USF Libraries
USF Digital Collections

A data link layer in support of swarming of autonomous underwater vehicles

MISSING IMAGE

Material Information

Title:
A data link layer in support of swarming of autonomous underwater vehicles
Physical Description:
Book
Language:
English
Creator:
Jabba Molinares, Daladier
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Acoustic communications
MAC sublayer
Logical link control sublayer
Multichannel communications
OFDMA
Hidden Markov Model
Dissertations, Academic -- Computer Science and Engineering -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Communication underwater is challenging because of the inherent characteristics of the media. First, common radio frequency (RF) signals utilized in wireless communications cannot be used under water. RF signals are attenuated in such as way that RF communication underwater is restricted to very few meters. As a result, acoustic-based communication is utilized for underwater communications; however, acoustic communication has its own limitations. For example, the speed of sound is five orders of magnitude lower than the speed of light, meaning that communications under water experience long propagation delays, even in short distances. Long propagation delays impose strong challenges in the design of Data Link Layer (DLL) protocols. The underwater communication channel is noisy, too. The bit error rate (BER) can also change depending on depth and other factors, and the errors are correlated, like in wireless communications.As in wireless communications, transducers for acoustic communication are half duplex, limiting the application of well-known detection mechanisms in Medium Access Control (MAC) layer protocols. Further, known problems like the hidden and exposed terminal problem also occur here. All these aspects together make the underwater communication channel to have the worst characteristics of all other known channels. Because of these reasons, underwater scenarios are complicated to implement, especially when they have underwater autonomous vehicles exchanging information among them. This dissertation proposes data link layer protocols in support of swarming of underwater autonomous vehicles that deal with the problems mentioned before. At the MAC sublayer, a MAC protocol called 2MAC is introduced. 2MAC improves the throughput of the network using the multichannel capabilities of OFDM at the physical layer. At the logical link control sublayer, a protocol named SW-MER is proposed.SW-MER improves the throughput and the reliability combining the well-known stop and wait protocol with the sliding window strategy, and using an exponential retransmission strategy to deal with errors. 2MAC and SW-MER are evaluated and compared with other protocols using analytical means and simulations. The results show that by using 2MAC, packet collisions are considerably reduced and the throughput improved. In addition, the use of SW-MER improves the packet delivery ratio over existing mechanisms. In general, the evaluations indicate that the proposed data link layer protocols offer a better communication alternative for underwater autonomous vehicles (UAV) than traditional protocols.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Daladier Jabba Molinares.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 153 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002069471
oclc - 608494828
usfldc doi - E14-SFE0003261
usfldc handle - e14.3261
System ID:
SFS0027577:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200409Ka 4500
controlfield tag 001 002069471
005 20100903135844.0
007 cr mnu|||uuuuu
008 100422s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003261
035
(OCoLC)608494828
040
FHM
c FHM
049
FHMM
090
TK7885 (Online)
1 100
Jabba Molinares, Daladier.
2 245
A data link layer in support of swarming of autonomous underwater vehicles
h [electronic resource] /
by Daladier Jabba Molinares.
260
[Tampa, Fla] :
b University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 153 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Communication underwater is challenging because of the inherent characteristics of the media. First, common radio frequency (RF) signals utilized in wireless communications cannot be used under water. RF signals are attenuated in such as way that RF communication underwater is restricted to very few meters. As a result, acoustic-based communication is utilized for underwater communications; however, acoustic communication has its own limitations. For example, the speed of sound is five orders of magnitude lower than the speed of light, meaning that communications under water experience long propagation delays, even in short distances. Long propagation delays impose strong challenges in the design of Data Link Layer (DLL) protocols. The underwater communication channel is noisy, too. The bit error rate (BER) can also change depending on depth and other factors, and the errors are correlated, like in wireless communications.As in wireless communications, transducers for acoustic communication are half duplex, limiting the application of well-known detection mechanisms in Medium Access Control (MAC) layer protocols. Further, known problems like the hidden and exposed terminal problem also occur here. All these aspects together make the underwater communication channel to have the worst characteristics of all other known channels. Because of these reasons, underwater scenarios are complicated to implement, especially when they have underwater autonomous vehicles exchanging information among them. This dissertation proposes data link layer protocols in support of swarming of underwater autonomous vehicles that deal with the problems mentioned before. At the MAC sublayer, a MAC protocol called 2MAC is introduced. 2MAC improves the throughput of the network using the multichannel capabilities of OFDM at the physical layer. At the logical link control sublayer, a protocol named SW-MER is proposed.SW-MER improves the throughput and the reliability combining the well-known stop and wait protocol with the sliding window strategy, and using an exponential retransmission strategy to deal with errors. 2MAC and SW-MER are evaluated and compared with other protocols using analytical means and simulations. The results show that by using 2MAC, packet collisions are considerably reduced and the throughput improved. In addition, the use of SW-MER improves the packet delivery ratio over existing mechanisms. In general, the evaluations indicate that the proposed data link layer protocols offer a better communication alternative for underwater autonomous vehicles (UAV) than traditional protocols.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Miguel A. Labrador, Ph.D.
653
Acoustic communications
MAC sublayer
Logical link control sublayer
Multichannel communications
OFDMA
Hidden Markov Model
0 690
Dissertations, Academic
z USF
x Computer Science and Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.3261



PAGE 1

A Data Link Layer In Support Of Swarming Of Autonomous Underwater Vehicles by Daladier Jabba Molinares A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Computer Science & Engineering College of Engineering University of South Florida Major Professor: Miguel A. Labrador, Ph.D. Ken Christensen, Ph.D. Wilfrido A. Moreno, Ph.D. Rafael Perez, Ph.D. William Stark, Ph.D. Date of Approval: October 16, 2009 Keywords: acoustic communications, mac sublayer, logical link control sublayer, multichannel communications, ofdma, hidden markov model Copyright 2009 Daladier Jabba Molinares

PAGE 2

Tomyfamily

PAGE 3

AcknowledgementsIwouldliketoacknowledgemyadvisorDr.MiguelLabrador,f orhisdirection,support, patienceandconsiderablecontributionstomyresearch.Th eenvironmenthehasforhis students,lledwithfreedom,resourcesandchallenges,ha screatedanidealresearch space.IwouldliketothanktheUniversityofSouthFlorida,Univer sidaddelNorteandLASPAUCOLCIENCIASforsupportingmystudies.Iwouldalsoliketos incerelythankthemembersofmycommitteeDr.Perez,Dr.Christensen,Dr.Stark,a ndDr.Morenofortheir valuablecommentsandsuggestions.Ithankallmyfriendsfortheircontinuousencouragementan dsupport;withthesepeople Iengagedindiscussions,workedonprojectsandexperiment s,andreceivedtheirhelp. Wordsarenotenoughtoexpressmythankstomyfamilyforthei rconstantandunconditionalsupportandlove,butespeciallytoSalome,mypiec eofheaven,andtomywife Milena,whoencouragedmetofulllthisdream.Sheiseveryt hingtome. Finally,IthankGodforguidingmylifeandmydreams,withou tHimnothingwouldbe possible.

PAGE 4

TableofContentsListofTables vii ListofFigures viii Abstract xiii Chapter1Introduction 1 1.1UnderwaterNetworks 1 1.1.1StaticTwo-DimensionalUnderwaterAcousticNetwork s forOceanBottomMonitoring3 1.1.2StaticThree-DimensionalUnderwaterAcousticNetworksforOcean-ColumnMonitoring4 1.1.3Three-DimensionalNetworksofAUVs51.1.4AUVNetworks 6 1.2UnderwaterCommunicationsandtheDataLinkLayer81.3MACSublayerinUnderwaterAdHocNetworks121.4LogicalLinkControlSublayerinUnderwaterAdHocNetworks 16 1.5Contributions 16 1.6OrganizationoftheDissertation18 i

PAGE 5

Chapter2LiteratureReview 19 2.1OrthogonalFrequency-DivisionMultipleAccess(OFDMA )19 2.2MediumAccessControl(MAC)Layer21 2.2.1ChannelClassication222.2.2NetworkArchitecture242.2.3UnderwaterMACProtocols25 2.2.3.1ANetworkingProtocolforUnderwaterNetworks[17]25 2.2.3.2ModiedMediaAccessControlDesignfor theAcoustic-BasedUnderwaterDigitalDataCommunication[18]27 2.2.3.3AdaptedMACAtoUnderwaterAcoustic Networks[19]28 2.2.3.4SlottedFAMA,ProtocolforUnderwater AcousticNetworks[21]30 2.2.3.5UWAN-MAN,anEnergy-EfcientMAC ProtocolforUnderwaterAcousticWirelessSensorNetworks[22]32 2.2.3.6AHybridMediumAccessControlProtocol forUnderwaterWirelessNetworks[3]34 2.2.3.7T-Lohi,aNewClassofMACProtocolsfor UnderwaterAcousticSensorNetworks[24]35 2.2.3.8AMACProtocolforUnderwaterSensor Networks[26]37 2.3LogicalLinkControl(LLC)Layer38 2.3.1LogicalLinkControlClassication39 ii

PAGE 6

2.3.1.1ARQProtocols392.3.1.2ForwardErrorCorrectionProtocols422.3.1.3HybridProtocols43 2.3.2UnderwaterLogicalLinkControlProtocols43 2.3.2.1OptimizationofaDataLinkProtocol[35]432.3.2.2AMulti-HopARQProtocol[37]452.3.2.3FEC-BasedReliableDataTransportProtocol forUnderwaterSensorNetworks[38]46 Chapter32MAC:AMultichannelMACProtocol49 3.12MACDescription 49 3.2NetworkTopologyandChannelAssignment513.3Scenario 53 3.42MACStateTransitions 54 3.4.1IdleState 54 3.4.2ChannelAssignmentState573.4.3Contention(Listen)State583.4.4WaitingforCTSState593.4.5ReceivingRTSState 60 3.4.6WaitingforACKState623.4.7WaitingforDataState623.4.8BackoffState 63 3.4.9AdjustedResponseState63 iii

PAGE 7

3.4.10BlockedtoSendState64 3.52MAC'sBackoffAlgorithm 64 Chapter4SW-MER:AStopandWaitWindow-BasedLogicalLinkC ontrol ProtocolWithExponentialRetransmissions684.1SW-MERDescription 69 4.2SW-MERStateTransitions 72 4.2.1SelectingPossiblePackets724.2.2UpdatingCopiesofEachPacketintheWindow734.2.3Listen 74 4.2.4WaitingforCTS 75 4.2.5SendingWindowofPackets754.2.6WaitingforACKVector754.2.7IdentifyingPacketsSentWithErrors764.2.8ReceivingRTS 76 4.2.9PreparingtheCTStobeSent764.2.10WaitingtoReceivePackets774.2.11CheckingPackets 77 4.2.12EnqueuingPacketswithoutErrors774.2.13VerifyingPacketstobeSent784.2.14GeneratingtheACKVector78 Chapter5AnalyticalModelsfortheProposedDataLinkLayer Protocols79 5.1AnalyticalModelfor2MAC 80 iv

PAGE 8

5.1.12MACRTS/CTSTransmissionProcess815.1.2ThroughputAnalysis83 5.1.2.1PacketTransmissionProbability835.1.2.2SaturationThroughput91 5.1.3ModelValidation 98 5.2AnalyticalModelforSW-MER100 5.2.1SW-MERTransmissionProcess1005.2.2ThroughputAnalysis1025.2.3ModelValidation 104 Chapter6PerformanceEvaluation 107 6.1ScenariosandParameters 107 6.2ChannelErrorModels 108 6.3PerformanceEvaluationfor2MAC1106.4PerformanceEvaluationforSW-MER1116.5PerformanceEvaluationfortheEntireDataLink116 Chapter7AnAdaptiveLogicalLinkSublayerProtocolinResp onsetoUnderwaterAcousticCommunication(UAC)ChannelChanges1207.1AdaptiveSW-MERDescription1217.2DeterminingtheChannelQuality1217.3StatesoftheAdaptiveLogicalLinkProtocol125 7.3.1Listen 126 7.3.2VerifyingtheStateoftheChannel126 v

PAGE 9

7.3.3UpdatingChannelStatusHistory1277.3.4UpdatingCopiesofPacketstobeSent1277.3.5WaitingtoReceivePackets1277.3.6CheckingPackets 128 7.3.7EnqueueingPacketsWithoutErrors1287.3.8VerifyingPacketstobeSenttotheUpperLayer1297.3.9GeneratingtheACKVector1297.3.10IdentifyingPacketsSentWithErrors129 7.4SW-MERProtocolExample1317.5ChannelErrorModels 132 7.6PerformanceEvaluation 132 7.6.1SimulationParameters1327.6.2ThroughputEvaluation133 Chapter8ConclusionsandFutureWork 140 8.1Conclusions 140 8.2FutureWork 142 ListofReferences 143 Appendices 149 AppendixA:TheStationaryDistributionoftheMarkovChain 150 AppendixB:TheThroughputEfciencyofSW-MER152 AbouttheAuthor EndPage vi

PAGE 10

ListofTablesTable1.1Nodepowerconsumption[3]. 9 Table5.1Notationsusedtoobtainthesaturationthroughpu t.93 Table5.2Parametervaluesusedtovalidatethesaturationt hroughput.99 Table5.3Notationsusedtoobtainthethroughputefciency .101 Table5.4Parametervaluesusedtovalidatethethroughpute fciency.105 Table6.1Parametervaluesusedtoevaluate2MACandSW-MER. 108 Table7.1Simulationparameters. 133 vii

PAGE 11

ListofFiguresFigure1.1Statictwo-dimensionalunderwateracousticnet works[1].4 Figure1.2Staticthree-dimensionalunderwateracousticn etworks[1].5 Figure1.3Three-dimensionalnetworksofautonomousunder watervehicles[1].6 Figure1.4AUVnetworks. 7 Figure1.5Reectioninwater. 10 Figure1.6Refractionofsoundinwater. 10 Figure1.7Hiddenterminalproblem. 13 Figure1.8Exposedterminalproblem. 13 Figure1.9Captureproblem. 14 Figure1.10Deafnessproblem. 15 Figure2.1OFDMA. 21 Figure2.2ComparisonbetweenOFDMandOFDMA.21Figure2.3ClassicationofMACprotocolsbasedonthenumbe rofchannels.22 Figure2.4ClassicationofMACprotocolsbasedontheirnet workarchitecture. 23 Figure2.5TopologyDiscoveryMessage(TDM)Propagationin anunderwater acousticnetwork[17]. 26 viii

PAGE 12

Figure2.6AdaptationofMACAprotocoltounderwateracoust iccommunications[20]. 29 Figure2.7SlottedFAMA[21]. 31 Figure2.8Hybridprotocolframestructure[3].34Figure2.9AMACprotocolusingRTS/CTShandshaking[26].38Figure2.10Classicationoflogiccontrolprotocols.39Figure2.11One-dimensionaln-hopacousticchannel[37].4 5 Figure2.12Two-dimensionalmulti-hopacousticsensornet work[37].45 Figure3.12MACtransmissionprocess. 50 Figure3.2Lineartopology. 52 Figure3.3Hexagontopology. 52 Figure3.4Nonagontopology. 53 Figure3.5Finitestatemachineof2MACprocessatthesender .55 Figure3.6Finitestatemachineof2MACprocessatthereceiv er.56 Figure3.72MAC,transmissiondiagram. 57 Figure3.8Contentionprocess. 60 Figure3.9RTS/CTScommunicationprocess.61Figure3.10Differentbackoffsusingpacketsize=150bytes and BER = 1 x 10 3 .65 Figure3.11Differentbackoffsusingpacketsize=300bytes and BER = 1 x 10 4 .65 Figure3.12Throughputwithdifferentbackoffalgorithms. 67 Figure4.1Firsttransmission,M=6andpackets2and3arrive witherrors.70 Figure4.2Secondtransmission,M=6andpacket3arrivewith errorsagain.70 ix

PAGE 13

Figure4.3Finitestatemachineoftheprocessinthesendera ttheLLClayer.73 Figure4.4Finitestatemachineoftheprocessinthereceive rattheLLClayer.74 Figure5.1Markovchainforthebackoffprocess[46].84Figure5.2Transmissionperiodsinanerror-pronechannel. 92 Figure5.3Throughputcomparisonwith BER = 1 x 10 3 .99 Figure5.4Throughputcomparisonwith BER = 1 x 10 4 .100 Figure5.5SW-MER,transmissionprocess. 101 Figure5.6Throughputefciencycomparisonwithawindowsi zeof8packets.105 Figure5.7Throughputefciencycomparisonwithawindowsi zeof16packets. 106 Figure5.8Throughputefciencycomparisonwithawindowsi zeof32packets. 106 Figure6.1Two-stateMarkovmodelrepresentation.109Figure6.2ThroughputoftheMACprotocols,packetsizeof12 00bits.111 Figure6.3ThroughputoftheMACprotocols,packetsizeof24 00bits.111 Figure6.4Throughputefciencyforawindowsizeof8datapa cketsusingthe Bernoullierrormodel. 113 Figure6.5Throughputefciencyforawindowsizeof16datap acketsusinga Bernoullierrormodel. 113 Figure6.6Packetdeliveryrateforawindowsizeof8datapac ketsusingthe Bernoullierrormodel. 114 Figure6.7Packetdeliveryrateforawindowsizeof16datapa cketsusingthe Bernoullierrormodel. 114 Figure6.8Throughputefciencyforawindowsizeof16packe tsusingthe shallowwatererrormodel. 115 x

PAGE 14

Figure6.9Packetdeliveryrateforawindowsizeof16datapa cketsusingthe shallowwatererrormodel. 115 Figure6.10ThroughputofthecombinedSW-MERand2MACproto colsusing packetsizeof1200bits,andBER=1 x 10 3 .117 Figure6.11ThroughputofthecombinedSW-MERand2MACproto colsusing packetsizeof1200bits,andMarkovmodelchannel.118 Figure6.12ThroughputofthecombinedSW-MERand2MACproto colsusing packetsizeof2400bits. 119 Figure7.1Transmissionbetweennodeslocatedatdifferent depths.122 Figure7.2Four-statemachinerepresentation.122Figure7.3Physicalanddatalinklayer,interaction.124Figure7.4Finitestatemachineofthesenderprocess.125Figure7.5Finitestatemachineofthereceiverprocess.126Figure7.6Exampleoftheproposedadaptivestopandwaitsli dingwindowbasedmechanism. 130 Figure7.7Lineartopology. 133 Figure7.8CopiesperpacketusingaBernoullierrormodel,a windowsizeof8 packets,apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof 2400bps. 134 Figure7.9CopiesperpacketusingaBernoullierrormodel,a windowsizeof8 packets,apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof 19600bps. 135 Figure7.10Copiesperpacketusingashallowwatererrormod el,awindowsize of8packets,apacketsizeof1200bits,anddatarateof2400b ps.135 Figure7.11ThroughputusingaBernoullierrormodel,awind owsizeof8data packets,apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof 2400bps. 137 xi

PAGE 15

Figure7.12ThroughputusingaBernoullierrormodel,awind owsizeof8data packets,apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof 19600bps. 137 Figure7.13Copiesperpacketusingashallowwatererrormod el,awindowsize of16packets,apacketsizeof2400bits,anddatarateof1960 0bps.138 Figure7.14ThroughputusingaBernoullierrormodel,awind owsizeof16data packets,apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof 19600bps. 138 Figure7.15Throughputusingashallowwatererrormodel,aw indowsizeof8 packets,andapacketsizeof1200bits.139 Figure7.16Throughputusingashallowwatererrormodel,aw indowsizeof8 packets,andapacketsizeof2400bits.139 xii

PAGE 16

ADataLinkLayerinSupportofSwarmingofAutonomousUnderw aterVehicles DaladierJabbaMolinares ABSTRACT Communicationunderwaterischallengingbecauseoftheinh erentcharacteristicsofthe media.First,commonradiofrequency(RF)signalsutilized inwirelesscommunications cannotbeusedunderwater.RFsignalsareattenuatedinsuch aswaythatRFcommunicationunderwaterisrestrictedtoveryfewmeters.Asaresult ,acoustic-basedcommunicationisutilizedforunderwatercommunications;however,a cousticcommunicationhasits ownlimitations.Forexample,thespeedofsoundisveorder sofmagnitudelowerthan thespeedoflight,meaningthatcommunicationsunderwater experiencelongpropagation delays,eveninshortdistances.Longpropagationdelaysim posestrongchallengesinthe designofDataLinkLayer(DLL)protocols.Theunderwatercommunicationchannelisnoisy,too.Thebit errorrate(BER)canalso changedependingondepthandotherfactors,andtheerrorsa recorrelated,likeinwirelesscommunications.Asinwirelesscommunications,trans ducersforacousticcommunicationarehalfduplex,limitingtheapplicationofwellknowndetectionmechanisms inMediumAccessControl(MAC)layerprotocols.Further,kn ownproblemslikethe hiddenandexposedterminalproblemalsooccurhere.Allthe seaspectstogethermakethe underwatercommunicationchanneltohavetheworstcharact eristicsofallotherknown xiii

PAGE 17

channels.Becauseofthesereasons,underwaterscenariosa recomplicatedtoimplement, especiallywhentheyhaveunderwaterautonomousvehiclese xchanginginformation amongthem.Thisdissertationproposesdatalinklayerprotocolsinsup portofswarmingofunderwater autonomousvehiclesthatdealwiththeproblemsmentionedb efore.AttheMACsublayer,aMACprotocolcalled2MACisintroduced.2MACimprov esthethroughputof thenetworkusingthemultichannelcapabilitiesofOFDMatt hephysicallayer.Atthe logicallinkcontrolsublayer,aprotocolnamedSW-MERispr oposed.SW-MERimproves thethroughputandthereliabilitycombiningthewell-know nstopandwaitprotocolwith theslidingwindowstrategy,andusinganexponentialretra nsmissionstrategytodeal witherrors.2MACandSW-MERareevaluatedandcomparedwith otherprotocolsusing analyticalmeansandsimulations.Theresultsshowthatbyusing2MAC,packetcollisionsareco nsiderablyreducedand thethroughputimproved.Inaddition,theuseofSW-MERimpr ovesthepacketdelivery ratiooverexistingmechanisms.Ingeneral,theevaluation sindicatethattheproposeddata linklayerprotocolsofferabettercommunicationalternat iveforunderwaterautonomous vehicles(UAV)thantraditionalprotocols. xiv

PAGE 18

Chapter1:Introduction1.1UnderwaterNetworksTheoceanisanscenariothathasstartedtogeneratelotofin terestintheresearchers. Therearedifferentsituationsunderseathatcanbeexplore d,anditisnecessarytohave anadequateunderwaternetworktocollecttheinformationt hatwillbeprocessedlater fordifferentpurposes.Obtaininginformationunderseass uchasbyexploration,tactical surveillanceordatacollectionisnecessaryfordisasterp revention,civilianapplicationsor formilitarytactics,andunderwaterdevicesaretheresour cesthatmustbeusedinorder tocollectthedata.Todothis,underwaternetworksmustbed eployedintheareasofthe oceanwhichneedtobeexamined.Someofthescenarioslikeex plorationanddatacollectionunderwater,mayrequireautonomousvehiclesmovin g,collectingandexchanging informationamongthem.InordertosupportswarmingofAuto nomousUnderwaterVehicles(AUVs),itisveryimportanttohaveconstant,andrelia blecommunicationamongthe AUVs.Theproblemisthatcommunicationunderwaterisveryc hallenging. First,radiofrequency(RF)signalsutilizedinwirelessco mmunicationscannotbeused underwaterbecauseelectromagneticwavesdonotpropagate wellinthatmedium.RF signalsareattenuatedinsuchwaythatRFcommunicationund erwaterwouldberestricted toveryfewmeters.Asaresult,acoustichasbeenusedinunde rwatercommunications. 1

PAGE 19

However,acousticcommunicationhasitsownlimitationsas well.Forexample,thespeed ofsoundisnotconstantunderwater;itvarieswithdepth,sa linity,andotherfactors.Also, thespeedofsoundisveordersofmagnitudelowerthanthesp eedoflight(radio),meaningthatcommunicationsunderwaterexperiencelongpropag ationdelays,eveninshort distances.Inaddition,theunderwatercommunicationchannelisnoisy andthebiterrorrate(BER) alsochanges,dependingonthedepthandotherfactors.Norm ally,theerrorsarecorrelated,likeinwirelesscommunications.Similarly,theamo untandtypeoferrorsrequire newdatalinklayerprotocols.Finally,asinwirelesscommunications,transducersforac ousticcommunicationarehalf duplex,limitingtheapplicationofwell-knowndetectionm echanismsinMAClayerprotocols.Alltheseaspectstogethermaketheunderwatercomm unicationchanneltheworst mediumcomparedwithothernetworks,andimposestrongchal lengesinthedesignof DataLinkLayer(DLL)protocols.Therearedifferentapplicationsinwhichunderwaternetwo rkscanbeapplied,andthe architecturedesignthatsupportsthosenetworkscanvaryd ependingontheparticular characteristicsofthecorrespondingscenario.Someofthe applicationsforunderwater communicationsarethefollowing: Environmentmonitoring Humanactivitiesaffectingthemarineecosystemsuchasche micalanalysisofindustrialwaste Underseaexplorations Detectionofunderwateroilelds 2

PAGE 20

Disasterprevention Monitoringofoceancurrentsandwinds,forexampletsunami s Assistednavigation Locationofdangerousrocksinshallowwaters Distributedtacticalsurveillance Intrusiondetection Forthedesignofunderwatercommunications,threeaccepte darchitecturesinunderwater communicationsaredescribedin[1].Thesearchitecturesa re:statictwo-dimensional underwateracousticnetworksforoceanbottommonitoring, staticthree-dimensional underwateracousticnetworksforocean-columnmonitoring ,andthree-dimensionalnetworksofAUVs.Afourtharchitectureinthisdissertationis presented,similartothethird onein[1],andwillbenamedasAUVnetworks.Allthesearchit ecturesareexplainedin thefollowingsections.1.1.1StaticTwo-DimensionalUnderwaterAcousticNetwork sforOceanBottom Monitoring AsshowninFigure1.1,clustersarebuiltbygroupsofnodesa ndoneofthenodesin everyclusterrepresentsacluster-head.Nodesinthisnetw orkareanchoredtothebottom oftheoceanandtheycancommunicatedirectlywiththeclust er-headorviamulti-hop. Therearetwotypeofcommunications,horizontalandvertic al.Thehorizontalisbetween thenodesinaclusterwiththeircluster-head.Thevertical isthecommunicationbetween 3

PAGE 21

thecluster-headofeveryclusterwiththesurfacestation. Thecluster-headiscalledthe underwatergateway.Inthisarchitecturethebasestationn otonlycanbeequippedwith acousticlinksbutalsowithradiofrequency(RF)communica tionsorsatellitetransmitter nr r r rrrrn r rnnr rr r r Figure1.1:Statictwo-dimensionalunderwateracousticne tworks[1]. tocommunicatewiththeonshoresinkand/orthesurfacesink .Therearesomeproblems presentedinthisarchitecture,likethepowernecessaryto transmitfromnodestothe underwatergatewayorfromthistothesurfacestation.Also multi-hoppathscanincrease boththecomplexityoftheroutingprotocolsandthesignali ngoverheadoverthenetwork. 1.1.2StaticThree-DimensionalUnderwaterAcousticNetwo rksforOcean-Column Monitoring Inthesenetworksallnodesarestatic,andareusedincaseth escenariocannotbewell modeledusingsensorslocatedatthebottomoftheocean.The principaldifferencefrom therstarchitectureisthatthisarchitecturedoesnotbui ldclustersandthereforethere isonlyverticalcommunication,asshowninFigure1.2.Here ,nodescanbeatdifferent 4

PAGE 22

n r rn r n n Figure1.2:Staticthree-dimensionalunderwateracoustic networks[1]. depths(usingoatingbuoysorwithcablesanchoredtothebo ttomoftheocean)tocollect information.Dataissenttothesurfacestationdirectlyor viamultihoppingthroughthe restofthenodes.Someproblemspresentedinthisarchitect urearesuchthatbuoysare vulnerabletoweather,canobstructshipsnavigatingonthe surface,andtheycanbeeasily detectedbytheenemyifthescenarioisformilitarypurpose s. 1.1.3Three-DimensionalNetworksofAUVsAsseeninFigure1.3,nodesinthesenetworksarealsoanchor edtothebottomofthe sea.Theycanbeatdifferentdepthscollectinginformation ,butthisinformationisstored untilAUVscancollectit.AUVsareequippedwithantennas,w hichrequiresnetwork coordinationamongthevehicles.Inthisarchitectureisne cessarytohavealgorithmsthat guaranteethecoordinationbetweentheAUVandthesensors, andtheircorrectfunctionalityinthescenario. 5

PAGE 23

n r n rn r n Figure1.3:Three-dimensionalnetworksofautonomousunde rwatervehicles[1]. 1.1.4AUVNetworksThisisthemostadequatearchitectureforthedatalinklaye rproposedinthisdissertation. AUVsareequippedwithcommunicationsystemstocommunicat ewitheachotherfor swarmingpurposes,andsensorstocollecttheinformationr equiredinscenariossuch asdetectingmine-likeobjectsandoilelds,collectingwa terqualityinformationand detectingintruders.Sensorseitheranchoredtotheoceanb ottomorattachedtobuoys arenotneededinthisarchitecture,asseeninFigure1.4.Th eapplicationusedinthis dissertation,aswarmofAUVsforseaexploration,hasthefo llowingdescription: SmallnumberofAUVs(lessthan10) AUV'sspeedistypicallylimitedto3-5knots Depthofaround200m 6

PAGE 24

Figure1.4:AUVnetworks. AUVsassumetheexistenceofalocalizationsystemthatthey willuseforautonomous navigation LowrateinformationexchangeamongtheAUVsfortelemetry, coordination,and planning ASyntheticApertureSonar(SAS)sensingandprocessingdev iceisassumed SASswathwidthof250misassumed Amaximumcommunicationrangeof1kmisassumedalthoughthe SASswath widthwilldeterminetherealvalue Forthisapplicationthecontinuousandreliabletransmiss ionoflocationdataamongAUVs formaintainingtheformationisofgreatimportance.Foral lthesearchitectures,itis necessarytohavereliableMACandLLCprotocolsinordertoe stablishadequatecommunicationamongthenodesinthenetwork.Itisimportantno tonlytosavetimeatthe 7

PAGE 25

momentofaccessingthemediumanddecidingwhichnodeisgoi ngtostartsendingdata, butalsotoreducethenumberofcollisionsatthemomentofse ndingpackets,andpacket lossduetothebiterrorrate,whichishighinunderwaternet works.Further,newmechanismsareneededtoimprovethethroughputduetolongpropag ationdelays. Thedatalinklayerproposedinthisdissertationaddresses theseproblems.First,anew multichannelMACsublayerprotocoltailoredtoworkwithth eOrthogonalFrequencyDivisionMultipleAccess(OFDMA)technologyisintroduced .TheproposedMACprotocolreducescollisionsandimprovesthethroughputofthe network.Second,anewLLC protocolisproposedtoimprovethereliabilityandthethro ughputofthetransmission.In general,thegoalsofthenewdatalinkprotocolsaretoreduc epacketcollisions,improve thethroughput,achievebetterchannelutilization,andin creasethepacketdeliveryratio. 1.2UnderwaterCommunicationsandtheDataLinkLayerThedatalinklayerisresponsibleforgroupingthebitsfrom thephysicallayerintological chunksofdatacalledframes,providingthemeanstotransfe rdatabetweentwoadjacent usersinanetwork,anddetectingandcorrectingpossibleer rorsthatcouldoccurinthe physicallayer.Thislayerissplitintotwosublayersandea chonehasitsowncharacteristicsandtaskstoexecute.ThesublayersaretheMAClayerand theLLClayer. Thefollowingareimportantissuestoconsiderinthedesign ofanunderwaterdatalink layer: Halfduplexradios.Sincenodescannottransmitandreceive atthesametime,stop andwaitprotocolsarethemostcommonlyimplementedLLCpro tocols. 8

PAGE 26

Burstchannelerrors.Inunderwatercommunicationserrors arecorrelated[2], meaningthattheyoccurinbursts. Energyconsumption.Energyconsumptionisacriticalissue notonlyinwireless butalsoinunderwatercommunicationsystems,sinceunderw atertransceivershave largepowerratios,andbatteriesarenoteasilyreplaceabl easonland.Table1.1 comparesthepowerconsumptionofthreedevices:amodemfor underwatercommunications(WHOI),RFsensor(Mica2),andRFwirelesscomp uternetworks (CiscoAironet). Table1.1:Nodepowerconsumption[3]. State Underwater RFSensor RFComputer Tx 50W 80mW 2.24W Rx 3W 30mW 1.35W Idle 80mW 30mW 1.35W Temporalandspatialuctuationoffrequency(spacetimeun certainty).Signal variesaccordingtothechannelgeometry(spatial)andinti me(temporal)inunderwatercommunications.Duetothehighlatencymedium(sl owpropagationin acousticchannels),collisionshappennotonlywithconcur renttransmissions,but alsowithtransmissionsatdifferenttimeanddistance. Propagationdelays.Largepropagationdelaysreducethene tworkthroughputand channelutilization,anditisevenworsewhentheexchangeo fseveralcontrolpacketsisrequiredtoestablishthecommunicationbetweennode sinanetwork.This imposesimportantchallengesinthedesignofefcientdata linklayerprotocols. Longtermfading.Oceanstructuressuchasreefsandmountai ns,largeships,offshoreunderwateroildrillingequipmentandinternalwaves overalongperiodof time,generatelongtermfadinginwater[4]. 9

PAGE 27

nnr nr rn nnr Figure1.5:Reectioninwater. Figure1.6:Refractionofsoundinwater. Shorttermfadingbecauseofmultipathanddelayspread.Inm ultipathpropagation anddelayspread,thereceivedsignalpowerchangesasafunc tionoftimebecause ofreection,diffraction,andscattering,causingsignal fadingforshortperiodsof time.Inunderwatercommunications,multipathformationd ependsnotonlyonthe reectiononthesurface(eitherbottom,onthewatersurfac e,oronanyobjectsin between),butalsoonadirectpathbetweentransmitterandr eceiver,andonrefractionorbendingofraysdependingonthedepth,pressureandt emperatureofwater atdifferentwaterlevels[4].Figures1.5and1.6represent thereectionsofrays andtherefractionscenariosinwater. 10

PAGE 28

ShorttermfadingbecauseoftimeselectivityandDopplersp read.Thisshortterm fadinggivesrandomsignaluctuationortimevariationone achchannelresponse. Thistimevariabilityiscausedbythescatteringonthewave sofsurfacewaterin shallowwatersituationsandscatteringondeepseawavesin deepwater.Inaddition,themotionofthetransduceralsocontributestoDoppl erspread[5].Onthe otherhand,becauseofthewayraysarrive(raysarriveatdif ferentangles),different Dopplershiftsappear,alsocreatingshorttermfading. InterSymbolInterference(ISI).Thisistotallydifferent fromnoise;itissignal distortioninwhichonesymbolinterfereswithsubsequents ymbols.Thisisan undesiredphenomenon,astheprevioussymbolshaveasimila reffectasnoise, thusmakingthecommunicationlessreliable.ISIisusually causedbymultipath propagationandtheinherentnon-linearfrequencyrespons eofachannel[6]. Synchronization.Achievingsynchronizationinunderwate rcommunicationsis challengingbecauseofthelongpropagationdelays. Bandwidthavailability.Bandwidthislimitedinunderwate rcommunicationscomparedwithwireless.Eventhoughthereisalongrangesystem operatingoverseveraltensofkilometers,thisislimitedtoafewkHzofbandwi dth[4].Forthatreason,MACprotocolsneedtobeefcient. Pathloss.Pathlossrepresentstheattenuationofthepropa gatedinformationsignal overdistance.Thisattenuationinwaterforhighfrequency radio,especiallyin electricallymoreconductivesaltwater,isextremelyhigh .Ifitisassumedthatan averageconductivityfromseawateris4mhos/meters,and0. 05mhos/metersin freshwater(tapwater),asaconsequencetheattenuationfo r2.4GHzwillbearound 1695dB/metersinseawater,and189dB/metersinfreshwater [4].Inunderwa11

PAGE 29

tercommunications,RFsignalsaremoreattenuatedthanino thermedia,making acousticcommunicationtransmissionthemethodofchoice. 1.3MACSublayerinUnderwaterAdHocNetworksMACprotocolsworkasaninterfacebetweentheLLCsublayera ndthephysicallayer. TheyprovideaphysicaladdresscalledtheMACaddress,anda naddressingmechanism thatisauniqueserialnumberforeverynodeinthenetwork,a llowingsuccessfuldelivery ofpacketsinasharednetwork.Theyalsoprovidechannelacc esscontrolmechanisms requiredforsharingthecommonchannelinthenetwork(mult ipleaccessprotocol).In general,MACprotocolsregulatetheaccessofanumberofnod estoasharedmedium, andareinchargeofprovidingfairaccesstoalltheuserssha ringacommonmedium whileachievingefcientchannelutilization.AMACprotoc olcontrolswhenanodecan sendapackettoanode(viaunicast)ortoasetofnodes(viamu lticastorbroadcast). Dependingonthetypeofnetwork,technologyutilized,netw orktopology,application requirements,andotherfactors,datalinklayerprotocols havetodealwithaspectssuch asQualityofService(QoS)support,energyefciency,sync hronization,anderrorcontrol (detectionandrecovery).Also,therearesystemconstrain tsthatareimportanttoconsider inthedesignofMACprotocols,suchasthehalfduplexnature ofthetransceivers,underwaterlocalizationofthenodesandtheavailablebandwidth Themostimportantissuestosolveinordertodesignunderwa terMACprotocolsarethe following: Hiddenterminalproblem.Thehiddenterminalproblemoccur swhenanodeis withintherangeoftheintendeddestinationbutoutofrange ofthesender.InFigure 12

PAGE 30

Figure1.7:Hiddenterminalproblem. Figure1.8:Exposedterminalproblem. 1.7,nodeAtransmitstoB,nodeCcannothearA,andalsotrans mitstoB;asa result,thereisacollisionatB.Thehiddenterminalproble mcanbesubstantially reducedbyusinganRTS/CTShandshake. Exposedterminalproblem.Theexposedterminalproblemocc urswhenanode ispreventedfromsendingpacketstoothernodesbecauseane ighborisalready transmitting.InFigure1.8,nodeBistransmittingtonodeA andnodeChasa packettotransmittoD;asaconsequence,Ccannottransmitt hedata,although nodeDcanreceivethetransmissionwithoutinterferencebe causeitisoutofrange ofB.Theexposedproblemcanbesolvedbyseparatingcontrol packetsanddata packetsindifferentchannels. 13

PAGE 31

Figure1.9:Captureproblem. Synchronization.SomedistributedMACprotocolsneedasyn chronizationmechanism.Theseprotocolsdonothaveacentralnodetocontrolwh etherthechannelis alreadybusyornot,whichnodeisusingthatchannel,andfor howlongthechannel willbeused.Alsoitisnecessarytohaveasynchronizationm echanismtocontrol whenallthenodesinthenetworkgotothesleepingmodetosav eenergy. Captureproblem.Acaptureproblemiswhentwonodessendpac ketssimultaneouslytothesamereceiverandthisnodecleanlyreceivesthe informationfromone ofthem,suchasinFigure1.9.NodesAandCsendpacketsatthe sametimeto nodeBbutbecausethesignalstrengthfromCishigherthanAa tB,Cpackets willbedecodedwithouterrors[7].Thissituationcancreat eunfairsharingofbandwidth,sinceCwillhavethechannelallthetime. Deafnessproblem.Anotherproblemoccurswhenanodetriest otransmitpacketsbuttheintendedreceiveriscurrentlytransmittingina ninactivechannelofa thirdnode(directionalantennasareused),andthesenderh astowaituntilthereceiverisavailable;thisproblemiscalledthedeafnesspro blem,asseeninFigure 1.10.Inthisgure,thereisanetworkwithatleastthreenod esA,B,C;Bstarts 14

PAGE 32

Figure1.10:Deafnessproblem. tosendDATAtoCaftersendingRTSandCTScontrolpacketstoe achother.Ais notawareofcommunicationbetweenBandCandattemptstosen danRTSpacket controltonodeBeventhoughnodeBisdeaftonodeA. Asmentioned,duetothecharacteristicsandthecomplicati onspresentedinthemedium, theMACsublayerinunderwatercommunicationsisacritical protocolthatmustassure communicationamongthenodesinaneffectiveandoptimalwa y,soastoreducecollisions,wasteoftimeandtoomuchenergyconsumption.Thes eproblemsstarttoincreasewhenthenumberofnodesthatcommunicatewitheachot heratthesametime isincreased.Intheliteraturesomesolutionsarepresente dusingnotonlyonebutmultichannelswithoneormultipleantennas.TheMACprotocolpro posedinthisdissertation providesanefcientsolutionthatcanbescalabletosevera lnodes,havingformations eitherlinearorpolygonal,withoutgeneratingmorecollis ionswhennodescommunicate witheachotheratthesametime. 15

PAGE 33

1.4LogicalLinkControlSublayerinUnderwaterAdHocNetwo rks TheLLCisasublayerthatbelongstothedatalinklayerinthe OSImodel.Theunderwateracousticmedium,likewirelessnetwork,isahalf-dup lexcommunicationmedium. Forthisreason,mostoftheprotocolsdesignedforunderwat ercommunicationsarebased onmodicationstothewell-knownstopandwaitAutomaticRe peat-ReQuest(ARQ) scheme.TheLLChasthefollowingfunctions: Implementsowcontrol,usingeitherthewell-knownstopan dwaitprotocol,orthe SlidingWindowstrategy. Implementserrorcontrol. Ensuresthatdataistransferredcorrectlybetweenadjacen tnodesintheunderwater network. Themostimportantissuesinthislayeraretoprovidearelia bledeliveryofthepackets, owcontrol,errordetectionandcorrection,betterthroug hputandchannelutilization[8]. Thelogicallinkprotocolproposedinthisdissertationuse sexponentialpacketretransmissionsguaranteeingamorereliablecommunicationfortheun derwatercommunication, andcombinesthestopandwaitschemewithawindowpackettra nsmissiontoimprove thethroughputandchannelutilization.1.5ContributionsTheprimaryresearchcontributionofthisdissertationisa newdatalinklayerinsupport ofswarmingofautonomousunderwatervehicles.Thisnewdat alinklayerisdividedinto 16

PAGE 34

twonewprotocolsworkingtogether,oneattheMACsublayera ndtheotheroneatthe LLCsublayer.Thecontributionsineachlayerare: AmultichannelMAClayerprotocolcalled2MAC,tailoredtow orkwithOFDMA technologyatthephysicallayer,isdesignedtoreducetheh iddenterminalproblem, exposedterminalproblem,captureproblem,deafnessprobl emandfairness,andimprovethethroughputoftheunderwaterchannel.Withthispr otocol,thenumberof nodesinthenetworkcanbeincreasedwithoutincreasingthe numberofcollisions. This2MACprotocolispresentedin[9]. Astopandwait,window-basedLLCprotocolcalledSW-MERisd esignedtoimprovethethroughputinunderwaterchannels.Also,SW-MERi ncludesanexponentialretransmissionstrategytoimprovethepacketdeli veryratio.TheSW-MER protocolisdescribedin[9]. Othercontributionsare: AnewbackoffalgorithmattheMAClayerthatismoreadequate forunderwater communications.Theproposedbackoffalgorithmisintrodu cedin[9]. Ananalyticalmodeltocalculatethesaturationthroughput of2MAC. Ananalyticalmodeltocalculatethethroughputefciencyo fSW-MER. AnimprovementofSW-MERthatadaptsitstransmissionwindo winresponseto underwateracousticcommunicationchannelchanges.Thead aptiveSW-MERis introducedin[10]. Anunderwaterchannelerrormodelgeneratedfromsynthetic tracesthatcorrespond tomovingnodesinshallowwater[9]. 17

PAGE 35

1.6OrganizationoftheDissertationTheremainderofthedissertationisorganizedasfollows.C hapter2reviewstheexisting literatureindatalinklayerprotocolsforunderwateracou sticcommunications.Chapter 3presentsthenewMACprotocol.Chapter4describesthenews topandwait,windowbasedprotocolfortheLLClayer.Chapter5presentstheanal yticalmodelforeachofthe proposeddatalinkprotocolsandacomparisonbetweenthean alyticalandexperimental evaluations.Chapter7introducesanimprovementofthepro posedlogicallinkprotocol, byadaptingthetransmissionwindowtodifferentchannelco nditions.Detailsofthescenariosusedfortheproposeddatalinkprotocols,theadapti veprotocolversion,andtheir performanceevaluationsusingsomecomparisonswithother traditionaldatalinklayer protocols,arepresentedinchapter6.Finally,chapter8su mmarizesthedissertationand presentsdirectionforfutureresearch. 18

PAGE 36

Chapter2:LiteratureReviewInthischapter,theOFDMAschemeisintroduced,andthepurp oseofhavingOFDMA atthephysicallayerisexplained.Later,anoverviewofMAC protocolsforunderwater communicationsispresentedincludingthetaxonomyandabr iefdescriptionofthemost importantprotocolsineachcategory.Similarly,thenexts ectionpresentsLLCprotocols forunderwatercommunications,includingtheclassicati onandsomeofthemostimportantprotocolsavailableintheliterature.2.1OrthogonalFrequency-DivisionMultipleAccess(OFDMA ) ThereareissuesinunderwaterMACprotocolssuchasthehidd enandexposedterminal, capture,anddeafnessproblem,thatwhencombinedwiththec haracteristicsofthechannelmaketheperformanceofacousticcommunicationsverypo or.Theseproblemscan besolvedtoimproveperformanceefciencyintermsofthrou ghput.In[11]itisshown whatisnecessarytogettheseimprovementsandsomeresults thathavebeenachieved. Ontheotherhand,insystemswhereonlyonecarrierisused,a singlefadewillcause thelinktofail,buthavingamulticarriersystemonlysomes ubcarrierswillbeaffected. OrthogonalFrequencyDivisionMultiplexing(OFDM)isamul ticarriertransmissionin whichasingledatastreamistransmittedthroughseverallo werratesubcarriers.OFDM 19

PAGE 37

splitsormultiplexestheradiosignalintoseveralsmallsu b-signalsthataretransmitted atthesametimeatdifferentfrequenciestothereceiver.Th edatacarriedisdividedinto severalparalleldatastreams,oneforeachsubcarrier,whe reeachsubcarrierismodulated withaconventionalmodulationschemeatalowsymbolorbitr ate.OFDMincreases theefciencyofdatacommunicationsbyincreasingthedata throughputsincethereare severalspacedsubcarriersmodulated.SystemssuchasthosethatuseOFDMaregoodcandidatesforun derwatercommunicationsinwhichhighdatarateisinvolvedwiththeexistenceo fhighdelayspreadchannels.Theycanprovidethesameorevenbetterperformanceef ciency,butofferalowercomplexityimplementationthanunderwaterMACprotocolsw ithoutusingOFDM.Severalpapersonthistopichavebeenwritten,suchas[12],[13 ]and[14]. OFDMofferssomeadvantages,likehighfrequencybandefci ency,performanceimprovementformultipathinterference,reductionofthesel ectivefadinganomaly,andthe increaseofsystemcapacity[15][16].OFDMallowsonlyoneu seratatimetousethe channel,asseeninFigure2.2.Thatmeansitisnotpossiblet ohaveseveralusersatthe sametimeusingthemedium.Tohaveseveralusersaccessingt hechannelsimultaneously, severaltransducerswouldhavetobeinstalledineachnodei nthenetwork,andthehardwareinfrastructurecanbecostly.Ontheotherhand,tradit ionalMACprotocolsarenot adjustedtotakeadvantageofusingmorethanonechannelsim ultaneously. OFDMAisanextensionofOFDM.Itisamulti-userOFDMschemet hatpermitsmultipleaccessonthesamechannel.OFDMAusesmultipleandclo selyspacedsubcarriers,asseeninFigure2.1,butthedifferenceisthatcarrier sinOFDMAaredividedinto groupsofsubcarriers.Thesubcarrierscreatedaredistrib utedbyOFDMAamongthe usersinsuchawaythatuserswilltransmitandreceiveatthe sametimeusingjustone 20

PAGE 38

n n nr n n Figure2.1:OFDMA. n nrr Figure2.2:ComparisonbetweenOFDMandOFDMA. channel,withhavingonlyonetransceiverineachnode(thec ostintermsofhardware isnotincreased).Eachusercanhaveagroupofsubchannels, reducingproblemslike fadingandinterferencepresentedinthenetwork.Usingana dequatetopologyanddata linkprotocolinthedatalinklayerwithOFDMA,everynodeca ntransmitsimultaneous datatodifferentneighbors,eliminatingtheproblemsinun derwatercommunicationsand increasingthethroughputinthenetwork.Thecostintermso fhardwareforsupporting theproposeddatalinkprotocolisequivalenttohavingasta ndardunderwaternetwork, sincerequirementsintermsofhardwareareminimal,needin gonlyasinglehalf-duplex transceiverpernode.2.2MediumAccessControl(MAC)LayerMACprotocolscanbeclassieddependingonthenumberofcha nnelsused:onechannel andmorethanone,asshowninFigure2.3,ortheirnetworkarc hitecture,asshownin 21

PAGE 39

n rn r nr rr r r!"r rr # Figure2.3:ClassicationofMACprotocolsbasedonthenumb erofchannels. Figure2.4.Thesewillbeexplainedinthefollowingsection s.Someoftheunderwater MACprotocolsthathavebeendesignedaredisplayedintheco rrespondingclassication anddescribedlater.2.2.1ChannelClassicationTheseprotocolsoffersomeadvantages,suchasmultiplefre quencychannelsthatimprove thecapacityofanunderwaternetwork,thereductionorelim inationofcollisions,andthe reductionofthehiddenandexposedterminalproblems.They mayusemorethanone channelbutonlyonetransceiverormultiplechannelswithm ultipletransceivers,where thenumberofchannelsandtransceiversdoesnotnecessaril ymatch.AsseeninFigure 2.3,protocolsaredividedintoonechannelormorethanonec hannelused.Intherst category,achannelisselectedfromthebeginningonwhichn odeswillcommunicatewith 22

PAGE 40

n r nnr nr nnr nr r n n !"#$ rr Figure2.4:ClassicationofMACprotocolsbasedontheirne tworkarchitecture. eachother.Inthesecondcategory,alsocalledmulti-chann elprotocols,everynodecan havenotonlymorethanonechannelbutalsoseveraltranscei vers.Twonodescouldbe transferringinformationthroughaspecicinterfaceandc hannel,andtheothertwonodes couldbeusingeitherthesamechannelbutadifferenttransd ucerorthesametransducer butadifferentchannel.Therearesomeissuesthatareimportanttoconsiderinthede signofmulti-channelMAC layerprotocolsforunderwaternetworks.Forexample,when nodestrytocommunicate witheachotherthroughaspecicchannel,takingintoaccou ntthatthereareseveralchannels(someofthemavailable)andthereisonlyoneNIC,nodes cannotknowwhichchannelsarebusyforothernodesinrealtimeandtheycancreatei nterferencewithothernodes; thisproblemiscalledthehiddenmulti-channelproblem.Ho wchannelscanbeassigned toeverynodeinthenetworkisacomplexproblem,especially ifthereisjustoneNICin eachofthem,sinceeveryNIConlycanuseonechannel. 23

PAGE 41

Innetworkswithanumberofnodesitisdifculttodetermine whentransducersarein sleepingorwakingmodeandalsotoestablishthecommunicat ionbetweennodes.Synchronizationamongthenodesinthenetworkusingeithersev eralchannelsormorethan oneNICateachnodeisalsoacomplexprocessandtime-consum ing.Anotherproblem occurswhenanodetriestotransmitpacketsbuttheintended receiveriscurrentlytransmittinginaninactivechannelofathirdnode.Inthiscase,t hesenderhastowaituntilthe receiverisavailable.ThisproblemiscalledaDeafnesspro blem. 2.2.2NetworkArchitectureInthiscategory,protocolscanbeclassiedintocentraliz ed,distributedandhybridprotocols.CentralizedMACprotocolsassumetheexistenceofa basestationinthenetwork architecture.Thebasestationhastheresponsibilityofbe ingthearbiterassigningchannelresourcestousersaccordingtoaspecicalgorithm.Ass uch,usersaretoldwhento transmitandreceiveinformation,forhowlong,atwhatrate ,atwhatpower,etc.Allthese calculationsareperformedbythebasestation,whichisass umedtobeapowerfuldevice intermsofcomputationcapabilities,energyresources,me mory,etc. Centralizedprotocolshaveadvantagesanddisadvantages. Amongthemostimportantadvantages,centralizedMACprotocolsallowforsimpleusers ,andtherearenohiddenand exposedterminalproblems.Onthedownside,centralizedme chanismsusuallycannot scaletoalargenumberofusers,needacentralandpowerfuls tation,andmaypresenta singlepointoffailure.Someofthemostimportantcentrali zedMACprotocolsareshown inFigure2.4.IndistributedMACprotocols,stationsmaked ecisionsaboutwhentoacquirethechannel,i.e.stationscontendforthechannel.Be causestationsdonotknow 24

PAGE 42

whenotherstationsacquirethechannel,collisionsarepos sible.Indeed,animportant componentofdistributedalgorithmsisacollisiondetecti onandresolutionalgorithm. Distributedschemesarescalablecomparedwithcentralize dmechanismsanddonotrequireanyspecialstationtomakedecisionsonbehalfofther est.Therefore,nosingle pointoffailureexists,andreliabilityisimproved.Howev er,distributedschemesmust dealwithcollisions,asexplainedbefore,andalsowiththe hiddenandexposedterminal problems.Hybridprotocolsareacombinationofcentralizedanddistr ibutedMACprotocols.Most ofthehybridprotocolshavebeendesignedbasedonrequestgrantmechanisms.Before startingtotransmit,nodeshavetosendarequestforpermis sion(channelreservation)to thebasestationtellinghowmuchbandwidththeyneedandfor howlong.Thebasestation utilizesdifferentalgorithmstoassigneachuseraportion ofthechannel.ContentionbasedprotocolssuchasALOHAarenormallyutilizedforther eservationphase. 2.2.3UnderwaterMACProtocolsThissectionprovidesanoverviewofthemostrelevantMACla yerprotocolsforunderwatercommunications.2.2.3.1ANetworkingProtocolforUnderwaterNetworks[17]Inthisprotocol,theauthorsfocusontwoobjectives.Ther stgoalistoreducemessage latencybyremovingdependenciesbetweendataandcontrolc ommunications.Thesecond oneistouseaproactiveapproachforachievingarobustande fcienttopologymainte25

PAGE 43

n r Figure2.5:TopologyDiscoveryMessage(TDM)Propagationi nanunderwateracoustic network[17].nanceandrouting.Thisprocessisdonebyacentricgatewayt hatworksasamasternode. Thereisaproactivealgorithmthathelpsthemasternodenot onlyforcreatingandmaintainingatreetopologyforthenetwork,butalsoforupdatin gcontrolinformationsuchas powerlevelandroutes.Inaddition,theprotocolreducesth eexpecteddelaygeneratedin thenetworkbyseparatingthecontroloverheadfromthedata delivery. Theprocessstartswithanetworkcongurationcyclegenera tedbythemasternodetransmittingTopologyDiscoveryMessages(TDM),calledacongu rationprobe,anditisdone viadedicatedchannels(dedicatedchannelsaresetbyapply ingCDMAcodes).Every nodeinthenetworkwillselectachannelandsendaresponset othemasternode.Asa consequence,themasternodewillknowwhichchannelhasbee nselectedforeverynode initsnetwork,asshowninFigure2.5.Also,informationabo uttopologyisprocessed ineachnode(topologydiscovery)andsentwhenresponding. Thistopologydiscovery willbedoneperiodically.Oncethetopologyhasbeencreate d,routingpathsarecreated byaproactiveapproachtakingintoaccounttheexistenceof amasternode(centralized 26

PAGE 44

routing).Duetotheroutingprocessthroughthemaster,the reisaglobalplanningofthe resourcesinthenetwork.Forexample,ifonenodehaslowbat terypoweravailable,the masternodewillupdatetheroutetrafcanduseanotherpath fortransmissions.Thereare twotypeofpathsinthenetwork:pathsthatendatthemastern odeandpathsbetween nonmasternodes.Thersttypeofpathisusedforroutingtra fcandthesecondfor exchanginginformationbetweennon-masternodes.ByusingCDMA,eachnodethathasassignedachannelfortrans mittingpacketsalso mustbeabletolistentotherestofthechannelsthatarebein gusedforothernodesin thenetwork.Havingthecontrolofthetransmissionrangeof eachnodeinthemaster nodewilloptimizethenodepowerreservesandpermitchanne lreutilization.Ontheother hand,creatingdedicatedtransmissionchannelsforeachno dewillreducethepropagation delayproblem.Oneofthedisadvantagesofthisprotocolisthatthecapture problem,alsocalledthenearfarproblem,ispresentedinwhichtheunfairsharingofband widthhasnotbeenwell solvedinacousticnetworks.Anotherproblemisthatthispr otocoliscentralized,based onamasternode,andthisisnotthebestoptionwhenmobilene tworksareimplemented. 2.2.3.2ModiedMediaAccessControlDesignfortheAcousti c-BasedUnderwater DigitalDataCommunication[18] Thisprotocolworkswithmultiplechannelsandatimerforea chofthemineverynode inthenetworktotransmitdatapackets.Eachnodejustcanus eonechanneltostarta transmissionwiththereceiver.Theprocessisdividedinto fourstages:carriersensing, transmissionframe,receivingframeanderrorcontrol.Int hecarriersensingstage,if 27

PAGE 45

carriersensingisdetectedtheagsforthatchannelareset to'0'andtimerstartsagain, otherwisewhenoverowoccursduetotheincreasingidletim eforeachchannel,the agisincreasedandalsothetimerstartsagain.Thetransmi ssionframestageoccursby requestingatransmissionsignalfromthesender,andonech annelamongthemisselected fortheprocess.Aprocedureforsearchingthecandidatecha nnelsisexecutedandasa result,thechannelwiththelongestidletimeisselectedfo rthetransmission. Oncetheframearrivestothereceiver,itwillsendeitheran ACKsignalduetoasuccessfultransmission,oraNACKsignal(NegativeACKduetof rameerrordetection), andgotothecarriersensingmode.WhentheNACKissent,ther eceivingframestage iscanceled.Theerrorcontrolstageisveriedbyusingasto pandwaitARQprotocolin whichanACKpacketindicatestothesenderthatthedatapack etarrivedwithouterrorsto thereceiver.Someofdisadvantagesinthisprotocolarethatthenumberof channelsusedbyevery nodefordatatransmissionandreceptionaffectsitsconnec tivityinthenetwork,generatingasaconsequenceproblemsintermsofthroughput;having localinformationaboutthe stateofthechannelsineachnodeisrequiredtomakedecisio ns;andaschedulingprocess forsendingdatapacketsthroughthechannelsmustbealsode ned. 2.2.3.3AdaptedMACAtoUnderwaterAcousticNetworks[19]AmodiedversionoftheoriginalMACAprotocolisproposedi nwhichthreetypeof packetsareutilized:RTS(RequesttoSend)andCTS(Clearto Send)controlpackets toestablishaconnectionbetweentwonodesinthenetwork,a ndDATApackets.The proposedprotocolhasbeendesignedtoworkinanunderwater acousticsensornetwork 28

PAGE 46

n (a) n n n rn n (b) n (c) nnr (d) Figure2.6:AdaptationofMACAprotocoltounderwateracous ticcommunications[20]. thatincludesamasternode.Nodesareconnectedwiththemas teroneinahierarchical manner,andthemasternodeisinchargeofsendingtheinform ationtoasurfacebuoy. Nodesstartinalow-powerstateuntiltheydecidetotransmi tdatapackets,inwhichthey willsendanRTSpacket.OncethetransmitterreceivesaCTSpacket,itstartstosend theDATApackettothereceiver,andlateranACKpacketwillbesentfromthereceiver asshowninFigure2.6(a). IfthesenderhasnotreceivedtheCTSduringacertainperiod oftime,itwillretransmit theRTSpacket,repeatingtheprocessforamaximumnumberof Kretransmissionsas showninFigure2.6(b).Afterthat,thesourcewilldecideth atthelinkisnotlongeravailableandwillgobacktoalow-powerstate.DuringthetimethatthereceiversendstheCTS,thesenderst artstotransmittheDATA packetandlaterreceivestheACK,allotherrequestsfromth erestofthenodestoeither thetransmitterorreceiverwillbedeclined.Theproblemha ppenswhenthesourcewhose transmissionhasbeendeclinedisnotinformed;itwillbere peatingitsrequestperiodi29

PAGE 47

cally,generatingasaconsequencealotofunnecessarypowe rconsumptioninthesource andalsoanincreaseofthepossibilityofcollisions.Toavo idthis,aWAITcommandis addedtoinformthesourcethatthedestinationisbusyandwi llsendtheCTSassoonas possible.ThisisshowninFigure2.6(c).AdeadlockproblemappearswiththeutilizationoftheWAITc ommand.Iftwonodes sendanRTSatthesametimetoeachother,theywillsendlater aWAITcommandand waitforeverfortheothernodetosendaCTS,asinFigure2.6( d).Theproblemissolved byassigningprioritytothepacketsthataredirectedtothe masternode[19]. Someproblemsstillhappeninthisprotocol.Althoughtheco ntrolpacketexchangesare reduced,largepacketcollisionscanstillbepresentinden senetworks.Problemslikescalabilitywillincreasethedelaybecauseseveralnodeswillt rytoaccessthenetworkatthe sametimewithhavingonechannel,andonlytwoofthemcansta rtthecommunication; therestofthenodeswillbewaitinguntilthetransmission nishes(theunfairsharingof bandwidth).2.2.3.4SlottedFAMA,ProtocolforUnderwaterAcousticNet works[21] Thisisbasedontheooracquisitionmultipleaccess(FAMA) protocolwhichcombines carriersensingandanexchangingofcontrolpacketsbetwee nsenderandreceiverbefore datatransmissionstarts.Thisprotocolusestimeslotstoc ontrolpacketsfromconsuming excessivetimeintheirtransmission,whichisaprobleminu nderwateracousticcommunications.Twoproblemsthatareimportanttodealwithinun derwatercommunications arelowbandwidthandlongpropagationdelay(causedbythel owspeedofthesound).In slottedFAMA,nodesstayidleuntiltheyhavetotransmitpac ketsorsensethechannel. 30

PAGE 48

Figure2.7:SlottedFAMA[21]. Neighborsofeithersenderorreceiverknowiftheycansendi nformationwithoutcreatingcollisionswiththesenderorthereceiverpackets.Thep rocessofsendingpackets, eithercontrolordatapackets,isbasedontimeslotsandthe datapackettransmissionis successfulwhenthesenderreceivesanACKcontrolpacket.T hecommunicationprocess startswhenanodewantstosenddatapacketstoareceiverand itsensesthechanneland doesnotdetectacarrier.Atthispoint,ithastowaituntilt henexttimeslottosendan RTScontrolpacketrequiringpermissionfromthereceiverf ordatapackettransmission. OncethereceiverreceivestheRTSithastowaituntilthenex tslottosendaCTScontrolpackettothesenderinformingthatitisacceptingthed atatransmission.Allofthe sender'sneighborsalsoreceivetheRTSpacket.Inthesames lotthatthesenderreceives theCTS,allofthereceiver'sneighborsalsoreceivetheCTS packet,andinthenextslot thesenderwillstarttotransmitdatapackets,asseeninFig ure2.7. IfduringtheslottimethesenderdidnotreceiveaCTS,acoll isionisassumedandthe senderwillgotoabackoffstateforarandomnumberofslots. Oncethisrandomnumber ofslotsiscovered,thechannelissensedandanRTSwillre-s endifnocarrierisdetected. Ifacarrierisdetectedinthechannelduringabackoffstate inthenode,thenodechanges immediatelyfromidletoreceivingandexecutestheoperati onsrequiredinthisstate. 31

PAGE 49

Oncetheoperationsareendedthenodewillgobacktoitsback offstate,re-startingits backofftime.Afterthepacketisreceived,errorsareveri edusingastandardcyclicredundancycheck(CRC)algorithm.Ifeverythingiscorrect,a nACKcontrolpacketis sent,otherwiseaNACKcontrolpacketissent,indicatingth atthepacketneedstobe retransmitted(theretransmissionisdoneinthenextslota fterthesenderreceivesthe NACK).Apriorityschememustbeusedtoavoidcollisionsamo ngthepackets(CTSand RTSmusthavethehighestpriority)andtogetabetterthroug hput.Thepriorityinsending packetsisalsoimportantforgettingfairnessinunderwate rcommunicationswhenremote transmissionsareinvolved.Thisprotocolpresentshighcomplexityandthesynchroniza tionprocesscanbecostly, especiallyfordenseadhocnetworks.Routingalgorithmsto updateroutingtables,and algorithmsfornetworkmaintenancearealsoanissueinSlot tedFAMA.Asaresult,scalabilityhasnotbeenwelladdressedinthisprotocol.2.2.3.5UWAN-MAN,anEnergy-EfcientMACProtocolforUnde rwaterAcoustic WirelessSensorNetworks[22] Oneoftheobjectivesforthisprotocolistosolvethesynchr onizationprobleminunderwaternetworkstakingintoaccountdensenetworkswithsmal lspacingamongitsnodes. Inaddition,inUWAN-MACtheideaof"sleepmode"ineverynod ehasbeenimplementedtosaveenergy.Theprocessofgettingtransmissionb etweennodesinthenetwork withthisprotocolisdividedintoaninitializationperiod andatransmissionperiod,and startswiththedeterminationof"listencycles"tosynchro nizethescheduleforlatertransmissionsintheunderwaternetwork. 32

PAGE 50

Intheinitializationperiodeverynodebroadcastsacontro lpacket(SYNCsignalthat includesitstransmissioncycleperiod)atthebeginningof itscycleperiod,goestoits sleepmodeturningoffitstransceiver(savingenergy)andi twillwaituntiltheendof itssecondcycleperiodfortheSYNCsignalofitsneighbors. Duringthetransmission periodanodewilltellitsneighborsthatitisgoingtotrans mitagainafterthatperiodof time.Thedatatransmissionperiodstartsaftertheinitial izationperiodiscomplete,which meansthateverynodeknowsthetransmissionscheduleofits neighborsandwhenthey havetowakeupagaintoreceivedatafromtheirneighbors.Th etransmissiondatapacket hasaheaderthatisdividedintothreepartsrelatedwiththe actionsthatmustbetakenfor thenodesthatreceivethispacket:data,missingandSYNC.T hedataeldrepresentsthe destinationofthepacket.Missingeldrepresentsthelist ofneighborsofthenodewhich havenotreceiveditssignal.TheSYNCeldpermitsthenodetoinformitsneighborsofchan gingitscycleperiod foranewoneandasaconsequence,itsneighborswillchanget heirwake-uptimesfor thatnode.Oncethenodeendsitsdatatransmission,itgoesi ntoanidlelisteningmodein ordertosaveenergy;ifithearssomethingitisgoingtochan geitsstatetoareceivemode (thisstepisfortakingintoaccountnewnodesthatcanjoint henetwork).Whenanew nodeappearsinthenetwork,thisnodeisgoingtosendaHELLO messageassoonasit receivesapacketfromsomenodeinthenetwork.ThisHELLOpa ckethasatimestamp containinginformationaboutitstransmissionschedule.N odesthatreceivethispacketare goingtosendbackanACKcontrolpacketintheirnextdatatra nsmissiontimeconrming thattheyhavereceivedtheHELLO,andupdatetheirtransmis sionschedulewiththenew node.Whenanodedoesnotreceivedataatspecicscheduledwake-u ptime,twothingsmay havehappened:eitherthereisabadchannelconditionfromt hesenderorthesenderhad 33

PAGE 51

Figure2.8:Hybridprotocolframestructure[3]. afailure.Thereceivernodeisgoingtoupdateitsmissingli standanannouncementof thissituationwillbesent.Iftheproblemwasabadchannelc ondition,thesenderisgoing toreceivethemissinglist,decodeitandknowthatithastos endaHELLOmessagein itsnextcycleperiod,re-schedulingitswake-uptimecorre spondingtothesender.Ifthe problemwasthatthesenderfailed,itwillnotsendaHELLOan dthereceiverwillwait fortheHELLOfortwoconsecutivecyclesofthesenderandlat eritwilldeletethewakeuptimecorrespondingtothesender(thisalsosavesenergy) Themainproblemwiththisprotocolisthatitusessleepingm odestosaveenergy,which worksperfectlyforsensornetworks,butnotformobilenode sinwhichconnectionsamong thenodesmuststayactiveallthetimeinordertoavoidsomen odeslosingcommunicationswiththeothers.2.2.3.6AHybridMediumAccessControlProtocolforUnderwa terWirelessNetworks[3] Inthisprotocolitisassumedthatallnodesinthenetworkca nheareachotherandlisten tothechannel.Itworkswithslottedframesthatincludesch eduledandunscheduledperiodscombiningTDMAwithanunscheduledchannelaccessmet hod,toprovidelowenergyconsumptionbyreducingcollisions.Withscheduledpe riodscollisionsarereduced andwithunscheduledperiodstheprotocolisadaptedtochan gingtrafcconditions.The scheduledperiodpermitsnodestodistributestateinforma tion.AsshowninFigure2.8, 34

PAGE 52

theframeisdividedintotwodifferentstructures.Therst portionoftheframeisdivided intoNscheduledslotsthatusetheTDMAprotocol,eachoftho seslotswillbeassigned toonenodeinthenetworkfortransmissionsforalongtime.T hesecondportionofthe frameisdividedintoNuunscheduledslotsthatwillbeassig nedbasedonadistributed approach;thisassignmentwillbeforashorttimeandcanbeu sedforseveralnodesin differentframes.Fortransmissions,thedenedtimeslotcoversthemaximuml engthpacketplusthelongest expectedpropagationdelay,ensuringthenodeswillnothav eproblemstocompletely receiveapacketbeforeanothernodestartstotransmit[3]. Itisassumedthatthenodes usehighqualityclocksandsynchronizationprotocolsandw illhaveslotsynchronization withtheirneighbors.Thisisimportantsincethemediumisa cousticandthechannel conditionsvaryovertime[23].Onedisadvantageistheassumptionofqualityclocksandsyn chronizationprotocols, whichisaproblemforunderwaternetworks.Alsoitisassume dthatallnodeswillhave slotsynchronizationwiththeirneighbors,andthereisnop rocessdenedinthisprotocol todothis.2.2.3.7T-Lohi,aNewClassofMACProtocolsforUnderwaterA cousticSensor Networks[24] T-LohiorToneLohiisareservation-basedprotocolforunde rwateracousticsensornetworks.Itwasdesignedasanenergy-efcientMACprotocolfo rshortrangeacousticnetworks,toexploitlowpowerwake-uphardwarewheneverenerg yconservationisneeded, andtoreducethepropagationdelayinunderwatertransmiss ions.Theenergyisconserved 35

PAGE 53

intwoways:therstbydoingdatareservationsinordertoen surenocollisionsindata packettransmissions,andthesecondbyusingawake-uptone receiverhardwaremechanismtosolvereservationcontentionsbyallowinglisteni ngwithlowpowerforwakeup tones.TheT-Lohiprocessisdonerstbyexecutingatone-basedres ervationandafterthat, adatatransfer.Inthereservationperiod(contentionperi od),nodesthatwanttosend datapacketsthroughthenetworkcompetetogetandreservet hechannelsendingashort tone.Afterthat,theylistenforaperiodoftimecalledthec ontentionround(CR)toknow whetherthereservationwassuccessfulornot.Nodesthathe arthistonewillbackoff inthedatatransmissionperiodandthenodethatobtainsthe righttotransmitwillhave adataslotreserved.Iftherewasonlyonenodecompetingfor thechannelintheCR, thereservationperiod(RP)endsandthechannelisassigned toit.Ifseveralnodeswere competingintheCR,eachofthemdetectcontentionandwillr etransmitagaininanother CRextendingtheRP,untiloneofthenodesgetsthechannel.Inthedatatransferperiod,awake-uptoneissentbythesend ertowakeupthereceivers. Oncethewake-uptoneisreceived,nodeshavetoscanthedata channeltodetectwhether thereisapreambleornot.Ifapreamblewasnotfoundinthech annelthetoneisconsideredasacontentionindicator(alowpowertonereceiverdev elopedbyWillsetal.[25] isusedbyT-Lohi),otherwisenodeswillhavetodecodethepa cketreceivedandverify iftheyarethedestination.Iftheyarethedestination,the ywillbeswitchedtoareceive mode,otherwisetheywillgobacktosleep.Therearediffere ntversionsofT-Lohidependingonthereservationmechanismimplemented:synchro nizedT-Lohi(ST-Lohi), conservativeunsynchronizedT-Lohi(cUT-Lohi),andtheag gressiveUT-Lohi(aUT-Lohi). Thisprotocolismoreadequateforsensornetworks,sinceno desmustgotoasleeping stateforsavingenergyconsumption.Insomecaseswhenthet rafcinthenetworkin36

PAGE 54

creases,duetotheabsenceofCTScontrolpacketsattherece iver,thehiddenterminal problemispresentcausinganincreaseofcollisionsandasa consequencethethroughput ofthenetworkdecreases.2.2.3.8AMACProtocolforUnderwaterSensorNetworks[26]Thisprotocolworkswithasingletransceiveranditisdistr ibuted.Theauthorsimprove thenetworkefciencybyproposingamulti-channelMACprot ocolbasedontheMACA protocolin[27].Itutilizesonechannelforcontrolpacket s(commoncode)andseveral channelsfordatapackettransmissionswheredifferentspr eadingcodesareapplied.By usingseveralCDMAschemes,nodescansimultaneouslytrans mitbyoverlappingin timeandspacedomains.Allnodesinthenetworkareassigned tothesamecommon channel(commoncode)andthecommoncodeismonitoredforan ypacketarrival.Once thesourcesendstheRTSandreceivestheCTS(RTS-CTShandsh aking),theoptimal spreadingcodeinwhichthedatapacketisgoingtobetransmi ttedischosenforboth nodes(senderandreceiver).Afterreceivingthedatapacke t,thedestinationdespreads thereceivedsignalandretrievesthedata.Attheend,thedestinationnodewillsendanACKpackettothe transmitter.Anexample ofthisprocessisshowninFigure2.9inwhichthecommoncode is c ,andthecodeselectedfordatatransmissionis Ctl .Afternode2sendstheCTStonode1,the Ctl codeis chosenbythetransmitterfordatatransmission.AsinotherunderwaterMACprotocols,oneofthedisadvantag esisthecaptureproblem. Inaddition,scalabilityproblemswhenanodedecidestojoi norleavefromanunderwater networkarepresent. 37

PAGE 55

nr nr nrnr Figure2.9:AMACprotocolusingRTS/CTShandshaking[26]. OtherMACprotocolsdesignedforunderwaternetworksaremu lti-clusterprotocolfor adhocmobileunderwateracousticnetworks[28],adistribu tedCDMAmediumaccess controlforunderwateracousticsensornetworks[29],dist ributedMACprotocolsfor underwateracousticdatanetworks[30],R-MAC[31],andane nergy-efcientMACprotocolforunderwaterwirelessacousticnetworks[32].2.3LogicalLinkControl(LLC)LayerProtocolsinthislayerhavebeencreatedtoimprovethethro ughputinthenetwork,and toreducetheamountofpacketsinerror.Duetoproblemslike thelongpropagationdelay present,badqualitychannel(severalpacketsinerror)and verylowthroughputefciency inunderwatercommunications,traditionalLLCprotocolsn eedtobemodiedinorderto bemoreadequateforunderwatercommunications.Thereares omeprotocolsthathave beenimplementedtoimprovetheproblemsjustmentioned.Th eclassicationofthe logicallinkprotocolsandsomeexamplesofthemshowninFig ure2.10willbeexplained inthefollowingsections. 38

PAGE 56

nnrr nnnr nrr n r nr r rr nnr nr r rr "n n r rn rr r nr rnr r rr Figure2.10:Classicationoflogiccontrolprotocols. 2.3.1LogicalLinkControlClassicationLogicallinkcontrolprotocolsareclassieddependingont heactionsrequiredforthe packetsarrivedtothereceiverwitherrors,eitheronlydet ectionordetectionandcorrection.TheseprotocolsareclassiedinAutomaticRepeatReq uest(ARQ),ForwardError Correction(FEC)andhybridprotocols,asshowninFigure2. 10. 2.3.1.1ARQProtocolsARQprotocolshavebeendesignedfordataerrordetectionso fpacketsthatarrivetothe receiver[33].Onceerrorsaredetected,packetsareretran smitteduntiltheyarrivewithout errorsorthemaximumnumberofretransmissionsisachieved 39

PAGE 57

Theproblemwiththeseprotocolsistheamountofpacketretr ansmissionsthatcanbe producedwhenthechannelerrorrateincreases.Theincreas eofpacketretransmissions willcausethethroughputinthenetworktodecreasetryingt omaintainhighreliabilityin thenetwork.In[33]itisexplainedwhichistheoptimalblocklengthforp acketsinordertoreduce thedataretransmissions.Itisbettertohavethelargestbl ocklengthtominimizethetime wastedsendingacknowledgmentsandtheassociateddelays, butalsoitisbettertohave thesmallestblocklengthtominimizetheerrorprobability ofthepacketsentandtominimizethetimewastedinretransmissions.Asaconsequenceofobtainingtheoptimalblocklength,thet hroughputefciencyismaximized;inotherwords,thewastedtimeisminimized.Theopt imizationofthethroughput efciencyvariesdependingonthetypeofARQprotocolandit willbeexplainedinthe followingsectionwheresomeexamplesofunderwaterlogica llinkprotocolsareshown. ARQprotocolsaredividedintostopandwait,gobackN,andse lectandrepeatprotocols. StopandWaitProtocols.Instopandwaitprotocols,thetran smittersendsonepacket anditwillonlytransmitthenextoneafterreceivingaposit iveacknowledgment fromthereceiver,tellingitthatthepacketreceivedwasco rrect.Ifthetransmitter eitherdoesnotreceivetheacknowledgment(timeout)orrec eivesanegativeone,it willretransmitthepackettothereceiver.Thisprocesswil lberepeateduntilapositiveacknowledgmentisreceived.Thisprotocoliseasytoim plement,howeverone problemisthetotaltimespentwhenamessagetosendissplit intosmallpackets (onlyonepacketcanbesentatatime).Anotherproblemisthetimelostbetweensendingapacketfro mthetransmitterand receivingthepositiveacknowledgment,becauseinthattim ethetransmittercannot 40

PAGE 58

sendanotherpacket.Furthermore,adeadlocksituationwil loccurifthetransmitter doesnotreceivethepositiveacknowledgment,oriftherece iverdoesnotreceive thepacketorthepacketcomeswitherrorsallthetime.Ifany ofthesesituations occur,thepacketwillhavetoberetransmitted,creatingan inniteloop,unlessa maximumnumberofretransmissionsisdened. GoBackNProtocols.GoBackNisaslidingwindowprotocol.Th isprotocolworks usingawindowofsizeNtospecifythemaximumnumberofpacke tsbeingsent (whenavailable)withoutwaitingforanacknowledgment.Th eprotocoleasilyhandlesbadframesandlostframesbymaintainingalistofthefr amestransmittedthat remainsunacknowledgedandmostrecentlytransmittedfram es[34].Inorderto havethelist,abufferisrequiredatthereceiver.Iferroro ccurs,sendertransmitsall framesfromframeinerrorandreceiverdiscardsallothers. Theimplementationof thisprotocolcanbedifcult,butthemostchallengingpart istondtheoptimum timeoutvaluethatprovidesefcienttransmissionforalle rrorrates. Smallertimeoutvalueseemstobeanexcellentdecisiononhi ghererrorratesbut itisbettertouseamarginallyhighervaluetooptimizethem .Anotherproblemin thisprotocolisthehighnumberofframesretransmitted,wh ichgeneratesbandwidthwasteandalongertimetocompletetransmissions.The causeofthisisthat foreverylostframe,alltheframesinthewindowtransmitte dpreviouslymustbe retransmitted. SelectiveRepeatProtocols.AsintheGoBackN,awindowsize isusedinSelectiveRepeatprotocols.Windowsareusednotonlyinthese nderbutalsointhe receiverandthosewindowsizesmustbeequal.Thedifferenc ewithGoBackNare thatthereceiveracceptsbufferout-of-orderframesarriv ingwithouterrors,anda 41

PAGE 59

retransmissionmechanismisusedtoaskfortheretransmiss ionofspecicpackets. Everyframesentbythetransmitterhasthedatainformation andanadditionalsequencenumber.Bytheinclusionofhesequencenumber,there ceiverknowsthe earliestframenotreceived,anditwillsendbackanACKwith thecorresponding number.Thesendertransmitsframesuntilitswindowisempty,movin gitswindowforevery packetacknowledgedinthereceiver.Later,itwillre-send frameswhichdidnot arriveandthenwillcontinuewhereithadleftoff.Bythetim ethesenderissending frames,thereceiverisalsollingitsreceivingwindowwit hthoseframes,frames thathavebeenveried.Oncethereceivingwindowisfull,allframeswillbesenttot heupperlayer[34]. Forwirednetworks,allofthemcanbeapplied,butselective repeatprotocolsare themostefcient(wirednetworksarefullduplex).Inunder waternetworks,becauseofthewaythatpacketscanbetransmittedviatransduc ers,onlyhalfduplex communicationcanbedone,meaningthatstopandwaitprotoc olsonlyareused. 2.3.1.2ForwardErrorCorrectionProtocolsTheseprotocolsareappliedtobothdetectandcorrecterror s.Theyaddredundantinformationtotheoriginalframesatthesender.Thisredundanti nformationisusedtoreconstructapproximationsorexactversionsofpacketswitherr ors.Oncethepacketisreceived bythereceiver,aFECtechniqueisappliedtobothdetectand correcterrors.Thisprocess helpstodecreasethenumberofretransmissionsthatcouldb edonebythesenderdueto immediatecorrectionsoferrorsexecutedbythereceiver. 42

PAGE 60

Thesetechniquesalsoreducethetimefortransmissionsbet weensenderandreceiver becausethesenderavoidswaitingfortheround-triptimepr opagationdelayneededto receivethenonacknowledgment(NAK)controlpacket,andas explainedbefore,frames arexedatthereceiver[34].2.3.1.3HybridProtocolsHybridprotocolscombineARQandFECtechniques.Theywerec reatedtoobtainamore efcientprocesstodecreasethenumberofretransmissions thatcanbedonebythesender, comparedwithARQandFECprotocols.Differentstrategiesa reappliedduringthepacket transmission,toidentifywhendetectionschemesareusedi nsteadofcorrectionschemes. 2.3.2UnderwaterLogicalLinkControlProtocolsAdescriptionofsomeLLCprotocolsandmechanismsforunder watercommunications areshowninFigure2.10,andareexplainednext.2.3.2.1OptimizationofaDataLinkProtocol[35]AcomparisonamongsomeARQprotocolsaredoneinthispaper. Inunderwateracoustic channels,theefciencyoftheARQschemesislimitedbyboth thepoorbiterrorrate (BER)performanceandthelongdelaygeneratedbythelowspe edofsoundpropagation, whichis1500m/s.In[35]statisticalanalysisismadetomax imizeefciencyofARQ schemesbasedonndingtheoptimalpacketsizetobesentasa functionofbiterror 43

PAGE 61

probability.Thestatisticalanalysisisappliedtothreed ifferentversionsofthestopand waitprotocolwhicharethewellknownstopandwait,andvers ionsin[33]and[36].It isassumedthateachpackethas N = N d + N oh bits,where N d isthenumberofdatabits, and N oh isthepacketoverheadwhosenumberofbitsisatleastthenum berofbitsused fortheCyclicRedundancyCheck(CRC)function. T p = NT isthepacketdurationwhere T = 1 = R isthebit(symbol)duration,and R representsthebit(symbol)rate.Moreover, T sync isthesynchronizationpreamblewhichprecedesthegroupof packetstransmitted. T d = l = c isthepropagationdelay, l isthedistancetransmitter-receiver,and c = 1500 m = s isthespeedofsoundunderwater.If m representsthenumberofpacketstobetransmitted, thenthetimeneededtotransmitthemandthereceptionofthe correspondinggroupof acknowledgmentsis T ( m )= m ( T p + T ack )+ T w (2.1) wherethetotalwaitingtimeis T w = 2 ( T sync + T d ) (2.2) andthedurationofanacknowledgmentisnegligiblewithres pecttothepacketduration, inotherwords T ack << T p .Thetimeoutofanstopandwaitprotocolthattransmits m packetshastobethesameastheround-triptime T ( m ) inordertoobtainthebestefciency.Otherconceptsappliedarethethroughputefcienc yofARQprotocolsthatisthe ratioofusefulpackettimeandthetotaltimespentontheave rageforasuccessfulpacket transmissioninwhichtheaverageisoverthenumberofretra nsmissions,and p thatis thepacketerrorprobability.Insteadoftransmittingonep acketeverytimelikeinthe traditionalstopandwaitprotocol,inprotocolsin[33]and [36]awindowof m packets aresentineachtransmission.ItisshownthattheMorrisver sion[33]presentsthebest throughputefciencyduetotheamountofnewpacketstransm ittedineverywindow. 44

PAGE 62

nr n !" Figure2.11:One-dimensionaln-hopacousticchannel[37]. 2.3.2.2AMulti-HopARQProtocol[37]Thisprotocolisanopportunisticacknowledgmentschemecr eatedforstopandwaitARQ protocolsandworksovertwodifferenttypesofunderwatern etworks:one-dimensional n-hopacousticchannelandtwo-dimensionalmulti-hopacou sticsensornetworks,asseen inFigure2.11andFigure2.12. n r r n r n r Figure2.12:Two-dimensionalmulti-hopacousticsensorne twork[37]. Thedesignproposedisaper-hophybridimplicit/explicita cknowledgmentschemeover amulti-hopchanneltakingintoaccountthatacknowledgmen tsingeneralaresentfrom thereceivertothetransmitterviatwodifferentprocedure s,ImplicitAcknowledgment 45

PAGE 63

(IMP)orExplicitAcknowledgment(EXP).TounderstandhowI MPworks,considera transmissionfromnode j tonode j + 1.Ifnode j hearsthatthenode j + 1isforwarding itspacket,therewillbeanimplicitacknowledgmentfrom j + 1to j .Ifnode j + 1does notforwardthepacketreceivedfrom j afteratimeout,meaningthatthenode j isnot acknowledged,thennode j willretransmitthepacket.EXPhappenswhennode j sends apackettonode j + 1,andwaitsfortheexplicitacknowledgmentthat j + 1hastosend it.Node j + 1transmitstheacknowledgmentifitspreviouspackettrans missionwas acknowledged,exceptthenode n + 1,whichisthelastnodeandonlyhastosend(ACK) packets.ThepurposeofusingeitherIMPorEXPinthehybridd esignistomaximizethe latencyefciency,and/ortheenergyefciencyfordatadel iveryinmulti-hopacoustic networksystems.Anotherpurposeistoreduceunnecessaryt ransmissions. Someofthedisadvantagesinthisprotocolarethehidden,de afnessandcaptureproblems stillpresent.Thepoorpacketdeliveryratioobtainedusin gthetraditionalstopandwait protocolandtakingintoaccountthebadqualitychannelinu nderwaternetworksisanotherproblem.Inaddition,thelongpropagationdelaypres entinunderwatercommunicationsnegativelyaffectsthethroughputefciency,especi allywhenstopandwaitprotocols areapplied.2.3.2.3FEC-BasedReliableDataTransportProtocolforUnd erwaterSensorNetworks[38] TheprotocolthatiscalledSegmentedDataReliableTranspo rt(SDRT)isahybridapproachofARQandFEC.SDRTisinchargeofreconstructinglos tpacketsinsteadof error-correctioninpackets.Theprocessofsendingpacket sstartswithgroupingthedata packetstosendinblocks,inwhichawindowof m blocksissentbythetransmitter.This 46

PAGE 64

numberguaranteesthatthereconstructionprocesscanbedo neinthereceiver,anditis obtainedbyestimationstakingintoaccountthespeedofthe mobilenodes,soundpropagationspeed,availablebandwidthanddistancebetweensen derandreceiver. Oncetheblocksaregenerated,theyareencodedusingtheFEC protocolcalledTornado[39] andtransferredintothenetwork.Datapacketsareforwarde dblockbyblockandhop byhopbyintermediatenodes.Afteranodestartstoreceiveb locksfromthesender,it waitsforthewindowofencodedpacketsneededthatguarante esthereconstructionofthe datapackets.Whenthewindowof m encodedpacketsissent,thesenderdoesnotstop transmittingsubsequentpackets;itcontinuessendingthe mbutataslowdatarate,waiting fortheacknowledgmentfromthereceiverrelatedtothewind owsent.Atthereceiver side,oncethewindowisreceived,ittriestoreconstructth eencodeddatapackets.Ifthis ispossible,itwillsendbackanacknowledgment,andthesep acketswillbeencodedand forwardedtothenexthop.Afterreceivingtheacknowledgme nt,thesenderwillstopthe processofsendingsubsequentpackets.Oneoftheproblemsinthisprotocolisthattheestimationof thenumberofblocksdependsontheavailablebandwidth.Withapoorbandwidth,whi chiscommoninunderwaternetworks,thenumbercouldbeone,executingthetradi tionalstopandwaitprocess inwhichanodethatsendsapacket,cannotdoanythingelseun tilitreceivesanacknowledgment.Thatmeanstimewillbewastedfornodeswhentheyar ewaitingforacknowledgments,especiallyinunderwateracousticnetworkswher elongpropagationdelaysare present.Otherproblemsarenotonlythethroughputofthene tworkbutalsothepacket deliveryratio,whichdecreaseifthenumberofpacketsalso decrease,andthescalability inthenetwork. 47

PAGE 65

OtherLLCprotocolsdevelopedforunderwaternetworksare: amulti-hopARQprotocol foranunderwateracousticchannel[40],efcienterrorrec overyusingnetworkcodingin underwatersensornetworks[41]andonapplyingnetworkcod ingtounderwatersensor networks[42]. 48

PAGE 66

Chapter3:2MAC:AMultichannelMACProtocolTheproposedMACprotocol2MAChasbeencreatedinordertore ducetypicalproblems presentinwirelessnetworkslikecollisions,andhiddenan dexposedterminalproblems, andtoimprovethethroughputperformanceintheunderwater channel. Thischapterdescribes2MACindetail,themostadequatenet worktopologyneededfor thisprotocol,howthechannelassignmentworkshavingOFDM Aatphysicallayer,possiblescenarioswhere2MACworks,andcomparisonswithothert raditionalMACprotocols usingdifferentbackoffapproaches.3.12MACDescription2MACisacontention-basedMACprotocolbasedontheMACAWpr otocol[43].Ithas beendesignedtoimprovetheperformanceefciencyinunder wateradhocacousticnetworkshavingneighborslocatedindifferentsub-channels.2MACusesafour-wayhandshakingaccessmethod(RTS/CTS/DA TA/ACK),anewcontrolpacketcalledBTS(BlockedToSend),anAdjustedRespon setime(ARS)towaitfor signalsfrombothneighbors,andalisten/contentiontimet oexchangedata,asshownin Figure3.1.Ineverytransmissionprocess, m datapacketsaretransmittedthrougheach 49

PAGE 67

nrn { { { (a) nrn { { { (b) Figure3.1:2MACtransmissionprocess. channel.ShortInter-FrameSpacing(SIFS)andDistributed Inter-FrameSpacing(DIFS) arealsousedinthepackettransmissionprocessasinIEEE80 2.11. BTSisacontrolpackettoinformthatachannelwillnotbeava ilableforcommunications withthecorrespondingneighbor.Onceasenderdecidestost artatransmissionandjust onechannelisgoingtobeused,thesenderstartsahandshaki ngprocessthroughthis channelbutatthesametimetransmitsaBTSintheotherchann eltotellitsotherneighbor forhowlongthechannelisgoingtobeunavailable.Thisissh owninFigure3.1(b)in whichNodeAhaspacketsforAbutnotforC.Theproposedprotocolworkswiththeassumptionofexisting OFDMAinthephysical layerthatcreatesmultiplecommunicationchannelsthatca nbeusedsimultaneouslyby theMAClayer(thereisjustonetransceiverineachnode).2M ACreliesinOFDMAatthe lowerlayeranddenestwosubchannelspernodeforsimultan eouscommunicationswith itsneighbors,andthreechannelsatthesametimefortheent irenetwork.Theassumptions for2MACareoutlinedbelow: Asinglehalf-duplextransceiver OFDMAinthephysicallayer 50

PAGE 68

Anetworkwithalinearorpolygontopologies,forswarmingp urposes Distancebetweennodesmustbeknownandstaywithoutvariat ionsbyhavinga controlsystemineachAUV Threechannelsusedsimultaneouslyinthenetwork Everynodehastwoneighborsandeachofthemworksinadiffer entchannel 3.2NetworkTopologyandChannelAssignmentByusingseveralsubchannelspernodesimultaneously,itis possibletohavedifferent topologies,especiallylinearorpolygontopologies,andt otransmitinformationtoeach oftheneighborswithoutcreatingcollisions.Figures3.2a nd3.3showlinearandhexagon topologieswiththeassignmentofthreesub-channelsthata llowsimultaneoustransmissionwithoutinterference.Forexample,inthelineartopol ogy,nodeAhasBandCasits neighbors,itcommunicateswithBthroughchannel2andwith Cthroughchannel1.C andBcancommunicatewithAandDrespectivelyatthesametim e,withoutexperiencing theexposedterminalproblemsinceAdoesnotsharethesamec hannelwithCandB; similarly,AandEcancommunicatewithCsimultaneouslywit houtcolliding.Asimilar situationwillhappenhavingeitherhexagonornonagontopo logies,asseeninFigures3.3 and3.4.Thereisarelationbetweenthenumberofchannelsinthenetw ork,twoasaminimum, andthenumberofsidesthatapolygontopologycanhave.Usin gthreechannelsinthe networkbuttwodifferentpernode,itispossibletohavepol ygontopologiesinwhich theirnumberofsidesisamultipleof3. 51

PAGE 69

Figure3.2:Lineartopology. 2 1 32 1 3 2 1 channels nr rr Figure3.3:Hexagontopology. Theoptimalnumberofchannelspernetworkrequiredtosetin apolygontopologycan beobtainedusingEquation3.1,inwhich S representsthenumberofsidesthatthepolygontopologyisdesiredtohaveand CH istheoptimalnumberofchannelsneeded.The valueof CH issuchthat n isthemaximumquotientvalue.Asanexampleforthehexagon topologyshowninFigure3.3,3or6channelscanbedened,bu tusingonly3,amaximumvalue n isobtainedwhen S isdividedby CH = 3,wherenodesusingthesame channelinatopologywith CH channelswillbe CH 1hopsawayfromeachother, avoidingcollisions.With3OFDMAsub-channelsthemaximum valueisobtainedfor n ,andthecommunicationamongthenodescanbeeffectivelydo ne. Anotherexampleishavingadodecagontopologyinwhicha3-c hannel,4-channelor12channelnetworkcanbesetup.With3channelsdenedbyOFDMA themaximumvalue 52

PAGE 70

2 13 2 3 1 2 1 3 nr Figure3.4:Nonagontopology. isobtainedfor n ,andthecommunicationamongthenodescanbeeffectivelydo ne.Linear topologiesarealsopossibleandtheyneedonlya3-channeln etwork;itdoesnotmatter howmanynodeswillbeinthelinearnetwork. n = max S CH (3.1) 3.3ScenarioWithOFDMAworkingatthephysicallayer,everyneighborwil luseadifferentchannel, twochannelspernode,andthetotalamountofchannelsuseds imultaneouslyforallnodes willbethree(forexplanationpurposesthenetworkwillbea polygontopology),asseen inFigure3.3.Itisimportanttohighlightthatbyusing2MAC overlinearorpolygon topologies,hiddenandexposedterminalproblems,capture anddeafnessproblemsare eliminated.Theseproblemswillhappenonlywhentwoadjace ntnodestrytocommunicatewitheachotheratthesametimebysendinganRTSpacket. Thissituationwillbe explainedlater. 53

PAGE 71

Inthistypeofdesignduetotherangeofcoverageineverynod e(eventhoughthereisjust onetransceiverpernode),AandCcanstartaconversationsi multaneouslywithBwithout collisionsatthereceiver(nohiddenterminalproblem);th isisbecausethecommunicationisthroughdifferentchannels.Also,Ccanstarttransm ittingtoDwithouttakinginto accountwhatishappeningbetweenAandBforthesamereason( theexposedterminal andcaptureproblemsareavoided).Themostimportantrestr ictionisthatnodesusingthe samechannelmustbe2hopsawayfromeachother.Inthenextse ctionitwillbeshown how2MACworks,takingadvantageofOFDMAoverthesetopolog ies. 3.42MACStateTransitions2MAChasthefollowingtenstates:Idle,ChannelAssignment ,Contention(Listen),WaitingforCTS,ReceivingRTS,WaitingforACK,WaitingforData ,Backoff,Adjusted Response,andBlockedtoSend.Figures3.5and3.6showthe2M ACsenderandreceiver statemachines,respectively.Thestatesdisplayedinthe guresthatrepresentthesender andreceiverprocessareexplainednext.3.4.1IdleStateWhenanodereceivesanRTSthroughoneofitschannelsinthis state,itgoestotheReceivingRTSstatetocheckifanRTSpacketiscomingfromtheo therchannel.Itstaysin theIdlestateuntilithaspacketstosend,whenitgoestothe ChannelAssignmentstate. IntheIdlestate,anodecanreceivenotonlyRTSbutalsoBTSp ackets.IfthenodereceivesaBTSpacket,itblocksthecorrespondingchannelfor thetimedenedintheBTS, 54

PAGE 72

Figure3.5:Finitestatemachineof2MACprocessatthesende r. 55

PAGE 73

Figure3.6:Finitestatemachineof2MACprocessattherecei ver. 56

PAGE 74

andonlytheotherchannelisavailablefortransmissionsdu ringthattime.Whenanode receivesaBTSpacketfrombothchannels,itgoestotheBlock edtoSendstate. 3.4.2ChannelAssignmentState n rn nn nn nn n !nn r r n r n#n$nn Figure3.7:2MAC,transmissiondiagram. Sinceeverynodehastwoneighborsandeachofthemhasadiffe rentchannelassigned, twoBTSeldsareusedtoavoidcollisions,oneforeverychan nel(similarusagelikeNAV inIEEE802.11),asseeninFigure3.7.Oncepacketsarrivefr omtheupperlayer,their destinationsareveriedandthecorrespondingchannelsar eselectedandactivateddependingontheirchannelBTSvalues.Iftwodatapacketsmust betransmittedatthesame time(oneforeveryneighbor),bothBTSareveried,otherwi seonlyonedatapacketis transmittedandthecorrespondingchannelBTSvalueisanal yzed.AnonzeroBTSvalue meansthatthechannelisunavailabletosendpacketsbecaus ethatneighborisalready usingit. 57

PAGE 75

Threepossibilitiescanhappen.Therstoneiswhentwo M datapacketsmustbetransmittedsimultaneouslyindifferentchannelsandatleaston eofthecorrespondingBTS valuesiszero.IfbothBTSarezero,allchannelsareactivat edandthetransmitterwillgo totheContentionstateinbothchannels,toseewhetherneig hborsaretransmittingornot andtostartatransmissionprocesslater.Otherwise,theda tapacketsthatcorrespondto theblockedchannel,thechannelwiththenonzeroBTSvalue, willgobacktothequeue foranothertransmissionprocess,andthenodewillgotothe Contentionstateonlyforthe channelwiththezeroBTSvalue;thischannelisactivated.I nthecasethatbothBTShave nonzerovalues,thenodewillgototheBlockedtoSendstate.Thesecondpossibilityoccurswhentwo M datapacketsmustbetransmittedbutboth BTSvaluesarenonzero,inthiscasethenodewillgotoaBlock edtoSendstate.Once itsBlockedtoSendstatenishes,thenodewillgobackagain totheChannelAssignment state.Thelastpossibilityiswhenjustone M datapackethastobetransmittedanditscorrespondingBTSiseitherazeroornonzerovalue.IfBTSiszero, thecorrespondingchannel isactivated,theotheroneisputonhold(itwillusedtosend BTSpackets),andthenode willgototheContentionstate,otherwiseitwillgotoaBloc kedtoSendstate. 3.4.3Contention(Listen)StateInthisstatethetransmitterwillbelisteningtotheactiva tedchannels(eitherbothorjust one)foratimeequaltoaroundtriptimeofacontrolpacket(C TSorRTS),dependingon whathappenedintheChannelAssignmentstate.Ifthechanne l(orchannels)arefree(no communicationoccursduringtheContentionstate),thenod esendsRTSsimultaneously 58

PAGE 76

toalltheactivatedchannelstostartacommunicationproce ss(aBTSwillbesentthrough theactivatedchannelthatisnotgoingtodoadatatransmiss ion),andgoestotheWaiting forCTSstate.Ifthechannelisbusy,thenodedefersitsrema iningcontentiontime,and receivestheneighbortransmission(s).Oncetransmission nishes,itgoesbacktothis statetonishthelisteningperiod.TheexampleinFigure3.8(d)showsanodeBnishingitsconte ntionperiodwithout receivinganypacketfromitsneighbors.Then,acommunicat ionprocessstartswiththe nodesitneedstotransmitdatapacketsbysendingRTSatthes ametimetoAandB(due tothepresenceofOFDMA).Whenoneofthechannelsisputonho ld,meaningthereare packetstosendtoonlyoneneighbor(theotherchannelwasac tivated),thetransmitter willsendtwodifferenttypeofpackets,theRTSfortheneigh boritwantstostartadata transmissionwith,andaBTSpackettotheotherneighbor,re layingthetimethatitis goingtobebusy,asseeninFigure3.8(a).InFigure3.8(b),N odeBstartsahandshaking withAbutaBTSwasnotsenttoCbecausepreviously,thechann elwasblocked. 3.4.4WaitingforCTSStateInthisstateanodewaitsuntileitheritreceivesbothCTSor atimeoutoccurs.AfterreceivingtherstCTSpacket,thetransmittergoestotheAdju stedResponsestatetowait forthesecondCTSarrival.Ifthesenderdoesnotreceiveit, itassumesthatacollisionoccurredinthatneighborandthecorrespondingchannelisdis abledfordatatransmissions. Whenatimeoutoccurs,thesenderenterstheBackoffstatefo racertainamountoftime. OncetheAdjustedResponsestatenishes,thetransmitters endsthedatapacketsthrough thecorrespondingactivatedchannelsandgoestotheWaitin gforACKstate. 59

PAGE 77

nnr { { (a) nrn { { (b) nnr { { { (c) nrn { { (d) Figure3.8:Contentionprocess. IntheexampleinFigure3.9(a),bothCTSarereceivedsimult aneously,thenBdoesnot spendtimeintheAdjustedResponsestate.However,inFigur e3.9(c)onceBreceivesthe CTSfromnodeCitwillbeintheAdjustedResponsestatetosee whetherornotaCTS packetwillcomefromA.TheFigure3.9(d)showshowatimeout canoccur;inthiscase BwillgotoaBackoffstate.3.4.5ReceivingRTSStateIfanodereceivesanRTSwhilebeingintheContentionstate, itdefersitsremainingtime untilthecommunicationprocessnishes.IntheReceivingR TSstate,whentherstRTS isreceivedthenodegoestotheAdjustedResponsestatetove rifythatanotherRTSis comingfromtheotherneighbor.IftwoRTSarereceived,aCTS istransmittedthrough itsactivatedchannelssimultaneously,otherwisejustone CTSissenttoanswertheRTS, 60

PAGE 78

nrn { { { (a) n rr { { n nn n n { n (b) nrn { { { (c) n rr { r (d) Figure3.9:RTS/CTScommunicationprocess. andaBTSistransmittedtoitsotherneighborcarryingtheti meitwillbebusy.Oncethe CTSaretransmitted,thenodegoestotheWaitingforDatasta te. AsshowninFigure3.8(c)forexample,CreceivesanRTSfromD .Then,aftertheAdjustedResponseperiodnishes,itsendstwopackets,aCTSt oDandaBTStoA(to indicateCisgoingtobebusyandthechannelwithAwillbeblo cked).Inthisway,A andBandEandFcanstartatransmissionwithoutgeneratinga nyproblemswiththe transmissionbetweenCandD.IntheexampleinFigure3.9(c) ,BreceivesCTSfrom AandC,anditstartsthedatapackettransmissionsimultane ouslytobothnodes.Thisis possiblebecauseOFDMAresidesinthelowerlayer. 61

PAGE 79

3.4.6WaitingforACKStateInthisstatethenodewaitsuntileitheritreceivesbothACK ,atimeoutoccurs,orjustone ACKarrived.AfterreceivingtherstACKpacket,ifonlyone channelisactivated,the transmitternishesitstransmissionprocess,channelsar esetuptotheavailablemode, andthenodegoestotheIdlestate.Ifafterreceivingthers tACKpacketbothchannels areactivated,thetransmittergoestotheAdjustedRespons estatetowaitforthesecond ACK.IfthesenderdoesnotreceivethesecondACK,itassumes thatacollisionoccurred inthatneighbor,itscorrespondingdatapacketwillberetr ansmittedinthenexttransmissionprocess,thechannelsaresetasavailable,andthenode goestotheIdlestate.Whena timeoutoccurs,thesendergoestoaBackoffstateforacerta inamountoftime. 3.4.7WaitingforDataStateOncetherstdatapacketisreceived,thereceivergoestoth eAdjustedResponsestateto verifythatanotherdatapacketiscomingfromitsotherneig hbor.Otherwise,atimeout occursandthenodegoeseithertoitsdeferredcontentionpe riodortoanIdlestate.Once theAdjustedResponsestatenishes,ACKforthecorrespond ingdatapacketsaresent througheachactivatedchannel,andthereceiverwaitsunti lthetimedenedinitsBTS nishestosetupitschannelsasavailableandstartBTSinze ro. AsshowninFigure3.8(d),nodeDdoesnothavetogototheAdju stedResponsestate becauseitreceivesdatapacketsfromEandCsimultaneously .Then,itsendsbothACKto CandEatthesametime. 62

PAGE 80

3.4.8BackoffStateInthecaseofcollisions,nodesexecuteaBinaryExponentia lBackoff(BEB)retransmissionalgorithmforcollisionrecoverythatisamodicat iontotheoneusedin[44]. ThesendergoestotheBackoffstateforacertainamountofti me,andthetransmitter startsagainaChannelAssignmentperiodforanewcontentio ntime.OncetheBackoff nishes,ifthenumberofretransmissionshavereachedthem aximumallowednumberthe packetisdroppedandthenodegoestotheIdlestate.Anodere ceivinganRTSwhilein theBackoffperioddefersitsremainingtimeuntilthecommu nicationprocessnishes. 3.4.9AdjustedResponseStateNodesinthisstatewaitforacertainperiodoftimecalledAd justedResponseState(ARS) toseeifanotherpacketiscomingornot.Sinceanodecanrece ivepacketsfromitsneighborsthroughdifferentchannels,thesepacketsdonotneces sarilyarriveatthesametime andagapisproducedbetweenbotharrivals.Tocontrolthisg ap,anARStimeisincluded. Thegoalofthisstateistogetbothchannelsofanodetostaye itherinthereceivingorthe sendingmodeatthesametimebyusinganAdjustedResponsepe riod.Withthisstate, anodeknowsiftherewasarequestornotfrombothchannelsan dlaterrespondssimultaneouslythroughbothchannels.Forexample,asinFigure3 .8(d),nodeDisreceiving RTSfromitsneighborsindifferentmoments.WiththeAdjust edResponseperiod,Dwill respondwithCTSatthesametimetoCandE.ThisARStimeisde nedasaconstant valueinthe2MACprotocol,andincludedasaportionoftheti meout(thetimeoutwillbe theARSplusthecostofacontrolpacketroundtriptime). 63

PAGE 81

3.4.10BlockedtoSendStateAnodeisinthisstatewhenBTSarereceivedfrombothitschan nelsorjustfromthe activatedchannel.ItusesthelargestoftheBTStimestoblo ckthechannelforthatperiod oftime.Oncethetimenishes,theBTSvaluesareresetandth enodegoestoaChannel Assignmentstateiftherearepacketstotransmit;otherwis eitgoestotheIdlestate. 3.52MAC'sBackoffAlgorithmThewell-knownBinaryExponentialBackoff(BEB)algorithm wasdesignedinitially todealwithcongestionin802.3networks,andnowtheyareal sousedwith802.11networks.Duringcongestionepochs,nodesareforcedtowaitlo ngerandlongeraftersuccessivecollisionsbydoublingthesizeofthecontentionwi ndoweachtime.Although thishasbeenshowntobeagoodmechanisminwirelesslocalar eanetworks,itsdirect applicationinlongpropagationdelayunderwaterchannels mightnotbeagoodidea.The longpropagationdelaysinunderwatercommunicationsderi vedfromthelowspeedof theacousticchannel,only1500 m = s ,affectsthetimespentforatransmissionprocess betweentwonodes,especiallyinafourhandshakeprocess(R TS/CTS/DATA/ACK).In otherwords,doublingthecontentionwindowateverycollis ionopportunitymaybetoo muchofanincreasesincealargecontentionwindowwillmake thenodewaitforavery longtime.Thenthebackofftimedoesnotneedtobeasbigasin wirelesstominimizethe collisionsduringcontentionamongmultiplenodes.AverysimplemodicationoftheBEBalgorithmthatusesafac torof1 : 25insteadof2 toreducethecontentionwindowisintroduced.The1 : 25valuepresentedintheproposed 64

PAGE 82

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 70 80 90 100 Offered Load in the network Throughput Backoff with 1.25 Backoff with 1.50 Backoff with 1.75 Backoff with 2.00 Figure3.10:Differentbackoffsusingpacketsize=150byte sand BER = 1 x 10 3 backoffalgorithmisobtainedbyvaryingtheinitialfactor 2inthesimulations,andanalyzingthethroughputbehaviorwhentheloadinthenetworki sincreased.InFigure3.10 andFigure3.11thethroughputinanunderwaternetworkfort heproposeddatalinkprotocolisshown,usingdifferentvaluesforthebackofftime. Asseen,thethroughputsare similardespitethesizeofthebackoffinterval.Thereason isthatin2MAC,collisions happenwithaverylowprobabilitybecausenodescanaccesst hechannelsimultaneously, andthereforeaminimumnumberofbackoffswillappear. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 70 80 90 100 Offered Load in the network Throughput Backoff with 1.25 Backoff with 1.50 Backoff with 1.75 Backoff with 2.00 Figure3.11:Differentbackoffsusingpacketsize=300byte sand BER = 1 x 10 4 65

PAGE 83

Theproposedbackoffalgorithmisalsocomparedwithothert raditionalbackoffalgorithmstoverifytheadvantageofusingtheproposedone.Iti scomparedwiththeone presentedin[43],whichincreasesthecontentionwindowby afactorof1 : 75.Thealgorithm,whichisdepictedinEquation3.2,increasesthecont entionwindowbyafactorof 1.25andlimitsthenumberofincrementstoasmallnumber,de pendingontheminimum andmaximumcongestionwindowvalues.Forthecasethatthe CW Min isequalto4,andthe CW Max isequalto9,thenumberof incrementsis3.Oncetheretransmittedpacketgoesthrough ,thecontentionwindowis notresetto CW Min ,asin802.11,butratherisreducedbythatvalueforeverysu ccessful packet.Giventhenumbersfor CW Min and CW Max utilizedforexample,thecontention windowgoesto CW Min injusttwosteps. 8><>: CW = min ( 1 : 25 CW ; CW Max ) whencollisionsoccur CW = max ( CW CW Min ; CW Min ) otherwise (3.2) Figure3.12showsthethroughputobtainedbysimulationsof thethreeschemesasafunctionoftheofferedloadusingpacketsof300bytes,anunderw aterchannelerrormodel derivedfrom[45],andthecombinedSW-MER/2MACdatalinkla yer. Asitcanbeseen,theproposedalgorithmimprovesthethroug hputespeciallyatlownetworkloads,dueto2MACconsiderablyreducingthenumberofc ollisionspresentinthe channel.Lookingatthegure,sincetheperformanceofthet hreealgorithmsissimilar athigherloads,onemaythinkthattheprotocolisnotworkin gwellwhenitissupposed toprovidemorebenets;however,thisisnottrue.Whathapp ensisthatathighloadsthe performanceisdominatedbythebadunderwaterchannel,mos tlybyretransmissionsof packetsinerrorduetothechannel,notduetocollisions. 66

PAGE 84

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 70 80 90 100 Offered Load in the network Throughput 1x10 -3 2MAC Backoff algorithm BEB MILD Figure3.12:Throughputwithdifferentbackoffalgorithms 67

PAGE 85

Chapter4:SW-MER:AStopandWaitWindow-BasedLogicalLink Control ProtocolWithExponentialRetransmissionsTheunderwatercommunicationchannelisverynoisy,andthe biterrorrate(BER)also changesdependingonthedepthandotherfactors,suchassal inity.Normallytheerrors arecorrelated,likeinwirelesscommunications,butfadin gisdeeperandlongerunder waterproducingmoreandburstiererrors.Further,asinwir elesscommunications,transducersforacousticcommunicationsarehalfduplex,negati ngthepossibilityofusing moreefcientowandcongestioncontrolmechanismsatthel ogicallinkcontrollayer. Asaresult,mostdatalinklayersusethesimplebutinefcie ntstopandwaitprotocol insteadofaslidingwindow-basedapproachatthelogicalco ntrolsublayer. Thischapterintroducesanewlogicallinkcontrolprotocol thatusesawindowstopand waitprotocol,andmultiplecopiesofpacketsaresentinexp onentialretransmissions namedSW-MER.TheproposedSW-MERprotocolincludesanexpo nentialretransmissionstrategythatprovidesahighlyreliableserviceoverh igherrorpronechannelsand astopandwaitwindow-basedmechanismtoincreasethechann elutilizationoverlong propagationdelaychannels.SW-MERisdescribedindetaila ndcomparisonswithother traditionallogicallinkcontrolprotocolsareshown. 68

PAGE 86

4.1SW-MERDescriptionTheSW-MERprotocolisacombinationofstopandwaitandslid ingwindowprotocols. Thetransmittersendsagroupof M packets(windowsizeofthetransmission)andthen waitsforthe(group)acknowledgment.Asintheregularstop andwaitprotocol,thesender isnotallowedtosendmorepacketsuntiltheacknowledgment isreceived.Inorderto implementtheSW-MERprotocol,eachpacketcontainsadditi onalinformation,suchas aconsecutivenumberrepresentingitspositioninthewindo w,anumberthattellshow manypacketsaremissingtonishthereceptionofthecurren twindow,andthenumberof timesthepacketisbeingrepeatedinthatwindow.Attheotherend,thereceiververiesthattheincomingpack etsareerror-freeandinsequence.Ifso,itsendsthemtotheupperlayerandacknowledg esthemallinoneACK packettothesender.ThisACKpacketcontainsavector V ofsize M inwhicheverypositioninthevectorreectsthestateofeverypacketreceived ,asfollows: V [ i ]= 8>>>><>>>>: 1 ifpacketiarrivedwitherrorsordidnotarrive 0 ifpacketiarrivedwithouterrors (4.1) UponreceivingtheACKwithvector V [ i ] ,thesenderretransmitsallpackets i thatwere notreceivedcorrectly.AsinanyARQprotocol,theSW-MERse nderalsowaitsforthe correspondingACKpacketbeforearetransmissiontimerexp ires.Ifatimeoutoccurs, thesenderretransmitsthesame M packetsthatweresentinthelastwindow;otherwise, onlythosepacketswith V [ i ]= 1areretransmitted,whicharesentintheinitialpositions ofthefollowingwindow.Ifthereisenoughspaceinthenextw indowafterincludingthe 69

PAGE 87

1 2 3 4 1 2 3 4 wrong ok wrong ok (0,1,1,0,0,0) (0,1,1,0,0,0) 5 6 5 6 okok Sender Receiver 1 2 3 4 1 2 3 4 wrong ok wrong ok (0,1,1,0,0,0) (0,1,1,0,0,0) 5 6 5 6 okok Sender Receiver Figure4.1:Firsttransmission,M=6andpackets2and3arriv ewitherrors. nnnnn nnnnn r r nnnnn nnnnn r r Figure4.2:Secondtransmission,M=6andpacket3arrivewit herrorsagain. retransmittedpackets,thesenderllstherestofthewindo wwithnewpackets.Inthecase ofretransmissionsofalreadyretransmittedpackets,thes enderrepeatsthosepacketsinan exponentialmanner,i.e.,packet i willberetransmitted2 n timesinthefollowingwindow untilamaximumof M times,where n isthenumberofwindowswherepacket i hasbeen receivedwitherrors.Figures4.1and4.2showgraphicallyh owthewindowmechanismof theSW-MERprotocolworksinthecaseofasenderthatwantsto send8packetsusinga windowofsize M = 6. Afterthersttransmission,thereceiversendsanACKpacke twithvector V sayingthat packets2and3arrivedwitherrors(asseeninFigure4.1).Up onthereceptionoftheACK 70

PAGE 88

packet,thesendertransmitspackets2and3twotimeseach,a ndllstherestofthewindowwithnewpackets.Duringthissecondtransmission,pack et3arriveswitherrorsagain (noneofthetwocopiesarrivedwithouterrors).Thereceive rsendstheappropriateACK packetwiththenewvector V (asseeninFigure4.2).Then,thesenderretransmitspacket 32 2 timesinthenextwindow.Theprocesscontinuesuntilthepac ketisreceivedcorrectly orthemaximumnumberofretransmissionsisreached,whichi ssetinourprotocoltothe windowsize M = 6.Withthisprocedure,morecopiesofthesamepacketaretra nsmitted incaseofrepeatederrorswiththeideaofincreasingtherel iabilityoftheprotocol. Animportantaspectoftheprotocolisthesizeofthebuffera tthereceiver.Ifthebuffer isnotdimensionedappropriately,packetsmightbedropped becauseoflackofspace. Inordertoguaranteethatpacketsaresenttotheupperlayer inorder,packetsthatarrivecorrectlyarestoredinmemoryandkeptthereuntilanyo fthemissingpacketsare retransmittedandreceivedcorrectly.Giventhesizeofthe window, M ,andtakinginto accountthatineachretransmissioncopiesofwrongpackets areincreasedexponentially, thesizeofthebuffer BS neededtoguaranteetheoperationoftheprotocolisgivenby Equation4.2as: BS = M ( log 2 M 1 )+ 2(4.2) Todeducethisformula,theworstcasecanbeanalyzedbyhavi ngonlyonepacketretransmitteduntilitsamountofcopiesgetsthewindowsize M ,usingthefollowingexample: with M = 8,thersttime8newpacketswillbesent,thesecondtime2co piesofthe packetarrivedwitherrorsareretransmittedincluding6ne wpackets,thethirdtime4 copiesaresentincluding4newpackets,andthelastretrans missionwillonlyhavecopies ofthepacketarrivedwitherrors.Intotal,atthereceiver1 8differentpacketsarrivethat 71

PAGE 89

needtobetemporarilysavedinthebuffer.Thenthebuffersi zeshouldbe BS = 8 ( log 2 8 1 )+ 2 = 18. 4.2SW-MERStateTransitionsTheproposedlogicallinklayerprotocolhasthefollowings tates:SelectingPossiblePackets,UpdatingCopiesofeachPacketintheWindow,Listen,Wa itingforCTS,Sending WindowofPackets,WaitingforACKVector,IdentifyingPack etsSentwithErrors,ReceivingRTS,PreparingtheCTStobeSent,WaitingtoReceive Packets,Identifyingwhich PacketsdidnotArrive,CheckingPackets,EnqueuingPacket swithoutErrors,VerifyingPacketstobeSenttotheUpperLayer,andGeneratingtheA CKVector.Figures4.3 and4.4showthesenderandreceiverstatemachinesforthene wprotocol,respectively. Thestatesandstatetransitionsneededinthesenderandrec eiverprocessatthelogical linkcontrollayerareexplainednext.4.2.1SelectingPossiblePacketsOncethenodehaspacketstotransmit,someofthemareselect edfromthecorresponding queuethathasallthepacketsgeneratedforitsapplication layer.Theamountofpackets selecteddependsonthenumberofpacketsdenedforthewind owtransmission,andwill bethesameforeachchannel.Thisnumberisaconstantassign edatthebeginningofthe entiretransmission,andisthesamevalueforallthenodesi nthenetwork.Thesepackets aresavedtemporarilyintwobuffers,oneforeachchannel,u ntilitisdecidedinthenext statewhichofthemwillbesent. 72

PAGE 90

n r r n nr r r r rn n r nr r !" n r nr r nr # rn r $r $#r $r %nr rn r r rn Figure4.3:Finitestatemachineoftheprocessinthesender attheLLClayer. 4.2.2UpdatingCopiesofEachPacketintheWindowForeachpacket,thetransmitterkeepsavaluethatrepresen tstheamountofcopiestobe sentineachwindowuntileitherthepacketisacceptedorthe maximumnumberoftrials isreached.Thewindowsizeineachtransmissionis m ,andeachwindowwillgroupnot onlyrepeatedpacketsbutalsonewones,ifthereisenoughsp acetodoit. Thersttimeapacketissent,onlyonecopyistransmitted.I nlaterretransmissionsof thesamepacket,theamountofcopiestobesentisincreasede xponentiallybytwo.This valuewillincreaseuntiltheamountofpacketstobetransmi ttedinthewindowisreached, andoncethishappens,thesameamountofcopieswillcontinu ebeingsentuntilthemax73

PAGE 91

nr nr r nn nr r nr nr! r #r r nr r nr !! $! #r!! !!! % nr && #n r ! n "nr !n !n nr r #nr r 'nr ( !$rr !nr Figure4.4:Finitestatemachineoftheprocessinthereceiv erattheLLClayer. imumnumberoftrialsisaccomplished.Increasingtheamoun tofcopiesforpacketretransmissionswillimprovetheprobabilityofapackettoar rivewithouterrorsinthereceiverandasaconsequence,increasesthereliabilityofth echannelthatisthemaingoal withthisproposedprotocol.Figure4.1and4.2showhowthee xponentialretransmission works,inordertogetamorereliablechannel.4.2.3ListenAfterthewindowofpacketstobetransmittediscreated,the nodestartsalistenprocessto knowwhetherornotothernodesarealsotransmittinginorde rtoavoidcollisions. 74

PAGE 92

Then,anRTSissentinwhichisindicatedthetotaltimeneede dforsendingwindowsof m packetstothereceiver.4.2.4WaitingforCTSThetransmitterwaitsforacertainamountoftime,denedas aCTStimeout,toreceive theCTSfrombothreceivers.Ifthetimeoutisreachedduetoa tleastoneCTSisnotreceived,thenodewillagaininitiateatransmissionprocess withthereceiver;otherwise, theCTSisreceivedinwhichthereceiverisinformingthatth eprocesstostartadata transmissionhasbeenacceptedeitherforonenodeorboth.4.2.5SendingWindowofPacketsInthisstate,thetransmitterstartstosendthewindowof m packetstobothreceivers. Whenthedatatransmissionnishes,thetransmitterwillgo thenextstatetoreceivethe correspondingacknowledgments.4.2.6WaitingforACKVectorThenodestaysinthisstateuntilatleastoneorbothACKvect orsarereceived,otherwise atimeoutisaccomplishedandthenodewillhavetostartthee ntiretransmissionprocess again. 75

PAGE 93

4.2.7IdentifyingPacketsSentWithErrorsAfteratleastoneACKvectorarrivestothesender,thevecto risreviewedtoidentify whichpacketsarrivedwithandwithouterrors.Packetsthat arrivedwithouterrorsare deletedfromthesenderqueue;theothersarekeptuntilamax imumnumberoftrialsis achievedortheyareaccepted.4.2.8ReceivingRTSBecauseeverynodehastwochannels,onepereachneighbor,i tcanreceiveRTScontrol packetsfromeitheroneorbothchannels.Oncethereceivers tartstoacceptanRTS,itwill waitforashorttimetoreceivetheRTSfromtheotherneighbo r.Attheendofthetime, thereceiverwillgotothenextstatetoanswerthesender.4.2.9PreparingtheCTStobeSentAftertheRTScontrolpacketsarrive,eitheroneortwowitht heircorrespondingtimefor eachwindowofdatapackets,thereceiverwillsavethemostt imebetweenbothRTS. Finally,theCTSpacketissenttothetransmitter,eitheron eortwo,asanacceptanceof atransmission. 76

PAGE 94

4.2.10WaitingtoReceivePacketsInthisstate,thereceiverwaitsuntileithereachwindowof packetsarrives,oratimeout occursbecauseatleastoneofthepacketsneverarrivesinth ecorrespondingwindow.If thetimeouthappens,thereceiverisabletoidentifywhichp acketsdidnotcomeinevery windowbecauseeverypacketsentbythetransmitterhastwoa dditionalelds.Therst onerepresentsthepositionofthepacketinthewindow,andt hesecondonerepresentsthe ordinalcopynumber.4.2.11CheckingPacketsThereceiververieswhetherthepacketsarrivedwithorwit houterrorsbyapplyinga well-knownCyclicRedundancyCheck(CRC)function,classi fyingpacketsintothose thatarrivedwithandwithouterrors.4.2.12EnqueuingPacketswithoutErrorsPacketsarrivingwithouterrorsaretemporarilyenqueuedi nthelogicallinklayeruntila consecutivenumberof m packetsarrivewithouterrors,andpacketsarrivingwither rors areeliminated.Thequeuehasamaximumnumberofpacketstob eenqueued.Afterthat value,everypacketthatarriveswithouterrorswillbedrop ped,andthecorresponding informationwillbelost. 77

PAGE 95

4.2.13VerifyingPacketstobeSentOnceaconsecutivenumberof m packetsisachievedinthequeue,theyaresenttothe upperlayeranddeletedfromthequeue.Thepurposeofsendin gthepacketstotheupperlayeristoverifywhetherthosepacketsbelongtothatno deasanaldestination,or whethertheymustberetransmittedtothenextneighbor.4.2.14GeneratingtheACKVectorInthisstate,thereceiverisinchargeofcreatinganACKcon trolpackettotellthetransmitterthestateofeachpacketreceived.ThisACKisavector inwhicheachpositionwill representavalueofeither0or1,where0meansthatthepacke tarrivedwithouterrors, and1meansthatthepacketsarrivedwitherrorsordidnotarr ive.Attheend,theACK vectorissenttothetransmitter. 78

PAGE 96

Chapter5:AnalyticalModelsfortheProposedDataLinkLaye rProtocols Forboth802.11wirelesslocalareanetworksandunderwater acousticnetworks,theMAC layerhasarelevantroleintermsofthethroughputandthere liabilityofthenetwork. Throughput,forexample,hasbeenafundamentalareaofrese archtoobtainamoreefcientMAClayer.Severalmodelshavebeenpresentedintheliteraturetorepr esentthethroughputinsaturatedloadconditionsfor802.11networks,butnotforunde rwaternetworks,inwhich [46]isafundamentalone.In[46]themodelworkswiththeass umptionofhavingonly collisionsinthenetworkandthereisanerror-freechannel condition.Inthatresearcha two-dimensionalMarkovchainisproposedtomodelthebacko ffprocessandthethroughputisdescribedintermsofthatbackoff.In[47],thethroug hputisderivedfrom[46] withthedifferencethatthechannelisnoterror-free.Itis assumedthatcontrolpackets suchasRTS,CTSandACKareverysmallandtheywillneverbeco rruptedbythechannel;asaconsequenceonlydatapacketswillbeaffectedbyth ewirelessmedium.In[48], thethroughputisdescribedasafunctionofbothcollisions anderror-pronechannels,but foraMACprotocolinwhichonlydataandacknowledgmentpack etsareinvolvedina transmissionprocessinwirelessnetworks.Inthischapter,ananalyticalmodelispresentedtoobtaint hethroughputfortheproposed datalinklayerprotocolforunderwaternetworks.Thisanal yticalmodelisbasedonthe 79

PAGE 97

discreteMarkovmodelpresentedin[46]and[48].Thediffer encestartsfromtheMAC protocolwherethesaturationthroughputisobtained.[46] and[48]modelthetraditional 802.11MACprotocolthatusesonechannelpernodetocommuni catewiththeothers,and here[9]istheprotocolthathasbeenmodeledinwhicheachno depresentstwochannels forthecommunication.Anotherdifferenceisthatforthesa turationthroughput,2MAC supposesthatRTS,CTS,andACKcontrolpackets,andalsodat apacketscanbecorruptedbytheerror-pronechannel.Inthenextsection,thedescriptionoftheanalyticalmodel fortheproposeddatalink protocolwillbedividedinto2MACandSW-MER.Intherstsec tion,theanalytical modelfor2MACispresentedandinthesecondsection,theana lyticalmodelforSWMERisintroduced.5.1AnalyticalModelfor2MACAnanalyticalmodelfor2MACisdevelopedtocalculatetheth roughputoftheprotocol, assuminganidealchannelcondition,anitenumberofnodes inthenetwork,andthe exponentialbackoffprocessmodeledasaMarkovchain.Fort hemodeldesign,itisalso assumedaconstantandindependentcollisionprobabilityo fapackettransmittedineach windowtransmissionforeverynodeinthenetwork,regardle ssoftheamountofretransmissions.Theremainderofthissectionisorganizedasfollows.Secti on5.1.1includesabriefdescriptionoftheRTS/CTStransmissionprocessofthe2MACpr otocol.Section5.1.2describesthethroughputanalysis,andSection5.1.3present sthevalidationoftheanalytical model. 80

PAGE 98

5.1.12MACRTS/CTSTransmissionProcessAsexplainedbefore,themechanismtoaccessthechannelimp lementedin2MACisalso aCarrierSenseMultipleAccesswithCollisionAvoidance(C SMA/CA)withRTS/CTS controlpackets,similartotheonedesignedforIEEE802.11 protocols.However,2MAC utilizesvetypeofframes,theseareRequesttoSend(RTS), CleartoSend(CTS),Data, Acknowledgment(ACK)andBlockedtoSend(BTS),asshowninF igure3.7.TheBTS frameisnecessarybecause2MACnodesusetwodifferentchan nels,oneforeachneighbor.IfthereisanodeAwithtwoneighborsBandCanditdecide stotransmitdataonly toB,CdoesnotknowwhathappensbetweenAandBbecauseitisl isteninginadifferent channel.Inthiscase,CreceivesaBTSframeinstead,withth etimethatChastodelayits transmission.Inadditiontotheframes,vetypesoftimingintervalshave beenimplemented.Three ofthem,theShortInterframeSpace(SIFS),theSlotTime(st ),andtheDistributedInterframeSpace(DIFS),arethesameusedinIEEE802.11.Theo thertwoareAdjusted Responsetime(ARS)andChannelAssignmentTime(CHA).ARSt imeisusedfora nodetowaitforsignalsfrombothneighbors.Oncethenodere ceivesaframefromone ofitsneighbors,itwillwaitforanARStimetondoutifanot herframeiscomingfrom theotherneighbor.NodeswaitforaCHAashorttimebeforest artingtheirtransmissions, toidentifywhetherbothchannelsarereadyornottosendpac ketsandalsotoavoidcollisions.Asexplainedinchapter3,whenanodewantstotransmiteithe ronewindowortwowindowsofpackets,thechannelsareactivatedinacertainperi odoftimecalledCHA.Once activated,theyaremonitored.Ifthechannelsareidlefora periodoftimecalledDIFS,the 81

PAGE 99

nodegeneratesarandombackoffintervalbeforethetransmi ssionoftheRTSstarts.This randombackoffintervalisnamedContentionperiodandisme anttoreducepacketcollisions.Otherwise,thebackoffalgorithmistriggeredaft erthecurrenttransmissiontime nishesplusmoreDIFSandCHAtimes,asshowninFigure3.7.O ncethedestination receivesanRTS,itspendsaperiodoftimeARStoverifyifthe reisanotherRTScoming fromtheotherneighborandanotherSIFS.Then,theCTSissen ttothetransmitter.After thetransmitterreceivestheCTS,itwaitsforanARSplusaSI FS,andsendsthedata.The receivertransmitsthecorrespondingACKvectorsoncethed ataisreceived,andafteran ARSplusSIFS.Thebackoffalgorithmadoptedisanexponentialbackoffsch eme,similartotheoneused in802.11.Thetimeisdividedinslots, st ,thebackofftimeisuniformlychoseninan interval(0, W -1),and W isthecontentionwindowthatwillbeexponentiallyincreas ed everytimearetransmissionfails.Atthersttransmission W startsat CW min ,whichis denedastheminimumcontentionwindow.Everytimearetran smissionfails, W isincreasedas W = 1.25 W uptoamaximumvalue CW max ,where CW max =1.25 m CW min and m isthemaximumnumberoftimesthebackoffwillbeexecuted,n amedasthemaximum backoffstage.Everytimethechannelissensedidle,thebac kofftimeisdecrementedin oneslottime,otherwisewhenatransmissionisdetectedthe nthebackofftimeisfrozen andreactivatedoncethechannelisidleagain.Thenodetran smitsonlywhenthebackoff slottimereachesthe0value. 82

PAGE 100

5.1.2ThroughputAnalysisTheanalyticalmodelforthethroughputissolvedinasimila rwaylikein[46].Oneofthe differencesisthatin[46]problemssuchashiddenandexpos edterminalarenotsolved, anidealchannelisassumed.In2MAC,hiddenandexposedterm inalproblemsdonot existbecausetheprotocolworkswithtwochannelsineachno de,oneperneighboras explainedbefore,solvingthepossibilityofoccurringthe seproblems. Toobtainthethroughputanalysis,axednumberofnodeswhe reeachnodealwayshas apacketavailabletotransmit,isassumed.Asaconsequence ,anon-emptytransmission queueofeachnodeisassumedallthetime.Inthenextsection,thebackoffprocessispresentedbymean sofaMarkovchainmodel. BasedontheMarkovchain,theprobabilitytforanodetobetransmittinginatransmissionperiodiscalculated.Attheend,thethroughputisobta inedasafunctionoft. 5.1.2.1PacketTransmissionProbabilityToobtainthepackettransmissionprobability, n nodesareassumedinthenetwork,and anintegertimescaleisadoptedwhere t and t + 1representconsecutiveslottimes st Eachnodehasapacketavailabletotransmitafterasuccessf ultransmission,andwaits forarandombackofftimegeneratedandsavedinabackofftim ecounter.Ineachslot timethechannelissensed.Ifthechannelisbusythenthebac kofftimecounterstops, otherwiseitisdecrementeduntilitreaches0.Asexplained before,thecontentionwindow W startsat CW min andincreasesuntilitreachesthemaximumcontentionwindo w CW max where CW max =1.25 m W ; m representsthemaximumbackoffstage,and W i =1.25 i W ,such 83

PAGE 101

n r Figure5.1:Markovchainforthebackoffprocess[46]. that i 2 ( 0 ; m ) .Ineachretransmissionattempt,apacketcancollidewitha constantand independentprobability p ,this p isalsodenedin[46]asthetransitionprobabilityfrom onestagetoanother.Likein[46],thebackoffprocessismod eledwithaMarkovchainin whichpij representsthesteadystateprobabilityforabackoffstage i andbackoffcounter j ,asshowninFigure5.1. Following[46],intheMarkovchainevery( i W i )representsabidimensionalstatewhere therstparameterrepresentsthebackoffstagei 2 (0, m )andthesecondparameteris thebackoffcounter(0, W i )wherei 2 (0, m ).Thenon-nullone-stepprobabilitiesderived in[46]fromtheMarkovchainare: 84

PAGE 102

P f ( i ; k ) j ( i ; k + 1 ) g = 1 ; k 2 ( 0 ; W i 2 ) ; i 2 ( 0 ; m ) (5.1) P f ( 0 ; k ) j ( i ; 0 ) g =( 1 p ) = W 0 ; k 2 ( 0 ; W 0 1 ) ; i 2 ( 0 ; m ) (5.2) P f ( i ; k ) j ( i 1 ; 0 ) g = p = W i ; k 2 ( 0 ; W i 1 ) ; i 2 ( 1 ; m ) (5.3) P f ( m ; k ) j ( m ; 0 ) g = p = W m ; k 2 ( 0 ; W m 1 ) (5.4) Theseprobabilitiesrepresenttheconditionalcollisionp robabilitiesasexplainedin[46]in which(i,j)representsthebackoffstage i withthebackofftimecounter j .Theequation5.1 meansthatatthebeginningofeachslottimethebackofftime counterisdecremented withprobabilityof1.TheprobabilityinEquation5.2indic atesthatafterasuccessful transmission,anewpackettobetransmittedrequiresthatt hetransmitterstartswithits backoffstage0.Thecorrespondingbackofftimecounterwil lbeinitializedwithanumber uniformlychosenintherange(0, W 0 -1). TheprobabilityinEquation5.3representswhenanunsucces sfultransmissionoccursat thebackoffstage i -1,thenthebackoffstateincreasesandthenewbackofftime counter isinitializedinavalueuniformlyselectedfrom(0, W i ).Finally,Equation5.4meansthat oncethebackoffstagereaches m ,itwillnotbeincreasedinanewpackettransmission. Itwillstaythereuntilthemaximumnumberoftrialsdenedb ytheprotocolisreached; thenthepacketwillbedropped. 85

PAGE 103

Applyingthenon-nullone-stepprobabilitiesinthebackof fmodelproposedin2MAC[9], theequationsarederivedasfollows: P f i ; k j i ; k + 1 g = 1 ; k 2 ( 0 ; W i 2 ) ; i 2 ( 0 ; m ) (5.5) P f 0 ; k j i ; 0 g = 1 p W ; k 2 ( 0 ; W 1 ) ; i 2 ( 0 ; m ) (5.6) P f i ; k j i 1 ; 0 g = p 1 : 25 i W ; k 2 ( 0 ; 1 : 25 i W 1 ) ; i 2 ( 1 ; m ) (5.7) P f m ; k j m ; 0 g = p 1 : 25 m W ; k 2 ( 0 ; 1 : 25 m W 1 ) (5.8) Thesteadystateprobabilitiesderivedin[46]fortheMarko vchainrepresentedbypi ; j ,for abackoffstage i andabackoffcounter j ,arethefollowing:pi ; 0 = p pi 1 ; 0 )pi ; 0 = p i p0 ; 0 0 < i < m (5.9)pm 1 ; 0 p =pm ; 0 )pm 1 ; 0 p =( 1 p ) pm ; 0 )pm ; 0 = p pm 1 ; 0 ( 1 p ) )pm ; 0 = p m p0 ; 0 ( 1 p ) (5.10) 86

PAGE 104

p0 ; 0 =( 1 p ) m j = 0pj ; 0 (5.11)t= m j = 0pj ; 0 (5.12) Theonlypossibilitytoentertheset f ( 0 ; 0 ) ; ( 0 ; 1 ) ;:::; ( 0 ; W 0 1 ) g isafterasuccessful windowpackettransmissionfrom f ( 0 ; 0 ) ; ( 1 ; 0 ) ; ( 2 ; 0 ) :::; ( m ; 0 ) g .ThesuccessfultransmissionhappenswiththeprobabilityshowninEquation5.13 ( 1 p ) m j = 0pj ; 0 (5.13) Acountervalueischosenfollowedintherange f 0 ; 1 ;:::; k ;:::; W 0 1 g ,whichhappens withprobability W 0 k W 0 .Then,having k 2 ( 1 ; W i 1 ) allthesteadystateprobabilitiesare obtainedasfollows:pi ; k = W i k W i 8>>>>>><>>>>>>: ( 1 p ) m j = 0pj ; 0 ; if i = 0; p pi 1 ; 0 ; if0 < i < m ; p (pm 1 ; 0 +pm ; 0 ) ; if i = m (5.14) 87

PAGE 105

BasedonEquations5.10and5.11,Equation5.14canbemodie dandthesteadystate probabilitycanbeobtainedasfollows:pi ; k = W i k W i 8>>>>>><>>>>>>:p0 ; 0 ; if i = 0;pi ; 0 ; if0 < i < m ;pm ; 0 ; if i = m (5.15) )pi ; k = W i k W i pi ; 0 ; i 2 ( 0 ; m ) ; k 2 ( 0 ; W i 1 ) (5.16) Usingthenormalizationconditionthatthesumofallsteady stateprobabilitiesshouldbe equalto1,thefollowingequationisobtained: 1 = m i = 0 W i 1 j = 0pi ; j = m i = 0pi ; 0 W i 1 j = 0 W i j W i = m i = 0pi ; 0 W i 1 j = 0 1 1 W i W i 1 j = 0 j # = = m i = 0pi ; 0 W i + 1 2 = 1 2 m i = 0pi ; 0 W i + m i = 0pi ; 0 # = 1 2 m i = 0pi ; 0 W i +p0 ; 0 ( 1 p ) # = = 1 2 m 1 i = 0pi ; 0 W i +pm ; 0 W m +p0 ; 0 ( 1 p ) # = = 1 2 m 1 i = 0 p i p0 ; 0 W i + p m ( 1 p ) p0 ; 0 W m +p0 ; 0 ( 1 p ) # (5.17) 88

PAGE 106

Byreplacingthebackofftimedenedin[9]thatis W i =( 1 : 25 ) i W = 5 4 i W inEquation5.17,thestationarydistributionoftheMarkovchainf orabackoffstage0andabackoffcounter0for2MACisobtained. 1 =p0 ; 0 2 m 1 i = 0 p i 5 4 i W + p m ( 1 p ) 5 4 m W + 1 ( 1 p ) # (5.18) )p0 ; 0 = 8 1 5 4 p ( 1 p ) 4 ( W + 1 ) 1 5 4 p + Wp 1 5 4 p m (5.19) TheprocesstoobtaintheEquation5.19fromEquation5.18is describedinAppendixA. ThepackettransmissionprobabilityortheprobabilitytthatanodetransmitsinarandomlychosentransmissionperioddisplayedinEquation5.1 2for2MAC,canberepresentedintermsof p and W .t=p0 ; 0 ( 1 p ) = 8 1 5 4 p 4 ( W + 1 ) 1 5 4 p + Wp 1 5 4 p m = = 8 4 ( W + 1 )+ Wp 1 ( 5 4 p ) m 1 ( 5 4 p ) )t( p )= 8 4 ( W + 1 )+ Wp 1 ( 5 4 p ) m 1 ( 5 4 p ) (5.20) Now,itisnecessarytoverifyiftiscontinuousintherange p 2 ( 0 ; 1 ) .Todothis,tis evaluatedin0 ; 1and4 = 5: 89

PAGE 107

t( 0 )= 8 4 ( W + 1 )+ 0 )t( 0 )= 2 ( W + 1 ) (5.21)t( 1 )= 8 4 ( W + 1 )+ W 1 ( 5 4 ) m 1 ( 5 4 ) (5.22) Toevaluatetin4 = 5,Equation5.20canberepresentedasfollows:t( p )= 8 4 ( W + 1 )+ Wp m 1 i = 0 5 4 p i (5.23) )t 4 5 = 2 ( W + 1 )+ Wm 1 5 (5.24) ConsideringtheresultsfromEquations5.21,5.22and5.24, itisprovedthattiscontinuousintherangeof p Thevalueoftiscalculatedinsuchawaythattheprobability p isobtainedasfollows: p = 1 ( 1 P c ) ( 1 p e ) (5.25) 90

PAGE 108

Thetransmissionerroreventgeneratedduetoerrorsinthec hannel p e iscalculatedinthe followingequation: p e = p rtse + p ctse + p datae + p acke p rtse p ctse p datae p acke (5.26) P c = 1 ( 1 t) n 1 (5.27) 1 P c representsthatthereisnocollisionwhenapacketwastrans mittedand1 P e means thattherearenobitswitherrorsinareceivedpacket.Thede scriptionsofthevariables canbeseeninTable5.1.5.1.2.2SaturationThroughputSaturationthroughputisdenedastheratioofsuccessfuld atatransmittedoverarandomlychosentransmissionperiod,asseeninEquation5.28. T = E[numberofchannelsidle] E[Datatransmittedinatransmissionperiod] E[lengthofthetransmissionperiod] (5.28) Toobtainthesaturationthroughputlikein[46],itisneces sarytoanalyzewhathappens inarandomlychosentransmissionperiod.Therearesevendi fferenttransmissionperiod situations,asshowninFigure5.2.Tocalculatethesetrans missions,itisassumedthatthe timerequiredeithertotransmitaCTS,anRTSoranACKcontro lpacketarethesame anditisnamedas T ctl 91

PAGE 109

n r r r r n r n r Figure5.2:Transmissionperiodsinanerror-pronechannel 92

PAGE 110

Table5.1:Notationsusedtoobtainthesaturationthroughp ut. T i Theidletransmissionperiod T suc Transmissionperiodinwhichthechannelissensedbusy becauseofasuccessfultransmission RTS RequestToSendcontrolpacket CTS ClearToSendcontrolpacket ACK Acknowledgmentcontrolpacket T col Timethatthechannelissensedbusybecauseofa collision T rts E Timethatthechannelissensedbusybecause T cts E Timethatthechannelissensedbusybecauseofan ofanRTSframeerrortransmission RTSframeerrortransmission T data E Timethatthechannelissensedbusybecause T ack E Timethatthechannelissensedbusybecauseofan ofadataframeerrortransmission ACKframeerrortransmission P tr Probabilityofhavingatleastonenode p Transmissionfailureprobabilityofanodeeither transmittinginatransmissionperiod foracollisionoratransmissionerrorevent P x Probabilitythatonlyonenodetransmitsonthe P c Probabilityofacollisionseenbyapacketbeing channel transmittedonthechannel P suc Probabilityofasuccessfultransmissionina H Packetheader transmissionperiod T 2 mac s Averagetimethatthechannelissensedbusy T 2 mac c Averagetimethatthechannelissensedbusybyeach becauseofasuccessfultransmission nodeduringacollision E [ Pkt tr ] Averagedatasizesuccessfullytransmittedin E [ W pkt ] Averagewindowpacketsizetobetransmitted atransmissionperiod E [ L st ] Averagelengthofatransmissionperiod w Numberofpacketstobetransmittedinawindow Pkt size Sizeofthedatapacketinbits n Numberofnodesinthenetworkthatcancontendthe channel,eachonetransmittingwithprobabilityt p datae Dataframeerrorprobability p rtse RTSframeerrorprobability p ctse CTSframeerrorprobability p acke ACKframeerrorprobability lrts SizeinbitsoftheRTSframe lcts SizeinbitsoftheCTSframe ldata SizeinbitsoftheDATAframe lack SizeinbitsoftheACKframe BER Biterrorprobability a Durationofanemptytransmissionperiod T rts DurationofanRTScontrolpackettransmission T cts DurationofaCTScontrolpackettransmission T txd DurationofaDATApackettransmission T txa DurationofanACKcontrolpackettransmission ARS Adjustedresponsetime CHA Channelassignmentduration P rts E Probabilitythatatransmissionerroroccurs P cts E ProbabilitythatanRTSframeissuccessfullytransmibecausetheRTSframeiscorruptedwhenonly ttedbutthecorrespondingCTSframeiscorrupted onenodeistransmitting becauseoftransmissionerrors P data E ProbabilitythatanRTSandCTSaresuccessfully P ack E ProbabilitythatanRTS,CTSanddataframesis transmittedbutthecorrespondingdataframeare successfullytransmittedbuttheACKframeis corrupted corruptedduetotransmissionerrors T prop Propagationdelay T ctl Durationofacontrolpackettransmission P col Probabilitythatatleasttwonodesstarttransmissions inasametransmissionperiod Therstcaseinthetransmissionperiodcorrespondstothet imethenodeisidle,since therearenopacketstotransmit.Thesecondcaserepresents thetimespentforanodeto transmitthedatastartingwiththeRTStransmissionandni shingwiththeACKreception.Thethirdcaseidentiesthattherewasacollisionatt hetimeanRTSwassent.The fourthcaserepresentsthattheRTSarrivedwitherrorstoth ereceiverbecauseoferrorsin thechannel. 93

PAGE 111

Thefthcasedenesthetransmissionperiodgeneratedwhen theCTSarrivedwitherrors tothesenderduetochannelproblems.Thesixthcaseistheti merequiredwhenthedata arrivedwitherrorstothereceptorbecauseoferrorsinthec hannelandthelastcaseis thetimespentwhentheACKarrivedwitherrorsatthesendera lsoduetoerrorsinthe channel.Thecorrespondingvariablesthatrepresentthesetransmis sionperiodsaredenedinTable5.1,andtheycanbeobtainedasfollows: T ctl = T txa = T cts = T rts T suc = T txd + 4 ( ARS + T prop )+ 3 ( T ctl + SIFS )+ CHA + DIFS (5.29) T cts E = 2 ( T ctl + ARS + T prop + SIFS ) (5.30) T col = T ctl + ARS + T prop + SIFS (5.31) T rts E = T col (5.32) T ack E = T data E = T suc (5.33) 94

PAGE 112

Asaconsequenceofthesituationspresentedinatransmissi onperiod,someothervariablesorparametersthataffectthethroughputhavetobede ned,followingasimilar strategyusedforerror-pronewirelessnetworksin[48].Th esevariablesarealsoshown inTable5.1,andthecorrespondingequationsneededtocomp utethemarethefollowing: P tr = 1 ( 1 t) n (5.34) inwhich ( 1 t) n istheprobabilityofnotransmissioninatransmissionperi odforthe n nodesinthenetwork.Theprobabilityof P tr and P x aredenedinTable5.1,andare neededtoknowtheprobability P col thatatleasttwonodestransmitinthesametransmissionperiodgeneratingcollisions. P col affectsthethroughputofthenetworkbecausethe collisionoftwonodeswillincrementthetotaltimeforthos enodestoobtainasuccessful transmission. P x = nt( 1 t) n 1 (5.35) P col = 1 ( 1 t) n nt( 1 t) n 1 = P tr P x (5.36) Thefollowingequationalsoaffectsthethroughputintheca sethatatransmissionissuccessful,soitdisplaystheprobabilitythatthesuccessful transmissioncanoccur. P suc = P x 1 p rtse 1 p ctse 1 p datae 1 p acke (5.37) 95

PAGE 113

Alloftheprobabilitiesincludedinasuccessfultransmiss ionprobabilityarecalculated basedon BER asfollows: p rtse = 1 ( 1 BER ) lrts (5.38) p ctse = 1 ( 1 BER ) lcts (5.39) p datae = 1 ( 1 BER ) ldata (5.40) Thedataframesizeisdenedas ldata = H + Pkt size p acke = 1 ( 1 BER ) lack (5.41) Toobtainthethroughput,theaveragedatasizesuccessfull ytransmittedshouldbeknown. Thisaveragedependsontheaveragewindowpacketsizethatm ustbesubmittedandis calculatedasfollows: E [ W pkt ]= w Pkt size (5.42) 96

PAGE 114

Then,theaveragedatasizesuccessfullytransmittedthatd ependsontheaveragewindow packetsizeisobtainedinthefollowingequation. E [ Pkt tr ]= E [ W pkt ] P suc (5.43) Theexpectedvalueofthelengthofthetransmissionperiodw here(1P tr )istheprobabilityofnotransmissioninatransmissionperiod(idletransm issionperiod),canbecalculatedasfollows: E [ L st ]=( 1 P tr )a+ T suc P suc + P col T col + T data E P data E + T ack E P ack E + + T rts E P rts E + T cts E P cts E (5.44) P rts E = P x p rtse (5.45) P cts E = P x 1 p rtse p ctse (5.46) P data E = P x 1 p rtse 1 p ctse p datae (5.47) P ack E = P x 1 p rtse 1 p ctse 1 p datae p acke (5.48) 97

PAGE 115

Everynodein2MAChastwochannels,oneperneighbor,andtos tartatransmissionitis onlynecessarytohaveatleastonechannelidle.Then,thenu mberofchannelsexpected tobeidleare: E[numberofchannelsidle] = numberofchannels P[atleastonechannelisidle] = = 2 ( : 25 + : 25 + : 25 )= 1 : 5 (5.49) Then,thesaturationthroughput T iscalculatedasfollows: T = E[numberofchannelsidle] E[Datatransmittedinatransmissionperiod] E[lengthofthetransmissionperiod] = 1 : 5 E [ Pkt tr ] E [ L st ] T = 1 : 5 E [ W pkt ] P suc ( 1 P tr )a+ T suc P suc + P col T col + T data E P data E + T ack E P ack E + T rts E P rts E + T cts E P cts E (5.50) 5.1.3ModelValidationResultsforthethroughputintheanalyticalmodelhavebeen comparedwiththeones obtainedbyrunningthesimulations.Theseresultsarebase dontheparametersincluded inTable5.2. 98

PAGE 116

Duetothecharacteristicsoftheproposedprotocol,everyn odehastwochannels,and amaximumoftwoneighbors,oneperchannel,duringthetrans missionprocess.This meansthatthenumberofnodescontendingthechannelwillbe n = 2. Table5.2:Parametervaluesusedtovalidatethesaturation throughput. Windowsize 8 ; 16 ; 32packets Datarate 19600b/s Biterrorrate 1 x 10 3 ; 1 x 10 4 Distance 50m ACKpacketsize 30bits Overhead 8bits Speedofsound 1500m/s Packetsize 150 ; 300 ; 600 ; 1200 ; 2400 ; 4800bits 500 1000 1500 2000 2500 3000 3500 4000 4500 0 50 100 150 200 250 300 Packet size Throughput simulation analytical Figure5.3:Throughputcomparisonwith BER = 1 x 10 3 Figures5.3and5.4showthatsimulationresultsarereallyc losecomparedwiththeresultsgivenbytheanalyticalmodel,meaningthattheanalyt icalmodelofthethroughputisaccurate,representingasaconsequencethethroughp utof2MAC.Thereisahuge throughputimprovementinFigure5.4sincetheBERis1 x 10 4 ,meaningthattheprobabilityofhavingerrorsinapacketarrivedtothereceiveri sverylowcomparedwiththe BERinFigure5.3,andtherewillindeedbefewerretransmiss ions. 99

PAGE 117

500 1000 1500 2000 2500 3000 3500 4000 4500 0 50 100 150 200 250 300 Packet size Throughput simulation analytical Figure5.4:Throughputcomparisonwith BER = 1 x 10 4 5.2AnalyticalModelforSW-MERThissectionpresentstheanalyticalmodeltocalculatethe throughputoftheSW-MER protocol.Thethroughputisobtainedasafunctionofthepac ketsize,bitrateandtheexpectedbiterrorprobability.Theanalysisassumesaniten umberofnodesinthenetwork. Theremainderofthissectionisorganizedasfollows.Secti on5.2.1includesabriefdescriptionofsomecomponentsneededforthetransmissionpr ocessatthelogicallinklevel forunderwatercommunications.Section5.2.2describesth ethroughputanalysis,and Section5.2.3presentstheanalyticalmodelvalidation.5.2.1SW-MERTransmissionProcessTheprocessimplementedinSW-MERforsendingpacketsissim ilartotheonedesigned in[33],howeverthedifferenceisinthenumberofcopiessen tforeverypacketarrived witherrors,inthenextwindowretransmission. 100

PAGE 118

Table5.3:Notationsusedtoobtainthethroughputefcienc y. T prop Packetpropagationdelay T txd Datapackettransmissiontime T procd Datapacketprocessingdelay T txa Acknowledgmentpackettransmissiontime T proca Acknowledgmentpacketprocessingdelay T tot Totaltimeneededtosendthedata n r Figure5.5:SW-MER,transmissionprocess. Asexplainedinchapter4,thisisexponentiallyincreasede verytimethepacketisretransmitted.Figure5.5showsthetimesinvolvedinthetran smissionofadatapacket usingtheSW-MERprotocol,andTable5.3describeseachofth evariables.Following thegure,thetotaltime T tot neededforthetransmittertosendawindowof w dataframes andreceivethecorrespondingacknowledgmentpacket,isca lculatedasfollows: T tot = T txd + T txa + 2 T prop + T procd + T proca (5.51) 101

PAGE 119

Factorsliketheswitchtimespenteitherinthetransmitter orthereceivertogotothe listeningstatetothesendingstatealsocanaffectthetota ltime.If T procd and T proca are assumednegligibleandtheswitchtimeisincluded,Equatio n5.51canberewrittenas follows: T tot = T txd + T txa + 2 ( T prop + T sw ) (5.52) Inthenextsectionthethroughputisderived.5.2.2ThroughputAnalysisTheanalyticalmodelforthethroughputissolvedasin[35], computingthethroughput efciencyoftheSW-MERprotocol.Itisassumedthat D f = D d + D oh inwhich D f isthe sizeofeachdataframeinbits, D d isthelengthofthedatatobetransmittedinbits,and D oh isthepacketoverhead. Thesizeoftheacknowledgeframeinbitsisrepresentedas D ack ,thespeedofthesound underwaterisby c where c = 1500 m = s d isthedistancebetweentransmitterandreceiver, and r isthedatarateinbits.Basedonthesevariables,thepropag ationdelaycanbecalculatedas T prop = d c ,thetransmissiontimeneededtosendawindowofdataframes as T txd = w D f r ,theswitchtimeas T sw = 16 r (formulatakenfrom[35]),andthetransmission timeneededtosendanacknowledgepacketas T txa = D ack r .Theequationforthethroughputefciencyisrepresentedinthefollowingequation: 102

PAGE 120

h= T d T f (5.53) T d representsthetimenecessarytotransmitawindowof w datapacketsand T f isthe averagetimeneededtotransmitawindowof w dataframessuccessfully. T d isobtained inEquation5.54,and T f inEquation5.55. T d = w D d r (5.54) T f = E [ w ] T tot (5.55) E [ w ] istheaveragenumberoftrialsneededtotransmitawindowof w packetssuccessfullyandiscalculatedasfollows, E [ w ]= k = 0 2 k 1 P 2 k 1 1 2 k 1 ( 1 P ) = (5.56) = k = 0 4 k 1 P 2 k 1 1 ( 1 P ) (5.57) P isthepacketerrorprobabilityandiscalculatedas P = 1 ( 1 BER ) D f ,where BER representsthebiterrorprobabilityorbiterrorrateinthe channel,and D f representsthe 103

PAGE 121

sizeofthedataframeinbits.Theprobabilityofgettingatl eastonepacketwithouterrors arrivingtothereceiverinthe2 k 1 trialiscalculatedas P ( 2 k 1 1 ) 2 k 1 ( 1 P ) Takingintoaccounttheequationspresentedbefore,thethr oughputefciencyinEquation5.53canberepresentedalsointermsofthelengthofthe datatobetransmittedas follows:h= D d [ 1 BER ] D d + D oh k = 0 4 k 1 1 ( 1 BER ) D d + D oh ( 2 ( k 1 ) 1 ) [ D d + e ] (5.58) inwhich e is, e = D ack w + 2 r ( T sw + T prop ) w + D oh (5.59) TheprocesstoobtaintheEquation5.58isdescribedinAppen dixB. 5.2.3ModelValidationTovalidatetheanalyticalmodel,theresultshavebeencomp aredwiththeonesobtained withthesimulationsinchapter4.Thevaluesoftheparamete rsusedtoobtaintheresults forboththesimulationprogramthatrepresentsthepropose ddatalinkprotocolandthe analyticalmodelcanbeseeninTable5.4. 104

PAGE 122

Table5.4:Parametervaluesusedtovalidatethethroughput efciency. Windowsize 8 ; 16 ; 32packets Datarate 2400b/s Biterrorrate 1 x 10 3 Distance 50m ACKpacketsize 30bits Overhead 8bits Speedofsound 1500m/s Packetsize 150 ; 300 ; 600 ; 1200 ; 2400 ; 4800bits Figures5.6,5.7,and5.8showthecomparisonoftheanalytic alandsimulationresults oftheSW-MERprotocol.Asitcanbeseen,theresultsarevery close,meaningthatthe analyticalmodelofthethroughputefciencyisaccurate,r epresentingthethroughputof SW-MER.Theresultsshownintheguresarejustforvalidationpurpo ses,thechapter6presentsthe evaluationoftheSW-MERprotocol. 500 1000 1500 2000 2500 3000 3500 4000 4500 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency simulation analytical Figure5.6:Throughputefciencycomparisonwithawindows izeof8packets. 105

PAGE 123

500 1000 1500 2000 2500 3000 3500 4000 4500 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency simulation analytical Figure5.7:Throughputefciencycomparisonwithawindows izeof16packets. 500 1000 1500 2000 2500 3000 3500 4000 4500 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency simulation analytical Figure5.8:Throughputefciencycomparisonwithawindows izeof32packets. 106

PAGE 124

Chapter6:PerformanceEvaluationThischapterpresentstheresultsoftheperformanceevalua tionofthe2MACandSWMERprotocols.Threesetsofresultsareincluded.Thersta ndsecondsetsincludethe evaluationoftheMACprotocolsandthelogicallinkcontrol protocolsalone,respectively. ThethirdsetpresentstheevaluationoftheSW-MERand2MACp rotocolstogether(the datalinklayer).6.1ScenariosandParametersInallexperimentsthemaximumnumberoftrialsappliedinth eprotocols[49],[33]and [36]isthemaximumassignedbythe802.11fordatapackets,4 .Toobtainthepacket errorprobability,basedontheBER,Equation6.1inwhich D f = D d + D oh wasapplied. Theevaluationisperformedusingtwoerrormodels.Therst onecorrespondstothe Bernoullimodelandthesecondonetothechannelcharacteri zationresultspresented in[45].Thescenariosinwhichtheprotocolswereevaluated arelinearnetworkswith fournodes.Theparametersusedintheevaluationscanbesee ninTable5.4.Theseerror modelsareexplainedinmoredetailnext. 107

PAGE 125

Table6.1:Parametervaluesusedtoevaluate2MACandSW-MER Windowsizes 8 ; 16packets Datarate 2400b/s Biterrorrate 1 x 10 3 ,1 x 10 4 Distance 50m ACKpacketsize 30bits Overhead 8bits Acousticspeed 1500m/s Packetsizes a)150 ; 300 ; 600 ; 1200 ; 2400 ; 4800bits Packetsizes b)100 ; 200 ; 300 ; 400 ; 500 ; 600bits sinchronizationtime 0 : 16s 6.2ChannelErrorModelsInordertoevaluatetheperformanceofthedatalinkcontrol protocolsunderconsideration,twochannelerrormodelswereutilized.Therstmodel correspondstothesimple BernoullimodelinwhichthePacketErrorProbability(PER) canbeeasilycalculatedas: PER = 1 ( 1 BER ) D f (6.1) where N isthenumberofbitsinthepacketandBERistheBitErrorRate ThesecondmodelnamedastheMarkovmodelerrorchannel,was developedbymodelingtheerrorsofanunderwaterchannelusingasimilarappro achastheonesdescribed in[50]and[51],whereatwo-stateMarkovchainisusedtomod eltheerrorsinwireless communicationschannels.Therststepinthemodelgenerat ionisthecharacterization ofthechannel.Itisimportanttotakeintoaccountthatreal measurementswerenotperformed;instead,thechannelcharacterizationresultspre sentedin[45]wereused,where realmeasurementsweretakenusingacousticcommunication sinshallowwater(depth 15-20m)atarangeof50m.From[45],thechannelimpulses(theamplitudesandthevalu es)wereusedtorecreate thechannelbehavior,includingtheeffectofmobilitycons ideringthedopplereffectwhile movingat1m/sec,adepthof4m,andadistanceof50m.UsingMa tlabsimulations,a 108

PAGE 126

Figure6.1:Two-stateMarkovmodelrepresentation. traceof10millionimpulseswasgeneratedusingOrthogonal Frequency-DivisionMultiplexing(OFDM)andBinaryPhaseShiftKeying(BPSK)modulat ionschemes.Thegeneratedimpulseresults,+1,-1and0weretakenasarepresent ationofbitwitherrors(+1 and-1values)andbitswithouterrors(0value).Then,witht histrace,thetransitionprobabilitymatrix A andtheerrorprobabilitymatrix B ofthetwo-stateMarkovchainshown inFigure6.1wereobtained.Giveninitialvaluesof A and B ,aHiddenMarkovModel (HMM)approachwastakentogeneratethechannelerrormodel [52].Inthisprocessthe well-knownBaum-Welchalgorithm[53]wasusedtondtheunk nownparametersofthe HMM.Theinitialvaluesofthe A and B matrices( A 0 and B 0 )aswellasthenalvalues, nalsteadystatematrices A ss and B ss ,foundbythemodelareasfollows: A 0 = 264 0 : 980 : 02 0 : 050 : 95 375 A ss = 264 0 : 81160 : 1884 0 : 00950 : 9905 375 (6.2) B 0 = 264 0 : 90 : 9 0 : 10 : 1 375 B ss = 264 0 : 99090 : 68 0 : 00910 : 32 375 (6.3) AfterdoingsomesimulationsusingtheMarkovmodelerrorch annelover10 ; 000 ; 000 bitssentbythetransmitter,theamountofbitswitherrorsa rrivedtothereceiverisobtained,andalsotheaverageBERgenerated.TheBERobtained is2 : 4 x 10 2 ,meaningthat thiserrorchannelishighinerrors. 109

PAGE 127

6.3PerformanceEvaluationfor2MACTheexperimentsattheMAClayercomparetheIEEE802.11prot ocolwiththeproposed 2MACprotocolusingbothachannelspeedof2400bpsandpacke tsizesof1200and 2400bits.Theprotocolswereevaluatedconsideringalinea rnetworkoffournodes50m apartsimulatingaswarmofAUVsmappingtheoceanoorwitho uterrorsandwithout thelogicallinkcontrolsublayer.Thetrafcinthenetworkwasgeneratedasfollows.Eachnode generatedowedtoall othernodes,sendingpacketsaccordingtoaPoissonprocess .Therateoftheowswasset sothattheloadinthenetworkwassettothespecicdesiredl evel.Inaddition,atagged owsendingpacketsfromnodeone(leftextremeofthenetwor k)tonodefour(rightmost node)wasestablishedandmonitored.Theperformanceresul tsshowninallgraphsare relatedtotheperformanceofthistaggedow,which,asaref erence,isindicatedinthe plotbyadottedlinewithamaximumpossiblethroughputof60 bpsor30bpsinthecase ofpacketsizesof2400and1200bits,respectively.Asimple networklayerwasincluded ontopoftheMAClayertorouteincomingpacketstothefollow ingadjacentnodeand theassumptionofanunderwaterchannelwithouterrors.The objectivewastoanalyze theamountofcollisionsproducedforeachMACprotocolunde rhightrafclevelinthe network.Figure6.2and6.3showthethroughputresultsoftheMACprot ocols.Asitcanbeseen fromthegures,thetrafcloadofthenetworkisincreasedf rom10%to100%and2MAC stillpresentsabetterthroughputperformancecomparedwi ththetraditionalCSMA/CA protocol.Thiscanbeeasilyexplainedbytheuseofmultiple channels,whichreducesthe numberofcollisions.Also,packetsaresentthroughbothch annelsin2MACincreasing 110

PAGE 128

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 Offered Load in the network Throughput (bps) 1200 bits 802.11 1200 bits 2MAC 30 bps flow Figure6.2:ThroughputoftheMACprotocols,packetsizeof1 200bits. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 20 40 60 80 100 Offered Load in the network Throughput (bps) 2400 bits 802.11 2400 bits 2MAC 60 bps flow Figure6.3:ThroughputoftheMACprotocols,packetsizeof2 400bits. thethroughput,butinCSMAonlyonechannelisusedforthetr ansmission. 6.4PerformanceEvaluationforSW-MERInthissection,theperformanceoftheproposedSW-MERprot ocoliscomparedwith thestopandwaitprotocolsdescribedin[49],[33]and[36]( shownintheplotsas sw 2, sw 3,and sw 4,respectively).Thestopandwaitdescribedin[49]worksb ysendingone 111

PAGE 129

packetallthetimefromthesender.Oncethepacketarrivest othereceiver,itveriesif thepacketarriveswithorwithouterrorsandsendsanacknow ledgmenttothetransmitter.Thenthetransmitterwillretransmitthepacketuntili tarriveswithouterrorstothe receiver.Thestopandwaitvariantproposedin[33]worksbytransmitt ingawindowof m packetsallthetime.Thetransmittersends m packetsineverytransmissionopportunityand waitsforthecorrespondingacknowledgmentssentbytherec eiverinonepacketthat includesalltheinformation.Oncetheacknowledgmentsare received,thetransmitter verieswhichpacketsarrivedwitherrors,andsendsanewwi ndowof m packetslled withthosereceivedinerrorfromthepastwindow,andnewpac ketsifspaceinthenew windowisavailable.Theprocesscontinuesuntilallthepac ketsaretransmitted.The stopandwaitversiondescribedin[36]isverysimilartothe onein[33],workingwitha windowofpackets.However,uponreceivingtheacknowledgm entpacket,thetransmitter sendsanewwindowcontainingonlythepacketsreceivediner ror.Oncethesepacketsare correctlyreceived,thesendersendsanewwindowwith m newpackets. Theevaluationoftheprotocolswasperformedusingthetwoe rrormodelsdescribedin thelastsection,andtheperformancemetricsutilizedwere throughputefciencyand packetdeliveryrate.Thislastmetricindicatesthelevelo freliabilityprovidedbyeach protocol.Theexperimentsuseachannelspeedof2400bpsand packetsizesof300,600, 1200,2400,and4800bitsfortheBernoullimodel,andpacket sizesof100,200,300, 400,500,and600bitsfortheMarkovchannelerrormodel.The protocolswereevaluated consideringapointtopointtransmissionbetweentwonodes 50mapartfromeachother andwithouttheexistenceoftheMACcontrolsublayer.Traf cwasgeneratedaccording toaPoissonprocess. 112

PAGE 130

600 1200 1800 2400 3000 3600 4200 4800 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency SW-MER sw2 sw3 sw4 BER = 1e-3 BER = 1e-4 Figure6.4:Throughputefciencyforawindowsizeof8datap acketsusingtheBernoulli errormodel. 600 1200 1800 2400 3000 3600 4200 4800 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency SW-MER sw2 sw3 sw4 BER = 1e-4 BER = 1e-3 Figure6.5:Throughputefciencyforawindowsizeof16data packetsusingaBernoulli errormodel. 113

PAGE 131

600 1200 1800 2400 3000 3600 4200 4800 0 10 20 30 40 50 60 70 80 90 100 Packet size Packet delivery rate SW-MER sw2 sw3 sw4 Figure6.6:Packetdeliveryrateforawindowsizeof8datapa cketsusingtheBernoulli errormodel. 600 1200 1800 2400 3000 3600 4200 4800 0 10 20 30 40 50 60 70 80 90 100 Packet size Packet delivery rate SW-MER sw2 sw3 sw4 Figure6.7:Packetdeliveryrateforawindowsizeof16datap acketsusingtheBernoulli errormodel. 114

PAGE 132

100 150 200 250 300 350 400 450 500 550 600 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Packet size Throughput efficiency SW-MER sw2 sw3 sw4 Figure6.8:Throughputefciencyforawindowsizeof16pack etsusingtheshallow watererrormodel. 100 150 200 250 300 350 400 450 500 550 600 0 10 20 30 40 50 60 70 80 90 100 Packet size Packet delivery rate SW-MER sw2 sw3 sw4 Figure6.9:Packetdeliveryrateforawindowsizeof16datap acketsusingtheshallow watererrormodel. 115

PAGE 133

Figures6.4and6.5showthethroughputefciencyofthelogi callinkcontrolprotocols usingawindowsizeof8and16packets,respectively.Asitca nbeseen,regardlessofthe BER,allprotocolspresentasimilarperformance.Thebene toftheSW-MERprotocol isinthepacketdeliveryratio,asshowninFigures6.6and6. 7,basedonachannelwith aBERof10 3 .Asexpected,theperformanceoftheprotocolsdecreaseswi ththepacket size.However,thesuperiorityinreliabilityoftheSW-MER protocolisdemonstrated. Thisisduetotheexponentialincreasepacketretransmissi onstrategyutilizedbySWMER.InthecaseofaBERof10 4 (resultsnotshownhere),allprotocolsexperienced similarperformance,indicatingthatSW-MERisbetteronly inthoseextremescenarios withverypoorqualitychannels.Inthecaseoftheunderwaterchannelerrormodel,similarex perimentswereperformedto comparetheresultswiththeBernoullierrormodel,usingth etrace,Markovmodel,and parametersdescribedinSection6.2.AsitcanbeseenfromFi gure6.8,thethroughput efciencyisverylowcomparedwiththeBernoullimodel,ind icatingthattheunderwater errormodelintroducesalargeramountoferrors.Thisismag niedconsideringthatthe experimentsutilizedsmallerpacketsizes.Thepacketdeli veryratiooftheprotocolsis showninFigure6.9.Fromtheselastplots,itcanbeconclude dthatinordertohavean acceptablethroughputandpacketdeliveryratioinanunder waterchannel,packetscannot belongerthan250bytes.Similarresultswerefoundusingsm allerwindowsizes. 6.5PerformanceEvaluationfortheEntireDataLinkTheperformanceofthetaggedowwasalsoassessedinthesam elinearnetworkusing thesameparametersinthelastsectionbutnowusingtheprop osedSW-MERand2MAC 116

PAGE 134

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 35 Offered Load in the network Throughput (bps) 1200 bits 802.11 BER 1200 bits Proposed Data Link BER 30 bps flow Figure6.10:ThroughputofthecombinedSW-MERand2MACprot ocolsusingpacket sizeof1200bits,andBER=1 x 10 3 protocolstogetherandincludingtheBernoulliandtheunde rwatererrormodels(shownin theplotasMM-MarkovianModel).Figure6.10showstheperfo rmanceofthetagged owwhensendingpacketsof1200bitswithaBER=1 x 10 3 .Itisshownthathavinga packetofthissize,theproposeddatalinkprotocolpresent sabetterperformanceinterms ofthroughputthanthetraditional802.11protocol.Figure 6.11presentsthethroughput performancebutusingtheunderwaterMarkovchannelerrorm odel.Althoughbothprotocolsareevaluatedwithamorerealisticchannelerrormod el,theproposeddatalinkprotocolstillhasabetterperformance.Thereasonofhavingab etterperformancewithboth channelerrormodelsisbecausetheproposeddatalinkproto colusestwochannelsfor simultaneoustransmissions,andwindowstopandwaitpacke ttransmissionsguaranteeing anincreaseinthethroughput.ThesamesituationispresentedinFigure6.12whenpacketso f2400bitsaresent.The proposeddatalinkprotocolpresentsagainabetterperform ancethanthetraditional802.11 protocol,usinganunderwatererrorchannelwithaBER=1 x 10 3 .Frombothgraphs, 117

PAGE 135

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 35 Offered Load in the network Throughput (bps) 1200 bits 802.11 MM 1200 bits Proposed Data Link MM 30 bps flow Figure6.11:ThroughputofthecombinedSW-MERand2MACprot ocolsusingpacket sizeof1200bits,andMarkovmodelchannel.ingeneralthreemainconclusionscanbedrawn.First,thepr oposedMACprotocol,as expected,improvestheperformanceovertheCSMA/CAprotoc ol.Second,ifthepacket sizeisincreased,howeverthethroughputpresentedby2MAC isbetter,theperformance ofthenetworkintheunderwaterchanneldecreasesbecauset heprobabilityofhavingan errorineachpacketsentincreases.Third,itisclearthatt heperformanceinallcasesis fairlypoorbutevenworseusingtheunderwaterchannelMark overrormodel,indicating thatmoreresearchisneededatalltheselayerstoimproveth eperformanceofacousticbasedunderwatercommunicationsystems. 118

PAGE 136

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 Offered Load in the network Throughput (bps) 2400 bits 802.11 BER 2400 bits Proposed Data Link BER 60 bps flow Figure6.12:ThroughputofthecombinedSW-MERand2MACprot ocolsusingpacket sizeof2400bits. 119

PAGE 137

Chapter7:AnAdaptiveLogicalLinkSublayerProtocolinRes ponsetoUnderwater AcousticCommunication(UAC)ChannelChangesInordertoimprovethethroughputandthereliabilityofthe underwatercommunication,theSW-MERprotocolwasproposedinchapter4,whichin cludesacombination ofstopandwaitandwindow-basedowcontrolstrategytoimp rovethecorresponding throughput,andanexponentialretransmissionstrategyto improvethepacketdelivery ratio.However,theretransmissionstrategyofSW-MERuses apredenednumberof packetsforretransmissionregardlessofthechannelcondi tion.Althoughthisstrategy improvesthepacketdeliveryratio,asdemonstratedinchap ter6,theremaybeoccasions wheretheprotocolmightbesendingunnecessarycopiesofth esamepacket. Thischapterpresentsanadaptivelogicallinkprotocoltha taddressesthisparticularproblem.Thenewprotocolbuildsuponthelogicallinkcontrolpr otocolpresentedinchapter 4,improvingtheperformanceoftheSW-MERprotocolbymakin gitadaptivetochannel conditionswithoutcompromisingthereliabilityofSW-MER .Inthenextsections,adescriptionoftheadaptiveprotocolisintroduced,includin ghowthequalityofthechannel isobtained,andsomecomparisonevaluationswiththeSW-ME Rprotocolaswell. 120

PAGE 138

7.1AdaptiveSW-MERDescriptionTheproposedprotocolincludesanadaptiveretransmission strategythatchangesthenumberofcopiestoberetransmittedaccordingtothequalityof thechannel.Acrosslayer approachbetweenthedatalinkandthephysicallayerisassu med,inwhichthephysical layerinformsthestateofthechanneltothedatalinklayere verytime,inordertoadapt theamountofcopiesofretransmittedpacketsineachwindow packettransmissionasa reactiontochannelchanges,toimprovethethroughputofth enetwork. TheprocessstartsattheMAClayerwhentheRTScontrolpacke tissent,andlaterthe CTSincludingavaluerepresentingthestateofthechanneli sreceived.Takingintoaccountthisstateandthepossiblepacketstobesentinthewin dowtransmission,thesender willllthewindowofpacketstobesentwiththosethatshoul dbesentandtheircorrespondingnumberofcopiestoberetransmitted.Eachnumbero fcopiescanexponentially increaseordecreasedependingonthechannelbehaviorhist orysavedinthesender.The adaptiveprotocolwillbeexplainedinmoredetailinthenex tsections. 7.2DeterminingtheChannelQualityDependingonthedistancebetweentransmitterandreceiver ,thechannelconditionsas seenbyeachonemightbedifferent.Further,ifthenodesmov e,theseconditionswill changewithtime.Figure7.1showsthescenariowheretwound erwaterunmannedvehiclesexploretheoceanseaatdifferentdepths.WhenthetransmittingUAVsendsinformationtotheadjacent AUV,itdoesnotknow aboutthechannelconditionspresentatthereceiver.Ifthe channelpresentserrors,those 121

PAGE 139

Figure7.1:Transmissionbetweennodeslocatedatdifferen tdepths. willbereectedinthedatapacketswhentheyarriveatthere ceiver.Thereceiverthen createsavectorofacknowledgmentsindicatingwhichpacke tswerereceivedwithand withouterrors,andsendsittothetransmitter.Savinginformationaboutthelasttransmissionisnotenoug htopredictwhatthechannel conditionwillbethenexttime.Increasingthenumberofcop iesexponentiallyallthetime mightnotbenecessary,andabetterthroughputcouldbeachi evedifthatnumberchanged accordingtothechannelquality.Basedonthisinformation ,thefour-statemachineshown inFigure7.2wasdesignedtodeterminethenumberofcopiest obesentperincorrect packet,whichconsidersthestatusofthechannelduringthe lasttwotransmissions(history)plusthecurrentpacket.Theentireprocessworksasfo llows: Figure7.2:Four-statemachinerepresentation. 122

PAGE 140

Ifthereceiverreceivedthelasttwopacketscorrectlyandt hecurrentoneisalso correct,thetransmitterdeterminesthatthechannelisint heGood-Goodstate.In thiscase,thesenderwilldecreasethenumberofcopiesperr etransmittedpacketby half,i.e., c = c 2. Ifthereceiverreceivedthelasttwopacketscorrectlyandt hecurrentoneisbad, thetransmitterdeterminesthatthechannelisintheGood-B adstate.Inthiscase, thesenderwillincreasethenumberofcopiesperretransmit tedpacketbytwo,i.e., c = c 2. Ifthelasttwopacketswerereceivedgoodandbad,andthecur rentoneisgood, thetransmitterdeterminesthatthechannelisintheBad-Go odstate.Inthiscase, thesenderwilldecreasethenumberofcopiesperretransmit tedpacketbyhalf,i.e., c = c 2. Ifthelasttwopacketswerereceivedbadandgood,andthecur rentoneisbad,the transmitterdeterminesthatthechannelisintheGood-Bads tate.Inthiscase,the senderwillincreasethenumberofcopiesperretransmitted packetbytwo,i.e., c = c 2. Ifthelasttwopacketswerereceivedbadandgood,andthecur rentoneisgood, thetransmitterdeterminesthatthechannelisintheGood-G oodstate.Inthiscase, thesenderwilldecreasethenumberofcopiesperretransmit tedpacketbyhalf,i.e., c = c 2. Ifthelasttwopacketswerereceivedgoodandbad,andthecur rentoneisbad,the transmitterdeterminesthatthechannelisintheBad-Badst ate.Inthiscase,the senderwillincreasethenumberofcopiesperretransmitted packetbytwo,i.e., c = c 2. 123

PAGE 141

n r n n r n n nnn nn n n Figure7.3:Physicalanddatalinklayer,interaction. Ifthereceiverreceivedthelasttwopacketsincorrectlyan dthecurrentoneisalso incorrect,thetransmitterdeterminesthatthechannelisi ntheBad-Badstate.Inthis case,thesenderwillincreasethenumberofcopiesperretra nsmittedpacketbytwo, i.e., c = c 2. Ifthereceiverreceivedthelasttwopacketsincorrectlyan dthecurrentoneisgood, thetransmitterdeterminesthatthechannelisintheBad-Go odstate.Inthiscase, thesenderwilldecreasethenumberofcopiesperretransmit tedpacketbyhalf,i.e., c = c 2. Figure7.3illustratesthecommunicationprocessbetweens enderandreceiverthatconveysthechannelqualityinformation. 124

PAGE 142

n r rn rn n r r nr !!nr "r# !r!n $ # !r nr n r !!nr!rr rr r rr! rr! !r % r rn rn rr r!n rn r rn r &nrr r! %r Figure7.4:Finitestatemachineofthesenderprocess. 7.3StatesoftheAdaptiveLogicalLinkProtocolTheadaptiveSW-MERhasthefollowingstates:Listen,Verif yingtheStateoftheChannel,UpdatingChannelStatusHistory,UpdatingCopiesofPa cketstobeSent,Waitingto ReceivePackets,CheckingPackets,EnqueueingPacketswit houtErrors,VerifyingPacketstobeSenttotheUpperLayer,GeneratingtheACKVector,a ndIdentifyingPackets SentwithErrors.Figures7.4and7.5showthesenderandrece iverstatemachinesfor thenewprotocol,respectively.Thestatesandstatetransi tionsneededinthesenderand receiverprocessesareexplainednext. 125

PAGE 143

nr nr r nn nr r nr nr! r # rr $r r nr r nr !! %! $r!! !! & nr '' r ! n "nr !n! rr !n nr r $nr r (nr ) !%rr !nr Figure7.5:Finitestatemachineofthereceiverprocess. 7.3.1ListenWhenanodedecidestosendpacketstoanothernode,itrstse lectsthepacketstobe transmittedfromthequeue.ThenanRTSissentwiththetotal timeneededtotransmit awindowof m packetsandreceivetheacknowledgmentvector. 7.3.2VerifyingtheStateoftheChannelOncetheRTSarrives,thereceiversavesthetimethatthetra nsmissionof m packetswill takeandpreparestheCTSpacket.Atthistimethephysicalla yeralreadyhasmeasured 126

PAGE 144

andidentiedthequalityofthechannel,andpassedthisinf ormationontothedatalink layer.ThisinformationisaddedintheCTSpacket,whichis nallysenttothetransmitter. 7.3.3UpdatingChannelStatusHistoryOncetheCTSpacketisreceived,thetransmitterextractsth echannelqualityinformation andusesthefour-statemachinepresentedinFigure7.2tode terminethenumberoftimes packetsinerrorneedtoberetransmittedinthenexttransmi ssionopportunity. 7.3.4UpdatingCopiesofPacketstobeSentBasedonthechannelqualityinformationandthefour-state machine,thesenderthen assemblesandtransmitsthenewwindowofpackets.7.3.5WaitingtoReceivePacketsInthisstate,thereceiverwaitsuntilthewindowof m packetsarrivesoratimeoutoccursbecauseatleastoneofthepacketsneverarrived.There ceiveridentiesthemissing packetsusingtwoinformationeldsintheheaderofthepack ets,onethatrepresentsthe positionofthepacketinthewindowandanotheronethatrepr esentstheordinalcopy number. 127

PAGE 145

7.3.6CheckingPacketsThereceiververiesifthepacketsarrivedwithorwithoute rrorsbyapplyingawellknownCyclicRedundancyCheck(CRC)function,andclassie sthemaccordingly. 7.3.7EnqueueingPacketsWithoutErrorsPacketsthatarrivedwithouterrors,aretemporallyenqueu edinthebufferofthelogical linklayer.Theystaythereuntilaconsecutivenumberof m packetsarrivewithouterrors. Thosepacketsthatwerereceivedincorrectlyareeliminate d.Intheproposedprotocol,it isveryeasytocalculatethebuffersize B atthereceivertoavoidpacketdrops,whichis givenbyEquation7.1inwhich w isthesizeofthewindowinnumberofpacketsand t is themaximumnumberoftrialsthatadatapacketcanberetrans mitted. t isobtainedfrom Equation7.2. B = w t (7.1) t = log 2 ( w )+ 1(7.2) 128

PAGE 146

7.3.8VerifyingPacketstobeSenttotheUpperLayerOnceaconsecutivenumberof m packetsarereceivedwithouterrors,theyaresenttothe upperlayeranddeletedfromthebuffer.Thepurposeofsendi ngthepacketstotheupper layeristoverifywhetherthosepacketsbelongtothatnodeo rmustberetransmittedtothe nextneighbor.7.3.9GeneratingtheACKVectorInthisstate,thereceiverisinchargeofcreatinganACKcon trolpackettotellthetransmitterthestateofeachpacketreceived.ThisACKpacketisa vectorinwhicheachpositionisaonebitvaluethatrepresentsthestatusofthepac ket,where0meansthatthe packetarrivedcorrectly,and1meansthatthepacketsarriv edwitherrorsordidnotarrive atall.Attheend,theACKvectorissenttothetransmitter.7.3.10IdentifyingPacketsSentWithErrorsOncetheACKarrivestothesender,thevectorisrevisedtoid entifywhichpacketsarrived withandwithouterrors.Packetsthatarrivedwithouterror saredeletedfromthesender's queue,theothersarekeptuntilamaximumnumberoftrialsis achievedortheyarereceivedcorrectly.Themaximumnumberoftrialsisalreadyde nedinEquation7.2. 129

PAGE 147

n rrrrr nn n rrrrr nn nr nn rrrrr rrrrr (a)Firsttransmission,M=6andpackets2,and3arrivewitherrors. n n n n n r nnnnn r n rnnnnn nnnnn rr rr rr (b)Secondtransmission,M=6andpacket3arriveswitherrorsagain. nnnnn r nnnnn r n r nnnnnnnnnn rrrr rr (c)Thirdtransmission,M=6andpacket9arriveswitherrors. n nnn rr rrr nnnn rr rrr nr n n nnnn nnnn n n (d)Fourthtransmission,M=6andpacket9and12arrivewitherrors. n r n r nr n n (e)Fifthtransmission,M=6andpacket9arriveswitherrors. n r n r n r n n (f)Sixthtransmission,M=6andpacket9arriveswithouterrors. Figure7.6:Exampleoftheproposedadaptivestopandwaitsl idingwindow-based mechanism. 130

PAGE 148

7.4SW-MERProtocolExampleFigure7.6showsanexampleofhowtheproposedprotocolwork susingawindowsizeof 6packets.TodeneatransmissionprocessasanRTS/CTS/DAT A/ACKtransmission,it isassumedthatthechannelwasina Good Good statethelasttime,andatransmission processstarts.Figure7.6(a)displaysthersttransmissi oninwhichthechannelstatus receivedintheCTSpacketfromthereceiverindicates Good .Then,thestatemachine staysin Good Good state,onecopyofpackets1 ; 2 ; 3 ; 4 ; 5 ; 6aresentand2 ; 3arrivewith errors.Thereceiversendsanacknowledgmentrelayingthis tothesender. Inthenexttransmissionprocess,Figure7.6(b),thetransm ittersendstwocopiesofthe packets2 ; 3becauseitreceiveda Bad channelstatusintheCTSpacketfromthereceiver. Also,thestatusofthestatemachinechangestoa Good Bad state.Asdisplayedin Figure7.6(c),thechannelstatusstillis Bad atthereceiver.Asaconsequence,thestate machinestatusgoestoa Bad Bad stateandtheamountofcopiesofeachpacketto besentwillexponentiallyincreaseby2.Fourcopiesofpack et3aresentduetohaving arrivedwitherrorthelasttime.Ontheotherhand,onlytwoc opiesofpacket8aresent becauseitisanewpacket.Figure7.6(d)showsthefourthtransmission.Thetransmitt erisinformedthatthechannelstatusis Good ,thenthestatemachinestatestatusisupdatedto Bad Good andthe amountofcopiesofeachpackettobesentisexponentiallyde creasedby2.Sincethe transmitterisgoingtosendnewpackets,onlyonecopyfrome achoneissent.InFigure7.6(e),onlyonecopyofpackets9 ; 11aresent,althoughtheyarrivedwitherrorsin theprevioustransmission.Thisisbecausethelasttimethe statusofthechannelstatewas Good Good andstillis Good 131

PAGE 149

7.5ChannelErrorModelsToevaluatetheperformanceofbothprotocols,thesametwoc hannelerrormodelsusedin Section6.2areconsidered.7.6PerformanceEvaluationTheperformanceevaluationoftheproposedadaptiveprotoc olispresentedinthissection. First,ageneraldescriptionoftheparametersusedinthesi mulationsispresented.Then, theSW-MERandtheadaptiveSW-MERprotocolarecomparedusi ngthenumberof packetretransmissionsandthroughputasthemainperforma ncemetrics. 7.6.1SimulationParametersThesimulationiscarriedoutusingtheparametersincluded inTable7.1.Nodeswith linearformationsaredesignedtoevaluatetheprotocol.Tw oerrormodelsareconsidered, theBernoullimodel,andasynthetictraceobtainedfrom[45 ].Thevaluesofdistance betweennodes,datarate,andotherrelatedparametersarep resentedinthetable.Forthe Bernoullimodelabiterrorrateof1 x 10 3 isused,controlpacketlength'sof30bitsfor RTS,CTSandACKareappliedinallthescenarios.Itisassumedalineartopologywith5AUVs,asshowninFigure 7.7.Thetrafcinthe networkwasgeneratedasfollows.Eachnodegeneratedowst oallothernodessending packetsaccordingtoaPoissonprocess,withlof4datapacketspersecond(expected arrivalinatimeinterval).Therateoftheowsisdetermine dsothattheloadinthenet132

PAGE 150

Table7.1:Simulationparameters. ErrorModel Parameters Values Bernoulli Speedofsound 1500m/sec Datarate 2400and19600b/s Range 50m Packetsperwindow 8 Packetsize 150and300bytes [45] Speedofsound 1550m/sec Datarate 2400b/s Distance 50m Packetsperwindow 8 Packetsize 150and300bytes workissettoaspecicdesiredlevelstartingfrom0to100%. Inaddition,ataggedow sendingpacketsfromnodeEtonodeDisestablishedandmonit ored.Thesimulations resultsoftheperformanceinallgraphsarerelatedtothepe rformanceofthattaggedow. AsimplenetworklayerisincludedontopoftheLLClayertoro uteincomingpacketsto thefollowingadjacentnode. Figure7.7:Lineartopology. 7.6.2ThroughputEvaluationInthissectiontheperformanceofbothprotocolsiscompare dusingthroughputandnumberofcopiesperpacketsentasthemainperformancemetrics .Thissecondmetricis 133

PAGE 151

denedastheaveragenumberofcopiesthatneedtobetransmi ttedperpacketsoitis nallyreceivedcorrectly.Figures7.8,7.9and7.10showtheamountofcopiesperpacket generatedbythetwo protocols.ThersttwoscenarioswereexecutedwiththeBer noullimodelusingaBERof 1 x 10 3 thatgeneratesachannelhighinerrors.Asitcanbeseen,hol dingthesamepacket sizeandincreasingtheloadinthenetworkcausemanypacket stoarrivewitherrors.Then thethroughputofeachprotocoldecreases,asseeninFigure s7.11and7.12.TheSWMERprotocol[9]reactsexponentially,increasingthenumb erofcopiesperpacketeach timeatransmissionprocessstarts,toguaranteeasuccessf ulpacketdelivery.Theprotocol doesnottakeintoaccountthatthechannelisnotinfailurea llthetimeliketheproposed adaptiveSW-MER.Asexpectedhowever,thechannelpresents alotoferrors,andthe proposedprotocoladaptstochannelchangesgeneratingfew eramountofcopiestobe sent,andimprovingthecorrespondingthroughput,asseeni nFigures7.11and7.12. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 Offered Load in the network Number of copies per packet No adaptive Data Link Adaptive Data Link Figure7.8:CopiesperpacketusingaBernoullierrormodel, awindowsizeof8packets, apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof2400bps. Inthecaseoftheunderwaterchannelerrormodel,similarex perimentswereperformed tocomparetheresultsusingthetrace,Markovmodel,andpar ametersdescribedinSec134

PAGE 152

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 Offered Load in the network Number of copies per packet No adaptive Data Link Adaptive Data Link Figure7.9:CopiesperpacketusingaBernoullierrormodel, awindowsizeof8packets, apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof19600bps. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 Offered Load in the network Number of copies per packet No adaptive Data Link Adaptive Data Link Figure7.10:Copiesperpacketusingashallowwatererrormo del,awindowsizeof8 packets,apacketsizeof1200bits,anddatarateof2400bps. 135

PAGE 153

tion6.2.Asitcanbeseenfromthethirdscenariodisplayedi nFigures7.10and7.15,the numberofcopiesgeneratedperpacketwiththeprotocolin[9 ]startstobebiggerthanthe amounttheproposedadaptiveSW-MERproducesassoonasthel oadofthenetworkis increased,affectingthethroughput.AsitisshowninFigur e7.15,thethroughputstarts todeterioratebutstillthesuperiorityoftheproposedada ptiveSW-MERisdemonstrated. Thisisasaconsequenceofthereductionofthenumberofcopi esretransmittedineach window,incrementingthenumberofnewpacketstobesent.FromtheresultsplottedinFigures7.8and7.9forexample,i tcanbeseenthattheadaptiveprotocolreducesthenumberofcopiesperpacketthatar egeneratedcomparedwith theamountgeneratedbySW-MER,especiallywhenthenetwork ishighlyloaded.In addition,athroughputimprovementcanbeobtainedevenusi ngthesynthetictracethat introducesalargeramountoferrorsintheorderof15%,assh owninFigures7.11,7.12 and7.16.Figures7.13and7.14showthatforawindowof16pac kets,howeverthe throughputofbothprotocolsaresimilar,theamountofpack etsgeneratedfortheprevious protocoldramaticallyincreasescomparedwiththeamountp roducedbytheproposed adaptiveSW-MER. 136

PAGE 154

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 Offered Load in the network Throughput No adaptive Data Link Adaptive Data Link Figure7.11:ThroughputusingaBernoullierrormodel,awin dowsizeof8datapackets,a packetsizeof2400bits,BER=1 x 10 3 ,anddatarateof2400bps. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 Offered Load in the network Throughput No adaptive Data Link Adaptive Data Link Figure7.12:ThroughputusingaBernoullierrormodel,awin dowsizeof8datapackets,a packetsizeof2400bits,BER=1 x 10 3 ,anddatarateof19600bps. 137

PAGE 155

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 5 10 15 20 25 30 35 40 45 50 Offered Load in the network Number of copies per packet No adaptive Data Link Adaptive Data Link Figure7.13:Copiesperpacketusingashallowwatererrormo del,awindowsizeof16 packets,apacketsizeof2400bits,anddatarateof19600bps 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 Offered Load in the network Throughput No adaptive Data Link Adaptive Data Link Figure7.14:ThroughputusingaBernoullierrormodel,awin dowsizeof16datapackets, apacketsizeof2400bits,BER=1 x 10 3 ,anddatarateof19600bps. 138

PAGE 156

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 7 8 9 10 Offered Load in the network Throughput No adaptive Data Link Adaptive Data Link Figure7.15:Throughputusingashallowwatererrormodel,a windowsizeof8packets, andapacketsizeof1200bits. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 2 4 6 8 10 12 14 16 18 20 Offered Load in the network Throughput 1x10 -2 No adaptive Data Link Adaptive Data Link Figure7.16:Throughputusingashallowwatererrormodel,a windowsizeof8packets, andapacketsizeof2400bits. 139

PAGE 157

Chapter8:ConclusionsandFutureWorkThisdissertationproposesanewdatalinkprotocoldesigne dforacoustic-basedunderwatercommunicationsystems,usingOFDMAtechnologyattheph ysicallayertosupport applicationsofswarmsofautonomousunderwatervehicles. Thischapterpresentsthe conclusionsdrawnfromthecontributionsshowninSection1 .5andoutlinesfuturedirectionsofthiswork.8.1Conclusions AnewMAClayerprotocol,called2MAC,designedtocoordinat eandtakeadvantageofthemultiplesub-channelsmadeavailablebytheu seofOFDMA,isincluded.2MACusesthreechannelstotransmitorreceivedata simultaneouslyfrom differentneighborswithoutcollisionsandusingonlyonet ransceiver.Theproposed MACprotocolreducescollisions,andeliminateshiddenter minal,exposedterminal andcaptureproblems,improvingasaconsequencethethroug hputofthenetwork. The2MACprotocolshowsitssuperiorityoverthewell-known 802.11protocol. Alogicallinkcontrolprotocol,calledSW-MER,isintroduc ed.SW-MERcombines stopandwaitandslidingwindowmechanismstoincreasethec hannelutilization andimplementsanexponentialretransmissionstrategytoi ncreasethepacketdeliv140

PAGE 158

eryratio,especiallyunderpoorchannelconditionsliketh eonescommonlyfound underwater.Byusingthewindowpackettransmission,thepr oposedLLCprotocol alsoimprovesthethroughputefciencyofthenetwork. ThecombinedSW-MERand2MACprotocolsarealsoevaluatedas acompletedata linkprotocol.Theproposeddatalinkprotocoloffersabett erperformanceinterms ofthroughputthanthe802.11protocolinbothchannelerror modelsusedforthe performanceevaluation. Anewbackoffalgorithmhasbeenproposedforthe2MACprotoc ol.Thethroughputofferedbytheproposedbackoffalgorithmoffersasimil arthroughputthan thetraditionalbackoffalgorithms,howeverthesizeofthe contentionwindowin thenewbackoffissmaller.Thisisanadvantagebecausethen odesdonothaveto spendalotoftimecontendingthechannelbeforestartingto transmit,especiallyin underwatercommunicationswherelongpropagationdelaysa represent. Ananalyticalmodelthatrepresentsthesaturationthrough putfor2MACisshown. AnanalyticalmodelthatrepresentsthethroughputfortheS W-MERprotocolis alsointroduced. AnimprovementofthelogicallinkprotocolnamedadaptiveS W-MERthatworks asacrosslayerwiththephysicallayerisshown.Itisalogic allinkprotocolthat adaptstounderwateracousticcommunicationchannelchang esforunderwater vehicles.Thesuperiorityintermsofthroughputoftheadap tiveprotocolwhen comparedwithSW-MERisdemonstrated. 141

PAGE 159

8.2FutureWorkTheresearchdevelopedinthisdissertationcanbeextended forfutureworkasfollows: TheunderwaterMarkovchannelerrormodeldesignedreprese ntsanunderwaterchannelforshallowwater.Anunderwaterchannelerrorm odelfordeepwater shouldbedesignedtocoverotherunderwaterscenarios. Theproposeddatalinklayerprotocolwasevaluatedoverlin eartopologies.Evaluationoftheproposeddatalinkprotocoloveradditionalne tworkscenariossuch asthoseincludingpolygonalformations,andusingotherun derwaterchannelerror modelsmustbedone. Theanalyticalmodelproposedforthesaturationthroughpu tonlyincludestheMAC sublayer.Amorepreciseanalyticalmodelrequirestheincl usionoftheLLCsublayer,torepresentthesaturationthroughputforthepropo seddatalinkprotocol. Theimplementationandevaluationoftheproposedprotocol shavebeenappliedin simulations.Implementingandevaluatingtheseprotocols overarealunderwater networkshouldgivemorepreciseresults. 142

PAGE 160

ListofReferences [1]I.F.Akyildiz,D.Pompili,andT.Melodia,“State-of-th e-ArtinProtocolResearch forUnderwaterAcousticSensorNetworks,”inProceedingsofthe1stACM internationalworkshoponunderwaternetworks,2006,pp.7–16. [2]M.Chitre,S.Shahabodeen,andM.Stojanovic,“Underwat eracousticcommunicationsandnetworking:Recentadvancesandfuturechallenge s,”MarineTechnology SocietyJournal,vol.42,No.1,pp.103–116,2008. [3]K.B.K.IIandP.Mohapatra,“AHybridMediumAccessContr olProtocolfor UnderwaterWirelessNetworks,”inProceedingsofthesecondworkshopon underwaternetworks,September2007,pp.33–40. [4]F.Schill,U.Zimmer,andJ.Trumpf,“VisibleSpectrumOp ticalCommunicationand DistanceSensingforUnderwaterApplications,”inProceedingsoftheAustralian conferenceroboticsandAutomation(Canberra,2004),2004. [5]B-C.KimandI-T.Lu,“ParameterstudyofOFDMunderwater communications system,”inProceedingsoftheIEEEOceansconferenceandExhibition,September 2000,pp.1251–1255. [6]S.Haykin,CommunicationSystems,4thed.JohnWileyandSons,2001. [7]A.ChandraandV.GummallaandJ.O.Limb,“WirelessMediu mAccessControl Protocols,”IEEECommunicationsSurveysandTutorials,2000. [8]J.W.Conard,“ServicesandProtocolsoftheDataLinkLay er,”Proceedingsofthe IEEE,vol.71,No.12,pp.1378–1383,1983. 143

PAGE 161

[9]D.JabbaandM.Labrador,“ADataLinkLayerinSupportofS warmingof AutonomousUnderwaterVehicles,”inProceedingsoftheMTS/IEEEOceans conference,Bremen,Germany,May2009. [10]——,“AnAdaptiveLogicalLinkProtocolinResponsetoUn derwaterAcoustic CommunicationChannelChangesforAutonomousUnderwaterV ehicles,”inProceedingsoftheMTS/IEEEOceansconference,Biloxi,Mis sissippi,October 2009. [11]M.Stojanovic,“Recentadvancesinhigh-speedunderwa teracousticcommunications,”IEEEJournalofOceanicEngineering,vol.21,No.2,pp.125–136,1996. [12]——,“LowComplexityOFDMDetectorforUnderwaterAcous ticChannels,”inProceedingsoftheIEEEOceansconference,September2006. [13]L.Freitag,M.Stojanovic,D.Kilfoyle,andJ.Preisig, “HighRatePhase-Coherent AcousticCommunication:AReviewofaDecadeofResearchand aPerspective onFutureChallenges,”in7thEuropeanConferenceonUnderwaterAcoustics,July 2004. [14]J.Huang,J.Sun,C.He,X.Shen,andQ.Zhang,“High-spee dunderwateracoustic communicationbasedonOFDM,”inMicrowave,Antenna,PropagationandEMC TechnologiesforWirelessCommunications,2005.MAPE2005 .IEEEInternational Symposium,August2005,pp.1135–1138. [15]R.F.Ormondroyd,“Arobustunderwateracousticcommun icationsystemusing OFDM-MIMO,”inProceedingsoftheMTS/IEEEOceansconference,June2007. [16]S.Roy,T.M.Duman,V.McDonald,andJ.G.Proakis,“High ratecommunication forunderwateracousticchannelsusingmultipletransmitt ersandspace-timecoding: Receiverstructuresandexperimentalresults,”IEEEJournalofOceanicEngineering, February2007. [17]G.G.XieandJ.Gibson,“Anetworkingprotocolforunder wateracousticnetworks,”NavalPostgraduateSchool,Tech.Rep.TR-CS-00-02,December2000. 144

PAGE 162

[18]J.-K.Yeo,Y.-K.Lim,andH.-H.Lee,“Modiedmac(media accesscontrol)protocol designfortheacoustic-basedunderwaterdigitaldatacomm unication,”IEEE,pp. 364–368,2001. [19]E.M.Sozer,M.Stojanovic,andJ.G.Proakis,“Underwat eracousticnetworks,”IEEEjournalofOceanicEngineering,vol.25,No.1,pp.72–83,2000. [20]H.DoukkaliandL.Nuaymi,“AnalysisofMACprotocolsfo runderwateracoustic datanetworks,”VehicularTechnologyConference,2005.VTC2005-Spring20 05 IEEE61st,vol.2,pp.1307–1311,2005. [21]M.MolinsandM.Stojanovic,“SlottedFAMA:AMACProtoc olforUnderwater AcousticNetworks,”inProceedingsoftheIEEEOceansconference,2006. [22]M.K.ParkandV.Rodoplu,“UWAN-MAC:AnEnergy-Efcien tMACProtocol forUnderwaterAcousticWirelessSensorNetworks,”IEEEJournalofOceanic Engineering,vol.32,No.3,pp.710–720,July2007. [23]A.A.SyedandJ.Heidemann,“Timesynchronizationforh ighlatencyacoustic networks,”inProceedingsoftheIEEEINFOCOMconference,April2006. [24]A.A.Syed,W.Ye,andJ.Heidemann,“T-Lohi:ANewClasso f MACProtocolsforUnderwaterAcousticSensorNetworks,”inTechnical ReportISI-TR-638,USC/InformationSciencesInstitute,2007.[Online].Available: http://www.isi.edu/~johnh/PAPERS/Syed07a.html [25]J.Wills,W.Ye,andJ.Heidemann,“Low-powerAcousticM odemforDense UnderwaterSensorNetworks,”inProceedingsoftherstACMinternational workshoponunderwaternetworks(WUWNet),2006. [26]M.Tao,S.Haoshan,andW.Yu,“AMACProtocolforUnderwa terSensorNetworks,” pp.144–148,2007. [27]R.K.Creber,J.A.Rice,P.A.Baxley,andC.Fletcher,“P erformanceofundersea acousticnetworkingusingRTS/CTShandshakingandARQretr ansmission,”inProceedingsoftheIEEEOceansconference,2001,pp.2083–2086. 145

PAGE 163

[28]F.Salva-GarauandM.Stojanovic,“Multi-clusterprot ocolforadhocunderwater acousticnetworks,”inProceedingsoftheIEEEOceansConference,September 2003,pp.91–98. [29]D.Pompili,T.Melodia,andI.F.Akyildiz,“ADistribut edCDMAMediumAccess ControlforUnderwaterAcousticSensorNetworks,”inSubmittedforPublication, August2006. [30]H.Doukkali,L.Nuaymi,andS.Houcke,“DistributedMAC protocolsforUnderwater AcousticDataNetworks,”VehicularTechnologyConference,2006.VTC-2006Fall 2006IEEE64th,pp.1–5,2006. [31]P.XieandJ.-H.Cui,“R-MAC:AnEnergy-EfcientMACPro tocolforUnderwater SensorNetworks,”inProceedingsoftheinternationalconferenceonwireless algorithms,systems,andapplications(WASA'07),2007. [32]V.RodopluandM.Park,“Anenergy-efcientMACprotoco lforunderwaterwireless acousticnetworks,”inProceedingsoftheMTS/IEEE,Oceansconference,2005, 2005,pp.1198–1203. [33]J.M.Morris,“OptimalBlocklengthsforARQErrorContr olSchemes,”IEEE TransactionsonCommunications,vol.COM-27,No.2,pp.488–493,1979. [34]A.S.Tanenbaum,ComputerNetworks,3rded.PrenticeHall,1996. [35]M.Stojanovic,“OptimizationofaDataLinkProtocolfo ranUnderwaterAcoustic Channel,”inProceedingsoftheIEEEOceansconference,2005,pp.68–73. [36]P.F.Turney,“AnimprovedStop-and-WaitARQLogicforD ataTransmissionin MobileRadioSystems,”IEEETransactionsonCommunications,vol.COM-29,No. 1,pp.68–71,1981. [37]H.-P.Tan,W.K.Seah,,andL.Doyle,“AMulti-hopARQPro tocolforUnderwater AcousticNetworks,”inProceedingsoftheIEEEOceansconference,2007. 146

PAGE 164

[38]P.XieandJ.-H.Cui,“AnFEC-basedReliableDataTransp ortProtocolforUnderwater SensorNetworks,”inProceedingsofthe16thinternationalconference,2007,pp. 747–753. [39]M.Luby,M.Mitzenmacher,A.Shokrollahi,D.Spielman, andV.Stemann,“Practical Loss-ResilientCodes,”inProceedingsoftheACMconference,1997,pp.150–159. [40]H.-P.Tan,W.Seah,andL.Doyle,“AMulti-hopARQProtoc olforanUnderwater AcousticChannel,”inProceedingsoftheIEEEOceansConference,2007. [41]Z.Guo,B.Wang,andJ.-H.Cui,“Efcienterrorrecovery usingnetworkcoding inunderwatersensornetworks,”inProceedingsoftheIFIPnetworkingconference, 2007,pp.227–238. [42]Z.Guo,“Onapplyingnetworkcodingtounderwatersenso rnetworks,”inProceedings oftheACMWUWNETconference,2006. [43]V.Bharghavan,A.Demers,S.Shenker,,andL.Zhang,“MA CAW:AMediaAccess ProtocolforWirelessLAN's,”inProceedingsoftheconferenceoncommunications architectures,protocolsandapplications,1994,pp.212–225. [44]P.Karn,“MACA-anewchannelaccessmethodforpacketra dio,”inARRL/CRRL AmateurRadio9thComputerNetworkingConference,September1990,pp.134– 140. [45]M.Chitre,J.Potter,andO.S.Heng,“Underwateracoust icchannelcharacterization formedium-rangeshallowwatercommunications,”inProceedingsoftheIEEE Oceansconference,November2004,pp.40–45. [46]G.Bianchi,“PerformanceAnalysisoftheIEEE802.11Di stributedCoordination Function,”IEEEJournalonSelectedAreasinTelecommunications,Wire lessseries, vol.18,No.3,pp.535–547,2000. [47]J.Yin,“TheAnalysisofPerformanceofIEEE802.11MACP rotocolUsingMarkov Chain,”IJCSNSInternationalJournalofComputerScienceandNetwo rkSecurity, vol.7,No.12,pp.27–37,2007. 147

PAGE 165

[48]Q.Ni,T.Li,T.Turletti,andY.Xiao,“Saturationthrou ghputanalysisoferror-prone 802.11wirelessnetworks,”WileyJournalofWirelessCommunicationsandMobile Computing(JWCMC),vol.5,No.8,pp.945–956,December2005. [49]M.Schwartz,TelecommunicationNetworks:Protocols,ModelingandAnal ysis. AddisonWesley,1987. [50]H.S.WangandN.Moayeri,“Finite-StateMarkovChannel s-AUsefulModelfor RadioCommunicationsChannels,”IEEETransactionsonCommunications,vol.43, pp.163–171,February1995. [51]W.TurinandR.V.Nobelen,“HiddenMarkovModelingofFl atFadingChannels,”IEEEJournalonSelectedAreasinCommunications,vol.16,No.9,pp.1809–1817, December1998. [52]L.R.RabinerandB.H.Juang,“AnIntroductiontoHidden MarkovModels,”IEEE ASSPMagazine,pp.4–16,1986. [53]L.E.Baum,T.Petrie,G.Soules,andN.Weiss,“AMaximiz ationTechnique OccurringintheStatisticalAnalysisofProbabilisticFun ctionsofMarkovChains,”TheAnnalsofMathematicalStatistics,vol.41,pp.164–171,February1970. 148

PAGE 166

Appendices 149

PAGE 167

AppendixA:TheStationaryDistributionoftheMarkovChainThestationarydistributionoftheMarkovchainforabackof fstage0andabackoffcounter 0for2MACiscalculatedasfollows: 1 =p0 ; 0 2 m 1 i = 0 p i 5 4 i W + p m ( 1 p ) 5 4 m W + 1 ( 1 p ) # = =p0 ; 0 2 W m 1 i = 0 5 4 p i + 5 4 p m ( 1 p ) + 1 ( 1 p ) # = =p0 ; 0 2 W 1 5 4 p m 1 5 4 p + 5 4 p m ( 1 p ) + 1 ( 1 p ) # = =p0 ; 0 2 W 1 p 5 4 p m + p 5 4 p m + 5 4 p m 5 4 p 5 4 p m 1 5 4 p ( 1 p ) + 1 ( 1 p ) # = =p0 ; 0 2 W 1 p 1 4 p 5 4 p m 1 5 4 p ( 1 p ) + 1 ( 1 p ) # = =p0 ; 0 2 W W p W 1 4 p 5 4 p m 1 5 4 p ( 1 p ) + 1 ( 1 p ) # = =p0 ; 0 2 W W p W 1 4 p 5 4 p m + 1 5 4 p 1 5 4 p ( 1 p ) # = =p0 ; 0 2 W 5 4 1 4 W p W 1 4 p 5 4 p m + 1 5 4 p 1 5 4 p ( 1 p ) # = =p0 ; 0 2 W 5 4 p W + 1 4 p W W 1 4 p 5 4 p m + 1 5 4 p 1 5 4 p ( 1 p ) # = =p0 ; 0 2 24 W 1 5 4 p + 1 4 p W 1 5 4 p m + 1 5 4 p 1 5 4 p ( 1 p ) 35 = =p0 ; 0 2 24 ( W + 1 ) 1 5 4 p + 1 4 p W 1 5 4 p m 1 5 4 p ( 1 p ) 35 (A.1) 150

PAGE 168

AppendixA:(continued)Then,thestationarydistributionis:p0 ; 0 = 8 1 5 4 p ( 1 p ) 4 ( W + 1 ) 1 5 4 p + Wp 1 5 4 p m (A.2) 151

PAGE 169

AppendixB:TheThroughputEfciencyofSW-MERThethroughputefciencyofSW-MERiscalculatedasfollows :h= T d T f (B.1) )h= w D d r h k = 0 4 k 1 P ( 2 k 1 1 ) ( 1 P ) i [ T txd + T txa + 2 ( T sw + T prop )] = = w D d r ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) i [ T txd + T txa + 2 ( T sw + T prop )] = = D d ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) i r w [ T txd + T txa + 2 ( T sw + T prop )] = = D d ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) i r w h w D f r + D ack r + 2 ( T sw + T prop ) i = = D d ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) ih D f + D ack w + 2 r w ( T sw + T prop ) i = = D d ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) ih D d + D oh + D ack w + 2 r w ( T sw + T prop ) i = = D d ( 1 P ) h k = 0 4 k 1 P ( 2 k 1 1 ) ih D d + D oh + D ack w + 2 r w ( T sw + T prop ) i (B.2) 152

PAGE 170

AppendixB:(continued) )h= D d [ 1 BER ] D d + D oh k = 0 4 k 1 1 ( 1 BER ) D d + D oh ( 2 ( k 1 ) 1 ) [ D d + e ] (B.3) inwhich e is, e = D ack w + 2 r ( T sw + T prop ) w + D oh (B.4) 153

PAGE 171

AbouttheAuthorMr.DaladierJabbaisaPh.D.candidateintheDepartmentofC omputerScienceand EngineeringattheUniversityofSouthFlorida.Mr.Jabbare ceivedtheBachelorofSciencedegreeinAugust1991fromtheDepartmentofSystemEngi neering,Universidad delNorte,Barranquilla,Colombia.HereceivedhisM.S.deg reeinComputerScience fromITESM(Mexico)andUNAB(Colombia)in2002.Healsorece ivedhisM.S.degree inComputerEngineeringfromUSF(US)in2007.HeisaCOLCIEN CIAS-LASPAU scholaratUniversidaddelNorte,Barranquilla,Colombia. Hisresearchinterestfocuseson underwateracousticandwirelessnetworks.