ArchitecturesandProtocolsforDynamicSpectrumSharinginHeterogeneousWirelessAccessNetworks
OliverHolland1,AlirezaAttar1,MaheshSooriyabandara2,TimFarnham2,HamidAghvami1,MarkusMuck3,andVladimirIvanov4,andKlausNolte5
1
2
3
4
5
CentreforTelecommunicationsResearch,King’sCollegeLondon,UK{Oliver.Holland,Ali.Attar,Hamid.Aghvami}@kcl.ac.ukToshibaResearchEuropeLtd.,Bristol,UK
{Mahesh.Sooriyabandara,Tim.Farnham}@toshiba-trel.com
MotorolaLabs,Gif-sur-Yvette,France(MarkusMuckhassincemovedtoInfineonTechnologies,Munich,Germany)markus.muck@gmail.com
IntelCorporation,CommunicationsTechnologyLab,St.Petersburg,Russiavladimir.ivanov@intel.com
Alcatel-LucentDeutschlandAG,BellLabs,Germanyklaus.nolte@alcatel-lucent.de
3.1Introduction
Theexplosivegrowthofcapacity-hungryapplicationsandtheincreasingprolifera-tionofmobileandwirelessdevicesareprovingchallengingforthewirelessindustry.Worldwide,almostallfrequencybandsoffavorablepropagationcharacteristicsforwirelesscommunicationsarecurrentlyallocatedtoatleastoneRadioAccessTech-nology(RAT)(see,e.g.,[1]–[4]);moreover,onlyatinyproportionofthatspectrumisallocatedtomobilecommunicationsandwirelessnetworkingmeans.Thisleadstoanincreasingrequirementforbit-rateperunitspectrumamongthesystemsanddevicesaccessingit.Thischapterfocusesonmeansfordynamicspectrumsharingtoenhancespectrumusageefficiency,investigatingvariousarchitecturesandprotocolsfromanumberofdifferentperspectives.
Spectrumsharingamongwirelesssystemscanoccureitherhorizontallyorverti-cally[5].Ifallradioshavethesamerighttoaccessaparticularband,asisthecaseinunlicensedbands,horizontalspectrumsharingisimplied.Ontheotherhand,ifonesystem(typically,theownerofthelicensetotheband)hasahigherprioritytoaccessthebandthanothersystems,theimplicationisverticalspectrumsharing.Generally,verticalspectrumsharingisenvisagedthroughsecondaryspectrumaccess,wherebythelicensedbandissharedbutonlywiththelicenseowner’sconsent,betweenthelicenseowner’snetwork,i.e.,theprimarysystem,andlowerprioritynetworks,i.e.,thesecondarynetworks.Meansforverticalspectrumsharingarethemainemphasisofthischapter.
E. Hossain (ed.), Heterogeneous Wireless Access Networks,
DOI: 10.1007/978-0-387-09777-0_3, © Springer Science+Business Media, LLC 2008
56O.Hollandetal.
Technicallyspeaking,spectrumcanbesharedintherealmsoftime,space,fre-quency,oracombinationoftheabove.Spectrumsharingintimeinvolvessystemsusingabandatdifferenttimeintervals,forexample,byexploitingalisten-before-talketiquette.Anotherpossibilityhere,whichiscomplicatedbythefactthatitrequiressynchronizationamongsystems,istouseidletime-slotsinaTimeDivisionMultipleAccess(TDMA)context.Spectrumsharinginthefrequencydomaininvolvestrans-mittingsuchthatnofrequencyoverlapoccursamongthecoexistingsystems.Inmanysuchcases,itmightbenecessaryforasystemtoconsolidateseveralsmalleridlefre-quencybandstocreateatransmissionopportunity,throughusingmulti-carriermod-ulationapproachessuchasOrthogonalFrequencyDivisionMultiplexing(OFDM).Spectrumsharinginthespatialdomainmightbeusedtobenefitfromtherealizationthatspectrumoccupancyvariesfromlocationtolocation.Henceasystemmightuseabandinaspecificlocationifothersystemsarenotpresentatthatlocation.
Transmissionopportunitiesarisingfromspectrumsharinginthedomainsoftime,space,frequency,oracombinationofthese,aregenerallycalledspectrumholes,orsometimes“whitespaces”[6].Spectrummightalsobesharedinthepowerdomainhowever.Forexample,asystemmightbepermittedtotransmitwithahigherpowerinanyoftheabovedomains(time,space,frequency,oranycombination),providingthatthesetransmissionsdonotbreachtheinterferencetolerancethresholdatothersystems’receivers.Transmissionopportunitiesinthiscontextareknownas“greyspaces”[6].Theimposedinterferencetoeachreceiverheremightbeminimizedbyusingspread-spectrumtechniques,forexamplethroughUltraWideBand(UWB)orthroughverywidebandCodeDivisionMultipleAccess(CDMA).Suchmethods,comprisingopenaccesstoothersystems’bandswithextremelylowtransmissionpowerspectraldensitythroughspreadspectrumtechniques,areknownasunderlaytransmission[7].
Anotherwaytoclassifyspectrumsharingtechniquesisbasedontheconsider-ationofwhethertheutilizedbandsareexpresslyallocatedorare(semi-)indepen-dentlychosenbydevices.Intheformercase,spectrumamongoperators,orpossiblyfromaspectrumpool,mightbedynamicallycoordinatedandallocatedtoRadioAc-cessNetworks(RANs)astriggeredbytrafficdemands.ThismethodisknownasDynamicSpectrumAllocation(DSA).Theotherapproachisthatofdevices/RANsselectingresources(semi-)independently,usuallyasasecondaryentityinthespec-trum.Thiscanbeeitherfullydecentralizedornetwork-assisted,andisoftenenvis-agedasbeingachievedthroughCognitiveRadio(CR)[6],[8].Severalterminologiesandapproachestothissharingparadigmmightbeidentifiedintheliterature,suchasDynamicSpectrumAccess(DSA)[9]andOpportunisticSpectrumAccess(OSA).TheumbrellatermDynamicSpectrumSelection(DSS)isemployedthroughoutthischapter.
3.1.1DemandsofApplicationsandServices
Amajorcauseofthespiralingspectrumdemandinrecentyearsistheemergenceofnovelwirelesslynetworkedapplicationsandservicessuchasvideostreaming,video
3ArchitecturesandProtocolsforDynamicSpectrumSharing57
telephony,conferencing,highcapacitydownloadsandpeer-to-peertransfers,de-mandingdatastreamingapplications(e.g.,GoogleEarth),andevermorecomplicatedwebsiteswithincreasedactivecontentrequiringgreatercapacity.Anothercauseofspectrumdemandistheincreasedproliferationandtake-upofvariouswirelesstech-nologiesamongthegeneralpublic.Unfortunately,asstatedbyShannon’sTheorem,thelawsofphysicsspecifyalimitontheamountofdatathatcanbetransmittedforagiventransmissionpowerandnoise(or,underspecificassumptions,noiseplusinter-ference)inafixedspectrumbandwidth.Hencetosatisfythedemandsofpioneeringapplicationsandservices,aswellastheincreasingdensityoftransceivers,theneedforgreaterspectrumusageefficiencyisultimatelyunavoidable.
Whereasinunlicensedspectrumsystems/devicesmightpreviouslyhavebeenreasonablyhappycontendingfortheavailablespectruminahaphazardmanner,newerservicessuchasVoice-over-IPandvideoconferencing,andotherservicessuchashighprioritydata,requireabetterguaranteeofQualityofService(QoS).Toimproveserviceerrorrateandreduceschedulingdelay,thisoftenimpliestheassur-ance,withaveryhighprobability,thatspectrumwillbeavailablewhenaskedforbythesysteminquestion.Theneedforsuchanassurancefurtherexacerbatesspectrumbandwidthneeds.
3.1.2InefficienciesofSpectrumUsageintheModern-DayWorld
Thespectrumallocationprocessasdeterminedbyregulatorsworldwidehashistori-callybeenultra-conservative.EachfrequencybandhasbeenallocatedtoonlyoneorasmallnumberofRATs,andonlythoseRATsareallowedtoaccessthespectrum.Thishascommonlyledtoasituationwherethespectruminquestionisused,onaverage,atonlyatinyproportionofitstheoreticalcapacityinanyonegeographicallocation.
Evenwithinthecontextofsystemsservingrelativelysimilarpurposes,betheyusingthesamestandardoralternativestandards,thereareoftensignificantdiffer-encesintheloadingsofthesesystemsandtheirassociatedspectrumatanyonetime.Throughrealizingspectrumsharing,spectrallylocalizedincreasesintrafficloadscanbedissipatedacrossmuchlargerspectrumranges.Theendresultofsuchaparadigmwouldbeafargreateraveragespectrumutilizationacrossmultiplebands,hencevastlyimprovedQoSforarangeofwirelessusers.Thisisareasoningbehindthegloballyincreasedinterestedinspectrumsharingsolutions.
3.1.3IncreasedFreedomsandTechnologicalAdvancesFacilitatingSpectrumSharing
Inthelightofthisinterestinspectrumsharinganditspotential,industryhasbeenlookingatvarioussolutionsforit.OnesignificantfacilitatorforspectrumsharingfromatechnicalpointofviewisSoftware-DefinedRadio(SDR),whichallowscharacteristicsofradiointerfacestobedefinedinsoftwaretherebyprovidingeas-ieradaptationtoRATssuchasmightapplyinadifferentspectrumbandormightbe
58O.Hollandetal.
moreappropriatefortheradiocharacteristicsorcapabilitiesofsystemsinaprospec-tiveband.Anotherfacilitatoris“Reconfigurability”,theobjectivebeingtosimplifyadaptationstoarangeoflayersoftheprotocolstack,includingtheassociatedradionetworkconfigurationchangesandRATadaptations.Givensuchtechnologicalad-vancesintandemwithincreasedinterestinspectrumsharing,regulatorshavebeguntolookmoreseriouslyatmakingspectrumallocationsandusageslessviscous.Theseandsimilarsuchadvancesessentiallybegthequestionof“when”not“if”spectrumsharingparadigmswillbecomemoreofareality.
Thepurposeofthischapteristoinvestigatethetechnicalrequirements,usagescenarios,architecturesandprotocolsforspectrumsharingintheDSSandDSAcontexts.Thissupplementsseveralcontributionsthatpresentgeneralsurveysinthefieldofspectrumsharing(see,e.g.,[6],[7]and[9]).Thischapterisorganizedasfollows.InSection3.2,methodsforspectrumsharinginthecontextofcentralizedcontrol,throughDynamicSpectrumAllocation(DSA),arediscussed.Section3.3investigatesdecentralizationofspectrumsharingdecisionsthroughDynamicSpec-trumSelection(DSS).Section3.4discussesGameTheoryasameansforresourceoptimizationinspectrumsharingscenarios,beforethechapterconcludes.
3.2CentralizedControl:DynamicSpectrumAllocation
ThefirstclassificationofspectrumsharingsolutionsinvestigatedisDynamicSpec-trumAllocation(DSA)whichgenerallyimpliescentralizedcontrolintheallocationofresourcesbyanentityorcollaboratinggroupofentities.3.2.1BasicConcept
DSAaimstomanagespectruminamulti-radioenvironmentbydynamicallyallo-catingitamongparticipatingnetworksovertime,space,orboth.Thiscanservetoexploitbothtemporalandspatialvariationsinspectrumrequirementsamongnet-worksinordertoachievehigheroverallspectrumusageefficiency.ConsidertheexamplescenarioinFig.3.1,depictingnormalizedtrafficloadsoftwonetworksatthesamelocationina24hourperiod.Notethateachnetworkoperatesinitsownlicensedband.Threeregionscanbeidentifiedinthisfigure.Inregion1,bothnet-worksencounterarelativelyhightrafficload,hencespectrumsharingbetweenthemwouldnotbeadvantageous.Inregion2,network2isfacingahightrafficloadwhilenetwork1isrelativelyidle.Thereforetheloadonnetwork2couldbealleviatedbydynamicallyallocatingsomeoftheidlebandsofnetwork1tonetwork2.Finally,inregion3,bothnetworksareoperatingwithalowtrafficdemand,hencealthoughidlespectrumexistsnogainwouldbeachievedthroughspectrumsharing.Notethatthesamescenarioisvalidifthehorizontalaxisisspaceinsteadoftime.
ThevalueofautilizedDSAprocesscanbemeasured,atleastpartially,bytheDSAgain.TheDSAgainisgivenbytheaverageloadincreasesupportedoverpar-ticipatingnetworkswhenDSAisemployed,ascomparedwiththeaverageloadsup-portedinfixedspectrumallocation.Asimpliedbytheexamplediscussedabove,the
3ArchitecturesandProtocolsforDynamicSpectrumSharing
59
1normalized traffic loadnetwork 1network 200region 1region 2time (hours)region 324Fig.3.1.Exampleoftemporalvariationintrafficloadsfortwoco-locatednetworks.
DSAgaindepends,amongotherfactors,onthecorrelationintrafficdemandsamongparticipatingnetworks.AlowercorrelationinthetrafficdemandsgenerallyindicatesahigherDSAgain.
3.2.2ArchitecturesandProtocolsforDSA
ThemainphasesofDSAoperation,whichareexecutedinacyclicfashion,areillus-tratedinFig.3.2.TheassociatedapproachestoDSA,withreferencetothesephases,arediscussedinthissection.TheemphasisisparticularlyonDSAtosatisfyfluctu-ationsintrafficrequirements;however,itisnotedthatinfuturemanifestationstheremaybeotherconceivablereasonsforspectrumallocationstobereassessed.
TheevolutionarystepsofDSAliquidityarecapturedinFig.3.3,asinspiredby[10].Fig.3.3(a)representsastaticspectrumallocationcase,whereclearlynoDSAistakingplace.Thisiscomparabletothepresentrealityintermsofspec-trumusageamongsystems.Fig.3.3(b)and(c)representthecontiguousandfrag-mentedDSAschemes,respectively,inbothcasesofwhichspectrumsharingistak-ingplaceattheRANlevel.Finally,Fig.3.3(d)representscell-by-cellDSA,wherespectrumissharedbothinthetemporaldomainandwithahighresolutioninthespa-tialdomain.ThecomplexityinachievingDSAincreasesfromFig.3.3(a)throughtoFig.3.3(d),wherebyDSAarchitecturesareanticipatedtoprogressivelyaspireto-wardsFig.3.3(d).Thisistheultimateobjective,presentingmostDSAgain.
StrategiestoaddressDSAinvolvejointnetworkcoordinationandmanagementofresourcesamongmultipleoperators/systemsand/orwithinmeta-operators.InthecaseofDSAbeingindependentlyappliedthrougharrangementsbetweendistinctentities(e.g.,networks),theexactarchitecturesmightoftenbeindependentlydevel-opedbythoseparticipatingentities.Theproceduresinthischapter(e.g.,Fig.3.2)are
60O.Hollandetal.
TrafficPredictionResourceRequestSignalingResourceAllocationCoordinationFig.3.2.PhasesofDSAoperation.
(a)(b)(c)(d)
Fig.3.3.Acomparisonofspectrumallocationapproaches:(a)fixedallocation,(b)contiguousDSA,(c)(highly!)fragmentedDSA,and(d)cell-by-cellDSA.NotethatthisfiguremightalternativelyberepresentedintermsofRANs(eitherofthesameRATordifferentRATs–see[10]).
stillnecessary,andofcoursethesharingprocessmustbeconsistentwithlocalregu-lationsfortheuseoftheassociatedspectrum.Somespecificationsapplicabletosuchcasesdoneverthelessexist:oneexampleofthisisthearchitectureandfunctionalrequirementsforsharednetworksina3Gcontext[11],inwhichsharingisspeci-fiedbetweenoperatorsattheRANorCore-Networklevel.Throughthisapproach,multipleoperatorscanmaintainindependenceamongtheirnetworkinfrastructures.
Thingsaremorecomplicatedinthecaseofspectrumbeingpooledbynet-workoperatorsforDSApurposes,whereofcourseawidelyunderstoodarchitec-ture/specificationforthesharingprocessisneeded.AnillustrativearchitectureforDSAinthisscenarioisshowninFig.3.4,wherebysharedspectrumisavailabletoallnetworks/operatorsviaaCentralizedSpectrumController(CSC),whichdy-namicallygrantsResourceAllocationstothenetworks/operatorsbasedonpredictedtrafficloadsandsubmittedResourceRequests.AssociatedwiththisprocessarethecomputationalandsignalingoverheadsforthecoordinationandcontrolofDSA,
3ArchitecturesandProtocolsforDynamicSpectrumSharing61
includingtheissuessuchastrafficpredictiontoappropriatelyidentifyrequiredre-sources(seelatersections).Notealsothattrafficpredictionmightbeamulti-stageprocess;e.g.,devicesmightinformthenetworkabouttheir(perhapsanticipated)trafficrequirements(e.g.,ananticipateddownloadbyanapplicationforanupgrade,orananticipatedstreamingeventthattheuserhassignedupforinadvance),andthenetworkmighttwinthiswithitsownknowledgeoftrafficrequirementsbasedonexperiencedloads(usingmechanismssuchasmentionedinthenextsection).Giventheserequirements,inadditiontootherssuchasperformingnetworkanddevicere-configurationstothenewspectrumconfiguration,DSAiscommonlyenvisagedasbeingperformedwitharelativelylowtime-resolution,atperiodicintervals.Real-timeDSA,wherethereisthefreedomtoupdateallocationsatanyinstance,presentsanumberofsignificantchallengeshenceisleftoutofscopeofthischapter.
Fig.3.4.ExampleDSAarchitectureinapooledspectrumenvironment.
TrafficPredictionforDSA
AfirstimportantoperationforDSA,asillustratedinFig.3.2,isthepredictionoffuturetrafficrequirementsbynetworks’RSCs.ThispredictionisnecessaryfortheCSCtodeterminewhichdynamicspectrumreallocationswouldbeappropriatetosatisfyloadvariationsamongnetworks,andwhethertheprocesses,workloads,andassociatedtemporarydetrimentaleffectsonnetworksandtheirdevices(e.g.,duetosignalingandreconfigurationrequirements)involvedinundergoingaspectrumreallocationwouldbejustifiedbylongertermgains.Twochoicesdeterminetheper-formanceofatrafficpredictionmethod:thechoiceofestimationparameters,and
62O.Hollandetal.
thechoiceofestimationalgorithm.Themostcommonestimationparametersaretheaverageandpeaktrafficrates[12];associatedestimationalgorithmsarediscussedasfollows.
Futuretrafficratesaregenerallypredictedbasedonasetofpastobservedrates.TwoapproachestoestimatefuturetrafficaretheAutoregressive(AR)andMovingAverage(MA)models[13].LetY={Yt|t∈T}denotethepredictedtrafficseriesandX={X1,X2,···}denotethesetofpastobservedvalues,wheretisthetimeindex.TheARestimateoftheseriesY,basedonppastobservedvalues,isdefinedby,
Yt=
pi=1
φiXt−i+Wt
wheretheφi’saretheparametersofthemodel,andWtistheerrorterm,commonly
takenfromaNormaldistributionwithzeromeanandvarianceσ2.TheARparame-tersφcanbecalculatedusingYule-Walkerequations[13],
γm=
pk=1
φkγm−k+σ2δm
whereγm=E[XtXt−m]istheauto-correlationfunctionofX,andδmisanimpulse
function.
TheMAestimateoforderqisdefinedby
Yt=Wt+
qi=1
θiWt−i
whereθarethemodelparameters(the“weights”ofthemovingaverage)andWare
i.i.d.errorsamples,takenfromaNormaldistributionwithzeromeanandvarianceσ2asbefore.
AmoregeneralapproachistheAutoregressiveMovingAverage(ARMA)model[13], whichcombinestheARandMAmodelstoproduce
Yt=Wt+
pi=1
φiXt−i+
qi=1
θiWt−i.
Alloftheabove-mentionedmodelsassumeergodicityandstationarityforthe
estimatedvalues.MoreadvancedgeneralizationssuchastheAutoregressiveInte-gratedMovingAverage(ARIMA)andFractional-AutoregressiveIntegratedMovingAverage(F-ARIMA)modelscanbeusedtoestimatenon-stationaryprocesses[14].F-ARIMA,andsometimesFractionalGaussianNoise(FGN),aregoodforestimat-ingtrafficifitexhibitsself-similarity[15].
Theuseoflinearornon-linearregressiontechniquesintheformofextrapolationand/orinterpolationisalsocommonpracticefortrafficprediction(see,e.g.,[12]).Thegeneralequationforlinearregressioncanbewrittenas
3ArchitecturesandProtocolsforDynamicSpectrumSharing63
Yt=α+βXt+Wt
whereαistheintercept,βistheslope,andWtistheerrorterm.Toestimatetheinter-ceptandslope,differentapproachessuchasLeastSquare,MaximumLikelihood,orBayesianmethodsmightbeused[13].Forinstance,usingpsamplesforestimation,undertheLSapproachtheaforementionedparametersarecalculatedas
(β=
p−1i=0
¯)(Yt−i−Y¯)Xt−i−X
p−1i=0
p−1i=0
¯)2(Xt−i−X
and
¯−βX¯α=Y
¯andY¯aretheaveragevaluesofXiandYi(i=t,t−1,···,t−p+1).UnderwhereX
non-linearregressiontechniques,asthenameimplies,thevaluesofYtareestimatedfromasubsetofobservedvariablesusinganon-linearfunction.Oneexampleofthisisexponentialregression[13].
Afurthertrafficestimationapproachistotakeadvantageofhistoricalpatternsinthetrafficthroughusingpastexperiencedtrafficloadsatcertaintimesofthedayoratcertaintimesoftheweek/yeartopredictexpectedloadatthattimeofday,weekoryear[12].Inadditiontobeingusedasastand-aloneprocess,thismightbecombinedwiththeabovestatisticalmethods,forexamplethroughaweightingmechanismamongtheresultsproducedbythem.Anotherapproachheremightbetocombinethehistoricalseriesofstatisticsfromcertainperiods(e.g.,timeseriesfromintervalsinpastbusyhoursforthesamedayofeachweek,perhapsthroughanaveragingmethod)andusethesecombineddatasets,fedintotheabovestatisticalmethods,toobtainmorepreciseestimatesoffuturerequirementsforcomingDSAintervals.Finally,itmustalsobenotedthatinformationaboutfutureeventsassistingtrafficpredictionmightcomefromterminalsthemselves,orfromotherentities,e.g.,throughsignalingaboutsubscribedservicesinfuturetimeintervals.Examplesofsuchservicesmightbeastreamedsportingevent,oraplannedoperatingsystemupgradedownload.
Clearly,thereisatradeoffbetweentheaccuracyandcomplexityoftrafficpre-dictiontechniques;moreover,sincedifferentDSAscenariosneedtomeetdifferentrequirements,itisnotpracticaltoselectoneapproachasbeingoptimalforallsce-narios.Somecomparativestudiesofperformanceandcomplexitycoveringalimitednumberofselectedtrafficestimationapproachesareavailableintheliterature(see,e.g.,[12],[16],and[17]).ThereisneverthelesstheneedforafarmoreexhaustivestudyoftrafficpredictioninaDSAcontext,encompassingimpliedrequirementsfromanetworkingpointofviewaswellasperformanceandcomplexity.SignalingExchangeforDSA
GivenaDSAprocess,resourceexchangesshouldbenegotiatedbetweenRANs,probablyviaacentralizedserver.ThiscorrespondstotheResourceRequestand
O.Hollandetal.
AllocationphasesofFig.3.2,wherebytrafficpredictionsfromnetworkswouldbepresentinResourceRequests.Networksshouldtheninformdevicesaboutwhichbands(andperhapswhichRATs)aretobeused,beforenetworks’BaseStations(BSs)anddevicestunetheirtransceivers,inacoordinatedway,tothenewlyal-locatedresources(inadditionto,perhaps,adaptingtheRATsapplied).MirroringthearchitectureforDSAinFig.3.4,anassociatedsignalingexampleisdepictedinFig.3.5.
Fig.3.5.ExamplesignalingexchangeforDSA.
InaDSAcontext,signalingforthepurposeofresourcecoordinationbetweenRSCsandtheCSCcaneasilybedoneoverexistingwirednetworks.Suchsignalingshouldnotbeanissue,providingthattheparticipatingnetworksareproperlyau-thenticatedandtheinformationisappropriatelyencrypted.Withregardstosignalingbetweennetworksanddevices,thismightbeachievedthroughoneoftwooptions.Eitheran“in-band”methodmightbeused,wherebyinformationissentusinganex-istingchannelwithinthenetworkthatthedeviceisconnectedtoatthatpointintime,oralternatively,perhapsinconjunctionwithanin-bandchannel(e.g.,toprovidein-formationtodeviceswhennotconnectedoratstart-up),an“out-band”methodmightbeused,wherebyanewwidelyunderstood(andlikelyextremelysimple)physicalchannel,outsideofoperators’spectrum,isdefined[18].Fortheformer,itisacaseofsimplymappingtheinformationrequirementtoappropriatechannelsfortheRATsinquestion,andprovidingthespecifiedmeansforthatinformationtobestructured.Forthelatter,aradiointerfaceandprospectivespectrumbandmustbedefined.Various
3ArchitecturesandProtocolsforDynamicSpectrumSharing65
ideashavebeenfloatedforthis,rangingfromtheuseofanentirelynewpurposefullydefinedchannelinwidelyavailablededicatedspectrum(see,e.g.,[19]),totheuseofapre-existingDigitalVideoBroadcasting-Handheld(DVB-H)channelalongwiththeDVB-HRAT[18].
Spatial,Temporal,andSpectralCoordinationforDSA
Aftertheestimationoffuturetrafficloadshencerequiredresources,andinconjunc-tionwiththenecessarysignalingoperations,anallocationofspectrumwilloccuras/ifassessedappropriatebytheCSC.AfirstcriticaloperationherefortheCSCisdeterminingtheresolutionandappropriatealignmentofallocatedspectrumslices,andalsotheappropriatenessofperformingDSAbetweencertaintypesofsystemswithdifferentchannelbandwidths.Withregardstoresolutionandalignment,ofcoursethismustbecompatiblewiththetunabilityoftheparticipatingsystemsanddevices,aswellasthechannelbandwidthsforeachRAT-thisisfurtherinformationwhichmightbeconveyedinResourceRequests.Moreover,thechannelbandwidthsoftheparticipatingsystemsshouldideallybedivisible;i.e.,ifthreesystemsarepar-ticipatingintheDSA,andthelargestchannelbandwidthisAfollowedbyBthenC,thenideally,A/BandB/Cshouldbothproduceintegers.Inanycase,performingDSAamongsystemswithvastlydifferentchannelbandwidthsmightnotbeworth-while.Tohighlightthis,reallocationofachannelofspectrumfromasystemthatuseslargechannelbandwidthstoasystemthatusessmallchannelbandwidthswouldren-dertherestofthelargerchannelunusabletothelargersystem,hencewouldwastealotofspectrum(asitwouldbeunlikelythatthesmaller-channelsystemwouldbeabletomakeuseofthewholereallocatedband);conversely,reallocationofachannelofspectrumfromasystemusingsmallerchannelstoasystemthatuseslargerchannelswouldoftennotbepossibleasitwouldrequiretoomanyofthesmallerchannelstobereallocatedinordertoprovideonesuchlargerchannel.Ifthechannelbandwidthsarethesamehowever,spectrumsharingismuchsimpler.In[10],aDSAmethodissuccessfullyoutlinedbetweenamodifiedDigitalVideoBroadcasting-Terrestrial(DVB-T)systemandaUniversalMobileTelecommunica-tionsSystem(UMTS)network,bothofwhichuse5MHzchannels.
SpatialcoordinationisalsoacriticalissueforDSA,animportantaspectofwhichistheplacementofBSs.In[20],throughMonte-Carlosimulations,thepossiblespectrumsharinggaininamulti-operatornetworkfordownlinkUMTS-FDD,usingspeechtraffic,wasassessed.Itwasshownthattherelativedisplacementofmulti-operatorBSshasasignificanteffectoncapacitygainforDSA.Co-locatedBSsre-sultinbetterperformancecomparedwithdisplacedBSs,sincethenear-fareffectandinterferenceareminimized.
Furthertothiscontext,thecoverageoverlapamongsystemsclearlyaffectsDSAperformance.Forexample,theaveragecellradiusforDVB-Tmightbeinthetensofkm,comparedwithacellularsystemsuchastheGlobalSystemforMobilecommu-nications(GSM)orUMTSwhichmayhaveanaveragecellradiusofonlyakmorless.HenceanallocationfromGSMtoDVB-TrequiresthatallocationtobemadeforalargenumberofGSMcellsinordertocoveroneDVB-Tcell,whichmightnot
66O.Hollandetal.
befeasibleassomeofthesecellsmightstillrequiretheresourcethatisbeingreal-located.Conversely,anallocationofoneDVB-TcelltoGSMmightbe(partially)wastedasitwouldlikelyremainunusedinmanyofthecoveredGSMcells.
Regardingspatialcoordination,[21]proposedamethodtogroupsetsofcells/APsamongsystemsusingthelargestcoveragesystemasareference,whichisusefulforthepurposeofmakingspectrumsharingdecisionsonacell-by-cellbasis;thisisde-pictedforanexemplarycaseinFig.3.6(a).AnalternativedepictioninFig.3.6(b)highlightsthattheissueismuchmorecomplicatedinsituationswheresomeofthenetworksarenotspatiallyaligned.ForthecaseofspectrumsharinginahomogenousenvironmentcomprisedofonlyUMTSnetworks,[22]showed,throughsimulation,thedestructiveeffectofsuchcoveragemisplacementsonoverallDSAperformance.However,nostudyhasyetprovidedananalyticalbasisforspatialcoordinationinaDSAcontext.
.. .. ….. .. . . ………. . . . …. . . . . . . …. . . .. ….. .. . . ……….. ….. .. . . ………. . . . …. . …. . ………. …. . . …. . . …………. …. . . …. . . ………. ……. …. . . …. . . ………. . . .. . …. .. . . . ….. . …. . .. . …. .. . . . ….. . …. .. .. . .. .. …. . . .. . …. .. . . . ….. . ……. ….. .. . . . …. . . . …. . . ……. …. . . …. .. . . . ….. . . ….. .. . . . …. ….. . .. ….. . . …. .. .. …. ….. .. . . . …. ….. …….. . . . . …. …. . . .. . …. .. . . . ….. . …. . ….. . .. ….. . . ….. . . . . …. . .. …. .. . . . . .. ….. . . ….. . . . . …. ….. .. …. .. . . . ….. . . ….. .. . …. .. …. .. . . . ….. ……. . . . ….. ……. . . . …….. . . ….. . . .. . . .. …. .. . . . ….. ……. . . . . . . . …. ….. . .. ….. . . ….. . . …….. . . ….. . . .. . .. .. . ….. .. . …. . . . ……. . ….. …….. . . ….. . . .. . .. .. . . . …. …. .. …. .. . . . …. . ……. . ….. . .. . ….. . . . …. .. . ….. . . . …….. .. ….. .. …….. …. . . . ……. . ….. . .. . . . . …….. . . ….. . . .. . .. .. . …….. …….. . . …….. …….. …….. . . …. .. ….. . . . …….. .. …….. ….. . . . …. . . . …….. ……. .. . ……. . ….. . .. . …. .. .. . . . ….. ……. .. . ….. . …….. .. ……. …….. …….. . . …. .. .. ….. . . . …….. . . …….. ……. . . .. . …….. .. ……. …. . . ….. . . . ….. ……. .. . …….. . …….. . . . . …….. …….. . . …. .. .. . . ….. ….. . . . . ….. …. . .. . . .. …. . .. . . .. …….. .. ……. …. . . ….. . ….. ……. .. . …….. . .. ….. .. . ….. ….. . . . ….. . ….. .. . ….. ….. . . . ….. . . ….. . . . . ….. …. . .. . . .. …….. .. ……. …. . . ….. . . . ……………. . . . ... . …. . .. .. . …. .. . …UMTS Cells….. .. . ….. ….. . . . ….. ….. . . . . ….. …. . .. . . . . ………. . . . . …….. . . ….. . ….. . .. ….. .. . …….. …. . . . ... . …. . .. .. . …. .. . DVB-T Cells….. .. . ….. ….. . . . ….. ... …. . .. .. . …. .. .... .. . ….. ………….. ….. ….. . .. . .. . . …….. .. . .. …. . . . .. . . . .. ….. .. . . …WiFiAPs. .. ….. ... . . UMTS CellsWhich set do these cells apply to?DVB-T System 1DVB-T System 2(a)(b)Fig.3.6.Methodforspatialcoordination(grouping)ofnetworkcells/APsforspectrumsharingusingthelargestcoveragesystemasareference:(a)casewherenetworksarealigned,and(b)casewherenetworksarenotaligned.Finally,temporalcoordinationamongnetworksinDSAalsohasaprofoundim-pactonperformance.Clearly,asynchronizationmechanismisrequiredsuchthatallthenetworksarecoordinated,inthetimedomain,regardingtheirresourcere-quests/trafficpredictions,andparticularlytheirspectrumtransactionsandanyre-quiredretuning/reconfiguration.ThisisparticularlyrelevantwheretheparticipatingspectruminDSAisbeingusedwithahighaverageload.SucharequirementmakesacentralizedentityforDSA,asdiscussedpreviously,allthemoreimportant.Regardingtemporalcoordination,thereexiststheDSAintervalτ,actingasthesynchronizedperiodofdynamicallocationsamongnetworksthroughtheCSCforthetimerequirementsofDSAcoordination,trafficprediction,reconfigurations,andsoon.Ifτistoolarge,however,thepredictedtrafficinnetworksmightbeoutdated3ArchitecturesandProtocolsforDynamicSpectrumSharing67
atthepointintimethereallocationofspectrumisupdated,orinaworstcasesce-nario,thespectrumthatanoperatorhasbeenkindenoughtoallowtobetemporarilyallocatedtoanotheroperatorcouldbeblockedfromtheoriginaloperatorshouldherequirethespectrumbackduetoanincreasedtrafficload.Ifτistoosmall,issueswillusuallybecomeapparent,suchascomputationalload,signaling,etc.,relatedtoDSAprocesses,aswellasperhapsanincreasedfrequencyofunnecessaryreconfigu-rations.Indecidingτandtheassociatedtrafficpredictionapproach,atradeoffanal-ysisisthereforenecessarytoaddresstheconflictingrequirementsofaccuratetrafficprediction,andcomputational/signalingandother(e.g.,reconfiguration)loads.
In[12]and[23],simulationstudieswereperformedtoaddresstheeffectoftheDSAintervalongainforvariousapproachestoDSAbetweenaDVB-TandaUMTSnetwork.Itwasobservedthatincreasingτtoaboveacertainthresholdintervalof4-5hourscausesanegativeDSAgain(i.e.,loss)duetooutdatedtrafficprediction;ontheotherhand,reducingτtobelow1-2hoursdoesnotprovideasignificantimprovementinDSAgainhenceisnotjustified.Notethatsuchresultscanonlybeinterpretedinthecontextof[12]and[23],giventhespecificassumptionssuchasthetrafficpattern,trafficpredictionmethod,DSAmethod,etc.OtherNotableArchitecturesforDSA
VariousotherarchitecturesandprotocolsforDSAhavebeenproposedintheliter-ature.Oneimportantrealizationhereisthehierarchicalarrangementthatappliesinthestructureofwireless/mobilenetworks,andofradioresourcesingeneral.Inviewofthis,[21]and[24]proposedhierarchicalspectrummanagementconceptsthatcanfacilitatebothDSAandDSSarchitectures.Anotherconcept,proposedin[25],istheuseofaSpectrumAllocationServer(SAS)operatingintheMediumAccessControl(MAC)layer,wherehybridFrequencyDivisionMultipleAccess(FDMA)/Multi-CarrierCodeDivisionMultipleAccess(MC-CDMA)systemsinamulti-vendoren-vironmentwerestudied.BuddhikotandRyan[22]developedacoordinatedspectrummanagementconceptbasedonaspectrumbroker,withaparticularfocusonhomoge-neousCDMAnetworks.TheyconsideredseveralinfrastructurescenariosalongthelinesofsharingofBSsandtheuseofcollocatedantennas.Spectrumsharingcanalsobedirectlylinkedtoeconomicconceptssuchasauctions.In[26],theauthorspro-posedatoken-basedsharingprotocolwhichisappropriateinOrthogonalFrequencyDivisionMultipleAccess(OFDMA)environments.Thismethodcanbeappliedinbothlicensedandunlicensedspectrumtocreateadistributedandreal-timechan-nelrentalprotocol.Variousauctionmechanismscanapply,wherebythisandsimilarsuchapproachesareinlinewithactivitiesintheInstituteofElectricalandElec-tronicsEngineers(IEEE)802.22[27],[28]andIEEE802.16h[29]standards.Somerelatedworkslookingatsimilarapproachesinclude[30]and[31].
Inamulti-radioenvironmentwithDSA,perhapsthemostimportantissuetoconsideristheinterferenceeffectofsystemsoneachother.Aninterestingandratherchallengingcharacteristicinthiscontextisthatofdynamicallychangingspectralneighbors,resultinginunforeseeninterferenceeffectsofsystemsoneachother.One
68O.Hollandetal.
exampleastowhythismighthappenisthatofpowercontrolincreasingtransmis-sionpowertherebyaffectingtherangeatwhichinterferenceonaparticularfrequencymightberelevant,aswellasaffectingthelevelofout-of-bandemission.Asanexam-ple,[32]performedasimulationstudycharacterizingthedifferenttypesofinterfer-encepresentinatwo-RANspectrumsharingscenarioinvolvingUMTSandDVB-Ttypesystems.
Reconfigurability[33],[34]andSDR[35]areimportantpracticalfacilitatorsforDSA.Throughreconfigurability,perhapsasfacilitatedbySDR,thecharacteristicsofRATsindevicesandBSscanbemoreeasilychangedtothosethatareemployedinspectrumthatistobeswitchedto:thisisanotherparadigmtospectrumsharingwherebyspectrummightbesharedwithoutrequiringchangestoitsbounds,respon-sibleparty,ordeployedtechnology.Moreover,ReconfigurabilityandSDRareusefulforDSAperse,asthroughadaptabilitytheyallowgreaterfreedominthedefinitionofspectrumrangesandRATsthatmightapplytosystems/devicesasaresultofDSA.
Manynationalandinternationalresearchinitiativespartlyorfullyaddressedvar-iousaspectsofspectrumsharing,includingSDRandReconfigurability.StudiesandconsortiainEuropethathavefocusedonflexiblespectrumusageandrelatedcon-ceptsincludeIST-DRiVE[36]andOverDRiVE[37],IST-SCOUT,IST-WINNERPhasesIandII[38],IST-E2RPhasesIandII[39],andMobileVCE[40],nodoubtamongothers.VariousotherstudiesintheUSandtheFarEasthavecomplementedtheEuropeaninterestinDSA.IntheUS,theSPEAKEASY[41],JTRS[42],andXGprojects[43],amongothers,haveassistedprogressinrelatedareastospectrumsharing,andinJapananumberofProjects,manyofwhicharedrivenbytheNationalInstituteofInformationandCommunicationsTechnology(NICT)[44],havemadeasignificantimpactinsuchfields.IntheUS,muchprogressonSDRandrelatedtechnologiesisfacilitatedbytheSDRForum[45].SomefurtherinterestingreadingonDSAandrelatedarchitecturesincludes[10],[12],[23],[46],[47]and[48]amongothersreferencesinthischapter.
SummaryofArchitecturalRequirementsforDSA
Finally,therequirementsforsuccessfulDSAoperationaresummarizedasfollows:•
DSAmustbeperformedonlyinthecontextofagreementbetweenparticipatingnetworks,andparticularlywiththeagreementofthenetworksthatownspectrumthatistobedynamicallyreallocated.
Fromanetworkarchitecturepointofview,itisanticipatedthatthereshouldbesomeformofcentralizedspectrumcontroller,whichmustknowaboutthespec-trumusagesandrequirementsofeachnetworkacrossthesharedspectrumateachlocation(e.g.,inanultimaterealizationofDSA,ineachcell).Suchinfor-mationshouldbeexchangedbetweenthespectrumcontrollerandthenetworksinquestionusingpredefinedsignalingmeans.
Thecentralizedcontrollershouldcoordinate,spatiallyandtemporally,thespec-trumallocationsamongparticipatingnetworks.Inmostconceptualizations,theseallocationsmustbeupdatedatunderstoodperiodicintervals.Thecentralized
•
•
3ArchitecturesandProtocolsforDynamicSpectrumSharing69
•
•
•
controllermustallocatespectrumusingintelligentalgorithms,inawaythatmin-imizesinterference(ofcoursewithintheboundsofregulations)betweenpartici-patingnetworks,andsatisfiestrafficrequirementsconveyedinresourcerequestsasbestaspossible.AllocationsmustbemadeinafairandefficientmannerandmadeinaccordancewithprearrangedQoSpolicies,andofcoursemustgivepref-erencetotheownerofthespectruminquestionthroughpoliciesagreedbythatowner.
Eachnetwork,throughsomeformoftraffic/loadprediction,mustestimateitsspectrumrequirementforatleastthecomingDSAinterval,andreportthistothecentralizedcontroller.Thisestimationmustbedoneinafairwayamongparticipatingnetworks,likelyusingamethodthatitagreedamongnetworks.Thereisashorttimeintervalduringwhichtheactualspectrumreallocationtakesplaceacrossmulti-operatornetworks/systems.TheDSAprocessshouldguar-anteenoserviceinterruptiontoactiveusersduringsuchintervals,e.g.,throughappropriatesynchronization/coordinationmechanisms.
Thereshouldbesufficientprovision/backupincaseofanyfailureinaparticipat-ingnetworkorthespectrummanager.Mirroringspectrumcontrollersmightbeprovided,orautomatedmechanismsmightbeemployedinnetworkstoreverttotheirdefault(licensed)spectrumallocationsshouldtheDSAprocesscrash.Thealgorithmsemployedinthespectrummanagershouldberobustagainstfailure,stable,andglitch-free.
3.2.3RegulatoryandStandardizationViewpointsonDSA
InrealizationoftheimportanceofnovelspectrummanagementtechniquessuchasDSAandtheirpotentialtoimprovesystemperformances,regulatorsaroundtheworldareintroducingnewconsultations,measuresandrulestothecontext.TheOf-ficeofCommunications(Ofcom)intheUKhasproposedtheconceptof“SpectrumUsageRights(SURs)”[49].Underthisidiom,spectrumusagesbylicenseholdersarenolongernecessarilytiedtospecificRATs.SURsalsointendtospecifyacceptableinterferencelimitsforgeographical,in-bandandout-of-bandinterference,therebytakinganapproachmoreinlinewithallowingnovelspectrumusagessolongasnounacceptableeffectsresult.Moreover,itwasplannedthroughSURsfortheinvolvedpartiesinspectrumdecisionstobeabletointeractdirectly,withminimalinvolvementoftheregulator.TheseareallpromisingdevelopmentsleadingclosertoDSAbeingemployedamongcollaboratingnetworks.OtherOfcomreportshaveproventobeheadinginasimilardirectionregardingspectrumliberalization(e.g.,[50]and[51]),ashasbeenrecommendedbyinfluentialstudiessuchasthe“ReviewofRadioSpec-trumManagement”(“TheCaveReport”)[52]andthe“IndependentAuditofSpec-trumHoldings”[53]intheUK.
ThroughmeanssuchasSURs,theOfcomvisionistoreplacecommandandcontrolmethodswithmarket-basedmechanismsin71.5%ofbandsby2010[50].Moreover,Europeingeneralappearstobeheadinginasimilardirection,withthepassingofthenon-legislativeresolution“TowardsaEuropeanPolicyonRadioSpec-
70O.Hollandetal.
trum”[54],basedonaninitialsubmittedreport“AMarket-BasedApproachtoSpec-trumManagementintheEuropeanUnion”.
OtherinternationalDSA-relatedregulatorydevelopmentsincludeproposalssuchasa“Real-TimeSpectrumAuctioningModel”presentedtotheFederalCommuni-cationsCommission(FCC)intheUSbyGoogleInc.,andthereleasebytheFCCofkeydocumentssuchas“PromotingEfficientuseofSpectrumThroughElimi-nationofBarrierstotheDevelopmentofSecondaryMarkets”[55].InIreland,theCommissionforCommunicationsRegulation(ComReg)hasrecentlypublishedaconsultationonDynamicSpectrumAccess[56],muchofthecontentofwhichalsorelatestowidervisionsforenhancedspectrumsharingincludingDSA.Manyotherregulatorsworldwidearefollowingsimilarpathstogreaterdynamicityandfreedominspectrumattributions.
Standardizationisanareathatiscurrentlywakinguptothepossibilitiescon-cerningDSA.OneparticulareffortinthisdirectionistheIEEEP1900seriesofstan-dards[57].Establishedin2005andnowunderthesponsorshipofIEEEStandardsCoordinatingCommittee(SCC)41,thesestandardsaredevelopingvariousmech-anismsinsupportofSDR,CR,andimprovedspectrumoperationparadigms.TheIEEEP1900.4standardhasbeenconcernedwithdevelopingagenericarchitecturetofacilitatemechanismssuchasDSAandDSS[56].3.2.4FutureOutlookforDSA
Technically,someformsofDSA,suchassharingamongoperatorswiththesameRAT,arealreadyachievable.Specificationofprocesses/architecturesforDSAop-erationinsystemsoftenstillneedtobeaddressed,forexample,providingmeanstocoordinateDSAwithinacompositenetworkoramongnetworks,andprovidingmeanstoinformdevicesofchangesinspectrumallocations.Nevertheless,givenanagreementtoallowDSAbetweendistinctnetworks/operators,thoseoperatorsmightuseindependentlyagreedmechanisms(ofcourse,inalignmentwithregula-toryrequirements—whichmighthavetobevalidatedbytheregulatorinsomecases),and/ormightusespecificationstoassistwheretheyexist(e.g.,[11]).
TherearetechnicalissuestobeaddressedintherealizationofsomeadvancedformsofDSA,suchasensuringappropriatereconfigurabilityandtunabilityofnet-works/devices.Forthemostparthowever,challengesintheDSAcontexthaveforsometimebeenlargelypoliticalratherthantechnical.First,thesepoliticalchallengesexistbetweenoperators:whyshouldanoperatorwhohaspaidasignificantpricetoacquireprimespectrumallowcompetitorstouseit?Operatorsdoneverthelessfore-seemechanismstoachieveapproachestoDSAthroughchargingthe“secondary”operator,akintospectrumleasing.Secondandperhapsmostsignificantly,politicalchallengestoDSAexistforregulators.Thisisbecauseregulatorsareresponsibleforassuringthatspectrumallocationsarereliable,andwanttoensurethatanynewtech-nicalapproacheswillnotleadtoasituationwherethosewhohavepaidforspectrumaredisadvantagedbyinterferencefromothersystems.Nevertheless,asindicatedinthepriorsection,therearesignificant,perhapsirreversiblesignsthatthepresentcon-servativesituationwithinregulationisrapidlychanging.
3ArchitecturesandProtocolsforDynamicSpectrumSharing71
Inthelightoftheobservationsthathavebeenpresented,itseemsthatDSApro-cessesarecertaintobecomewidelyemployed-thequestionissimplyhowlongitwilltake.Tothisend,DSAmechanismswillberealizedlongbeforetheeventoffull“Mi-tola”CR,whichcanbeconsideredtheultimateandmostchallengingachievementinspectrumsharing.Givenrecentadvances,variousformsofDSAcanbeexpectedtobewidelyusedlongbeforetheendofthe2010sdecade.
3.3DecentralizationofSpectrumSharing
Nextinvestigatedarearchitecturesandprotocolsforthedecentralizationofspec-trumsharingdecisions,wherebytheterm“decentralization”isusedinthesenseofsecondarydevicesorsystemsindependently(orperhapssemi-independently)choos-ingspectrumtousebasedonusageopportunitiestheyhaveidentified.Suchdecen-tralizationisassumedtobeachievedthroughmechanismssuchasOpportunisticSpectrumAccess/DynamicSpectrumAccess[6],[9],wherebyforsimplicitytheseandallsimilarsuchtermsarecollectedunderonegenericterm“DynamicSpec-trumSelection(DSS)”.Adegreeofcentralizationindecisionswilloftenneverthe-lessremain;forexample,DSSmightbeperformedonlywithinconstraints/policiesasconveyedbythenetwork/operatoroftheprimarysystemoranassociatedentity.Suchconstraints/policiesmightspecify,forexample,themaximumallowableinter-sensingperiod,themaximumallowablesecondaryaccesstransmissionpower,thebands(perhapsdynamically)thatareallowedtobeopportunisticallyused,andanacceptableopportunisticusagemechanism(e.g.,aMAC).3.3.1BasicConceptofDSS
DSS,similarlytoDSA,isabasisforsecondaryaccesstospectrumofalicensedsys-tem.DSSmight,however,alsobeperformedinthecontextofunlicensedspectrum,akintowhatisalreadyachievedthroughIEEE802.11networks.TherearefewerchallengestotherealizationofDSSinunlicensedspectrum,particularlybecauseitisnotcontroversialtothespectrumlicenseowner(asthereisnone!),transmissionpowerisusuallylimited(tominimizeinterference),andthereisnoimposedRATwhichmustbeadheredto.HenceonlyDSSinthelicensedspectrumdomainiscon-sideredhere.
DSS,unlikeDSA,isbasedonadevice-centric(orsometimessecondary-systemcentric)approach,wherebysecondarydevices/systemschoosewhichbandstheywilluseratherthanbeingallocatedthem.Inmakingthischoice,atransmittingDSS-enabledsecondaryradiomustoperatesuchthattheimposedinterferencetoallpri-maryradiosisbelowarequiredthreshold.Ifsuchaconditionismet,forinstancethroughthedistanceand/orphysicalobstaclesbetweenthesecondarytransmitterandprimaryreceiver(s)beingsignificantenoughtocausesufficientpathloss/shadowing,orthoughtheDSSradiointelligentlyusingopportunitieswithintheRATofthepri-marysystem(e.g.,idlefrequencydivisions,timeslots,codes,etc.),theDSSdeviceisallowedtotransmitintheprimaryband.
72O.Hollandetal.
TwocategoriesofDSSarenegotiatedandopportunisticaccess.Intheformercase,asignalingchannel,likelytobewirelessinnature,providesaninteractionmediumbetweentheprimaryandsecondarysystem,wherebyseveralsignalingso-lutionsforDSShavebeenproposedintheliterature,e.g.,[19],[58]and[59],andareunderinvestigationaspartoftheIEEEP1900.4standardizationeffort[57].Notethatsuchasignalingchannelmightalsoplayapartinacontrolledformofopportunisticaccess,whereidentificationofaccessopportunitiesisdirected/constrainedbysig-nalingfromtheprimarysystemoranassociatedentity.Inpureopportunisticaccesshowever,directinteraction(i.e.,signaling)isnotpossiblebetweentheprimaryandthesecondarysystems.OneexampleofsuchisTVbands,whichareconsideredforsecondaryaccesswithintheIEEE802.22standard[27],[28].Insuchcases,itistheresponsibilityofthesecondarysystemtoreliablyidentifysecondaryusageopportu-nities,andtovacatethebandassoonastheprimarysystemreappearsorinterferencetoitisanticipatedtobeanissue[48].Thealternativeofusing“underlaytransmis-sion”isextremelycontroversial.
3.3.2ArchitecturesandProtocolsforDSS
ThemainphasesinvolvedinDSSoperation,whichareemployedcyclically,areillus-tratedinFig.3.7.TheassociatedapproachestoDSS,withreferencetothesephases,arediscussedinthissection.NotealsothatsucharepresentationmightbeextendedtothecontextoftheCognitionCycle,introducedbyMitola[60].
OpportunityIdentificationResourceAccessCognition&SignalingResourceReleaseFig.3.7.PhasesofDSSoperation.
OpportunityIdentification
AfirstmajorrequirementinDSSistheidentificationoftransmissionopportunities,i.e.,whiteorgreyspacesinthespectrum.Thisisparticularlyimportantgiventheneedtoavoidinterferencetoprimarysystems.
3ArchitecturesandProtocolsforDynamicSpectrumSharing73
Thegoaloffindingsecondaryspectrumaccessopportunitiescanbeachievedthrough“spectrumsensing”,aclassificationoftechniquesforwhichisprovidedin[61].Therearetwogeneralapproachestospectrumsensing:energydetection,andfeaturedetection.Energydetectioniseasiertoperform,butishamperedbe-causesourcesofenergyotherthantheprimarysystem,i.e.noise,arepresentinthechannel.Featuredetectionisabletomoreaccuratelydetectthepresenceofsig-nals,althoughhasanotableprocessingrequirementandlatency.Inanycase,ithasbeensuggestedthatthelinkbudgetanalysisforaccuratespectrumsensing,giventhenoise,interferenceandchannel(i.e.,shadowingandfading)uncertaintiesmakesitalmostimpossibleforindividualCRstoidentifyopportunitiesreliably[61].Simplyconsideringafademargininthelinkbudgetanalysistosolvethisisnotsufficient,duetouncertaintyoffadingdistributionmodelsamongotherreasons.
Twoarchitecturalapproachestotacklethisproblemarecooperativesensing[63],andtodecouplethesensingtaskfromdevicesthroughusinganinfrastructureofsen-sors[].ThecooperativesensingapproachreducestheerrormargininthesensingtaskbycollatingmeasurementsofCRswhichareatsomewhatdifferentlocationsandtimes(henceexperiencingdifferentfadesand,dependentonpurpose,differentshad-owingcharacteristics).Thedegreeofsensingcomplexity(sensitivity)ateachnode,andthechannelcharacteristics(e.g.,coherencetimeandbandwidth)determinethedegreeofcooperationrequired[65].
Cooperativesensingfacesseveralchallenges.Specifically,cooperationisnotpossiblewithoutareliablesignalingchannelbetweensecondarydevices,wherebythischannelwouldmostcommonlyhavetobeachievedusingwirelessmeans.Afrequentproposalisfordedicatedspectrumtoexistforsuchsignaling(see,e.g.,[18]and[19]),spectrumwhichhasnotbeenallocatedbyanyregulatorybodysofar(in-deed,arelatedproposalhashadtobesomewhatwatereddownintheWorldRadio-communicationsConference(WRC)2007[66]).Somedubiousalternativesincludeusingtheopportunisticaccessitselfforsignaling,therebyreducingtheinherentrelia-bility,andtouseofunderlaytransmission.Furtherchallengestocooperativesensingarethesheeramountofsignalingoverheadlikelytoberequired,andthequestionofreliabilityofsensedinformation.Asnotedin[63],cooperativesensinggainisaffectedbythepresenceoffailing/maliciousnodes.
Despitesuchissues,aformofcooperativesensingarchitectureisalreadyclosetocommercialrealizationwithintheIEEE802.22standard.OnereasonthatthisispossibleisbecauseofthestaticnatureofprimaryandsecondarytransmittersinthecontextofIEEE802.22,thusreducingrequiredsensingoverhead.Sensinginthiscaseisperformedintwostages.First,afastsensingalgorithmisusedbyallCus-tomerPremiseEquipments(CPEs)(i.e.,homeAPs),wheretheresultsofthisareaggregatedatthecentralized802.22BS.Thisfastsensingalgorithmtypicallytakeslessthan1msperchannel.Next,theBSdecidesifmoresensitive,∼25msperchan-nel,finesensingisnecessaryonsecondaryaccesschannelstolookforsignaturesofprimarysystems.Thisensuresthattheprimarysystemexperiencesnointerfer-ence[28].
ConsideringtheapproachdepictedinFig.3.8ofusingadedicatedinfrastructureofsensorsasthespectrumsensingmeans,anumberoftechnicalchallengesexist.
74O.Hollandetal.
First,thereisthequestionofwhowouldimplementandmaintainsuchasensingnet-work.Next,thequestionofreliabilityofinformationalsoapplieshere.Furthermore,thereisthequestionofhowthegatheredinformationissignaledamongsensorsandto/fromdevices,thatofsensorenergy,andalsotheissueofnetworkroll-outtimeandcost,giventhenecessitytocoverlargeareasandtooptimizesensorsforgeographiclocale.
PrimaryNetworkSensorNetworkSecondaryNetworkFig.3.8.Multi-layerednetworkarchitecturedecouplingspectrumsensingfromthesecondaryspectrumaccessdevices.
Variousalternativemethodstopreventinterferencetoprimaryreceivers,otherthanspectrumsensing,exist.Considerthecasewherethesecondarytransmitterismuchclosertotheprimarytransmitterthantheprimaryreceiver.Insuchcircum-stances,itispossiblefortheCRtoknowtheprimarysystem’sinterferencechar-acteristicsandbenefitfromusingadirtypapercodingtechniquetotransmitwith-outcausingdetectableinterferencetotheprimaryreceiver.Ithasbeenshownthat,throughthis“cognitiveradiochannel”strategy,rateregionsofsomewherebetweentheinterferencechannelandMultiple-InputandMultiple-Output(MIMO)channelcanbeachieved[67].Ofcourse,issuesofdelay(reactivity)andprocessingloadareachallengehere.
Asanotherapproach,secondarytransmittersmighttrytodeterminethelocationsofprimaryreceivers.Throughthisapproach,[62]proposedtointroduce“NoTalkZones”forsecondarydeviceswithinacertainrangeofeachprimaryreceiver.Toaccountforchanneluncertainty(e.g.shadowingandfading),thisrangeshouldbesignificantlylargerthantherangeoftheprimarysystem.Ananalysisoftrade-offwasperformedin[62]toaddressdifferentoptionsinthisscenario.
Anothermeansforidentifyingthepresenceoftheprimarysystemistheuseofdedicatedwirelesssignalingchannelsperse.Differentapproacheshereincludetheadvertisingofavailableresourcesbytheprimarynetwork[8],[18],orbroadcastingofbeaconsassociatedwithoccupiedbands(e.g.,[58]).Alongsimilarlines,aded-
3ArchitecturesandProtocolsforDynamicSpectrumSharing75
icatedsignalingchanneltermedtheResourceAwarenessChannel(RAC)hasbeenproposedin[19],utilizedbyprimaryreceiverstobroadcastinformationabouttheoc-cupancyofthebandstheyarereceivingdataon.Inaccordancewithreciprocity,thesecondarytransmittersmightalsogleansomelimitedinformationaboutthechannelconditionstoprimaryreceiverstoenhanceopportunitycharacterization.
Finally,givenanyopportunityidentificationmechanism,itsperformancemustbecontinuallymonitored,likelywiththeinvolvementofaregulatorybody.Tech-nicalboundsfortheoperationofsuchamechanismmustalsobespecifiedinitsselectionordevelopment.Toservesuchpurposes,aquantitativemetricforinducedinterferenceisofimportance.TheInterferenceTemperature,proposedbytheFCCintheUS[68],mightbesuchametric;however,theFCChasrecentlydropped(withstrongreservations)thisproposalduetoalukewarmreceptionoftheconceptfromsomecommentingparties[69].Anotherrecentinterference-basedsuggestionforspectrumusage,albeitnotdirectlyaimedatsecondaryspectrumusage,hasbeenthatofSURsproposedbyOfcomintheUK[49].ResourceAccessandRelease
Havingidentifiedanappropriatespectrumtransmissionopportunity,thenexttaskforDSSdevices/systemsistoefficientlyuseit.ThisphasecomprisesdefinitionofanappropriatenetworkingandRATsolutionforthesecondarysystem,bothofwhichmayfeedonsolutionsthathavealreadybeendevelopedinothercontexts.Becauseofcross-layereffects,thedefinitionofsuitablehigherlayersmightalsoapply.OneconsiderationinthechoiceofanappropriateRATisthat,givenuseofasensingmechanismforopportunityidentification,thespectrummustfrequentlyberescannedtoensurethattheprimarysystemhasnotreappeared.Hencetimeperiodsmustbesetasideforthescanningproceduretobere-performed,duringwhichthesecondarysystemmustremainsilentintheband.IEEE802.22[27],[28]givesafirstrealizationofanappropriateRATwithinthiscontext.
Regardingnetworkingforthesecondarysystem,differentapproachessuchasCORVUS[70]andCogMesh[71]wereproposed,andIEEE802.22mightagainbereferredtoasanexampleofacentralizednetworkingapproach.[70]proposedaran-domaccessproceduresimilartothatusedinunlicensedspectrumaccess,wherebythesecondaryuserscancreate“SecondaryUserGroups(SUG)”tocoordinatecom-munication.MembersofaSUGmightcommunicatewitheachotherinanad-hocfashion,oralternativelymightaccessafixedinfrastructureviaadedicatedaccesspoint.Throughusingtheadaptiveclusteringtechnique,[71]proposedtocreateasecondaryadhocnetworkbyjoiningsuchclusters.Eachclusterisformedbasedontheavailabilityofalocalcoordinationchannelinasharedspectrumscenario.Theproblemsofneighbordiscovery,clusterformation,networkformationandnetworktopologymanagementinthisscenarioareaddressed.
WithregardstoResourceRelease,secondaryaccesstospectrummustcurtailassoonasitbecomesapparentthattheprimarysystemispresent,orinterferencetoitislikelytobeaboveathreshold.Moreover,tominimizetheprobabilityofinter-ferencetotheprimarysystem,throughcoordinationmechanisms,maximumusage
76O.Hollandetal.
ofprimarybandsthatarealreadybeingopportunisticallyusedmustbemadebythesecondarysystem(s),beforeanynewbandsareaccessed.Cognition
AfinalimportantaspectofDSS,particularlyinthecaseofCR,isthatofthecog-nitionlevelofdevices.Aquickmentionofthisapplieshere:issuessuchasCRandcognitionareexpandeduponasthemainpurposeofthefollowingchapter.
ThecognitivecapabilitiesofDSSenableddevicesmightdifferconsiderably,wherethecomplexityofaCRisindirectrelationtothelevelofcognitionineachphaseoftheCognitionCycle:Observing,Orienting,Planning,Deciding,Learning,andfinallyActing[60].Policies,whichassistinidentifyingspectrumopportuni-tiesandhowtheymustbeused,arenecessarytobound/guidethecognitionofCRs.Definitionsofpoliciesintheformofmachine-understandablepolicydescriptionlan-guageareprovidedin[72],[73];moreover,thecreationofpoliciesisfurthersimpli-fiedbyanaccuraterepresentationofcontextsuchastheRadioKnowledgeRepre-sentationLanguage(RKRL)[60].GenericmethodsthatcanbeusedtoformpoliciesincludetheCommonInformationModel(CIM)[74],Ponder[75],andDEN-ng[76].
Ingeneral,forpurposessuchasintelligentspectrumopportunityidentificationandintelligentRAT/MACdefinition,thecognitionlevelofaCRgoesindirectrela-tiontotheperformanceimprovementitcommands.Clearlyhowever,ahigherlevelofcognitionisassociatedwithmorecomplexsoftware/hardwarerequirements,henceincreasesthecostofthedevice.Asinspiredby[77],thelevelofcognitionfordiffer-entpurposes,versuscomplexity,isdepictedinFig.3.9.OtherNotableArchitecturesforDSS
AnumberoffurthernotablearchitecturesforDSShavebeenpublishedinthelit-erature,someexamplesofwhicharebrieflydiscussedasfollows.OFDMAiscon-centrateduponinitially,asthisrepresentsthefutureofmulti-useraccessinavastarrayofwirelesssystemsandiscapableofsharedspectrumusagethroughdifferentsub-carriersbeingallocatedtodifferentsystems.ADSSmethodinthiscontextwasdevelopedin[78],wherethroughusingbinaryallocationvectorstomarktheutilizedsub-carriersoftheOFDMA-basedsystem,aswellasanefficientphysicallayerde-tectionmechanismtermedthe“boostingprotocol”,spectralcoexistenceofprimaryandsecondarysystemsispossible.Othertopicsdiscussedin[78]includeMACis-sues,synchronization,andschedulingalgorithms.QoSinsuchOFDMA-basedDSSscenarioswasinvestigatedin[79],whichhasanemphasisonthetradeoffbetweenfairnessandQoSofparticipatingnodesinasharedspectrumpool.
TheDARPANeXtGeneration(XG)programhasaimedtodevelopandex-pandtechnologiesandsystemconceptsforimplementingdynamicspectrumsharingthroughopportunisticaccess[43].Recentfielddemonstrationshaveshowncom-municatingradiosthroughXGprotocolstobeeffectiveatavoidingoccupiedfre-quencybands[80].TheXGprojectalsoaimstodevelopaframeworkformachine-understandablepoliciesmanagingradiobehavior,facilitatingdevicestomakeintel-
3ArchitecturesandProtocolsforDynamicSpectrumSharing77
level of cognitionconceptualdecision makingautonomoussymboliclearningnumericenvironmentawareevent-drivenfeaturedetectionspectrumsensingcoherentdetection(match filter)energydetectionself awareawarenesslevel of complexityFig.3.9.Levelsofcognitionandcomplexity.
ligentopportunisticdecisions.MoreabouttheXGframeworkandprotocolscanbefoundfrom[73],[81]and[82].
FurtherinterestingliteratureonarchitecturesforDSSisobtainedbyreferringtotheaforementionedreferencesinthischapter.SummaryofArchitecturalRequirementsforDSS
Finally,therequirementsforsuccessfulDSSoperationaresummarizedasfollows:•
Inlicensedspectrum,thereliableidentification(through,e.g.,spectrumsensing)oftransmissionopportunitiesmustapply,suchthatthereisnotangibleinter-ferencetoprimarysystems.Thisisachallengebecauseofuncertaintiessuchasnoise,fadingandshadowing.Cooperativesensingmightbeasolutioninop-portunisticDSSscenarios,thevariousoverheads(signaling,computationaletc.)ofwhichshouldbecarefullyaddressed.Anotheroptionmightbetodecouplethesensingtaskfromsecondarysystemsthroughtheuseofasensornetwork.Amongfurtheroptions,secondarydevicesmaybewardedoffofutilizedbandsbytransmissionsfromprimarysystems.
Secondarytransmittersinlicensedspectrummustkeep(periodically)checkingthattheprimarysystemhasnotappeared.Undercommonlyassumedsensingmeans,thishasimplicationsfortheMAC/RATasemployedbythesecondarysystem,asthesecondarysystemmustbesilencedduringeachsensingperiod.Inlicensedspectrum,inthevastmajorityofforeseeablescenarios,DSSmustoperatewithinconstraintsasspecifiedbytheownerofthatspectrum.Various
•
•
78O.Hollandetal.
•
•
•
mechanismsforconveyingtheseconstraintsmustbespecified,suchasarecur-rentlybeingconceivedthroughnetwork-devicedistributedpolicieswithintheIEEEP1900.4workinggroup[57].
Bothstructuredandunstructured/ad-hocnetworkarchitecturesforsecondarysystemscanbeenvisaged.Astructuredarchitecturemightbemoresuitableiftheprimarysystem’sspectrumusagecharacteristicsinthechannelarenotfrequentlychanging(inspaceandtime),forexampleintheIEEE802.22architectureforsecondaryaccessofTVbands.Unstructuredsecondarysystemarchitecturesaremoresuitabletobenefitfromlocalized(intime/space)changestothespectrumusageoftheprimarysystem.
Areliablesignalingchanneltocoordinatetheusageofresourcesandforcoopera-tivespectrumsensingisachallengeinDSSduetotheunavailabilityofdedicatedbandsforsuchsignaling.Suchsignalingmeanswouldlikelyneedtobecreated.Finally,thenetworkingprobleminsecondarysystemsisanotherissuetobecon-sidered.Differentmeansforthishavebeenoutlinedintheliterature.
3.3.3ArgumentsforandAgainstDSS
NumerousresearchershaveidentifiedpotentialusesforDSS,someofthemostwide-rangingofwhicharewithintheCRidiomasintroducedbyMitola[8].Otherlessambitioususessimplyrevolvearounddevicesbeingabletocommunicatedirectly,bypassingnetworksandthusmakingthecommunicationfarmoreefficient,user-friendlyandcheaper.Forinstance,ifyouarecallingorsendingatextmessagetosomeoneacrossthestreet,whyisthecommunicationnotsentdirectlywithonetrans-missionoveralocallinkthroughaDSSmechanism,insteadofbeingtransmittedtoaBS,throughoneormorenetworks,andthentransmittedagainbyaBStothere-cipient?SuchobservationsrepresentverystrongargumentsforDSS.
TherearealsovariousotherargumentsastowhyDSSmightbeused.DSSisfasterandmoreflexible,potentiallyachievingfarbetterspectrumusageefficiencyinthesharingmechanism.Forinstance,throughDSSmechanisms,theloads(i.e.,signalingandcomputational)andotherissuesinperformingnetwork-widedecisionscanbelargelyavoided,asdecisionsaremadeatthe“grass-roots”oflocalradiotransmissionlinks.Thisleadstodecisionsbeingmadeinafarmoretimelymanner,thusimprovingreactivitytodynamicchangesinspectralloads.Moreover,throughthespectrumdetectionorotherawarenessmechanismsinDSS,farmoredetailedlocalinformationcanbeleveraged,achievingamuchenhancedspatialaccuracyintheoptimalityofspectrumsharingdecisions.
Ontheothersideofthecoin,centralizedapproachesareusuallyeasiertodevelopandmaintain.DSAtakescomplexityawayfromdevicesthusmakingthemsimplerandcheapertoproduce,whereasintheDSScase,complexityisbroughttotheedgeofthenetwork,i.e.,totheaccessnetworkandtodevices.DependentontherangeoftasksandnatureoftheDSSscenario,suchcomplexitymightvarysignificantlyonacase-by-casebasis.Forexample,ifasensornetworkisusedforspectrumsensingasillustratedinFig.3.8,thecomplexitywillmostlyberelatedtoadaptivetransception
3ArchitecturesandProtocolsforDynamicSpectrumSharing79
capabilities.Ontheotherhand,ifspectrumsensingislefttodevicesandsecondarysystems,thecomplexityandcostofdevicesandsystemswillincreasedramatically.
AsanotherargumentagainstDSS,ofcourseitiscontroversialbecauseoftheincreasedpotentialforinterferencetoprimarysystems.Thisisdiscussedinotherpartsofthischapter.
3.3.4RegulatoryandStandardizationViewpointsonDSS
RegulatoryandstandardizationviewpointsarebothprogressingtowardsDSSbe-comingareality.Inunlicensedspectrum,throughIEEE802.11andmorerecentlythroughemergingIEEE802.16networks,meansforwhatmightbeconsideredaformofDSShavealreadybeenexistenceforsometime.DSSinlicensedspectrumisbothtechnicallyandpoliticallymorecomplicated,becauseoftheneedtoavoidinterferencetotheownerofthatspectrum.Standardsinthiscontextareneverthelessdeveloping:someofthesearetheIEEE802.22standardwhichallowssecondaryaccesstolocallyunusedTVbands[27],[28],IEEE802.11ywhichisintendingtoextendWi-Fitothescopeofsecondaryaccesstolicensedspectrum[83],andtheIEEEP1900.4standardwhichaimstoprovideagenericarchitecturetoassistde-cisionmakingandinformationprovisioninbothDSSandDSA[57].Moreover,effortssuchasDARPAXGarelookingtostandardizetheirdevelopingCRtech-nology[43],[73],[81],[82].
Becauseoftheneedtoavoidinterference,DSSintheregulatorydomainisfarmorechallengingthanDSA,henceregulatorsarefarmorecautionsaboutacceptingit.Nevertheless,theFCCintheUShasrecentlyannouncedarulechangeasregardsCR,allowingsecondaryaccesstoTVbands[84].Animposedrequirementisthatalldevices/systemscertifiedtouselocallyvacantTVbandsmustemployCRtech-nologytoavoidinterferencetoprimarysystems.Thisisamongotherdevelopmentssuchthe“CognitiveRadioTechnologiesProceeding”[85],andadocumentconsid-eringthepromotionofsecondaryspectrummarkets[55].Moreover,inmid-2006theSenateCommerceCommitteeintheUSadoptedthe“AdvancedTelecommunica-tionsandOpportunityReformActof2006”.ThisrequiredtheFCCtocontinuerule-makingproceduresregardingtheopeningofVeryHighFrequency(VHF)andUltraHighFrequency(UHF)TVbands(54MHz-698MHz)tousebywirelessbroad-bandservicesandotherDSSenableddevices.Furthermore,interestingproposalshavebeensubmittedtotheFCC,includingthoserelatedtousingtheFMbandformass-communicationintrains[86],andonesubmittedtoGoogleInc.relatedtoasecondaryspectrumauctioningmodel[87].
TheoutlookinothercountiesissimilarasregardsprogresstowardsDSS.ThisisevidentfromvariousstudiesintheUKandIreland[51],[56],[88],aswellasthesupportinJapanandEuropeofanumberofDSSrelatedprojects[39],[44].Moreover,theimplicationofrecentconsultationsanddecisionsisthatregulatorsworldwideareheadinginabroadlysimilardirectiontowardstheacceptanceofDSS,althoughtherateofprogress,asidefromsecondaryaccesstoTVbands,islikelytobeslow[88].
80O.Hollandetal.
3.3.5FutureOutlookforDSS
DSSbysecondarysystemsinradioandTVbandsismucheasiertoachieve,asprimaryradio/TVtransmissionsaregenerallywithahighpowerandfixedlocationsoaremucheasiertodetectandavoid.Moreover,thoughthesituationmaychange,thekindsofsecondarysystemsthathavethusfarbeenenvisagedtousesuchbandsinasecondaryaccessfashionarerelativelystationary.HenceitisperhapsnosurprisethatthesebandsareexpectedtoprovidesomeoftheearliestcommercialrealizationsofDSSthroughCR,possiblyevenasearlyasin2009-10intheUS.
Insituationswhereeitherorboththeprimaryandsecondarysystemsemploymovingtransmitters,andparticularlyiftheprimarysystemsareofrelativelylowtransmissionpower,progresstowardstherealizationofDSSisexpectedtobeslow.Thisisbecauseitmustensuredthattheprimarysystemsexperiencenointerference:therewouldbemassiveoppositionfromownersofprimaryspectrum,especiallyincellularbands,shouldsecondaryaccessdegradetheirservices.Giventheneedforgreaterchecksinsuchcontexts,complexitiesresultontheregulatoryside:e.g.,thequestionofwhatdurationofinterferencebeingmarginallyaboveathresholdforaspectrumbandisconsideredunacceptable(thisisimportantasregardsthefrequencyofspectrumsensingbythesecondarysystem),andwhatconstitutesanacceptablethreshold,foreachspecifickindofallocatedband.Furthermore,howisitensuredbytheregulatorthatmechanismsemployedareabletoavoidcausinginterference(i.e.,whichmechanismsshouldtheregulatorenforceasbeingacceptableforthis),andhowitisknownthattheywillremainreliableandwon’tbe‘hacked’tomis-behave?ItislikelythatalltheseanswersmustbeprovidedbytheregulatorbeforeDSScanbecomeareality,andthetypesofanswerswilllikelybeverydifferentfordifferenttypesofprimarybandsandpotentiallyalsofordifferenttypesofsecondarysystemsusingthosebands.HencetherearemanyunknownsandchallengesintheDSScontext,someofwhichtechnicalsolutionsareyettobeprovidedfor,andsomeofwhichtheeventualsolutionsforwillalmostcertainlybecontroversialtosomestakeholders.
Tosummarize,thedevelopmentofDSSisexpectedtobesignificantlydelayedinmanycontexts.OfcomintheUKhasanticipatedtherealizationofafull“Mitola”CR,the“ultimate”inadvancedDSS-relatedcapability,asbeingaslateas2030[88].Finally,significantaspectsofDSAandDSSarecomparedinTable3.1.
3.4ResourceOptimizationinSpectrumSharingUsingGameTheoreticalApproaches
Inthissection,analyticalapproachestoradioresourceoptimizationinbothDSAandDSSdomainsarediscussed,basedonGameTheory.Gametheoryischosenbecauseitprovidesasetofpowerfulmathematicaltoolsforanalysisof“conflictandcoopera-tionbetweenintelligentrationaldecisionmakers”[].RadioresourcemanagementinDSAandDSScontextscanbemodeledbygametheory,wherebythenetworksareeithercooperating(asinDSAornegotiatedDSS),orarecompetingforresources
3ArchitecturesandProtocolsforDynamicSpectrumSharingTable3.1.CharacteristicsofDSAandDSScompared.
DynamicSpectrumAllocation(DSA)
Management
Centralized
DynamicSpectrumSelection(DSS)Centralized/Decentralized
81
Intelligence
Centrallylocatedwithinthenetwork
Reliesonlocallydistributedinformationgathered/processedbyoneorseveraldevices/BSsand/orasensornetwork?Decisionmakingmightbecentralizedwithinthe
DSSnetwork,dependentonthearchitectureNewerdevicesrequired,oftenbasedonSDR,andoftenwithacapabilitytosenseandaccessspectrumindifferentbandsaccordingtothedefinedspecificationSpectrumactivitydetectionandadaptivetransmissionrequired(indevicesandBSsifemployed).InmanycaseswillcompriseSDRorarelatedsolution(implyingexcessive
digitalsignalprocessing)Likelycooperationbetweenmultipleintelligentdevices;mustbedonewithahighprobabilityofreliabledetectionofspectrumholesAmountofsignalingoverheadmightbecomeprohibitiveasnetworkdimension(particularlynumberofnodesperarea)increases
Usuallymustbereal-time
Changesrequiredmainlyatthenetworkside;anoperatorupgrade
maybenecessary.Greater
ImplementationComplexityreconfigurability/tunability
forbothnetworksanddevicesmightberequiredforadvancedsolutions
ProcessingComplexity
Trafficpredictionalgorithmsnecessary
Signaling
Overhead/InformationExchange
Somesignalingrequiredbetweenthecentralizedspectrumcoordinatorandnetworks,andtoconveypoliciestodevicesAssystemsizegrows,centralizeddecisionmakerbecomesheavilyloaded.Ahierarchicalarchitecturemightimprovethis
Nonreal-timeand
tosomeextentreal-time
Scalability
ReactivitytoContextChanges
82O.Hollandetal.
hencecreatingconflict(asinopportunisticDSS).Theplayers,whichbygametheorydefinitionarethedecisionmakers,arethecentralizedspectrummanagemententitiesintheDSAcase.IntheDSScase,devices,orsometimesthetransmission-receptionpairs,aretheplayers.TheDSAdomainisdiscussedfirst.3.4.1TheDSACase
Themulti-operatorarchitectureinFig.3.4canbeusedasthebaselineforanalyzingtheDSAcase,wherebydifferentRANSpectrumControllers(RSCs)interactviatheCentralizedSpectrumController(CSC)tosatisfytheirtrafficdemandsforaspecificresourceallocationperiod.TheCSCresemblesamarketplace,wheredifferentplay-ers(i.e.,theRSCs)bargaintoobtaingoods(i.e.,spectrum)inordertomaximizetheirpayoffs,whichmightbethroughputorpercentagereductioninblocking/droppingprobability.TheassumptionofrationalityoftheplayersimpliesthatifusingstrategyA(e.g.,usingaspecificbandortransmissionpowerlevel)resultsinpayoffX,andusingstrategyBresultsinpayoffY,andY>X,therationalplayerchoosesstrategyB.
SinceinaDSAcontextdifferentplayersarecooperatingtoachievetheirgoals,cooperativegametheoryisappropriate.AssumethereareMdifferentnetworksco-operatinginacontiguousDSAcase,asdepictedinFig.3.3(b).Supposethegoalofeachplayer(i.e.,network)istomaximizeitsrevenuefromtheavailablespectrum,whereeachnetworkhasNm(m=1,2,...,M)unitsofspectrum.EachparticipatingnetworkpredictstheneedforDmunitsofspectrum(ascalculatedfromanticipatedtrafficload)inthenextDSAinterval,where,inthiscase,theGSMchannelsizeof200kHzischosenasthespectrumunit.IfDm>Nmfornetworkm,thatnet-workneedstoborrowDm−Nmunitsofspectrumfromanothernetwork.WelettheborrowingofeachunitofspectrumhaveacostCb,therevenuefromprovidingserviceperunitofspectrumbeRs,andtheinitialwealthofnetworkmbeWm(0).Dependingonthetrafficpatternofeachnetwork,theplayershoulddecidehowmanyspectrumunitstouse,rentorborrowsuchthatitswealthattheDSAintervali
−1
(Dm−Nm)CbWm(i)=Wm(i−1)+DmRs−Dm
Nm−1
ismaximized.Considerasimplescenariowhereanoperatorhas10unitsofspectrum(i.e.,Nm=10),whereborrowingonespectrumunitcosts$0.5,andwheretheinitialwealthoftheoperator,Wm(0)=$1.Fig.3.10plotsthetotaloperatorrevenuesinthenextperiodbasedonthedifferentamountsofrequiredspectrum(Dm)andtherevenueperspectrumunit(Rs).Alsorepresentedistheplanecorrespondingtozerorevenueinordertoidentifythedomains(strategies)thatimproverevenue.
Thechoiceofstrategywouldlikelybebasedonlong-termgoals,suchastomax-imizerevenueafteraspecific(high)numberofDSAperiods.Otherutilityfunctionsmightalsobedefined,dependingontheoptimizationgoal.Moreover,sinceeachplayerisrationalandhe/sheknowsthatotherusersarealsorational,theplayerswillonlyselectastrategythatmaximizestheirutilityamongthesetofpossiblestrategies.
DmNm
3ArchitecturesandProtocolsforDynamicSpectrumSharing83
Ifadominantstrategyexists,thegamewillfinallyreachanequilibriumpoint,knownasaNashEquilibrium(NE)[].AstrategyissaidtobePareto-optimalifthepayoffofnousercanbeimprovedwithoutdecreasingthepayoffofsomeoralloftheotherusers.Dependingonthecharacteristicsofthegame,theNEmightbePareto-optimalornot.ThegamemightalsohavemultipleNEsorPareto-optimalsolutions.
Fig.3.10.TotaloperatorrevenuegiventherevenueperspectrumunitandspectrumdemandperDSAinterval.
Therearedifferentcooperativegamesolutionsthatcanbeusedtoanalyzethisscenario;forexample,NashBargainingSolutions(NBS)provideafair,unique,andPareto-optimalresult[].Ifitisassumedthatwithoutanycooperationbetweentheplayerstheycanachievethepayoffdenotedbythevectorv∈M,thenbycoordinationofresourceusagesthroughtheNBSthepayoffoftheplayersincreasesproportionallytotheirminimumpayoffrequirementvectorv.Therefore,asarguedin[90],theNBSprovidesaproportionalfairnesssolutionparadigm.
DenotetheboundedsetofallfeasiblepayoffallocationsbyF,andtheallocationvector(inM)forthebargainingproblem(F,v)byφ(F,v).ItwasshownthatthereexistsauniqueNBSforabargainingproblem,whichcanbecalculatedby[]
φ(F,v)∈argmax
x∈F,x>v
Ki=1
(xi−vi)
orequivalentlyby[90]
φ(F,v)∈argmax
x∈F,x>v
Ki=1
ln(xi−vi).
SincethesolutionobtainedisuniqueandPareto-optimal,anysetofrationalplayers
willchoosethissameresult.
84O.Hollandetal.
TheplayersintheaboveDSAgamewillcoexistforalongperiodoftime(oftheorderofmonthstomanyyears),hencetheycanlearnfromhistory(i.e.,fromtheoutcomesoftheDSAoperationsinpreviousperiods)astooptimalcoexistencestrategies.Forexample,ifinacertainDSAperiod,oneRSCtriestoincreaseitspay-offbymisguidingtheothernetworksregardingtheresourcesavailableorrequired,otherplayerscanpunishsuchbehaviorbystoppingthecooperativeapproachandbyusingasimilargreedystrategy.Suchgreedybehaviorwill,clearly,resultinthereductionofpayoffforallplayers.Hence,rationalplayerswillavoidmisleadingoneanotherinacooperativegame,andwillbehavehonestly.3.4.2TheDSSCase
FortheDSScase,itispossibletoextendthesamelinesofargumentpresentedforDSA.In[91],ascenarioisoutlinedfornegotiatedDSS,wherebyanadhocnetworkofCRsistryingtoaccesstheresourcesofaprimarycellularOFDM-basednetwork.TheprimarysystemguaranteesalevelofQoSforitsusers,definedasaminimumbitrateandtargetBER.Usingtheachievableratesaspayoffs,andthechoiceofOFDMsub-channelandtransmissionpowerasthestrategy,ithasbeenshownin[91]thataNBSsolutionexistsandthepowerallocationstrategythatachievesthissolutionwasderived.
ThedifferencesbetweennegotiatedDSSandDSAintermsofresourceallocationstrategyareasfollows.First,toavoidinterference,theresourceallocationperiodintheDSScaseisgenerallymuchshorterthanthatinDSA,oftheorderofmillisecondsinsomeextremecases.ThisDSSintervalisgenerallybasedontheprimarysystem’sresourceallocationperiod.Aseconddifferenceisthechoiceofutilityfunction(pay-off).IntheDSAcase,theutilitycouldbealongtermgoalsuchasreducingdroppingorblockingprobability;however,inDSStheutilityisusuallyashorttermgoalsuchasmaximizingtheachievabledatarateatanygiventime.
FurtherreadingontheuseofGameTheoryforspectrumsharingcanbeobtainedfrom,e.g.,[92],inadditiontopreviouslymentionedreferences.
Conclusion
Spectrumisinsignificantdemand,butevidentlyitisusedwithextremelyloweffi-ciency.Tofulfilltherequirementsofpioneeringwirelessservicesandapplications,noveltechnicalsolutions,aswellasmoreflexiblespectrumregulations,areneeded.Thechallengeofimprovingspectralefficiencythroughsecondaryspectrumaccesshasbeenstudiedinthischapter,thefocusbeingonDynamicSpectrumAlloca-tion(DSA)(centralizeddecisionmaking)andDynamicSpectrumSelection(DSS)(largelydecentralizeddecisionmaking).ArchitecturesandprotocolsforspectrumsharinghasbeeninvestigatedintheparadigmsofDSAandDSS(particularlyop-portunisticDSS),andmajorchallengesandopenissueshavebeenidentifiedineachcase.Asafinalcontribution,radioresourceoptimizationbasedonGameTheoryhasbeendiscussed,ofrelevancetoboththeDSAandDSScontexts.
3ArchitecturesandProtocolsforDynamicSpectrumSharing85
Itisclearthatbyimprovingspectrumutilizationthroughspectrumsharing,moreresourcescanbemadeavailabletowirelesstechnologies;moreover,spectrumshar-ingmightbeasourceofcostreductionforoperatorsandserviceproviders,andthereforeawaytoincreaseprofit.Soundeconomicandbusinessmodelsarenev-erthelessrequiredforspectrumsharing,inadditiontosolutionstovarioustechnicalchallenges,inordertomaximizeitsbenefits.Moreover,thequestionofwhattypesofservicescanbesufficientlyprovidedinvarioussharedspectrumenvironmentsmustbeaddressed,andbusinesscasesmustbeinvestigated.
Clearly,farmoreworkisneededtoovercomevariousissues,technicalandpoliti-cal,andparticularlyintheDSSdomain,beforethefullpotentialofspectrumsharingcanberealized.Nevertheless,significantstridesarecurrentlybeingmadetowardsspectrumsharinginthetechnical,regulatory,andstandardizationdomains.Givensuchprogress,itisentirelyreasonabletoexpectsimplifiedsolutionsforbothDSAandDSStobewidelyemployed,inanumberofcountries,withinadecade.
Acknowledgment
SomeoftheworkreportedinthischapterformedpartoftheDeliveryEfficiencyCoreResearchProgrammeoftheVirtualCentreofExcellenceinMobile&PersonalCommunications,MobileVCE(http://www.mobilevce.com).ThisresearchhasbeenfundedbyEPSRCandbytheIndustrialCompanieswhoareMembersofMobileVCE.FullydetailedtechnicalreportsonthisresearchareavailabletoIndustrialMembersofMobileVCE.SomeofthisworkhasbeenperformedwithintheprojectE2RII,whichhasreceivedfundingfromtheEuropeanCommunity’sSixthFrame-workprogramme.Thisworkreflectsonlytheauthors’views;theCommunityisnotliableforanyusethatmaybemadeoftheinformationcontainedherein.Thecontri-butionsofcolleaguesfromtheE2RIIconsortiumareherebyacknowledged.
References
1.USFrequencyAllocationTable,accessiblefromhttp://www.fcc.gov/oet/spectrum/table/,accessedJan.2008.
2.UKFrequencyAllocationTable,accessiblefromhttp://www.ofcom.org.uk/radiocomms/isu/ukfat/,accessedJan.2008.
3.GermanyFrequencyAllocationTable,accessiblefromhttp://www.bundesnetzagentur.de/media/archive/1820.pdf,accessedJan.2008.4.JapanFrequencyAllocationTable,accessiblefromhttp://www.tele.soumu.go.jp/e/search/share/plan.htm,accessedJan.2008.
5.N.Devroye,P.Mitran,andV.Tarokh,“CognitiveDecompositionofWirelessNetworks,”inProc.CROWNCOM2006,Mykonos,Greece,Jun.2006.
6.S.Haykin,“CognitiveRadio:Brain-EmpoweredWirelessCommunications,”IEEEJ.Sel.AreasCommun.,vol.23,no.2,pp.201-220,Feb.2005.
7.A.TonmukayakulandM.Weiss,“SecondaryUseofRadioSpectrum:AFeasibilityAnal-ysis,”inProc.ResearchConferenceonCommunication,Information,andInternetPol-icy,Arlington,VA,USA,Oct.2004.
86O.Hollandetal.
8.J.Mitola,“CognitiveRadioforFlexibleMobileMultimediaCommunications,”inProc.IEEEInt’l.WorkshoponMobileMultimediaCommun.,SanDiego,CA,USA,Nov.1999,pp.3-10.
9.I.Akyildiz,W.Lee,M.Vuran,andS.Mohanty,“NeXtGeneration/DynamicSpectrumAccess/CognitiveRadioWirelessNetworks:ASurvey,”ComputerNetworks,vol.50,no.13,pp.2127-2159,Sep.2006.
10.P.Leavesetal.,“DynamicSpectrumAllocationinCompositeReconfigurableWireless
Networks,”IEEECommun.Mag.,vol.42,no.5,pp.72-81,May2004.
11.3GPP,“NetworkSharing;Architectureandfunctionaldescription(Release6),”3GPPTS
23.251V6.6.0,Mar.2006.
12.IST-DRiVEDeliverableD09,“DynamicSpectrumAllocationAlgorithmIncludingRe-sultsofDSAPerformanceSimulations,”Jan.2002.
13.A.PapoulisandS.U.Pillai,Probability,RandomVariablesandStochasticProcesses
(4thEdition).NewYork:McGraw-Hill,Jan.2002.
14.T.C.Mills,TimeSeriesTechniquesforEconomists.NewYork:CambridgeUniversity
Press,1990.
15.W.Leland,M.Taqqu,W.Willinger,andD.Wilson,“OntheSelf-SimilarNatureofEth-ernetTraffic(ExtendedVersion),”IEEE/ACMTrans.Netw.,vol.2,no.1,pp.1-15,Feb.1994.
16.H.FengandY.Shu,“StudyonNetworkTrafficPredictionTechniques,”inProc.Int’l.
ConferenceonWirelessCommunications,NetworkingandMobileComputing,Wuhan,China,Sep.2005,pp.995-998.
17.S.Fernandes,V.Teichrieb,D.Sadok,andJ.Kelner,“TimeSeriesAppliedtoNetwork
TrafficPrediction:ARevisitedApproach,”inProc.AppliedModelingandSimulation(AMS2002),Cambridge,MA,USA,Nov.2002.
18.O.Hollandetal.,“DevelopmentofaRadioEnablerforReconfigurationManagement
withintheIEEEP1900.4WorkingGroup,”inProc.IEEEDySPAN2007,Dublin,Ireland,Apr.2007,pp.232-239.
19.O.Hollandetal.,“AUniversalResourceAwarenessChannelforCognitiveRadio,”in
Proc.IEEEPIMRC2006,Helsinki,Finland,Sep.2006,pp.1-5.
20.M.Pereirasamyetal.,“DynamicInter-OperatorSpectrumSharingforUMTSFDDwith
DisplacedCellularNetworks,”inProc.IEEEWCNC2005,NewOrleans,LA,USA,Mar.2005,pp.1720-1725.
21.A.AttarandA.H.Aghvami,“AFrameworkforUnifiedSpectrumManagement(USM)
inHeterogeneousWirelessNetworks,”IEEECommun.Mag.,vol.45,no.9,pp.44-51,Sep.2007.
22.M.BuddhikotandK.Ryan,“SpectrumManagementinCoordinatedDynamicSpectrum
AccessBasedCellularNetworks,”inProc.IEEEDySPAN2005,Baltimore,MD,USA,Nov.2005,pp.299-307.
23.P.Leaves,J.HuschkeJ,andR.Tafazolli,“ASummaryofDynamicSpectrumAllocation
ResultsfromDRiVE,”inProc.IST2002,Thessaloniki,Greece,Jun.2002,pp.245-250.24.O.Holland,Q.Fan,andA.H.Aghvami,“ADynamicHierarchicalRadioResourceAllo-cationSchemeforMobileAd-HocNetworks,”inProc.IEEEPIMRC2004,Barcelona,Spain,Sep.2004,pp.1005-1010.
25.R.KulkarniandS.Zekavat,“TrafficAwareInter-VendorDynamicSpectrumAllocation:
PerformanceinMulti-VendorEnvironment,”inProc.Int’l.Conf.onCommun.andMo-bileComputing,Vancouver,Canada,2006,pp.85-.
26.D.Grandblaiseetal.,“CreditTokenbasedRentalProtocolforDynamicChannelAllo-cation,”inProc.CROWNCOM2006,Mykonos,Greece,Jun.2006.
3ArchitecturesandProtocolsforDynamicSpectrumSharing87
27.IEEE802.22Homepage,http://grouper.ieee.org/groups/802/22/,ac-cessedJan.2008.
28.C.Cordeiroetal.,“IEEE802.22:AnIntroductiontotheFirstWirelessStandardbased
onCognitiveRadios,”JournalofCommunications,vol.1,no.1,pp.38-47,Apr.2006.29.IEEE802.16’sLicense-Exempt(LE)TaskGroupHomepage,http://grouper.
ieee.org/groups/802/16/le/index.html,accessedJan2008.
30.J.Huang,R.Berry,andM.Honig,“Auction-BasedSpectrumSharing,”SpringerMobile
NetworksandApplications,vol.11,no.3,pp.405-408,Jun.2006.
31.V.Rodriguez,K.Moessner,andR.Tafazolli,“MarketDrivenDynamicSpectrumAllo-cationoverSpaceandTimeamongRadio-AccessNetworks:DVB-TandB3GCDMAwithHeterogeneousTerminals,”SpringerMobileNetworksandApplications,vol.11,no.6,pp.847-860,Dec.2006.
32.C.Hamacher,“SpectralCoexistenceofDVB-TandUMTSinaHybridRadioSystem,”
inProc.IST2001,Barcelona,Spain,Sep.2001,pp.203-208.
33.M.Mehtaetal.,“ReconfigurableTerminals:AnOverviewofArchitecturalSolutions,”
IEEECommun.Mag.,vol.39,no.8,pp.82-,Aug.2001.
34.D.Grandblaise,K.Moessner,P.Leaves,andD.Bourse,“ReconfigurabilitySupportfor
DynamicSpectrumAllocation:FromtheDSAConcepttoImplementation,”inProc.WorkshoponMobileFutureandSymposiumonTrendsinCommunications,Bratislava,Slovakia,Oct.2003.
35.W.Tuttlebee(ed.),SoftwareDefinedRadio:EnablingTechnologies.London:Wiley,
May2002.
36.IST-DRiVEProjectWebsite,http://www.ist-drive.org,accessedJan.2008.37.IST-OverDRiVEProjectWebsite,http://www.comnets.rwth-aachen.de/$\\
sim$o\\_drive,accessedJan.2008.
38.IST-WINNERProjectWebsite,http://www.ist-winner.org,accessedJan.
2008.
39.IST-E2RProjectWebsite,http://e2r.motlabs.com,accessedJan.2008.
40.MobileVCEConsortiumWebsite,http://www.mobilevce.com,accessedJan.
2008.
41.R.LackeyandD.Upmal,“Speakeasy:TheMilitarySoftwareRadio,”IEEECommun.
Mag.,vol.33,no.5,pp.56-61,May1995.
42.JointTacticalRadioSystem(JTRS)Homepage,http://enterprise.spawar.
navy.mil/body.cfm?type=c\\&category=27\\&subcat=60,accessedJan.2008.
43.DARPAXGProgramHomepage,http://www.darpa.mil/ATO/programs/
XG/,accessedJan.2008.
44.NationalInstituteofInformationandCommunicationsTechnology(NICT)Website,
http://www.nict.go.jp,accessedJan.2008.
45.SDRForumWebsite,http://www.sdrforum.org,accessedJan.2008.
46.I.Akyildiz,S.Mohanty,andJ.Xie,“AUbiquitousMobileCommunicationArchitecture
forNext-GenerationHeterogeneousWirelessSystems,”IEEECommun.Mag.,vol.43,no.6,pp.S29-S36,Jun.2005.
47.S.SouissiandE.Callaway,“MethodandApparatusforDynamicSpectrumAllocation,”
USPatent6553060,Apr.2003.
48.J.Peha,“ApproachestoSpectrumSharing,”IEEECommun.Mag.,vol.43,no.2,pp.
10-12,Feb.2005.
49.Ofcom,“SpectrumUsageRights:TechnologyandUsageNeutralAccesstotheSpec-trum,”Consultation,http://www.ofcom.org.uk/consult/condocs/sur/,2006.
88O.Hollandetal.
50.Ofcom,“SpectrumFrameworkReview,”Consultation,http://www.ofcom.org.
uk/consult/condocs/sfr/,2004-2005.
51.Ofcom,“AStudyintoDynamicSpectrumAccess–FinalReport,”
http://www.ofcom.org.uk/research/technology/overview/emer\\_tech/dsa/dsafinal.pdf,Mar.2007.
52.M.Caveetal.,“ReviewofRadioSpectrumManagement,”forUKDept.TradeandIn-dustryandH.M.Treasury,Mar.2002,accessiblefrom
http://www.ofcom.org.uk/static/archive/ra/spectrum-review/index.htm.
53.M.Caveetal.,“IndependentAuditofSpectrumHoldings,”forUKH.M.Treasury,Dec.
2005,accessiblefromhttp://www.spectrumaudit.org.uk.
54.EuropeanParliamentResolution,“TowardsaEuropeanpolicyonradiospectrum,”
ITRE/6/37236,Feb.2007.
55.FCCETDocketNo.04-167,“PromotingEfficientUseofSpectrumThroughElimination
ofBarrierstotheDevelopmentofSecondaryMarkets,”2004.
56.ComReg,“DynamicSpectrumAccess,”DocumentNo.07/22,Apr.2007.
57.IEEESCC41/P1900Website,http://www.scc41.org,accessedJan.2008.
58.A.P.Hulbert,“SpectrumSharingThroughBeacons,”inProc.IEEEPIMRC2005,Berlin,
Germany,Sep.2005,pp.9-993.
59.D.RaychadhuriandX.Jing,“ASpectrumEtiquetteProtocolforEfficientCoordination
ofRadioDevicesinUnlicensedBands,”inProc.IEEEPIMRC2003,Beijing,China,Sep.2003,pp.172-176.
60.J.MitolaIII,PhDdissertation,“CognitiveRadio:AnIntegratedAgentArchitecturefor
SoftwareDefinedRadio,”RoyalInstituteofTechnology,Sweden,2000.
61.S.Shankar,C.Cordeiro,andK.Challapali,“SpectrumAgileRadios:Utilizationand
SensingArchitectures,”inProc.IEEEDySPAN2005,Baltimore,MD,USA,Nov.2005,pp.160-169.
62.A.Sahai,N.Hoven,andR.Tandra,“SomeFundamentalLimitsonCognitiveRadio,”in
Proc.AllertonConf.onCommun.,Control,andComputing,Champaign,IL,USA,Oct.2004.
63.S.Mishra,A.Sahai,andR.Brodersen,“CooperativeSensingAmongCognitiveRadios,”
inProc.IEEEICC2006,Istanbul,Turkey,Jun.2006,pp.1658-1663.
.A.Sahai,S.Mishra,R.Tandra,andN.Hoven,“SensingforCommunication:TheCaseof
CognitiveRadio,”inProc.AllertonConf.onCommun.,Control,andComputing,Mon-ticello,IL,USA,Sep.2006.
65.A.Sahai,R.Tandra,S.Mishra,andN.Hoven,“FundamentalDesignTradeoffsinCog-nitiveRadioSystems,”inProc.Int’l.WorkshoponTechnologyandPolicyforAccessingSpectrum(TAPAS),Boston,MA,USA,Aug.2006.
66.ITUWorldRadiocommunicationsConference2007(WRC2007),Geneva,Switzer-land,Oct.-Nov.2007,accessiblefromhttp://www.itu.int/ITU-R/index.asp?category=conferences\\&link=wrc-07,accessedJan.2008.
67.N.Devroye,P.Mitran,andV.Tarokh,“AchievableRatesinCognitiveRadioChannels,”
IEEETrans.Inf.Theory,vol.52,no.5,pp.1813-1827,May2006.
68.FCCETDocketNo.03-237,InterferenceTemperature,NoticeofInquiry/NoticeofPro-posedRulemaking,2003.
69.FCCOrderFCC-07-78A1,“IntheMatterofEstablishmentofanInterferenceTemper-atureMetrictoQuantifyandManageInterferenceandtoExpandAvailableUnlicensedOperationinCertainFixed,MobileandSatelliteFrequencyBands,”May2007.
70.D.Cabricetal.,“ACognitiveRadioApproachforUsageofVirtualUnlicensedSpec-trum,”inProc.IST2005,Dresden,Germany,Jun.2005.
3ArchitecturesandProtocolsforDynamicSpectrumSharing
71.T.Chenetal.,“CogMesh:ACluster-BasedCognitiveRadioNetwork,”inProc.IEEE
DySPAN2007,Dublin,Ireland,Apr.2007,pp.168-178.
72.L.Berlemann,S.Mangold,andB.Walke,“Policy-BasedReasoningforSpectrumShar-inginCognitiveRadioNetworks,”inProc.IEEEDySPAN2005,Baltimore,MD,USA,Nov.2005,pp.1-10.
73.BBNTechnologies,“XGPolicyLanguageFramework,”Version1.0,2004.
74.DistributedManagementTaskForce(DMTF)Website,http://www.dmtf.org,ac-cessedJan.2008.
75.N.Damianou,N.Dulay,E.Lupu,andM.Sloman,BookChapter:“ThePonderPolicy
SpecificationLanguage,”inLectureNotesinComputerScience.Springer,2001.
76.J.Strassner,“DEN-ng:AchievingBusiness-DrivenNetworkManagement,”inProc.
IEEE/IFIPNOMS2002,Jeju,Korea,Apr.2002,pp.753-766.
77.A.Attar,A.Sheikhi,H.Abiri,andA.Mallahzadeh,“ANewMethodforCommunication
SystemRecognition,”IranianJournalofScience&Technology,TransactionB,Engi-neering,vol.30,no.B6,pp.775-788,2006.
78.T.WeissandF.Jondral,“SpectrumPooling:AnInnovativeStrategyfortheEnhancement
ofSpectrumEfficiency,”IEEECommun.Mag.,vol.42,no.3,pp.8-14,Mar.2004.
79.T.Weiss,M.Spiering,andF.Jondral,“QualityofServiceinSpectrumPoolingSystems,”
inProc.IEEEPIMRC2004,Barcelona,Spain,Sep.2004,pp.345-349.
80.F.Seeling,“ADescriptionoftheAugust2006XGDemonstrationsatFortA.P.Hill,”in
Proc.IEEEDySPAN2007,Dublin,Ireland,Apr.2007,pp.1-12.81.BBNTechnologies,“TheXGVision,”Version2.0,2004.
82.BBNTechnologies,“TheXGArchitecturalFramework,”Version1.0,2003.
83.IEEE802.11y,http://grouper.ieee.org/groups/802/11/,accessedJan.
2008.
84.FCCETDocketNo.04-186,“UnlicensedOperationintheTVBroadcastBands,”2004.85.FCCETDocketNo.03-108,“CognitiveRadioTechnologiesProceeding,”http://
www.fcc.gov/oet/cognitiveradio/,accessedJan.2008.
86.FCCETDocketNo.06-161,RequestforCommentsonaRequestforaWaiverofPart
15oftheCommissionsRules,2006.
87.FCCDA07-2197,CommentsSoughtonGoogleProposalsRegardingServiceRulesfor
700MHzBandSpectrum,May2007.
88.OfcomCognitiveRadioStudy,accessiblefromhttp://www.ofcom.org.
uk/research/technology/research/emer\\_tech/cograd,accessedJan.2008.
.R.Myerson,GameTheory:AnalysisofConflict(3rdEdition).Cambridge:Harvard
UniversityPress,1999.
90.H.Yaiche,R.Mazumdar,andC.Rosenberg,“AGameTheoreticFrameworkforBand-widthAllocationandPricinginBroadbandNetworks,”IEEE/ACMTrans.Netw.,vol.8,no.5,pp.667-678,Oct.2000.
91.A.Attar,M.R.Nakhai,andA.H.Aghvami,“CognitiveRadioGame:AFrameworkfor
Efficiency,FairnessandQoSGuarantee,”inProc.IEEEICC2008,Beijing,China,May2008.92.M.Halld´orsson,J.Halpern,L.Li,andV.Mirrokni,“OnSpectrumSharingGames,”in
Proc.ACMSymposiumonPrinciplesofDistributedComputing,St.Johns,Canada,Jul.2004.
因篇幅问题不能全部显示，请点此查看更多更全内容
Copyright © 2019- gamedaodao.com 版权所有 湘ICP备2022005869号-6
违法及侵权请联系：TEL:199 18 7713 E-MAIL:2724546146@qq.com
本站由北京市万商天勤律师事务所王兴未律师提供法律服务