DO AALOHA Slotted ALOHA for Ad Hoc Netw orking Using Smart Antennas Harkirat Singh and - PDF document

DOA-ALOHA:SlottedALOHAforAdHocNetworkingUsingSmartAntennasHarkiratSinghandSureshSingh,DepartmentofComputerSciencePortlandStateUniversity,Portland,OR97201harkirat,singh
@cs.pdx.eduAbstract—ThispaperdevelopsanovelslottedALOHAproto-col(Direction-Of-ArrivalALOHA)foruseinadhocnetworkswherenodesareequippedwithsmartantennas.TheprotocolreliesontheabilityoftheantennaandDOAalgorithmstoidentifythedirectionofthedesiredsignalandthedirectionoftheinterfererstomaximizeSINR(SignaltoInterferenceandNoiseRatio)atthereceiver.TheperformanceoftheprotocolisevaluatedusingjointsimulationinOPNETandMatlab.WecomparetheperformanceofournewprotocolagainstrecentdirectionalMAC(MediumAccessControl)[3]protocol.WeshowthatDOA-ALOHAachievessignicantlyhigherthroughputthan[3]despiteitssimplicity.Theimpactofusingdifferentnumberofantennaelementsisalsostudiedforthisenvironment.I.INTRODUCTIONRecently,therehasbeenincreasinginterestindevelopingMACprotocolsforuseinadhocnetworkswherenodesareequippedwithdirectionalantennas.Antennamodelsusedincludesectoredxedbeamantennas,idealizedadaptivear-rayantennas,andsteerabledirectionalantennas.Aspreviousresearchershaveshown,usingdirectionalantennasincreasesthroughputbecauseofbetterspatialreuseofthespectrum(see[10],[1],[3],[6]).However,wenotethatthesepreviousworkshavenotfullyexploitedthebenetsofadaptivearrayantennas(orsmartantennas)suchastheabilitytoformnullsinthedirectionofinterferers(resultinginhighSINR)andtheabilitytodeterminethedirectionoftransmitters(DirectionofArrival).Weshowthatbyexploitingthesecapabilitiesofsmartantennas,asimpleprotocolcanyieldthroughputsthatare2x–4xhigherthanoneoftherecentprotocols[3].Wealsonotethatoursimulationsuserealisticantennamodelsunliketheidealizedmodelsusedinmany(withtheexceptionof[6])papersand,despitethis,ourprotocoloutperformsmostoftheseexistingprotocols.Thekeyprobleminexploitingthecapabilitiesprovidedbyasmartantennaatareceiverisindeterminingthedirectionoftheinterferingsignals(soastoplacenullsinthosedirections)andformingabeamtowardsthetransmitter.Inordertodothis,wehavedevelopedamodiedversionoftheslottedALOHAprotocolinwhichasmallinitialportionoftheslotisusedforndingthedirectionofvarioustransmitters.Thisisdonebyrequiringeachtransmittertotransmitapuretonetowardsitsintendedreceiverforashortintervalpriortotransmittingthepacket.ThereceiverrunsaDOA(DirectionOfArrival)algorithmwhichprovidesinformationThisworkisfundedbytheNSFundergrantANIR-0125728.aboutthereceivedsignalstrengthanddirectionofthedifferenttransmitters.Thisinformationisthenusedateachreceivertoguidebeamforming(beamandnulls)fortheremainingdurationoftheslot.Uponcorrectpacketreception,areceiversendsanACKusingthealreadyformedbeams.WeimplementedsmartantennainMatlabandinterfaceditwiththephysicallayerofOPNET.Inourstudyweusedrealisticantennapatternswiththesidelobes.Wecarriedoutextensivesimulationsandobtainveryhighthroughputinthesingleaswellasmulti-hopcase.Theremainderofthepaperisorganizedasfollows:inthenextsectionwedescribeoursystemmodel.SectionIIIdescribesthecontextforthisworkinrelationtootherresearch.SectionIVdescribeourprotocolDOA-ALOHAinmoredetail.SectionVsummarizeskeyresults.II.SYSTEMMODELWeassumethateachnodeisequippedwithasmartantennasystemwhichiscomposedofalineararrayofelements.Forsimplicityweassumethattheantennaarrayisperpendiculartothex-yplaneinwhichthenodeslie.Thereasonforthisassumptionisthatthebeamformedbytheantennaissymmetricabouttheantennaaxisandisthusindependentofthedirectioninwhichanodeis“facing”1.Figure1providesaschematicofanadaptivearraysmartantennasystem.Asillustratedinthegure,theantennaelementsareseparatedfromeachotherbyaknowndistance.Weassumethatthetransmitterislocatedfarenoughawayfromthereceiverthatallthesignals\n arrivingatthedifferentantennaelementsareparallel.However,sincetheelementsareseparatedbydistance,thephaseofthedifferentsignalsisdifferent.Let\rdenotethephaseandgainthatisaddedtoeachsignal\n .Then ,theoutputsenttothereceiver,canbewrittenas,\n \n ! " $#!%\n '&)(+*-,+.(0/2143'57698! " where䀀:;=isthephasepropagationfactor,@isthewavelength,#isanarbitrarygainconstant,and " isAWGNnoise.Theweights\rusedinthispaperonlyshiftthephaseofthesignalandleavetheamplitudeuntouched.1Ifweplacetheantennaarrayinthex-yplanethenthebeampatternformedinsomedirectionwilldependontherelativealignmentoftheantennawithrespecttothexandyaxes.Thismakestheanalysisfarmoredifcultwithoutgaininganyadditionalgenerality. Ifthedesireddirectionis
,thentherepresentationfortheweightsis,&*,+.(/1'8Foramorecomprehensivediscussion,pleasesee[5].InFigure2weshowthedifferentantennapatternsformedbyalineararrayof\n \relementswhen
.Wenotethatasthenumberofelementsincreases,thebeamwidthbecomesnarroweranddirectivityoftheantennaincreases.Further,wenotethat,ratherthanonebeam,usingalineararrayresultsintwobeamsthatcanleadtogreaterinterference.Aswenotedintheintroduction,anotherbenecialfeatureofsmartantennasistheabilityoftheseantennastoformnullsingivendirections.Infact,givenelements,anantennacanformupto\rnulls.However,theshapeofthedesiredbeamcanchangedependingonthenumberofandthedirectionofthenulls.InthisworkweusetheMMSE(MinimumMeanSquareError)algorithmtodetermineweightstoformnullsappropriately[5].Figure3illustratestwocaseswhenusing antennaelementswith!beingthedesireddirection.Intherstcase,weareformingonlytwonullswhereasinthesecondcaseweareformingsixnulls.Ascanbeseen,theshapeofthebeamandsidelobeschanges.wwww123MSReceiverS1(t)S2(t)S3(t)SM(t)dVariable gain andAntenna elementsSignal received fromqFig.1.Schematicofasmartantenna(adaptivelineararray).III.RELATEDWORKRecentlytherehavebeenseveralpapersthathavelookedattheproblemofMACdesignforadhocnetworkswherenodesareequippedwithdirectionalantennas.Thedirectionalantennamodelsusedincludeswitchedbeamantennas(thean-tennaissectoredandoneofthesesectorsisuseddependingthedirectionofthecommunicatingnode),multi-beamantennas(heremorethanonebeamcanbeusedsimultaneously),andadaptiveantennaarrays(herethebeamcanbemadetopointinanydirectionasdescribedinsectionII).[2]developsslottedscheduling-basedMACprotocolsfornodesequippedwithdirectionalantennas.Thedirectionalantennaconsideredisamulti-beamadaptivearrayantenna(MBAA)whichiscapableofformingmultiplebeams.Theprotocolsassumethatnodescanengageinseveralsimulta-neoustransmissions.Thekeycontributionofthepaperisthedevelopmentofaneighbortrackingschemethatisthenusedtoscheduletransmissionsbyeachnodeinadistributedway. 0.005 0.01 0.0153021060240902701203001503301800Array Pattern with 8 Elements DOA = 45 DegreesTheta(Degree)Normalized Pattern 0.005 0.01 0.015 0.02 0.0253021060240902701203001503301800Array Pattern with 16 Elements DOA = 45 DegreesTheta(Degree)Normalized PatternFig.2.Antennapatternsfor8and16antennaelements.[11]alsodevelopsMACprotocolsforthisscenario.Theyassumethatthegainofthedirectionalantennaequalsomnidirectionalantenna.Anantennaiscomprisedof4antennaelementswhichcantransmitin90degsectorsonly.NodesaredependentonGPSorsomeotherdevicetohavepositioninformationofthesenderandreceiver.TheypresentaMACprotocolbasedonDirectionalRTS/CTSoracombinationofDRTS/O-CTS(OmnidirectionalCTS).Theprotocolassumesthatifatransmissionishappeninginsomedirectionthenitwilldeferalltransmissionsinthatdirection.Similarly,[4],[12]consideraswitchedbeamantennamodelwithnonullingorDOAinformation.Theyuseadirectionalantennaatthereceiver.AsecondprotocolbasedonDRTS/DCTSassumestwoseparatechannels:onefordataandanotherforsignaling.Thesepapersdonotconsidernullingornarrowerbeamwidths.In[3]anovelmulti-hopRTSisproposedtoestablishlinksbetweendistantnodesand1(directionalgain)isassumedtobehigherthan(omnidirectionalgain).ThedirectioninwhichthemainlobeistobeorientedisdeterminedbytheMACprotocol(whichinturnisprovidedthisinformationbythenetworklayerwhichisassumedtobeneighbor-aware).Theauthorsnotethatnodealignmentnegatesthebenetsachievedduetodirectionalantennas,however,unaligned 0.2 0.4 0.6 0.8 13021060240902701203001503301800Desired = 45 deg, Nulls = 25, 70 deg 0.2 0.4 0.6 0.8 13021060240902701203001503301800Desired = 45 deg, Nulls = 10, 20, 30, 70, 80, 90 degFig.3.Patterswith8antennaelementsand2or6nulls.routesenhancesthespatialreuse.Theyshowthattheirprotocolhasa4-5xthroughputascomparedwith802.11.In[8],[7]theauthorsassumethateachnodemaintainsneighborAngle-SINRtable(AST)andtheyprovidealinkstatebasedtable-drivenroutingandMACprotocol.BasedonASTanodecalculatesanafnityforananglewhichprovidesmaximumSINR.BasedonthisaNLSTableisformed.Nodesinthebeamformedregionremainintheomnimodebuttheymakenullsinthedirectionofongoingtransmissions.[9]usesdirectionaltransmissionsforcontrolanddatapack-ets.Itusesadirectional-NAVtablefortransmissionschedulingandcollisionavoidance.However,theydonotexploitthecapabilitiesofthesmartantennas,suchasbeamsteeringandtheplacementofnullsinthedirectionofinterferers.In[6]anodecachesAOAinformationbasedonsignalsreceivedandnodesremaininpromiscuousmodetocachesignals.802.11specicationssaythatRTSneedstobetrans-mitted7times,soanodewilltransmit4directionalRTSandremainingthe3asomni-directionalRTSifthereisnoresponsetothedirectionalRTS.Acircularantennawith6elementsisassumed,andanodeiscapableofelectronicallysteeringtheboresighttowardsaspecicdirection.Aconstantbeamwidthof45degassumed.However,itwasobservedthatastheboresightchanges,thesidelobepatternchangesdrastically.Thelessonisthattheeffectsofsideandbacklobescannotbeignoredintheevaluationofnetworkperformancewithdirectionalantennas.[6]showsthatusinganidealantennaresultsinamaximumthroughputof2.2Mbpswhileusingarealisticantennahasamaximumthroughputofonly1.4Mbps.Thisfact,thatantennapatternsmatterinevaluatingMACbehavior,isonethathaslargelybeenignoredbyagreatmanyauthors.Followingthelessonof[6],weuserealisticantennapatternsinourstudies.IV.PROTOCOLDESCRIPTION:DOA-ALOHAInthissectionwedescribethebehaviorofourprotocol.However,beforedoingthis,weneedtomakethefollowingassumptions:(1)Nodesareawareoftheangularlocationofeachoftheirneighbors(asin[3])sincethisinformationisneededattransmitterstoformdirectedbeamstowardsreceivers;(2)Forsimplicity,weassumethatallnodesusethesameconstanttransmitpower.Packet TransmissionDOAACKOne SlotFig.4.StructureofaslotinDOA-ALOHA.Figure4showstheformoftheslotsusedinDOA-ALOHAandasshown,eachslotisbrokenintothreeminislots.Ouralgorithmworksasfollows:1)TherstminislotinaslotiscalledtheDOA-minislotanditisherethatanodeidentiestheangulardirectionofalltransmittersthatitcanhear.Alltransmitterstrans-mitasimpletone(i.e.,asinewave)duringtheDOA-minislottowardstheirintendedreceivers.Thesignalreceivedatsomereceiveristhusthecomplexsumofallofthesetones.ThereceiverrunsaDOAalgorithm(suchasMUSIC[5])todeterminetheangulardirectionofeachofthetransmittersandthereceivedpowerfromeachtransmitter.2)OnceareceiverdeterminestheDOAofalltransmittersitcanhear,itformsitsdirectedbeamtowardstheonethathasthemaximumpowerandformsnullsinalltheotheridentieddirections.3)Thesecond(andlargest)minislotisthepackettransmis-sionslotanditisherethatthepacketsaretransmitted.Afterthereceiverhasformeditsbeamandnullsasde-scribedabove,itreceivesthepacketfromthetransmitter.Afterreceivingthepacket,itlooksattheheaderandrejectsthepacketifitwasnottheintendeddestination.4)ThelastminislotistheACKslotwherethereceivertransmitsanACKusingthealreadyformedbeamtothesender(ifthepacketwasnotrejectedandcorrectly received).WhenatransmitterdoesnotreceiveanACK,itretransmitsthepacketatalatertime(thisisexactlyasinALOHA).abdcNode d mistakenly formsa beam towards a becausea's signal is strongerthan b's signal at da has a packet for cb has a packet for dFig.5.Falsebeamforming.Theintuitionbehindthereceiverbeamforminginthedirec-tionofthemaximumsignalisthat,becauseofthedirectivityoftheantenna,thereisahighprobabilitythatitistheintendedrecipientforthepacket.However,wenotethatincases,asinFigure5,thereceiverincorrectlybeamformstowardsbecause'ssignalisstrongerthan
's.Whilethisisnotaseriousprobleminmostcases,wecanenvisionscenarioswherethe
transmissiongetsstarvedduetoalargevolumeoftrafc.Anoptimizationwehavethereforeimplementedisasingle-entrycacheschemewhichworksasfollows:Ifanodebeamformsincorrectlyinagiventimeslot,itremembersthatdirectioninasingle-entrycache.Inthenextslot,ifthemaximumsignalstrengthisagaininthedirectionrecordedinthesingle-entrycache,thenthenodeignoresthatdirectionandbeamformstowardsthesecondstrongestsignal.–Ifthenodereceivesapacketcorrectly(i.e.,itwastheintendedrecipient),itdoesnotchangethecache.–Ifitreceivesapacketincorrectly,itupdatesthecachewiththisnewdirection.Ifthereisnopacketinaslotfromthedirectionrecordedinthecache,thecacheisreset.ThissimplemechanismensuresthatincasessimilartoFigure5,connectionsarenotstarved.However,wecanconstructmorecomplexscenarioswhereasingle-entrycachewillfailtopreventstarvation.Inthesecases,moresophisticatedmultiple-entrycachingschemesarerequired.However,inoursimula-tions,weonlyusethesingle-entrycachingschemebecausetheprobabilityofmorecomplexscenariosresultinginstarvationareveryrare.V.RESULTSOPNETprovidesanexcellentphysicallayermodelbuthasadrawbackinthatithasaveryidealisticdirectionalantennamodel.Toovercomethisdrawbackweimplementedthesmartantennamodel(foralineararrayofantennaelements)inMatlabandinterfaceditwiththephysicallayerofOPNET.WeinvokeMatlabtocompute's(sectionII)basedonactualreceivedsignalstrength97 ateachantennaelementTABLEIOPNETSIMULATIONPARAMETERS.SimulationParametersBackgroundNoise+ambientNoise-143dBPropagationmodelFreespaceBandwidth1,000kHzMinfrequency2,402MHzDataRate2000kbpsCarrierSensingThreshold+3dBMinimumSINR9dBBitErrorBasedonBPSKModulationcurveMaximumradiorange250masobtainedfromOPNET.WealsomodiedOPNET'sradiopipelinestagewiththesimulationparametersdisplayedinTableI.WeevaluatetheperformanceofDOA-ALOHAusing5x5mesh(asusedin[3])withfourpre-denedows.Figure6showsthenetworktopologyandowsusedfortwooftheseexperiments.Forthethirdexperiment,weusedarandomnodeplacementonthegridwhereanode'spositionisrandomlyshiftedinthex-axisandy-axisbyaddingadisplacementrandomlyselectedfrom[-150m,+150m]andtheowsareasinFigure6(b).ThetrafcisCBR(ConstantBitRate)whichincreases(perow)from75kbpsto2Mbps.Thepacketsizeis512bytes.Figure7plotstheaggregatethroughputasafunctionofthedatarateofoneow(forFigure6(a))fortwoantennasystems–onewith8elementsandonewith16elements.Figure8doesthesameforFigure6(b)andFigure9correspondstotherandommeshtopologycase.Weused10differentcasesforrandomows(Figure6(b))andrandomlyselectednodes.Inordertomakethecomparisonasfairaspossible,weusedtheexactsameparametersinourexperimentsasthosedescribedin[3].(a) Four flows (some alignment)(b) Randomly selected flowsFig.6.5x5gridtopologyusedtocompareperformancewith[3].Weobservethatusing16antennaelementsasopposedto8elementsmakesabigdifferenceinaggregatethroughput.Thisisbecausethebeamwidthwhenusing16elementsissmallerthanwhenusing8elementswhichresultsinmoresimultaneoustransmissions/slot.FortheowsinFigure6(a),(whenowsarealigned),wedidnotnoticemuchdifferenceintheperformanceof16and8antennaelementsbutforFigure6(b)andforrandomtopologieswedoseeasignicantdifference.Thereasonisthatwhenowsarenotaligned,thereisagreaterpotentialforspatialreusewith16antennaelements(duetoitssmallerbeamwidth).Wenotethatthe TABLEIIMeshFigure6(a)MeshFigure6(b)RandomMesh16Elements8Elements16Elements8Elements16Elements8Elements(
)()(
)()(
)()DOA-ALOHA2500kbps21503000200042503750[3](
\n )800kbps100010000200400600800100012001400160018002000050010001500200025003000Sending rate (Tx) vs Aggregate Throughput Sending Rate (Kbps) Aggregate Throughput (Kbps)8 elements(a) Fig.7.Performanceofourprotocolin6(a)02004006008001000120014001600180020000500100015002000250030003500Sending rate (Tx) vs Aggregate Throughput Sending Rate (Kbps) Aggregate Throughput (Kbps)8 elements(b) Fig.8.Performanceofourprotocolin6(b)beamwidthusedin[3]is
.Inourcase,thelineararraycreatestwosymmetricbeamsandwedenebeamwidthforourprotocolasthesumofthesetwobeams.For16antennaelements,wenotedanaveragebeamwidthof
foreachofthetwosymmetricbeamsformedwithalineararraythusgivingusaneffectivebeamwidthof
(addingthetwobeams).Theeffectivebeamwidthwhenusing8antennaelementsisapproximately\r
.TableIIsummarizesourresultsandcomparesthemwith[3].Weobservethatourprotocolis2x–3xbetterwhenweuse8elementsandismuchbetter(3x–4x)for16elements.VI.CONCLUSIONInthispaperwehavepresentedDOA-ALOHA,aversionofslottedALOHAthatusesDOAinformationatthereceivertobeamforminawaythatmaximizesSINR.Wecomparetheperformanceofourprotocolagainst[3]andshowthatourprotocolhasathroughputof2x–4xhigherthanthe[3].In0200400600800100012001400160018002000050010001500200025003000350040004500Sending rate (Tx) vs Aggregate Throughput Sending Rate (Kbps) Aggregate Throughput (Kbps)8 elementsFig.9.Performanceofourprotocolinrandomgridtopologies.thefutureworkwewillexaminetheperformanceof802.11whenusingsmartantennas.ACKNOWLEDGEMENTSWewouldliketothankOPNETfortheirtechnicalsupport.REFERENCES[1]J.YouA.Nasipuri,S.YeandR.Hiromoto.Amacprotocolformobileadhocnetworksusingdirectionalantennas.InIEEEWCNC,2000.[2]LichunBaoandJ.J.Garcia-Luna-Aceves.Transmissionschedulinginadhocnetworkswithdirectionalantennas.InACM/SIGMOBILEMobiCom2002,23–28Sep2002.[3]RomitRoyChoudhury,XueYang,RamRamanathan,andNitinH.Vaidya.Usingdirectionalantennasformediumaccesscontrolinadhocnetworks.InACM/SIGMOBILEMobiCom2002,23–28Sep2002.[4]ZhuochuanHuangandChien-ChungShen.Acomparisonstudyofomnidirectionalanddirectionalmacprotocolsforadhocnetworks.InIEEEGlobecom2002,2002.[5]J.C.LibertiandT.S.Rappaport.SmartAntennasforWirelessCommunications.PrenticeHall,1999.[6]RajivBagrodiaMineoTakai,JayMartinandAifengRen.Directionalvirtualcarriersensingfordirectionalantennasinmobileadhocnet-works.InACM/SIGMOBILEMobiHoc2002,Oct2002.[7]S.Roy,D.Saha,S.Bandyopadhyay,T.Ueda,andS.Tanaka.Anetwork-awaremacandroutingprotocolforeffectiveloadbalancinginadhocwirelessnetworkswithdirectionalantenna.InACMMobihoc'03,1–3June2003.[8]S.HorisawaS.Bandyopadhyay,K.HausikeandS.Tawara.Anadaptivemacanddirectionalroutingprotocolforadhocwirelessnetworksusingesparantenna.InACM/SIGMOBILEMobiHoc2001,Oct2001.[9]GentianJakllariThanasisKorakisandLeandrosTassiulas.Amacprotocolforfullexploitationofdirectionalantennasinad-hocwirelessnetworks.InACMMobihoc'03,1–3June2003.[10]JamesWardandJr.R.T.Compton.Improvingtheperformanceofaslottedalohapacketradionetworkwithanadaptivearray.IEEETransactionsonCommunications,40(2):292–300,February1992.[11]V.ShankarkumarY.B.KoandN.H.Vaidya.Mediumaccesscontrolprotocolsusingdirectionalantennasinadhocnetworks.InIEEEINFOCOM2000,March2000.[12]ChavalitSrisathapornphatZhuochuanHuang,Chien-ChungShen.Amacprotocolbasedondirectionalantennaandbusy-toneforadhocnetworks.InIEEEMILCOM2002,2002.