drawback.netcms

KI-KünstlicheIntelligenz(2019)33:111–116https://doi.
org/10.
1007/s13218-019-00591-4EDITORIALSpecialIssueonSmartProductionMartinRuskowski1·TatjanaLegler1·MichaelBeetz2·GeorgBartels2GesellschaftfürInformatike.
V.
andSpringer-VerlagGmbHGermany,partofSpringerNature20191ResearchOverviewTodayalotofdiscussionisheldaboutthepositionsandrolesofAIresearchinEuropecomparedtotheUSandChina.
WhereasUScompaniesfocusonconsumerdrivenapplicationslikesocialmediaandsearchmachinesandChinaispredominantinstatedrivenAIprojectslikefacerecognition,thereseemstobeaconsensusthatthebusinesstobusinessareaandespeciallyproductionisthestrengthofEuropeandthusshouldbethemajorapplicationfieldforEuropeanAI.
ProductionisstronglycoupledwithautomationandthelateststepintheindustrialdevelopmentofthepastcenturiesisIndustrie4.
0asanoutputoftheGermanhigh-techstrat-egy.
InthepastyearsmanypeopleconsideredIndustrie4.
0tobeatechnologicalchallengeonly.
Butitbecameapparentthatthereplacementofonetechnologywithanotherdoesnotleadtonewbusinessmodelsnorcostreduction.
Insteadofthis,Industrie4.
0mustbediscussedonamoreabstractlevel.
FromtheverybeginningtheuseofAImeth-odswasapartoftheIndustrie4.
0idea.
ButAIdoesnotmeanmoresophisticatedautomation.
Onthecontrary,insomeapplicationswehavealreadyreachedapeakofauto-mationlevel,sincehighlyautomatedsystemstendtobelessresilientthansystemswithhumanworkers.
NotonlyElonMuskmadetheobservationthat"humansareunderrated"inafamoustweetwhenTeslawasstrugglingwithsettinguptheModel3production[22].
Otherthantooptimizeautomationprocessesasymptoti-cally,Industrie4.
0hastoaddresscompletelyotherfieldsinproduction.
Todaymostinefficienciesoccur(andmostmoneyislost)inthenon-productiveprocesseswithinpro-duction,i.
e.
thetransportofdata,brokendataflows,dupli-cateentryofdata,manualreprogrammingofmachinesetc.
AImustbeusedtobuildintelligentmachinesinaveryspecialway.
Weneedautonomousmachinesinaway,thatallinformationneededtooperatethemachineisincludedinit.
Asoftodaytheinterfaceofamachinetoahigherlevelsupervisorycontrolmainlyconsistsofbinaryswitchesandcorrespondingparameters.
Theuseofthemachine(i.
e.
theprogrammingoftheinterfaceonthesupervisorycontrolside)needsalotofinternalknowledgeaboutwhatbitstosetatwhichstepandwhattoexpectastheanswerfromthemachine.
Theintroductionoffield-buscommunicationhasonlyreplacedtheformercablesbetweenmachines–thecommunicationconceptitselfusuallyisstillthesameasinearlydaysofelectro-mechanicalsystems.
Opposedtothis,intelligentmachineshavetobehaveinaformofamulti-agentsystem.
Theintelligenceofamulti-agentsystemdoesnotresultfromasophisticatedintelli-genceinsidethesinglemachine.
Thewholesystemconsist-ingofautonomousmachineswillbehaveinanintelligentwayalthoughthesinglemachinehasalimitedscope.
ManypeoplestillassociateAIinproductionmainlywithroboticsasintelligentmachines.
Butwhenlookingintothedetailstherearefarmorefieldsofapplication.
Someofthemainareasareamongothers:RoboticsVisualinspectionQualitycontrolDataanalyticsPredictivemaintenanceProductionplanningMultiagentsystemsHumanmachineinteraction*MartinRuskowskimartin.
ruskowski@dfki.
deTatjanaLeglertatjana.
legler@dfki.
deMichaelBeetzbeetz@cs.
unibremen.
deGeorgBartelsgeorg.
bartels@cs.
unibremen.
de1DeutschesForschungszentrumfürKünstlicheIntelligenz(DFKI),TrippstadterStrae122,67663Kaiserslautern,Germany2InstituteforArtificialIntelligence(IAI),UniversityofBremen,AmFallturm1,28359Bremen,Germany112KI-KünstlicheIntelligenz(2019)33:111–1161.
1AIfromtheEngineeringPointofViewHistorically,AIisseenasacomputersciencetechnology.
Whenitcomestoproductionwemeetthephysicalworldwithallitsimplicationsandchallenges.
ForanengineeritisinterestingtolookatAIfromadifferentangle.
Humansalwaysactinaclosedloop—wecollectinputdatafromourenvironment,trytounderstandit,makedecisionsbasedonthisunderstandingandactaccordingtothesedecisions.
Whenwecomparethisbehaviortoatechnicalcontrolloopwefindsimilaritiessuchassensors,stateobservers,controllersandactuators.
ThusfromanengineeringpointofviewAIcanbeconsideredasclosedloopcontrolinaverygeneralizedway.
Thisleadstoasignificantobservation—inaphysi-calenvironmentAIincorporatesthroughanapplica-tion.
Neuralnetworks,decisiontrees,automateddeci-sionmakingsystems,augmentedandvirtualrealityandmanyothertechnologieswhichusuallyareconsideredtobeAIaremerelymethodstoachieveAI—ormorespe-cificallymachinesbehavinginwayweconsidertoappearintelligent.
TheanalogytoclosedloopcontrolprovidessomehintsofthelimitationsforAIbasedsystems.
Propertieslikeobservabilityandcontrollabilityarewellknowninengi-neering.
Observabilitydescribestheabilitytoreconstructtheinternalbehaviorofasystempurelyontheobservationofitsoutputs.
Ontheotherside,controllabilityisimpor-tantfortheabilitytomanipulateallpartsofanexistingsysteminadesiredwayjustfromthesysteminputs.
AIinaclosedloopcanonlybesuccessfulifasystemhasbothproperties.
Evenmoreimportantisthestabilityofaclosedloopsystem,sinceitcanbedifficulttodesignastablecontrolloopevenforsimplesystems.
TheproperdesignofanAIbasedfeedbackhascompletelydifferentrequirementsforstabilitythanconventionalsystems,thusrequiringsignificantfutureresearch.
AIapplicationscanbedividedintothreemainareas:The"visible"onesrelatedtomanipulationtasksandrobots,thoserelatedtothemere"hidden"tasksofproduc-tionplanningandinter-machinecommunicationincyber-physicalproductionsystemandfinallythefieldofhumanmachineinteractionandnewuserinterfaces.
1.
2CyberPhysicalProductionSystemsSincetheintroductionofthetermIndustrie4.
0in2011[16]researchhasprocureditselfwiththequestionhowtoachieveanadaptive,self-configuringandthereforeflex-ibleproduction.
Oneofthekeycomponentstoachiev-ingthatconceptarecyber-physicalproductionsystems(CPPS)[13].
Theymakeuseofsensorsandactuatorstocaptureandinfluencethephysicalworldandareintercon-nectedvialocalnetworksortheinternet.
Sincecommuni-cationnolongerhastobehierarchical,therearetenden-ciesawayfromtheclassicautomationpyramidtowardsadecentralized,partiallyself-organizingnetwork[15].
Duetotheoftencriticalroleoftheunits,faulttoleranceandtheabilitytorespondandadapttoadisruptiveeventisstillamajorresearchtopic[7].
Productionsystemsareusuallycontrolledbyprogramma-blelogiccontrollers(PLC).
Theirprogramminglanguageshadbeenderivedfromelectricalcircuitryandwiring[24].
Sincethe1980sfieldbuseshavebeenusedtoconnectPLCbasedautomationsystems.
Unliketocomputerscience,wherethedevelopmentconvergedintoTCP/IPasthepri-marycommunicationtechnology,automationstuckwithahugenumberofdifferentandincompatiblefieldbuses[33].
Butonethingiscommontothesesfieldbuses:Theymerelyreplacediscretewiringwithasinglecablebutdonotchangetheconnectionprinciple.
Whereasincomputersciencesophisticatedprotocolsandclient/serverbasedcommuni-cationhasbeenstate-of-the-artforalongtimeafieldbusmainlyemulatesdigitalandanaloguesignalsonbothsidesoftheconnection[8].
Onedependencycanbeconsideredascrucial:Automa-tionofmachinesreliesondeterministicreal-timebehavior[18,34].
Variableruntimesoftasksandunpredictablecom-municationdelaysarenotacceptable.
Butreal-timebehaviormustnotbeconfusedwithhighCPUpower,asitismainlyrelatedtothearchitectureoftheoperatingsystem[35].
WellknowndesktopoperatingsystemslikeLinuxandWindowsarenotsuitableforreal-timecontrolanddedicatedreal-timeoperatingsystems(RTOS)arerequiredinstead.
ThisisoneofthemainparadigmsforEdgecomputinginautomation[9].
Whendefinedreactiontimesareneeded,adeterministictransferofsimplesignalsisveryrobustandreliable.
Eve-rybodywhohasdevelopedanembeddedreal-timesoftwarealsoknowsabouttheproblemsindebuggingacomplexmultitaskingsystem.
Incontrasttothis,thesimplicityofPLCprogrammingandthepopularIEC61131-3program-minglanguagesalongwithtoday'ssophisticatedengineeringsystemsmakereal-timeprogrammingmoreaccessibleandeasiertounderstand[14].
Nonethelessweencounteradrawbackofthisstructurewhenitcomestocyber-physicalproductionssystems.
Wheneverthecommunicationisbasedonsimplebinarysig-nalsthecommunicationpartnersneedaninternalknowledgeofeachothersbehavior.
Acommandissentviasettingabitandthereplyreceivedbyadifferentbitandflowcontrolismainlyperformedwithsimplestatemachines[31].
Usuallyseveralmachinesarecombinedtoformaclusterofdevices,aproductioncell.
Oftentherearereal-timedependencies113KI-KünstlicheIntelligenz(2019)33:111–116betweenthedevices,i.
e.
betweenamachinetool,ahandlingrobotandaconnectingPLC.
Motionsmustbesynchronized,switchtimescoordinatedandmanymoreconditionsmet[37].
Weoftenfindconfigurationswithoutanexplicitmasterfunctionality,merelythestatecontrolofthewholeproduc-tioncellisdistributedamongtheindividualmachines.
Thesteptowardssmartproductionrequiresanewapproachtotheinterfacingofproductioncells.
Itisneces-sarytoencapsulateallfunctionsofthecellsothereisnoneedforreal-timecommunicationtotheoutside.
Allreal-timedependenciesshallbehandledwithinthecellitself.
Theresultingentitycanthenbecalledacyber-physicalproduc-tionmodule(CPPM)[27].
Anecessarystepisthedefinitionofamastercontrollerinterfacingtotheindividualmachinesprovidingaservice-orientedinterfacetotheorchestrationlayerofthecyber-physicalproductionsystem.
Theresultingsystemarchitectureshowsacertainanalogytomulti-agentsystem[19].
Forthebroadacceptanceofcyber-physicalproductionsystemsastandardizationofthearchitecture,theinterfacesandtheorchestrationiscrucial.
Anumberofcontributionsinthiseditionshowpossibleapproachesfortheseinterfacesandfortheorchestrationlayerpavingthepathtowardsrealsmartproduction.
1.
3AIBasedRoboticsSimilarlytotheengineeringperspective,whitepapersandroad-mapsforAI-basedroboticsidentifysmartproductionasoneofitskeytargetdomains[2,30].
Indeed,economicpressure,thedesiretobringbackproductiontohigh-wagecountries,aswellassupportinganagingwork-forcemoti-vateintenseresearchactivitiestowardsmoreintelligentroboticagentsontheshopfloor.
Importantresearchtrendsthatcurrentlyexperiencemassiveresearchinterestincludecloudrobotics,easyprogrammingthroughimitationlearn-ing,accomplishingcomplexmanufacturingtaskswithouttheneedforfixtures,digitaltwinrepresentationsoffacto-riesandmanufacturingprocesses,adaptablemanipulationsolutions,objectperceptioninunstructuredenvironments,andmachinelearningtolearngraspposes,failuremonitors,objectrecognition.
ThecurrenttechnologicalwaveinAIistoalargeextenddrivenbyautomatedmachinelearningtechnologies,inpar-ticulardeeplearning.
Givenmassiveamountsofannotatedtrainingdata,supervisedmachinelearningtechniqueshavebeensuccessfullyappliedtoreal-worldperceptiontasksandevensimplemanipulationtaskssuchasbinpickingandfetchingalargevarietyofobjects[6,20,28].
Thesetech-nologiesenabledeveloperstoimplementhigh-performanceperceptionandactioncapabilitieswithreasonableprogram-mingefforts.
Anotherwayofeasingtheprogrammingofrobotsistheincreasedapplicationofimitationlearningtoassemblyandothermanipulationtasks.
Imitationlearningcanbeseenasaformofprogrammingbydemonstration[5],wherethelearningsystemusesdeepermodelsofactions,includingintentionsandstructuredmotionmodels.
[26]Amorerecenttrendistorealizeimitationlearningmethodsthroughsim-ulation-basedvirtualrealityenvironmentsinsteadofbeinglimitedtovision-basedobservationdata.
Thisisapromisingapproachbecausebyaccessingtheunderlyingdatastruc-turesofthesimulationenginesonecanoftengenerateaccu-ratelyannotatedlearningdatathatconstitutegroundtruthandwouldotherwisebehardtoobtain[12].
Anotherimportanttrendisknowledge-basedrobotpro-gramming[32].
ThisapproachisinparticularpromotedbyGilPratwhostatesthat:"Robotsarealreadymakinglargestridesintheirabilities,butasthegeneralizableknowledgerepresentationproblemisaddressed,thegrowthofrobotcapabilitieswillbegininearnest,anditwilllikelybeexplo-sive.
"[25]Formanyyears,theapplicationofknowledge-basedprogrammingtechniqueshavebeenhinderedbyknowledgerepresentationtechniquesbeingtooabstract.
Recently,newtechniqueshavebeenproposedthatrepresentsymbolicknowledgeatgeometriclevelwhichisnecessaryforproperlyparameterizingrobotmotionsforaccomplish-ingmanipulationtasksandavoidundesiredsideeffects[4].
Anotherapproachtomaketheprogrammingofsomanymanipulationapplicationfeasibleistocrowdsourcetheprogrammingtasks.
Amajorbarrierincurrentrobotpro-grammingisthatrobotsaretypicallyprogrammedforanindividualcombinationoftasks,robots,andenvironments.
Currently,thereislittlere-useofcodefromoneapplicationtoanotherone.
Tosurpassthisbarrier,cloudroboticshasproposedthatdevelopersprovidecodepiecesandcomputa-tionservicesinmoregeneralformssuchthattheycanbere-usedbyothers.
ThisapproachwaspioneeredintheEUFP7projectRoboEarth[36],andfurtherpushedinKenGold-berg'sinitiativeforcloudrobotics[17].
Today,weseethathigh-techITcompanies,includingAmazon,Google,andMicrosoft,areallproposingtheirowncloudplatforms123.
Oneofthemainissuesthatslowdowntherealizationofnewproductionprocessesinthefactoryflooristheneedfordesigningandcreatingfixturesthatmaketherobots'manipulationactionsreliableandfast.
Removingtheneedforsuchfixturesrequirestohavebetterandmoreflexiblehand-eyecoordination,aswellashighersinglearmanddual-armmanipulationcapabilities[1,21].
1https://aws.
amazon.
com/robomaker/2https://cloud.
google.
com/cloud-robotics/3https://azure.
microsoft.
com/en-us/114KI-KünstlicheIntelligenz(2019)33:111–116Anotherresearchdirectionisthedevelopmentofmanipu-lationrobotsthatassisthumansintheirmanipulationtasks.
Suchrobotshaveparticularhighdemandsontheircognitivecapabilities[3].
Thisisbecausetheydonotonlyhavetoplanandexecutetheirownactionsbutratherunderstandwhattheirhumanco-workersneedintermssupport.
Theserobotsareparticularlyimportantformanipulationtasksthatarepotentiallyhazardousorergonomicallyunhealthy.
Astheserobotsshareworkspaceswithhumanco-workerstheyneedtoguaranteethesafetyofthehumans[10,11].
Examplesofparticularlyexpressiveandpowerfulcognitivecapabilitiesforsuchrobotsthatarecurrentlyunderresearchincludethelearningofhumanpreferencemodelsandsimulation-basedmechanismsforperceptivetaking[23,29].
ThisspecialissueofKIpresentsseveralofthelead-ingdevelopmentsthatwillhelppushAI-basedroboticsalongthepathoutlinedbypublicroad-mapsandeconomicdemand.
2AbouttheSpecialIssueCyber-physicalsystemsandAI-basedrobotsareincreasinglyimportantbecauseagrowingnumberofindustrialapplica-tionshastoflexiblychangeoratleastcustomizeitsproduc-tionrelativelyoften.
Hence,supportingfrequentadaptationswithoutsignificantadditionalinvestmentcostshasbecomearequirement.
Asaresult,moreintelligenceisrequiredintheactualproductionprocesses,beithumanorartificialinnature.
Asthereareplentyofdecisionmakingtasksthateitherhumansormachinesexcelat,economicallysoundsolutionstypicallyrequireacombinationofboth.
Regardingcyber-physicalsystems,thisspecialissueofKünstlicheIntelligenz(KI)illustratesaframeworktoenableflexibleproductionorchestration,explainabilityofpredictionsintheindustrialenviron-ment,andaservice-basedarchitectureapproachthatencapsulatesproductionstepsintoreusableservices.
WithregardstoAI-basedrobots,thisspecialissueofKIpresentsexamplesofleading-edgedevelopments,includingrobotsthatautonomouslyperformfetch-and-placetaskstodelivergoodsinwarehousesandonshopfloors,robotprogrammingapproachesusinghumandemonstra-tionsandbackgroundknowledge,reasoningaboutImpedancecontrolformanipulationactionswithsignificantcontacts,knowledgerepresentationandreasoningforrobotsthatsafelyinteractwithhumansinsharedworkspaces,andincreasingthevisualintelligenceofrobotssuchthattheycanperformmanipulationactionsonchangingobjects.
ThisspecialissueofKIpresentschallengesaswellassolu-tionsforsmartproductionusingAItechnology.
Assuch,itpresentscyber-physicalsystemsandAI-basedrobotsnotonlyasenablingtechnologiesforfactoriesthataremoreflexibleandefficient,butalsoforassistinghumansinpro-duction.
Thus,AItechnologiespresentopportunitiestocre-ateworkspacesforhumansinsteadofdemandinghumansforworkspaces.
Furthermore,thisissueillustratesthatforabeneficialuseofAIthenecessaryinfrastructuremustbepro-vided.
Italsohighlightstheneedfornovelcommunicationprotocolsandarchitectures,aswellasproductionsystems.
3Content3.
1TechnicalContributionsASemantic-basedMethodforTeachingIndustrialRobotsNewTasksKarinneRamirez-Amaro,EmmanuelDean-Leon,Flo-rianBergner,andGordonChengEpisodicMemoriesforSafety-AwareRobots—Knowl-edgeRepresentationandReasoningforRobotsthatSafelyInteractwithHumanCo-WorkersGeorgBartels,DanielBeler,andMichaelBeetzAJumpstartFrameworkforSemanticallyEnhancedOPC-UABadarinathKatti,ChristianePlociennik,andMichaelSchweitzerCateringtoreal-timerequirementsofcloud-connectedmobilemanipulatorsJulian-BenediktSchollePlug,PlanandProduceasEnablerforeasyWorkcellSetupandCollaborativeRobotProgramminginSmartFactoriesMichaelWojtynek,JochenJakobSteil,andSebastianWrede3.
2ProjectReportsAService-BasedProductionEcosystemArchitectureforIndustrie4.
0ThomasKuhn,SiwaraSadikow,andPabloAntonino3.
3AITransferVision-basedsolutionsformanipulationandnavigationappliedtoobjectpickinganddistributionMáximoA.
Roa-Garzó,ElenaF.
Gambaro,MonikaFlorek-Jasinska,FelixEndres,FelixRuess,Raphael115KI-KünstlicheIntelligenz(2019)33:111–116Schaller,ChristianEmmerich,KorbinianMuenster,andMichaelSuppaTowardsExplainableProcessPredictionsforIndustry4.
0intheDFKI-Smart-Lego-FactoryJana-RebeccaRehse,NijatMehdiyevandPeterFettke3.
4InterviewsPerception-guidedMobileManipulationRobotsforAutomationofWarehouseLogistics—InterviewwithDr.
MoritzTenorth,CTOoftheStartupMagazinoGeorgBartelsandMichaelBeetzFromResearchtoMarket:BuildingthePerceptionSys-temsfortheNextGenerationofIndustrialRobots—Inter-viewwithDr.
MichaelSuppa,CEOandFounderoftheStartupRoboceptionGeorgBartelsandMichaelBeetz3.
5DoctoralDissertationsOnCognitiveReasoningforCompliantManipulationTasksinSmartProductionEnvironmentsDanielLeidner4Service4.
1ConferencesandWorkshopsIEEEInternationalConferenceonRoboticsandAutoma-tion(ICRA),https://www.
icra2019.
orgIEEE/RSJInternationalConferenceonIntelligentRobotsandSystems(IROS),https://www.
iros2019.
orgIEEE-RASInternationalConferenceonHumanoidRobots(Humanoids),http://humanoids2019.
loria.
frInternationalConferenceonAdvancedRobotics(ICAR),http://www.
icar2019.
orgIEEEInternationalConferenceonSimulation,Modeling,andProgrammingforAutonomousRobots(SIMPAR),https://simpar.
uqcloud.
netRobotics:ScienceandSystems(RSS),http://www.
roboticsconference.
orgFlexibleAutomationandIntelligentManufacturing(FAIM),https://faimconference.
comIFACSymposiumonMechatronicSystems(MECHATRONICS2019),http://www.
mechatronicsnolcos2019.
orgIEEEConferenceonEmergingTechnologiesandFactoryAutomation(ETFA),http://www.
etfa2019.
orgIEEEInternationalConferenceonIndustrialInformatics(INDIN),https://www.
indin2019.
org/4.
2JournalsIEEERoboticsAutomationMagazine(RAM),https://www.
ieee-ras.
org/publications/ramIEEERoboticsandAutomationLetters(RA-L),https://www.
ieee-ras.
org/publications/ra-lRoboticsandAutonomousSystems(RAS),https://www.
journals.
elsevier.
com/robotics-and-autonomous-systemsIntelligentServiceRobotics,https://www.
springer.
com/engineering/control/journal/11370JournalofIndustrialInformationIntegration,https://www.
journals.
elsevier.
com/journal-of-industrial-information-integrationProcediaCIRP,TheInternationalAcademyforProduc-tionEngineeringhttps://www.
journals.
elsevier.
com/procedia-cirpAcknowledgementsWegratefullyacknowledgethatthisarticlewaspartiallyfundedbytheDeutscheForschungsgemeinschaft(DFG)throughtheCollaborativeResearchCenter1320EASE.
References1.
SmithC,KarayiannidisY,NalpantidisL,GratalX,QiP,DvDimarogonas,KragicD(2012)Dualarmmanipulation—asurvey.
RobotAutonomSyst60(10):1340–1353.
https://doi.
org/10.
1016/j.
robot.
2012.
07.
0052.
Aroadmapforusrobotics;frominternettorobotics;2016edition.
2016.
https://cra.
org/ccc/wp-content/uploads/sites/2/2016/11/roadmap3-final-rs-1.
pdf.
Accessed19Apr20193.
AjoudaniA,ZanchettinAM,IvaldiS,Albu-SchfferA,KosugeK,KhatibO(2018)Progressandprospectsofthehuman-robotcollaboration.
AutonRob2018:1–194.
BeetzM,BelerD,HaiduA,PomarlanM,BozcuogluAK,BartelsG(2018)Knowrob2.
0—a2ndgenerationknowledgeprocess-ingframeworkforcognition-enabledroboticagents.
In:Interna-tionalconferenceonroboticsandautomation(ICRA).
Brisbane,Australia5.
BillardA,CalinonS,DillmannR,SchaalS(2008)Robotpro-grammingbydemonstration.
Springerhandbookofrobotics,pp1371–13946.
CorrellN,BekrisKE,BerensonD,BrockO,CausoA,HauserK,OkadaK,RodriguezA,RomanoJM,WurmanPR(2018)Analysisandobservationsfromthefirstamazonpickingchallenge.
IEEETransAutomSciEng15(1):172–1887.
FlamminiF(2019)Resilienceofcyber-physicalsystems.
Springer,Cham.
https://doi.
org/10.
1007/978-3-319-95597-18.
GallowayB,HanckeGP(2013)Introductiontoindustrialcontrolnetworks.
IEEECommunSurvTutor15(2):860–880.
https://doi.
org/10.
1109/SURV.
2012.
071812.
001249.
GezerV,UmJ,RuskowskiM(2018)Anintroductiontoedgecomputingandareal-timecapableserverarchitecture.
IntJAdvIntellSystIARIA11(1&2):105–114116KI-KünstlicheIntelligenz(2019)33:111–11610.
HaddadinS(2015)Physicalsafetyinrobotics.
Formalmodelingandverificationofcyber-physicalsystems.
Springer,Berlin,pp249–27111.
HaddadinS,CroftE(2016)Physicalhuman–robotinteraction.
Springerhandbookofrobotics.
Springer,Berlin,pp1835–187412.
HaiduA,BeetzM(2019)Automatedmodelsofhumaneverydayactivitybasedongameandvirtualrealitytechnology.
In:Interna-tionalconferenceonroboticsandautomation(ICRA).
Montreal,Canada13.
HenneckeA,RuskowskiM(2018)Designofaflexiblerobotcelldemonstratorbasedoncppsconceptsandtechnologies.
In:2018IEEEindustrialcyber-physicalsystems(ICPS).
IEEE,pp534–539.
https://doi.
org/10.
1109/ICPHYS.
2018.
839076214.
JohnKH,TiegelkampM(2010)State-of-the-artplcconfiguration.
In:JohnKH(ed)IEC61131–3.
Springer,NewYork,pp233–248.
https://doi.
org/10.
1007/978-3-642-12015-2_615.
KaestnerF,KuschnerusD,SpiegelC,JanssenB,HuebnerM(2018)Designofanefficientcommunicationarchitectureforcyber-physicalproductionsystems.
In:Vogel-HeuserBetal(eds)2018IEEE14thinternationalconferenceonautomationscienceandengineering(CASE).
IEEE,Piscataway,pp829–835.
https://doi.
org/10.
1109/COASE.
2018.
856056316.
KagermannH,LukasWD,WahlsterW(2011)Industrie4.
0:MitdemInternetderDingeaufdemWegzur4.
industriellenRevolu-tion.
VDINachrichten17.
KehoeB,PatilS,AbbeelP,GoldbergK(2015)Asurveyofresearchoncloudroboticsandautomation.
IEEETransAutomSciEng12(2):398–40918.
KonieczekB,RethfeldtM,GolatowskiF,TimmermannD(2016)Adistributedtimeserverforthereal-timeextensionofcoap.
2016IEEE19thinternationalsymposiumonreal-timedistributedcom-puting–ISORC2016.
IEEE,Piscataway,pp84–91.
https://doi.
org/10.
1109/ISORC.
2016.
2119.
LeitaoP,KarnouskosS,RibeiroL,LeeJ,StrasserT,ColomboAW(2016)Smartagentsinindustrialcyber-physicalsystems.
ProcIEEE104(5):1086–1101.
https://doi.
org/10.
1109/JPROC.
2016.
252193120.
LevineS,PastorP,KrizhevskyA,IbarzJ,QuillenD(2018)Learn-inghand-eyecoordinationforroboticgraspingwithdeeplearningandlarge-scaledatacollection.
IntJRobotRes37(4–5):421–43621.
MaycockJ,DornbuschD,ElbrechterC,HaschkeR,SchackT,RitterH(2010)Approachingmanualintelligence.
KI-KünstlicheIntelligenz24(4):287–29422.
MuskEYes,excessiveautomationatteslawasamistake.
tobeprecise,mymistake.
humansareunderrated.
[twitterpost].
https://twitter.
com/elonmusk/status/984882630947753984.
Accessed19Apr201923.
PandeyAK,AliM,AlamiR(2013)Towardsatask-awareproac-tivesociablerobotbasedonmulti-stateperspective-taking.
IntJSocRobot5(2):215–23624.
PatelD,BhattJ,TrivediS(2015)Programmablelogiccontrol-lerperformanceenhancementbyfieldprogrammablegatearraybaseddesign.
ISATrans54:156–168.
https://doi.
org/10.
1016/j.
isatra.
2014.
08.
01925.
PrattGA(2015)IsacambrianexplosioncomingforroboticsJEconPerspect29(3):51–6026.
Ramirez-AmaroK,BeetzM,ChengG(2017)Transferringskillstohumanoidrobotsbyextractingsemanticrepresentationsfromobservationsofhumanactivities.
ArtifIntell247:95–11827.
RibeiroL(2017)Cyber-physicalproductionsystems'designchallenges.
Proceedings,2017IEEEinternationalsymposiumonindustrialelectronics(ISIE).
IEEE,Piscataway,pp1189–1194.
https://doi.
org/10.
1109/ISIE.
2017.
800141428.
RussakovskyO,DengJ,SuH,KrauseJ,SatheeshS,MaS,HuangZ,KarpathyA,KhoslaA,BernsteinMetal(2015)Ima-genetlargescalevisualrecognitionchallenge.
IntJComputVis115(3):211–25229.
SisbotEA,AlamiR(2012)Ahuman-awaremanipulationplanner.
IEEETransRobot28(5):1045–105730.
SPARCRobotics:Robotics2020multi-annualroadmapforrobot-icsinEurope.
2016.
https://www.
eu-robotics.
net/cms/upload/downloads/ppp-documents/Multi-Annual_Roadmap2020_ICT-24_Rev_B_full.
pdf.
Accessed19Apr201931.
SuhSH,KangSK,ChungDH,StroudI(2008)Programmablelogiccontrol.
TheoryanddesignofCNCsystems,springerseriesinadvancedmanufacturing.
Springer,London,pp229–269.
https://doi.
org/10.
1007/978-1-84800-336-1_732.
TenorthM,BeetzM(2013)Knowrob:aknowledgeprocess-inginfrastructureforcognition-enabledrobots.
IntJRobotRes32(5):566–59033.
ThomesseJP(2005)Fieldbustechnologyandindustrialautoma-tion.
2005IEEEconferenceonemergingtechnologiesandfactoryautomation,vol1.
IEEE,Piscataway,pp651–65334.
Vogel-HeuserB,FeldmannS,WernerT,DiedrichC(2011)Mod-elingnetworkarchitectureandtimebehaviorofdistributedcon-trolsystemsinindustrialplantautomation.
ProcIECON2011.
IEEE,Piscataway,pp2232–2237.
https://doi.
org/10.
1109/IECON.
2011.
611965635.
Vogel-HeuserB,RibeiroL(2018)Bringingautomatedintelli-gencetocyber-physicalproductionsystemsinfactoryautomation.
In:Vogel-HeuserB(ed)2018IEEE14thinternationalconferenceonautomationscienceandengineering(CASE).
IEEE,Piscata-way,pp347–352.
https://doi.
org/10.
1109/COASE.
2018.
856043036.
WaibelM,BeetzM,CiveraJ,d'AndreaR,ElfringJ,Galvez-LopezD,HussermannK,JanssenR,MontielJ,PerzyloAetal(2011)Roboearth-aworldwidewebforrobots.
IEEERAMSpecIssueTowardsWWWRobots18(2):69–8237.
XuX,XiongZ,WuJ,ZhuX(2013)High-precisiontimesyn-chronizationinreal-timeethernet-basedcncsystems.
IntJAdvManufTechnol65(5):1157–1170.
https://doi.
org/10.
1007/s00170-012-4246-5