


Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2

Figure 12 The test result of path tracking

is used asmonitoring signals LMUs are responsible for signaldetection and TDOA estimation The location informationserver calculates the userrsquos position through the proposedalgorithms Furthermore based on the basic TDOA algo-rithm in most location systems a new location algorithmis raised It reduces synchronization demands among dis-tributed antennas which is verified to have a good per-formance on position estimation In addition to avoid theeffects of NLOS and multipath overlap we also proposean improved optimization algorithm by selecting the bestTDOAs Simulation results show that the proposed algorithmhas performance gain and improves the efficiency and accu-racy of the location system in C-RAN architecture

Competing Interests

The authors declare that they have no competing interests

Acknowledgments

This work is supported by National Key Research and Devel-opment PlanGrants no 2016YFB0502000The authorswouldlike to acknowledge the helpful pieces of advice from ProjectPartner Professor Deng Zhongliang in Beijing University of

Posts and Telecommunications and Professor Liu HuapingWang Youxiang PhD Qiu Jiahui PhD Ma Yue ZhangWenhao and Chen Yi

References

[1] P J Voltz and D Hernandez ldquoMaximum likelihood time ofarrival estimation for real-time physical location tracking of80211ag mobile stations in indoor environmentsrdquo in Proceed-ings of the Position Location andNavigation Symposium (PLANSrsquo04) pp 585ndash591 April 2004

[2] C-P Yen and P J Voltz ldquoIndoor positioning based on statisticalmultipath channel modelingrdquo EURASIP Journal on WirelessCommunications and Networking vol 2011 article no 189 2011

[3] Q Liu and Y Ma ldquoA wireless location system in LTE networksrdquoin Proceedings of the 2nd IEEE International Conference onConsumer ElectronicsmdashTaiwan (ICCE-TW rsquo15) pp 160ndash161June 2015

[4] SWei LQi L Yiqun BMeng CGuoli andZHongke ldquoSmallcell deployment and smart cooperation scheme in dual-layerwireless networksrdquo International Journal of Distributed SensorNetworks vol 2014 Article ID 929805 7 pages 2014

[5] S Sesia I Toufik and M Baker LTEmdashThe UMTS Long TermEvolution John Wiley amp Sons 2011


[6] H Wu Y Chen N Zhang et al ldquoThe NLOS error mitigationjoint algorithm in hybrid positioning system combining DTMBand GPSrdquo in Proceedings of the China Satellite NavigationConference (CSNC rsquo12) pp 287ndash296 Guangzhou China May2012

[7] W Kim J G Lee and G-I Jee ldquoThe interior-point methodfor an optimal treatment of bias in trilateration locationrdquo IEEETransactions on Vehicular Technology vol 55 no 4 pp 1291ndash1301 2006

[8] R Ye and H Liu ldquoUWB TDOA localization system receiverconfiguration analysisrdquo in Proceedings of the International Sym-posium on Signals Systems and Electronics (ISSSE rsquo10) pp 205ndash208 September 2010

[9] Y Zhou L Jin C Jin and A Zhou ldquoFIMO a novelWiFi locali-zation methodrdquo inWeb Technologies and Applications vol 7808of Lecture Notes in Computer Science pp 437ndash448 SpringerBerlin Germany 2013

[10] A Catovic and Z Sahinoglu ldquoThe Cramer-Rao bounds ofhybrid TOARSS andTDOARSS location estimation schemesrdquoIEEE Communications Letters vol 8 no 10 pp 626ndash628 2004

[11] M R Gholami S Gezici and E G Strom ldquoA concave-convexprocedure for TDOAbased positioningrdquo IEEECommunicationsLetters vol 17 no 4 pp 765ndash768 2013

[12] R Ye S Redfield and H Liu ldquoHigh-precision indoor UWBlocalization technical challenges andmethodrdquo in Proceedings ofthe IEEE International Conference on Ultra-Wideband (ICUWBrsquo10) pp 1ndash4 IEEE Nanjing China September 2010

[13] B Kempke P Pannuto and P Dutta ldquoHarmonia widebandspreading for accurate indoor RF localizationrdquo in Proceedingsof the 1st ACM MobiCom Workshop on Hot Topics in Wireless(HotWireless rsquo14) pp 19ndash23 New York NY USA September2014

[14] O Simeone A Maeder M Peng O Sahin and W Yu ldquoCloudradio access network virtualizing wireless access for dense het-erogeneous systemsrdquo Journal of Communications and Networksvol 18 no 2 Article ID 7487951 pp 135ndash149 2016

[15] J Wu Z Zhang Y Hong and Y Wen ldquoCloud radio accessnetwork (C-RAN) a primerrdquo IEEE Network vol 29 no 1 pp35ndash41 2015

[16] R Wang H Hu and X Yang ldquoPotentials and challenges ofC-RAN supporting multi-RATs toward 5G mobile networksrdquoIEEE Access vol 2 pp 1200ndash1208 2014

[17] C H Knapp and G C Carter ldquoThe generalized correlationmethod for estimation of time delayrdquo IEEE Transactions onAcoustics Speech and Signal Processing vol 24 no 4 pp 320ndash327 1976

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

[19] S He and S-H G Chan ldquoWi-Fi fingerprint-based indoorpositioning recent advances and comparisonsrdquo IEEE Commu-nications Surveys and Tutorials vol 18 no 1 pp 466ndash490 2016

[20] S Bozkurt S Gunal U Yayan andV Bayar ldquoClassifier selectionfor RF based indoor positioningrdquo in Proceedings of the 23rdSignal Processing and Communications Applications Conference(SIU rsquo15) pp 791ndash794 IEEE Malatya Turkey May 2015

[21] K He Y Zhang Y Zhu W Xia Z Jia and L Shen ldquoA hybridindoor positioning system based on UWB and inertial naviga-tionrdquo in Proceedings of the International Conference on WirelessCommunications and Signal Processing (WCSP rsquo15) pp 1ndash5Nanjing China October 2015

Submit your manuscripts athttpswwwhindawicom

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



allocation and flexible selection of RF antennas which wouldguarantee both communication quality and positioning accu-racy Meanwhile high-precision synchronization among dif-ferent base stations for TDOA measurement is easily solvedby involving the LMU (location measurement unit) whichcan independently measure and calculate the TDOAs ofdifferent antennas By the way both stand-alone LMU andbase station with LMU function are supported in our system

In this paper we also propose a new method thatcalculates and analyzes the TDOAs of sounding referencesignals (SRSs) in uplink channel of C-RAN According tothe TDOAs of SRS and classical nonlinear least squares(NLLS) algorithm a new location algorithm based on theLTE location system is proposed It combines informationfrom different antenna clusters provided by LMUs whichdoes not require synchronization among BSs In addition wepropose an improved algorithm by optimizing the selectionof TDOAs in order to enhance the accuracy of the usersrsquolocation Both the experiment test and simulation resultsshow the performance gain of the proposed algorithms andverify the efficiency

The remainder of the article is organized as followsA wireless location system in LTE networks is given inSection 2 including the system model the performance ofSRS correlation for TDOA estimation and the proposedlocation algorithms In Section 3 the simulation and testresults are described Finally Section 4 concludes the paper

2 A Wireless Location System inLTE Networks

21 System Model Generally in order to ensure the locationperformance location systems need to be low-latency andhigh-precision especially for multiuser systems So the loca-tion procedure shall be completed in a short time (eg 10ms)in LTE networks including channel measurement large-scale computing and information feedback Meanwhilewith the development of positioning technology a softwareupdate of the related equipment comes naturally Traditionalcellular network architecture obviously cannot support thesefeatures However in C-RAN architecture operators canrapidly deploy or upgrade their networks and make full useof base station and antennas resources

In this section we introduce the user location mecha-nism in LTE system with the C-RAN architecture whereUplink-Time Difference of Arrival (U-TDOA) technology isapplicative Here location measurement of SRSs and channelmeasurement functions are achieved in LMUs RF antennasor RRHs are flexibly selected and managed and locationalgorithms and computing are centrally realized in BBU poolFigure 1 gives the LTE location system model in C-RANarchitecture

Firstly 119872 (119872 isin N) receiver units are used for collectingthe LTE signals from the target User Equipment (UE) orMobile Terminals (MT) In other words there are119872 antennaclusters deployed and 119873119898 (119873119898 isin N 119873119898 ge 2) antennas incluster-119898 (119898 = 1 2 119872) and the antenna gap is usuallyset as 1 to 10 meters for more accurate location measurement

By the way each location antenna cluster should contain atleast two antennas to insure the available measurement ofTOATDOA

Secondly there are 119872 wireless location measurementunits (LMUs) which connect the location servers in thissystem In cluster-119898 a LMU receives SRSs from each antennain this cluster extracts the time differences of SRSs fromdifferent antennas and calculates TDOAs It is worth men-tioning that there is no need for synchronization betweendifferent LMUs or base stations

Thirdly LMUs report location measurement data to thelocation information server including TDOAs and relatedantenna clustersrsquo information The server collects time dif-ferences from 119872 clusters and gets user location informationby the positioning algorithm In this architecture the timedelay of location information transmitting will be less thantraditional cellular architecture

Figure 2 describes functions and signal processing inthis location system The RF front ends receive the LTEsignals and transmit the signals to LMUs LMUs process thebaseband analog signals through Analog to Digital signalsconversion (AD) at first then exact single userrsquos SRS andcalculate TDOAs Finally location information server esti-mates the userrsquos position through the location algorithm andstores the information in databases of BBU pool

22 SRS Correlation Performances for TDOA EstimationTwo methods for TDOA estimation are widely used Oneis calculating the cross-correlation between two receivedsignals and getting TDOA [17] But it does not work well inmultipath environments especially for the indoor localization[18] Another is estimating TOA at first by cross-correlationbetween the received signal and the transmitted pseudoran-dom sequence And then the difference between the twoTOA estimations is calculated assuming that all receivers aresynchronized In this paper TDOA is detected and estimatedbased on the second method upon which the TDOA errormodel is built

SRS is a reference signal in LTE networks It is usuallyused to figure out the channel quality of uplink path MobileTerminals (MT) send SRS at the last symbol of a slot A MTcan transmit SRS signal every 2 subframes at the most andevery 32 frames (320 subframes) at the least A signalingparameter transmitted by eNodeB named as SRS-Config-Index tells UEs the periodicity of SRS transmission and theperiod can be 2 5 10 20 40 80 160 and 320 milliseconds(ms) [5] SRS is generated by Zadoff-Chu sequences whichare good candidates as their ideal correlation properties

In this section we evaluate SRS correlation performancein LTE networks by simulations In the simulations SRS isgenerated by a MT Then SRS and useful information dataare combined in a frame according to 3GPP LTE standardsThe eNodeB receives the uplink signals and maps frequencysignals to time domain obtaining the Single-Carrier Fre-quency Division Multiple Access (SC-FDMA) signals Thenthe received signals are correlated with the local SRS whichis the same with SRS generated by the MT So the correlationpeak is found The peak time is what we want to get for ourlocation system which is helpful for obtaining the accurate



LTE signalantenna

cluster-M


LMULMULMU

Hub



RRHRRH

RRH RRH

RRH

RRH RRHUE

RRHRRH UE

UE

UE





calculation

digitalsignals

SRS parameters


RF frontend

RF frontend

AD

AD

Multiuser baseband

analog signals

Multiuser baseband

Multiuserbaseband

digitalsignals

Multiuser baseband

analog signals



Single userrsquos

Single userrsquos


calculation

Single userrsquos

TDOAs amp antenna

information



LMU-1

LMU-M


cluster-MAntenna







X 9333e minus 05Y 2698

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

86 8884 92 949 98 10 10296Time (s) times10

minus5


X 9333e minus 05Y 2684

X 9292e minus 05Y 2638

times10minus5

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

1029694 98 109288 98684Time (s)



X 9333e minus 05Y 2698

X 9334e minus 05Y 1358

X 9333e minus 05Y 1356

times10minus5

9332 93339331 9335 93369334Time (s)

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n in

tens

ity


X 9396e minus 05Y 1966e + 04

X 9396e minus 05Y 1981e + 04

X 9396e minus 05Y 2146e + 04

times104

times10minus5

0

05

1

15

2

25Th

e cor

relat

ion

inte

nsity

93945 93955 93965 93975 9398593935Time (s)

(b) Oversampling




r = h (x) + 120576 (1)




r = y + h (x0) minus H0x0y = H0x + 120576 (3)



0Rminus1y (4)



y1015840 =[[[[[[[

y1y2y119872

]]]]]]]

H10158400 =[[[[[[[

H01H02

H0119872

]]]]]]]

1205761015840 =

[[[[[[[

1205761

1205762

120576119872

]]]]]]]

(5)





(3) and (5)








TDOA (119894) gt 119879upper 119894 = 1 2 1198730 (7)


1198601 1198602 1198601198621198721198731

sub 119879 (1) 119879 (2) 119879 (1198731) (8)




119875 (119895) = sum1198621198721198731119894=1 119901119895 (119894)1198621198721198731

119895 = 1 2 119869 (9)

where

119901119895 (119894) =


119895 = 1 2 119869 119894 = 1 2 1198621198721198731 (10)


dis (119894) = 1003817100381710038171003817pos (119894) minus pos1198601003817100381710038171003817 119894 = 1 119871 (11)





1198731


119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =
















01

015

02

025

03

035

Estim

ate e

rror

(m)









0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





LTE signalantenna

cluster-M


LMULMULMU

Hub



RRHRRH

RRH RRH

RRH

RRH RRHUE

RRHRRH UE

UE

UE





calculation

digitalsignals

SRS parameters


RF frontend

RF frontend

AD

AD

Multiuser baseband

analog signals

Multiuser baseband

Multiuserbaseband

digitalsignals

Multiuser baseband

analog signals



Single userrsquos

Single userrsquos


calculation

Single userrsquos

TDOAs amp antenna

information



LMU-1

LMU-M


cluster-MAntenna







X 9333e minus 05Y 2698

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

86 8884 92 949 98 10 10296Time (s) times10

minus5


X 9333e minus 05Y 2684

X 9292e minus 05Y 2638

times10minus5

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

1029694 98 109288 98684Time (s)



X 9333e minus 05Y 2698

X 9334e minus 05Y 1358

X 9333e minus 05Y 1356

times10minus5

9332 93339331 9335 93369334Time (s)

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n in

tens

ity


X 9396e minus 05Y 1966e + 04

X 9396e minus 05Y 1981e + 04

X 9396e minus 05Y 2146e + 04

times104

times10minus5

0

05

1

15

2

25Th

e cor

relat

ion

inte

nsity

93945 93955 93965 93975 9398593935Time (s)

(b) Oversampling




r = h (x) + 120576 (1)




r = y + h (x0) minus H0x0y = H0x + 120576 (3)



0Rminus1y (4)



y1015840 =[[[[[[[

y1y2y119872

]]]]]]]

H10158400 =[[[[[[[

H01H02

H0119872

]]]]]]]

1205761015840 =

[[[[[[[

1205761

1205762

120576119872

]]]]]]]

(5)





(3) and (5)








TDOA (119894) gt 119879upper 119894 = 1 2 1198730 (7)


1198601 1198602 1198601198621198721198731

sub 119879 (1) 119879 (2) 119879 (1198731) (8)




119875 (119895) = sum1198621198721198731119894=1 119901119895 (119894)1198621198721198731

119895 = 1 2 119869 (9)

where

119901119895 (119894) =


119895 = 1 2 119869 119894 = 1 2 1198621198721198731 (10)


dis (119894) = 1003817100381710038171003817pos (119894) minus pos1198601003817100381710038171003817 119894 = 1 119871 (11)





1198731


119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =
















01

015

02

025

03

035

Estim

ate e

rror

(m)









0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




X 9333e minus 05Y 2698

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

86 8884 92 949 98 10 10296Time (s) times10

minus5


X 9333e minus 05Y 2684

X 9292e minus 05Y 2638

times10minus5

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n pe

ak in

tens

ity

1029694 98 109288 98684Time (s)



X 9333e minus 05Y 2698

X 9334e minus 05Y 1358

X 9333e minus 05Y 1356

times10minus5

9332 93339331 9335 93369334Time (s)

0

500

1000

1500

2000

2500

3000

The c

orre

latio

n in

tens

ity


X 9396e minus 05Y 1966e + 04

X 9396e minus 05Y 1981e + 04

X 9396e minus 05Y 2146e + 04

times104

times10minus5

0

05

1

15

2

25Th

e cor

relat

ion

inte

nsity

93945 93955 93965 93975 9398593935Time (s)

(b) Oversampling




r = h (x) + 120576 (1)




r = y + h (x0) minus H0x0y = H0x + 120576 (3)



0Rminus1y (4)



y1015840 =[[[[[[[

y1y2y119872

]]]]]]]

H10158400 =[[[[[[[

H01H02

H0119872

]]]]]]]

1205761015840 =

[[[[[[[

1205761

1205762

120576119872

]]]]]]]

(5)





(3) and (5)








TDOA (119894) gt 119879upper 119894 = 1 2 1198730 (7)


1198601 1198602 1198601198621198721198731

sub 119879 (1) 119879 (2) 119879 (1198731) (8)




119875 (119895) = sum1198621198721198731119894=1 119901119895 (119894)1198621198721198731

119895 = 1 2 119869 (9)

where

119901119895 (119894) =


119895 = 1 2 119869 119894 = 1 2 1198621198721198731 (10)


dis (119894) = 1003817100381710038171003817pos (119894) minus pos1198601003817100381710038171003817 119894 = 1 119871 (11)





1198731


119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =
















01

015

02

025

03

035

Estim

ate e

rror

(m)









0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in





y1015840 =[[[[[[[

y1y2y119872

]]]]]]]

H10158400 =[[[[[[[

H01H02

H0119872

]]]]]]]

1205761015840 =

[[[[[[[

1205761

1205762

120576119872

]]]]]]]

(5)





(3) and (5)








TDOA (119894) gt 119879upper 119894 = 1 2 1198730 (7)


1198601 1198602 1198601198621198721198731

sub 119879 (1) 119879 (2) 119879 (1198731) (8)




119875 (119895) = sum1198621198721198731119894=1 119901119895 (119894)1198621198721198731

119895 = 1 2 119869 (9)

where

119901119895 (119894) =


119895 = 1 2 119869 119894 = 1 2 1198621198721198731 (10)


dis (119894) = 1003817100381710038171003817pos (119894) minus pos1198601003817100381710038171003817 119894 = 1 119871 (11)





1198731


119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =
















01

015

02

025

03

035

Estim

ate e

rror

(m)









0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in






1198731


119882(119894) =119870

sum119896=1

119872

sum119898=1

119908119896119898 (119894) (12)

where

119908119896119898 (119894) =
















01

015

02

025

03

035

Estim

ate e

rror

(m)









0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




0 05 1 15 2 25 3 350

02

04

06

08

1

Estimation error

CDF

clustN = 1

clustN = 2

clustN = 3

clustN = 4

clustN = 5


005

01

015

02

025

03

035

04

Estim

ate e

rror

(m)

(3 1)(2 1)(0 1)

(45 1)(35 1)









0

01

02

03

04

05

06

07

08

09

1

CDF

1 clustN = 3

2 clustN = 3

1 clustN = 4

2 clustN = 4

1 clustN = 5

2 clustN = 5







4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




4 channels4 ADCs

FPGA

ARM

VC0clockbuffer




5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)








5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




5

10

15

Leng

th o

f lab

(m)


5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)



5

10

15

Leng

th o

f lab

(m)

5

10

15

Leng

th o

f lab

(m)





1 148 586 879 2502 691 853 927 0863 264 770 100 3004 854 958 100 0475 720 880 100 0536 495 838 950 190Average 535 793 920 130





4 Conclusions



Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


Acknowledgments



References






























Advances in

FuzzySystems


Volume 2014












Journal of

Journal of





Advances in

Multimedia


Biomedical Imaging



Advances in


RoboticsJournal of










Advances in




Council board

Table

Table

Table

Table

TablePillar

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Table

Window

Wall

Wall

Real pathTest path

Cluster 1Cluster 2



Competing Interests


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



researcharticle a wireless location system in lte networks

Documents