USF Libraries
USF Digital Collections

A control layer algorithm for ad hoc networks in support of urban search and rescue (USAR) applications

MISSING IMAGE

Material Information

Title:
A control layer algorithm for ad hoc networks in support of urban search and rescue (USAR) applications
Physical Description:
Book
Language:
English
Creator:
Ramarathinam, Venkatesh
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Link stability
Link breakage
Energy efficiency
Ad hoc routing protocols
Dissertations, Academic -- Computer Science -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Ad hoc networks have gained significant importance and gathered huge momentum within the wireless network research community. We explore the novel idea of applying ad hoc networking for urban search and rescue operations. Several algorithms have been proposed and implemented for routing in ad hoc networks and their performance have been thoroughly analyzed. But none of the prior work deals specifically for search and rescue operations, which entail certain specific criteria such as prevention of node loss, maximizing the area of coverage and constant and instantaneous access to a main controller. In this thesis, we propose a centralized and adaptive algorithm tailored for efficient performance of mobile nodes assisting in search and rescue operations. The proposed algorithm assists in finding and maintaining stable links between the mobile nodes and base station, while optimizing the area of coverage and energy efficiency of the nodes. The algorithm is implemented using ns (network simulator), and its performance is compared with that of a widely used ad hoc routing protocol, Ad hoc On-demand Distance Vector (AODV) routing protocol. We use frequency of link breakages, network throughput and routing overhead as our performance metrics. This algorithm can also be extended to provide support for routing among mobile nodes.
Thesis:
Thesis (M.S.C.S.)--University of South Florida, 2004.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Venkatesh Ramarathinam.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 78 pages.

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 - 001680986
oclc - 62496390
usfldc doi - E14-SFE0000604
usfldc handle - e14.604
System ID:
SFS0025294:00001


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 Ka
controlfield tag 001 001680986
003 fts
005 20060215071056.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051207s2004 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000604
035
(OCoLC)62496390
SFE0000604
040
FHM
c FHM
049
FHMM
090
QA76 (ONLINE)
1 100
Ramarathinam, Venkatesh.
2 245
A control layer algorithm for ad hoc networks in support of urban search and rescue (USAR) applications
h [electronic resource] /
by Venkatesh Ramarathinam.
260
[Tampa, Fla.] :
b University of South Florida,
2004.
502
Thesis (M.S.C.S.)--University of South Florida, 2004.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 78 pages.
3 520
ABSTRACT: Ad hoc networks have gained significant importance and gathered huge momentum within the wireless network research community. We explore the novel idea of applying ad hoc networking for urban search and rescue operations. Several algorithms have been proposed and implemented for routing in ad hoc networks and their performance have been thoroughly analyzed. But none of the prior work deals specifically for search and rescue operations, which entail certain specific criteria such as prevention of node loss, maximizing the area of coverage and constant and instantaneous access to a main controller. In this thesis, we propose a centralized and adaptive algorithm tailored for efficient performance of mobile nodes assisting in search and rescue operations. The proposed algorithm assists in finding and maintaining stable links between the mobile nodes and base station, while optimizing the area of coverage and energy efficiency of the nodes. The algorithm is implemented using ns (network simulator), and its performance is compared with that of a widely used ad hoc routing protocol, Ad hoc On-demand Distance Vector (AODV) routing protocol. We use frequency of link breakages, network throughput and routing overhead as our performance metrics. This algorithm can also be extended to provide support for routing among mobile nodes.
590
Adviser: Labrador, Miguel.
653
Link stability.
Link breakage.
Energy efficiency.
Ad hoc routing protocols.
0 690
Dissertations, Academic
z USF
x Computer Science
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.604



PAGE 1

AControlLayerAlgorithmforAdhocNetworksinSupportofUrbanSearchandRescue(USAR)ApplicationsbyVenkateshRamarathinamAthesissubmittedinpartialfulllmentoftherequirementsforthedegreeofMasterofScienceinComputerScienceDepartmentofComputerScienceandEngineeringCollegeofEngineeringUniversityofSouthFloridaMajorProfessor:MiguelLabrador,Ph.D.SrinivasKatkoori,Ph.D.KimonValavanis,Ph.D.DateofApproval:March30,2004Keywords:linkstability,adhocroutingprotocols,energyefciency,linkbreakagecCopyright2004,VenkateshRamarathinam

PAGE 2

ACKNOWLEDGEMENTSItakethisopportunitytoexpressmysincerethankstoDr.MiguelLabrador,forgivingmethiswonderfulopportunityofworkingonthisproject.Iamalsogratefultohimforhisextendedsupportandguidancethroughoutthecourseofthiswork,andformakingmystudyatUSFapleasantandexcitingeducationalexperience.MysincerethankstoDr.KatkooriandDr.Valavanis,forbeinginmycommitteeandfortheirvaluablecommentsandsuggestions.Ittakesmorethanwordstoexpressmythankstomyfamilyfortheirconstantmotivationandsupport,withoutwhichthisworkwouldnothavebeenpossible.Ithankallmyfriendsfortheircontinuousencouragementandsupport.

PAGE 3

TABLEOFCONTENTSLISTOFTABLESiiiLISTOFFIGURESivABSTRACTviCHAPTER1INTRODUCTION11.1Background21.2UrbanSearchAndRescue(USAR)41.2.1CommunicationsinUSAROperations51.3Motivation61.4ContributionofthisThesis71.5OrganizationofthisDocument7CHAPTER2BACKGROUNDANDLITERATUREREVIEW82.1DesignofAdHocRoutingProtocols82.1.1OverviewofRoutingMethods82.2ReviewofAdhocRoutingProtocols112.2.1DestinationSequencedDistanceVector(DSDV)112.2.2DynamicSourceRouting(DSR)122.2.3AdhocOn-demandDistanceVector(AODV)132.2.4TemporallyOrderedRoutingAlgorithm(TORA)142.2.5AssociativityBasedRouting(ABR)142.2.6SignalStability-basedAdaptiveRouting(SSA)152.2.7LocationAidedRouting(LAR)162.3EvaluatingPerformanceandSuitabilityIssues162.4SimulatorandOtherToolsUsed18CHAPTER3PROPOSEDALGORITHM-DESIGNANDDEVELOPMENT203.1DesignConsiderationsandChoicesfortheNewAlgorithm203.2ApproachandDesignoftheAlgorithm213.2.1OutlineoftheProblem213.2.2AlgorithmattheMainController263.2.3LogicattheIndividualNodes283.2.4AlphaEstimation303.3ChoiceofAODVastheRoutingLayerProtocol313.4VariationsPossibleintheProposedAlgorithm333.5SampleModelBasedExplanationoftheAlgorithm35i

PAGE 4

CHAPTER4IMPLEMENTATIONANDEXPERIMENTALRESULTS424.1Implementationinns-2.26424.1.1PhysicalLayer424.1.2DataLinkLayer444.1.3ControlLayer444.1.4RoutingLayer464.2SimulationScenario464.2.1MobilityModel474.2.2CommunicationModel474.3Results484.3.1SimulationModel1:GroupMobility484.3.2SimulationModel2:ClusterBasedRandomWaypointModel514.3.3SimulationModel3:RandomWaypointModel(Indepen-dentDirection)524.3.4SimulationModel4:RandomWalkModel544.3.5SimulationModel5:RandomWaypointModel(ClusterBasedMobility)574.3.6SimulationModel6:RandomWaypointModel594.3.7AreaofCoverage61CHAPTER5CONCLUSIONANDFUTUREWORK665.1Conclusion665.2FutureWork66REFERENCES68ii

PAGE 5

LISTOFTABLESTable3.1FormatofanInformationTableMaintainedatIndividualNodes24Table3.2FormatofanUpdateTableMaintainedattheMainController24Table3.3InformationTableatNode1AlongwithCompositeThresholdValues34Table3.4InformationTableatNode1and2attimet=1.0s36Table3.5UpdateTableatNodes1and2attimet=1.2s36Table3.6DataStructureatMainControllerattimet=1.2s38Table3.7InformationTableatNode1and2attimet=50.0s38Table3.8UpdateTableatNodes1and2attimet=50.4s39Table3.9DataStructureatMainControllerattimet=50.4s39Table3.10InformationTableatNode1and2attimet=100s40Table3.11UpdateTableatNodes1and2attimet=100.8s40Table3.12DataStructureatMainControllerattimet=100.8s41Table4.1ComparativeDistanceCoveredbyMobileNodesinModel163Table4.2ComparativeDistanceCoveredbyMobileNodesinModel263Table4.3ComparativeDistanceCoveredbyMobileNodesinModel364Table4.4ComparativeDistanceCoveredbyMobileNodesinModel464Table4.5ComparativeDistanceCoveredbyMobileNodesinModel565Table4.6ComparativeDistanceCoveredbyMobileNodesinModel665iii

PAGE 6

LISTOFFIGURESFigure1.1StandardWirelessModelwithMobileNodesandFixedInfrastructure2Figure1.2AnAdHocNetworkof3MobileNodes3Figure1.3ExampleTopologyof6MobileNodes4Figure2.1Table-driven,On-demandandHybridAdHocProtocols10Figure2.2DynamicSourceRouting(DSR)MethodofROUTEREQUEST12Figure2.3DynamicSourceRouting(DSR)MethodofROUTEREPLY13Figure3.1ScenarioConsistingofMainControllerand6MobileNodes21Figure3.2LayeredApproachforDesignoftheProtocol22Figure3.3SampleScenarioof6NodeswiththeirTransmissionRange23Figure3.4FlowchartfortheLogicatLocalNodes25Figure3.5ModelTopologywithNodesinaLinearChain33Figure3.6SampleModel:StartingTopology36Figure3.7SampleModel:NetworkTopologyattimet=50.0s38Figure3.8SampleModel:NetworkTopologyattimet=100.0s40Figure4.1MobilityModel1(StraightLineMovementAlongtheDiagonal)48Figure4.2ComparativeLinkStabilityAnalysisforMobilityModel150Figure4.3ThroughputofAllFlowsforModel150Figure4.4ThroughputatMainControllerforModel150Figure4.5NetworkOverhead(RoutingandHelloPackets)forModel150Figure4.6MobilityModel2(2ClustersMovingTowardsPresetDestination)52Figure4.7ComparativeLinkStabilityAnalysisforMobilityModel253Figure4.8ThroughputofAllFlowsforModel253Figure4.9ThroughputatMainControllerforModel253iv

PAGE 7

Figure4.10NetworkOverhead(RoutingandHelloPackets)forModel253Figure4.11MobilityModel3(NodesMovingatanInter-nodalAngleof45)54Figure4.12ComparativeLinkStabilityAnalysisforMobilityModel355Figure4.13ThroughputofAllFlowsforModel355Figure4.14ThroughputatMainControllerforModel355Figure4.15NetworkOverhead(RoutingandHelloPackets)forModel355Figure4.16MobilityModel4(RandomWalkModel)56Figure4.17MobilityModel5(NodesMovingin2DifferentClusters)57Figure4.18ComparativeLinkStabilityAnalysisforMobilityModel458Figure4.19ThroughputofAllFlowsforModel458Figure4.20ThroughputatMainControllerforModel458Figure4.21NetworkOverhead(RoutingandHelloPackets)forModel458Figure4.22ComparativeLinkStabilityAnalysisforMobilityModel560Figure4.23ThroughputofAllFlowsforModel560Figure4.24ThroughputatMainControllerforModel560Figure4.25NetworkOverhead(RoutingandHelloPackets)forModel560Figure4.26MobilityModel6(RandomWaypointModel)61Figure4.27ComparativeLinkStabilityAnalysisforMobilityModel662Figure4.28ThroughputofAllFlowsforModel662Figure4.29ThroughputatMainControllerforModel662Figure4.30NetworkOverhead(RoutingandHelloPackets)forModel662v

PAGE 8

ACONTROLLAYERALGORITHMFORADHOCNETWORKSINSUPPORTOFURBANSEARCHANDRESCUE(USAR)APPLICATIONSVenkateshRamarathinamABSTRACTAdhocnetworkshavegainedsignicantimportanceandgatheredhugemomentumwithinthewirelessnetworkresearchcommunity.Weexplorethenovelideaofapplyingadhocnetworkingforurbansearchandrescueoperations.Severalalgorithmshavebeenproposedandimplementedforroutinginadhocnetworksandtheirperformancehavebeenthoroughlyanalyzed.Butnoneofthepriorworkdealsspecicallyforsearchandrescueoperations,whichentailcertainspeciccriteriasuchaspreventionofnodeloss,maximizingtheareaofcoverageandconstantandinstantaneousaccesstoamaincontroller.Inthisthesis,weproposeacentralizedandadaptivealgorithmtailoredforefcientperformanceofmobilenodesassistinginsearchandrescueoperations.Theproposedalgorithmassistsinndingandmaintainingstablelinksbetweenthemobilenodesandbasestation,whileoptimizingtheareaofcoverageandenergyefciencyofthenodes.Thealgorithmisimplementedusingns(networksimulator),anditsperformanceiscomparedwiththatofawidelyusedadhocroutingprotocol,AdhocOn-demandDistanceVector(AODV)routingprotocol.Weusefrequencyoflinkbreakages,networkthroughputandroutingoverheadasourperformancemetrics.Thisalgorithmcanalsobeextendedtoprovidesupportforroutingamongmobilenodes.vi

PAGE 9

CHAPTER1INTRODUCTIONIntherecentpast,wehavewitnessedatremendousgrowthindeploymentofwirelessnetworktechnologydrivenbytheneedforubiquitousserviceandrapiddevelopmentsintelecommunicationsinfrastructure.Mobilehostssuchasnotebookcomputers,featuringpowerfulCPUsandgigabytesofdiskspacearenoweasilyaffordableandbecomingquitecommonineverydaylife.Atthesametime,hugeimprovementshavebeenmadeinwirelessnetworkhardware,andeffortsarebeingmadetointegratethetwointoameaningfulresourcesuchastheInternet.Wearewitnesstolargescaleproliferationofmobilecomputingandwirelesstechnologyinourday-to-daylivesintheformofvarioushardwareinterfacesandtechnologydevices,runningnumerousapplicationscateringspecicallytowirelesstechnology.TheuseofcellphonesandPDA'sformobilevideoconferencing,GPSbasedtrackingsystemsandremotewirelesssensorsurveillancegivesusanindicationofthegrowthandproliferationofwirelesstechnologyintoday'sworld.Theincreaseddemandandusageofmobiledevices,directlycorrelatestotheinateddemandformobiledataandinternetservices.AccordingtoastudybyCahnersIn-StatGroup,thenumberofsubscriberstowirelessdataserviceswouldreach1.3billionbyendof2004,andthenumberofwirelessmessagessentpermonthwillreach244billionbyDecember2004[1].Butthesedevicesandtechnologyusethestandardwirelessnetworkmodelofabasestation,repeaters,accesspoints,andwirelessnodes(Figure1.1).Oftentimeshowevermobileuserswillwanttocommunicateinsituationsinwhichnoxedwiredinfrastructureisavailable,becauseitmaynotbepossibletoprovidethenecessaryinfrastructureorbecausetheexpediencyofthesituationdoesnotpermitthisinstallation.Thetermadhocnetworksrefertosuchacollectionofwirelessmobilehostsformingatemporarynetworkwithouttheaidofanyestablishedinfrastructureorcentralizedadministration[2](Figure1.2).1

PAGE 10

Figure1.1StandardWirelessModelwithMobileNodesandFixedInfrastructure1.1BackgroundThehistoryofadhocnetworksdatesbacktotheDARPAradiopacketnetworkin1972[3],whichwasprimarilyinspiredbytheefciencyofthepacketswitchingtechnology,suchasband-widthsharingandstoreandforwardrouting,anditspossibleapplicationinmobilewirelessenvi-ronment[1].But,itwasnotuntiltheearly90'swhenresearchintheareaofadhocnetworksgainedsignicantmomentumandwidespreadattention.Thiscouldbeattributedtothesurgeincheapavail-abilityofnetworkhardware,microcomputerrevolution,andtheincreasingnumberofapplicationsthatrequiredanadhocnetworkkindofsetup.MANET'sorMobileAdhocNetworkshavegainedsignicantmomentumastheyarethesolutionforprovidingnetworkservicestomobileusersatplaceswherethereisnoinfrastructureoranexistinginfrastructureneedswirelessextensions.Someofthecommonapplicationsofadhocnetworksare:conferencehalls,classrooms,searchandrescueoperations,vehicularcommunication,wirelesssurveillanceandmilitaryoperations.Inanadhocnetwork,everynodeactsasarouter,andforwardspacketstowardsthedestination.Itisaself-organizednetworkwhereeverynodecooperatestoprovideconnectivityandservices.ConsiderthetopologyshowninFigure1.3,with6mobilenodesforminganadhocnetwork.Eventhoughthereisnodirectlinkbetweennode1andnode6,packetsfromnode1aredeliveredtonode6throughintermediatenodes2,3,4and5.Suchanarrangementoffersmanyadvantages,someofwhicharelistedbelow:2

PAGE 11

Figure1.2AnAdHocNetworkof3MobileNodesAd-hocnetworksprovideforfastandinstantaneoussetup,reducingthecostduetolackofxedinfrastructure.Whenappliedtoscenariossuchassearchandrescueoperationsadhocnetworksresultinbetterperformancethanconventionalwirelessnetworksbecauseoftheirnon-hierarchicaldis-tributedcontrolandmanagementmechanisms[4].Withproperpowercontrolmechanismsandrelaying,theyprovideforinherentscalabilityandincreasedareaofcoverage,whereeachnodeaddstotheexistingnetworkcapacity.Providesforincreasedmobilityandgreaterexibilityinnetworktopologyasanadhocnet-workcouldbebroughtupandtorndowninaveryshorttimeAdhocnetworksarewellsuitedtouseunlicensedbandsandprovideforincreasedspectrumreusepossibility.Associatedwiththeseadvantagesandapplicationpossibilitiesaresomeinherentdrawbacksthatholdad-hocnetworksfarfrombeingdeployedonlarge-scalecommercialbasis.Somefundamentalad-hocnetworkingproblemsthatremainunsolvedorneedoptimizedsolutionsarelistedbelow.Thenodesinanadhocnetworkcanmovearbitrarilywhichresultsinaverydynamictopologyandfrequentlinkbreakages,disruptingcommunicationbetweennodes.Oftentimesnodesoperatinginanadhocnetworkrelyonbatteryforenergy,thusforthesenodesenergy-efcientprotocolsbecomeacriticaldesigncriterion.3

PAGE 12

Figure1.3ExampleTopologyof6MobileNodesBandwidthutilizationisanothersignicantfactorforconcern,thusnecessitatingreducedroutingoverheadandgoodcongestioncontrolmechanisms.Wirelessnetworksarepronetophysicalandinformationsecurityvulnerabilities,andismoredominantinadhocnetworks,wherethereisnocontrolmechanismandcentralizedadminis-tration.1.2UrbanSearchAndRescue(USAR)Traditionallytherehavebeendifferenteldsofrescueoperations[5]:Woodland/wildlandsearch&rescueMountainsearchandrescueWatersearchandrescue(includesdiving)Urbansearch&rescue(includesconnedspacerescue)Combatsearch&rescueCavesearchandrescueOftheabovementionedeldsofrescueoperations,UrbanSearchandRescuefocussesonlo-catinglifeandresourcesatmanmadestructures,suchasbuildingscollapsedinanearthquake,oratdisastersitesaffectedbyarticialornaturalcalamities.Recentresearchinvestigatestheuseofrobotsinsuchscenarios.Thesedisastersitesposeseveralsituationalhazardsthatdrasticallyaffectstheefciencyofhumanrescueteams.Someofthemostcommonofthesehazardsandlimitationsarelistedbelow[6]:4

PAGE 13

Disastersitesareinherentlyunsafeandmovementinsidethesesitesisextremelyrestrictedduetoavailabilityofonlysmallornoentryvoids,toexploretherubble.Vibrationsduetomobilitymightfurtheraffectthefoundationofthecollapsedconstructionandcouldtriggerasecondarycollapse.Disastersitesareusuallycontaminatedbydamagetowater/sewagedistributionsystems,toxicgasspill,bodyuidsandotherhazardousmaterialsandgases.Oftentimes,rescueoperatorswouldrstneedtoextinguishtheblazingreinthesesitesbeforeproceedingwithanykindofrescueoperations.Thissoakstheentiresiteandleavesitwetandslippery.Alltheabovementionedfactorscoupledwiththelackofenoughtrainedskilledprofessionalsmakesitimperativetolookforothereffectivemeanstocarryoutrescueoperations.Theuseofmobilerobotsprovidesaneffectivealternativeforimprovedefciencyinurbansearchandrescueopera-tions.Duetosmallersizesandrobustdesign,robotscanexploredisastersitesthatposenumeroushazardthreatsandarenotconduciveforexplorationbyreliefworkers.1.2.1CommunicationsinUSAROperationsUsuallywehaveateamofautonomousmobilerobotssurveyingadisastersiteforlifeandresources[7].Communicationbetweentheserobotsiscriticaltotheperformanceinsearchandres-cueoperations.Thiswouldfacilitatetele-operationoftherobotsandalsoprovidesvitalinformationonthendingsbytherobotstothemaincontroller.TheIEEE802.11standard[8]forwirelessLANiscurrentlyusedasthecommunicationplatformfortheserobots.Butcommunicationbetweentherobotsisheavilydisruptedbyinterferencesduetometalandotherstructuraldebris,andisalsoaffectedbytheuseofavailableradiochannelsbyrescuepersonalandmedia[6].Also,whenop-eratingautonomously,themobilerobotsneedtoconstantlycommunicatewiththemaincontrollersituatednearthedisastersite.Thisseverelyaffectstheavailableareaofmobilityforthemobilerobots.Anyrobotthathasmovedoutofthetransmissionrangeofthemaincontrollercouldbeassumedtobetrappedinthedebrisortobelost.Thisseverelyaffectstheperformanceoftheserobotsandalsoinducesnancialdamage.Thusforbetterperformanceresultsandeffectiveusage5

PAGE 14

ofrobotsforsearchandrescueoperations,constantcommunicationbetweenthemobilerobotsandthemaincontrollerneedstobeensured,whilealsoaccountingformaximizedareaofcoverage.Byprovidingaconstantcommunicationlinkbetweenthemobilerobotsandthemaincontroller,itisensuredthattherobotsdonotgetlost;thetermnodeconnectivityisintroducedheretodenotethesame.Nodeconnectivityisdenedastheabilityofanodetocontinueorstopitsmobilitywithoutbreakingawayfromthenetworkofnodes,whileremaininginconstantcommunicationwiththemaincontroller.Forminganadhocnetworkofthemobilerobotsandthemaincontrollereffectivelyaddressestheissueofmaximizedareaofcoverage.Byforminganadhocnetwork,intermediatenodesactasarouterforwardingpacketstowardsthedestination.Bythismethod,robotscontinuetheirmobilitybeyondthetransmissionrangeofthemaincontroller,ifthereexistsanintermediatenodethroughwhichitcanestablishaconnectionwiththemaincontroller.However,forminganadhocnetworkofmobilerobotsdoesnotaddresstheissueofnodeconnectivity.Itisessentialtoensurenodeconnectivityinapplicationswherelossofanode(mobilerobotinthecaseofurbanandsearchandrescueoperations)couldbedetrimentaltotheperformanceofthesystem.1.3MotivationVastmajorityoftheresearchworkdoneintheareaofadhocnetworkshasbeenfocussedondesigninganddevelopingroutingprotocolstoaddresstheissuesofnodemobility,overheadandenergyefciency.Therehasbeenanincreasedattentionindevelopingroutingprotocolsthatconsidertheissueoflinkstability.[9]and[10],discussdesignoflinkstabilitybasedroutingprotocol,whereroutestodestinationareselectedbasedonthestrengthofsignalsreceivedfromneighboringnodesordurationforwhichthelinkhasbeenactive.Butnoneoftheexistingworksguaranteelinkstabilityorprovidenodeconnectivity.Nodeconnectivityisofutmostimportanceinscenarioswherethemobilityofthenodeiscompletelyautonomous,orbreakageofanodefromanetworkisdetrimentaltothecostandperformanceofthesystem.Thoughtheexistingprotocolssuggestselectionanduseofstablelinksforcommunication,theydonotguaranteelinkstabilityoraccountfornodeconnectivity.Thisworkaddressestheissueofnodeconnectivityinadhoc6

PAGE 15

networks,andprovidesareliablesolutionforUrbanSearchandRescue(USAR)applications,wherethemobilityofthenode(robots)isstoppedbeforebreakingawayfromthenetwork.1.4ContributionofthisThesisThisthesisinvestigatesmethodsofprovidingstablelinksandconstantnodeconnectivityformobileadhocnetworks.Themaincontributionsofthisworkare:Providesacomprehensivereviewofthecurrentandpastworkdoneintheareaofroutingprotocolsformobileadhocnetworksandoutlinestheadvantagesanddisadvantagesoftheseprotocolswhenappliedtosearchandrescueoperations.Introducestheconceptofcontrollayerforadhocnetworks,inter-operatingbetweentherout-ingandthedatalinklayers,andsuggeststhedesignofacontrollayeralgorithmtoprovidefornodeconnectivityandlinkstability.Evaluatestheperformanceoftheproposedalgorithmbycomparingitwithexistingprotocolsusingstandardperformancemetrics.Implementsthealgorithminthens-2.26simulator,andismadeavailableforfurtherresearchinthisarea.1.5OrganizationofthisDocumentThischapterprovidesabriefoutlineandmotivationforthiswork.Thenextchapterprovidesanoverviewofthebackgroundrelatedtothismaterial,andreviewsthepastandcurrentworkrelevanttothiseld.Thethirdchapterdiscussesdesignconsiderationsandexplainsthealgorithmwiththehelpofowchartsandpseudocode,andmakeuseofasamplemodeltoillustratetheworkingofthealgorithm.Thefourthchapterdescribesthesimulationsetupandimplementationdetails,analyzesresultsandevaluatesthealgorithmbasedonstandardperformancemetricsandbycomparingitwithexistingprotocols.Thenalchapterconcludesthisworkandenliststhescopeforfuturework.7

PAGE 16

CHAPTER2BACKGROUNDANDLITERATUREREVIEWThischapterreviewssomeofthewellresearchedandpublishedworksonadhocroutingpro-tocolsanddiscusstheirsuitabilityforapplicationtourbansearchandrescueoperations.Italsoprovidesabriefoverviewofthesimulatorandothertoolsusedtoimplementtheproposedalgo-rithm.2.1DesignofAdHocRoutingProtocolsRoutingisaverycriticalaspecttotheperformanceofanadhocnetwork,andrightly,vastma-jorityoftheresearchinthisareahasfocussedondesigningefcientroutingprotocols.Thedesignofaroutingprotocolforadhocnetworksneedstoaccountforthehighlydynamicnetworktopology,batterypowerofthemobilenodes,minimizedroutingoverheadandshouldprovidestablelinks.Also,thedesignoftheroutingprotocolneedstobedecentralized,becauseinanadhocnetworkeverynodeperformsthesamefunctions,andthereisnocentraladministratorofthenetwork.2.1.1OverviewofRoutingMethodsTheproblemofroutinginanadhocnetworkissimilartothedistributedversionoftheclassicalshortestpathproblem[11].Everynodeinthenetworkmaintainsalistofpreferredneighbors(nexthopnodes)foreachdestinationinthenetworkintheformofaroutingtable.Everydatapacketcontainsadestinationnodeidentierinitsheader.Anodereceivingadatapacketchecksthedestinationheadereld,andifitisnottheintendeddestination,forwardsthepackettothenexthopneighbortowardsthedestinationbasedontheinformationinitsroutingtable.Thedesignofthevariousadhocroutingprotocolsdifferinthemannerinwhichtheneighborinformation,routingtables,andnexthoptowardsdestinationareestimated,maintainedandupdated.Essentially,allroutingprotocolsattemptinachievingtheoptimalpathtowardsdestination.Routingprotocolsalso8

PAGE 17

differinthemannerinwhichtheyestimateoptimalpath.Someprotocolsusedistancebetweenneighborsasameasureofdeterminingthenexthoptowardsadestination,whilesomeprotocolsselectthenexthopbasedonthedurationforwhichalinktothatnodehasbeenactive.Routingprotocolsusethestandardnexthoproutingmethodsandcanbeclassiedintothetwoprimarycategoriesoflink-stateanddistance-vectorrouting.Inalink-stateapproacheachnodeinthenetworkmaintainsaviewofthecurrentnetworktopology,andacostisassociatedtoeverylinkinthenetwork.Inorderforallnodestohaveaconsistentviewofthenetwork,eachnodeforwardsitscostinformationoftheoutgoinglinkstoallothernodes,usingaprotocolsuchasooding.Uponreceivingthisinformation,thenodesupdateitsviewofnetworktopologyandappliesashortestpathalgorithmtoselectthenexthoptowardseachdestination.Aftertheinitialset-uptimeofthenetwork,linkcostupdatesarebroadcastedonlywhenthereisachangeinthenetworktopology.Thismightintroducetemporarystaleroutes,butitisusuallycorrectedbythetimeittakesamessagetotraversethediameterofthenetwork[12].Linkstateprotocolsprovideverygoodscalabilityandtheroutingoverheadisminimalevenathighernodedensity.Anexampleoflink-stateroutingistheOpenShortestPathFirst(OSPF)[13]protocol.Distance-vectorroutinguseshopcounttodeterminethebestpathtowardsadestination.Thecalculatedroutewouldbeoptimalintermsofthenumberofhopsrequiredtowardsthedestination,butmostoftenthebestroutetowardsadestinationneednotnecessarilybetheonewithshortestnumberofhops.Convergencetimeisthetimeittakesforallnodesinthenetworktomakechangestoitsviewofthenetworktopology,anddistance-vectorroutingachievesconvergencebyperiodi-callysendingitsroutingtablestoallnodesinthenetwork.Innetworkswithlargenumberofnodes,distancevectorroutingtakestimetoconvergeandalsoresultsinincreasedoverheadandlowerbandwidthutilization.RoutingInformationProtocol[14]isaclassicexampleofdistance-vectorrouting.Routingprotocolsforadhocnetworkscanbeclassiedasproactive,reactive,andhybridproto-cols(SeeFigure2.1).Table-drivenorproactiveroutingprotocolssendsitsroutingtableasperiodicupdatestoitsneighborsortoallnodesinthenetwork.Bysendingperiodicupdatestheprotocolensuresthateverynodeinthenetworkhasapictureofthecurrentnetworktopology,andeliminatesthepresenceofstaleroutes.Butoneoftheinherentdrawbacksofthismethodisthatitincreasestheroutingoverheadandthereforereducestheavailablebandwidthfordatapackets.Theincrease9

PAGE 18

Figure2.1Table-driven,On-demandandHybridAdHocProtocolsinroutingoverheadisdirectlyproportionaltothenumberofnodesinthenetworkandtotherateofupdates.Thus,thetabledrivenapproachesdonotprovidescalabilityandarealsonotaptforscenarioswithlittleornonodemovements.Thedifferentproactiveprotocolsdifferinthenumberoftablesmaintained,andthemethodbywhichchangesinthenetworkarebroadcasted.DestinationSequencedDistanceVector(DSDV)[15]andWirelessRoutingProtocol(WRP)[16]areexamplesofproactiveortabledrivenapproaches.On-demandorreactiveprotocols,createroutesonlywhenrequiredbythesourcenode.Whenanoderequiresaroutetoadestination,itinitiatesaroutediscoveryprocesswithinthenetwork.Allnodesinthenetworkreceivethispacket,andeitherthedestinationnode,oranyothernodethathasroutetothisdestinationrepliesbacktothesourcewiththecompleterouteinformation.Eachnodereceivingthispacketcachestherouteinformationandusesitforfuturepacketstowardsthisdestination.Thisapproachsignicantlyreducesroutingoverhead,andnegatestheeffectofroutingloops.Therearemanyadhocroutingprotocolsinthiscategory.DynamicSourceRouting(DSR)[2],AdhocOn-demandDistanceVector(AODV)Routing[17]andTemporallyOrderedRoutingAlgorithm(TORA)[18]areafewexamples.Hybridroutingprotocolsincorporatethemeritsofon-demandandproactiveroutingprotocols.Nodesaregroupedintoclusters,andfornodeswithinthesamecluster,atable-drivenroutingpro-10

PAGE 19

tocolisused,whiletocommunicatewithadestinationnoderesidingoutsidetheclusterofsourcenode,anon-demandsearchmethodisused.2.2ReviewofAdhocRoutingProtocolsThissectionreviewssomeofthemostcommonadhocroutingprotocols:DestinationSe-quencedDistanceVector(DSDV)[15],DynamicSourceRouting(DSR)[2],AdhocOn-demandDistanceVector(AODV)[17],TemporallyOrderedRoutingAlgorithm(TORA)[18],AssociativityBasedRouting(ABR)[19],SignalStability-basedAdaptiveRouting(SSA)[10],andLocationAidedRouting(LAR)[20].Someoftheotherproposedprotocolsarediscussedin[21,22,16,23,9,24]andthereferenceslistedtherein.2.2.1DestinationSequencedDistanceVector(DSDV)TheDestinationSequencedDistanceVector(DSDV)routingprotocolwasintroducedin[15].DSDVisadistancevectorroutingprotocolthatimplementsmodicationstotheBellman-Fordroutingmechanismasspeciedby[14],tomakeitsuitableforadynamicandself-startingnetworkmechanisms.Everynodeinthenetworkmaintainsaroutingtablewithentriesforeverydestinationinthenetworkalongwiththecorrespondingnumberofhopstowardsthedestination.Alsoeachentryismarkedbyasequencenumberdistributedbythedestinationnode.Thesequencenumberscouldbeusedtodistinguishstaleroutesfromnewroutes,therebyavoidingtheformationofroutingloops.DSDVrequireseachnodetoadvertisetoitsneighborsitswholeroutingtable,includingthenexthopinformationtoreachallotherdestinationsinthenetwork.DSDVsendsupdatesperiodi-cally(proactive)orintheeventofalinkchangeandusessequencenumbersinordertousethemostrecentinformation.Moreovertheupdatessentoutmightbeuctuating,andwouldcreatetemporaryroutingloops.Todampenuctuations,settlingtimeiscalculatedforeverynodeintheroutingtable,andupdatesaresentoutafterthissettlingtimeisreached.DSDVcanbebeneciallyappliedtoadhocnetworksandguaranteesloopfreepathstoeachdestinationatallinstants.11

PAGE 20

Figure2.2DynamicSourceRouting(DSR)MethodofROUTEREQUEST2.2.2DynamicSourceRouting(DSR)DynamicSourceRouting(DSR),presentedin[2]isanon-demand(reactive)routingprotocolthatisbasedonsourcerouting,wherepacketsleaveasourcenodetowardsthedestination,withthecompleterouteinformationintheirheaders.Eachnodemaintainsalocalcacheofsourceroutesofwhichitisaware.Whenanodehastosendapackettoadestination,itrstchecksitlocalcacheforasourcerouteforthatdestination.Ifithasasourceroute,itwilluseittosendthepackettowardsthedestination.Ifthereisnoentryforthedestinationintheroutecache,thenthesourcenodeinitiatesaROUTEREQUESTtothatnode.EachnodereceivingthispacketappendsitsinformationintheROUTERECORDoftheREQUESTpacket.Ifthereceivingnodehasaroutetotheintendeddestination,itappendsthisinformationtotheROUTERECORDandforwardsthepackettothenodethatinitiatedtheroutediscovery.However,ifthereceivingnodedoesnothavearoutetothedestination,itbroadcaststheROUTEREQUESTpackettoallnodesinitsoutgoinglinks.Toavoidmultiplebroadcastofthesamepacket,anodeforwardsapacketonlyifithasnotseenitalreadyanditsinformationisnotappendedtotheROUTEREQUEST.Figure2.2illustratestheformationofrouterecordastherouterequestmessagepropagatesthroughthenetwork.WhentheROUTEREQUESTisreceivedbythedestinationnodeoranintermediatenodewithroutetothatdestination,itgeneratesaROUTEREPLYwiththereversedinformationfromtherouterecordandsendsittowardsthedestination.Figure2.3illustratestheroutereplymethodofDSR.Forroutemaintenance,DSRmakesuseofROUTEERRORpacketstoinformnodesinthenetworkabouta12

PAGE 21

Figure2.3DynamicSourceRouting(DSR)MethodofROUTEREPLYstalerouteorbrokenlink,anduponreceivingarouteerrorpacketallnodesupdatetheinformationintheirroutecachetoreectthecurrentnetworktopology.2.2.3AdhocOn-demandDistanceVector(AODV)AdhocOn-demandDistanceVector(AODV)discussedin[17]isareactiveprotocolbasedonDSDVandDSR.Itborrowstheideaofon-demandroutingfromDSRandusesthemechanismofhop-by-hoproutingandsequencenumbersfromDSDV.Eachnodeinthenetworkmaintainsaroutingtablewithentriesforeveryothernodeinthenetworkalongwiththesequencenumberoftheroutereplyreceivedfromthatnode.AnodeSthatrequiresaroutetodestinationD,initiatesaroutediscoveryandbroadcastsaROUTEREQUESTpackettoitsneighbors,whichinturnbroadcasttherequesttowardsthedestination.UnlikeDSR,whereeachnodereceivingtherouterequestappendsitsinformationtotherouterecord,nodesreceivinganAODVrouterequest,initiatesaROUTEREPLYifitisthedestinationD,orhasaroutetothedestination.Else,thenodedoesare-broadcastoftherouterequest.EachnodethatforwardstheROUTEREQUESTcreatesareverserouteforitselfbacktonodeS.ROUTEREPLYcontainsthenumberofhopstowardsthedestinationandthesequencenumberforthedestination,mostrecentlyseenbythenodegeneratingthereply.EachnodethatparticipatesinforwardingthisreplybacktothesourceorinitiatoroftheROUTEREQUEST(nodeS)createsaforwardroutetothedestinationD.ThusAODVuseshop-by-hoproutingwhereeachnodere-membersonlythennexthoptowardsadestinationandnotthecompleteroutetoadestination,as13

PAGE 22

wouldbethecasewithsourcerouting.RoutemaintenanceinAODVissimilartothatofDSRandisachievedbyperiodicbroadcastofhellomessagesorthroughlinklayerdetection.2.2.4TemporallyOrderedRoutingAlgorithm(TORA)TemporallyOrderedRoutingAlgorithm(TORA)[18]usesacontrolledoodingmechanismtodiscovermultipleroutesfromasourcetoadestinationonademandbasis.Inordertoreduceover-head,TORAsometimesutilizessub-optimalpathsinsteadoftriggeringanewoodingprocedure.Itisaloop-freedistributedroutingalgorithmbasedontheconceptoflink-reversal.ThemaindesignideaofTORAislocalizationofcontrolmessagestoaverysmallsetofnodesneartheoccurrenceofatopologicalchange.TORAusesrelativeheightsofnodeasametricforpathestimation,andeachnodeviewsthenetworkasadirectedacyclicgraph(DAG).WhenanoderequiresaroutetoadestinationitbroadcastsaQUERYpacketcontainingtheaddressofthedestinationforwhichitrequiresaroute.Thispacketpropagatesthroughthenetworkuntilitreachesthedestinationnodeoranyothernodethathasaroutetothisdestination.ThereceivingnodethenpropagatesaUPDATEpacketwithitsheightsetrelativetothedestination.EachforwardingnodethensetsitsheightgreaterthantheheightoftheneighborfromwhichitreceivedtheUPDATEpacket.Inthismanneraseriesofdirectedlinksiscreatedfromthesourcetothedestination.Whenanodedetectsabrokenlinktooneofitspreviousneighbors,itsetsitheightsothatitisatalocalmaximumwithrespecttoitsneighborsandtransmitsanUPDATEpacketwiththisinformation.BecauseTORAusesinter-nodalcoordinationforroutemaintenance,thereisapotentialfortemporaryroutingloopsanductuations,howeverthesearejusttemporaryandrouteconvergencewouldeventuallyoccur.2.2.5AssociativityBasedRouting(ABR)AssociativitybasedRouting(ABR)[19]isasourceinitiatedon-demandroutingprotocol.Asso-ciativityisameasureofthenodesconnectivitywithitsneighbors.Nodesthathavegreaterassocia-tivityhavebeenstableforalongerduration,whilenodeswithlowerassociativityindicategreatermobilityrates.Eachnodeinthenetworkbroadcastsperiodicbeacons,andeveryneighborreceivingthisbeaconkeepsacountofthefrequencyatwhichthesebeaconshavebeenreceived.Largenum-berofbeaconsexchangedbetweennodesindicatethatthelinkhasbeenstableforalongerduration.14

PAGE 23

AsuggestedmodicationofABRmakesuseofbatterypowerlifeofthenodes,signalstrengthandrouterelayingloadinestimationofoptimalpathtowardsdestination.TherearethreephasesoftheABRprotocol:(a)routediscoveryphase,(b)routereconstructionphase,and(c)routedeletionphase.ABRlaysemphasisonlongevityofroutesandsignicantlydiffersfromthetraditionaldis-tancevectorroutingalgorithmsthatconcentrateonestimatingroutesbasedonshortestpathtowardsdestination.2.2.6SignalStability-basedAdaptiveRouting(SSA)SignalStability-basedrouting[10]usessignalstrengthandstabilityoftheindividualhostsasrouteselectioncriteria.Selectingthemoststablelinks,linksthatexhibitstrongestsignalsformaximumamountoftime,leadstolongerlivedroutesandlessermaintenance.SSAfollowstheon-demandroutingparadigm.ItdiffersfromAssociativityBasedRouting(ABR)inregardtoitsroutediscoverymechanism.UnlikeABR,whereroutesareselectedpurelybasedonlinksthatarestableforlongerduration,SSAuseslinkstabilitycoupledwithsignalstability.Theauthorsclaimthatselectionofroutebasedonlyonlinkstabilitymaynotalwaysyieldbetterperformance,duetothefactthatnodesarehighlymobileinanadhocnetworkandhasunpredictablemobilitypatterns.InSignalStability-basedAdaptiverouting,sourcenodesinitiatesaroutediscoverywhenithasdatatosendtoadestinationthatisnotintheroutingtable.Intermediatenodesforwardthepacketsonlyifitisreceivedoverastrongchannelandhasnotbeenforwardedearlier.Forwardedpacketshavethehostinformationappendedtoitsheadersandthesourcenodegetsthecompleteroutewhenitreceivesaroutereplyfromthedestination.Functionally,theSSAprotocolconsistsoftwoprotocols-ForwardingProtocol(FP)andDynamicRoutingProtocol(DRP).DRPmakesextensiveuseoftheunderlyingdevicedriverinterface,whichforwardsthesignalstrengthinformationtotheroutingprotocol.Unlikeconventionalroutingprotocols,twotablesaremaintainedinSSA:theroutingtableandsignalstabilitytable.DRPmaintainsthesetablesandFPdoestheworkoflookupandforwardingthepacketsontheappropriatechannels.15

PAGE 24

2.2.7LocationAidedRouting(LAR)Whilevariousroutingprotocolsusedifferentparameterssuchasnumberofhops,bandwidthofalinkandsignalstrengthofalinktoestimaterouteinformation,LocationAidedRouting(LAR)[20]utilizeslocationinformationtoimproveperformanceofadhocwirelessnetworks.Thisclearlyimpliesthatthenodesneedtohaveamethodofestimatingtheircurrentlocation,sayviaGPS.ThisassumptioncouldseverelyhandicaptheapplicationofLARtoscenarioswhereitmightnotbepracticallyfeasibletogatherlocationinformation.LARalsomakesanotherassumptionthatasendernodeisawareofthedestinationlocationandvelocity.Inaddition,withoutconsideringsignalstrength,batterypowerlife,andconnectivityinformation,communicationperformancewillsufferifdataareroutedbasedonlocationinformationalone[25].2.3EvaluatingPerformanceandSuitabilityIssuesAvastdealofresearchhasalsobeenfocussedonanalyzingtheperformanceoftheproposedprotocolsusingdifferentperformancemetricsandvariousscenarios.Someofthemostcommonperformancemetricsarethroughputofthenetwork,packetdeliveryratio,routeestablishmentde-lay,convergencetime,routingoverheadandlinkstability.Alsothesemetricsareevaluatedfordifferentscenariossuchasvariousmobilityrates,randompausetimesandunderdifferentloadcon-ditions.ComprehensiveanalysishavealsobeendonetostudytheperformanceofTCPandUDPovertheseprotocols.[26,27,28,29]providesaperformancereviewofAODV,DSDV,DSRandTORAunderdifferentsettings.[30,31]reviewssomeofthepositionbasedroutingprotocols.In[32]theauthorshavedoneacompleteanalysisconsideringvariousnetworkpossibilitiesandperfor-mancemetrics,ontheperformanceofTCPoveradhocnetworks.[33]alsoincludesaperformancecomparisonofTCPoveradhocroutingprotocols,butanalyzestheperformanceoverstaticadhocnetworks.[34]comparestheperformanceofhybridvs.proactiveadhocroutingprotocols,while[35]reviewstheperformanceofon-demandvs.table-drivenroutingprotocols.Howeveroneoftheimportantconclusionsfromtheanalysisdonein[26]thatinuencesthedesignchoiceisthatforslowermobilityrates,oftheorderof1meter/second,performanceofAODVandDSRaresimilarandarebettercomparedtootherroutingprotocols.16

PAGE 25

Currently,communicationbetweenagroupofautonomousmobilerobotsisachievedbyformingawirelessLANoftherobots[36].ThismodelofcommunicationisbasedontheIEEE802.11wirelessstandard.TheideaofformingawirelessLANofmobilerobotswhenappliedtourbansearchandrescueoperationshasseveraldrawbacks.Someofthereasonsattributedtothesamearelistedbelow.SettingupawirelessLANrequiresinfrastructure(accesspoint)andpowertopoweruptheaccesspoint,whicharenotreadilyavailableatdisastersites.Oftentimes,inUSARoperations,themobilerobotsmaneuveringthedisastersitewouldneedtomaintainconstantcommunicationwithastationarycontroller,transmittingsearchndingsandlocationinformation.Themaincontrollerisusuallystationaryandprovidesscopefortele-operationandanalyzesndingsoftherobotstoprovidemeaningfulinformationtothereliefworkers.Toensurethisconstantcommunicationwiththemaincontroller,themobilerobotsneedtostaywithinthetransmissionrangeofthemaincontroller.Thisseverelyaffectstheareaofcoverageandtheperformanceofthesystem.FormingawirelessLANofmobilerobotsrequiresaninitialsetuptimeNodeconnectivityisanotherimportantaspectinUSARoperations.Nodesmovingawayfromthetransmissionrangeofthemaincontrollerareconsideredlost,unlesstheyuseinherentpositionawarenessprotocolstotraceroutebacktothemaincontroller.Lossofarobotcouldinducenancialdamageaswellasperformancedegradation.AnetworkofmobilerobotsformingawirelessLANusingIEEE802.11standardsdoesnotprovidefornodeconnectivity.However,mostoftheabovementionedissuescouldberesolvedbyforminganadhocnetworkofmobilerobots,insteadofformingawirelessLAN.Theideaofapplyinganadhocnetworkingtoateamofmobilerobotsisdiscussedin[37].[38]reportstheimplementationofalinkstateroutingprotocolinwirelessmodemsforapplicationinmobilerobot-basednetwork.Byforminganadhocnetworkofmobilerobots,theareaofcoverageisdrasticallyincreased,asitallowsmulti-hoprouting,androbotscanmaintaincommunicationwiththemaincontrollerthroughintermediatenodes.Alsothereisnosetuptimeorneedforanyxedinfrastructure(accesspoint).Forminganad17

PAGE 26

hocnetworkalsoresultsinreducednetworkcongestionasroutingremainsdistributed.Further,theuseofmulti-hoproutingprovidesmanyalternateroutesforcommunicationwiththemaincontroller.Thoughforminganadhocnetworkofmobilerobotsseemtobeaplausiblesolution,noneoftheexistingroutingadhocprotocolsaddresstheissueofnodeconnectivity.Forawirelessadhocnetworkwithahighlydynamicnetworktopology,theuseofoptimalpathroutingalgorithmswouldnotyieldgoodresultsastheroutesarehighlyvolatileandthenetworktopologyataninstantisnotthesameasthetopologyat.Shortestpathroutingisnotofmuchusewhenthelinksarehighlyunstable.Rather,routingdecisionneedstobebasedonsignalstrengthbetweennodes,durationofactivelinksandthebatterypowerlifeofthenodes.SSA[10]andABR[19]adhocprotocolsaddressestheseissues.Thoughtheseprotocolssuggestroutingdatathroughstablelinks(SSAusessignalstabilityanddurationoflinkasroutingmetrics,whileABRmakesroutingdecisionsbasedonlinkstabilityordurationforwhichalinkhasbeenactive),theydonotguaranteenodeconnectivity.Thisissueofproviding/guaranteeingnodeconnectivity,whilemaintainingtheotherbenetsfromimplementinganadhocnetworkisaddressedinthiswork.Insteadofcreatinganentirelynewroutingprotocoltoaddressthisissue,theproposedsolutiontakesadvantageofthetheresearchedworkandsuggestsacontrollayeralgorithmtoensurenodeconnectivity.Theideaofacontrollayer-basedapproach,choiceofroutinglayerprotocol,andthealgorithmareexplainedindetailinthenextchapter.2.4SimulatorandOtherToolsUsedTheproposedalgorithmisimplementedusingtheNetworkSimulator-ns(version2.26).nsisadiscreteeventsimulatordevelopedattheLawrenceBerkeleyNationalLaboratory(LBNL)[39]withextensionsfromtheMONARCHprojectatCarnegieMellon[26].Thecurrentimplementationofnssupportssimulationofreal-timewiredandwirelessnetworks.Itisafreelydistributed,opensourcetool,writteninC++,andusesOTclasacommandandcongurationinterface.nsprovidesacommonplatformforsimulating/comparingdifferentnetworkmodels.ItprovidessupportforFreeBSD,Linux,Solaris,WindowsandMacplatforms.Fordetailedinformationaboutthissimula-torthereaderisreferredto[39]and[26].nscurrentlyincludesextensionsforthemostcommonad18

PAGE 27

hocroutingprotocols,802.11MACdatalinklayer,andalsofree-spaceandtworaygroundmodelsforthephysicallayer.Inaddition,theNetworkAnimator(NAM)isusedforvisualizationanddebuggingpurposes.NAMisaTcl/TKbasedanimationtoolforviewingnetworksimulationtracesandrealworldpackettraces.Itsupportstopologylayout,packetlevelanimation,andvariousdatainspectiontools.TheNAMdevelopmenteffortwasanongoingcollaborationwiththeVirtualInterNetworkTestbed(VINT)project.Currently,itisbeingdevelopedattheInformationSciencesInstitute(ISI)attheUniversityofSouthernCalifornia(USC),aspartoftheSAMANandConserprojects.NAMisavisualizationtoolusedtoreadlargelesofdatasetsandbeextensibleenoughsothatitcouldbeusedtoviewdifferentnetworksituations.nsgeneratesaNAMreadabletraceleformattoviewananimationofthesimulatedexperiments.MoreinformationonNAMcanbefoundin[40].19

PAGE 28

CHAPTER3PROPOSEDALGORITHM-DESIGNANDDEVELOPMENTInthischapterthedesignofthealgorithmisdiscussedalongwithanelaborationoftheparam-etersandfunctionalspecications.Italsoexplainsthealgorithmwiththehelpofasamplemodel.3.1DesignConsiderationsandChoicesfortheNewAlgorithmInSection2.2someofthemostcommonandpopularroutingprotocolsforadhocnetworkswerereviewed[17][15][2][10][20]and[41].Thoughtheseaddressthebasicissuessurroundinganadhocnetwork,theydonotserveastherightmodelforrobotassistedsearchandrescueoperationswhichentailanaddeddimensionoffunctionality-NODECONNECTIVITY.Thedesignedprotocolmakessurethatthenodes(mobilerobots)donotgetlostandaconnectionwiththemaincontrolleralwaysexists.LetustaketheexampleofthescenariodepictedinFigure3.1.Thereisastationarymaincontroller,and6mobilenodes(robots).Robots1,2,3and4arewithinthetransmissionrangeofthemaincontroller(denotedbyacircle),whilerobots5and6areoutsideitstransmissionrange.Thisdoesn'tnecessarilymeanthatrobots5and6havelosttheircommunicationwithmaincontroller.Forexample,robot6canstillcommunicatewiththemaincontrollerthroughrobot2.Hererobot2actsasarouter,transmittingpacketstoandfromrobot6.Similarlyrobot5cantransmititspacketstothemaincontrollerthroughrobot3or4.Butasitcanbeseen,robot2ismovingoutsidethetransmissionrangeofthemaincontroller.Thisnotonlybreaksthecommunicationlinkbetweenrobot2andthemaincontroller,butalsothelinkbetweenrobot6andthemaincontroller,asrobot2wasservingasthelinkbetweenthesetwonodes.NoneoftheexistingroutingprotocolsaddressthisissueasdiscussedinSection1.3.Thenewdesignoftheroutingprotocolhastoensurethatinadditiontotheexistingdemandsofadhocnetworkssuchasnodemobility,linkstability,energyefciencyandreducedroutingover-head,ithadtoprovideforNODECONNECTIVITY.Basedontheinformationfromthisprotocol,20

PAGE 29

Figure3.1ScenarioConsistingofMainControllerand6MobileNodesanodeshouldstopitsmobilitywhenitsconnectionwiththemaincontrollerisindangerofbeinglost,whichisindicatedbyathresholdbasedonsignalstrengthandbatterpowerofthenodes.Al-ternatively,analgorithmoperatingbetweentheroutinganddatalinklayercouldbedesigned,asrepresentedinFigure3.2.Thisideaofhavinganinter-operatinglayertoprovidefornodecon-nectivity,allowstheuseofexistingMAClayerandroutinglayerprotocols.Inthiscase,theIEEE802.11standardprovidesforanefcientMAClayerprotocolandtheAdhocOn-demandDistanceVector(AODV)actsastheroutinglayerprotocol.AnexplanationfortheselectionofAODVastheroutinglayerprotocolisprovidedinSection3.3.3.2ApproachandDesignoftheAlgorithm3.2.1OutlineoftheProblemAsdetailedinSection1.3,nodeconnectivityiscriticaltotheperformanceinsearchandrescueoperationsusingrobots.Lossofrobotsaredetrimentaltobothcostandeffort,alike.Figure3.3providesaninsighttotheproblem.Toprovidefornodeconnectivity,thedesignofthealgorithmneedstoensureconstantcommunicationlinkbetweenthemobilenodesandthemaincontroller.Inadditiontomaintainingconstantlinkbetweentherobotsandthemaincontroller,thedesignofnewalgorithmshouldalsoensurethefollowing:21

PAGE 30

Figure3.2LayeredApproachforDesignoftheProtocolIncreaseinroutingoverheadshouldbeminimal,asincreasedoverheadleadstocongestioninthenetworkandwastageofbatterypowerAreaofcoverageshouldnotbeconnedandrestrictedforprovidingnodeconnectivityEnergyofthemobilenodesshouldbeconsidered,asanodemightbewellwithinthetrans-missionrangeofthemaincontroller,yethaveverylittleornobatterylife.Suchnodesshouldnotbeusedtorelaypacketsandshouldreturntothemaincontroller.Nodesatthresholdofaconnectionshouldstoptheirmobilityifbreakingawayfromthisconnectionwoulddisruptitscommunicationwiththemaincontroller.Forexample,robot2inFigure3.1isatthreshold,asitsdirectionofmovementisawayfromthemaincontroller.Itsmobilitywouldbreakitsconnectionwiththemaincontrolleraswellastheconnectionofrobot6withmaincontroller,sincerobot2connectsrobot6.Nodesshouldconstantlymonitorthesignalstrengthofthepacketsreceivedfromitsneigh-bors.22

PAGE 31

Figure3.3SampleScenarioof6NodeswiththeirTransmissionRangeIfafterstoppingitsmobility,anodeisabletoestablishalinktothemaincontrollerthroughanothernodethathasjustmovedwithinitstransmissionrange,itshouldstartitsmobilityandcontinuetomoveinthesamedirectionatthetimeofitstoppingitsmobility,unlessthereisatele-operatedchangeindirection.Thedesignedalgorithm,completelyimplementedinthecontrollayershowninFigure3.2,considersalltheabovelistedrequiredfunctionalities.Themainfunctionsandassumptionsmadeinthedesignoftheproposedalgorithmarenextsummarized:Therobotsandmaincontrolleraredeployedatthesamelocationatthestartofsearchandrescueoperation.Themaincontrollerremainsstationarythroughouttheentireoperationandtherobotsstartexploringthedisasterzone.Everynodebroadcasts1periodichellomessageswithitsenergyinformationinastandardpacketformat.ThehellomessagesareexchangedonceeveryHELLO INTERVAL2. 1Anodereceivingabroadcastmessagedoesnotforwardthepacket,thusthepacketisdeliveredonlytothenodesthatarewithinthetransmissionradiusofthesendernode2HELLOINTERVAListhetimeintervalbetweensuccessivehellomessagesbroadcastedbytheindividualnodes23

PAGE 32

Table3.1FormatofanInformationTableMaintainedatIndividualNodes NeighborID NormalizedEnergylevel NormalizedSignalPower Table3.2FormatofanUpdateTableMaintainedattheMainController NeighborID CompositeThreshold Eachnodethatreceivesthehellomessagesfromitsneighbors,calculatesthesignalstrengthatwhichthepacketwasreceived.Aninformationtable(seeTable3.1)ismaintainedateverynode,where,foreachnodeinthenetworkanentryismaintained.Thistablehaseldscorrespondingtotheneighboringnodeid,itsnormalizedenergylevel,andthenormalizedpowerlevelatwhichthehellopacketwasreceivedfromthatnode.Atperiodicintervalseverynodeinthenetwork,computesthecompositethreshold.foreachofitsneighboringnodes,basedonthevaluesintheinformationtable.Thisvalueofcompositethresholdisstoredalongwiththecorrespondingneighboridintheupdatetable(seeTable3.2).Eachnodethenforwardsitsupdatetabletothemaincontroller.Compositethresholdreferstoacombinedvalueofenergyandsignalstrengthcalculatedus-ingtheequationnrnrnr "! #%$ &'()'*(+-,n.nr+/#nr0/ "! 123()'"(+,wherethevalueofnr0nisdynamicallyestimated(Section3.2.4).Maincontrollerreceivesupdatetablesfromallnodesatperiodicintervalsandperformsalocalcomputationoneachofthereceivedtablestoseeifeverynodeinthenetworkhasaconnectiontothemaincontroller.Themaincontrollerloopsthrougheverynodeinthenetworktocheckifithasadirectcon-nectionwiththemaincontrollerorthroughanyotherintermediatenode,orifithassuchaconnection,butisatthelinkthreshold3.Themaincontrollersendsamessagetothesenodes,withitsmobilityagsettofalse(Section3.2.2). 3Linkthresholdistheminimumcompositethresholdforthenetwork,belowwhichthelinkisconsideredaweaklink.24

PAGE 33

Figure3.4FlowchartfortheLogicatLocalNodesNodesthatreceivemessagesfrommaincontrollerwithitsmobilityagsettofalse,stoptheirmobility(ifnotstoppedalready),andwaitforacertainperiodoftimetoseeifitreceivesamessagefromthemaincontrollerwithmobilityagsettotrue,inwhichcaseitcontinuesitsmobility.Ifthiswaittimeexpires,andthenodesdonotreceiveamessagefromthemaincontrollerwithmobilityagsettotrue,theybegintomovetowardsthemaincontrollerasapreemptivemeasure.Thedesignedcontrollayeralgorithm,usesacentralizedmechanismtomonitormobilityofthenodes,andallnodesusetheunderlyingdistributedadhocroutingprotocoltoexchangehellopacketsandsendthecomputedinformationtabletothemaincontroller.However,itisthemaincontrollerthatdecidesonmobilityofallthenodesbasedonthedataintheinformationtable,andhencethisapproachisclassiedascentralized.Butthisassumptioncouldbeeasilyjustied.InthecaseofUSARapplications,allthemobilerobotsneedtohaveaconstantcommunicationlinkwiththemaincontrolleratalltimes,thusmakingitacentralizedmodeofoperation.Thepossibilityofleavingthecontrolofdecidingonmobilitytothelocalnodes,asinadistributedmechanismisdiscussedin25

PAGE 34

Section3.4.Figure3.4explainsthelogicattheindividualnodesintheformofaowchart.Thealgorithmatthemaincontrollerisindependentofthelogicattheindividualmobilenodesandisdiscussedinthenextsection.3.2.2AlgorithmattheMainControllerThemaincontrollerreceivesupdatetablesfromeverynodeinthenetworkonceeveryUP-DATE INTERVAL4.Theinformationinthesetablesiscopiedintoadatastructuremaintainedatthemaincontroller.Asampleformatofthisdatastructureisgivenbelow:structBASESTATION OBJECT4intnumber of neighbors;boolmobility FLAG;doubleneighbor id[n];doublecomposite threshold[n];longintupdate seqno;5instances[n]Asitcanbeseenfromtheaboveformat,instanceoftheabovestructureiscreatedatthemaincontroller,whereisthenumberofnodesinthenetwork.Thusforeachnodeinthenetworkthemaincontrollermaintainsadatastructurecomprisingofnumberofneighborsofthatnode,mobilityagthatspeciesifthatnodeismobileornot,anarrayofneighborid's,compositethresholdvaluesandthesequencenumberoftheupdatepacketexpectedfromthatnode.UseofsequencenumbersforupdatepacketsissimilartotheimplementationofsequencenumbersforTCPpackets.Entriesinthisdatastructurearemodiedasandwhenupdatepacketsarereceived.ThemaincontrollerrunsanalgorithmonceeveryMONITOR INTERVAL5tocheckformobilitystatusofeverynodeinthenetwork.Thealgorithmloopsthroughtheinformationtableofeverynodeinthenetwork,andchecksforaconnectiontothemaincontrollerthatisabovethecompositethreshold.Ifnosuchlinkexists,thenitndstheneighborofthisnodethathasthemaximumvalueofcompositethresholdandlooksintheinformationtableoftheneighbornodeforaconnection(Thenodethathasthemaximumvalueofcompositethresholdamongallneighbors,shouldhaveitscompositethresholdgreaterthanthelinkthreshold).Thealgorithmstopsthemobilityofthemobile 4UPDATEINTERVAListhetimeintervalbetweensuccessiveupdatepacketssentbytheindividualnodes5MONITORINTERVAListhetimeintervalbetweensuccessiveloopsthroughthealgorithmatthemaincontrollertocheckformobilitystatusofthenodesinthenetwork26

PAGE 35

node,ifafterrecursivelyiteratingthroughtheinformationtableofallneighbors,itdoesnotndadirectormulti-hopconnectiontothemaincontroller.Thecompletealgorithmispresentedbelow.nodes visited=NULL;forallnodeinthenetworkforeachneighborofthisnodeif((neighbor==maincontroller)&&(composite threshold[neighbor]6link threshold))4continue mobility(node);5else4parent=ndparent(node);nodes visited=nodes visited+node;mobility=check connection(parent);if(mobility==TRUE)continue mobility(node);elsestop mobility(node);5check connection(node id)4foreachneighborofthisnodeif((neighbor==maincontroller)&&(composite threshold[neighbor]6link threshold))4returnTRUE;5else4parent=ndparent(node);nodes visited=nodes visited+node;mobility=check connection(parent);5if(mobility==TRUE)returnTRUE;elsereturnFALSE;5ndparent(node)4select(PARENT7neighborsofnode)suchthatcomposite threshold[PARENT]==max(composite threshold8neighbors)&&composite threshold[PARENT]69link threshold&&PARENT:nodes visitedif(nosuchnodeexists)returnNULL;27

PAGE 36

elsereturnPARENT;5stop mobility(node)4if(mobility FLAG[node]==TRUE)setmobility FLAG[node]=FALSE;send message(node);5start mobility(node)4if(mobility FLAG[node]==FALSE)setmobility FLAG[node]=TRUE;send message(node);5Ifthereisnodirectlinkbetweenanyofthenodesandthemaincontroller,thealgorithmre-cursivelyloopsthrougheverysafeneighbor6tondalinktothemaincontroller.Thisensuremaximumsafeareaofcoveragewithoutloosingcontactwiththemaincontroller.Bymakingsurethatallnodeshaveadirectorroutedlinktothemaincontroller,thealgorithmalsoensurethatthereisacommunicationlinkbetweenindividualnodes,i.e.atreestructureofthenetworkisalwaysmaintained.3.2.3LogicattheIndividualNodesThealgorithmatthemobilenodeshasbeenrepresentedbyaowchartinFigure3.4.Thelogicattheindividualnodesrelatestotriggeringdiscreteeventsandregularintervalsoftime.Thealgorithmworksasfollows.Eachnodebroadcastsperiodichellomessages,alongwiththelocalnodesbatterypower,onceeveryHELLOINTERVAL.Onreceivingahellomessagefromoneofitsneighbors,itestimatesthesignalpoweratwhichthepacketwasreceived,andstoresthisinformationalongwiththebatterypowerleveloftheneighborfromthehellomessage,inthelocalinformationtable. 6Safeneighborsareneighbornodesthathaveacomposite thresholdvaluegreaterthanlinkthreshold28

PAGE 37

OnceeveryUPDATEINTERVAL,nodescalculatecompositethresholdforallentriesintheinformationtableandgeneratetheUPDATETABLE.Thisupdatetableisthenforwardedtothemaincontroller.ALPHAvalueisestimatedattheindividualnodes(referSection3.2.4).foreachHELLO INTERVAL4send hello packets();5foreachUPDATE INTERVAL48;7ofinformation table)estimate composite threshold();generate update table();forward update table to basestation();estimate ALPHA();5foreachpacket received4if(type==hello)4neighbor=packet.node id;power=packet.recd power();energy=packet.energy;insertdataininformation table;5elseif(type==basestation message)4if(mobility FLAG==STOP&&nodeismobile)stop node();elseif(mobility FLAG==START&&nodeisstopped)start node();55Functioncallsend hello packets()sendsabroadcasthellomessagetoallneighbornodes.Asthenameimpliesfunctioncallgenerate update table()estimatesthecompositethresholdvalueforneighbornodesintheinformationtableandstoresitintheupdatetable,whichisthenforwardedtothebasestation.Functionestimate ALPHAcalculatesthevalueofALPHAasexplainedinthenextsection.29

PAGE 38

3.2.4AlphaEstimationCompositethresholdreferstoacombinedvalueofenergyandsignalstrengthcalculatedfromthevaluesintheinformationtableofamobilenode.Itisameasureofqualityofthelinkwiththecorrespondingneighbor.Itiscalculatedusingtheequation:composite threshold=(ALPHA*normalized energylevel)+(1-ALPHA)*normalized powerlevelwhereALPHAisdynamicallyestimated.Batterypowerandsignalpowervaluesstoredintheinformationtableofthenodesneedtobenormalizedtothesamescalebeforebeingusedintheabovegivenequation(Batterypowerismeasuredinjoulesandisusuallyapositivenumberintherange(0-maximumpowerofthebattery),whilesignalstrengthismeasuredindecibelsisoftheorderof=< >?,whererangeofxdependsonthewirelessinterfaceandotherchannelparameters.Theyareconvertedtoratios,denotedasafractionofthemaximumenergyandmaximumsignalstrengthpossible.%@nr *! #$ &'()A*'9BDC*BFEFGHAIKJLCBDMNGOQPRIKEFMNCGC*IKSB TVUWMTVXQTBFC*BDEKGHAYIKZDZFM[P]\0BrJ"IFE3^_OQMNZrC*IKS`B#nr0/ "! 123()A*'9ZDMNGCU\ZD^_EKBDCG^_OLJQIKE3^_OQMNZr\0MNC*a TVUWMT#X"TZDMNGCU\ZD^bEFBDCG^bOcYIKZDZDMNPR\0BrJ"IKEUCH\0MNCa'JQEKIT^bOQM[ZrCIKSBThemaximumpossiblevaluesfornodeenergyandsignalstrengthofanylinkarepre-determinedandremainconstantfortheentiredurationofthesimulation.Themaximumpossiblevaluefortheenergylevelwouldbetheenergyvalueofthebatterieswhenfullycharged.Andthemaximumpossiblevalueforsignalstrengthforanylinkwouldbethesignalpowerlevelcalculatedbetweentwonodesthatareveryclosetoeachotherassumingidealtransmissionconditions.Astaticvalueforalphaprovidesaconstantweightfactorfortheenergyandpowerlevelsofnodesandlinks,irrespectiveofthecurrentnetworktopology.Forexample,atthestartofthesimulation,allnodeswouldhaveenergyvaluesclosetothemaximum.Inthiscase,itwouldbebettertohaveanALPHAvalueofd0.2,thusgivingmoreweighttotheneighboringlinkpowerlevel.Similarly,whenallthenodesareincloseproximitytothemaincontroller,thesignalstrengthofthereceivedpacketswouldbeclosetomaximum.AnALPHAvalueofd0.8wouldbebetter,asthecalculatedcompositethresholdwillbemorebiasedtowardsenergyvalues.30

PAGE 39

AsimplemethodtodynamicallyestimateALPHAvalueisadopted.TheprocedureislistedasanalgorithmandisrunonceeveryUPDATE INTERVAL.Itistoberememberedthatonceeveryupdateinterval,eachnodegeneratestheupdatetableandsendsittothemaincontroller.selectbest nodefromupdatetablewhere:best node=nodewithmaximumcomposite thresholdvalue;selectbest energylevelandbest powerlevelfrominformationtablewhere:best energylevel=energylevel[best node];best powerlevel=powerlevel[best node];selectlocal maxenergyandlocal maxpowerfrominformationtablewhere:local maxenergy=maximumenergyvalueintheinformationtable;local maxpower=maximumpowervalueintheinformationtable;if(local maxenergy6best energylevel)ALPHA=ALPHA+(1-best energylevel/local maxenergy)if(local maxpower6best powerlevel)ALPHA=ALPHA-(1-best powerlevel/local maxpower)Itistobenotedthattheenergyandpowerlevelvaluesofthenodewithmaximumcompositethresholdneednotnecessarilybethemaximumenergyandpowerlevelvalues.Thus,ifthemax-imumenergyvalueintheinformationtableisgreaterthantheenergyvalueofthenodewiththemaximumcompositethreshold,ALPHAvalueisincreasedbythefractionofthedifferencebetweenthesetwovalues.AnincreaseinALPHAvaluewouldresultinhigherweightagefortheenergyval-uesinthecomputationofcompositethreshold.Similarly,ifthemaximumvalueforpowerlevelintheinformationtableisgreaterthanthepowerlevelofthenodewithmaximumcompositethreshold,ALPHAvalueisdecreasedbythefractionofdifferencebetweenthesetwovalues.AlphaestimationandthecompletealgorithmisexplainedusinganexampleinSection3.5.3.3ChoiceofAODVastheRoutingLayerProtocolTheabovediscussedalgorithmprovidesfornodeconnectivity,butitdoesnotprovideforrout-inginformationbetweennodes.Theforwardingofupdate tablestomaincontrolleronceevery31

PAGE 40

UPDATE INTERVALreliesontheunderlyingroutingprotocolforthetransmission.Thefunction-alitiesattheMAClayerrequiredbyanadhocnetworkcontrollayer,areverysimilartothatrequiredbyawirelessnetwork.TheIEEE802.11standardforwirelessLAN'swaschosenastheMediumAccessControl(MAC)layerprotocol,whiletheroutinglayerprotocolhadtobechosenfromoneoftheseveralprotocolsdesignedspecicallyforadhocnetworks.Thefollowingconsiderationsweremadeinchoosingtheroutingprotocol.Theselectedroutingprotocolshouldhaveaveryminimalroutingoverhead.Thissuggestselectionofon-demandroutingprotocols,sincetheyhavemuchloweroverheadwhencom-paredtoproactiveprotocols[34][27].Scalabilityisanimportantfactorwhenmakingthechoiceofasuitableroutingprotocol.Increaseinthenumberofnodesshouldnotaffecttheperformanceofthesystem.Performanceoftheselectedprotocolshouldbewellanalyzedandcomparedwithotherstan-dardprotocols,asthisgivesusabaseforcomparingtheperformanceoftheproposedalgo-rithm.Also,theroutingprotocolshouldhavegoodperformanceinscenariosthatinvolveconstanttopologychange,andrandomnodemobility,asthesecharacterizethemobilityofrobotsinUSARoperations.BasedontheabovementionedfactorstheAdhocOn-demandDistanceVector(AODV)waschosenastheroutingprotocol[17].TheAdhocOn-DemandDistanceVector(AODV)algorithmenablesdynamic,self-starting,multi-hoproutingbetweenparticipatingmobilenodeswishingtoestablishandmaintainanadhocnetwork.Itusesperiodicbroadcastofhellomessagestomaintainroutesofestablishedlinks.Energyinformationisappendedtothehellomessages,andthepowerlevelatwhichthispacketisreceivedbytheneighboringnodesisestimatedasanindicatorforlinkstrength.Thustheimplementationoftheproposedalgorithmcausesonlyasmallincrementalincreaseinroutingoverhead,andisstudiedinSection4.3Alsofromperformanceanalysisofroutingprotocolsforadhocnetworksin[26],itcanbeseenthatAODVperformswellinscenariosinvolvingrandomnodemobilityandhigherratesofmovement.Also,atlowerratesofmobilityAODVandDSRperformbetterthantheotherprotocols.32

PAGE 41

Figure3.5ModelTopologywithNodesinaLinearChainAlso,theperformanceofAODVhasbeenwellanalyzedandcomparedwithotherstandardprotocols[26,27,28,29,33,35]forvariousscenariosandisalsoimplementedinns-2.26.AODVprovidestherightplatformforbuildingthecontrollayeralgorithms,withoutincurringanymajorroutingoverhead,anditsimplementationinnsgivesagoodbasetocompareandevaluateitsperformanceagainstotherstandardroutingprotocols.3.4VariationsPossibleintheProposedAlgorithmCertaindesignchoicesneedalittlebitmoreexplanation.Insteadofcomputingthecompositethresholdateverynode,andsendingthisinformationtothemaincontrollerasanupdatetable,eachnodecouldsenditsinformationtable(consistingofneighborid,energyleveloftheneighbor,andreceivedsignalpowerlevel),andtheALPHAvalue,andleavethecomputationtothemaincontroller.Thiscouldhelpinreducingthecomputationtimeattheindividualnodes,whileitwouldincreasetheroutingoverhead.AscanbeseenfromTable3.1,informationtablehasanextradoubleeldwhencomparedtoupdatetable(Table3.2).Thus,dependinguponthefrequencyofUPDATEINTERVAL,andthenumberofnodesinthenetwork,therewouldbesignicantincreaseintheoverheadduetotransmittingtheinformationtable.Ifisthenumberofnodesinthenetwork,thenforeveryegfhjikml n.okglmprqsiuttherewillbeanextravwyxbytesofoverhead,wherevbytesisthedefaultsizefora! z|{/(variable.33

PAGE 42

Table3.3InformationTableatNode1AlongwithCompositeThresholdValues NeighborID NormalizedSignalStrength NormalizedEnergy CompositeThreshold 2 0.51 0.92 0.4692 3 0.57 0.81 0.4617 4 0.63 0.73 0.289 5 0.56 0.82 0.4592 BS 0.45 0.91 0.4095 Insteadofhavingthemaincontrollervalidatethemobilityofnodesbyiteratingthroughtheupdatetablesreceivedfromeachnode,thenetworkcouldbeoodedwithupdatetablesfromeachnode,sothateverynodeinthenetworkhasacopyoftheupdatetableoftheothernodes.Bydoingthis,thecontrolofdecidingonmobilityislefttotheindividualnodesratherthanthemaincontroller.Thiswouldcorrespondtothedistributedimplementationoftheproposedalgorithm.Thoughthismightseemplausibleduetothefact,thateachnodecouldstopitsmobilityasandwhenitdetectsaweaklinktothemaincontroller,thisapproachhascertaininherentdrawbacks.Considertheexampleofanetworkwithnodesandamaincontroller,locatedinalinearchain(Figure3.5),suchthateachnodeisattheendpointofthetransmissionrangeofitsneighboringnode.Insuchatopology,whenusingoodingtechniquestobroadcastinformationtabletoallnodesinthenetwork,thereare}~,=+Rx€packetsbroadcasted,whileifeachnodetransmitsitsupdatetabletothemaincontroller,thereisatotalof CU‚yƒnpacketstransmittedforeveryUPDATE INTERVAL.Thesignicantincreaseinnumberofpacketsexchangedandtheconsequentriseofcongestioninthenetworkwhenooding,isonereasonwhyeachnodesendsitsupdatetabletothemaincontrollerandthemaincontrollerdecidesonmobilityofthenodes.Moreover,keepingthecomputationalcomplexityatthemobilenodestoaminimum,helpsincreasebatterylife.ALPHAvaluescouldbekeptstatic,ratherthanhavingdynamicallyestimatedvalues(Sec-tion3.2.4).ButwithstaticALPHAvaluesthecalculatedcompositethresholddoesn'treectthecurrentnetworkbehavior.Forexample,letsassumeanetworkof5nodesandamaincontroller,withstaticALPHAvalueof0.5ateachnode.LetTable3.3bethevaluesintheinformationtableatnode1,alongwiththeircompositethresholdvalues.34

PAGE 43

Ascanbeseenfromtheinformationtablefornode1,node4hasthebestlinkwiththisnode(normalizedsignalstrengthof0.63),whiletheneighbornodewiththebestenergyisnode2(normalizedenergyvalueof0.92).Thetablealsoshowsthecompositethresholdfortheneighborsofnode1,calculatedwithiut#fg„…i9†<=‡bˆ.Thisresultsinnode2havingthemaximumvalueforcompositethreshold,andwouldbechosenastheimmediateparent7forthisnode.While,node4havingabettersignalstrength,shouldhavebeenmadetheparentneighborforthisnode.AniutVf„…ivalueof0.1wouldbiasthecalculationofcompositethresholdtothenodewithbetterpowerlevel,whileaniwtVf„…ivalueof0.9wouldbiasthecalculationtowardstheneighbornodewithbetterenergylevel.Thus,inordertobalancethebiasingfactor,ALPHAvaluesaredynamicallyestimatedasexplainedinSection3.2.4.Eachnodehasitsowniut#f„‰ivalueandisestimatedbasedonthepreviousiwtVf„…ivalue,andthecurrentdataintheinformationtable.Thenextchapteranalyzestheeffectofdynamicestimationvs.staticALPHAvaluesontheperformanceofthealgorithm.3.5SampleModelBasedExplanationoftheAlgorithmThissectionprovidesanillustrationbasedexplanationofthealgorithm.Letusconsidertheex-ampleofaŠ-<=<‹rˆ=<-
PAGE 44

Figure3.6SampleModel:StartingTopologyTable3.4InformationTableatNode1and2attimet=1.0s NODE1 NODE2 NeighborID Energylevel SignalPower NeighborID Energylevel SignalPower 2 0.95 0.9 1 0.95 0.87 BS 0.96 0.92 BS 0.93 0.9 rangeofthemaincontrollerind50seconds.Figure3.6presentsthenetworktopologyatŒ9.Nodes1and2areclosetothemaincontrollerandtheirdirectionofmovementsareindicated.Hellomessagesareexchangedbetweenthenodes,anddataintheinformationtablegetsupdatedforeachreceivedhellomessage.Table3.48showsthedatastoredininformationtableattimet=1.ItistobenotedthatenergyandsignalpowervaluesarenormalizedasexplainedinSection3.2.4,whileupdatingthecontentsofinformationtable.Attimet=1.2sthefunctioncalltosendupdatetableisevoked,thatcalculatescompositethreshold,estimatesALPHAvalue,andsendstheupdatetabletomaincontroller.UsingtheequationforcompositethresholdgiveninSection3.2.4,theupdatetablesaregeneratedattheindividualnodes(Table3.5). 8Thevaluespresentedinthesetablesarenotsimulated,norrepresentactualvaluestakenfromexperiments,butwereassignedtofacilitateeasyunderstandingofthealgorithmTable3.5UpdateTableatNodes1and2attimet=1.2s NODE1 NODE2 NeighborID CompositeThreshold NeighborID CompositeThreshold 2 0.925 1 0.91 BS 0.94 BS 0.915 36

PAGE 45

Tofurtherillustratethecalculationofcompositethreshold,letustaketheexampleofnode1.Node2isaneighborofnode1withenergylevelof0.95andsignalpowerof0.9.Hencethecompositethresholdforthisneighborofnode1wouldbe<=‡bˆVŽ<-‡_=ˆ-+<=‡bˆVŽ<=‡b=+whichis<-‡_-‘=ˆ.Aftercalculationofcompositethreshold,eachnodeestimatesitsALPHAvalue.Atnode1,theneighborthathasthemaximumvalueforcompositethresholdisthemaincon-troller(fromupdatetable).Thusthebest energylevelandbest powerlevel(Section3.2.4)corre-spondtotheenergyandsignalpowervaluesofthisnode,i.e,0.96and0.92respectively.Also,fromtheinformationtable,itcanbeseenthatthelocal maxenergyandlocal maxpowercorrespondtothevaluesofmaincontroller.ThisimpliesthatthecurrentvalueofALPHAisproperlybiasedbetweenenergyandsignalstrength,andthustheALPHAvalueforthisnoderemainsthesame.Againamongtheneighborsofnode2,themaincontrollerhasthemaximumvalueforcom-positethreshold.Variablesbest energylevelandbest powerlevelcorrespondtovalues<-‡_-’and<=‡_respectively.Butthevaluesforlocal maxenergyandlocal maxpowercorrespondto0.95and0.9respectively(Table3.4).Despitehavingagreaterenergyvalue,node1doesn'thavethemaximumcompositethresholdvalue.Thusthealphavalueisbiasedtogivemoreweighttoenergyvalueinthecalculationofcompositethreshold.Sincemax(normalizedenergy)6normalizedenergyofnodewithmax(compositethreshold),iut#f„‰i9iut#f„‰i“€,“{*”Œ #$ &'()'"(•-}%–nr n"?—$ &'+9<-‡_ˆ€“€,“<-‡_=’-•=<-‡_-ˆ=+9<-‡_ˆ-‘=ALPHAvaluefornode1remainsat0.5,whilenode2nowhasanALPHAvalueof0.521,increasedweightforenergyvalueofneighbors.AfterALPHAestimation,eachnodesendsitsupdatetabletothemaincontroller.Onreceivingthesetablesfromtheindividualnodes,themaincontrollerupdatesitsdatastructuretoreectthecurrentnetworktopology.ThusatŒ™9-‡_‘-”,themaincontrollerhasadatastructuresimilartoTable3.6AteveryMONITOR INTERVAL(=1.5inthiscase)themaincontrollerrunsitslocalalgorithmtocheckforlinksfromallnodestoitself.FromTable3.6,itcanbeseenthatnodes1and2,bothhaveadirectconnectionwiththemaincontroller,andisabovethelinkthreshold(=0.32).Sobothnodescanremainmobile,andthemobility agforthesenodesissettotrue.37

PAGE 46

Table3.6DataStructureatMainControllerattimet=1.2s NodeID NeighborID CompositeThreshold Mobile 1 2 0.925 Yes BS 0.94 2 1 0.91 Yes BS 0.915 Figure3.7SampleModel:NetworkTopologyattimet=50.0sFigure3.5representsthenetworktopologyatŒF]Œj9šˆ-<=‡b<=”.Table3.7showstheinforma-tiontableatnodes1and2,updatedbasedonthehellomessagesreceivedfromtheirneighborsatŒD›Œœ9ˆ-<=‡b<=”.Againatt=50.4s,thefunctioncalltoupdatetableisevoked.SinceALPHAvaluesaredynam-icallyestimatedeveryUPDATEINTERVAL,usingthevalueof0.5and0.521fornodes1and2,estimatedatt=1.2swouldnotbeappropriate.ThusALPHAisassumedtobe0.2att=50.0s.ThegeneratedupdatetableisshowninTable3.8.ALPHAvaluesareestimatedinthesamewayasinthepreviouscaseandisnotshownhere.Att=50.4s,thenodesforwardtheirupdatetablestothemaincontroller,wherethelocaldatastructureisupdatedbasedonthedatareceivedfromthenodesinthenetwork.TheupdateddatastructureatthemaincontrollerisshowninTable3.9Table3.7InformationTableatNode1and2attimet=50.0s NODE1 NODE2 NeighborID Energylevel SignalPower NeighborID Energylevel SignalPower 2 0.75 0.85 1 0.72 0.8 BS 0.8 0.2 BS 0.8 0.1 38

PAGE 47

Table3.8UpdateTableatNodes1and2attimet=50.4s NODE1 NODE2 NeighborID CompositeThreshold NeighborID CompositeThreshold 2 0.83 1 0.784 BS 0.32 BS 0.24 Table3.9DataStructureatMainControllerattimet=50.4s NodeID NeighborID CompositeThreshold Mobile 1 2 0.83 Yes BS 0.32 2 1 0.784 Yes BS 0.24 Themaincontrollerexecutesitslocalalgorithmatt=51s(MONITOR INTERVAL=1.5s).AscanbeseenfromTable3.9node1hasadirectionwiththemaincontroller,butthecompositethresholdvalueisatthelinkthreshold.So,thealgorithmloopsthroughtheupdatetablesofotherneighborsofthisnode,whichhaveacompositethresholdvaluegreaterthanlinkthreshold,foraconnectiontothemaincontroller.Theonlyotherneighborfornode1isnode2,anditscompositethresholdvalueisgreaterthanlinkthreshold.Hencethealgorithmcheckstheneighborslistofnode2forastronglinktothemaincontroller.Butnode2hasanevenweakerconnectiontomaincontroller(compositethreshold=0.24),andhencethealgorithmstopsthemobilityofnode1sinceitsonlyconnectiontothemaincontrollerisatthreshold.Themobilityagofnode1issettofalse,andamessageissenttothenodewiththemobilityag.Similarly,thealgorithmlooksforaconnectionfromnode2tothemaincontroller.Thedirectlinkfromnode2toBSisbelowthelinkthreshold(=0.24),andthealgorithmchecksforaconnectiontomaincontrollerthroughotherneighborsofthisnodewhichhavethresholdvaluesgreaterthanlinkthreshold.Neighbornode1hasaconnectionwiththemaincontroller,whichisatlinkthreshold.Butsinceitsmobilityhasalreadybeenstopped,andisnotindangerofbreakingawayfromthemaincontroller,thealgorithmsetsnode1astheparentnodeof2,andnode2continuestohaveitsmobilityagsettotrue.Figure3.8presentsthenetworktopologyatt=100s.Itistoberememberedthatnode1hasitsmobilityagsettofalse,andnode2hasitsmobilitybasedonnode1,i.e,node2hasalinktothemaincontrollerthroughnode1.ALPHAvalueisassumedtobe0.3att=100s.39

PAGE 48

Figure3.8SampleModel:NetworkTopologyattimet=100.0sTable3.10InformationTableatNode1and2attimet=100s NODE1 NODE2 NeighborID Energylevel SignalPower NeighborID Energylevel SignalPower 2 0.5 0.4 1 0.3 0.328 BS 0.6 0.2 Basedonthehellomessagesexchangedbetweenthenodesatt=100s,theinformationtableofeachnodegetsupdatedandisshowninTable3.10.Itistobenotedthatthereisnoentryformaincontrollerinthetableofnode2,andthisisduetothefactthatnode2hasmovedwellbeyondthetransmissionrangeofmaincontrollerandthehellomessagesbroadcastedbythemaincontrollerarenotreceivedatnode2.Att=100.4s,theupdatetablefunctionisevokedbyeverynode,whichgeneratestheupdatetable(Table3.11)forthatnode,estimatesALPHAvalueandsendsouttheupdatetabletomaincontroller.ThecontentsofthedatastructureatthemaincontrollergetupdatedonreceivingthesetablesformthemobilenodesandisshowninTable3.12.Node1hasitsmobilityagsettofalseandthealgorithmatthemaincontrollerisnotabletondanybetterlinktothemaincontroller(thisoccursifanodewithastronglinktothemaincontrollermoveswithinthetransmissionrangeofnode1).Butnode2,nowhasonlyoneneighbor,node1,Table3.11UpdateTableatNodes1and2attimet=100.8s NODE1 NODE2 NeighborID CompositeThreshold NeighborID CompositeThreshold 2 0.28 1 0.32 BS 0.32 40

PAGE 49

Table3.12DataStructureatMainControllerattimet=100.8s NodeID NeighborID CompositeThreshold Mobile 1 2 0.28 No BS 0.32 2 1 0.32 Yes andthislinkhasacompositethresholdvalueequaltolinkthreshold.Node2canstillcommunicatewiththemaincontrollerthroughnode1,butboththelinksareatthresholdlimits.Thealgorithmatthemaincontrolleriteratesthroughthedatacollectedfromupdatetablesandchecksformobilityofthenodes.Atimerisattachedtoeverynodetokeeptrackofthedurationforwhichithasbeenstopped.Basedonthistimevalue,themaincontrollercanissuecallbackfunctionstothenodesrequestingthemtomovetowardsthebase.Theexampletopologypresentedinthissectionprovidesthebasicideaofthealgorithmde-tailedintheprevioussections.Thenextchapterdiscussestheimplementationparameters,actualvaluesusedforsimulations,changestothebasealgorithmdoneduringimplementation,simulationscenarioandnallyanalyzestheresultsofsimulationandefciencyoftheproposedalgorithm.41

PAGE 50

CHAPTER4IMPLEMENTATIONANDEXPERIMENTALRESULTSThischapterdescribestheimplementationoftheproposedcontrollayeralgorithminns-2.26,explainsthesimulationsetup,andanalyzestheperformanceofproposedalgorithmbasedonlinkduration,overheadandthroughput.4.1Implementationinns-2.26ns(networksimulator)isacomprehensivetooltosimulatewiredandwirelessnetworks.Wire-lessextensionsfromtheMONARCHprojectatCarnegieMellon[26]providesfornodemobility,realisticphysicallayerincludingaradiopropagationmodelsupportingpropagationdelay,captureeffectsandcarriersense,radionetworkinterfaceswithpropertiessuchastransmissionpowerandantennagainandimplementsIEEE802.11MediumAccessControl(MAC)protocolusingtheDis-tributedCoordinationFunction(DCF).4.1.1PhysicalLayerThewirelessinterfaceworkslikethe914MHzLucentWaveLANDirect-SequenceSpread-Spectrum(DSSS)radiointerface[42].WaveLANismodelledasashared-mediaradiowithanominalbitrateof2Mb/s,andanominalradiorangeof250m.Eachmobilenodeusesanomni-directionalantennawithunitygain.Theantennaispositionedtobeatthecenterofthemobilenodeand1.5metersabovetheground(X =0,Y =0,Z =1.5,Gr=1.0,Gt=1.0,whereGrandGtarethegainofthetransmitterandreceiverantennasrespectively).TheSharedMediainterfaceisinitializedwiththefollowingparameterstomakeitworklikethe914MHzLucentWaveLANDSSSradiointerface.CarrierSenseThreshold(CSThresh )=-‡_ˆ-ˆ=-*>ƒƒWattsor-78dBm42

PAGE 51

ReceivingThreshold(RXThresh )=’-‡_-ˆ=‘->ƒ}or-64dBmBandwidth=2MbpsFrequency=914MHzThesignalpropagationmodelcombinesbothafree-spacepropagationmodelandatwo-raygroundreectionmodel.Thefree-spacemodelisusedwhenthetransmitteriswithinthecross-overdistanceofthereceiver,andthetwo-raygroundmodelisusedotherwise.Thefree-spacemodelusesFriisequation(Equation4.1)toestimatethereceivedsignalpower,wherethesignalattenuatesas1/!"x.fAŸ9 s¡€¢£¡Ÿ‰f¢ }¤-#+x!x(4.1)where, isthewavelength,fAŸisthereceivedsignalpowerinWatts(ordBm),¡¦¢isgainofthetransmitter'santenna,¡Ÿisthegainofthereceiver'santenna,f¢isthetransmittedsignalpowerinWatts(ordBm)anddisthedistancebetweenthereceiverandtransmitterantennasmeasuredinme-ters.Inthetwo-raygroundreectionmodel[43],thereceivedsignalpowerisinverselyproportionalto!"§orattenuatesas-•=!"§,andisestimatedusingequation4.2.fAŸ9¡¦¢¨¡Ÿ‰f¢„¢x„rŸx+ !§t(4.2)where„¢and„Ÿaretheheightsofthetransmitterandreceiverantennasrespectively.Thecross-overdistanceiscalculatedusingequation4.3.3%”*”*)'r!Q/”Œn"#–w9¤€„¢„Ÿ  (4.3)Eachmobilenodehasavelocityandpositioninformationassociatedtoit.Thepositionofanodeiscalculatedasafunctionoftimeandisusedbythepropagationmodeltoestimatethereceivedsignalpower.Whenapacketisreceivedatamobilenode,itsreceivedpowerisestimatedbythephysicallayerinterface.Thepacketisdiscardedifthereceivedpowerisbelowthecarriersensethreshold,andismarkedaserrorbeforebeingpassedtotheMAClayer,ifthereceivedsignalpowerisbetweenthecarriersensethresholdandreceivethreshold.Otherwise,thepacketishandedtotheMAClayerasagoodpacket.43

PAGE 52

4.1.2DataLinkLayerThelinklayerofns-2.26implementsthecompleteIEEE802.11standard[8]MediumAccessControl(MAC)protocoltomodelthecontentionofnodesforthewirelessmedium.Themainfunctions/featuresofthislayerarelistedbelow:TheIEEE802.11MACstandardimplementsDistributedCoordinationFunction(DCF),tousebothvirtualandphysicalsensingmechanismstoreducetheprobabilityofcollisionsduetothehiddenterminalproblem.ThetransmissionofanyunicastpacketisprecededbyexchangeofRequest-to-Send/Clear-to-Send(RTS/CTS)thatreservesthewirelesschannelforthetransmissionofadatapacket.Thesenderreceivesanacknowledgementfromthereceiver,foreverycorrectlyreceivedpacket,untilwhichtimeitretransmitsthepacketatspecicinterval.BroadcastpacketsarenotprecededbyRTS/CTSexchange,butaresentonlywhenthevirtualandphysicalcarriersenseindicatethatthemediumisclearforcommunication.Broadcastisunreliablein802.11becausenoacknowledgementissent.Itimplementsabinaryback-offmechanismtoreducethelikelihoodofcollisions.AdetailedstudyoftheIEEE802.11standardanditsimplementationinns-2.26,canbefoundin[8]and[39]respectively.4.1.3ControlLayerExtensionswereaddedtothesimulatortoimplementthecontrollayeralgorithm.Separatecon-trollayermodulesweredevelopedforthemaincontrollerandthemobilenodes.ALPHAestimationiskeptlocal,andeachnodemaintainsalocalcopyofALPHAcalculatedonceeveryUPDATEIN-TERVAL,basedonthedatainitsinformationtable.Thecontrollayeralgorithm(Figure3.2)reliesontheunderlyingMAClayerfortheexchangeofbroadcasthellopacketsandtheestimationofthesignalpower,andontheroutinglayerforthetransmissionofupdatetablestothemaincontroller.WeassumeaHELLOINTERVALandUPDATEINTERVALof1second,andaMONITORIN-TERVALof1.5seconds.Atimevariablefromastandardrandomdistributionisaddedtoboth44

PAGE 53

HELLOINTERVALandUPDATEINTERVALtoreducethecontentionforthemedium.Hellomessagesareexchangedonceeverysecond,andthemobilenodesgeneratetheupdatetableandforwardittothebasestationatthesameinterval.Thebasestation,runsitsalgorithmtocheckforsafemobilityofthenodesonceeveryMONITORINTERVAL,i.e.1.5seconds.Thereceivingthreshold(RXThresh )forthesharedmediaradiointerfaceis’=‡b=ˆ-‘=>ƒ},whichimpliesthatpacketsreceivedwithsignalpowerlessthanRXThresh arediscardedduetolowpowerleveloraremarkedaspacketsinerror.BasedonthisvalueofRXThresh weassignalinkthresholdvalueof’-‡_=ˆ-‘=>r,i.e,packetsreceivedthroughalinkatasignalpowerlessthanlinkthresholdareconsideredweaklinks.Suchlinksarenotcountedfor,whenthecontrollayeralgorithmchecksforconnectiontothemaincontroller.Themaximumsignalpoweratwhichapacketcanbereceivedwouldbewhentwonodesareataveryclosedistance,andcanbecalculatedusingFriisequa-tion(Equation4.1).Withatransmissionpowerof0.2818Watts(f¢=0.2818Wattsfora100mtransmissionrangeassumingAT&T'sWaveLANPCMCIAcard),distancebetweennodesas0.1m,unittransmitterandreceivergain,andusinga914MhzWaveLANPCMCIAcard,themaximumreceivedsignalpowercouldbe0.0585Watts(substitutingvaluesinEquation4.1).Thusforalinktobeconsideredastronglink,packetsneedtobereceivedatasignalpowerlevelintherangefrom0.0585Wattsto’=‡b=ˆ-‘=>rWatts.Thismaximumandminimumvalueforsignalpowerisusedinestimatingthenormalizedsignalpower.Thecompositethresholdisestimatedbasedonthenormalizedsignalpower,andnormalizedbatterypower.Initiallyallnodeshaveabatterypowerof600Joules.NormalizedbatterypowerattimeiscalculatedasaratiooftheEKBTVUMNC*MNCGPU^_^bBDEKHYIKBDEIKJ;C*BDMNGOQPRIKE3C*IKS`BU^Q^bMTBr T#UWMTVXQTPU^_^bBDEKHYIKBDEc*%.Toestimatethenormalizedsignalpower,themaximumandminimumvaluesforsignalpowertoestablishstablelinksaredivided(0.03652Wand<=‡_’-=ˆ-‘-*>ƒ}W)into20divisionsonalogscale.Eachdivisionhasavalueof0.05,thusbasedonthereceivedsignalpower,thenormalizedsignalpowercanrangefrom0to1.Itistobenotedthatthederivedvalueof0.0585Wformaximumpossiblesignalpowerisnotused,andinsteadthemaximumpossiblesignalpowerOF0.03652Wisutilized.Normalizedsignalpowerisassignedavalueof1,ifthereceivedsignalpowerisabove0.03652W.45

PAGE 54

4.1.4RoutingLayerTheadhocroutingprotocolsDSR,DSDV,AODVandTORAhavebeenimplementedinns,withAODVbeingtheroutingprotocolchosenforthiswork(seeSection3.2.4).AselaboratedinSection3.3,wechooseAODVastheroutingprotocol.TheimplementationofAODVinnsfollowsthestan-darddescribedin[17].Howevertheimplementationdiffersslightlyfromthespecicationsinthemethodofdetectinglinkbreakages.Insteadofusinghellomessagestodetectlinklayerbreakages,theauthorsin[26]analyzetheeffectofusingphysicallayermethods.Thissmallchangemodiesthebasebehavioroftheprotocolbecauseiteliminatestheoverheadofthehellomessages.Thoughthissavestheoverheadduetoperiodichellomessages,itincreasestheconvergencetimeandmodi-esthebasicbehavioroftheprotocol.However,sincetheproposedcontrollayeralgorithmmakesuseofhellomessagestopropagateenergyandpowerinformation,theoriginalimplementationofAODVisutilizedthatusesperiodichellomessagestodetectlinklayerbreakages.4.2SimulationScenarioTheevaluationoftheproposedalgorithmisbasedonthesimulationof9wirelessnodes(8mo-bilenodesorrobotsand1stationarynodeormaincontroller)forminganadhocnetwork,movingoveranareaof600mx600matspacefor600secondsofsimulatedtime.TheseareveryrealisticnumberforUSARapplications.Eachnodehas600Jofenergyatthestartofsimulationwithadrainof0.5Wforeverypacketreceivedandadrainof0.4Wforeverypacketsent.ThephysicalcharacteristicsofthemobilenodesnetworkinterfaceismodelledtoapproximatetheLucentWave-LANDSSSradiointerface.Thephysicalchannelreplicatesthetwo-raygroundpropagationmodel.Throughput,linkstability,andoverheadareusedasperformancemetrics.Intuitively,thecontrollayeralgorithmshouldprovideforstablelinksthroughoutthecourseofthesimulation.Byensuringnodeconnectivity,thecontrollayeralgorithmalsoensuresthatthereisaconstantcommunicationlinkbetweenallnodesinthenetwork,i.e.,atreestructureofthenetworkisalwayspreserved.LinkstabilityisevaluatedbyestablishingUDP/TCPowsbetweenallnodesandcontinuouslymoni-toringthestatusoftheseows.Thecorrectnessofthealgorithmistestedfor6differentmobilitymodelswiththesamecommunicationmodelandphysicalparameters.46

PAGE 55

4.2.1MobilityModelItisimperativetoevaluatetheproposedalgorithmutilizingdifferentmobilitymodelsastheperformanceofanadhocnetworkcanvarysignicantlywithdifferentmobilitypatterns[44].Asurveyofthevariousmodelsispresentedin[44].Theselectedmobilitymodelneedstorepresentthemovementbehaviorofrobotsinsearchandrescueoperations.Themobilityofautonomousrobotsiscontrolledbyitssensorinputsanditsprogrammedresponsebehavior.Thisessentiallymeansthattherobotsmovearoundinarandomfashion,notfollowingapre-denedpattern.Therandomwaypointmobilitymodel[45]replicatesthisbehavior:nodesmovetowardsaparticulardestinationataselectedspeed,pauseforpausetime,selectarandomdestinationandanewspeedandstartmovingtowardsthisnewdestination.Thisisrepeatedthroughoutthecourseofthesimulation.Randomwaypointmodel(2clusterbasedmodelsandagenericmodel),randomwalkmodel,agroupmobilitymodel(columnbasedmobilitywhereallnodesmovetowardsthesamedestinationatthesamespeed)andaspecialcaseoftherandomwaypointmobilitymodelareusedtoevaluatetheperformanceoftheproposedalgorithm.Inthemodiedversionofrandomwaypointmodel,nodesselectaparticulardestinationandmovetowardsthisdestinationatasetspeedanddonotmoveanyfurtherafterreachingthedestination,whileinthebasemodelafterabriefpausetimetheywouldpickanotherdestinationandstartmovingtowardsit.Intheresultssection(Section4.3),themobilitymodelofthenodesisrstintroduced,followedbyananalysisofthethroughput,linkstabilityandoverheadgraphsforeachmodel.4.2.2CommunicationModelAllthemobilenodesareconnectedtothemaincontrollerthroughFTPowsthatexistfortheentiredurationofsimulation(Packetsizeis512bytes).SimilarlyFTPconnectionsareestablishedfromthemaincontrollertotheindividualnodeswiththesameparameters.Monitoringtheseowsgivesinformationonthestabilityoftheselinksandthedurationforwhichtheowhasbeenactive.Similarly,tostudytheeffectoftheproposedalgorithmontheoveralllinkstabilityofthesystem,verylowbitrateUDPconnectionarecreatedbetweenallthemobilenodesofthenetwork(0.8Kbitsor100bytespersecond).Thisformsacompletelyconnectedgraphstructureofnnodes,wherethetotalnumberoflinks/owsisequalto},=+,whichisatotalof72owsinthesimulations47

PAGE 56

Figure4.1MobilityModel1(StraightLineMovementAlongtheDiagonal)(8mobilenodesandamaincontroller).Thetrafcmodelusedisarepresentativeofthekindofcommunicationdesiredbetweenrobotsoperatingindisastersites.Alltherobotsconstantlycom-municatewiththemaincontroller(FTPconnections),andalsokeepsendingminimalinformationtoneighbornodes(CBRconnections).4.3ResultsThissectiondiscussesthesimulationresultsandanalyzestheperformanceoftheproposedalgorithm.6mobilitymodelshavebeenusedandabriefdescriptionofeachofthemodelsprecedetheperformanceanalysisforthatmodel.4.3.1SimulationModel1:GroupMobilityTherstmodelimplementsagroupmobilitypattern(columnmobilitymodel)[44].Themaincontrollerisstationedatthelocation(0,0)andthe8mobilenodesstartat(0,0)andmovealongthediagonaltowards(600,600)atarateof1m/s(Figure4.1).Theidealtransmissionrangeofthe48

PAGE 57

LucentWaveLANDSSSradiointerfaceis250m.Thisimpliesthatnodesmovingat1m/swouldbreakitscommunicationwiththemaincontrollerataround250seconds.ThiswouldbethecasewhenthenetworkusesanyoftheexistingWLANs.ButwhenthenetworkisoperatedwiththeAODVbasedcontrollayeralgorithm,thereshouldbeaconstantcommunicationlinkfromallthenodestothemaincontroller.SimulationswererunforthismobilitymodelwiththeexistingAODV(whichisreferredtoasbaseAODV)andAODVwiththecontrollayeralgorithm.Theresultsofsimulationarepresentedbelow.Themostimportantofallplotsisthelinkstabilityplot.Asdiscussedintheprevioussectionwehaveatotalof72ows(ncompletelyconnectednodes,n(n-1)links):16FTPowsfromthemobilenodestothemaincontrollerandviceversaand56CBRows,ataverylowdatarate,inter-connectingallthe8mobilenodes.FlowID1-56denotestheUDP/CBRconnectionsbetweenthemobilenodes,whileowID's57-64denotesconnectionsfromthemaincontrollertothemobilenodesandowID's65-72denotesconnectionfrommobilenodestomaincontroller.Thecommu-nicationbetweenmobilenodes(CBRows1-56)wouldalwaysbeactiveasthenodesaremovingtogether.However,ataround250secondsinthesimulationallthenodeswouldloosetheircommu-nicationwiththemaincontroller,indicatingabreakintheows57-72.ButwhenusingAODVwiththecontrollayeralgorithm,thisisnotthecase,andthelinkstabilityplotinFigure4.2indicatesthataconstantcommunicationlinkisestablishedforallowsthroughoutthedurationofthesimulation.ThiscanbeveriedfromthethroughputplotsinFigures4.3and4.4,theformerrepresentstheoverallnetworkthroughputforallows,whilethelattercorrespondstothethroughputatthemaincontroller.FromthelatterplotitisevidentthatthroughputofbaseAODVdropstozeroat250seconds,whentheconnectionswiththemobilenodesareterminated.HoweverAODVwithcontrollayerensuresasteadythroughputthroughoutthedurationofthesimulation.FromtheoverallthroughputplotinFigure4.3,itcanbeseenthatthenetworkthroughputdoesnotdroptozeroforbaseAODVandthisisduetotheCBRowsbetweenthemobilenodes.Asinthepreviouscase,AODVwithcontrollayeralgorithmprovidesasteadythroughputthroughoutthedurationofsimulation.Itistobenotedthatfortherst250secondsofsimulationwhenallowsareactive,thethroughputachievedbyboththeprotocolsarecomparable,indicatingthatthenetworkthroughputisnotaffectedbytheoverheadduetothecontrollayeralgorithm.49

PAGE 58

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 1)AODV AODV with Control Layer Figure4.2ComparativeLinkStabilityAnalysisforMobilityModel1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 1AODV AODV with Control Layer Figure4.3ThroughputofAllFlowsforModel1 0 10000 20000 30000 40000 50000 60000 70000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 1AODV AODV with Control Layer Figure4.4ThroughputatMainControllerforModel1 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 1AODV AODV with Control Layer Figure4.5NetworkOverhead(RoutingandHelloPackets)forModel150

PAGE 59

Figure4.5plotstheoverallroutingoverhead(whichincludesthehellopackets,updatepacketssenttothemaincontroller,andAODVcontrolpackets)ofthetwocomparedprotocols.AODVwithcontrollayeralgorithmintroducesameanincreaseinoverheadof1200bytespersecond,attherateof4.8Kbps.Obviouslythisincreaseinoverheadisveryminimalwhencomparedwiththenetworkbandwidthof2Mbpsandtheperformanceimprovementsachievedbyusingthecontrollayeralgorithm.Fromthesimulationresultsandplotsformodel1,itisclearthatAODVwithcontrollayeralgorithmachievesconstantnodeconnectivityandsteadythroughputwithoutincurringanymajorincreaseinroutingoverhead.4.3.2SimulationModel2:ClusterBasedRandomWaypointModelInthismodelaclusterbasedmovementofthenodesissimulated.ThenetworktopologyisasshowninFigure4.6.Nodes1,2,3and4formaclusterandmovetowardsthex-axisaty=600.Similarlynodes5,6,7and8formaclusterandmovetowardstheyaxisatx=600.Othersimulationparametersarethesameasinthepreviousmodel.Intuitively,whenusingthebaseAODValgorithm,thenodeswouldloosetheirconnectivitywiththemaincontrollerataround250seconds.Howeverunlikeinthepreviousmodelwheretheowsamongmobilenodeswasalwaysactive,therewouldnotbeanyconnectionbetweenthenodesofthetwoclustersaftert=250s.Butthenodeswithinthesameclustercouldcommunicatewitheachotherthroughoutthecourseofthesimulation.Thusthelinkstabilityplotshouldindicatealternatingstabledurationsof600sand250sforows1to56andthedurationofows57to72wouldbearound250s.Thelinkstabilityplotformodel2ispresentedinFigure4.7.Theresultsareasexpectedandeveninthismodel,AODVwithcontrollayeralgorithmprovidesstablelinksforalltheows.Itisseenthattheimplementationofcontrollayeralgorithmnotonlyimprovesthestabilityofthelinkswiththemaincontroller,butalsothelinksbetweenthemobilenodes.Theoverallnetworkthroughput,throughputatmaincontrollerandoverheadformodel2areshowninFigures4.8,4.9and4.10respectively.Asinthepreviousmodel,AODVwiththecontrollayeralgorithmensuresconstantthroughputthroughoutthecourseofthesimulation,whilethethroughputofthebaseAODValgorithmdropsat250s,whenthemobilenodesbreakawayfrom51

PAGE 60

Figure4.6MobilityModel2(2ClustersMovingTowardsPresetDestination)themaincontroller.Similarly,theincreaseinoverheadduetocontrollayerpacketsisminimalandisoutweighedbytheachievedperformanceimprovement.4.3.3SimulationModel3:RandomWaypointModel(IndependentDirection)Inthismodel,themaincontrollerisstationedatthecenterofthedisastersite(300,300)andallthemobilenodesstartfromthislocationandmovetowardstheperimeterofthesimulationarea.Eachnodemovesinadifferentdirection,atanangleof45degreesfromitsneighbors.Intuitively,allnodeswouldloosetheirconnectionwiththemaincontroller(ows57-72)ataround250seconds(maximumtransmissionrangeis250mandnodesaremovingat1m/s),howeveralltheinter-nodalconnections(ows1-56)wouldbreakataround350seconds(rightangletriangleswith250msideswouldhaveanhypotenused350meters).ThelinkstabilityplotinFigure4.12showsthisbehavior.WhenusingthebaseAODV,thecommunicationlinkwiththemaincontrollerisbrokenat250s(ows57-72),whiletheinter-nodalcommunication(ows1-56)breaksat340s.HoweverAODVwithcontrollayeralgorithmensuresconstantlinkstabilitythroughoutthedurationofsimulation.52

PAGE 61

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 2)AODV AODV with Control Layer Figure4.7ComparativeLinkStabilityAnalysisforMobilityModel2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 2AODV AODV with Control Layer Figure4.8ThroughputofAllFlowsforModel2 0 10000 20000 30000 40000 50000 60000 70000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 2AODV AODV with Control Layer Figure4.9ThroughputatMainControllerforModel2 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 2AODV AODV with Control Layer Figure4.10NetworkOverhead(RoutingandHelloPackets)forModel253

PAGE 62

Figure4.11MobilityModel3(NodesMovingatanInter-nodalAngleof45)ThethroughputplotsshowninFigures4.13and4.14representthesamebehavior.Thethrough-put,whenusingbaseAODV,dropsto0.1Mbpsafter250secondsofsimulationandtozeroafter350seconds.Theinitialdropisduetothebreakageoflinksbetweenthemobilenodesandthemaincontrollerwhiletheseconddropisduetothebreakageoflinksbetweenthemobilenodes.However,AODVwithcontrollayeralgorithmensuresasteadythroughputfortheentirecourseofsimulation.Figure4.15comparesthenetworkoverheadduetoAODV,helloandcontrollayerpackets.Clearlythereisameanincreaseof1200bytesinoverhead,butthisincreaseisinsignicantcomparingtheavailablebandwidthandtheachievedperformanceimprovements.4.3.4SimulationModel4:RandomWalkModelInthismodelthemaincontrollerisstationedat(0,0)andalltheothermobilenodesstartfromthislocation(Figure4.16).Themobilenodesselectarandomdestination(6250m)andmoveto-wardsthisdestinationatarateof1m/s.Notwonodesmovetowardsthesamedestination.Therandommobilityofthenodesmakesittoughtopredictthelinkstabilityandthroughputofthe54

PAGE 63

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 3)AODV AODV with Control Layer Figure4.12ComparativeLinkStabilityAnalysisforMobilityModel3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 3AODV AODV with Control Layer Figure4.13ThroughputofAllFlowsforModel3 0 10000 20000 30000 40000 50000 60000 70000 80000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 3AODV AODV with Control Layer Figure4.14ThroughputatMainControllerforModel3 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 3AODV AODV with Control Layer Figure4.15NetworkOverhead(RoutingandHelloPackets)forModel355

PAGE 64

Figure4.16MobilityModel4(RandomWalkModel)network.However,itisimperativethattheresultsforAODVwithcontrollayeralgorithmshouldensurenodeconnectivitythroughoutthedurationofsimulation.Acomparativelinkstabilityanaly-sisofthenetworkispresentedinFigure4.18.Ascanbeseenfromtheplot,whenusingbaseAODVallnodesloosetheircommunicationlinkwithbasestationat250s,whileAODVwithcontrollayeralgorithmensuresstablelinksthroughoutthesimulation.Theinter-nodalcommunicationremainsstablefortheentiresimulationtime,whenusingAODVandAODVwithcontrollayeralgorithm.But,duringthetimewherecompleteconnectivityexistsinbothalgorithms(0-250seconds),thenetworkthroughputgraphinFigure4.19showsamarginaldifferenceinthethroughputachievedbythetwodifferentprotocols.AODVwithcontrollayerhasalowerthroughput(0.1MbpslesserthanbaseAODV),andthiscouldbeattributedtotheoverheadincurredbysendingthecontrollayerpackets.AsimilardifferenceinthroughputisobservedinFigure4.20andalsofortheothermobil-itymodels.Figure4.21plotstheroutingprotocoloverheadofthetwomethodsandlikeinpreviousmodelsthereisameanincreaseof1200bytesinoverheadwhenusingAODVwithcontrollayeralgorithm.56

PAGE 65

Figure4.17MobilityModel5(NodesMovingin2DifferentClusters)4.3.5SimulationModel5:RandomWaypointModel(ClusterBasedMobility)ThemodelpresentedinFigure4.17issimilartomobilitymodel2,butforthefactthatthereare6nodesinoneclusterand2intheother.However,inmobilitymodel2,therewere4nodesineachcluster,andthecommunicationbetweenthenodesofthetwoclusterswasdependentontheirconnectionwiththemaincontroller.Whenthenodesineachclusterlosttheirlinkwiththemaincontroller,itresultedinalinkbreakagebetweenthetwoclusters.ButinthemodelshowninFigure4.17,nodesinthetwoclusterscancommunicateevenaftertheircommunicationwiththemaincontrolleristerminated.Forexample,nodes2and3canmaintainacommunicationlinkforacertaindurationevenaftertheircommunicationwiththemaincontrolleristerminated.Ifthereisaconnectionbetweenonenodeinaclusterandanothernodeintheothercluster,thenallnodesineachclustercanroutepacketsthroughthesenodes(nodes2and3inthiscase).But,whenthese2nodesindifferentclustersloosetheirconnection,thetwoclustersbecomecompletelyindependent.LinkstabilityplotshowninFigure4.22conrmsthiskindofabehavior.Asinearliermodels,themobilenodesbreakawayfromthemaincontroller(ows57-72)after250secondsintosimulation57

PAGE 66

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 4)AODV AODV with Control Layer Figure4.18ComparativeLinkStabilityAnalysisforMobilityModel4 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 4AODV AODV with Control Layer Figure4.19ThroughputofAllFlowsforModel4 0 10000 20000 30000 40000 50000 60000 70000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 4AODV AODV with Control Layer Figure4.20ThroughputatMainControllerforModel4 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 4AODV AODV with Control Layer Figure4.21NetworkOverhead(RoutingandHelloPackets)forModel458

PAGE 67

whenusingbaseAODV.Theclustermobilitybehaviorisrepresentedbyows1-56anduntil410seconds,aconnectionexistsbetweenallthemobilenodeseitherdirectlyorthroughothernodesinthesamecluster.Howeverat410seconds,thetwoclustersbecomeindependentandweonlyhaveinter-clustercommunication.Flows1-14representtheCBRconnectionsfromtheclusterwith2nodes(thatisthereasonwhyweseemanyconnectionsbeingbrokenat410seconds)andows15-56representtheCBRconnectionsfromtheclusterwith6nodes(moreowsremainconnectedfortheentiredurationofthesimulation).Likeinthepreviousmodels,AODVwithcontrollayeralgorithmensuresasteadyconnectionofallthemobilenodesthroughoutthedurationofsimulation.ThroughputgraphsforthismodelarepresentedinFigures4.23and4.24.Throughputatthebasestationdropstozeroat250seconds,whenusingAODV,whileAODVwithcontrollayerensuresastablethroughput.TheoverheadplotforthismodelispresentedinFigure4.254.3.6SimulationModel6:RandomWaypointModelModel6representstheRandomWaypointModeldiscussedin[44].Themobilitylewasgener-atedusingthesetdesttooldistributedwiththenssimulator.Eachnodeselectsarandomdestinationandmovestowardsit,andonreachingthedestination,pausesforthegivenpausetimeandthenpicksanotherdestinationandmovestowardsit.Thispatternisrepeatedfortheentiredurationofsimulation.Also,thespeedofeachnodeisdifferentwithrespecttoothernodes.ThenodescanbeexpectedtomoveinamannerasshowninFigure4.26,butitistobenotedthatunlikeotherpre-viousmodelsthisguredoesnotrepresenttheactualnodedestinationanddirection.FromthelinkstabilitygraphshowninFigure4.27,itisevidentthatallthenodesremainconnectedthroughoutthedurationofsimulationwhenusingbothAODVandAODVwithcontrollayeralgorithm.TheperformanceofboththeprotocolsaresimilarandcanbeveriedfromthethroughputgraphsinFigures4.28and4.29.Also,likeinpreviousmodelsthereisaveryminimalincreaseinoverhead,ameanincreaseof1200bytes(Figure4.30).Thus,usingAODVwithcontrollayeralgorithminmodelswherethenodesstaywithinthetransmissionrangeandanyoftheexistingprotocolswouldensurenodeconnectivity,doesnotaffecttheperformance.Fromthesimulationresultsofthe6modelsitisveryevidentthatAODVwiththe59

PAGE 68

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 5)AODV AODV with Control Layer Figure4.22ComparativeLinkStabilityAnalysisforMobilityModel5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 5AODV AODV with Control Layer Figure4.23ThroughputofAllFlowsforModel5 0 10000 20000 30000 40000 50000 60000 70000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 5AODV AODV with Control Layer Figure4.24ThroughputatMainControllerforModel5 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 5AODV AODV with Control Layer Figure4.25NetworkOverhead(RoutingandHelloPackets)forModel560

PAGE 69

Figure4.26MobilityModel6(RandomWaypointModel)proposedcontrollayeralgorithmensuresconsistentnodeconnectivityoftheentirenetworkbymakingsurethatallthenodesstayconnectedwiththemaincontrollereitherdirectlyorthroughothernodes.4.3.7AreaofCoverageThissectionanalyzestheperformanceoftheproposedcontrollayeralgorithmforthedistancecoveredbythemobilenodeswhenusingthevariousmobilitymodels.Thedistancetraversedbythemobilenodesisestimatedasthemaximumdistanceatwhichthenodescanestablishcommunicationwiththemaincontroller.However,whenusingthebaseAODValgorithm,thenodeswouldstillcontinuetheirmobility(evenafterreachingthetransmissionrangeofthemaincontroller)untiltheyreachtheirpresetdestination.ButinthecaseofAODVwithcontrollayeralgorithm,thismeasure(distancecovered)isanindicatorofthedistancetraversedtowardsthedestination,atwhichthemobilenodesarestopped.Unlikemobilitymodels1to5,nodesinmodel6keepchangingtheirdestinationandrateofmobility,andhencethevaluespresentedinTable4.6areanapproximationof61

PAGE 70

0 100 200 300 400 500 600 700 800 0 10 20 30 40 50 60 70 80 Duration of flow (seconds) Flow ID (72 flows)Link duration of data flows (Estimation of link stability for model 6)AODV AODV with Control Layer Figure4.27ComparativeLinkStabilityAnalysisforMobilityModel6 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 100 200 300 400 500 600 Throughput (mbps) Time (sec)Overall Throughput for Mobility model 6AODV AODV with Control Layer Figure4.28ThroughputofAllFlowsforModel6 0 10000 20000 30000 40000 50000 60000 70000 0 100 200 300 400 500 600 Throughput (bytes) Time (sec)Throughput at Main Controller for Mobility model 6AODV AODV with Control Layer Figure4.29ThroughputatMainControllerforModel6 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 Overhead (bytes) Time (sec)Routing protocol overhead for Mobility model 6AODV AODV with Control Layer Figure4.30NetworkOverhead(RoutingandHelloPackets)forModel662

PAGE 71

Table4.1ComparativeDistanceCoveredbyMobileNodesinModel1 NodeID AODV(meters) AODVwithControlLayer(meters) 1 256.03 127.81 2 255.93 248.65 3 255.77 368.48 4 256.38 486.13 5 256.07 569.49 6 246.20 595.96 7 250.02 592.83 8 256.20 559.12 Table4.2ComparativeDistanceCoveredbyMobileNodesinModel2 NodeID AODV(meters) AODVwithControlLayer(meters) 1 253.41 126.86 2 241.85 243.14 3 253.51 360.58 4 250.59 471.12 5 255.66 136.25 6 251.65 251.68 7 198.20 373.46 8 255.58 326.03 thedistancecoveredbythemobilenodeswithrespecttoitsnalpositionatwhichitcouldestablishacommunicationlinkwiththemaincontroller.Table4.1comparesthedistancetraversedbythemobilenodesinmodel1usingbaseAODVandAODVwithcontrollayeralgorithm(Section4.3.1).WhenusingthebaseAODValgorithm,allthemobilenodesbreaktheircommunicationwiththemaincontrolleraftertraversingadistancebetween246.2metersand256.03meters.However,byusingAODVwithcontrollayeralgorithm,nodes5-8traverseadistancegreaterthan559.12meters,withnodes1-4beingstoppedatequallyspacedintervalstoenablemulti-hoproutingandconstantcommunication.Also,nodes6and7continuetheirmobilitythroughoutthedurationofsimulation(600seconds),whileconstantlycommunicatingwiththemaincontroller.TheproposedcontrollayeralgorithmclearlyoutperformsthebaseAODVprotocolwithrespecttothedistancecoveredbythenodesinmobilitymodel1.Thedistancetraversedbythemobilenodesinmodel2(Section4.3.2)isshowninTable4.2.ThemaximumtraverseddistancewhenusingthebaseAODValgorithmis255.7metersbynode5,afterwhichallthenodesbreaktheircommunicationwiththemaincontroller.However,when63

PAGE 72

Table4.3ComparativeDistanceCoveredbyMobileNodesinModel3 NodeID AODV(meters) AODVwithControlLayer(meters) 1 249.00 116.62 2 197.92 173.01 3 230.12 153.26 4 241.48 161.86 5 249.49 183.00 6 249.82 193.78 7 249.38 203.60 8 248.92 213.43 Table4.4ComparativeDistanceCoveredbyMobileNodesinModel4 NodeID AODV(meters) AODVwithControlLayer(meters) 1 251.62 124.04 2 255.24 243.75 3 255.10 356.39 4 256.25 441.14 5 256.53 349.85 6 254.19 337.13 7 255.90 276.59 8 255.13 380.04 usingAODVwithcontrollayeralgorithm,themaximumtraverseddistancewithoutbreakingthecommunicationlinkwiththemaincontrolleris471.12metersbynode4,whiletheminimumtra-verseddistanceis126.86metersbynode1.Again,formobilitymodel2,theproposedcontrollayeralgorithmachievesgreaterareaofcoveragewhileensuringconstantcommunicationwiththemaincontroller.InTable4.3thedistancetraversedbythemobilenodesinmobilitymodel3(Section4.3.3)iscompared.Unlikeintheprevioustwocases,thedistancetraversedbythemobilenodesusingthebaseAODValgorithmishigherthanthedistanceachievedusingAODVwithcontrollayeral-gorithm.Thisbehaviorisattributedtotwosignicantfactors:(a)mobilitypatternand(b)linkthreshold.Inmobilitymodel3(Figure4.11),allthenodesmovetowardstheperipheryofacirclewiththemaincontrolleratthecenter.Suchamobilitypatterndoesnotprovidescopeformulti-hoprouting,andAODVwithcontrollayeralgorithmstopseachnodeatthelinkthresholdtoensureconstantcommunicationwiththemincontroller.Also,thisworkusesalinkthresholdof’=‡_-ˆ=‘->rWatts,whilethereceivingthreshold(RXThresh )forthebaseAODVissetto’-‡_-ˆ=‘->ƒ(Watts.Per-64

PAGE 73

Table4.5ComparativeDistanceCoveredbyMobileNodesinModel5 NodeID AODV(meters) AODVwithControlLayer(meters) 1 253.05 127.05 2 254.26 241.90 3 256.32 200.49 4 256.36 313.63 5 256.30 433.37 6 256.20 394.29 7 250.87 504.07 8 255.90 449.26 Table4.6ComparativeDistanceCoveredbyMobileNodesinModel6 NodeID AODV(meters) AODVwithControlLayer(meters) 1 447.10 300.18 2 86.04 86.07 3 289.08 298.02 4 191.32 191.76 5 303.30 307.15 6 231.06 229.76 7 115.06 115.00 8 136.89 135.40 formanceoftheproposedcontrollayeralgorithmcouldbeimprovedbyincreasingthelinkthresholdvaluetothereceivingthreshold(RXThresh ).Tables4.4and4.5comparethedistancetraversedbythemobilenodeswhenusingbaseAODVandAODVwithcontrollayeralgorithm,formobilitymodels4and5(Sections4.3.4and4.3.5)respectively.Inbothcases,AODVwithcontrollayeralgorithmresultsinlargerdistancesbeingtraversedbythemobilenodes.Inmobilitymodel6,allthenodes(exceptfornode1)remainwithinthetransmissionrangeofthemaincontrollerandhencethedistancetraversedbythenodeswhenusingbaseAODVissimilartothedistancetraversedbythenodeswhenusingAODVwithcontrollayeralgorithm.Thus,theproposedcontrollayeralgorithmresultsinsignicantincreaseinthedistancesbeingcoveredbythemobilenodeswhileensuringconstantNODECONNECTIVITY.65

PAGE 74

CHAPTER5CONCLUSIONANDFUTUREWORK5.1ConclusionInthisthesistheideaofusingadhocnetworkinginsupportofUSARapplicationsispro-posed.Anewcontrollayeralgorithmisdesignedtoguaranteenodeconnectivityandensuresteadythroughputwithaveryminimalincreaseinoverhead.Nodeconnectivityiscriticaltotheperfor-manceofanetworkwherelinkbreakagesornodelosscouldincurheavyperformanceandnancialdamages,suchasanetworkofautonomousrobotscollaboratinginsearchandrescueoperations.Theproposedalgorithmisimplementedinns-2.26simulator,anditsperformanceisanalyzedusing6differentmobilitymodels.Thesimulationresultsshowthatthecontrollayerbasedalgorithmguaranteesnodeconnectivitywithveryminimalincreaseinoverhead.5.2FutureWorkThisistherstworkattemptingtoensurenodeconnectivityforadhocnetworksandsuggesttheideaofacontrollayerbasedsolution.Inthisworktheproposedalgorithmwasveriedanditsusefulnessandperformancewasevaluated.However,theideaofnodeconnectivityandcontrollayerbasedsolutionintroducesavastscopeforfutureresearchwork.Forexample,Basedontheinformationcollectedatthemaincontroller,agraphicaltoolcouldbedesignedtoprovidetherescuepersonnelwithexactinformationonthepositionandstatusofthemobilenodes,stabilityofthelinksandalsoanydatareceivedfromthenodes.Thiswouldalsofacilitateintele-operatingtherobots.TheproposedmodelusesanHELLOINTERVALandUPDATEINTERVALof1second,andtheMONITORINTERVALis1.5seconds.Itwouldbeinterestingtostudytheeffectofvaryingthesevaluesontheperformanceofthesystem.Itisessentialtoselecttherightvaluefortheseparameters,asasmallerintervalleadstonetworkcongestionandincreasedoverhead,whilelargeintervalsmighttoleadtostoppinganodewellafteritbreaksawayfromthenetwork.Theupdatepacketssentby66

PAGE 75

themobilenodeshaveaeldforsequencenumberswhichisnotusedinthiswork.However,theperformanceimprovementachievedbyusingsequencenumberforupdatepacketscouldbestudied.AperformancecomparisonofTCPandUDPfortheproposedsolutioncouldbedone,andalsoanalyzetheperformanceofvariousTCPversionsfortheproposedmethod.Also,insteadofsendingtheupdatepacketsseparately,theycouldbepiggy-backedwiththedatapackets.Theperformanceoftheproposedmethodathighmobilityratescouldbestudied.Theperformanceoftheproposedmethodcouldbecomparedwithexistingadhocroutingsolutionsthataiminselectingstablelinksforroutingpacketsbasedonsignalstability,locationstabilityandenergyofthenodes.67

PAGE 76

REFERENCES[1]I.Chlamtac,M.Conti,andJ.J.-N.Liu,“MobileAdhocNetworking:ImperativesandChal-lenges,”AdHocNetworks,vol.1,pp.13–64,July2003.[2]D.B.JohnsonandD.A.Maltz,“DynamicSourceRoutinginAdhocWirelessNetworks,”inMobileComputing,ImielinskiandKorth,Eds.KluwerAcademicPublishers,1996,vol.353.[3]J.A.FreebersyserandB.Leiner,ADoDPerspectiveonMobileAdHocNetworks,2nded.,ser.Adhocnetworking,editedbyCharles.E.Perkins.Addison-Wesley,2001,vol.1,chapter2,pp.29–51.[4]“Ashortdescriptiononadhocnetworks,”http://www.ka9q.net/papers/wcnc99/.[5]“NewYorkSearchandRescue(FAQ),”http://newyorksearchandrescue.org/faq.html.[6]R.Murphy,J.Casper,J.Hyams,M.Micire,andB.Minten,“MobilityandSensingDemandsinUSAR,”inConferenceoftheIEEEIndustrialElectronicsSociety(IECON),Nagoya,Japan,October2000.[7]“RemotecontrolledRobotssearchWorldTradeCenterrubble,”http://www.tms.org/pubs/journals/JOM/0112/News/News3-0112.html.[8]“WirelessLANMediumAccessControl(MAC)andPhysicalLayer(PHY)Specications,”TheInstituteofElectricalandElectronicsEngineers,NewYork,IEEEStd.802.11-1997,July1997.[9]M.Gerharz,C.D.Waal,M.Frank,andP.Martini,“LinkStabilityinMobileAdHocNet-works,”inProceedingsofthe27thAnnualIEEEConferenceonLocalComputerNetworks(LCN'02),Tampa,FL,November2002,pp.30–39.[10]R.Dube,C.D.Rais,K.-Y.Wang,andS.K.Tripathi,“SignalStabilitybasedAdaptiveRouting(SSA)forAd-HocMobileNetworks,”IEEEPersonalCommunications,vol.4,pp.36–45,February1997.[11]M.SchwartzandT.Stern,“RoutingTechniquesusedinComputerCommunicationNetworks,”IEEETransactionsonCommunications,vol.COM-28,pp.539–552,April1980.[12]J.M.McQuillan,I.Richer,andE.C.Rosen,“TheNewRoutingAlgorithmfortheARPANET,”IEEETransactiononCommunications,vol.COM-28(5),pp.711–719,May1980.[13]J.Moy,“Openshortestpathrstprotocol,”TheInternetEngineeringTaskForce(IETF),RFC2328,April1998.[14]C.Hendrick,“Routinginformationprotocol,”TheInternetEngineeringTaskForce(IETF),RFC1058,June1988.68

PAGE 77

[15]C.E.PerkinsandP.Bhagwat,“HighlydynamicDestinationSequencedDistance-vectorRout-ing(DSDV)forMobileComputers,”inProceedingsoftheACMSIGCOMMSymposiumonCommunication,ArchitectureandProtocols,vol.24,October1994,pp.234–244.[16]S.MurthyandJ.J.Garcia-Luna-Aceves,“ARoutingProtocolforPacketRadioNetworks,”inProceedingsoftheACMFirstInternationalConferenceonMobileComputingandNetworking(MOBICOM'95),November1995,pp.86–95.[17]C.E.Perkins,Belding-Royer,andS.Das,“AdhocOn-DemandDistanceVector(AODV)Routing,”TheInternetEngineeringTaskForce(IETF),RFC3561,July2003.[18]V.D.ParkandM.S.Corson,“AHighlyAdaptiveDistributedRoutingAlgorithmforMobileWirelessNetworks,”INFOCOM,pp.1405–1413,April1997.[19]C.-K.Toh,“Associativity-BasedRoutingforAdHocMobileNetworks,”WirelessPersonalCommunications,vol.4-2,pp.103–139,March1997.[20]Y.-B.KoandN.Vaidya,“Location-AidedRouting(LAR)inMobileAdHocNetworks,”inProceedingsoftheFourthACM/IEEEInternationalConferenceonMobileComputingandNetworking(MobiCom'98),Dallas,Texas,October1998,pp.66–75.[21]J.J.Garcia-LunaAceves;M.Spohn,“Source-treeroutinginwirelessnetworks,”inProceed-ingsoftheSeventhInternationalConferenceonNetworkProtocols(ICNP'99),October1999,pp.273–282.[22]D.CamaraandA.A.Loureiro,“ANovelroutingalgorithmforadhocnetworks,”inProceed-ingsofthe33rdInternationalConferenceonSystemSciences,Maui,Hawaii,January2000.[23]I.StojmenovicandX.Lin.,“Power-awarelocalizedroutinginwirelessnetworks,”IEEETransactionsonParallelandDistributedSystems,vol.12,pp.1122–1133,October2001.[24]Z.J.HaasandM.R.Pearlman,ZoneRoutingProtocol(ZRP)AHybridFrameworkforRoutinginAdhocNetworks,2nded.,ser.AdhocNetworking,editedbyCharles.E.Perkins.Addison-Wesley,2001,vol.1,chapter7,pp.221–253.[25]C.K.Toh,AdHocMobileWirelessNetworks.UpperSaddleRiver,NewJersey:PrenticeHall,2002.[26]J.Broch,D.A.Maltz,D.B.Johnson,Y.-C.Hu,andJ.Jetcheva,“APerformanceComparisonofMulti-HopWirelessAdHocNetworkRoutingProtocols,”MobileComputingandNetwork-ing,pp.85–97,1998.[27]S.R.Das,C.E.Perkins,andE.E.Royer,“Performancecomparisonoftwoon-demandroutingprotocolsforadhocnetworks,”inProceedingsoftheIEEEConferenceonComputerCommu-nications(INFOCOM),March2000,pp.3–12.[28]P.Johansson,T.Larsson,N.Hedman,B.Mielczarek,andM.Degermark,“Scenario-basedperformanceanalysisofroutingprotocolsformobileadhocnetworks,”inProceedingsofthe5thannualACM/IEEEInternationalConferenceonMobileComputingandNetworking,Seattle,Washington,August1999,pp.195–206.[29]A.Boukerche,“APerformancecomparisonofroutingprotocolsforadhocnetworks,”inPro-ceedingsoftheIEEEParallelandDistributedComputingforWirelessandMobileSystems,LNCS,Springer-Verlag,Berlin,August2001,pp.195–206.69

PAGE 78

[30]M.Mauve,J.Widmer,andH.Hartenstein,“Asurveyonposition-basedroutinginmobileadhocnetworks,”IEEENetworkMagazine,vol.15(6),pp.30–39,November2001.[31]S.Giordano,I.Stojmenovic,andL.Blazevie,“Positionbasedroutingalgorithmsforadhocnetworks:ataxonomy,”2001.[32]G.HollandandN.Vaidya,“AnalysisofTCPPerformanceoverMobileAdHocNetworks,”inProceedingsofthe5thannualACM/IEEEInternationalConferenceonMobileComputingandNetworking,Seattle,Washington,1999,pp.219–230.[33]VenkateshRamarathinamandMiguelLabrador,“PerformanceanalysisofTCPoverstaticadhocnetworks,”inProceedingsoftheISCA15thInternationalConferenceonParallelandDistributedComputingSystems(PDCS),September,2002,pp.410–415.[34]P.Sholander,A.Yankopolus,P.Coccoli,andS.Tabrizi,“ExperimentalComparisonofHybridandProactiveMANETRoutingProtocols,”inProceedingsoftheMilitaryCommunicationsConference(MILCOM),October2002.[35]J.RajuandJ.Garcia-Luna-Aceves,“AComparisonofOn-demandandTable-drivenroutingforAdhocwirelessNetworks,”inProceedingsoftheIEEEInternationalConferenceonCom-munications,JUNE2000,pp.1702–1706.[36]A.WineldandO.Holland,“Theapplicationofwirelesslocalareanetworktechnologytothecontrolofmobilerobots,”JournalofMicroprocessorsandMicrosystems(Elsevier),Vol23/10,pp.597–607,August2000.[37]A.Wineld,“Distributedsensinganddatacollectionviabrokenadhocwirelessconnectednetworksofmobilerobots,”DistributedAutonomousRoboticSystems,pp.273–282,2000.[38]“DARPAprojectdevelopscompactadhocnetworkingradios,”RoboticsUpdate,vol.3,no.1,pp.1122–1133,April2003.[39]K.V.KevinFall,TheNSManual(formerlynsNotesandDocumentation),theVINTProjectAcollaborationbetweenresearchersatUCBerkeley,LBL,USC/ISI,andXeroxPARC.[40]“NAM:NetworkAnimator,”http://www.isi.edu/nsnam/nam/.[41]I.Stojmenovic,“Position-basedroutinginadhocnetworks,”IEEECommunicationsMaga-zine,vol.40,no.7,pp.128–134,July2002.[42]LucentTechnologies,“WaveLAN/PCMCIAcarduser'sguide,”October1996.[43]TheodoreS.Rappaport,WirelessCommunications:PrinciplesandPractice.NewJersey:PrenticeHall,1996.[44]T.Camp,J.Boleng,andV.Davies,“ASurveyofMobilityModelsforAdHocNetworkRe-search,”WirelessCommunicationsandMobileComputing(WCMC):SpecialissueonMobileAdHocNetworking:Research,TrendsandApplications,vol.2,no.5,pp.483–502,2002.[45]D.JohnsonandD.Maltz,DynamicSourceRouting(DSR)inAdhocWirelessNetworks,2nded.,ser.MobileComputing.KluwerAcademicPublishers,1996,pp.153–181.70