USF Libraries
USF Digital Collections

An architecture for global ubiquitous sensing

MISSING IMAGE

Material Information

Title:
An architecture for global ubiquitous sensing
Physical Description:
Book
Language:
English
Creator:
PEREZ, ALFREDO JOSE
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Cellular Networks
Distributed Systems
Location-based Services
Mobile Sensor Networks
Sensor Placement
Dissertations, Academic -- Computer Science -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: A new class of wireless sensor networks has recently appeared due to the pervasiness of cellular phones with embedded sensors, mobile Internet connectivity, and location technologies. This mobile wireless sensor network has the potential to address large-scale societal problems and improve the people's quality of life in a better, faster and less expensive fashion than current solutions based on static wireless sensor networks. Ubiquitous Sensing is the umbrella term used in this dissertation that encompasses location-based services, human-centric, and participatory sensing applications. At the same time, ubiquitous sensing applications are bringing a new series of challenging problems. This dissertation proposes and evaluates G-Sense, for Global-Sense, an architecture that integrates mobile and static wireless sensor networks, and addresses several new problems related to location-based services, participatory sensing, and human-centric sensing applications. G-Sense features the critical point algorithms, which are specific mechanisms to reduce the power consumption by continous sensing applications in cellular phones, and reduce the amount of data generated by these applications. As ubiquitous sensing applications have the potential to gather data from many users around the globe, G-Sense introduces a peer-to-peer system to interconnect sensing servers based on the locality of the data. Finally, this dissertation proposes and evaluates a multiobjective model and a hybrid evolutionary algorithm to address the efficient deployment of static wireless sensor nodes when monitoring critical areas of interest.
Thesis:
Disseration (Ph.D.)--University of South Florida, 2011.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by ALFREDO JOSE PEREZ.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 108 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
usfldc doi - E14-SFE0004889
usfldc handle - e14.4889
System ID:
SFS0028145:00001


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 22 Ka 4500
controlfield tag 007 cr-bnu---uuuuu
008 s2011 flu ob 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0004889
035
(OCoLC)
040
FHM
c FHM
049
FHMM
090
XX9999 (Online)
1 100
PEREZ, ALFREDO JOSE.
0 245
An architecture for global ubiquitous sensing
h [electronic resource] /
by ALFREDO JOSE PEREZ.
260
[Tampa, Fla] :
b University of South Florida,
2011.
500
Title from PDF of title page.
Document formatted into pages; contains 108 pages.
Includes vita.
502
Disseration
(Ph.D.)--University of South Florida, 2011.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: A new class of wireless sensor networks has recently appeared due to the pervasiness of cellular phones with embedded sensors, mobile Internet connectivity, and location technologies. This mobile wireless sensor network has the potential to address large-scale societal problems and improve the people's quality of life in a better, faster and less expensive fashion than current solutions based on static wireless sensor networks. Ubiquitous Sensing is the umbrella term used in this dissertation that encompasses location-based services, human-centric, and participatory sensing applications. At the same time, ubiquitous sensing applications are bringing a new series of challenging problems. This dissertation proposes and evaluates G-Sense, for Global-Sense, an architecture that integrates mobile and static wireless sensor networks, and addresses several new problems related to location-based services, participatory sensing, and human-centric sensing applications. G-Sense features the critical point algorithms, which are specific mechanisms to reduce the power consumption by continous sensing applications in cellular phones, and reduce the amount of data generated by these applications. As ubiquitous sensing applications have the potential to gather data from many users around the globe, G-Sense introduces a peer-to-peer system to interconnect sensing servers based on the locality of the data. Finally, this dissertation proposes and evaluates a multiobjective model and a hybrid evolutionary algorithm to address the efficient deployment of static wireless sensor nodes when monitoring critical areas of interest.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor:
Labrador, Miguel A..
653
Cellular Networks
Distributed Systems
Location-based Services
Mobile Sensor Networks
Sensor Placement
690
Dissertations, Academic
z USF
x Computer Science
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.4889



PAGE 1

AnArchitectureforGlobalUbiquitousSensing by AlfredoJ.Perez Adissertationsubmittedinpartialfulllment oftherequirementsforthedegreeof DoctorofPhilosophy DepartmentofComputerScienceandEngineering CollegeofEngineering UniversityofSouthFlorida MajorProfessor:MiguelA.Labrador,Ph.D. RafaelPerez,Ph.D. KenChristensen,Ph.D. WilfridoMoreno,Ph.D. AlfredoWeitzenfeld,Ph.D. AdrianaIamnitchi,Ph.D. DateofApproval: March31,2011 Keywords:CellularNetworks,LocationBasedServices,MobileSensorNetworks,Distibuted Systems,SensorPlacement Copyright c 2011,AlfredoJ.Perez

PAGE 2

TableofContents ListofTablesiv ListofFiguresiv Abstract viii Chapter1Introduction1 1.1WirelessSensorNetworks2 1.2StaticWirelessSensorNetworks2 1.2.1WirelessNodes3 1.2.2Basestations5 1.3MobileSensorNetworks7 1.3.1LocationTechnologies8 1.3.2CellularNetworks8 1.3.3MobileDevices9 1.4UbiquitousSensing11 1.5ProblemStatement13 1.6Contributions14 1.7StructureoftheDissertation16 Chapter2LiteratureReview17 2.1ArchitecturesforLocation-basedServices17 2.1.1Network-basedLocationProviderArchitectures18 2.1.2LocationProvider-basedLocationProviderArchitectures19 2.1.3Mobile-basedLocationProviderArchitectures19 2.2ArchitecturesforUbiquitousSensing22 2.3AnEnablingArchitectureforUbiquitousSensing25 Chapter3G-SenseArchitecture26 3.1HardwareArchitecture26 3.2Client-sideSoftwareArchitecture28 3.2.1SensorCommunicationandOSLayer29 3.2.2LocationManagementComponent29 3.2.3SensorManagementComponent30 3.2.4ServerCommunicationManagementComponent31 i

PAGE 3

3.3Server-sideSoftwareArchitecture33 3.3.1MobileWSNManagementComponent33 3.3.2StaticWSNManagementComponent34 3.3.3ServerSensingManagementComponent34 3.3.4DataCollectionandAnalysisComponents35 3.3.5DataVisualizationComponent35 3.4TheCriticalPointAlgorithms35 3.5APrototypeApplication40 3.5.1SystemArchitecture41 3.5.2HardwareInfrastructure43 3.5.3SoftwareInfrastructure44 3.5.4MobileClient46 3.5.5ServerApplication48 3.5.6ControlStation50 3.5.7StaticWirelessSensorNetwork52 Chapter4Geotella54 4.1Requirements54 4.2SystemArchitecture55 4.3ProtocolMessages55 4.3.1ProtocolHeader56 4.3.2Acknowledgements58 4.3.3Join58 4.3.4Accept58 4.3.5Reject58 4.3.6Ping59 4.3.7Pong59 4.3.8Geoquery59 4.3.9Geomessage60 4.3.10GeoqueryResponse60 4.3.11GeomessageResponse60 4.4ProtocolStates61 4.4.1Initalization61 4.4.2Ready62 4.4.3Geoquery/Geomessage62 4.4.4Maintenance63 4.4.5Failure63 4.5ProtocolImplementation64 Chapter5RelayPlacementinWSNsusingMultiobjectiveOptimization65 5.1MultiobjectiveOptimization65 5.2OptimizationModel67 5.3TheM-RESTAlgorithm70 5.3.1RepresentationoftheIndividuals72 5.3.2CrossoverandMutationOperators73 ii

PAGE 4

5.3.3FitnessFunction74 5.4LocalSearchHeuristics74 5.4.1CycleReductionSearch76 5.4.2BreathFarthestFirstSearch78 5.5Evaluation79 5.6RelatedWork84 Chapter6ConclusionsandFutureWork87 6.1Conclusions87 6.2FutureWork88 ListofReferences90 iii

PAGE 5

ListofTables Table1.1Commerciallyavailablemotes5 Table1.2Evolutionofcellularnetworkstechnology9 Table1.3Stateoftheartcellularphonesandcharacteristics11 Table1.4Classicationofapplicationsinubiquitoussensing12 Table3.1CommunicationprotocolsinG-Sense32 Table3.2Changeofdirectionscriticalpointalgorithmresults40 iv

PAGE 6

ListofFigures Figure1.1Hardwarecomponentsofawirelesssensordevice3 Figure1.2StargateNetbridgebasestation6 Figure1.3ApossiblescenarioofaWSNdeployment7 Figure1.4Hardwarecomponentsofacellularphonedevice10 Figure2.1Network-basedlocationproviderarchitectures18 Figure2.2Locationprovider-basedlocationproviderarchitectures19 Figure2.3Mobile-basedlocationproviderarchitectures20 Figure2.4ASMS/MMS-basedLBSarchitecture21 Figure2.5TraXmiddleware22 Figure3.1G-Sense'shardwarearchitecture27 Figure3.2G-Sense'sclient-sidesoftwarearchitecture29 Figure3.3G-Sense'sserver-sidesoftwarearchitecture34 Figure3.4Threebasiccriteriaforcriticalpointalgorithms36 Figure3.5ThecriticalpointalgorithmcodiedasaJavamethod37 Figure3.6Exampleofthecriticalpointalgorithminatrackingsession39 Figure3.7Generalarchitectureofthesystemprototype42 Figure3.8Hardwareforsystemprototype45 Figure3.9Mobileclientimplementation46 v

PAGE 7

Figure3.10Structureofdatagramssentbythemobileclient47 Figure3.11Serverapplicationimplementation48 Figure3.12Structureofdatagramssentbytheservertotheclient49 Figure3.13Webinterfaceforthemaincontrolstation50 Figure3.14Geo-Alertnoticationsequence51 Figure3.15StaticWSNimplementation52 Figure3.16Intrusionasseenbythecontrolstationandmobileclient53 Figure4.1Geotella'speer-to-peerarchitecture56 Figure4.2Geotellamessages57 Figure4.3Geotellaprotocolstates61 Figure4.4Geotella-connectedserversasshownincontrolstation64 Figure5.1RelationshipsinParetooptimization67 Figure5.2AprobleminstanceandWSNrelaytree68 Figure5.3AgeneralevolutionaryalgorithmcodiedasaJavamethod71 Figure5.4TheM-RESTalgorithmcodiedasaJavamethod72 Figure5.5Representationoftheindividuals73 Figure5.6Examplesofcrossoverandmutationoperations75 Figure5.7Cyclereductionprocedure78 Figure5.8AParetofrontbyM-RESTalgorithm79 Figure5.9TwoefcientrelayplacementsfoundbyM-RESTalgorithm80 Figure5.10Deployedrelaysinthe11x11grid.81 Figure5.11DeployedRelaysinthe35x35grid82 Figure5.12NormalizedParetofrontsinthe11x11Grid.83 vi

PAGE 8

Figure5.13NormalizedParetofrontsinthe35x35Grid.83 vii

PAGE 9

Abstract Anewclassofwirelesssensornetworkshasrecentlyappearedduetothepervasinessofcellularphoneswithembeddedsensors,mobileInternetconnectivity,andlocationtechnologies. Thismobilewirelesssensornetworkhasthepotentialtoaddresslarge-scalesocietalproblems andimprovethepeople'squalityoflifeinabetter,fasterandlessexpensivefashionthancurrentsolutionsbasedonstaticwirelesssensornetworks.UbiquitousSensingistheumbrella termusedinthisdissertationthatencompasseslocation-basedservices,human-centric,and participatorysensingapplications.Atthesametime,ubiquitoussensingapplicationsare bringinganewseriesofchallengingproblems. ThisdissertationproposesandevaluatesG-Sense,forGlobal-Sense,anarchitecturethatintegratesmobileandstaticwirelesssensornetworks,andaddressesseveralnewproblemsrelated tolocation-basedservices,participatorysensing,andhuman-centricsensingapplications. G-Sensefeaturesthecriticalpointalgorithms,whicharespecicmechanismstoreducethe powerconsumptionbycontinoussensingapplicationsincellularphones,andreducethe amountofdatageneratedbytheseapplications.Asubiquitoussensingapplicationshavethe potentialtogatherdatafrommanyusersaroundtheglobe,G-Senseintroducesapeer-to-peer systemtointerconnectsensingserversbasedonthelocalityofthedata.Finally,thisdissertationproposesandevaluatesamultiobjectivemodelandahybridevolutionaryalgorithm toaddresstheefcientdeploymentofstaticwirelesssensornodeswhenmonitoringcritical areasofinterest. viii

PAGE 10

Chapter1:Introduction Sinceitsbeginnings,thehumanperceptionoftheworldhasbeenshapedbytheinstruments thatareavailabletogatherdatafromtheenvironment.Naturehasprovidedhumanitywith asetofbasicsensorseyes,ears,nose,tongue,andskin,whichwehaveutilizedoverthousandsofyearstoexploretheworld.Advancesincomputers,micro-electro-mechanicalsystemsMEMS,andtheInternethaveprovidednewclassesofdevicesandtechnologiesthat haveexpandedourabilitytosensetheworldandcapturedataofinterestwiththenalgoalof extractinginformation. WirelessSensorNetworksWSNs areelectronicartifactsdeveloped forthatpurpose.TherearemanypracticalexamplesofWSNs,fromtrackingturtles[1]to betterunderstandtheirbehavior,tothemonitoringofbridgestructuresandhighways[2]. Wirelesssensornetworksprovideaexibleapproachtocapturedatafromthesurroundings, usingsmall,relativelycheapdevicesthatcanbeplacedeverywhere.However,severalissues haveavoidedthewidespreaddeploymentofWSNs,especiallyinlargescales.Ithasbeenobservedthatthestaticplacement,maintenanceanddeploymentcosts,communicationlimitationsandprogrammingdifculties,havedelayedandevencanceledmanyWSNdeployments. Ontheotherhand,thedevelopmentofcellulartechnologyoverthepastyearshasgivenhumankindadevicethatisfoundeverywhere.Combiningsensortechnologywithcellularphones andInternetconnectivity,inablinkofaneye,humankindhasamobile,pervasiveandpotentiallyenormouswirelesssensornetwork,whichnotonlyovercomesmostofthelimitations ofstaticWSNs,butalsoopensupthepossibilitytoaddresslarge-scalesocietalproblems,not possiblebefore. 1

PAGE 11

ThischapterbeginswithabriefoverviewofthetechnologyofstaticandmobileWSNs.Then, anintroductiontoubiquitoussensingwithcurrentapplicationsandchallengesisdiscussed. Next,theproblemstatementisintroduced.Finally,thechapterendswiththecontributions andthestructureofthedissertation. 1.1WirelessSensorNetworks Wirelesssensornetworksarecomputernetworkscomposedofsmall,battery-powereddevicesthataredeployedinareasofinterestforsensing,monitoring,andreportingdataabout events.Initially,WSNswereutilizedtoreportdataaboutenvironmentalvariables,butother applicationshaveemerged,andnow,WSNsarebeingusedinsecurity,military,health,construction,andmanyotherdomains. Thedevelopmentofwirelesssensornetworkscanbetrackedbacktothelate70'swithDARPA's DistributedSensorNetworksprogram[3].However,itwasnotuntilthelate90'swhenadvancementsinthetechnologyofMicro-Electro-MechanicalSystemsMEMS,microcontrollersandcommunications,thattheconceptofWSNsstartedtoemergeasaresearchareain computernetworksandpervasivesystems,andinitialapplicationsweredeveloped.Currently twotypesofWSNsexist:staticWSNsandmobileWSNs. 1.2StaticWirelessSensorNetworks AstaticWSNisasensornetworkthatisdeployedonaparticularsitetomonitortheareaof interest.Oncethewirelesssensordeviceshavebeendeployed,theydonotmove;insteadthey remainatthesameplaceforthelifetimeoftheapplication.WSNsconsistofwirelessnodes usuallycalledmotes,sensors,actuators,andbasestations.Thesecomponentsarebriey describednext. 2

PAGE 12

Figure1.1:Hardwarecomponentsofawirelesssensordevice 1.2.1WirelessNodes Wirelessnodesormotesaresmall,battery-poweredcomputerswithcommunicationcapabilities.InastaticWSN,motesarethedevicesthataredeployedthroughoutthemonitoredarea andtheirmainobjectiveistosenseandreportthevariablesortheeventsofinterest. Amoteiscomposedbythefollowinghardwarebuildingblocksgure1.1[4]: Memory:Usuallythememoryofamotevariesbetween4KBand32MB,beingthe deviceswithlessmemorythemostpopularduetotheirprice. Processor:Usuallythemotes'processingunitsarenotgeneralpurposemicroprocessorsbutmicrocontrollers,sincetheyaremoreenergy-efcienttoperformthetasks athand.ThemostpopularmicrocontrollersaretheonesbuiltbyATMELandTexas Instruments;however,thereareafewmotesthatutilizeARM-basedgeneralpurpose microprocessors. Radio:Duetotheirenergyconstraints,motesuseradioswithshortrangeconnectivity. UsuallysuchradiosfollowtheIEEE802.15.4[5]standardthatspeciesthephysical andMAClayersforlow-ratewirelesspersonalareanetworks.RadiosforWSNshave 3

PAGE 13

amaximumtransmissionrangethatvariesbetween20-30metersindoors,and75-100 metersoutdoors. Battery:Providestheenergytothemote.Thisisoneofthemostcriticalcomponents ofaWSNnodesincemostapplicationsexpecttheWSNtobeoperativeforlongperiodsoftime,sometimeyears.Topowercontinuouslymotes,inthelastyears energy harvesting methodshavebecomeoneimportantareaofresearchinWSNs. Sensors:Devicesthatmeasurethevariablesofinterest.Duetothedevelopmentof MEMStechnologyoverthelastyears,thereareseveralsensorsthatcanbeintegrated eitherinthesameboardofthemoteorinspecializedsensorboardsthatcanbeconnectedbyusingdigitaland/oranalogports. Dependingonthevendorandhardwarecharacteristicsofthemotes,theycanbeprogrammed usingdifferentprogramminglanguagesandtools.Themostpopularprograminglanguages arethefollowing: NesC:TheNetworkEmbeddedSystemCNesCisacomponent-based,event-driven programminglanguagethathasthesyntaxoftheCprogramminglanguage.Aprogram inNesChastwocomponents:modulesandcongurations.Modulesprovideandutilize interfacesdevelopedusingaC-likesyntax.Congurationsconnector"wire"themodules.ApplicationsdevelopedusingNesCareexecutedinanoperatingsystemcalled TinyOSexecutedwithonly4KBofRAMmemory.Duetothesmallmemory,TinyOS doesnotperformthetraditionalmemorymanagementfunctionsfoundinmodernoperatingsystems.Currently,TinyOS/NesCisopensourcesoftwaremaintainedbythe TinyOScommunity.Thewebsiteofthecommunityishttp://www.tinyos.net. C:Itsrstversionwasreleasedin1972atBellLabsbyDennisRitchie,oneoftherst developersoftheUnixOperatingsystem.In1989theCprogramminglanguagewas standardizedbyANSI/ISO.CurrentstandardistheISO-IEC-9899-1999.Thestandard andvariationsofthisprogramminglanguagehavebeenutilizedtoprogrammotes. 4

PAGE 14

Table1.1:Commerciallyavailablemotes MoteName Company Remarks TMote Sentilla TI8MHzCPU,radioisa250kbps2.4GHzIEEE802.15.4 compatible,NesCprogrammable Mica2 Memsic ATMEGA128LCPU,with4KRAM.Radiois315/433/916MHz frequency.NesCprogrammable MicaZ Memsic ATMEGA128LCPU,with4KRAM.Radiois2.4GHz IEEE802.15.4compatible,NesCprogrammable Imote2 Memsic IntelPXA271416MHzprocessorwithMMXcapabilities,32MB RAM,radioisa250kbps2.4GHzIEEE802.15.4.Thestandardis programmableinNesC,butthereisaMicrosoft.NETversion TelosB Memsic TI8MHzCPU,radioisa250kbps2.4GHzIEEE802.15.4 compatible,NesCprogrammable IRIS Memsic ATMEGA1281CPU,with4KRAM.Radiois2.4GHz IEEE802.15.4compatible,NesCprogrammable SUNSPOT SunMicrosystems 180MHz32bitARM920Tcore-512KRAM/4MFlashCPU. Radiois2.4GHzIEEE802.15.4compatible.J2MEProgrammable Perk Sentilla TIMSP430microcontrollerwithTI/ChipconCC2420 lowpowerwirelessradio,J2MEProgrammable Java:Javaisanobject-orientedprogramminglanguagedevelopedbySunMicrosystemsandrstreleasedin1995.TheJavaprogramminglanguageiscomposedbya seriesofspecications,beingtheJ2MEspecicationusedforportableandembedded devicestheoneutilizedtoprogramJava-basedmotes. Someofthecurrentlyavailablemotesaresummarizedintable1.1.MostofthemareprogrammedusingNesC/TinyOSandonlyafewareprogrammedinlanguagessuchasJavaor C.However,thereisatendencytobuildnewhardwarethatcanbeprogrammedusingJavaor CsincetheyavoidthesteeplearningcurveofNesC. 1.2.2Basestations BasestationsorsinknodesarespecialnodesintheWSNthatcollecttheinformationsent bythemotes.Theycanworkaloneorbeconnectedtoothertypeofnetworks.Whenused inthelattercase,basestationsaregatewaysthatsendthesensedinformationthroughother networks,suchastheInternet.Usually,basestationsarecomputersthathavemorepowerful characteristicsandresourcesthanthemotes.Figure1.2showstheMemsicStargateNetbridge 5

PAGE 15

Figure1.2:StargateNetbridgebasestation basestationnode,withawebcamconnectedtoit.Thisbasestationisacomputerrunningthe Linuxoperatingsystemanditisequippedwitha450MhzARMCPUand64MBRAM. Basestationscomeequippedwithseveralcommunicationinterfaces.Oneinterfacemustbe ofthesametypeofthemotessothebasestationreceivesandcollectsthedatasentbythe motes,whileotherinterfacesmaybeEthernet,WiFi,cellular,oranyothertypethatallows thecommunicationwithothernetworks.Basestationscanalsoprovideotherfunctionalities suchastheorganizationofthenodes,sendingcommandsbackandforth,andrespondto externalqueries.InatypicalWSNdeployment,thereisatleastonebasestationnode,but dependingonthesizeofthedeployednetworkorthesensingtask,theremightbemorethan onesinknodeinthenetwork.Ingure1.2thebasestationhasone802.15.4interfaceandan Ethernetinterface,connectingthebasestationtoalocalareanetwork. Apossiblescenarioforstaticwirelesssensornetworksisshowningure1.3[4].ThisexampleshowstheintegrationofWSNswithothertypesofnetworks,suchaswirednetworks, cellularnetworksandwirelessad-hocnetworks.Also,thisguredepictsthenetworktopol6

PAGE 16

Figure1.3:ApossiblescenarioofaWSNdeployment ogyofaWSN.UsuallystaticWSNsaredeployedinatreetopology,rootedatthebasestation node.Thisfactisbecausenotallthenodesneedtobeindirectcommunicationrangewiththe basestationastheycanmultihopthroughothermotes,relayingdataupwardsuntilreaching thebasestation.Acompletesurveyontheevolutionandstate-of-theartofstaticWSNscan befoundin[6]. 1.3MobileSensorNetworks Recently,anewkindofwirelesssensornetworkhasstartedtoemergeasaresultofthewidespreaduseofcellulartechnologyoverthelastdecade.Powerfulmobiledeviceswithfast processors,Internetconnectivityandstoragearemanufacturedtodaywithcharacteristicsthat outperformtheonesofthemotesinstaticWSNs.Combiningtheavailablesensorsintoday's phonesmicrophone,camera,GPS,andaccelerometersandthewidespreadadoptionof cellulartechnology,cellularphoneshavebecomeaninterestingplatformforWSNsresearch withmanypracticalapplications.MobileWSNsarepossiblethankstotheavailabilityof locationtechnologies,cellularnetworksandmobiledevices.Thesethreetechnologiesare brieydescribednext. 7

PAGE 17

1.3.1LocationTechnologies Locationtechnologiesprovidethecoordinatesofthegeographicalsitewheresensingistakingplace.Withoutlocationtechnologiestoreferencewheresensinginformationiscaptured itmakesnosensetocollectdata,sincethesedataarealwaysreectingsomethingofthereal world. TheGlobalPositioningSystemGPSisbyfarthemostutilizedmethodforobtainingthecoordinatesofadevice.MostofthecurrentlyavailablecellulardeviceshaveGPSchipembedded,obtainingGPSxeswithgoodaccuracy.TheGPSsystemworksbyhavingcertainnumberofsatellitessurroundingtheEarththatperiodicallysendbeaconpacketstotheground.By receivingthesepacketsfromatleastfoursatellites,aGPSchipisabletocalculatethecurrent coordinateofthedevice.ThemajorlimitationoftheGPSsystemisthatthissystemisunable workproperlyinsidebuildings,duetosignalattenuation.Inordertoaddressthisproblem, indoorlocationtechnologiessuchastheonedevelopedbySkyhook[7]areavailable. Theaccuracyofthelocationplaysanimportantroleinasensingapplication.Sometimesa coarse-grainedlocationisenoughsincethevariablesthatarebeingsenseddonotchange drasticallyfromonezonetoanother,whichisthecase,forexample,withthetemperature; however,ne-grainedlocationisneeded,forexample,inhealth-relatedandtrackingapplications. 1.3.2CellularNetworks Cellularnetworksareradio-basedcommunicationnetworksthatarebuiltuponcells,where eachcellisservedbyoneormorebasestationsxedatacertainlocation.Thesebasestations areinterconnectedamongthemselvesviaberopticormicrowavelinks,providingcommunicationcoverageoverageographicalarea.Smalldevicessuchasmobilephonesareconnected 8

PAGE 18

Table1.2:Evolutionofcellularnetworkstechnology Technology 1G 2G 3G 4G 5G Start/Deployment 1970/1984 1980/1999 1990/2002 2000/2010 2010/2015? DataBandwidth 2Kbps 14.4Kbps2Mbps 200Mbps+1Gbps 64Kbps 1Gbps Standards AMPS TDMA,CDMA, WCDMA WiMAX, Unied GSM-GPRS, CDMA2000 LTE,WiFi Standard? EDGE,1xRTT Comm.Technology Analog Digital CDMAandIP UniedIP UniedIP Multiplexing FDMA TDMA,CDMA CDMA CDMA CDMA tooneofthesecellsatatime,allowingthemtoplacecallsand/orsendandreceivedatapackets. TherstcellularnetworkwasdevelopedbyMotoroladuringthe70's,supportingvoicetrafc only.Thisrstnetworkutilizedanalogcommunication,andwasknownastheAdvanced MobilePhoneSystemAMPS.Itutilizedfrequencydivisionmultiplexingtodividethebandwidthinsmallchannels,eachoftheseusedbyadifferentdevice.Sincethen,severaltechnologieshavebeendeveloped,eachofthemgroupedtogetherintogenerationsofcellular communicationsystems.ThecurrentgenerationsofcellularnetworksarebasedontheGlobal SystemforMobileCommunicationsGSMandCodeDivisionMultipleAccessCDMA technologies,supportingdatatrafcaswellasvoicetrafc.Table1.2showsasummaryof technologiesforeachgenerationofcellularnetworks[8]. 1.3.3MobileDevices Thesensingtaskinamobilesensornetworkisperformedbythemobiledevices.Usually, thesearepocket-sizeportablecomputerswithmorepowerfulcapabilitiesthantheirstatic WSNcounterparts.ThemostpopularmobiledevicesinamobileWSNarecellularphones; however,thesearenottheonlymobiledevicesthatcanbeutilizedformobilesensing,since anetbook,oranyotherembeddedcomputerwithacellularantennaand/orad-hocnetwork connectivitycanserveasasensingdevice. 9

PAGE 19

Figure1.4:Hardwarecomponentsofacellularphonedevice Theinternalarchitectureofacellularphonedeviceissimilartothearchitectureofamote; however,ascellularphonesweredesignedtohandlevoicecalls,theyhavespecializedcircuits digitalsignalprocessorsthatperformcertaintasks.Thegeneralhardwarearchitectureofa cellularphoneisdepictedingure1.4[9]. Fromthesoftwareperspective,themostcommonlanguagetoprogramcellularphonesisthe Javaprograminglanguage,usedinJ2MEandGoogleAndroidplatforms.Currentoperating systemsformobiledevicessupportmultitaskingwhichallowsapplicationstoruninthebackground,importantinapplicationsdevelopedforcontinuoussensing.Similarastheirstatic WSNcounterpart,cellularphonesarealsobatterypowered,whichcallsforapplicationsto managethisresourcecarefully.Table1.3[10]showsthecharacteristicsofvestate-of-the artcellularphones. 10

PAGE 20

Table1.3:Stateoftheartcellularphonesandcharacteristics Device NexusOne IPhone4 N900 Cliq BlackBerry8900 Manufacturer Google/HTC Apple Nokia Motorola RIM ProcessorClock 1Ghz 1Ghz 600Mhz 528Mhz 512Mhz RAMMemory 512MB 512MB 256MB 256MB 128MB Storage 32GB 32GB 32GB 32GB 32GB MicroSD MicroSD MicroSD MicroSD Communication 3G/WiFi 3G/WiFi 3G/WiFi 3G/WiFi 3G/WiFi GPS yes yes yes yes yes Acceloremeter yes yes yes yes no OperatingSystem Android iOS Maemo Android BlackBerryOS Prog.Interface Java ObjectC C/C++ Java Java 1.4UbiquitousSensing UbiquitoussensingUSencompassestheintegrationofdifferentsensordatasourcesstatic andmobileWSNs,anditisthenextstepintheevolutionofwirelesssensornetworksresearch.UScanbedividedinthreemajorareas[15]: Location-BasedServicesLBS:Systemsandapplicationsthatutilizeonlythelocationofthedevicetoobtaininformation.Usuallythesesystemsarerequest-response systems.LBSsystemsaredescribedin[9,16]. Human-CentricSensingHCS:Systemsdevelopedtomonitorthephysiologicalvariablese.g.pulse,temperature,heartelectricalactivity,breathrateanddepthofaperson.Thegoalofthesesystemsistomonitorthesevariablesandidentifythecontext, activitiesandpossiblehealththreateningsituationsofapersonoragroupofpeople. Examplesofthesesystemsaredescribedin[23]provideexamplesofthisclassof systems. ParticipatorySensingsystemsPS:inparticipatorysensing,communitiescollectand sharesensordatausingmobilephonesandstaticWSNs.Thisconceptcanbeextended toavarietyofvariables,forexample,airqualityandnoise.Thegoalofthesesystems istoextractusefulinformationtoimprovethequalityoflifeofthepeopleandtheir communities.Examplesofthesessystemsaredescribedin[30]. 11

PAGE 21

Table1.4:Classicationofapplicationsinubiquitoussensing DataDelivery/DataCollection Continuous Event-based Real-time LBS/HCS/PStracking PS/HCSemergencies Delay-tolerant HCS/PSdatastudy HCS/PSDatastudy WhereasUSapplicationscanbegroupedinthethreemajorgroupsmentionedbefore,abetter characterizationoftheclassesofapplicationsinUSisbygroupingthembytheirdatacollectionapproachcontinuousorevent-based,anddatadeliveryreal-timeordelay-tolerant. Thedatacollectioncategoryspecieshowthedataiscapturedfromthesensors.Assuch, thedevicemaycollectdatacontinuously,orthedatacanbecapturedbasedoneventsthat indicatethemomentorthecontexttoperformthedatacollection.Oncethedataarecaptured, theuploadingtotheserverscanhappenassoonaspossiblewithfeedbackfromtheserver immediatelyreal-time,oritcanbedeliveredtotheserverslaterdelay-tolerant. Withthesecategories,USapplicationscanbeclassiedasshownintable1.4.Incontinuous real-timeapplications,thedeviceisalwayscollectingdataanditsendssuchdatatotheserver assoonaspossible.TrackingapplicationsinLBS,HCS,andPSfallintothiscategory.Onthe otherhand,PS/HCSrelatedtoemergencymanagementfallintotheevent-basedreal-time categorysincetheydonotneedtocollectthedataallthetimebutonlywhencertainconditionsthatdenetheemergencyaremet.Finally,HCS/PSapplicationsthatdonotrequire immediatefeedback/actionfromtheserversfallintothedelay-tolerantdatadeliverycategory. LBSsystemsweretherstclassofsystemsdevelopedinUS,withapplicationsrangingfrom thetrackingofpersonsandobjects,uptondingthebestroutestotravelbyavehiclefrom oneplacetoanother.However,inthelastyearsPSapplicationsarebecomingpopular,through websitessuchasPanoramioandGooglemapsthatgivepeopleaplacetosharepictures,videos andcommentsbyplacingthemoveramapthatshowsthegeographicalplacewherethecontentwasgenerated.Inthenearfuture,newapplicationscombiningcollectedsensordataof anykindandsocialnetworkswillbecomepopular. 12

PAGE 22

1.5ProblemStatement UbiquitousSensingapplicationshavethepotentialtogatherdatafromstaticand/ormobile sensorslocatedanywhereintheworld.Thisbringsinterestingchallengestoovercomein ordertosuccessfullydeploythesenetworks.Assuch,thechallengesthatthisdissertation addressesare: CurrentarchitecturesforUSmostlysupportoneclassofapplications.Thedesignof ageneralarchitectureforUSisimportantbecauseitwillallowthereutilizationof elements,insteadofdesigninganddevelopingwholesystemsfromscratch,accelerating thedeploymentofUS-relatedprojects.Moreover,insomecontext,suchasmilitaryor emergencymanagement,theapplicationsareacombinationofmorethanoneofthe categoriesshownintable1.4.Theseapplicationsrequireageneralpurposearchitecture,insteadofhavingseparatesystems. Utilizationofcellularphonesforcontinuoussensingandreporting.Ascellularphones arebatterypowered,mechanismsareneededtomakeUSapplicationspowerefcient, whilebeingabletoperformtheirsensingtasksandreportinginareliableandtimely manner. IntegrationofstaticandmobileWSNsandtheefcientdeploymentofstaticsensor nodes.AlthoughstaticWSNsweredesignedasnetworkswithpotentiallythousands ofcheapdevices,currentlydeployedWSNsremainrathersmallnetworks.Asstatic WSNsareanimportantcomponentofubiquitoussensing,astheycanbedeployedin areaswhereamobileWSNsisnotsuitablee.g,rivers,factories,forests,volcanoes, thecarefulplacementoftheWSNnodesisanimportantproblem.Giventhelocations thataredesiredtosense,howmanynodesareneededandwherethesenodeshavetobe placedsotheleastamountofenergyisdissipated? 13

PAGE 23

Scalability.Ubiquitoussensingapplicationshavethepotentialtogatherinformation frommanyusersaroundtheworldsimultaneously.Consideranapplicationdeployed worldwidethatcollectslocationinformationfromeverysinglecellularphoneandeach ofthesedevicessendsabout40bytesofdatapersecond.Astherearearound5billion ofmobilecellularsubscriptions[35]aroundtheworld,theapplicationwouldgenerate 20billionbytesaround18GBoftrafcpersecondintheInternet.Howcanubiquitoussensingsystemsbescalableforglobalsensing? Otherimportantissues,notaddressedinthisdissertation,arerelatedtoprivacyandsecurity, incentiveforparticipation,anddatavalidityandvisualization,amongthemostimportant ones. 1.6Contributions Thisdissertationaddressesthemainproblemsdescribedbeforeandincludesthefollowing contributions: G-Sense,anewarchitectureforglobalsensing. ThisdissertationpresentsG-Sense[15], forGlobal-Sense,anarchitecturethatintegratesmobileandstaticwirelesssensornetworksinsupportoflarge-scalelocation-basedservices,participatorysensing,and human-centricsensingapplications.G-Senseincludesspecicmechanismstocontrol theamountofdatageneratedbytheseapplicationswhilemeetingtheapplicationrequirements.IncomparisonwithpreviousUSarchitectureswhichreliedinclient-server systems,G-Senseutilizesahybridclient-serverandpeer-to-peerarchitecturetomanage thescalabilitywhileenablingdatacollectionfromdifferentsources,includingstatic WSNsandotherservers.Inthisdissertationthedesignofthearchitectureisdescribed, alongwithanexampleofasystemthatimplementsG-Sense. Criticalpointalgorithms. Themaingoalofthesealgorithmsistoreducethepower consumptionofthecellulardevicewhenutilizedforcontinuoussensing.Thecritical 14

PAGE 24

pointalgorithmsdecidewhentoupdatethelocationdatatotheserver,thereforethe batterypowerofthedeviceisextended,andlessdataissentoverthecellularnetwork. Theutilizationofthecriticalpointsalgorithmsaccomplishesasecondaryobjective, whichistheminimizationofdatastoredindatabases. Geotella,ageographic-awarepeer-to-peersystem. InordertomakeG-Sensescalable, apeer-to-peerarchitectureisintroducedbasedonthelocalityofthedata.Thepeer-topeersystemthatimplementsthearchitectureiscalledGeotella.Geotellainterconnects individualserversandcreatesasensingoverlayovertheInternetinwhicheachofthe serversbecomesaplaceforsensordataaggregation,providesthefunctionalitytodistributetasksamongseveralplaces,andcollectsdatafromsuchtasks. AmultiobjectiveapproachtotherelayplacementprobleminstaticWSNs. Inthisdissertation,amultiobjectiveParetooptimizationmodelhasbeendevelopedtoexplore thetradeoffintheminimizationoftwogoalswhendeployingastaticWSN:thenumber ofsensorsandtheenergydissipatedbytheWSN.Tosolvetheoptimizationmodel,a hybridevolutionaryalgorithmisproposedandimplemented.Theresultingoptimization algorithmprovidesaframeworktosolveamultiobjectiveversionoftheSteinerTree ProblemaNP-CproblemndinganapproximationtotheParetofront.Theproposed approachservesasatoolfortheWSNdesigner/engineertoexplorethetradeoffbetweenthetwoobjectives. AnewbookonLocation-BasedSystems. Partofthedevelopmentexperienceobtained duringthestudyoflocation-basedsystemsandmobilesensinghasbeenpublished in2010byTaylorandFrancis,Chapman&Hall/CRCinthebook "Location-Based InformationSystems:DevelopingReal-TimeTrackingApplications" ,writtenbyMiguel A.Labrador,AlfredoJ.PerezandPedroM.Wightman.Thisbookisatextbookforan undergraduate/graduatecourseinlocation-basedinformationsystems. 15

PAGE 25

1.7StructureoftheDissertation Therestofthedissertationisstructuredasfollows.Chapter2containstheliteraturereview ofcurrentarchitecturesforubiquitoussensing.Chapter3introducestheG-Sensearchitecture andthecriticalpointalgorithms.GeotellaisintroducedinChapter4.Chapter5presentsthe multiobjectiveapproachtotherelayplacementproblemalongwiththeproposedalgorithmto optimizetheParetomodel.Chapter6concludesthedissertationandpresentspossibleways toextendtheresearchinthisareainthefuture. 16

PAGE 26

Chapter2:LiteratureReview Beforetheconceptofmobilesensornetworksbecamepopular,LBSprojectsandapplications hadbeendevelopedastherstclassofapplicationsthatmadeuseofthesensorsincluded inthecellulardevices.However,LBSsystemswereseenonlyasapplicationsdesignedto requestdatabasedonthecurrentlocationofthedevice[16,20,21,36],ortokeeptrackof personsorobjectse.g.peopletracking,eetmanagement[9,17].Asaconsequence, LBSarchitecturesdidnotincludeinterfaceswithstaticWSNs.Ontheotherhand,theUS communityhasdevelopedmobileWSNsarchitecturesfocusedmainlyindelay-tolerantapplications[34,37,38].Assuch,currentarchitecturesforUShavebeendesignedtoaddress certaintypeofapplicationsinsteadofdevelopingageneralpurposearchitecture.AsframeworksforstaticWSNshavebeendevelopedoverthelasttenyears[39],thischaptermainly describesarchitecturesthatsupportLBSandmobileWSNs. 2.1ArchitecturesforLocation-basedServices Duringthelastdecade,theimprovementoflocationsystemshaspushedthedevelopment ofLBSapplications.Forthedevelopmentofthesesystems,threemajorarchitectureshave beenproposed.Thesearethenetwork-based,mobile-based,andlocationprovider-based architectures[9].Thethreearchitecturesaredescribednext. 17

PAGE 27

Figure2.1:Network-basedlocationproviderarchitectures 2.1.1Network-basedLocationProviderArchitectures Innetwork-basedarchitectures,theaccessoflocationdatabythird-partyserviceproviders andthepositioningtechniquestolocatethedevicearecontrolledbythenetworkprovider. Onthesesystems,theaccuracyofthelocationdependsofthetechnologydeployedbythe network,whichisbasedinanyofthefollowinglocationtechnologies:Cell-Id,EnhancedObservedTimeDifferenceE-OTD,ObservedTimeDifferenceofArrivalOTDDoA,Uplink TimeDifferenceofArrivalU-TDoA,orAssistedGPSA-GPS. Figure2.1depictsthetypicalarchitectureforanetwork-basedLBSsystem.Inthegure,the locationofauserisobtainedthroughanynetwork-centricpositioningmethodandathirdpartyserviceproviderobtainsthelocationofthemobiledevicethroughtheGatewayMobile LocationCenter.TheLBSproviderexecutestheservice,returningtheinformationtothe mobileuserortoexternaluserse.g.parentstrackingtheirkids. Becausemostnetworkoperatorsdidnotmakethelocationdataaffordabletothirdusers[18] andthehighcostofdeployingadvancedpositioningmethodsworldwide,network-basedLBS systemswerenotasuccess.However,thereweresomesystemsdeployed,suchastheNTTDoCoMoLBSsystem,whichwastherstcommercialLBSsystemavailable[16]. 18

PAGE 28

Figure2.2:Locationprovider-basedlocationproviderarchitectures 2.1.2LocationProvider-basedLocationProviderArchitectures Underthisarchitecture,anindependententitycollectstheusers'locationsandservesaslocationprovidertothird-partyapplications.Asshowningure2.2,locationprovider-based architecturesworkbyobtainingthedevice'slocationeitherfromthemobiledeviceitselfor fromthenetworkprovider'slocationtechnology,andaftercachingthepositiondata,they providethelocationstoexternalLBSserviceproviders.Onthesesystems,theprivacycanbe controlledbytheuser,ashecansetrulestowhomitslocationwillbedisclosed.Examplesof thesesystemsareGoogleLatitude[40],YahooFireEagle[41]andFacebookPlaces[42]. 2.1.3Mobile-basedLocationProviderArchitectures Inmobile-basedarchitecturesgure2.3,thedeviceutilizesself-positioningmethodse.g. GPStoobtainitslocationandtheuserexplicitlyallowsexternalservicestoaccessthelocationdata.Inthesesystems,thecellularnetworkactsonlyasthecommunicationlinkbetween themobiledeviceandtheInternet.AsLBSsystemsbasedonthisarchitectureobtainsthe locationdirectlyfromthedevice,theyneedthedevelopmentofLBSmiddlewarewithcomponentsforbothmobiledeviceandserviceprovider. 19

PAGE 29

Figure2.3:Mobile-basedlocationproviderarchitectures Thecommunicationmethodshaveimprovedaccordinglywiththeevolutionofcellularnetworksanddevices.Initialmobile-basedLBSreliedonSMSshort-messagingsystemand MMSmultimediamessagingsystemtocommunicatewithLBSproviders.Nowadayswith 3GtechnologiesandmobileInternetconnectivity,LBSmiddlewareutilizetheTCP/IPprotocolstack. Figure2.4depictsamobile-basedarchitecturedevelopedtosendgeotaggedmultimediadata usingtheSMS/MMSplatform[17].Aserviceinvocationusingthisarchitecturestartswhen thedevicesendsageotaggedMMSmessage.Themessageistransformedintoane-mail message,anditisforwardedtoane-mailserverbythecellularnetworkprovider.Uponreceptionbythee-mailserver,theserviceproviderpullsthemessageandprocessthequery. Theprocessnisheswhentheserviceproviderexecutesthequeryandsendstheresponse backtothemobiledevice. Whereasexibleinthetypeofdatathatitcanhandle,thisarchitecturegeneratessignicant overheadinthemessagesduetotheMIMEformatrequiredbytheMMSsystem.Thismakes thearchitecturenotsuitableforcontinuousupdates,sincethedeviceneedsextraenergyto transformmultimediadatatobase64encodingandsendthemessagewiththeoverheadover thewirelessnetwork.Unexpecteddelayscausedbythenetworkloadinthecellularprovider's MMSsystem,makesthisarchitecturenotsuitableforthedevelopmentofreal-timesystems. 20

PAGE 30

Figure2.4:ASMS/MMS-basedLBSarchitecture TheTraXframework[18]gure2.5isanotherexampleofanimplementationofmobilebasedarchitectures.Developedasaclient-servermiddleware,thisframeworkiscomposedby fourmajorcomponents,thesearethepositioning,thepositionmanagement,theservice,and theapplicationmanagementlayers. Thepositioningandpositionmanagementlayersimplementmethodstoobtainthelocation informationandtrackthedevice.Moreover,thepositionmanagementlayerdenesthelocationupdatemechanismsavailabletotheapplicationimplementedbyboththeserverand device.Theservicelayeroftheframework,whichisimplementedontheserverinterfaces withservicesofferedbyotherprovidersusingwell-deneddataexchangemethodssuchas Webservices.Finally,theapplicationlayercorrespondstotheapplicationsthatcanutilize theframework.TraXutilizesHTTP/TCPforallitsdatatransfersanditassumesacentralized serverthatreceivesallthelocationinformation.ClearlydenedasaLBSframework,TraX doesnotdeneinterfacestoaccessstaticWSNsorothersensorsavailableinthedevice. Developedasaframeworktosupportreal-timeLBSapplicationsforJ2ME-basedmobile phones,theLAYSICframework[43]isamobile-basedarchitecturesthatutilizestheHTTP andtheUDPprotocolstosupportintelligentLBSapplications.Inthisarchitecture,theusers 21

PAGE 31

Figure2.5:TraXmiddleware loginusingHTTPandthelocationupdatesaresentusingtheUDPprotocol.Tosendother anyotherdata,LAYSICutilizesHTTP.Theframeworkincludesmodulestosaveenergy inthemobiledevice,bynotperformingupdateswhenthecellularsignalistooweak,the utilizationofastatemachinethatdeactivatestheGPSwhentheuserremainsatthesame siteorwhentheuserisinsidebuildings,andmechanismstodecidewhentouploaddata totheserver.AsLAISYCwasdevelopedasaLBSframework,thereisnosupportforstatic WSNsandtheframeworkdoesnotdenehowtoaccessotheravailablesensorsinthecellular device.Fromtheserver-sidepointofview,LAYSICdoesnotdeneothercommunication interfaceswithexternalservers.Theframeworkalsoassumesthatithasauniquecentralized server,whichdoesnotallowsystemsdevelopedunderthisframeworktoscalewhenmany usersareconnected. 2.2ArchitecturesforUbiquitousSensing UbiquitousSensingisatermforthestudyontheintegrationofdatacollectedthroughstatic andmobilesensornetworksthatareanywhere[44].Itaddressestechnicalaspectssuchas privacy,networking,andinformationextraction,andsociologicalaspectssuchasusage, participation,andcollaborationbycommunities.Asitwasdescribedinsection1.4,USapplicationscanbeclassiedintolocation-basedsystemsLBS,human-centricsystemsHCS, andparticipatorysensingsystemsPS. 22

PAGE 32

FromthepointofviewofLBSsystems,ubiquitoussensingexpandmobile-basedlocation providersystemstosupportstaticWSNsandothersensorsthatcanbeconnectedtothemobiledevice.Nevertheless,theyaddressratherdifferentapplications.UsuallyinLBSsystems, usersrequestinformationregardingtheirgeographicalcontextandeachuserisseenasan independententityfromeachother.ThisisdifferentinUSapplicationssinceusersmayshare orcollaboratewiththedatathattheycollect.Thus,onecandescribeLBSsystemsasoneof thetypeofsystemsthatcanbedevelopedinUSwhentheusersaretreatedindependently,and theonlysensorthatthemobiledeviceusesisaGPSorothersystemthatprovidesitslocation. Ontheotherhand,mostoftheUSarchitecturesdesignedsofarhavebeendevelopedtosupportdelay-tolerantmobileWSNsapplications,whichareonlyonetypeofapplicationsthat canbedevelopedinUS.Withintheseapplications,twomethodologiestocharacterizethe roleassumedbythepeopleinthesesystemshavebeenproposed.Thesearethe Opportunistic Sensing [33]and ParticipatorySensing [45]methodologies. Intherst,thedataarecollectedwithouthumanintervention,thisis,theapplicationidenties byitselfthecorrectcontextandcapturesthedata,whileinthesecondtype,thepersonconsciouslytakesthemeasurement.Thissectionpresentsanoverviewofcurrentarchitecturesin supportofthesemethodologies. TheMetroSensearchitecture[34]proposedatDartmouthCollegeisaframeworkdeveloped forUSusingopportunisticdatacollection.Threemajorcomponentsconstitutethisarchitecture,thesearethesensors,thesensoraccesspoints,andtheservers.Inthisarchitecture,the cellulardevicesandWSN'sbasestationsarepartofthesensoraccesspointswhichcollectthe datameasuredbythesensors,andtheservershavetheresponsibilitytostoreandprovidedata analysis.MetroSenseutilizecellularphonesmostlyasdatamulesthatcollectinformation fordelay-tolerantsensingtasks,andtheyuploadtheinformationwhencommunicationwith serversispossible. SimilartoMetroSense,theCarTelproject[46]proposedbytheMITisanopportunisticdelaytolerantarchitecturethatcollects,analyzesandvisualizessensordatafromembeddedcom23

PAGE 33

putersmountedoncars.Thesystemconsistsoffourcomponents,thesearetheon-carmanagementmodulecalledICEDB,whichcollectsandprioritizesinformationonthemounted computers,aprotocolcalledCafNetthatperformscarry-and-forwarddatadeliverytothecentralserverandusingasdatatransportintermittentlyconnectednetworks,andthevisualization portalwhichservesasthecontrolstationandvisualizationcomponent.Thisarchitecturedoes notincludesstaticWSNsassensingelements,anditisnotdesignedforreal-timesensing. UndertheUrbanSensingproject[37],researchersatCENSUCLAhavebuiltanarchitecture tosupportPSapplicationsinurbanenvironments.Fouractorscomposethearchitecture,these arethesensors,subscribers,registries,andmediators.BecauseofthefocusinPSnetworks, thisprojectaddressesissuessuchasprivacy,vericationandauthenticationofcollecteddata, andrulesfordissemination.Assuch,thisarchitecturecollectsdatausingdelay-tolerantmobileWSNsandreliesinHTTPtotransferdatatotheserver. WiththeconceptofvirtualtriplinesVTP[47],researchersattheUniversityofCaliforniaBerkeleydevelopedtheMobileMillenniumProjecttoestimatetrafconroadsusingcellular phonesmountedincars.VTPweredesignedtoaddressprivacyandavoidcontinuoussensing,astheVTPweredesignatedspacesinroadswherethecarreporteditsvelocityandother variablestotheserver,usingtheHTTPprotocol. MicrosoftResearch'sSenseWebproject[48]developedanarchitecturethatintegratesWSNs aroundtheplanetthroughtheutilizationofwebservices.ThegoalinSenseWebwastodevelopadatacollectorwebsitewheresensorinformationcouldbeeasilyshared.Fourmajor componentscomposethisarchitecture,thesearetheGeoDBageo-indexeddatabasethat storesmetadataaboutavailablesensordatasources,theDataHubawebserviceinterface thatexposesmethodstostoresensordata,theIconDcreatesimagesandiconsthataggregatethedatacollected,andtheSensorMapwebsiteforend-userstoperformqueriesand showresultsonweb2.0maps.Whereasexibleduetotheuseofwebservices,themajor drawbackofthisarchitectureistheutilizationofthemforcellularphonesincontinuoussensing,sincewebservicesrequiretheHTTPprotocol. 24

PAGE 34

2.3AnEnablingArchitectureforUbiquitousSensing Whereasmostofthepreviousarchitectureshavefocusedonsupportingdelay-tolerantapplicationandsystems,real-timeapplicationsareveryimportantsincetheycanalertaboutthe changeoftheconditionsintheenvironmentand/oraperson'shealthstate.Bysupporting thesereal-timesystems,anarchitecturewouldallowtoassessdangeroussituationsandtake controlofthembeforetheybecomecatastrophic.ThereforeanenablingarchitectureforUS shouldsupportbothtypesoftrafc:real-timeanddelay-tolerantsensing,andbothinascalablemanner. Fromthispointofview,LBSarchitecturesseemthemostsuitabletosupportUS,however sucharchitecturesdonotdeneinterfaceswithstaticWSNs,andtheysufferfromthescalabilityproblem.Whereasafederatedarchitecturewouldscalesuchsystems,inrapidchanging scenariossuchasbattleeldsorinareaswithcatastophes,amoreexibleapproachtocollect real-timedataisneeded. ThemajorcontributionsofG-Senseasanenablingarchitectureforubiquitoussensingisto provideaframeworktosupportbothreal-timeanddelay-tolerantapplicationsinascalable mannerbymeansofapeer-to-peersystemandalsoaddresstheefcientintegrationofstatic wirelesssensornetworks.Inthenextchaptersthecomponentswillbedenedanddescribed. 25

PAGE 35

Chapter3:G-SenseArchitecture ThischapterpresentstheG-Sensearchitecture[15],atwo-tierclient-serverandpeer-to-peer architecturethatsupportsthedevelopmentofLBS,PS,andHCSapplicationsataglobalscale.Thearchitectureintegratesmobilesensingdevices,staticwirelesssensorsnetworks, andserverstoperformdatacollectionfromtheenvironmentandtheindividual,analyzethe dataatdifferentlevelsofthearchitecture,makeestimationsandinferences,andprovidefeedbackandvisualizationcapabilities.Thischapterbeginswithspecicationofthehardware andsoftwarecomponentsofthearchitecture.Next,thecriticalpointalgorithmsareintroduced.ThechapterendswiththedescriptionofasystemprototypethatimplementstheGSensearchitecture. 3.1HardwareArchitecture G-Sense'shigh-levelhardwarearchitectureconsistsofthefollowingcomponentsgure3.1: Sensingdevices:Thispartofthearchitectureconsistsofalltypesofsensorsutilizedby theapplicationsandtheinterfacesavailabletotransferthedatafromthesensorstothe rstlevelintegratingdevice,whichiseitherthemobiledeviceorthebasestation.This pertainstothedatacollectionfunction. Firstlevelintegrator:Thisisthedevicethatcollectsalldatasentbythesensors.Inthe architectureitisrepresentedbyeitheracellularphoneoralaptop,orthebasestation inthecaseofaWSN.Thiscomponentmayperforminitialdataanalysisoverthedata collectedbythesensorsforimmediatefeedback. 26

PAGE 36

Figure3.1:G-Sense'shardwarearchitecture 27

PAGE 37

Datatransportnetwork:ThedatatransportnetworkisthecombinationofIP-based networkingtechnologiesthatmakepossiblethetransferofdatafromtherstlevel integratordevicetotheservers. Servers:Thiscomponentstoresthedatacollectedfromthemobilenodesandstatic WSNnodes.Inaddition,itmakesadditionalprocessingonthedatatoperformestimationsandinferencesthatcannotbeperformedneitherinthemobilenodenorthebase stationduetodataavailabilityand/orthecomplexityofthealgorithms.Thiscomponent alsosupportsdatavisualizationusingWeb2.0toolsforreportingpurposes. Basedonthishardwarearchitecture,G-SenseprovidesasoftwarearchitecturethatcansupportdifferentUSapplicationsthroughahybridclient-serverandpeer-to-peerarchitecture, wheretheclient-serverpartabstractsthebasestationandmobilesensingnodesasclients thatconnectstoaserver,whichisamorepowerfulcomputersystemintermsofprocessing,energy,communication,andstoragecapabilities.Thearchitectureshowningure3.1 ismeanttohavearatherlocalscopeandbereplicatedinasmanyothersitesasnecessary. Then,serversfromdifferentsitesareinterconnectedamongthemselvesusingapeer-to-peer system,sothesensingtasksaredistributedforloadbalancing,trafcltering,andscalability purposes.Thispeer-to-peersystem,whichwillbeexplainedinthenextchapter,istheone thatallowsfortheglobaldeploymentofLBS,PS,andHCSapplicationsandmanagesthe trafcinthenetwork.Thesoftwarearchitecturethatsupportsthisplatformisdescribednext. 3.2Client-sideSoftwareArchitecture G-Sense'ssoftwarearchitecturefortheclientdevicesisshowningure3.2.Itconsistsof fourmajorcomponentsknownastheSensorCommunicationandOSlayer,LocationManagement,SensorManagement,andServerCommunicationManagementcomponents.The SensorCommunicationandOSlayeristheclosesttothehardwareandiscommontoallthe othercomponents,whicharemorespecicintermsofthefunctionthattheyperform. 28

PAGE 38

Figure3.2:G-Sense'sclient-sidesoftwarearchitecture 3.2.1SensorCommunicationandOSLayer TheSensorCommunicationandOSlayerabstractthecommunicationwiththesensorsthat areconnectedtothemobiledevice.Currently,cellularphonesandlaptopscomeequipped withWi-FiandBluetoothinterfacesthatcanbeusedtointegrateexternalsensorsalready inthemarketwithsuchcommunicationcapabilities.Examplesofcommerciallyavailable devicesthatcanbeusedtomeasurebodyvariablesaretheZephyr'sBioHarnessstrap,the NoninPulseOxymeter,andtheAliveHeartandActivityMonitor,amongothers.Morespecializedsensors,suchasthosemeasuringgases,temperature,humidity,etc.,areusuallybuilt inspecialcircuitboardsequippedwithsuchnetworkinterfaces.Othersensors,suchasaccelerometersandGPSchipsarecomingalreadyintegratedinthemobiledevice. 3.2.2LocationManagementComponent TheLocationManagementcomponentprovidesthemobilesensingapplicationwithinformationaboutthepositionofthedevice.Aslocationinformationisimportanttoanymobile 29

PAGE 39

sensingapplication,thecarefulmanagementofthisinformationisrequired.Forthisreason G-Sensemanageslocationinformationinaspecialcomponentthatconsistsofthreemodules. LocationAcquisitionModule:Thismoduleobtainsthepositionofthedevice.Example implementationsofthismodulearetheJavaLocationAPIJSR-179or293forJava ME,theAndroid'sLocationAPI,anycustommadeGPSwrapper,orlocationservices, suchastheoneprovidedbySkyhook,orsimilarcompanies.Therefore,thissubcomponentabstractsthepositioningmethodsanddevicesGPSthatthemobilesensing devicecanusetoobtainalocation. LocationEstimationModule:Themainobjectiveofthismoduleistocombinethe locationinformationfromdifferentsourcesthatthedevicegathersusingtheLocation Acquisitionmoduleandprovideabetterestimateoftheunit'sposition.Thismodule couldutilizetechnologieslikedeadreckoning,orusehistoricaldataaboutthetravel patternsoftheusertoobtainorimprovethepositionoftheunit. CriticalPointManagerModule:TheCriticalPointManagerdecideswhetherthecurrentsensedvaluesareworthtosendtotheserverornot.Thisdecisionisbasedupon themobiledevice'spositionanditisoneofthemostimportantmodulesintheclient's softwarearchitecture,asitismeanttoreducetheamountoftrafconthenetwork, reduceenergyconsumptionintheclientwhilesatisfyingtherequirementsoftheapplication.Thismoduleisdetailedinsection3.4. 3.2.3SensorManagementComponent TheSensorManagementComponentconsistsoftheDataAcquisitionmodule,theFeature Detectionmodule,andtheEventNoticationManager. DataAcquisitionModule:ThedataacquisitionmoduleconsistsofanAPItocontrol thesensors.Forexample,thisAPIshouldincludefunctionalitytoquerythesensorsand 30

PAGE 40

obtainmeasurements,turnonoroffthesensors,changethefrequencyofthequeries, andsoforth.Theimplementationofthismoduleshouldreturnanobjectthatrepresents themeasurementalongwiththeunitsthatareutilizedtoobtainsuchmeasurement.An exampleimplementationofthismoduleistheJSR256:MobileSensorAPIforJava ME[49]. FeatureDetectionModule:UsingthedataobtainedbytheDataAcquisitionmodule, thefeaturedetectionmoduleprovidesthefunctionalitytolearnanddetectbehaviors and/orfeaturesthatareusefulfortheapplicationandtheuser.Thismoduleismeantto performaninitialanalysisonthedataandprovideimmediatefeedbacktotheuser,if necessary.Dataminingand/orarticialintelligencetechniquesarenormallyusedfor thesetasks. EventNoticationManager:Uponthedetectionoffeaturesofinterest,thismodule providesthefunctionalitytonotifythemobileapplicationofthefeaturethathasbeen recognized.Thresholdscanbesetupinthismoduleaccordingtotheuser'sneedsorthe applicationtosendimmediatealertstotheuser,caregiver,orwhoeverisdesignatedor appropriate. 3.2.4ServerCommunicationManagementComponent Thiscomponentmanagesthecommunicationwiththeserveraswellasthesecurityandprivacypoliciesforthemobilesensingapplication.ItconsistsoftheCommunicationManagementmodule,SecurityandPrivacymodule,andtheSessionManagementmodule. CommunicationManagerModule:Thismoduleprovidesstandardwaystotransferdata fromtheclientstotheserverandviceversa.Standardinterfacesexisttoutilizereliableandunreliabletransportprotocols,whicharechosenaccordingtotheapplication requirementsseetable3.1.Forexample,continuousreal-timedatae.g.,GPSxes 31

PAGE 41

Table3.1:CommunicationprotocolsinG-Sense DataDelivery/DataCapture Continuous Event-based Real-time Unreliable Reliable Delay-tolerant Reliable Reliable everysecondinatrackingapplicationaresentusinganunreliableprotocolsuchasthe UDPprotocol.Ontheotherhand,ifreliabilityisrequired,e.g.,amedicalapplication, transferofloggedsensingdata,andsessionmaintenance,TCPorHTTPshouldbe used. SecurityandPrivacyModule:SecurityandprivacyarekeyaspectsinanyLBS,PS,or HCSsystem.Thismoduleincludestheinterfacesandmechanismsneededtoprovide securityandguaranteeprivacy.Encryptionalgorithms,useridentityrandomization mechanisms,andsecuritypoliciesandtheirenforcingmechanismsaresomeexamplesofthemechanismsincludedinthismodule.Thesemechanismsareapplicationdependentandshouldbesetbytheuserorthesystemadministrator. SessionManagerModule:Thisconsistsofthemethodsneededtoexchangesession datawiththeserver.Sessiondataconsistsofusers'data,devices,sessionids,andsimilardataplusindividualsessioninformationsuchasthenumberofpacketssent,the sessionkey,andthelike. WhereasmostofthearchitecturesinthepreviouschapterusedtheHTTPprotocoltotransfer theirdata,theutilizationofthisprotocolinreal-timedatareportingapplicationsisnotrecommendedsinceitrequiresmorepower.ThereasonisthatunlessHTTPpipeliningwhich isnotavailableinAPIsincurrentcellularphonesisused,thecellulardeviceneedstoset upanewTCPsessionwiththeservereverytimeitsendsanewupdate,duetothestateless natureoftheHTTPprotocol.Thethree-wayhandshakeintheTCPconnectionestablishment processrequirestheclienttosendtwopacketsbeforetherstdatapacket,thusatleastthree packetsarerequired.SincetheTCPheadersizeis128bytes,theaverageHTTPheadersize 32

PAGE 42

is200bytes,andwithanaverageenergyconsumptionof13J/KBin3Gdatauploads[50],at least7.4WisneededtouploaddataeverysecondifHTTPisused. 3.3Server-sideSoftwareArchitecture G-Sense'sserver-sidesoftwarearchitectureisshowninFigure3.3.Fromthebottomup,the architecturestartswiththeOperatingSystemandtheApplicationServercomponents.The applicationserverisaruntimeenvironmentforserverapplications.Examplesofthesearethe JavaPlatformEnterpriseEditionJ2EEapplicationserver,Microsoft'sInternetInformation ServicesIIS,andtheApacheHTTPServer.Atthesamelevel,therearethespatialand relationaldatabasesneededtostoreallthedata. Thefollowinglayerconsistsofthreecomponentsthatmanagethecommunicationofthe serverwithmobileandstaticsensorsandotherservers.ThesearetheMobileWSNManagementcomponent,theStaticWSNManagementcomponent,andtheServerSensingManagementcomponent.ThefollowingfourcomponentsaretheDataCollectionandAnalysis components,theDataVisualizationComponent,andtheSensingApplication.Allthesecomponentsaredescribednext. 3.3.1MobileWSNManagementComponent TheMobileWSNManagementcomponentmanagestheconnectivitywiththemobilesensing devices.ThefunctionalityofthiscomponentmatchesthoseincludedintheServerCommunicationManagementcomponentintheclient-sidesoftwarearchitecture.Inaddition,it includestheTaskManagementModule,whichexecutespoliciesoverthereceiveddatafrom themobiledevicestodecidewhethertostorethedatainthedatabase,invokeadataanalysis algorithm,ornotifyotherdevicesinthesystem. 33

PAGE 43

Figure3.3:G-Sense'sserver-sidesoftwarearchitecture 3.3.2StaticWSNManagementComponent TheStaticWSNManagementcomponentintegratesstaticwirelesssensornetworksintheGSensearchitecture.ItconsistsoftheCommunicationManagementmodule,WSNNetwork Servicesmodule,andTaskManagementmodule.TheCommunicationManagementmodule providesthebasictransportandsessionmanagementfunctionalitytoconnectandtransfer datatoandfromthebasestationoftheWSN.TheWSNNetworkServicesmoduleincludes algorithmsthatcannotberuninthebasestationbecauseofthelackofdataand/orprocessing capacity.ExamplesofalgorithmsthatcouldrunasservicesforaWSNaretopologycontrol andtopologymaintenancealgorithms.ThelastmoduleisTaskManagement,whichalso executespoliciesandmakesdecisionsbasedonthereceiveddata. 3.3.3ServerSensingManagementComponent Thiscomponentinterconnectsasensing-awareapplicationwithothersensingapplicationsin otherservers.Withthisfunctionality,asensingtaskcanbedistributedamongseveralservers, 34

PAGE 44

whichreducesandbalancestheloadinaserverandprovidesservicereliabilityincaseof sensinginhostileenvironmentssuchasawarzone.ThiscomponentconsistsoftheTask Managementmodule,whichprovidessimilarfunctionalityastheoneintheMobileWSN, andtheCommunicationManagementmodulethatprovidesconnectivityamongservers.The CommunicationManagementmoduleisresponsiblefortheestablishmentandmaintenanceof thepeer-to-peerarchitecture.Byinterconnectingserversinapeer-to-peerfashion,asensing cloudiscreated,allowingserverstoshareinformation. 3.3.4DataCollectionandAnalysisComponents Thesetwoservercomponentsusethedatacomingfromstaticandmobilesensorsandother servers,andhistoricaldatastoredinthedatabasetoperforminference,correlation,anddata analysistasks.Contrarytothedataanalysistasksperformedatthemobiledevice,whichare basedonlyonlocalandindividualdata,thesecomponentshaveacompletepictureofthe situationandthereforeareablemakedeeperandglobalanalyses. 3.3.5DataVisualizationComponent Thenalcomponentofthearchitectureisthedatavisualizationcomponent.ThisisimplementedinthearchitectureusingWeb2.0toolssuchastheGoogleWebToolkitandopen sourcemappingapplications,whichalongwithgeographicmarkuplanguagessuchasKML showthedataingeographicbrowserse.g.,GoogleEarth,NASAWorldwind. 3.4TheCriticalPointAlgorithms Reducingtheamountofunnecessarytrafcandtheenergyconsumptioninthemobiledevice arecriticalaspectsinLBS,PS,andHCSapplications,inparticularthosethatperformcontinuoussensing.Sinceoneofthemostexpensivefunctionsinresource-constraineddevicesis 35

PAGE 45

Figure3.4:Threebasiccriteriaforcriticalpointalgorithms communications,reducingtheamountofunnecessaryorredundantdatahasthedoubleeffect ofsavingbandwidth,particularlyimportantinbandwidthscarcenetworkssuchaspublic cellularnetworksandwirelessmobileadhocmilitarynetworks,andsavingenergy. Thecriticalpointalgorithm[51]decideswhethertosendsensinginformationtotheserver ornot,whilemaintainingtheaccuracyofthelocationinthesensingapplication.Thisdecisionisbasedonasetofmeasurementsthatcontainlatitudeandlongitudevalues,aswell astimestampinformation.Thegoalistoreducethenumberofupdates,usingthefollowing threebasiccriteria: Changeofdirection:Informationismarkedascriticalsenttotheserverifthereisa considerablechangeinthedirectionoftheuser.Thisinformationismeasuredasthe differenceinazimuthvaluesbetweenthelastcriticalpoint,thecurrentpositionandthe lastpositionofthedevice.Thisisshowningure3.4left. Distance-based:Informationismarkedascriticalwhenadistancethresholdisreached withrespecttothelastcriticalpointgure3.4center. Time-based:Informationismarkedascriticalwhenatimethresholdisreachedbetweenthelastcriticalpointandthecurrentlocation.gure3.4right. 36

PAGE 46

Figure3.5:ThecriticalpointalgorithmcodiedasaJavamethod 37

PAGE 47

Figure3.5showsthegeneralcriticalpointalgorithmcodiedasaJavaprocedure.Thethresholdsandcurrentlocationofthemobiledevicearepassedasparametersandalsothealgorithmcachesthelastlocationofthedevice.Asthresholdscanchangefromapplicationto application,theyplayanimportantroleinthealgorithm.Forexample,ifthecriticalpoint algorithmissetinsuchwaythatthecriticalpointwillevaluateonlydistances,ahighvalue onthethresholdswillsendveryfewupdates,makingitlesssuitabletoreal-timetrackingapplications.Ontheotherhand,alowthresholdvalue,willmaketheapplicationtosendmany unnecessarylocationupdateswastingpreciousresources.Listing3.5alsodepictsconditional evaluationswherecombinationsoftheparametersandothervariablescanbeutilizedtoevaluateifthecurrentlocationiscritical.Thethresholdsforthealgorithmaresetexternallyand theybasedmainlyonthetransportationmodesand/orothercriteriabasedontherequirements ofthesensingapplication.Thethresholdplayanimportantroleinthealgorithm:ifthevalues aresettoolow,manyunnecessarywillbeperformed;howeverifthresholdsaresettohigh thenthereisariskofnottrackingtheusereffectively. Figure3.6showstheeffectofapplyingthecriticalpointalgorithminanapplicationused totrackanindividualwhilewalkingthroughtheTampacampusoftheUniversityofSouth Florida.Inthisexample,thedistancethresholdwassetto20metersandthetimethreshold to30seconds.Thismeansthatthedevicewillsendalocationupdateeverysecondonlyif itismovingataratefasterthan20m/sKm/h.Figure3.6ashowsallthecoordinates sentbythedevicetotheserverwithoutthecriticalpointalgorithm;therewere386coordinatesrecordedduringasevenminuteswalk.Figure3.6bshowsthesamewalkutilizingthe criticalpointalgorithm.Inthiscase,only20locationsroughly5%weremarkedascritical pointsandsenttotheserver. Withtheseparameters,thealgorithmwastestedtwotimeswhiledrivingacar.Intherst experimentthealgorithmrecorded768xespersecondandonly218%updates weresenttotheserver.Inthesecondexperiment,247xesxpersecondwererecorded, andonly58%weresenttotheserver.Withoutthecriticalpoint,using13J/KBin3G 38

PAGE 48

aTripwithoutthecriticalpointalgorithm.Updatessentevery second. bCriticalpointsusingdistance-basedandtime-basedCP Figure3.6:Exampleofthecriticalpointalgorithminatrackingsession 39

PAGE 49

Table3.2:Changeofdirectionscriticalpointalgorithmresults Trip TotalNumberofLocations LocationsMarkedasCritical Percentage 1 72 26 35.61% 2 363 56 15.42% 3 489 65 13.29% 4 208 73 35.09% 5 357 62 17.37% 6 2330 159 6.8% 7 1022 139 13.60% 8 811 137 16.89% datatransmissionasreportedin[50],40bytesperlocationupdatepacketwithoutthe overheadinthetransportlayerprotocol,theaveragepowertosendtheupdatesis0.5W.With thecriticalpointalgorithm,theaveragepowerrequiredbytheapplicationtosendthedatais 0.14W. Usingaversionofthealgorithmwiththechangeofdirectioncriticalpointcriteriaonly,eight tripswereconducted.Intheseexperiments,thealgorithmsentonaveragelessthan20%of thetotalnumberoflocations.Theresultsareshownintable3.2.Whenthecriticalpoint algorithmisimplementedasachangeofdirection,itsimplieslinesegments,inasimilar approachtotheperpendiculardistanceroutineingeographicinformationsystems[52].As such,thecriticalpointalgorithmbasedonthispolicyisusefulfordelay-tolerantapplications. With13J/KB[50]in3Gdatatransmission,40bytesperlocationupdatepacket,andtaking asexamplethetripatrowsixofthetable,thecellularphonerequiresagain1183Jtosend theupdateswithouttheutilizationofthecriticalpoint.Usingthecriticalpointinthiscase requires80J. 3.5APrototypeApplication G-SensehasbeenprototypedinamilitaryapplicationthatcombinesLBS,PS,andHCS, integratesstaticandmobilesensingclients,andimplementsthepeer-to-peerarchitecture 40

PAGE 50

fordatatrafcmanagement,reliability,andscalabilitypurposes.Theapplicationsupports militarydeploymentsbyprovidingthefollowingmainservices: Real-timetrackingofsoldiers. Real-timehealthstatusofeachsoldierwithtemperature,pulserate,breathingdepth, andEKGinformation,ifnecessary.Thesesensorsareautomaticallyactivatedbythe applicationeitherperiodicallyorwhennecessarybasedonthelocallysenseddata,type ofactivity,andothervariables. Integrationofstaticwirelesssensornetworkswithintrusiondetectioncapabilities. Uponanintrusiondetection,thesystemautomaticallygeneratesGeo-Alertstothose soldierscloseenoughtotheevent,andtothemaincontrolstation. Real-timesituationalawarenessfeedbackbasedonlocation.ThisGeo-Alertcapability isimplementedintheserver. Datavisualizationallowsauthorizedusersateachserverlocationtoseetheusersand thevariablesofinterestinreal-time. Theservicesdescribedaboveareservicesperdeploymentsite.Sincetherecanbemanydeploymentsworldwideatthesametime,globalinformationisneededtocoordinatemilitary tasksappropriately.Thepeer-to-peerarchitectureandtheGeotellaprotocoldescribedinthe nextchapterprovidethisfunctionality.Noticethatthissamedistributedarchitecturecould beusedformanyotherPSapplications.Ifyouconsiderthateachindividualsystemcollects CO 2 datafromaPSapplicationinacity,thedistributedsystemwouldbeabletoshowa worldwidepollutionmap. 3.5.1SystemArchitecture Thesystemprototypeconsistsoffoursoftwarecomponents.Therstcomponentrunsatthe mobileclientdeviceandimplementsG-Sense'sclient-sidesoftwarearchitecture.Thesecond 41

PAGE 51

Figure3.7:Generalarchitectureofthesystemprototype componentimplementsthearchitecturefortheserver,thethirdisthecomponentrunningat thebasestation,andthefourthistheWeb2.0application,whichprovidesthevisualizationat themaincontrolstation.Figure3.7depictsageneraloverviewofthesystem'sarchitecture, thenetworks,andthecommunicationprotocolsinvolvedinthesystem.Thecommunication protocolsusedforthesystemareHTTPandUDP.HTTPisusedbythemobileclientsto loginandobtainasessionkey,whichisutilizedtosenddataoverUDP.ReliabilityinUDP isprovidedbytheimplementationofacknowledgmentsmessages,atboththeserverandthe mobileclient.Thisgivetheadvantagetosendreliable/unreliabledatawithlessoverhead.The controlstationconnectswiththeapplicationserverusingAJAX/HTTPtoretrievetheinformationandshowitinGoogleEarthandGoogleMaps.Thisarchitecturecanbereplicatedin asmanyplacesasnecessaryandtheserversareinterconnectedusingtheGeotellasystem. 42

PAGE 52

3.5.2HardwareInfrastructure Thehardwareshowningure3.8forthesystemprototypeconsistsofthefollowingdevices: PanasonicCF19/CF51Toughbook:RuggedizedlaptopsequippedwithinternalGPS receiverandBluetoothradio.Usedasrstlevelintegratorandvisualizationdevices. SanyoPro200Cellulardevice:CellularphoneequippedwithinternalGPSreceiverand Bluetoothradio.Usedasrstlevelintegratordevice. DellPowerEdge860:EquippedwithanIntelQuadcoreXeonprocessor,4GBRAM and500GBHD.Usedasaservertier. MSP410MoteSecurityPackage:SensorpackagecomposedbyeightMica2seetable1.1staticsensornodes.Eachmoteincludesfourseparatepassiveinfraredsensors PIRarrangedorthogonallyfor360-degreecoverage,and2-axismagneticeldsensors.UsedasstaticWSNmotes. MemsicStargateNetbridge:TheStargateservesasrstlevelintegratordevicetoconnectthestaticWSNwiththeserver.AlsoitintegratesaLogitechQuickcamPro4000 cameratocapturephotosuponanintrusiondetection.Thebasestationisequippedwith anIntelIXP420XScaleprocessorrunningat266MHz,with64MBRAMmemoryand USBstorage. ZephyrBioharnessBT:Cheststrapthatthesoldierwearstocollectheartrate,respirationrate,breathamplitude,skintemperature,posture,3D-acceleration,andECG amplitude.ItisconnectedtotherstlevelintegratordevicesviaBluetooth. PollutionMonitoringBoard:AnArduino-basedsensingboarddevelopedasaprototype intheLocation-awareInformationSystemsLaboratoryatUSF[53].Theprototype 43

PAGE 53

boardmeasures CO and CO 2 levels,airquality,andtemperature.Connectedtotherst levelintegratordevicesviaBluetooth. 3.5.3SoftwareInfrastructure Duetothedifferentdevicesinthesystemprototype,thefollowingprogramminglanguages andtoolsareused: J2SE/J2MEtodevelopthemobileclientinthelaptops/cellularphones. Java2EnterpriseEditiontodeveloptheserverapplication. TinyOS/NesCtodevelopthesoftwareinthemotes. Cprogramminglanguageforthedevelopmentofthebasestationsoftware. SunGlassshServerv2asruntimeenvironmentoftheserverapplication. Postgres8.2withPostGIS1.3.2asthedatabasesystemintheserver. GoogleWebToolkitGWT1.5.4todevelopthemaincontrolstationusingtheAJAX technology. GoogleEarth/MapswithGoogleEarthplugintoembedGoogleEarthintheAJAXweb client. NetbeansIDEversion6.5tosupportthedevelopmentoftheJava-basedsoftware. EclipseIDE3.4tosupportthedevelopmentinGoogleWebToolkit. MicrosoftWindowsXPastheOSforthelaptops. MicrosoftWindowsServerastheOSfortheserver. Debian/Linuxwithkernel2.6.24astheOSforthestargate. 44

PAGE 54

aPanasonicCF19Toughbook bSanyoPro200 cMemsicStargateNetbridge dMSP410mote eZephyrBioharnessBT fPollutionmonitoringboard Figure3.8:Hardwareforsystemprototype 45

PAGE 55

Figure3.9:Mobileclientimplementation 3.5.4MobileClient Themobileclientisthesoftwarecomponentthatrunsatthelaptopsandcellulardevices. TheclientwasdevelopedusingtheJavaprogramminglanguage;however,thelaptop'simplementationwasdoneinJ2SE,whileJ2MEwithCLDC1.1/MIDP2.0wasusedforthe cellulardevice.Asshowningure3.9theclientsaresimilarbuttherearesomedifferences intheaccessofthelocationandthesensors.SincethereisnoimplementationoftheJava LocationAPIinJ2SE,alocationproviderobjectwasdevelopedtoaccesstheGPS.This locationproviderutilizestheserialportandtheJavaCommAPI2.0toaccesstheGPS.The BlueCoveJavaAPIwasutilizedtoconnectthelaptoptotheZephyrBioharnessBTandthe pollutionmonitoringboard,usingBluetooth'sSerialPortProleSPP.TheJavaAPIfor BluetoothJSR82wasutilizedtoaccessthesedevicesfromthecellularphone. Asshowningure3.7,theUDPprotocolisusedtosendsensordatatotheserver.When desired,theimplementationprovidesreliabilitybyusingacknowledgmentsatbothendpoints. 46

PAGE 56

Figure3.10:Structureofdatagramssentbythemobileclient TheapplicationsendsoneUDPdatagrampersecondtotheserverwithoutusingthecritical pointalgorithm.DataaresampledfromtheGPSandsensorboardsat1Hz,andthisfunctionalitycanbeactivatedbythelocalapplicationorbythemaincontrolstation.Independent ofbeingactiveornot,themobileclientalwayssendsthelocationdatatotheserver,with theotherdataattachedasrequired.TheshortestUDPdatagram'slengthis28bytes,and thelongestdatagramlengthis92byteswhichissentwhenareliabledatagramwithboth pollutionandvitalsignsdataareattached.Figure3.10depictsthestructureofthemessages sentbytheclient.Tolimittheamountofupdatessenttotheserver,adistanceandtimebasedcriticalpointalgorithmwasutilized.Thethresholdsfordistanceandtimewereset to20metersindistanceand30secondsintime.Aconditionalevaluationwasaddedbased ontheaccuracyofthelocation.Ifanewxwasobtainedwithinthedistancethresholdwith betteraccuracythanthelastcritical,thisnewxwasmarkedandsent.Thealgorithmsent between5%and28%ofthetotalxes. 47

PAGE 57

Figure3.11:Serverapplicationimplementation 3.5.5ServerApplication TheserverapplicationreceivesdatafromthestaticWSN,mobileclients,andotherservers. Italsoservesasdatabackendforthemaincontrolstation.TheserverapplicationimplementsG-Sense'sserverarchitectureusingtheJ2EEspecicationandtheGoogleWebToolkit framework.Assuch,thisapplicationrunsintheSunGlassshV2applicationserverand utilizesthePostGISdatabaseasdatarepository.Theserverimplementationisshowningure3.11anditiscomposedbythemobileWSNmanagement,staticWSNmanagement,the servermanagementandvisualizationcomponents. Asshowninthegure,thecommunicationprotocolsusedwiththemobileclientarethe HTTPandUDP.HTTPisusedbytheclientstoexchangesessionkeysanddownloadles fromtheserver.UDPisusedbytheserverwithreliabledeliveryexceptforACKresponses, andthemessagesthattheserverissuescanbeapushmessagefortheclienttoinitiatethe downloadofale,theactivation/deactivationofthevitalsignsdatacollection,theactiva48

PAGE 58

Figure3.12:Structureofdatagramssentbytheservertotheclient tion/deactivationofpollutioninformation,ACKs,andsituationalawarenessmessagessent frommainthecontrolstation.TheelectionofUDPhastheadvantageofusinglessenergy attheclientsincelessoverheadissentbysensordataupdate.Alsothereisnoneedforhigh degreeofreliabilitysincelocationpacketsaresentfrequently. Theserver'simplementationrequirestwothreadstomanagethecommunicationwithUDP: oneistheUDPlistenerandanotheristheUDPsender.Eachofthesecomponentshavequeues wherethemessagesarecachedwhenawaitingforACKs,orawaitingtobesent.Athirdthread isusedbythesessionmanagertoupdatelocalobjectsallocatedwhenmobileclientslogin, andsendupdates.Thisthreadisalsousedtostoredatainthedatabase.Sincethecommunicationbetweenthemaincontrolstationandthemobileclientsisasynchronous,whenthe maincontrolstationissuesamessage,theservletlistenersatthevisualizationcomponent invokethesessionmanagertocreatethemessagesforthemobileclients.Thestructureofthe messagesareshowningure3.12. Thelasttwocomponentsattheserver-sideapplicationarethestaticWSNmanagementand theservermanagementcomponents.TherstnotiesthevisualizationandmobileWSN managementuponanintrusiondetectionbythestaticWSNmotes,andthesecondmanages theconnectivitybetweensensingservers.TheGeotellaprotocolisusedforthistask. 49

PAGE 59

Figure3.13:Webinterfaceforthemaincontrolstation 3.5.6ControlStation Thecontrolstationisthecomponentthatexecutesatacommander'slaptopusingweb2.0 technology.Thistechnologyencompassesframeworkstodevelopinteractivewebapplicationsthatbehaveasdesktopapplications,butrunningwithininawebbrowser.TheseapplicationsobtaindatafromawebserverusingasynchronousHTTPrequestswithXMLobjects.Todevelopthemaincontrolstationprototype,theGoogleWebToolkit1.5.3frameworkwasutilized.ThisframeworktranslateJavasourcecodetoJavascript,andcreatescode supportedbymajorbrowsers.InconjunctionwiththeGoogleWebToolkit,theprototype utilizesGoogleEarthandGoogleMapsembeddedinthewebapplicationtoshowtheactive soldiers'sessions.Thisinterfaceisusedtoactivate/deactivatethedatacollectionofvital signsorpollutiondata,andalsotosendGeo-Alertstothemobileclients.Thecontrolstation alsoreceivesintrusionnoticationsandthepicturestakenbythestaticWSN.Ascreenshotof thewebinterfaceisshowningure3.13. 50

PAGE 60

Figure3.14:Geo-Alertnoticationsequence TheGeo-Alertcapabilityenablesthesystemtosendmessagesfromthemaincontrolstation tomobileclientslocatedwithinageographicalarea.Whenthecommanderwantstogenerate aGeo-Alert,he/sheclickstwotimesintheGoogleEarthglobe.Therstclickisthecenter oftheareaandthesecondclickcalculatesthereachingdistanceoftheGeo-Alerts.When userwritesthemessageandclicksabutton,theGeo-Alertissenttotheserver.Then,using theGeo-Alert'scentercoordinatesandradius,theSessionManagersubcomponentmakes ageographicalquerytondallthemobiledevicesthatareinsidetheareaandgeneratesa noticationmessageperclientinsidethearea.Subsequently,allmessagesarepassedtothe UDPsenderwhichsendseachmessageusingUDPandawaitsfortheconrmationfromeach oftheclients.Ifnoconrmationisreceivedwithincertainamountoftime,theUDPsender resendsthemessage.Thesequenceisdepictedingure3.14.Allcontrolstationoriginated messagesfollowthisthissequence. 51

PAGE 61

Figure3.15:StaticWSNimplementation 3.5.7StaticWirelessSensorNetwork ThestaticWSNintegrationtotheprototypesystemenablesthemonitoring,detection,storage,andnoticationofintrusionalerts.ThenetworkiscomposedbyeightMemesicMSP410 motes,aStargateNetbridgebasestationthatcollectsdatafromthemotes,andawebcam attachedtothebasestationthattakesapicturewhenanintrusionisdetected.Thebasestation isconnectedtotheapplicationserverwhichmaintainsthedatabaseofintrusionsandpictures andisresponsibleforthenoticationofintrusionstomobileclientsandthecontrolstation. Figure3.15showstheimplementationofthestaticWSNcomponent. Twoclassesofmessagesaresentbythemotestothebasestation,therstaretheintrusion detectionmessage,andthesecondarenoticationmessagesoflowbatterywhenanyofthe motesisnearlyoutofpower.Periodically,theprogramrunningatthemotesreadvaluesfrom passiveinfraredsensorsPIR,andsendsamessagetothebasestationwhenthePIRreading isabovesomethreshold.Thismessagealsoincludesmeasurementsfromthemagnetometers. Afterreceivingtheintrusionmessage,thebasestationutilizesthewebcamtotakeapicture ofthearea,andthenitinvokesthecURLlibrarytosendviaHTTPtheintrusionmessageand uploadthepicturetotheserver.Then,theserverperformsasimilarprocessastheoneforthe Geo-Alertcapability,notifyingallusersthatarewithincertaindistanceoftheintrusion.Finallywhenthemobileclientsreceivethenoticationmessage,theydownloadviaHTTPthe 52

PAGE 62

Figure3.16:Intrusionasseenbythecontrolstationandmobileclient imagele.ThemaincontrolstationalsogetsnotiedoftheintrusionviaAJAX.Figure3.16 showsthenoticationmessagesasreceivedbythecontrolstationandamobileclient. 53

PAGE 63

Chapter4:Geotella OneofthemostimportantaspectsofbuildingaglobalsystemlikeG-Senseishowtomanage thelargeamountoftrafcgeneratedbyallthesensingdevices.Thischallengehasbeenpartiallyaddressedbythealgorithmspresentedinthepreviouschapter.However,ifthesystem isimplementedinmanyplacesatthesametimeacentralizedsystemwouldnotbeappropriate.InordertomakeG-Sensescalable,apeer-to-peerarchitectureisintroducedbasedon thelocalityofthedata.ThischapterdescribesGeotella,asystemthatutilizesahierarchical peer-to-peerapproachtomanagethescalabilityinanubiquitoussensingsystem,anddeliver geographical-basedqueriesandmessages. 4.1Requirements ThefollowingaretherequirementstoconstructaP2PsystemforG-Sense: Interconnectsensingservers:TheP2Psystemwillbeimplementedintheserversto diminishnetworktrafc,andprovidescalability.TheP2PsystemmustworkonaIPbasednetwork. Providemeanstotrackaserverbasedonitsgeographicallocation:Fromanyofthe servers,usersshouldbeabletodiscoverwhichserversareavailableinagivengeographicalarea. 54

PAGE 64

Utilizetheservers'locationdatatodispatchsensingtasks:Byusingtheserver'sgeographicaldata,theP2Psystemshouldprovidemeanstodistributesensingtasksand collecttheresultsfromsuchtasks. Utilizetheservers'locationdatatosendgeographicallyorientedmessages:Byusing theserver'sgeographicaldata,theP2Psystemshoulddelivergeographicallylocated messagestotheserversanduserslocatedintheareaofthemessage. 4.2SystemArchitecture InGeotella,thesystemusesahierarchicaloverlay[54]wherepeersareorganizedingroups. Betweengroups,thesystemutilizesadistributedhashtableDHTtodivideageographicalareaintonon-overlappingzones.Foreachzone,serversareconnectedtoapeerthatis responsibleforthezone,asshowningure4.1.Thesezone'speersareresponsibleforthe maintenanceoftheDHT. Allserversandpeershaveacopyofthelewiththegeographicalzonedivision,andwithin eachgroup,thepeerthatisconnectedtotheDHTservesasalocalserverdirectorywithinthe zone.Allserversthatareassociatedwithazoneutilizethepeertondotherserversinthe zone,updatetheirlocationsinthecasethattheserversaremoving,andreceivegeomessages andgeoqueriesfromotherzones. 4.3ProtocolMessages Geotella'smessagesaremeanttomaintaintheDHT,managetheintrazonedirectory,and delivergeoqueriesandgeomessages.Sincethesystemutilizestheimplementationofan availableDHTe.g.,OpenChord[55]andtheHTTPprotocol,thissectiononlydescribes thecustomizedUDPdatagramsthatthesystemsupports. 55

PAGE 65

Figure4.1:Geotella'speer-to-peerarchitecture 4.3.1ProtocolHeader Theheaderfortheprotocol'smessageshas24bytesinsizeandcontainsthelocationlatitude,longitudeoftheserverthatsendsthemessage.Theeldsareexplainedasfollows: messid:Eachheadersentbyaserverhasauniquemessageid.Thisnumberisalong integeranditisgeneratedusingthetimestampinmilliseconds. mess_type:Thiseldidentiesthetypeofpayloadinthemessage. lat:thelatitudeofthepeersendingthemessage. lng:thelongitudeofthepeersendingthemessage. cov_area:thecoveringdistanceoftheserverthatsendsthemessage.Thecovering distanceofaserverisdenedasthedistancetothefarthestmobileclientorstaticWSN connectedtothepeerfromtheserver'slocation. 56

PAGE 66

Figure4.2:Geotellamessages 57

PAGE 67

4.3.2Acknowledgements Theacknowledgmentsarecomposedbytheheaderandapayloadof8bytesthatspeciesthe messageidofthemessagethatneedstobeacknowledged.Themess_typeforACKsis0. 4.3.3Join OnceaserverobtainstheIPaddressofthepeerthatkeepstheintrazonedirectory,itsends thismessagetosuchpeertojointhezone.Ajoinmessageisdenedbytheheaderandapayloadof4bytesthatindicatesaportthattheserverutilizesforresponses.Thismessagesets themess_typeto1,andtheresponseofthemessageiseitheranacceptorarejectmessage. Thislealsohasthedesignationofaserverthatshouldtakecontrolofthezoneincasethe peerfails. 4.3.4Accept Ifapeeracceptsaserver,itrespondswithanacceptmessagethatcontainstheURLofale withIPaddresses,locationinformationandportnumbersofthecurrentserversconnectedto apeer.Themess_typeoftheacceptmessageis2andthepayloadisfourbytesandtheURL. ThefourbytesindicatethelengthoftheURL,withamaxlengthof1000bytes. 4.3.5Reject Ifapeerdoesnotacceptaserveritrespondswitharejectmessage.Themess_typeofthe rejectmessageis3andnopayloadisset. 58

PAGE 68

4.3.6Ping Apingmessagesentbyaserverhasthefunctionalitytobeakeep-alivemessagewithinthe zone.Themessageisdenedbytheheaderandapayloadof4bytesthatindicatesaport thataserverutilizesforresponses.Themess_typeofthepingmessageis4.Thismessage isissuedafternoHTTPGETtoobtainanupdateofthelocaldirectoryisnotansweredbythe peer. 4.3.7Pong Whenapeerreceivesapingmessage,itrespondswithapongmessagethatcontainstheURL speciedbytheacceptmessage.Themess_typeofthepongmessageis5andthepayloadis fourbytesthatindicatesthelengthoftheURLandtheURL.ThemaxlengthoftheURLis 1000bytes. 4.3.8Geoquery Thefunctionalityofthegeoquerymessageistodelivergeographicallocalizedqueries.The message'sstructureisasfollows: messid:Thiseldidentiesageoquerymessage.Themessidisthetimestampinmillisecondsofwhenthemessagewasgenerated. lat:Latitudeofthecenteroftheareawherethegeolocatedqueryissent. lng:Longitudeofthecenteroftheareawherethegeolocatedqueryissent. covdist:Theradiusofthegeographicallocatedquery.Thelat,lngandthecovdist deneacircumferencewherealltheserversandpeerswhicharewithintheareaare notied. 59

PAGE 69

ip_address:TheIPoftheinitialserverthatoriginatedthegeoquerymessage. port:Theportthattheoriginalpeerutilizestowaitformessageresponses. size:Thelengthofthemessagequerysent.Aquerymusthaveatleast1characterand lessthan1000characters. query:Thecontentsofthegeolocalizedquery. Themess_typeofthegeoqueryis6andthemessagemustbeacknowledgebytheserversthat accepttocollectdataforthegeoquery. 4.3.9Geomessage Similartothegeoquery,thegeomessagedeliverslocalizedmessages.Themess_typeofgeomessageis7. 4.3.10GeoqueryResponse Thismessageisissuedbyaserverthathascollecteddataforalocalizedgeoquery,anditis meanttonotifytheoriginalserverthatgeneratedthegeoqueryoftheresults.Thepayloadfor themessageisdenedbythemessage_idoftheoriginalgeoquery,thesizeoftheURL,and theURLofthelewiththeresults. 4.3.11GeomessageResponse Thismessageisissuedbyaserverthathasreceivedageomessageandiswithintheareaof thegeomessage.Thepayloadforthemessageisdenedbythemessage_idoftheoriginal message. 60

PAGE 70

Figure4.3:Geotellaprotocolstates 4.4ProtocolStates ThissectionexplainsGeotella'sserversandpeerstates,asdepictedingure4.3. 4.4.1Initalization InGeotella,allservershavealocationandazonedirectorywhichisidenticalforeachonein thesystem.WhenaserverwantstojoinGeotella,itperformthefollowingsteps: Usingitszonedirectoryandownlocation,theserverndstheidentierofthecurrent zonethatitiscontained. Withtheidentier,itcontactsapeerintheDHTtosearchforthecorrespondingpeer withinthezone. IfthereisnovalueassociatedwiththiskeyintheDHT,thereisnopeerinthezone. Theserverbecomesthezone'speeranditstoresitsIPaddress,UDPportandlocation informationintheDHT. Ifthereisavalue,thepeerobtainstheIPaddressandUDPportofthecurrentgateway inthezone.Then,itcontactsthezone'speerandissueaJoinmessage.Ifthepeeracceptstheserver,itrespondswithanacceptmessage,otherwiseitrespondswithareject message. 61

PAGE 71

4.4.2Ready Iftheserverbecomesthezone'speeroritisacceptedbyapeer,itenterstheReadystate. Inthisstate,theserver/peercanissuegeoalertsandgeoqueries,andifitisthezone'speer, itbecomesthezone'sdirectory.Inthisstate,thezone'speercanreceivejoin,accept,DHT messagesandHTTPrequests.Thepeeralsodesignatesoneofitsavailableserversasthe replacementofthezone'speerincasethepeerfails.TheserversperiodicallyissueHTTP GETmessagestoobtainanupdatecopyofthezone'sdirectorywiththeIPaddressesofother serversinthezone.Thislealsohasthedesignationoftheserverforreplacementincasethe peerfails.Atthispoint,aservercanissuegeoqueries/geomessages. 4.4.3Geoquery/Geomessage Inthisstate,aserverissuesgeolocalizedqueriesandmessages.Giventhecoordinatesand radiusofageoquery/geomessage,aserverperformthefollowingoperations: Verifytheareaofthemessage.Iftheareaiswithinthelocalzone,usethelocalzone directorytondoutwhichserversoverlapwiththeareaofthemessageandsendthe geoqueriestosuchserversdirectly.Iftheareaofthemessageoverlapsotherzones,use apeertoobtainfromtheDHTtoobtaintheIPandportinformationofeachzonethe messageoverlaps,andsendthemessagetoeachzone'speer. WaitforACKs.Onceaserversendsthegeoquery/geomessages,itwaitsforthereceptionofACKsfromtheserversthataccepttocollectdataorbroadcastthemessage. Waitforresponses.AfterreceivingtheACKs,theserverwaitsforeachoftheservers thatissuedtheACKsageoquery/geomessageresponse.Ifthegeoqueryresponsesare received,usetheprovidedURLtodownloadtheresultsofthegeoqueries. 62

PAGE 72

Whenapeerreceivesageoquery/geomessagefromaserverthatisnotinitscurrentzone,it usestheintrazonedirectorytondtheserversthatoverlapswiththeareaandthenitforwards themessagetosuchservers.Also,dependingonthepeer'spolicies,itcandropthemessage. 4.4.4Maintenance Inthisstate,theserversandpeersmaintainthepeer-to-peernetwork.Theactionsinthisstate areasfollows: ThepeerperformsthecorrespondingmaintenancetasksfortheDHT. Thepeerchoosesoneoftheserversinitszoneasitsreplacementincaseoffailure. Incasethepeerhasnotreceivedanyupdatefromapeerinawhile,itissuesaping messagetotheserver.Iftheserverdoesnotreplies,thepeerremovestheserverfrom itslocalserverlist. Theserversinthezoneusingpingmessagesupdatetheirlocationandsensingareain caseofchanges,orincaseaHTTPGETrequestcannotbecompliedbythepeer. 4.4.5Failure Incaseofthepeerfailsinazone,thedesignatedserverbecomesthepeer'szone.Thisnewly peerjoinstheDHTandalltheserversinthezoneperformajointothepeer.Ifboththepeer andthedesignatedserverfailsatthesametime,theserversperformwaitforarandomnumberofsecondsbeforetheygototheinitializationstate.Atthispoint,leavingthesystemis assumedasafailure. 63

PAGE 73

Figure4.4:Geotella-connectedserversasshownincontrolstation 4.5ProtocolImplementation Geotellahasbeenincorporatedtothesystemprototypedescribedinsection3.5tointerconnectseveralG-Sensesystems.WithGeotella,mobileclientsareassignedtoaparticularGSensesystemthatrepresentsagroupofsoldiersindeterminedgeographicalarea.Fromthe controlstation,acommandercanfollowanyoftheserversofageographicalzone,withtheir currentlocationandtheircoveragearea,asshowningure4.4. Currently,theintrazoneprotocolhasbeendeveloped,andithasbeenincorporatedintothe serverapplicationrunningattheSunJavaApplicationserver.Fortheinterzone,theplanisto utilizeOpenChord[55],whichisaJavaimplementationoftheChordDHT. 64

PAGE 74

Chapter5:RelayPlacementinWSNsusingMultiobjectiveOptimization ThischapteraddressestheplacementprobleminWSNs,ordeterminingtheoptimalplacementofwirelessrelaynodestomonitorareasofinterestwiththegoalofminimizingtwo objectivessimultaneously:thenumberofsensorsdeployedandtheenergydissipationofthe sensornetwork. Thesetwoobjectiveshavebeenconsideredbecauseitisdesirabletodeploythefeweramount ofsensors,tominimizethemonetarycostofthedeployment,whiledissipatingtheleastamount ofenergy,tomaximizethelifetimeofthenetwork.BytakingintoaccounttheMinimum TransmissionEnergyMTEapproachexploredbyHeinzelmanetal.[56],thechapterexploresthetradeoffintheplacementofrelays,allowingthenetworkdesignertodecideamong differentefcientplacements,undertheParetoapproach.ThisproblemiscalledtheMultiobjectiveConnectedRelayNodePlacementProblemM-RNPc-P.Tosolvetheproposed multiobjectivemodel,thischapterpresentstheMemeticRelaySteinerTreeM-RESTalgorithmwhichisahybridevolutionaryalgorithmutilizestwoheuristicstondPareto-efcient solutions. 5.1MultiobjectiveOptimization Inatypicaloptimizationprocess,itisdesiredtomaximizeorminimizesomefunctionthat representsameasuredpropertyorattributeofasystemoranobject.Acommonexampleis thecostofanairlineticket.Apassengerusuallychoosestheticketthatcoststheleastamount ofmoney,however,thelowcostoftheticketdoesnotcomefree:theserviceintheairline 65

PAGE 75

mightbetheworst,theremightbehiddencosts,orthebookedightmightbenottheefcient intermsoftimechangeofplanesinsteadofadirectight. Exampleslikethisarecommoninreal-lifeandengineering,wheretheminimizationofmonetarycostsmightcompromisesecurity,theenvironment,orcustomersatisfaction.Assuch, real-worldproblemsusuallydealwiththesimultaneousoptimizationofmorethanonecriteria,andmostofthetimesuchcriteriaarecompeting. Whendealingwithtwoormoreattributestooptimize,anengineermightchooseoneoffour options:theengineermightignoresomeofthecriteria,hecancombinethemintoasingle optimizationproblembymultiplyingthembyweightfactorsandthenadding,hecanrank theminorderofpreference,orhemightchoosetooptimizeallsimultaneously. ParetoorMultiobjectiveOptimization[57]correspondstothelastoption:aframeworkin whichthefunctionstobeoptimizedaretreatedasequals,sothereisnopreferenceforanyof thefunctionsduringtheoptimizationprocess.Thesolutionofaproblemunderthisframeworkisasetofsolutionswhichoptimizesimultaneouslytheobjectives,andthedecisionon whichsolutiontouseistakenaftertheoptimizationprocesshasnished. Mathematically,thisprocesscanbedescribedasminimizationcontext: MinimizeZ = F x ; x 2 X f .1 where X f isthefeasibleset, F x = f f 1 x ; f 2 x ;:::; f k x g and k isthenumberoffunctions tobeoptimized. InParetooptimization,thedominancerelationdetermineswhenasolutionisbetterthan another.Giventwosolutions x y 2 X f ,therelationshipstatesthat y isdominatedby x denotedas x y if Z x i Z y i where i=1,..k Z x i = f i x ^ Z x 6 = Z y .When y hasatleastone objectivebetterthan x ,thenbothsolutionsaresaidtobenon-comparable,sinceneither x nor y dominateseachother.Thisisdenotedas x y 66

PAGE 76

Figure5.1:RelationshipsinParetooptimization Finally,asolution x iscalled Paretooptimal ifthereisno x 2 X f suchthat x x .The goalinParetoOptimizationistondthe Paretooptimalset composedbyallParetooptimal solutions.Anapproximationtothissetiscalleda Paretoefcientset Theserelationshipsareshowningure5.1.Thegureontheleftrepresentsthesolution space,whereeachsolutioninthiscaseiscomposedbytwovariables.Theimageofthese variablesthroughthetwofunctionsthatareminimizedisdepictedinthemiddlegure.A quadrantovereachpointrepresentstheareaoftheobjectivespacethatsuchpointisdominating.Thereforec',e',f'arethepointsintheobjectivespacethatbelongstotheoptimalPareto frontandthesolutionsthatcorrespondtothesepointsareshowningreenontheleftsideof thegure.Becausethedominancerelationimposesapartialorderingonthesolutionspace, thesolutiontotheproblemisaset.Inthiscase,theParetooptimalsetwouldbecomposedby solutionsc,e,andf. 5.2OptimizationModel IntheWSNRelayPlacementproblem,adesignerwantstomonitorareasofinterestusingthe leastamountofrelaysanddissipatingtheleastamountofenergyaspossible.Thisscenariois depictedingure5.2,wherethe a,b,c,d,e,f and g representtheplacescriticalpointsthatare 67

PAGE 77

Figure5.2:AprobleminstanceandWSNrelaytree desiredtobemonitored,theroundedbluepointsrepresentadeploymentandtheblackpoint representsthebasestation.Ageneralinstanceoftheproblemisspeciedbythelocationof thebasestationandcriticalpoints,themaximumcommunicationandsensingrangeofWSN nodesandthecharacteristicsofthegrid. Withthisparameters,ageometricgraph, G = V ; E ; R Comm ; R Sense ,isdened.Onthisgraph, V isthesetofvertexesthatrepresentthesensors, E isthesetofedges,whichrepresentthe connectivityamongthesensors, R Comm istheradiotransmissionrangeofthesensors,and R Sense isthesensingareaofthesensors,wherebotharesymmetricdiskscenteredatthenode's positionwiththeirrespectiveradii.Eachvertexhasageometriccoordinateassociatedtoit andanopenballwithradius R Comm .Thisopenballisasetthatcontainsallthevertexeswith distancelessthan R Comm respecttoanode,whichrepresentsthecommunicationareaofthe sensors.Thenodesintheopenballaretheonlyneighborsthatthenodecancommunicate withdirectly.TheformaldenitionoftheopenballispresentedinEquation5.2,asdened in[58]. B r x = f y : d x ; y < R Comm g ; x ; y 2 V .2 68

PAGE 78

where d ; representstheEuclideandistanceandthesetofedges E isdenedastheunion oftheopenballsdenedforeachvertex. Theareaofinterestisdividedingridsandtheset C = f c 1 ; c 2 ;:::; c m g isdenedtorepresentthecriticalpoints,orthosepointswithintheareaofdeploymentthataredesiredtobe monitored.Intheintersectionofthehorizontalandverticallines,therearethegridpoints. Eachgridpointhasalistofcriticalpointsthatthegridpointcansense.Acriticalpoint c i iscontainedinthelistofsensedpointsofthegridpoint v k if d c i ; v k R Sense ,where d.,. againrepresentstheEuclideandistance. Lettheset G cover = f v 1 ; v 2 ;:::; v l g beaminimumsubsetofthenodesof G thatcoversthe set C .AnapproximationtotheM-RNPc-Pproblemistondaspanningtree T G that connectsthenodesin G cover withthebasestationthroughrelaynodes.Thus,thetreemust satisfythefollowingtwoproperties: Thetree T isrootedatthebasestation. Thetree T mustcoverallcriticalpoints. Twofunctionsrelatedtoanygiventree T arethedeploymentsizeforthetree, S T ,andthe dissipatedenergyofthetree, E T .Thesefunctionsaredenedas: S T = k T k .3 E T = v 2 T ^ v 2 G cover e v ; T .4 where k : k denotesthenumberofnodesin T and e v standsforthedissipatedenergyfrom v tothesinknodeinthetree T .Thedissipatedenergy e v ; T iscalculatedasfollows[56]: e v ; T = x ; y 2 P v ; sink ; T E Telec x ; y + E amp x ; y + E rec .5 69

PAGE 79

where E Telec x ; y = E elec k istheenergydissipatedbythesender'selectronicsonsendinga messageof k bits; E amp x ; y = e amp k d x ; y 2 istheenergydissipatedbythesender'sradio and E rec x ; y = E elec k istheenergyspentbythereceiver'selectronics. P v ; sink ; T denotes thepathfrom v tothebasestationinthetree T .Therefore,theequationthattheproposed algorithmoptimizesis Min Z = f S T ; E T g .6 where T isadeploymentasitwasstatedbefore. 5.3TheM-RESTAlgorithm InordertosolvethemodeldescribedbyEquation5.6,ahybridmultiobjectiveevolutionaryalgorithmisproposedandimplemented.AlsoknownasMemeticalgorithms[57],they combinetraditionalevolutionarytechniqueswithheuristics.Evolutionaryalgorithmsare techniquesforsolvingoptimizationproblemsbasedonnaturalevolutioninwhichonlythe characteristicsofthemostadaptedindividualsofapopulationinanenvironmentaretheones thattendtosurvive. Evolutionaryalgorithmsmimicnaturebyhavingasetofdatastructuresthatrepresentthe individualsandtwooperationscalledcrossoverandmutation.Bycombiningandselecting effectivelythedatastructures,theoperatorsofthealgorithmndefcientsolutionstothe problem.Therefore,inordertoapplyevolutionaryalgorithmsinaparticularproblem,itis necessarytodenethesolution'sindividualsrepresentation,thecrossoverandmutation operators,andthetnessfunction.Inthecontextofmultiobjectiveoptimization,evolutionary algorithmshavebeentraditionallyusedbecausetheycaneasilyexploreseveralregionsofthe feasiblespacesimultaneously,andthisisnotpossiblewithothermetaheuristictechniques suchassimulatedannealingortabusearch. 70

PAGE 80

Figure5.3:AgeneralevolutionaryalgorithmcodiedasaJavamethod Figure5.3showsthecodeofanevolutionaryalgorithmasaJavamethod.Here,line10generatestheinitialpopulationbysamplingatrandomthefeasiblespace.After,line13ofthe algorithmevaluateshowgoodisanindividualwithinitspopulation.Bestevaluatedindividualsarethenselectedinline14togenerateanewpopulationusingthecrossoverandmutation methodslines15and16.Theprocessisrepeateduntilsomestopcriteriaisreached,usually acertainnumberofgenerationsisreachedornoimprovementintheoptimizationprocess. Memeticalgorithmsextendevolutionaryalgorithmsbyprovidingtheindividualsawayto improvethemselves.Thisprocessisperformedbyincorporatingknowledgeoftheproblem throughheuristics.Thereforeahybridevolutionarymemeticalgorithmincludesthetraditionalcomponentsofanevolutionaryalgorithmplusheuristics.Theheuristicsarepartofthe proceduresoflocalsearchinamemeticalgorithm,whichallowindividualstoimproveand thealgorithmtondbettersolutionsandconvergeinlessgenerations. TheproposedalgorithmcalledtheMemeticRelaySteinerTreeorsimplyM-REST,Java codeshowningure5.4algorithmisanevolutionaryalgorithmwithtwolocalsearches thatutilizesanelitistpopulationtokeeptheParetoefcientsolutionsfoundduringtheoptimizationprocess.Itutilizesadifferentrandomweightedlinearfunctiononeveryiterationto evaluatethesolutionsandsearchParetoefcientsolutionsindifferentregionsofthefeasible 71

PAGE 81

Figure5.4:TheM-RESTalgorithmcodiedasaJavamethod space.Whereasmemeticalgorithmshavebeenutilizedinthepast[59]tooptimizemaximum coverageinanareausingWSNs,thisisthersttimetoourknowledgeahybridevolutionary algorithmhasbeenutilizedtominimizethenumberofrelaystomonitorpointsofinterest.In thefollowingsectionsofthischapter,thecomponentsofM-RESTareexplained. 5.3.1RepresentationoftheIndividuals Giventheset G cover ,atree T isrepresentedasavectorofpathsfromthesinknodetoeach ofthevertexesintheset.ThisisdepictedinFigure5.5.Inthisgure, G cover containsnodes 2,3,5,7,and10andtheindividualhassixpaths,startingfromthesinknode0.Therelay 72

PAGE 82

Figure5.5:Representationoftheindividuals treerepresentedbythegurehas11sensorsandadissipatedenergyof349J,assumingthat E elec = 50 nJ = bit e amp = 100 pJ = bit = m 2 and k = 2000bitsbytes[56]. 5.3.2CrossoverandMutationOperators Giventwoindividuals,thecrossoveroperatorindicateshownewsolutionsoffspringsare generated.TheM-RESTalgorithmusesasinglepointcrossover,whichtakestwoindividuals, I 1 and I 2 ,andproducestwonewindividualsasfollows: Theoperatorselectsacrossoverpointatrandom,whichisthenumberofoneofthe availablepaths. Theoperatorgeneratesanewindividualbycombiningthepathsfrom I 1 uptothecrossover pointwiththerestofthepathstakenfromthecrossoveruntiltheendfrom I 2 Asecondindividualiscreatedutilizingthesecondstepbutswitchingtheindividuals, i.e., I 2 becomes I 1 andviceversa. Figure5.6ashowsanexampleofthecrossoveroperationwithtwoindividualsandpath2as thecrossoverpoint.Pathsarenumberedfrom0to2. Givenanindividual,themutationoperatortakesarandompointintheindividualandgeneratesanewpathfromthesinknodetothecorrespondingcoveringnode.Figure5.6bshows 73

PAGE 83

anexampleofthemutationprocesswithoneindividualandrandompoint1,whichcreatesa newpathfromthesinktocoveringpoint15. 5.3.3FitnessFunction ThetnessfunctionutilizedintheproposedalgorithmistakenfromtheRandomDirectionsMultiobjectiveGeneticLocalSearchRD-MOGLSalgorithm[60].RD-MOGLSutilizesa weightedfunctiontoassigntnessvaluestoindividuals.Thisfunctionisdenedas: S l Z ; L = k i = 1 l i z i = k i = 1 l i f i x .7 where0 l i 1, l i 2 L l i = 1, f i x istheevaluation i th functionoftheindividual x and k isthenumberoffunctionstobeoptimized.Anormalized L vectorischosenatrandomon aniterationforsearchingnewsolutionsinadirectionofthefeasiblespace. Eachiterationofthealgorithmtriestominimize S l Z ; L byselectingforcrossoverstepthe individualsthathavelowerevaluationof S l Z ; L .Thealgorithmkeepsanelitistpopulation thatcontainsthebestindividualsfoundsofarandtheadmissiontothiselitistpopulationis givenbythedominanceandnon-comparablerelation.Whenthealgorithmrunsforachosen numberofiterationsoranotherstopcriteriaisreached,itstopsandreturnsaParetofront. 5.4LocalSearchHeuristics Theadvantageofmemeticalgorithmsoverevolutionaryalgorithmscomesfromtheincorporationoflocalsearchproceduresthatcontainknowledgeoftheproblembeingsolved.Inthe caseoftheproblemathand,twoheuristicshavebeendevelopedasimprovementtechniques forindividuals.Therstlocalsearchprocedureisaheuristicthattakesintoaccounttheconnectivityinformationofthepathsinanindividualtodiscovernewpaths.Thesecondheuristic isamoregeometricalapproach,asitconsidersthatthegraph G modelsanEuclideanspace. 74

PAGE 84

aCrossoveroperation. bMutationoperation. Figure5.6:Examplesofcrossoverandmutationoperations 75

PAGE 85

WecalltherstheuristictheCycleReductionSearchandthesecondheuristictheBreadth FarthestFirstSearch. Whenanindividualentersthelocalsearchprocedure,arandomnumberisgeneratedfrom auniformrandomnumbergeneratorintherange[0,1].Iftherandomnumberislessthana xedprobability,theindividualisgoingtobeimprovedusingtheCycleReductionSearch, elsetheindividualwillbeimprovedusingtheBreadthFarthestFirstSearch. Thetwoheuristicsareneededbecausetheyhelptocreatetreesandndshorterpaths.Since therstheuristicisnotrelatedtothespacegeometryoftheproblem,theremightbesituationswherethepathsarebetterreducedusingthesecondheuristic,moreover,theutilization ofbothlocalsearchprocedurescombinedimprovethesolutionsinbothobjectiveswhich cannotbeperformedbyusingonlyonelocalsearchmethod. 5.4.1CycleReductionSearch TheinitialpopulationofindividualsfortheevolutionaryalgorithmisgeneratedbyperformingaDepthFirstSearchfromthesinknodetoeachof v 2 G cover .Whenexecutingthisinitial searchthereisnopreferenceintheorderofvisitingnodeswhenexpandingthesearchtree, thereforeitispossiblethattwopathsofthesameindividualcrosseachother,creatingacycle. Althoughnewpathscanbegeneratedbyjoiningpathsatintersectionpoints,twoquestions arisewhentryingtocreatepathsinthismanner: Ifthepathsintersectinmorethanonepoint,whatpointshouldbechosentogenerate thenewintersectionpath? Howcanweselectanintersectionpointsuchthatthenewpathdoesnotcontaincycles? AsolutiontothesetwoquestionsiscalledtheCycleReductionSearchandwasdescribed in[61]wheretheauthorssolvedamultiobjectiveproblemformulticasting.TheCycleReductionSearchworksasfollows: 76

PAGE 86

Let w beanintersectionpoint,and X and Y bepathsin G withthesinknodeasthestarting node.Let D w bethe distanceofintersectionfunction denedas: D w : = abs Pos X w )]TJ/F41 11.9552 Tf 10.95 0 Td [(Pos Y w .8 where abs standsfortheabsolutevalue, Pos X w isthepositionoftheintersectionpointin path X ,and Pos Y w isthepositionoftheintersectionnodeinpath Y Bycalculatingthisfunctionforallintersectionpointsbetween X and Y ,andchoosingthe position w thatmaximizes D w ,itisguaranteedthatthenewpathwillnothavecycles.This isbecausewhenchoosingthisnode,thelargestcyclebetweenbothpathsisidentiedand reduced,sonootherloopwillbegeneratedwhenjoiningthepathsatthispoint.Theproofof thiscanbefoundin[61].Byapplyingthemethodamongallpairsofpathsintheindividual, thelongestcyclesareremovedandanimprovedindividualisreturned. Figure5.7showsanexampleoftheCycleReductionSearchprocessusingtwopathsfrom oneindividual.Theprocesstakesplaceinvesteps.Intherststepthealgorithmndsall intersectionpoints.Giventhepointsandtheirpositions,duringsteptwothealgorithmcalculatesthedistanceofintersection.Intheexample,thechosenintersectionnodesarenode 8andnodeandnode19.Next,inthestepthreethesubpathsfromthesinknodetoeachof thechosennodesinstep2areobtainedandtheirhopcountisevaluated.Instepfour,the newpathsarecreatedbychoosingforeachoftheintersectionnodesinstep3,thesubpath fromthesourcetotheintersectionnodewithlesshopcountandjoiningitwiththeremaining nodesfromtheintersectionnodetothedestinationintheoriginalpathofthelongesthop subpath.Forexample,takingnode19instepfourofgure5.7,thesubpath0,19hasonly onehopcountcomparedwiththe14hopsoftheothersubpath.Thenewpathiscreatedby joining0,19with19,10,22,5. Theworsttimecomplexityofthisalgorithmis O N 2 n where N isthenumberofpathsin theindividualand n isthenumberofnodesinthenetwork. 77

PAGE 87

Figure5.7:Cyclereductionprocedure 5.4.2BreathFarthestFirstSearch Givenanindividual,thesubgraph G I V 0 ; E 0 isdenedusingthenodesfromtheindividual andtheconnectivityofthenodesinit.TheBFFSisamodicationoftheBFSalgorithm wherethesearchtreein G I isexpandedusingBFSbutthenodesintheexpansionareordered indescendingorderbytheirEuclideandistancetothesinknode.Thesearchtreeisrootedat thesinknode. Oncethesearchtreeisbuilt,backwardpathsarecalculatedforallnodes v 2 G cover tothe sinknodeinthesearchtreeandthesearecomparedwiththeircorrespondingpathintheindividual.Ifthebackwardpathisbetterthanthepathforthatnode,thepathintheindividual isreplacedbythebackwardpath.Apathisconsideredbetterifithasabettersmallerhop count. 78

PAGE 88

Figure5.8:AParetofrontbyM-RESTalgorithm Theworsttimecomplexityofthisalgorithmis O V + E where V isthenumberofgrid pointsinthedeploymentand E isthenumberoflinks. 5.5Evaluation TheM-RESTalgorithmwasimplementedinJavaJDK6andaninitialtestcasewiththeinstanceshowningure5.5wasexecuted,obtainingaParetofrontdepictedingure5.8.This instancehadeighttargetpointsandthebasestationwasplacedinthemiddleofthearea.On thisinstanceandinallthefollowingexperiments,ithasbeenassumedthat E elec = 50 nJ = bit e amp = 100 pJ = bit = m 2 and k = 2000bitsbytes[56].M-RESTwasexecutedwith100 iterationsandtheParetofrontfoundonthisinstanceshowsaninterestingissue:theefcient numberofrelaysfoundwasninerelays,howeverithasthegreatestenergydissipationamong alltheefcientdeploymentsintheParetofront.Itisinterestingtoobservethatwithonlytwo 79

PAGE 89

Figure5.9:TwoefcientrelayplacementsfoundbyM-RESTalgorithm morerelaynodesrelaystheenergydissipationreducesalmost100J.Fromthatpoint, addingmorerelaynodesdoesnotsignicantlyimprovestheenergydissipation. Theresultsobtainedsuggestthatitwouldbebettertochoosethedeploymentwith11or12 relaysastheyprovideabalanceamongrelaysandenergydissipation.Nevertheless,such solutionswithmorerelaysmayreectatotallydifferentphysicalplacementoftherelays, asitcanbeseeningure5.9.Inthegure,therelayplacementontheleftisthesolutionwith ninerelays,andontheright,itistheefcientsolutionwith12relays. InordertobenchmarkM-REST,anapproximationalgorithm[62]AlgorithmAhasbeen implementedinJavaJDK6.Thisproposedalgorithmndsanapproximationtotheoptimal numberofrelaysgivensensingpoints,utilizingtheminimumspanningtreeandshortest pathalgorithmsprovidinganapproximationratioof2.Thisalgorithmwasimplementedto compareM-REST'sresultsonthenumberofrelays,anditwaschosensinceitsoptimization modelissimilaronthenumberofnodestotheM-RNPc-P.Fourtestinstanceswererandomly createdtodeployrelaysovertwogrids:asmall11x11gridandagreater35x35grid.Each gridhadaverticalandhorizontalseparationof20meters,withasensingrangeof20meters andamaximumcommunicationrangeof100meterspernode. Oneachgrid10,25,40and200sensingpointswerechosenatrandom,aswellasthepositionofthebasestation.M-RESTwasexecutedtentimes,eachtimewasrunfor100iterations. AlgorithmAwasalsoruntentimesoneachinstanceandonlytheminimalvaluefoundis 80

PAGE 90

Figure5.10:Deployedrelaysinthe11x11grid. showninthegraphs.Theresultsofbothgridsareshowningures5.10and5.11.Thebest resultsforAlgorithmAareshowninthegraphsastherhombusandtheM-RESTresultsare shownastheboxplots. FromthegraphsitcanbeobservedthatM-RESToutperformedAlgorithmAinallcases. Inthe11x11grid,theminvaluefoundbyAlgorithmAontentargetnodesisthemaximum ofallvaluesfoundbyM-REST;for25targetnodes,thebestfoundvaluebyAlgorithmA correspondstothe.75percentileofallthevaluesfoundbyM-REST.Fromthatpointon,MRESTperformedmuchbetterthantheapproximationalgorithm,almostbyafactoroftwo.In gure5.11,M-RESToutperformedcompletelyAlgorithmA. UsingthebestParetoefcientsolutionsfromtheinstancesabove,gures5.12and5.13shows thenormalizedParetofronts.Thesefrontshavebeennormalizedtoshowallfourinthesame gure.Fromgure5.13itisobservedthatthereisnottoomuchtradeoffininstanceswith10 and200targetsastherateofdecreaseinenergyissmoothincreasingthenodesdoesnotsignicantlydecreasestheenergy.Howeverthisisnotthecaseforthe25and40targetnodes instances,whereaminimumincreaseinthenumberofrelayshasdecreasedsignicantly 81

PAGE 91

Figure5.11:DeployedRelaysinthe35x35grid theenergy:intheoriginaldataforthe40targetinstance,havingonemoreavailablerelay decreasestheenergydissipationby300J,comparedtothesolutionwithlessrelays.Thisis similarinthe25targetinstance:anincreaseofonenodefromtheefcientsolutioninterms ofnodesmeantadecreaseof40Jinthedeployment. Asimilarsituationwasobservedinthe35x35gridgure5.13.Thegreatesttradeoffwas observedintheinstancewith25targets,asthereisasteepdecreaseintheenergydissipation. Herehavingaplacementwithonemorenode,meantadecreaseofalmost1kJintheenergy comparedtothesolutionwithlessrelays.Fromthatpointon,therewasnotasignifative decreaseastherateissmooth.Inthe40targetcase,havingaplacementwithsixmorerelays nodesmeantasteepdecreaseof800Jreectedinthenormalizedgraphinrangeof0to0.3 inthenumberofrelays.Forthetennodecase,addingonemorenodefromtheefcientone inrelaysmeantasavingof80Jandfromthatpointontherewasnotasignicantgainby addingmorerelaystothedeployment. Finallyforthe200targetsinthe35x35grid,havingarelayplacementwithonemorenode comparedtotheefcientsolutionintermsofnodesmeant100Jlessinenergy.Bychoosing 82

PAGE 92

Figure5.12:NormalizedParetofrontsinthe11x11Grid. Figure5.13:NormalizedParetofrontsinthe35x35Grid. 83

PAGE 93

arelayplacementwithsixmorenodes,itwasobservedthattheenergycanbedecreasedin almost300J.Fromthatpointontherewasnosubstantialimprovementbyhavingplacements withmorenodes. 5.6RelatedWork InordertosupportapplicationrequirementsinstaticWSNsdeployments,theplacementof nodesplaysanimportantrole.InstaticWSNdeployments,typicallytheobjectivestobe optimizedhavebeenthenumberofnodes[63,71],areacoverage[74],connectivity[80],networklifetime[64,67,84],anddelity[88],whilesupportingthesensingrequirements.Whenthedeploymentsareperformedinaeld,twomajormethodologies intheplacementofthenodesarepossible:thedeterministicandrandomnodeplacement. Intherstapproach,thepositionofeachofthenodesisdeterminedpreviousdeployment, anditisusefulwhenthenodesareexpensiveorhighprecisionisneededbythesensingapplication.Thelatterapproachisusefulinsituationswherenopredeneddeploymentsare possibleduetorapidchangesattheareaand/ordeploymentsinroughenvironments,suchas contaminated,disaster,andcombatzones[91]. SincecurrentcommercialstaticWSNsremainsmallnetworksduetothepriceoftheWSNs motes,mostofthedeploymentsutilizethedeterministicapproach.Here,byperformingthe carefulplacementofrelays,theWSNsadministratorsimposethenetworktopologywhile meetingthedesireddeploymentgoalsandobjectives.Basedonthehomogeneity/heterogeneityofthedeployedmotes,therelayplacementproblemsareusuallycategorizedinsingletieredandtwo-tieredproblems[86],withmanyvariationsasdifferentdeploymentgoalsmay exist[91].Inthesingle-tieredproblem,itisassumedthattherelaysensorsarefromthesame typeandhavethesamecapabilitiesasthesensornodes,thereforethesedeploymentsareat networks,whereasinthetwo-tieredproblem,relaynodesserveasgatewaysforoneorother sensors,performingusuallydataaggregation. 84

PAGE 94

Multiobjectiveapproachesfortwo-tierednetworkshavebeeninvestigatedin[72,73]where theauthorsexploretherelationshipbetweencoveragewithothervariablessuchasnumberof sensors[72],andenergyefciency[73].In[92]theauthorsexploretherelationshipsbetween severalenergyparametersanddensityofthemeasurementstoprovidefullcoverageofan agriculturalareausingtwo-tiernetworks. Insingle-tierednetworks,multiobjectiveoptimizationhavebeenutilizedtoexplorethetradeoffbetweencoverage,detectionofeventsandenergy[93]byplacingaWSNnetworknodes ina3Dspace.Thismodelassumesthatthereisaxednumberofnodestobedeployedand thealgorithmndsthelocationofthenodes.Jourdanetal.[84]proposedabiobjectivemodel tomaximizecoverageandminimizelifetimeinaatdeploymenttoprovidefullcoverageof anarea,usingaxednumberofnodes.In[71]theauthorsexplorethetradeoffbetweennumberofnodes,coverage,andsurvivabilitytoprovidefullcoverageofindustrialinstallations. Garciaetal.[94]exploresthetradeoffbetweennetworklifetimeandnumberofnodesusing aatnetworktoprovidefullcoverageofacompletearea.Intheaboveresearch,theauthors assumecompletecoverageofthearea,whichmakestheseproblemssimilartotheminimum connecteddominatingsetproblem[4]. Forthesingle-tieredprobleminthemonitoringofcriticalpointsonly,Chengetal.[63]considerahomogeneoussensornetworkonwhichallnodeshavethesametransmissionpower. Intheirwork,theauthorsattempttoplacethefeweramountofrelaysensorsinanareainterconnectingallsensingnodes.SincetheirmodelisformulatedasaSteinerMinimumTree withMinimumNumberofSteinerNodesSMT-MSPproblem,whichisshowntobeaNPCompleteproblem,anapproximationalgorithmisproposed.Bymakingageneralizationof thismodel,Lloydetal.[69]proposeanoptimizationmodelwhererelaynodeshavelonger transmissionrangethanthesensingrangeofthenodes,minimizingthenumberofsensors. ThismodelisoptimizedusinganapproximationalgorithmbasedontheMinimumSpanning TreeMSTalgorithm. 85

PAGE 95

Misraetal.[62]proposeanapproximationalgorithmforaconstrainedrelaymodelinthe single-tieredsensorplacementproblem,wheretherelaysensorscanonlybeplacedinintersectionlinesinagrid.Still,thisconstrainedversionoftheproblemisclassiedasNPHard,sincesolvingitefcientlymeanssolvingtheSteinerTreeProblem.Arelatedmodel toM-RNPc-PwasproposedbyHimsoonetal.[64]investigatingthepowerallocationina cooperativerelaynetworktomaximizethenetworklifetime,usingaxednumberofnodes asaconstrainttondthebestlocationsofthenodes.Sincetheirmodelisasingleobjective model,theevaluationofofthetradeoffbetweenpowerandnumberofnodesisperformed bychangingthenumberofnodes'constrainteverytimethealgorithmisexecuted.Thisisan disadvantagebecausesomeapproximationtothenumberofnodeshastobeknownbeforehand.Alsotheyneedtoexecutethealgorithmseveraltimes.SinceintheproposedM-RNPcPmodelbothobjectivesareoptimized,noguesshastobeperformedbeforetheoptimization processthereforeoneexecutionisenoughtogetanapproximationtotheoptimalsolutions. 86

PAGE 96

Chapter6:ConclusionsandFutureWork 6.1Conclusions Ubiquitoussensinghasreceivedagreatdealofinterestintheresearchcommunityduringthe pastyearsduetothecapacitytoutilizecellulardevicesasmotesinmobilewirelesssensor networks.ThecombinationofstaticWSNsandthepervasinessofcellulardevices,have thepotentialtoaddresslarge-scalesocietalproblemsandimprovethequalityoflifeofthe individualinabetter,lessexpensive,andfasterwaythancurrentsolutionbasedinstatic WSNsonly. USsystemsincludelocation-basedsystems,human-centricsystems,andparticipatorysensingsystems.Tosupportallthesesystems,thisdissertationproposestheG-Sensearchitecture. G-SenseintegratesstaticWSNs,mobileWSNsanddistributedserverstoprovideacommon groundinthedevelopmentofUSapplications.ThemajorcontributionsofG-Senseasan enablingarchitectureforubiquitoussensingistoprovideaframeworktosupportbothrealtimeanddelay-tolerantapplicationsinascalablemannerbymeansofapeer-to-peersystem andalsoaddresstheefcientintegrationofstaticwirelesssensornetworks. AspartofG-Sense,thecriticalpointalgorithmswerefeaturedasmechanismstoreduce thepowerconsumptionincontinuoussensingapplications,andreducetheamountofdata generatedbytheseapplications.Byutilizingthesealgorithms,amobiledevicesendsbetween 5%and30%ofthetotalamountofdatathatwouldhavetobeuploadedifnocriticalpoint algorithmhadbeenused.AsanexampleofanimplementationofG-Sense,asystemprototypehasbeenspeciedanddeveloped.ToaddresstheglobaldeploymentofUSsystems,this 87

PAGE 97

dissertationproposestheGeotellapeer-to-peersystemtointerconnectsensingserversinan scalablefashion.Geotellatakesadvantageofthedata'sgeographicalcontexttoconnectpeers andsendgeographical-basedqueriesandmessages.Finally,thisdissertationproposesamultiobjectiveapproachtoaddresstherelayplacementprobleminstaticWSNs.Theapproach minimizessimultaneouslythenumberofnodesandthedissipatedenergyofthenetwork. Themultiobjectivemodelisoptimizedbyaproposedhybridevolutionaryalgorithm.The algorithmshowedaoutstandingperformancewhencomparedtoanapproximationalgorithm thatminimizesthenumberofrelays.Alsotheproposedalgorithmshowedthetradeoffamong thenumberofrelaysforaplacementandthedissipatedenergy:choosingaplacementslightly morerelaynodesmightsignicantlydecreasetheenergydissipationofthedeployedwireless sensornetwork. 6.2FutureWork Ubiquitoussensingisafairlynewareawithstillmanyunresolvedchallenges.Thefollowing listprovidesabriefdescriptionofthemostimportantaspectstoaddressasfuturework: Validityofthedata. Mechanismstoguaranteethevalidityofthedataareveryimportant.ImagineaPSapplicationtomeasurethepollutionindexindifferentcountries thatwillbeusedtoeitherchargeorprovidefundstothecountriesaccordingtotheir indexes.Thesedatamightbemanipulatedtoprovideanunrealpictureofthepollution indexinacountry. Securityandprivacy. Mechanismstoensuresecurityandprivacyareveryimportantin alltheseapplications.Someapplicationsrequiretoprotecttherealpositionaswellas thepersonalinformationoftheuser.Achievingthesegoalsinanenergy-efcientand simplemannerformobileclientdevicesisverychallenging.Location-basedsecurityis alsoaninterestingresearchtopic. 88

PAGE 98

Activitydetermination. Algorithmstoautomaticallydeterminethetypeofactivitythat theuserisdoingareimportantforalltheseapplicationsaswell.Accelerometerandpositioningdatahavebeenusedindifferentstudiestodeterminethemodeoftransportation,thetypeofactivityandpositionoftheuserwalking,jogging,sitting,sleeping, etc.,andothers.Forsomeapplicationsthecombinationofthesedatapluspersonal datamightbeusedtomakemoreaccurateestimations.Further,severalofthesedata couldalsobeusedtotriggersensingtasksatmoreappropriatetimes,reducingthe amountofdataandsavingmoreenergy. Learningandfeatureextractionalgorithms. HCSapplicationscansendandstorelots ofdataaboutaparticularuser.Thedatabyitselfisnotveryuseful.Mechanismsneed tobedevisedtotranslatethesedataintoinformationabouttheusers,theirbehaviors, theirpreferences,etc.,sothiscontextualinformationcanbeusedtoenhancetheservice providingeffectiveandtimelyfeedback. AIalgorithmsforresource-constraineddevices. Learningandfeatureextractionalgorithmsarefairlycomputationallyexpensiveandtherefore,areusuallyruninservers. However,someoftheseapplicationsmightbenetifsomeofthesealgorithmscouldbe runintheclientdevice. Datacorrelation. Globalapplicationsprovidethecapabilitytocollectdataworldwide. Correlationandvisualizationmechanismsareneededtounderstandandseetheeffect thatthedatainoneplaceoftheglobemayproduceonothers. Incentivemechanisms. Someparticipatorysensingapplicationswillneedsomesortof incentivemechanismfortheuserstoparticipate. Datavisualization. Showingavariableofintereste.g.,pollution,temperature,etc.in amapwillneedofestimationandinferencetechniquestocompletethemapwhenthere isasmallnumberorincompletesetofsamples. 89

PAGE 99

ListofReferences [1]TurtleNet.[Online].Available:http://prisms.cs.umass.edu/dome/turtlenet [2]T.Harms,S.Sedigh,andF.Bastianini,Structuralhealthmonitoringofbridgesusing wirelesssensornetworks, IEEEInstrumentationMeasurementMagazine ,vol.Vol.13, No.6,pp.14,2010. [3]C.ChongandS.Kumar,Sensornetworks:evolution,opportunities,andchallenges, ProceedingsoftheIEEE ,vol.Vol.91,No.9,pp.1247,2003. [4]M.LabradorandP.Wightman, TopologyControlinWirelessSensorNetworks .New York,NY:SpringerScience+BusinessMediaB.V.,2009. [5]IEEE802.15WPANTaskGroup4TG4.[Online].Available:http: //www.ieee802.org/15/pub/TG4.html [6]J.Yick,B.Mukherjee,andD.Ghosal,WirelessSensorNetworkSurvey, Computer Networks ,vol.Vol.52,No.12,pp.2292,2008. [7]Skyhook:Howitworks?[Online].Available:http://www.skyhookwireless.com/ howitworks/ [8]S.Akhtar,G-4GNetworks:EvolutionofTechnologies,Standards,andDeployment, EncyclopediaofMultimediaTechnologyandNetworking ,2009. [9]M.Labrador,A.Perez,andP.Wightman, Location-BasedInformationSystems: DevelopingReal-TimeTrackingApplications .TaylorandFrancis,Chapman& Hall/CRCComputerandInformationScienceSeries,2010. 90

PAGE 100

[10]Google, "NexusOneOwner'sGuide" ,2010.[Online].Available:http://www.google. com/googlephone/NexusOneOwnersGuide.pdf [11]Apple, "iPhone4TechnicalSpecications" ,2011.[Online].Available:http: //www.apple.com/iphone/specs.html [12]Nokia, "NokiaN900TechnicalSpecications" ,2011.[Online].Available:http: //www.nokiausa.com/nd-products/phones/nokia-n900/specications [13]Motorola, "MotorolaCliqXTTechnicalSpecications" ,2010.[Online].Available: http://www.motorola.com/Consumers/US-EN/Consumer-Product-and-Services/ Mobile-Phones/ci.MOTOROLA-CLIQ-XT-with-MOTOBLUR-US-EN.alt [14]Blackberry, "BlackBerry8900" ,2009.[Online].Available:http://us.blackberry.com/ smartphones/blackberrycurve8900/ [15]A.Perez,M.A.Labrador,andS.Barbeau,GSense:AScalableArchitectureforGlobal SensingandMonitoring, IEEENetworkMagazine ,vol.Vol.24,No.4,pp.57,2010. [16]K.Virrantaus,J.Markkula,A.Garmash,Y.Terziyan,J.Veijalainen,A.Katanosov,and H.Tirri,DevelopingGISSupportedLocation-BasedServices,in Proceedingsofthe InternationalWorkshoponWebGeographicalInformationSystems ,2001,pp.423. [17]S.Barbeau,M.A.Labrador,P.Winters,R.Perez,andN.Georggi,Ageneral architectureinsupportofinteractive,multimedia,location-basedmobileapplications, IEEECommunicationsMagazine ,vol.Vol.44,No.11,pp.156,2006. [18]A.Kupper,G.Treu,andC.Linnhoff-Popien,TraX:ADevice-CentricMiddleware FrameworkforLocation-BasedServices, IEEECommunicationsMagazine ,vol.Vol. 44,No.9,pp.114,2006. [19]A.Kuepper, "Location-basedservices" .Wiley,2005. [20]A.Hand,J.Cardiff,P.Magee,andJ.Doody,Anarchitectureanddevelopment methodologyforlocation-basedservices, JournalofElectronicCommerceResearch andApplications ,vol.5,no.3,pp.201,2006. 91

PAGE 101

[21]V.Vianello,C.diFlora,andC.Prehofer,AComparisonofGISArchitecturesfor ImplementingIndoorLocation-basedServices, JournalofSoftware ,vol.4,pp.664 674,2009. [22]J.Gu,L.He,andJ.Yang,Lamoc:ALocationAwareMobileCooperativeSystem, 2009. [23]F.Hu,Y.Wang,andH.Wu,MobileTelemedicineSensorNetworkswithLow-Energy DataQueryandNetworkLifetimeConsiderations, IEEETransactionsonMobile Computing ,vol.Vol.5,No.4,pp.404,2006. [24]S.Barbeau,M.A.Labrador,P.Winters,R.Perez,andN.Georggi,TheTravelAssistant Device:UtilizingGPS-EnabledMobilePhonestoAidTransitRiderswithSpecial Needs, Proceedingsofthe15thWorldCongressonIntelligentTransportationSystems 2008. [25]I.Anderson,J.Maitland,S.Sherwood,L.Barkhuus,M.Chalmers,M.Hall,B.Brown, andH.Muller,Shakra:TrackingandSharingDailyActivityLevelswithUnaugmented MobilePhones, MobileNetworksandApplications ,vol.Vol.12,No.2,pp.185, 2007. [26]G.Castelli,M.Mamei,A.Rosi,andF.Zambonelli,ExtractingHigh-LevelInformation fromLocationData:TheW4DiaryExample, MobileNetworksandApplications ,vol. Vol.14,No.1,pp.107,2009. [27]Y.ChuandA.Ganz,WISTA:AWirelessTelemedicineSystemforDisasterPatient Care, MobileNetworksandApplications ,vol.Vol.12,No.2,pp.201,2007. [28]S.Kumar,K.Kambhatla,F.Hu,M.Lifson,andY.Xiao,Ubiquitouscomputingfor remotecardiacpatientmonitoring:asurvey, InternationalJournalofTelemedicineand Applications ,pp.119,2008. [29]U.Varshney,PervasiveHealthcareandWirelessHealthMonitoring, MobileNetworks andApplications ,vol.Vol.12,No.2,pp.113127,2006. [30]M.Demirbas,C.Rudra,A.Rudra,andM.A.Bayir,iMAP:Indirectmeasurementofair pollutionwithcellphones, IEEEInternationalConferenceonPervasiveComputingand CommunicationsPerCom2009 ,2009. 92

PAGE 102

[31]S.B.Eisenman,N.D.Lane,E.Miluzzo,R.A.Peterson,G.S.Ahn,andA.T.Campbell, TheBikeNetmobilesensingsystemforcyclistexperiencemapping,2007. [32]S.Gaonkar,J.Li,R.Choudhury,L.Cox,andA.Schmidt,Micro-Blog:sharingand queryingcontentthroughmobilephonesandsocialparticipation, Proceedingsofthe 6thinternationalconferenceonMobilesystems,applications,andservicesMobiSys'08 2008. [33]N.Lane,S.Eisenman,M.Musolesi,E.Miluzzo,andA.Campbell,Urbansensing systems:opportunisticorparticipatory?in Proceedingsofthe9thworkshoponMobile computingsystemsandapplications ,ser.HotMobile'08,2008,pp.11. [34]S.B.Eisenman,N.D.Lane,E.Miluzzo,R.A.Peterson,G.S.Ahn,andA.T.Campbell, Metrosenseproject:People-centricsensingatscale,in Proceedingsof4thACM ConferenceonEmbeddedNetworkedSensorSystemsSensys'06 ,2006. [35]Theworldin2010:I.c.t.factsandgures, ITU-DMarketInformationandStatistics 2010.[Online].Available:http://www.itu.int/ITU-D/ict/material/FactsFigures2010.pdf [36]I.JunglasandR.Watson,Location-basedservices, CommunicationsoftheACM vol.51,no.3,pp.201,2008. [37]CENSUrbanSensing.[Online].Available:http://research.cens.ucla.edu/urbansensing/ projects/nd/ [38]CambridgeMobileUrbanSensing.[Online].Available:http://www.escience.cam.ac. uk/mobiledata/ [39]I.Chatzigiannakis,G.Mylonas,andS.Nikoletseas,waystobuildyourapplication: AsurveyofmiddlewareandsystemsforWirelessSensorNetworks, Proceedingsofthe IEEEConferenceonEmergingTechnologiesandFactoryAutomation ,2007. [40]GoogleLatitude.[Online].Available:http://www.google.com/latitude/ [41]YahooFireEagle.[Online].Available:http://reeagle.yahoo.net/ [42]Facebookplaces.[Online].Available:http://www.facebook.com/places/ 93

PAGE 103

[43]S.Barbeau,R.Perez,M.Labrador,A.Perez,P.Winters,andN.Georggi,LAISYC: ALocation-AwareFrameworktoSupportIntelligentReal-timeApplicationsforGPSEnabledMobilePhones, IEEEPervasiveComputing ,2010. [44]Ubiquitoussensornetworks, ITU-TTechnologyWatchReport ,2008.[Online]. Available:http://www.itu.int/dms_pub/itu-t/oth/23/01/T23010000040002PDFE.pdf [45]J.Burke,D.Estrin,M.Hansen,A.Parker,N.Ramanathan,S.Reddy,andM.B. Srivastava,Participatorysensing, Proceedingsof4thACMConferenceonEmbedded NetworkedSensorSystemsSensys'06 ,2006. [46]V.Bychkovsky,K.Chen,M.Goraczko,H.Hu,B.Hull,A.Miu,E.Shih,Y.Zhang, H.Balakrishnan,andS.Madden,TheCarTelMobileSensorComputingSystem, in Proceedingsof4thACMConferenceonEmbeddedNetworkedSensorSystems Sensys'06 ,2006. [47]B.Hoh,M.Gruteser,R.Herring,J.Ban,D.Work,J.Herrera,A.Bayen,andM.A.Q. Jacobson,Virtualtriplinesfordistributedprivacy-preservingtrafcmonitoring, in Proceedingofthe6thinternationalconferenceonMobilesystems,applications,and services ,ser.MobiSys'08,2008. [48]A.Kansal,S.Nath,J.Liu,andF.Zhao,Senseweb:Aninfrastructureforsharedsensing, IEEEMultimedia ,vol.Vol.14,No.4,pp.8,2007. [49]JCP,"jsr256",2009.[Online].Available:http://www.jcp.org/en/jsr/summary?id=256 [50]N.Balasubramanian,A.Balasubramanian,andA.Venkataramani,Energyconsumption inmobilephones:ameasurementstudyandimplicationsfornetworkapplications, in Proceedingsofthe9thACMSIGCOMMconferenceonInternetmeasurement conference ,2009,pp.280. [51]S.Barbeau,M.Labrador,A.Perez,P.Winters,N.Georggi,D.Aguilar,andR.Perez, DynamicManagementofReal-TimeLocationDataonGPS-enabledMobilePhones, in Proceedingsofthe2ndInternationalConferenceonMobileUbiquitousComputing, Systems,ServicesandTechnologies ,ser.UBICOMM'08,2008. [52]W.ShiandC.Cheung,PerformanceEvaluationofLineSimplicationAlgorithmsfor VectorGeneralization, TheCartographicJournal ,vol.Vol.43,No.1,pp.27,2006. 94

PAGE 104

[53]D.Mendez,A.Perez,M.A.Labrador,andJ.Marron,P-Sense:aparticipatorysensing systemforairpollutionmonitoringandcontrol,in IEEEInternationalConferenceon PervasiveComputingandCommunicationsPerCom2011 ,2011. [54]J.Buford,H.Yu,andE.Lua, P2P:NetworkingandApplications .MorganKaufmann Publishers,2009. [55]OpenChord:AJavaimplementationofChordDHT.[Online].Available: http://open-chord.sourceforge.net/ [56]W.Heinzelman,A.Chandrakasan,andH.Balakrishnan,Energy-Efcient CommunicationProtocolforWirelessMicrosensorNetworks,in Proceedingsofthe 33rdHawaiiInternationalConferenceonSystemSciences ,vol.8,2000. [57]J.KnowlesandD.Corne,MemeticAlgorithmsforMultiobjectiveOptimization:Issues, MethodsandProspects,in RecentAdvancesinMemeticAlgorithms .Springer,2004, pp.313. [58]M.Penrose, RandomGeometricGraphs .OxfordUniversityPress,2003. [59]K.P.FerentinosandT.A.Tsiligiridis,Amemeticalgorithmforoptimaldynamicdesign ofwirelesssensornetworks, ComputerNetworks ,vol.33,no.2,pp.250258,2010. [60]A.Jaszkiewicz,Geneticlocalsearchformulti-objectivecombinatorialoptimization, EuropeanJournalofOperationalResearch ,vol.137,no.1,pp.50,2002. [61]Y.Donoso,A.Perez,C.Ardila,andR.Fabregat,OptimizingMultipleObjectives onMulticastNetworksusingMemeticAlgorithms, InternationalTransactionsOn ComputerScienceandEngineering ,vol.20,no.1,pp.192,2005. [62]S.Misra,S.Hong,G.Xue,andJ.Tang,Constrainedrelaynodeplacementin wirelesssensornetworks:formulationandapproximations, IEEE/ACMTransactionson Networking ,vol.18,no.2,pp.434447,2010. [63]X.Cheng,D.Du,L.Wang,andB.Xu,RelaySensorPlacementinWirelessSensor Networks, WirelessNetworks ,vol.14,no.3,pp.347355,2009. 95

PAGE 105

[64]T.Himsoon,W.Siriwongpairat,H.Z.Han,andK.Liu,LifetimeMaximizationby CooperativeSensorandRelayDeploymentinWirelessSensorNetworks,in Proceedings oftheIEEEWirelessCommunicationsandNetworkingConferenceWCNC2006 vol.1,2006,pp.439. [65]Q.Wang,G.Takahara,H.Hassanein,andK.Xu,Onrelaynodeplacementandlocally optimaltrafcallocationinheterogeneouswirelesssensornetworks,in Proceedingsof theIEEEConferenceonLocalComputerNetworksLCN05 ,2005. [66]Q.Wang,K.Xu,G.Takahara,andH.Hassanein,Locallyoptimalrelaynodeplacement inheterogeneouswirelesssensornetworks,in ProceedingsoftheIEEEConferenceon LocalComputerNetworksLCN05 ,2005. [67]Q.Wang,K.Xu,H.Hassanein,andG.Takahara,Minimumcostguaranteedlifetime designforheterogeneouswirelesssensornetworksWSNs,in Proceedingsofthe48th AnnualIEEEGlobalTelecommunicationsConferenceGlobecom05 ,2005. [68]B.Hao,H.Tang,andG.Xue,Fault-tolerantrelaynodeplacementinwirelesssensor networks:formulationandapproximation,in ProceedingsoftheWorkshoponHigh PerformanceSwitchingandRoutingHPSR04 ,2004. [69]E.LloydandG.Xue,RelayNodePlacementinWirelessSensorNetworks, IEEE TransactionsonCompututers ,vol.56,no.1,pp.134,2007. [70]X.Han,X.Cao,E.Lloyd,andC.Shen,Fault-tolerantrelaynodesplacementin heterogeneouswirelesssensornetworks,in Proceedingofthe26thIEEE/ACMJoint ConferenceonComputersandCommunicationsINFOCOM07 ,2007. [71]D.JourdanandO.deWeck,Multi-objectivegeneticalgorithmfortheautomated planningofawirelesssensornetworktomonitoracriticalfacility,in Proceedingsof theSPIEDefenseandSecuritySymposium ,2004. [72]J.Jia,J.Chen,G.Chang,Y.Wen,andJ.Song,Multi-objectiveoptimizationfor coveragecontrolinwirelesssensornetworkwithadjustablesensingradius, Computers andMathematicswithApplications ,vol.57,no.11-12,pp.17671775,2009. [73],Energyefcientcoveragecontrolinwirelesssensornetworksbasedonmultiobjectivegeneticalgorithm, ComputersandMathematicswithApplications ,vol.57,no. 11-12,pp.17561766,2009. 96

PAGE 106

[74]S.DhillonandK.Chakrabarty,Sensorplacementforeffectivecoverageandsurveillance indistributedsensornetworks,in ProceedingsoftheIEEEWirelessCommunications andNetworkingConferenceWCNC2003 ,2003. [75]T.Clouqueur,V.Phipatanasuphorn,P.Ramanathan,andK.Saluja,Sensordeployment strategyfortargetdetection,in Proceedingsofthe1stACMinternationalWorkshopon WirelessSensorNetworksandApplicationsWSNA02 ,2002. [76]C.HuangandY.Tseng,Thecoverageprobleminawirelesssensornetwork,in ProceedingsoftheACM9thAnnualInternationalConferenceonMobileComputing andNetworkingMobiCom03 ,2003. [77]S.Meguerdichian,F.Koushanfar,M.Potkonjak,andM.Srivastava,Coverageproblems inwirelessad-hocsensornetworks,in Proceedingsofthe20thInternationalAnnual JointConferenceoftheIEEEComputerandCommunicationsSocietiesINFOCOM 01 ,2001. [78]D.Pompili,T.Melodia,andI.Akyildiz,Deploymentanalysisinunderwateracoustic wirelesssensornetworks,in ProceedingsoftheACMInternationalWorkshopon UnderwaterNetworksWUWNet2006 ,2006. [79]E.BiagioniandG.Sasaki,Wirelesssensorplacementforreliableandefcientdata collection,in Proceedingsofthe36thAnnualHawaiiinternationalConferenceon SystemSciencesHICSS03 ,2003. [80]K.KarandS.Banerjee,Nodeplacementforconnectedcoverageinsensornetworks, in ProceedingsoftheWorkshoponModelingandOptimizationinMobile,AdHocand WirelessNetworksWiOpt03 ,2003. [81]J.Bredin,E.Demaine,M.T.Hajiaghayi,andD.Rus,Deployingsensornetworkswith guaranteedcapacityandfaulttolerance,in Proceedingsofthe6thACMInternational SymposiumonMobileAdHocNetworkingandComputingMobiHOC05 ,2005. [82]S.ToumpisandL.Tassiulas,Packetostatics:deploymentofmassivelydensesensor networksasanelectrostaticsproblem,in Proceedingsofthe24thIEEEConferenceon ComputerCommunicationsandNetworkingINFOCOM05 ,2005. 97

PAGE 107

[83]S.ToumpisandG.Gupta,Optimalplacementofnodesinlargesensornetworksunder ageneralphysicallayermodel,in Proceedingsof2ndIEEEConferenceonSensorand AdHocCommunicationsandNetworksSECON05 ,2005. [84]D.JourdanandO.deWeck,Layoutoptimizationforawirelesssensornetwork usingamulti-objectivegeneticalgorithm,in ProceedingsoftheVehicularTechnology ConferenceVTC2004 ,2004. [85]A.Konstantinidis,K.Yang,Q.Zhang,andD.Zeinalipour-Yazti,Amultiobjective evolutionaryalgorithmforthedeploymentandpowerassignmentprobleminwireless sensornetworks, ComputerNetworks ,vol.54,no.6,pp.960976,2010. [86]K.Xu,Q.Wang,H.Hassanein,andG.Takahara,Optimalwirelesssensornetworks wsnsdeployment:minimumcostwithlifetimeconstraint,in Proceedingsofthe IEEEInternationalConferenceonWirelessandMobileComputing,Networkingand CommunicationsWiMob05 ,2005. [87]Y.Hou,Y.Shi,andH.Sherali,Onenergyprovisioningandrelaynodeplacement forwirelesssensornetworks, IEEETransactionsonWirelessCommunications ,vol.4, no.5,pp.25792590,2005. [88]X.ZhangandS.Wicker,Howtodistributesensorsinarandomeld?in Proceedingsof the3rdInternationalSymposiumonInformationProcessinginSensorNetworksIPSN 04 ,2004. [89]D.Ganesan,R.Cristescu,andB.Beferull-Lozano,Power-efcientsensorplacementand transmissionstructurefordatagatheringunderdistortionconstraints,in Proceedingsof the3rdInternationalSymposiumonInformationProcessinginSensorNetworksIPSN 04 ,2004. [90]Q.Wang,K.Xu,G.Takahara,andH.Hassanein,Deploymentforinformationoriented sensingcoverageinwirelesssensornetworks,in Proceedingsofthe3rdInternational SymposiumonInformationProcessinginSensorNetworksIPSN04 ,2006. [91]M.YounisandK.Akkaya,Strategiesandtechniquesfornodeplacementinwireless sensornetworks:Asurvey, AdHocNetworks ,vol.6,no.4,pp.621655,2008. 98

PAGE 108

[92]K.P.FerentinosandT.A.Tsiligiridis,Adaptivedesignoptimizationofwirelesssensor networksusinggeneticalgorithms, ComputerNetworks ,vol.51,no.4,pp.10311051, 2007. [93]C.KangandJ.Chen,Anevolutionaryapproachformulti-objective3Ddifferentiated sensornetworkdeployment,in Proceedingsofthe11thAnnualConferenceCompanion onGeneticandEvolutionaryComputationConferenceGECCO09 ,2009. [94]G.Molina,E.Alba,andE.Talbi,Optimalsensornetworklayoutusingmulti-objective metaheuristics, JournalofUniversalComputerScience ,vol.14,no.15,pp.2549 2565,2008. 99