2. Knowledge-Based Modeling and Simulation 2.1. Mo

KBS:AnArtificialIntelligenceApproachtoFlexibleSimulationY.V.ReddyandMarkS.FoxCMU-RI-TR-82-1RoboticsInstituteCarnegie-MellonUniversityPittsburgh,Pennsylvania1521314September1982Copyright@1982Carnegie-MellonUniversityThisresearchwassupportedinpartbytheWestinghouseCorporation,andbytheCMURoboticsInstitute.iTableofContents1.Introduction2.Knowledge-BasedModelingandSimulation2.1.ModelingEntitiesandRelations2.2.RulesofBehavior2.3.ModelLibraries2.4.SimulationViaModelInterpretation3.AnExample4.ModelInstrumentation5.Circuit-BoardProductionExample6.ModelConsistencyandCompleteness7.Multi-levelSimulation8.Conclusion9.References124891111151622272930iiListofFiguresFigure2-1:MachineSchemaFigure2-2:MachineSchemawithvaluesFigure2-3:MachineSchemawithfacetsFigure2-4:MachineSchemawithcharactersFigure2-5:Continuous-MachineSchemaFigure2-6:nc-drillSchemaFigure2-7:nc-drill-1withPART-OFFigure2-8:LoadSchemaFigure2-9:load-ruleSchemaFigure2-10:AgentHierarchyFigure2-11:AnExampleofanEventNoticeFigure3-1:discrete-machineSchemaFigure3-2:machine1SchemaFigure3-3:machine2SchemaFigure3-4:queue1SchemaFigure3-5:SchemaDefinitionfortheCurrentModelFigure3-6:arrival-rule1SchemaFigure3-7:prime-ruleSchemaFigure3-8:CalendarSchemaAfterExecutionofPrimeEventFigure3-9:event1SchemaFigure3-10:event3SchemaFigure4-1:DataGatheringRulesFigure4-2:DisplayInstrumentationFigure5-1:work-areaSchemaFigure5-2:discrete-machineSchemaFigure5-3:LoadRuleforDiscreteMachinesFigure5-4:OperationSchemaFigure5-5:InitializationofSimulationDisplayFigure5-6:SimulationSnap-shotFigure5-7:ModelQueryFigure5-8:DisplayofAnOrderQueryFigure5-9:PcboardFactoryLayoutFigure5-10:LayoutofasingleWorkAreaFigure6-1:for-allschemaFigure6-2:there-existsschemaFigure7-1:machiningSchemaFigure7-2:millingSchemaFigure7-3:grindingSchemaFigure7-4:polishingSchema556778889101112121313131414141515161717181919202122232425272728282929Abstract‘ThisreportdescribesKBS.aKnowledge-BasedSimulatjonsystem.ThereportdescribestlieuseofSRI,,anAI-basedknowledgerepresentationsystemformodelling(e.g.,factoryorganizations),anditsinterpretationfordiscretesimulations.KBSprovidesfacilitiesforinteractivemodelcreationandalteration,simulationmonitoringandcontrol,graphicsdisplay,andselcctivcinstrumcntrition.Italsoallowstheusertodefineandsimulateasystematdifferentlevelsofabstraction,andtochcckthecornpletcixssandconsistencyofamodel,hencereducingmodeldebuggingtime.11.IntroductionInthesummerof1980webeganthestudyofproblemsinmanagingcomplexorganizationssuchasjob-shopfactories.Ourpurposewastodiscoverwhereintelligentsystemsmayaidintheachievementoforganizationalgoals.OuranalysisresultedintheformationoftheIntelligentManagementSystem(IMS)Project(Fox,1981).IMSisalong-termprojectconcernedwithapplyingartificialintelligencetechniquesinaidingprofessionalsandmanagersintheirdaytodaytasks.ResearchinIMSisproceedinginmanyareas,including:job-shopscheduling,flexiblesimulation,processdiagnosis,organizationmodeling,anduser-interfaces.ThispaperdiscussesflexiblesimulationresearchinIMS.Acommonlyoccurringprobleminmanagementistheinabilitytoanswer“whatif”questionsreadily.Inasurveyofquestionsposedbymanagersinthreeplants,manywereconcernedwiththeeffectofproposedchangesinfactoryorganization.Somecouldbeansweredbaseduponpreviousexperience,somebyanalysis,butmanywentunanswered.Whydothesequestionsremainunansweredwhentoolsexistforanalyzingorganizations.Inparticular,simulationsystemsareusedtomeasureperformanceofexistingorproposedsystemswhicharetoocomplextobestudiedanalytically.Itisthecostofconstructionthatlimitstheiruse,andresultingsystemsarelittleusedexceptwhenrunningthesameorsimilarsimulationsagain.Moreimportantly,simple“whatif”questionscannotbeansweredreadily.Amanagerrequiresanintermediary,suchasasystemanalyst,toanswerthem.Thereisadefiniteneedformoresophisticatedtoolsforanalyzingorganizations,andforprovidingusabletoolsdirectlytothemanagersandprofessionals.InthispaperwedescribeKBS,aKnowledge-BasedSimulationsystem‘.Ourreasonsforcreatingyetanothersimulationsystemarenumerous.Inparticular,issueswehaveexploredinKBSinclude:.0creatingasystemmodelinglanguagethatcansimultaneouslysupportmultipleapplicationsinadditiontosimu!ation.Thuseliminatingtheneedandcostofmaintainingmultiplemodels.0representingthebehaviorofsystementitiesdirectlyinthemodel.Thisadmitstotalflexibilityincreatingandalteringentitiesandtheirbehavior,withoutalteringthesimulationmodelinterpreter.0allowingthesystemtobeselectivelyinstrumented.Thisrestrictsdataanalysistoareasofinterest,andprovidessupportofgraphicsdisplays.0representingthesystematmultiplelevelsofabstraction.Thisallowstheusertospecifythelevelofsimulationandthedetail-levelofresults.0consistencyandcompletenesschecking.Muchtimeisspentverifyingthatmodelsareconsistentandcomplete.Wehavedevelopedacheckerwhichdetectsmodelincompletenessandinconsistencies.0providinginteractiveaccesstothemodelbuilding‘Theterm“Knowledge-BasedSimulation’’hasappearedbeforeinanotherartificialintelligenceapproachtosimulation.andsimulationsystem.ThisappearstheresearchofKlahr8Fought(1980),indescribin