A Concurrent Multi Target Tracker: Benchmarking and Portability

 Internet & Technology

 2 views
of 21
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Description
With the current advances in computing and network technology and software, the gap between parallel and distributed computing environment is gradually becoming narrower. Consequently, parallel programs run on parallel as well as distributed systems.
Share
Tags
Transcript
  Northeast Parallel Architectures Centerat Syracuse University AConcurrentMultiTargetTracker:Benchmarking andPortability  1 SalimHariri,RajeshYadav,BalajiThiagarajan,Sung-YongPark,MaheshSubramanyan,RajashekarReddyandG.C.Fox NortheastParallelArchitecturesCenterDepartmentofElectricalandComputerEngineering SyracuseUniversity  SCCS#659  1 ThisresearchisfundedbyRomeLaboratory(contractnumberF-30602-92-C-0063),Rome,NY   1INTRODUCTION  2  Abstract Withthecurrentadvancesincomputingandnetworktechnologyandsoftware,thegapbetween parallelanddistributedcomputingenvironmentisgraduallybecomingnarrower.Consequently,parallelprogramsrunonparallelaswellasdistributedsystems.However,programmingand portingcomplexapplicationstosuchenvironmentischallengingtaskandnotwellunderstood.Inthispaper,weuseaconcurrentmultitargettrackerasarunningexampletoanalyze andevaluateperformanceoftwodierentparallelimplementationsonparallelanddistributed systems.Wehavebenchmarkedboththeseimplementationsondierentarchitecturesthatvaryfromanetworkofworkstations(SUN,IBMRS6000)toparallelcomputers(CM5,iPSC 860)usingdierentparallel/distributedmessagepassingtools(PVM,p4,EXPRESS,HORUS).Wehavealsocomparedtheperformanceofthesetoolstoprovideimportantcommunication primitivesforhighperformancecomputingapplicationssuchassend/receive,broadcastand circularcommunication. 1Introduction  Currentadvancesintechnologyisunifyingparallelanddistributedcomputingsuchthatthe distinctionbetweenthetwoeldsisbecomingmoreandmoreblurred.Forexample,anap-plicationdevelopedbasedonmessagepassingprogrammingparadigmcanrunonparalleland distributedenvironmentswithoutanymajorchanges.Suchaparadigmcancapitalizeonex-istingarchitecturesandonenhancedfacilitiesincomputing,networkingandcommunication technologytoprovideecient,cost-eective,scalable,andhigh-performancesolutions.Soft-waredevelopmentforcomplexapplicationthatrunonparallelanddistributedsystemsisa non-trivialprocessandrequiresathoroughunderstandingoftheapplicationandarchitectures;currently,mostofthewellstudiedapplicationsaresmallandarenotcomplexenoughtohelp researchersunderstandtheprogramming,portabilityandperformanceissuesassociatedwith softwaretoolsforparallelanddistributedcomputingenvironments.Themainobjectiveofthispaperistoinvestigatetheseissuesbyexperimentingwithacomputeintensiveapplication.Weselectthemultitargettrackingapplicationastherstapplicationinabenchmarking suitebeingdevelopedatNPACofSyracuseUniversitytoevaluateperformanceofsoftware toolsaswellasparallel/distributedplatforms.Thetrackerhasbeenwrittenusingdierentmessagepassingtools ?  ]thatincludePVM ?  ],p4 ?  ],EXPRESS ?  ],andPICL ?  ].The trackerhasbeenportedonthenCUBE,CM5,InteliPSC860,IBM-SP1architecturesanda clusterofSUN4andIBMRS6000workstations.Tosimplifytheportabilityissueinaheterogeneousenvironmentwehavedevelopeda uniformstructureofthetrackeralgorithmthatcanbeeasilycodedusingdierentmessage passingtools.Bystudyingtheperformanceofthetoolsandtrackerondierentparalleland distributedsystems,wewillbeabletoidentifythebestcomputingenvironmenttorunthatapplication.Theremainingsectionsofthepaperareorganizedasfollows:Insection2wediscussthe sequentialimplementationofthetracker.Section3discussestwodierentimplementationsoftheconcurrentmultitargettracker.Section4detailstheexperimentalresultsoftheparallel NPAC/ECE,SyracuseUniversity  2MULTITARGETTRACKINGSYSTEM  3  anddistributedimplementationsofthetracker.Italsocomparestheperformanceofmes-sagepassingtools(likePVM,p4,EXPRESS,HORUS)toprovideimportantcommunication primitivesforhighperformancecomputingapplications.Section5presentsasummaryand directionsoffutureresearch. 2MultiTargetTrackingSystem  ThetrackerdemonstratesthemultitargettrackingcapabilitiesthatisrequiredbyaBattle ManagementCommandControlandCommunicationSystem.Itusesanextended3stage Kalmanlteringformalismwhichistheprimary\tool"usedtoprovideandsortrealistic data.Thislteringformalismisgeneralandcanbeusedinproblemsrelatedtopattern recognition,signalandimageprocessing.The3stageltermodelhashelpedthedevelopmentofaconcurrentversionofthetracker ?  ].Themultitargettracker,showninFigure  ??  ,isdesignedtoprovideanestimationoflaunch vehicleparametersforindividualtargets/missilesinmulti-targetscenarios.Thesystemdealswithamassraidscenarioandisdesignedtoprocesssituationswithvaryingnumberoftargetsandlaunchsites.ThetrackerreceivesinputfromtheEnvironmentGeneratorandSynthesizermodule(seeFigure  ??  )intermsofsensorscansandtargetinformation.Themultipletargettrackingsystemhastwogeostationarysensorswhichscanspeciclaunchsitesformissilesortargetslaunchedfromthesurfaceofearth.Thelaunchsitesarespeciedintermsoflatitudesandlongitudes.Thedatafromthesetwogeostationarysensorsarefedtotwofocalplanetracking(FPT)modules(2dimensionaltracking)at5secondintervals.Thefocalplane trackingmodulesprocessthisdatausingkinematiclteringalgorithmsandtrackpruningand predictionalgorithms.Theoutputofthismoduleisaninitialpredictionoftrajectoriesoflaunchedmissiles.Thisdataisthenfedtoa3D(threedimensional)trackingsystemwhich usesthedatafromthetwofocalplanetrackingmodulestopruneduplicatetracks(ifany),extendexistingtracks,prunebadtracksandinitiatenewtracks.Theoutputofthesystemisalistoftargettrajectories.Theconcurrentmultitargettracker(CMTT)isoneofthemodulesthattsintothe battlemanagementcommandcontrolandcommunicationsystem(seeFigure  ??  ).Thissystem consistsofasetofcomponentswhichinteractwitheachotherbyexchanginginformation fordataprocessing.ThemaincomponentsoftheoverallsystemincludeanEnvironmentGeneratorandSynthesizer,theTrackerSystem,DecisionControlSystemandanExternalGraphicsCommunicationModule,asshowninFigure  ??  .TheMTTsystemgetsinformationfromtheEnvironmentGeneratorandSynthesizer.ItprocessesthisdataandgeneratestheparametersrequiredbytheDecisionControlSystem.TheDecisionControlSystemusesthisinformationalongwithdatafromtheEnvironmentGeneratorandSynthesizertomakedecisionsonhowtomanagetheexistingbattlemanage-mentcommandandcontrolscenario.ThisinformationisfedtoanExternalGraphicsCom-municationmodule.Thismoduleactsasanextensionorfront-endtotheDecisionControlModule.ForexampleinaMissileTrackingsystemthefront-endcanservesasavisualization toolordatainterpreterwhichshowsthetrajectoryandpositionoflaunchedmissiles.The FireControlModuleperformstheactionsdirectedbytheDecisionControlSystem. NPAC/ECE,SyracuseUniversity  2MULTITARGETTRACKINGSYSTEM  4  2 D Mono Tracking 2D Mono Tracking 3D Tracking System Surface of Earth Sensor 2 Sensor 1 Figure1:Multi-TargetTrackerSystem   Multi-target Tracker Multi-target Tracker Environment Generator  and Synthesizer Decision Control SystemControl  Fire External GraphicsCommunication Figure2:BattleManagementCommandControlScenario  NPAC/ECE,SyracuseUniversity  2MULTITARGETTRACKINGSYSTEM  5  Inwhatfollowswediscusstheissuesinvolvedinsingletargettracking,multitargettracking andconcurrentmultitargettracking. 2.1SingleTargetTracking(STT) Trackingofasingletargetinvolvesthreemainsteps ?  ]:1.Focal-planekinematicKalmanltering 2.Estimationofboosterlaunchparametersfromfocal-planestatevectorsand 3.Combinationofindividualparametersusingasecondlter.Thelterintherststep(alsocalledafocalplanelter)processes2Dmeasurementsfrom thesensortoformestimatesofprojectedkinematicquantitiesasseeninasensorfocalplane.Position  (x) ,velocity  (v) ,acceleration  (a) andjerk  (j=da/dt) areusedforthestatevariablesinthelter.Stochasticcontributionsareintroducedtothejerktoallowtheltertorespond totargetstravellingalonglargelyunconstrainedtrajectories.Thisdynamicuncertaintyisthe onlyfreeparameterofthefocalplanelter.Sincethelterislinear,allgainandcovariance matricescanbecomputedandtabulatedduringinitializationofthetrackingprogram.Thisimprovestheperformanceofactualltering.Themainoutputofthefocalplanelterisan estimateforthereducedstatevectorconsistingofprojectedpositionsandvelocitiesatscan  t k . x  =( x;v;a;da=dt )(1)where  x  speciestheposition, v  thevelocity, a  theaccelerationand  da/dt denotesthejerk introducedinthelter.Inthesecondstepofthe3stageltermodelthestatevectorinEquation1isusedtoprovide anestimateofthelaunchparametersofthetarget.Anindividualtrajectoryisspeciedbya 4-componentparametervectorasshownbelow: p  =(   l ;   l ;   l ;t l )(2)where    l denotesthelatitudeofthelaunchsite,  l thelongitudeofthelaunchsite,  l denotestheinitiallaunchazimuth,and  t l thelaunchtime.Thethirdstepsimplyinvertsthe relationusingtheNewtonRaphsoniterationmethod ?  ].TheNewtonRaphsoninversionpro-videsrealisticparametercovarianceestimatesaswellasvaluesoftheparametersthemselves.Forthemulti-targettrackingdescribedinthenextsection,thetaskoflteringanindividualtrackisessentiallyconcerningwithresolvingtrack-hitambiguities.Thefocalplanelterin SingleTargetTrackingisinfacttheprimarytoolusedinsortingoutthenatureofmultiple targetthreats. 2.2MultiTargetTracking(MTT) Givenareasonablelterforprocessingsensordatafromasingletarget,theessentialproblem ofmulti-targettrackingisthatofassociatingindividualsensorreportswithdistinctunder-lyingtracks.Thisobjectiveisaccomplishedthroughthreeadditionalmodules:TrackSplitProcessor,TrackInitiator,andReportFunction. NPAC/ECE,SyracuseUniversity
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks