1. Trang chủ
  2. » Thể loại khác

DSpace at VNU: Outage analysis in cooperative cognitive networks with opportunistic relay selection under imperfect channel information

9 107 0

Đang tải... (xem toàn văn)

THÔNG TIN TÀI LIỆU

Thông tin cơ bản

Định dạng
Số trang 9
Dung lượng 0,93 MB

Các công cụ chuyển đổi và chỉnh sửa cho tài liệu này

Nội dung

The impact of ICI1 on the opportunistic relay selection in dual-hopcognitivenetworksi.e.,withoutconsideringthedirect channel,whichselectstherelaywiththemaximumend-to-end 1 The impact of

Trang 1

jo u r n al ho m e p a g e :w w w e l s e v i e r c o m / l o c a t e / a e u e

Khuong Ho-Van∗

Telecommunications Engineering Department, Ho Chi Minh City University of Technology, 268 Ly Thuong Kiet Str., District 10, Ho Chi Minh City, Viet Nam

a r t i c l e i n f o

Article history:

Received 8 February 2015

Accepted 7 August 2015

Keywords:

Opportunistic relay selection

Imperfect channel information

Cognitive radio

Performance saturation

a b s t r a c t

1 Introduction

Traditional static spectrum allocation is not flexible and

induceslowspectrumutilizationefficiency[1].Thisissueturned

withhighradiospectrumdemandofemergingwirelessservices

requires appropriate solutions to mitigate current spectrum

under-utilization.Cognitiveradiotechnology,which allows

sec-ondary/unlicensed users (SUs) to opportunistically access the

spectruminherentlyallotted toprimary/licensedusers (PUs),is

arightsolutiontothesecriticalissues[2].Nevertheless,inorder

toassuretransparentcommunicationofPUs,SUsmustlimittheir

transmitpowerforacceptableinterferenceatPUs,andthus,

reduc-ingthecommunicationcoverageofsecondarytransmitters.With

theadvantageofwide radiocoverage,relayingtechniqueshave

recentlybeenincorporatedintoSUstocomplementthedrawback

oftheshortradiorangeof SUs[3].Therelayingprocesscanbe

assistedbymultiplerelaysforhighperformancebut low

band-widthefficiencyduetotherequirementoforthogonalchannelsfor

differentrelaysinordertopreventmutualinterference.Assuch,

selectingasinglerelayamong allpossiblecandidatesaccording

∗ Corresponding author Tel.: +84 1229900719.

E-mail address: khuong.hovan@yahoo.ca

to a certain criterion is preferred tooptimize system resource utilization(e.g.,powerandbandwidth),incomparisonwith multi-relayassisted transmission whileremaining thesame diversity order[4].Furthermore,channelstateinformationisvery impor-tantin theprocessofsystemdesign optimization(e.g.,optimal signaldetection).However,itisinevitablethatthis information cannotbecollectedwithoutanyerror.Therefore,theimpactof imperfectchannelinformation(ICI)ontheoutageperformanceof relayselection criteriaincognitiverelayingnetworksshouldbe thoroughlyinvestigatedbeforepracticalimplementation

The impact of ICI1 on the opportunistic relay selection in dual-hopcognitivenetworks(i.e.,withoutconsideringthedirect channel),whichselectstherelaywiththemaximumend-to-end

1 The impact of ICI on cognitive radio networks was studied in different aspects; for example, dual-hop relaying with relay selection (e.g., [7,6,5]), direct trans-mission (i.e., no relay) [8], the amplify-and-forward relay selection (e.g., [10,9]), relay non-selection (e.g., [11,13,15,12,14]) Moreover, several relay selection crite-ria in cognitive radio networks are suggested without investigating the impact

of ICI in [16–18,27,28,26,29,23,30,24,31,38,32,35,36,34,25,33,37,39,44,43,42,40,41] The current paper concentrates on the opportunistic relay selection in decode-and-forward cooperative cognitive networks, and thus, the literature related to the aspects studied in [11,16–18,27,28,26,29,23,30,24,31,38,32,35, 36,34,25,33,37,44,43,42,40,7,6,8,10,9,13,15,12,14,39,41,5] should not be further surveyed.

http://dx.doi.org/10.1016/j.aeue.2015.08.004

1434-8411/© 2015 Elsevier GmbH All rights reserved.

Trang 2

signal-to-noiseratio(SNR),isinvestigated in[45].Nevertheless,

thisworkdoesnotinvestigateICIonallfadingchannels

simulta-neously,thatis,ICIoninterferencechannels(i.e.,channelsfrom

SUstoPUs)whileperfectchannelinformation(PCI)on

transmis-sionchannels(i.e.,betweenSUs);orICIontransmissionchannels

butPCIoninterferencechannels.Also,[45]onlyconsiders

indepen-dentpartially-identical(i.p.i.)fadingdistributions(i.e.,relaysare

assumedtobecloselylocated).TheeffectofICIonthereactiverelay

selection,which selectstherelayamong allpossiblecandidates

(i.e.,allrelays areassumedtocorrectlydecodesource

informa-tion)withthelargestSNRtothedestination,andtheLth-worst

relayselection,whichselectstheLth-worstrelay,isinvestigated

in[46,47].However,both[46,47]arelimitedtothecaseofICIon

interferencechannelswhilePCIontransmissionchannels,andthe

i.p.i.fadingdistributionassumption.TheeffectofICIonthepartial

relayselection,whichsimplyselectstherelaywiththelargestSNR

fromthesource,isstudiedin[42,48–50].Nevertheless,for

analy-sissimplicity,theworksin[42,48–50]imposeseveralassumptions

suchasonlyICIoninterferencechannelswhilePCIontransmission

channels,i.p.i.fadingdistributions,anddual-hoprelaying

The opportunistic relay selection is proved to be

capacity-optimal[16],andhence,itisinterestingtopredictits

information-theoreticperformancelimit(i.e.,outageprobability).Motivatedby

theabove,thispaperthoroughlyanalyzesitsoutageperformance

Thecontributionsofthispaperaresummarizedbelow:

• Proposeanexactclosed-formoutageprobabilityexpressionfor

cooperativecognitivenetworkswithopportunisticrelay

selec-tion,neglectingallassumptionsof[45].Morespecifically,this

expressionisapplicabletoageneralscenario:ICIonall

chan-nelsconcurrently,i.n.i.fadingdistributions,maximumtransmit

powerconstraint,interferencepowerconstraint,theusageofthe

directchannel

• Derivetheperformancelimitofcooperativecognitivenetworks

withopportunisticrelayselection,whichprovesnodiversitygain

achievableinthepresenceofICI

• Perform numerouscomparisons betweendual-hop and

coop-erativecognitivenetworks withopportunistic relayselection,

whichdemonstrateasignificantgainofutilizingthedirect

chan-nelinrelayingcommunicationsatalmostnoexpenseofsystem

resources(e.g.,powerandbandwidth)

• Provide numerousresultsto haveuseful insightsinto system

performancesuchasperformancesaturationphenomenonand

considerableperformance deterioration due to ICI, significant

performanceimprovementwithrespecttotheincreaseinthe

numberofrelays

• Outline an interference probability2 expression to reflect the

effect of channel information imperfection on the quality of

serviceof primaryusers Illustrativeresultsare alsoprovided

to show that the interference probability is proportional to

thenumberofrelays,whichconflictswithoutageperformance

improvement of secondary users when the number of relays

increases,establishingtheperformancetrade-offbetweenthe

primarynetworkandthesecondarynetworkwithrespecttothe

numberofrelays

Therestofthecurrentpaperisstructuredasfollows.The

sys-temmodelunderconsiderationispresentedin Section 2.Exact

2 Interference probability is defined as the probability that the interference power

constraint is invalidated Some works proposed the interference probability

expres-sion for the partial relay selection (e.g., [42,48,50]) and the reactive relay selection

(e.g., [46]) To the best of the author’s knowledge, the interference probability

expression for the opportunistic relay selection have not been presented in open

andlimit outageanalysisframework fortheopportunisticrelay selectionincooperativecognitivenetworksaswellasindual-hop cognitivenetworksiselaboratelydescribedinSection3.The inter-ferenceprobabilityexpressionisoutlinedinSection4.Resultsand discussionsontheoutageperformanceofthesenetworksaswell

astheinterferenceprobabilityareprovidedinSection 5.Finally, thepaperisclosedwithusefulconclusionsinSection6

2 System model

Acooperativecognitivenetworkwithopportunisticrelay selec-tionisdemonstratedinFig.1.Cooperativerelayingisimplemented

inthesecondarynetworkinwhichinformationtransmissionfrom thesourceSstothedestinationSdishelpedbytheselectedrelaySb

inthegroupofKrelays,S={S1,S2, ,SK}.Weassumecognitive radiostooperateintheunderlaymechanism(e.g.,[16,19–22]),and hence,SsandSbinterferethePU,namelyPp,buttheinterference levelatPpmustbelowerthanthemaximuminterferencepower,

I,thatcanbetoleratedbyPp

Weinvestigatefrequency-flatandi.n.i.Rayleighfading chan-nels.Therefore,thechannelcoefficient,htr,betweenatransmitter andareceiver,wheretandrdenotetheindicesofthe transmit-terandthereceiver,respectively(thespecificvaluesoftandrwill

bespecifiedlater),canbemodelledasacircularsymmetric com-plexGaussianrandomvariablewithzeromeanand1/tr-variance, i.e.htr∼CN(0,1/tr).Sinceourworkinvestigatesi.n.i.fading distri-butions,alltr’s,∀{t,r},arenotnecessarilyequal.Therefore,itis moregeneralandpracticalthanmostexistingworksonrelay selec-tionwherei.p.i.(i.e.,tr’sarepartitionedintogroupsofidentical value.Forexample,sr’s,rd’s,rp’swithr∈{1, ,K}areassumed

tobeidenticalin[7,16,23,24,26,29,42,45–48])orindependentand identical(i.e.,tr’s,∀{t,r},areequalin[25,27,28,31])fading distri-butionsareassumedforsimplicityofperformanceanalysis

It is inevitablethat PCIis impossibly availableowing tothe limitationsofchannelestimationalgorithms.Assuch,inorderto support performance analysis,we should model channel infor-mation imperfection appropriately In this work, we apply the well-known channel information imperfection model used in

[8,10,13,39,42,45,46,48,49,51].Accordingtothis model, thereal channelcoefficient,htr,isrelatedtotheestimatedone, ˆhtras

ˆhtr=trhtr+

1−2

Phase 1

Phase 2

Primary receiver

Secondary network

Pp

Trang 3

where εtr is the channel estimation error, t∈

s,1, ,K

,r∈



p,1, ,K,d

.Aselaboratelydiscussedin[8],allrandom

vari-ables {ˆhtr, htr, εtr} are modelledas CN(0,1/tr) Moreover,the

correlationcoefficient0≤tr≤1isaconstant,characterizingthe

averagequalityofchannelestimation

Asexposed inFig.1,cooperative relayingwithopportunistic

relayselectiontakesplaceintwophases.Inthefirstphase,Ss

trans-mitsthesignalvswithtransmitpowerPs(i.e.,Ps=Evs{|vs|2}where

EX{·}denotestheexpectationovertherandomvariableX).Acertain

secondarytransmitterSt,t∈

s,1, ,K

mustobeythe under-laymechanismononehand(i.e.,Pt|htp|2≤I),andthemaximum

transmitpowerPdesignedforit(i.e.,Pt≤P)ontheotherhand.For

maximumtransmissioncoverage,theupper-boundofthetransmit

powerischosen,i.e.Pt=min(I/|htp|2

,P).Duetotheunavailability

ofchannelinformation,thetransmitpowermustbesetwiththe

estimatedchannelinformation(e.g.,[8])asPt=min(I/|ˆhtp|2,P).

ThereceivedsignalatSr,r∈

1, ,K,d

,canbeexpressedas

usr=hsrvs+nsrinwhichnsr∼CN(0,N0)istheadditivewhite

Gauss-iannoise(AWGN)atthesecondaryreceiverSr.Plugging(1)intousr,

oneobtains

usr= ˆhsr

srvs−



1−2

sr

whichgeneratesthesignal-to-noiseratio(SNR)atSrinthefirst

phaseas

sr= Evs{|ˆhsrvs/sr|2}

Evs,nsr{|nsr−

1−2

srεsrvs/sr|2}=ˇsr|ˆhsr|

2

BydenotingI=I/N0,P=P/N0,=I/P,x=|ˆhsp|2,theˇsrterm

in(3)canberepresentedinacompactformas

ˇsr= srPs

(1−2

sr)Ps+sr2

srN0

=

srI

sr2srx+(1−2sr)I ,x>

srP

(1−2sr)P+sr2sr

,x< (4)

Opportunisticrelayselectionoptsforarelay,namelySb,with

themaximumend-to-endSNR.Therefore,theSNRatSdthrough

therelayingchannelisexpressedas

sbd=max

whereR=

1,2, ,K

andkdistheSNRoftheSk−Sdchannel

Theexpressionofkdcanbeinferredinthesamemannerassr,i.e

where

ˇkd=

kdI

kd2

kdyk+(1−2

kd)I ,yk>

kdP

(1−2

kd)P+kd2

kd

,yk<

(7)

withyk=|ˆhkp|2.Itisnotedthatopportunisticrelayselectioncanbe

implementedinadistributedmannerusingthetimermethodin[4]

whereeachrelaySksetsitstimerwiththevaluethatisinversely

proportionaltomin(sk,kd)andtherelaywiththetimerthatruns

outfirstisselected

Inthesecondphase,theselectedrelaySbdecodesthesource

signalandre-encodesthedecodedinformationbeforeforwarding

toSd Then,Sd candecodethesourceinformationby

selection-combiningthereceivedsignalsinbothphases.Therefore,thetotal

SNRatS canberepresentedas =max( , ).Itisrecalled

thattheimplementationoftheselectioncombininginthispaper

issimplerthanthemaximumratiocombiningwithoutsignificant performancedegradation[52].Therefore,theselectioncombining

ispreferableinpracticalapplications.Moreover,sincebothsdand

sbd containa commontermx=|ˆhsp|2,theyarecorrelated Fur-thermore,thequantitiesmin(sk,kd)fordifferentk insidethe maximumoperatorin(5) arecorrelatedsincetheyalsocontain

x.TheseSNRcorrelations(i.e.,betweensdandsbd,andbetween min(sk,kd)andmin(si,id),∀(k,i))maketheperformance analy-sisinnextsectionscomplicatedbutaccurate.Itisalsonotedthatfor analysissimplicity,[45]assumedthatmin(sk,kd)isuncorrelated withmin(si,id),∀(k,i)whilethisdoesnothold

3 Outage performance analysis

Thissectionfirstlypresentsanexactoutageanalysisin coopera-tivecognitivenetworkswithopportunisticrelayselection,whichis thenusedtoinfersystemperformancelimits.Theproposed analy-sisframeworkisrelativelygeneral,andhence,basedonit,asimilar analysisindual-hopcognitivenetworkswithopportunisticrelay selection(e.g.[45])canbeperformedtoexposetheadvantageof utilizingthedirectchannelinrelayingcommunicationswithout time-consumingsimulations

3.1 Exactanalysis Theoutageprobabilityisdefinedastheprobabilitythate2e

isbelowathreshold0,i.e.PCC

o =Pr{e2e≤0}where0=22R−1 withRbeingtherequiredtransmissionrateandPr

X

denotesthe probabilityoftheeventX.Sincee2econtainstwocorrelated quan-tities,sdandsbd,PCC

o mustbeevaluatedintermsofconditional probabilities,i.e

PCCo =Pr

max (sd,sbd)≤0



=Ex



Pr

max (sd,sbd)≤0 x

=Ex

⎪Pr



sd≤0 x





Pr

sbd≤0 x

Since ˆhtr∼CN

0,1/tr



,theprobabilitydensityfunction(pdf) and the cumulative distribution function (cdf) of ˆhtr 2

are expressed as f ˆhtr 2(z) =tre−tr z and F ˆhtr 2(z) =1−e−tr z for

z≥0.Therefore,itimmediatelyfollowsthat

=Pr



ˇsd ˆhsd 2

≤0 x

=F ˆhsd 2



0

ˇsd

x



whereQsd=e−sd0ˇsd hasacommonformasQtr=e−tr 0ˇtr ,whichis furthersimplifiedafterusing(4)and(7)as

Qtr=e−ˇtrtr0 =

Atre−

tr2

tr0

I z ,z= ˆhtp 2

>

Btr ,z= ˆhtp 2

<

(10)

with

Atr =e(2tr −1) 0

Btr =e



1 − 2

tr +tr2 tr

P



 0

(11)

Trang 4

=Pr



max

k∈R (min (sk,kd))≤0 x

=

K



k=1

Pr

min (sk,kd)≤0 x

=

K



k=1



1ưPr

min (sk,kd)≥0 x

=

K



k=1



1ưPr

sk≥0 x

Pr

kd≥0 x

Since kd is independent of x, the conditional probability

Pr

kd≥0 x

becomes the unconditional one Pr

kd≥0



Therefore,(12)canbefurtherrewrittenas

=

K



k=1



1ưPr

sk≥0 x

Pr

kd≥0



Inserting(3)and(6)into(13)andaftersomebasic

manipula-tions,oneobtains

=

K



k=1



1ưPr



ˇsk ˆhsk 2

≥0

x



Pr



ˇkd ˆhkd 2

≥0



=

K



k=1



1ưPr

ˆhsk 2

≥ 0

ˇsk

x



Pr

ˆhkd 2

≥ 0

ˇkd



=

K



k=1





1ưF ˆhsk 2

  0

ˇsk

x



Eyk



1ưF ˆhkd 2

 0

ˇkd

yk



(14) Usingthegeneralnotationin(10)forbothQskandQkd,onecan

simplify(14)as

=

K



k=1



1ưQskEyk



Qkd



=

K

 k=1

(1ưQkdQsk) (15)

In(15),Qkd=Eyk



Qkd



.Itsexactclosed-formrepresentationis obtainedbyusingthecompactformofQkdin(10)as

Qkd=

 ∞



Akdeưkd

2

kd 0

I ykkpeưkp ykdyk+

  0

Bkdkpeưkp ykdyk

= Akdkp

whereAkd and Bkd aregiven in(11)withappropriate subscript

substitutions,and

Ck =1ưeưkp 

Dkd =kd

2

kd0

(17)

Usingthefactthat

K



k=1

(1ưak)=1+ (ư1)K

k∈R

ak

+

Kư1



u=1

(ư1)u Kưu+1

s =1

Kưu+2

s =s +1

···

K





k∈A

ak, (18)

whereA=

R [s1], ,R [su]3

,toexpandtheproductin(15),one obtains

=1+ (ư1)K

k∈R

Qkd



k∈R

Qsk

+

Kư1

 u=1

(ư1)u Kưu+1

s 1 =1

Kưu+2

s 2 =s 1 +1

···

K



s u =suư1+1



k∈A

Qkd



k∈A

Plugging(9)and(19)into(8)resultsinthesimplifiedformof

PCC

PCC

o =T∅+ (ư1)KT R



k∈R

Qkd

+

Kư1

 u=1

(ư1)u Kưu+1

s 1 =1

Kưu+2

s 2 =s 1 +1

···

K



s u =suư1+1

T A



k∈A

whereW= {∅,A, R} and

T W=Ex (1ưQsd)

k∈W

Qsk

!

=Ex



k∈W

Qsk

!

ưEx

 k∈{W,d}

Qsk

It is apparent that the derivation of the exact closed-form expressionofPCC

o iscompletedafteranalyticallyevaluating ˆT Gin

(21)whereGiseitherWor{W,d}.Towardsthisend,weuse(10)

andthenevaluatetheresultingintegralsas ˆ

T G=Ex



k∈G

Qsk

!

=

 ∞



speưsp x

k∈G

Askeưsk

2

sk 0

I xdx

+

  0

speưsp x

k∈G

Bskdx

=sp



k∈G

Ask

 ∞



e

ư

"

sp+0I

k∈G

sk 2 sk

# x

dx

k∈G

Bsk

  0

speưsp xdx=sp

EGeưEG

k∈G

Ask+Cs



k∈G

Bsk, (22)

where{Ask,Bsk}andCsaregivenin(11)and(17)withappropriate subscriptsubstitutions,respectively,and

EG=sp+0

I



k∈G

Itisworthemphasizingthatalthoughtheopportunisticrelay selection in this paper is elaborately discussed (e.g., [45]), the derivationofitsexactclosed-formoutageprobabilityexpression

in(20)hasnotbeenreportedinanyopenliteratureforthe gen-eralcaseofi.n.i.fadingdistributions,cooperativerelaying,andICI

onallchannelsconcurrently.Oncontrary,[45]considersdual-hop relaying,i.p.i.fadingdistributions,andICIeitheroninterference channels or transmission channels.To the best of the author’s knowledge,(20)istotallynovelandrepresentedinavery conve-nientandcompactformfortheanalyticalevaluation.Furthermore,

Trang 5

iseasilyextendedtothecorrespondinganalysisindual-hop

cog-nitive networks withopportunistic relay selection Indeed, the

outageprobability of dual-hop cognitivenetworks with

oppor-tunisticrelayselectionisgivenby

PDH

sbd≤0



ByfollowingthesameprocedureofderivingPCC

obtains

PDH

o = ˆT∅+ (ư1)KTˆR



k∈R

Qkd

+

Kư1



u=1

(ư1)u

Kưu+1

s 1 =1

Kưu+2

s 2 =s 1 +1

···

K



s u =suư1+1

ˆ

T A



k∈A

Itisinterestingtodiscoverfrom(20),(21),(25)thatT W < ˆT W

andPCC

o <PDH

o Inotherwords,cooperativerelayingunder

con-siderationin thispaperis alwaysbetterthandual-hoprelaying

in[45]atnosignificantcostofsystemresources(e.g.,powerand

bandwidth),and hence, the former shouldbe usedin practice

Itisalsonotedthat(25)accountsforthecorrelationamongthe

quantitiesmin(sk,kd)in sbdfor differentkand hasnotbeen

reportedinanyopenliteraturewhile[45]ignoresthiscorrelation

inperformancelimitanalysis

3.2 Performancelimitanalysis

Tohave insightsinto systemperformance limits,we should

investigatetheoutageperformance inthehighSNR regime,i.e

P→∞.SinceI=P,thisregimealsoimpliesI→∞.Therefore,

basedon(4)and(7),onecanapproximatethequantityQtrin(10)

as4

QtrP,I→∞

Using(26),wecanalsoapproximatethequantitiesin(9),(15),

(16)as

{Qkd,,}P,I→∞→ {Akd,1ưAsd,

K

 k=1

(1ưAskAkd)} (27)

Substituting(27)into(20),weobtaintheoutageperformance

limit (i.e., in the high SNR regime) of cooperative cognitive

networkswithopportunisticrelayselectionas

PCCo =(1ưAsd)

K



k=1

Followingthesameprocedureofderiving(28),onecanachieve

theoutageperformancelimitofdual-hopcognitivenetworkswith

opportunisticrelayselection(i.e.,inthehighSNRregime)as

PDHo =

K



k=1

Itisobservedfrom(28)and(29)thattheperformancelimits

ofbothdual-hopandcooperativecognitivenetworkswith

oppor-tunisticrelayselectiondependonlyonconstantsAtr,which are

functionsofcorrelationcoefficientstr.Assuch,thesenetworks

experience performance saturation phenomenon at high SNRs

X, Y, 

is the short-hand representation which denotes

(equivalently,nodiversitygaincanbeachievable)andtheir satu-rationlevelsonlydependonthequalityofchannelestimators(i.e.,

tr).Inotherwords,ICIcompletelydestroystheadvantageofthe relayselectionintermsofdiversitygain(i.e.,zerodiversityorder) ThisiscontrasttothecaseofPCI(i.e.,tr=1),wherethediversity orderisalwaysnon-zero.Indeed,insertingtr=1into(28)and(29)

resultsinzerooutageprobabilityandhence,thenon-zerodiversity orderisachievable.However,thesaturationlevelofcooperative relayingissmallerthanthatofdual-hoprelaying.Thiscanbeseen againfrom(28)and(29)thatPDHo /PCCo =1/(1ưAsd),andhence,the saturationlevelofcooperativerelayingis1/(1ưAsd)timessmaller thanthatofdual-hoprelaying.Theseresultsencourageutilizing thedirectchannelinrelayingcommunications

4 Interference probability

TheinterferencepowerItatthePUcaused bythesecondary transmitterStcanbeexpressedas

It=Pt|htp|2

=min

"

I

|ˆhtp|2,P

#

|htp|2

Due to channel information imperfection (i.e., htp /= ˆhtp), It

can exceed the maximum interference power I, violating the interferencepowerconstraintwhichisacriticalrequirementfor guaranteeingthequalityofserviceofPUs.Theprobabilitythatthe interferencepoweratthePUexceedsIisdefinedastheinterference probability,J.Fortheopportunisticrelayselectionunder consider-ation,therearetwocaseswhichcausetheinterferencepowerat thePUtoexceedI:

• Case1:Thiscasecorrespondsthefirstphaseofcooperative relay-ingprocessasshowninFig.1.Thesourcemodifiesitstransmit powerPsinaccuratelyandcausesaninterferencepowerIsin(30)

tobelargerthanI.

• Case2:Thiscase happenswhenthesourcedoesnotinterfere withthePUinthefirstphase.However,whentheselectedrelay

Sbisactiveinthesecondphase,itadjustsitstransmitpowerPb

incorrectlyandcausesaninterferencepowerIbin(30)tobelarger thanI.

Basedonthetotalprobabilitylaw,theinterferenceprobability

isgivenby

J=Pr

Is>I

+Pr

Is≤I,Ib>I

=Pr

Is>I

+

K

 b=1

Pr

Is≤I,Ib>I Sb is selected

Pr

Sb is selected

, (31) where the event {Sb is selected} is min(sb,bd)> min(sk,kd),∀k∈R\b.

Substituting{Is,Ib}in(30),{sb,sk}in(3),and{bd,kd}in

(6)into(31)resultsinanexpressionwithmanycorrelatedrandom variables.Therefore,itisimpossibletoobtainanapproximate/exact closed-formexpressionofJ.Consequently,weborrowsimulations

inthenextsectiontoinvestigatetheeffectofimperfectchannel informationontheprimarynetwork

5 Results and discussion

Thissectionprovidesnumerousresultstocorroboratethe valid-ityoftheproposedanalyticalexpressions,investigatetheimpact

ofICIontheoutagebehaviorofcooperativecognitivenetworks withopportunisticrelayselectionandontheinterference proba-bility,andhighlighttheadvantageofthedirectchannelinrelaying

Trang 6

Fig 2.Outage probability versusP.

communications.Thefadingpowerofthet−rchannelisassumed

as1/tr=d−˛tr wheredtristhedistancebetweenthetransmittert

andthereceiverrwhile˛isthepath-lossexponent.Withoutlossof

generality,wechoose˛=3andfixedusercoordinates:Ssat(0,0),

Sdat(1,0),Ppat(0.6,0.7),

Sk

5 k=1at{(0.4612,−0.1765),(0.4867,

−0.1373),(0.5182,0.0621),(0.5282,0.1780),(0.4686,−0.0431)}.In

thesequel,‘Ana.’,‘Sim.’,‘CC’,and‘DH’standfor‘Analysis’,

‘Simula-tion’,‘CooperativeRelaying’,and‘Dual-hopRelaying’,respectively;

three groups of relays are considered: K=1 for {S1}, K=3 for



Sk

3

k=1,K=5for

Sk

5 k=1

Fig 2 demonstrates the outage behavior of the

oppor-tunistic relay selection in both dual-hop and cooperative

cognitive networks with respect to the variation of P=P/N0

for =0.2 and R=1bit/s/Hz Two availability levels of

channel information are illustrated: PCI (i.e., tr=1, ∀{t,

r}) and ICI with randomly selected and mutually

dif-ferent correlation coefficients (sd=0.9144, sp=0.9231,



kd

5

k=1={0.9037,0.9107,0.9681,0.9007,0.9749},

kp

5

{0.9095, 0.9438, 0.9040, 0.9556, 0.9345}, 

sk

5

{0.9168,0.9627,0.9094,0.9662,0.9528}) It is observed that

theexact analysis(i.e.,(20)and (25))perfectlyagrees withthe

simulationforthewholerangeofPwhiletheperformancelimit

analysis(i.e.,(28)and(29))isinaperfectagreementwiththe

sim-ulationatlargevaluesofP(e.g.,P≥30dB),verifyingthevalidity

oftheproposedexpressions5.Inaddition,inthecaseofICI,both

cooperativeanddual-hopcognitivenetworkswithopportunistic

relayselectionsuffertheerrorfloorphenomenoninthehighSNR

regime, which is already discussed and analytically proved in

Section3.2.Inotherwords,opportunisticrelayselectiondoesnot

contributeanydiversitygaininthecaseofICI,whichiscontrastto

thecaseofPCIwherethediversityorderisalwaysnon-zero6.Tobe

morespecific,inthecaseofPCI,bothnetworksobtainanon-zero

5 As discussed in Section 3.2, PCI makes zero outage probability in the high SNR

regime and hence, no performance limit for this case is shown in Fig 2.

6 To see no error floor even at very low outage probabilities, we extend the plot

of Fig 2 for the case of PCI at high SNRs The results are illustrated in Fig 3 It is seen

that both cooperative and dual-hop cognitive networks with opportunistic relay

selection do not experience performance saturation phenomenon and the former

is superior to the latter, especially at high SNRs This comes from the fact that the

former has a higher diversity order than the latter It is also noted that for PCI, [16]

proposed the exact and asymptotic outage analysis in dual-hop cognitive networks

with opportunistic relay selection but assumed the statistical independence of terms

in  More specifically, [16] assumed that min( ,  ) is uncorrelated with

10−10

10−9

10−8

10−7

10−6

10−5

10−4

10−3

10−2

10−1

100

P (dB)

Sim.: K=1 & DH Exact Ana.: K=1 & DH Sim.: K=1 & CC Exact Ana.: K=1 & CC Sim.: K=3 & DH Exact Ana.: K=3 & DH Sim.: K=3 & CC Exact Ana.: K=3 & CC Sim.: K=5 & DH Exact Ana.: K=5 & DH Sim.: K=5 & CC Exact Ana.: K=5 & CC

Fig 3. Outage probability versusPfor perfect channel information.

diversityorderandtheachievablediversityorderofcooperative relayingislargerthanthatofdual-hoprelaying(i.e.,theoutage probabilitycurveoftheformerhasalargerslopethanthatofthe latter).Also,thesaturationleveloftheformerisconsiderablylower thanthat ofthelatter.Briefly, theformer issignificantly better than thelatterfor anysystemparameters underconsideration This observation shows the importance of utilizing the direct channelinrelayingcommunicationswithoutanyadditionalcost

ofsystemresources(e.g.,bandwidthandpower).Moreover,both networksaredrasticallydegradedbyICI,especiallyathighSNRs Nevertheless, their performance can be considerably improved withtheincreaseinthenumberofrelayssincethemorerelaysare available,thehigherchanceofselectingthebestrelayis

InthealmostsamecontextasFig.2exceptP=20dB,Fig.4

investigatestheimpactoftheoutagethreshold0(ortherequired transmission rate R) onthe outage performance of the oppor-tunisticrelayselectioninbothdual-hopandcooperativecognitive networks It is seen that the simulation perfectly matches the analysis,verifyingtheaccuracyoftheproposedexpressions Addi-tionally,theoutageperformanceofbothnetworksissignificantly deterioratedwithrespecttotheincreaseintheoutagethreshold

0=22R−1.Thismakessensebecausegivenoperationconditions, themorestringentthesystemperformancerequirement(i.e.,the largeroutagethreshold),thehigheroutageprobabilitythesystem suffers.Also,theperformanceoftheopportunisticrelayselection

isconsiderablyenhancedwithbetterqualityofchannel estima-tion(i.e.,fromICItoPCI),whichexposestheimportanceofchannel estimationincognitiveradionetworks.Moreover,similartoFig.2, increasing the number of involved relays can further improve systemperformance.Foralloperationparametersunder consider-ation,cooperativerelayingisalwaysbetterthandual-hoprelaying, whichisalreadyprovedinSection3

InthealmostsamecontextasFig.2exceptP=20dB,Fig.5

investigatestheeffectoftheproportionalfactorontheoutage performanceofbothdual-hopandcooperativecognitivenetworks

min( si ,  id ),∀(k, i) while this does not hold since both contain a common term

x = ˆh 2 , as discussed in Section 2.

Trang 7

0.5 1 1.5 2

10−4

10−3

10−2

10−1

100

R (bits/s/Hz)

Sim.: PCI Ana.: PCI Sim.: ICI Ana.: ICI

Color codes blue: K=1 & DH magenta: K=1 & CC green: K=3 & DH red: K=3 & CC cyan: K=5 & DH black: K=5 & CC

Fig 4. Outage probability versus R.

withopportunisticrelayselection.Itisseenthatthesimulationand

theanalysisareinanexcellentagreement,againconfirmingthe

validityoftheproposedexpressions.Additionally,theoutage

per-formanceofbothnetworksissignificantlyimprovedwithrespectto

theincreasein.Thisisreasonablebecausetheincreasein=I/P

isequivalenttotheincreaseinIandthus,inducingthePUmore

tolerablewiththeinterferencefromSUs.Therefore,SUscan

oper-atewithhightransmitpowers,eventuallymitigatingtheiroutage

probability.Moreover,theoutageperformanceofthe

opportunis-ticrelayselectionisconsiderablyenhancedwithbetterqualityof

channelestimationandthehighernumberofinvolvedrelays,

espe-ciallyathighSNRs.Furthermore,takingtheadvantageofthedirect

10−4

10−3

10−2

10−1

100

μ

Sim.: PCI Ana.: PCI Sim.: ICI Ana.: ICI Color codes

blue: K=1 & DH magenta: K=1 & CC green: K=3 & DH red: K=3 & CC cyan: K=5 & DH black: K=5 & CC

Fig 5. Outage probability versus .

10−6

10−5

10−4

10−3

10−2

10−1

100

ρ

Sim.: K=1 & DH Exact Ana.: K=1 & DH Sim.: K=1 & CC Exact Ana.: K=1 & CC Sim.: K=3 & DH Exact Ana.: K=3 & DH Sim.: K=3 & CC Exact Ana.: K=3 & CC Sim.: K=5 & DH Exact Ana.: K=5 & DH Sim.: K=5 & CC Exact Ana.: K=5 & CC

Fig 6. Outage probability versus .

channelalwaysimprovesthesystemperformanceforalloperation parametersunderconsideration

Itisrecalledthatthecorrelationcoefficienttrcontrolsthe qual-ityofthechannelestimatorandthelargertr,themoreaccurate theestimatedchannelinformation.Consequently,inorderto inves-tigatetheeffectofICIontheoutageperformanceofcooperative cognitivenetworkswithopportunisticrelayselection,weshould investigatetheoutageprobabilitywithrespecttotr.Withoutloss

ofgenerality,weassumealltr’stobeequal,i.e.tr=,∀{t,r}in

Fig.6,whichdemonstratestheoutageprobabilityasafunctionof forP=20dB,=0.2,R=1bit/s/Hz.Itisobservedthatthe simula-tionexcellentlymatchestheanalysis,againvalidatingtheaccuracy

oftheproposedexpressions.Inaddition,ICIsignificantlydegrades theperformanceofcognitiveradionetworks.Morespecifically,the systemisalwaysinoutageas<0.5,andaslightimprovementof estimatedchannelinformationaccuracy(e.g.,=0.9→1.0) signifi-cantlyreducestheoutageprobability(e.g.,PCC

o isreducedmorethan

104timesforK=5).However,theperformancedegradationdueto ICIcanbecomplementedbyincreasingthenumber ofinvolved relays.Moreover,cooperativecognitivenetworksaremorerobust

toICIanddrasticallyoutperformsdual-hopcounterpartforany systemparameters

10−3

10−2

10−1

100

I (dB)

K=1 K=3 K=5

Fig 7.Interference probability versusI.

Trang 8

variationofI=I/N0forP=10dB,R=1bit/s/Hz,andICIwith

ran-domly selected and mutually different correlation coefficients

(

kd

5

k=1={0.9037,0.9107,0.9681,0.9007,0.9749},



kp

5

k=1={0.9095,0.9438,0.9040,0.9556,0.9345},

sk

5

{0.9168,0.9627,0.9094,0.9662,0.9528}, sd=0.9144, sp=

0.9231).Itisrecalledfrom(31)thatbothdual-hopandcooperative

relaying schemes results in the same interference probability

since (31) is independent of the direct channel Therefore, the

interferenceprobabilityinthisfigurerepresentsforbothschemes

ItisseenthattheincreaseofIreducestheinterferenceprobability

ThisisreasonableinthesensethatthelargerI,themore

inter-ferencepowerthePUcantolerate.Therefore,givenothersystem

parameters,theprobabilitythattheinterferencepowerexceedsI

decreases.However,theinterferenceprobabilityisproportionalto

thenumberoftherelays.Inotherwords,thequalityofservicein

theprimarynetworkisdegradedwithrespecttotheincreaseinthe

numberoftherelays.Thisconflictswiththeoutageperformance

trendinthesecondarynetworkwheretheoutageperformanceis

improvedwithrespecttotheincreaseinthenumberoftherelays

Asaresult,thereisaperformancetrade-offbetweentheprimary

networkandthesecondarynetworkwithrespecttothenumber

ofrelaysundertheconditionofimperfectchannelinformationon

allchannelsconcurrently

6 Conclusions

Thispaperproposesanexactandlimitoutageanalysis

frame-workforcooperativecognitivenetworkswithopportunisticrelay

selectionunderageneralscenario:imperfectchannelinformation

forallchannelsconcurrently,i.n.i.Rayleighfadingchannels,and

bothmaximumtransmitpowerconstraintandinterferencepower

constraint.Thisframeworkisstraightforwardlyextended tothe

correspondinganalysisindual-hopcognitivenetworkswith

oppor-tunisticrelayselectionforcomparisonconvenienceandemphasis

oftheimportanceofthedirectchannelwithouttime-consuming

simulations.Numerousresultsdemonstratethat(i)channel

infor-mationimperfectionsignificantlyimpactstheoutageperformance,

especially forhighSNRsand smallnumberof relays;(ii)

relay-ingcognitivenetworksexperienceperformancesaturationathigh

SNRsandthesaturationlevelonlydependsonthequalityof

chan-nelestimator;(iii)theopportunisticrelayselectioninthecaseof

imperfectchannelinformationdoesnotbringanydiversitygainfor

bothcooperativeanddual-hopcognitivenetworks;(iv)increasing

thenumberofrelays candramaticallyimprovetheoutage

per-formanceirrespectiveofchannelinformationimperfectiondegree

butalsodegradethequalityofserviceofprimaryusers;(v)the

outageperformanceofrelayingcognitivenetworksisconsiderably

enhancedwithtakingadvantageofthedirectchannelwithoutany

significantcostofsystemresources(e.g.,powerandbandwidth)

Acknowledgement

Thisresearchis funded by VietnamNationalFoundation for

Scienceand Technology Development (NAFOSTED) under grant

number102.04-2014.42

References

[1] FCC Spectrum policy task force report ET Docket 02-135; 2002.

[2] Mitola III J Cognitive radio: an integrated agent architecture for software

defined radio Sweden: Royal Institute of Technology (KTH); 2000.

[3] Laneman JN, Tse DNC, Wornell GW Cooperative diversity in wireless

networks: efficient protocols and outage behavior IEEE Trans Inf Theory

2004;50(12):3062–80.

[4] Bletsas A, Khisti A, Reed DP, Lippman A Simple cooperative diversity method

[5] Si J, Huang H, Li Z, Hao B, Gao R Performance analysis of adaptive modula-tion in cognitive relay network with cooperative spectrum sensing IEEE Trans Commun 2014;62(7):2198–211.

[6] Chamkhia H, Omri A, Bouallegue R The impact of imperfect channel state infor-mation on the performances of relay selection schemes in underlay cognitive networks In: Proc IWCMC 2014 p 174–9.

[7] Sun H, Pour MN Decode-and-forward relay selection with imperfect CSI in cognitive relay networks In: Proc IEEE MILCOM 2014 p 416–21.

[8] Suraweera HA, Smith PJ, Shafi M Capacity limits and performance analysis of cognitive radio with imperfect channel knowledge IEEE Trans Veh Technol 2010;59:1811–22.

[9] Zhong B, Zhang X, Li Y, Zhang Z, Long K Impact of partial relay selection on the capacity of communications systems with outdated CSI and adaptive transmis-sion techniques In: Proc IEEE WCNC 2013 p 3720–5.

[10] Chen J, Si J, Li Z, Huang H On the performance of spectrum sharing cognitive relay networks with imperfect CSI IEEE Commun Lett 2012;16:1002–5 [11] Ho-Van K, Sofotasios PC, Freear S Underlay cooperative cognitive networks with imperfect Nakagami-m fading channel information and strict transmit power constraint: interference statistics and outage probability analysis J Com-mun Netw 2014;16(1):10–7.

[12] Khuong HV, Sofotasios PC Exact bit-error-rate analysis of underlay decode-and-forward multi-hop cognitive networks with estimation errors IET Commun 2013;7(18):2122–32.

[13] Bao VNQ, Duong TQ, Chintha T On the performance of cognitive underlay mul-tihop networks with imperfect channel state information IEEE Trans Commun 2013;61(12):4864–73.

[14] Khuong HV Impact of imperfect channel information on the perfor-mance of underlay cognitive DF multi-hop systems Wirel Pers Commun 2014;74(2):487–98.

[15] Khuong HV Exact outage analysis of underlay cooperative cognitive networks with maximum transmit-and-interference power constraints and erro-neous channel information Trans Emerg Telecommun Technol 2013;24(7-8): 1–17.

[16] Lee J, Wang H, Andrews JG, Hong D Outage probability of cognitive relay networks with interference constraints IEEE Trans Wirel Commun 2011;10(2):390–5.

[17] Jing T, Zhu S, Li H, Xing X, Cheng X, Huo Y, Bie R, Znati T Cooperative relay selection in cognitive radio networks IEEE Trans Veh Technol 2015;64(5): 1872–81.

[18] Xing Z, Yan Z, Zhi Y, Jia X, Wenbo W Performance analysis of cognitive relay networks over Nakagami-m fading channels IEEE JSAC 2015;33(5): 865–77.

[19] Chen Y, Ge J, Bu Q Outage and diversity analysis of cognitive relay networks with direct link under interference constraints over Nakagami-m fading In: Proc IEEE CIT 2014 p 88–93.

[20] Hanif M, Yang HC, Alouini MS Receive antenna selection for underlay cogni-tive radio with instantaneous interference constraint IEEE Signal Process Lett 2015;22(6):738–42.

[21] Li D Opportunistic DF-AF selection for cognitive relay networks IEEE Trans Veh Technol 2015 [in press].

[22] Mokari N, Parsaeefard S, Saeedi H, Azmi P, Hossain E Secure robust ergodic uplink resource allocation in relay-assisted cognitive radio networks IEEE Trans Signal Process 2015;63(2):291–304.

[23] Si J, Li Z, Huang H, Chen J, Gao R Capacity analysis of cognitive relay networks with the PU’s interference IEEE Commun Lett 2012;16(12):2020–3 [24] Si J, Li Z, Chen X, Hao BJ, Liu ZJ On the performance of cognitive relay networks under primary user’s outage constraint IEEE Commun Lett 2011;15(4): 422–4.

[25] Guimaraes FRV, da Costa DB, Tsiftsis TA, Cavalcante CC, Karagiannidis GK Multi-user and multi-relay cognitive radio networks under spectrum sharing constraints IEEE Trans Veh Technol 2014;63(1):433–9.

[26] Tourki K, Qaraqe KA, Alouini MS Outage analysis for underlay cognitive networks using incremental regenerative relaying IEEE Trans Veh Technol 2013;62(2):721–34.

[27] Hong JP, Hong B, Ban TW, Choi W On the cooperative diversity gain in underlay cognitive radio systems IEEE Trans Commun 2012;60(1):209–19.

[28] Yan Z, Zhang X, Wang W Exact outage performance of cognitive relay networks with maximum transmit power limits IEEE Commun Lett 2011;15(12): 1317–9.

[29] Liping L, Zhang P, Zhang G, Qin J Outage performance for cognitive relay networks with underlay spectrum sharing IEEE Commun Lett 2011;15(7): 710–2.

[30] Bao VNQ, Duong TQ Exact outage probability of cognitive underlay DF relay networks with best relay selection IEICE Trans Commun 2012;E95-B(6):2169–73.

[31] Zhang X, Yan Z, Gao Y, Wang W On the study of outage performance for cogni-tive relay networks (CRN) with the Nth best-relay selection in Rayleigh-fading channels IEEE Wirel Commun Lett 2013;2(1):110–3.

[32] Asaduzzaman, Kong HY, Lyum K Cooperative relaying in interference limited cognitive radio networks In: Proc IEEE WiMob 2010 p 280–5.

[33] Zhong B, Zhang Z, Chai X, Pan Z, Long K, Cao H Performance analysis for oppor-tunistic full-duplex relay selection in underlay cognitive networks IEEE Trans Veh Technol 2015 [in press].

[34] Guimaraes FRV, da Costa DB, Benjillali M, Tsiftsis TA, Karagiannidis GK Best relay selection in cooperative spectrum sharing systems with multiple primary users In: Proc IEEE ICC 2013 p 2661–5.

Trang 9

[35] Chang CW, Lin PH, Su SL A low-interference relay selection for

decode-and-forward cooperative network in underlay cognitive radio In: Proc IEEE

CROWNCOM 2011 p 306–10.

[36] Zhang Y, Xie Y, Liu Y, Feng Z, Zhang P, Wei Z Outage probability of

cogni-tive relay network with transmit power and interference constraints In: Proc

WPMC 2012 p 1–5.

[37] Liu Y, Wang L, Duy TT, Elkashlan M, Duong TQ Relay selection for

secu-rity enhancement in cognitive relay networks IEEE Wirel Commun Lett

2015;4(1):46–9.

[38] Sakran H, Shokair M, Nasr O, El-Rabaie S, El-Azm AA Proposed relay selection

scheme for physical layer security in cognitive radio networks IET Commun

2012;6(16):2676–87.

[39] Zhong B, Zhang Z, Zhang X, Wang J, Long K Partial relay selection with

fixed-gain relays and outdated CSI in underlay cognitive networks IEEE Trans Veh

Technol 2013;62(9):4696–701.

[40] Chamkhia H, Hasna MO, Hamila R, Hussain SI Performance analysis of relay

selection schemes in underlay cognitive networks with decode and forward

relaying In: Proc IEEE PIMRC 2012 p 1552–8.

[41] Kim KJ, Duong TQ, Tran XN Performance analysis of cognitive

spectrum-sharing single-carrier systems with relay selection IEEE Trans Signal Process

2012;60(12):6435–49.

[42] Giang NH, Bao VNQ, Le HN Cognitive underlay communications with

imper-fect CSI: network design and performance analysis In: Proc IEEE ATC 2013.

p 18–22.

[43] Seyfi M, Muhaidat S, Liang J Relay selection in cognitive radio networks with

interference constraints IET Commun 2013;7(10):922–30.

[44] Minghua X, Aissa S Cooperative AF relaying in spectrum-sharing systems: out-age probability analysis under co-channel interferences and relay selection IEEE Trans Commun 2012;60(11):3252–62.

[45] Ding H, Ge J, da Costa DB, Jiang Z Asymptotic analysis of cooperative diversity systems with relay selection in a spectrum-sharing scenario IEEE Trans Veh Technol 2011;60:457–72.

[46] Zhang X, Xing J, Yan Z, Gao Y, Wang W Outage performance study of cog-nitive relay networks with imperfect channel knowledge IEEE Commun Lett 2013;17(1):27–30.

[47] Tourki K, Qaraqe KA, Abdallah M Outage analysis of incremental opportunistic regenerative relaying with outdated CSI under spectrum sharing constraints In: Proc IEEE WCNC 2014 p 851–6.

[48] Thanh TL, Bao VNQ, An B On the performance of outage probability in underlay cognitive radio with imperfect CSI In: Proc IEEE ATC 2013.

p 125–30.

[49] Wu Q, Zhang Z, Wang J Outage analysis of cognitive relay networks with relay selection under imperfect CSI environment IEEE Commun Lett 2013;17(7):1297–300.

[50] Haiyan H, Zan L, Jiangbo S, Lei G Underlay cognitive relay networks with imper-fect channel state information and multiple primary receivers IET Commun 2015;9(4):460–7.

[51] Ahn KS, Heath RW Performance analysis of maximum ratio combining with imperfect channel estimation in the presence of cochannel interferences IEEE Trans Wirel Commun 2009;8(3):1080–5.

[52] Vucetic B, Yuan J Space-time coding John Wiley & Sons, Inc.; 2003.

...

chan-nelestimator;(iii)theopportunisticrelayselectioninthecaseof

imperfectchannelinformationdoesnotbringanydiversitygainfor

bothcooperativeanddual-hopcognitivenetworks;(iv)increasing

thenumberofrelays candramaticallyimprovetheoutage... qual-ityofthechannelestimatorandthelargertr,themoreaccurate theestimatedchannelinformation.Consequently,inorderto inves-tigatetheeffectofICIontheoutageperformanceofcooperative cognitivenetworkswithopportunisticrelayselection,weshould...

frame-workforcooperativecognitivenetworkswithopportunisticrelay

selectionunderageneralscenario:imperfectchannelinformation

forallchannelsconcurrently,i.n.i.Rayleighfadingchannels,and

Ngày đăng: 16/12/2017, 06:30

TỪ KHÓA LIÊN QUAN

🧩 Sản phẩm bạn có thể quan tâm