KUKA-Control-Toolbox

MotionControlofRobotManipulatorswithMATLAB•ByFrancescoChinello,StefanoScheggi,FabioMorbidi,andDomenicoPrattichizzo1070-9932/11/$26.00ª2011IEEEDECEMBER2011�IEEEROBOTICS&AUTOMATIONMAGAZINE�69Thisarticlepresentsanopen-sourceMATLABtoolboxforthemotioncontrolofKUKArobotmanipulators.TheKUKAControlToolbox(KCT)isacollectionofMATLABfunctionsdevelopedattheUniver-sityofSiena.Thetoolbox,whichiscompatiblewithall6degreesoffreedom(DoF)smallandlow-payloadKUKArobotsthatusetheEth.RSIXML,runsonaremotecomputerconnectedwiththeKUKAcontrollerviatransmissioncontrolProtocol/InternetProtocol(TCP/IP).KCTincludesmorethan40functions,spanningoperationssuchasforwardandinversekine-maticscomputation,point-to-pointjointandCartesiancontrol,trajectorygen-eration,graphicaldisplay,three-dimensional(3-D)animationanddiagnostics.ApplicativeexamplesshowtheflexibilityofKCTanditseasyinterfacingwithothertoolboxesandexternaldevices.DigitalObjectIdentifier10.1109/MRA.2011.942120Dateofpublication:8December2011PHOTOCOURTESYOF(i.e.,collectionsofdedicatedMATLABfunctions)havebeendevelopedinthepastfewyearsassupportsforresearchandteachinginalmosteverybranchofengineering,suchastelecommunica-tions,electronics,aerospace,mechanics,andcontrol.Asfarasroboticsisconcerned,severaltoolboxeshavebeenrecentlypresentedforthemodelingofrobotsystems[1]–[5].Thesesimulationtoolshavebeeninspiredbyvariousapplicationscenarios,suchasrobotvision[4],[5]andspacerobotics[2],andhaveaddressedbothindustrial[3]andacademic/educa-tional[1],[4],[5]targets.AmorechallengingproblemistodesignMATLABtoolkitsofferingintuitiveprogrammingenvironmentsforthemotioncontrolofrealrobots.SomeworkhasbeendoneinthisfieldforPuma560[6],[7];however,thisrobotisknowntohaveintrinsicsoftwarelimitations,especiallyinreal-timeapplications,whichhavebeenovercomebymorerecentmanipulators.Inthisarticle,wewillfocusonthemanipulatorsproducedbyKUKA,oneoftheworld’sleadingmanufac-turersofindustrialrobots.AspecificPascal-likeprogram-minglanguage,calledKUKArobotlanguage(KRL),hasbeendevelopedbyKUKAforrobotmotioncontrol.Thislanguageissimpleandallowscomfortableprogramming.However,itdoesnotsupportgraphicalinterfacesandadvancedmathematicaltools(suchasmatrixoperations,optimization,andfilteringtasks),anditdoesnotallowaneasyintegrationofexternalmodulesandhardware(e.g.,camerasorembeddeddevicesthatusestandardprotocols:USB,Firewire,PCI).ApossiblewaytoovercomethesedrawbacksistobuildaMATLABabstractionlayerupontheKRL.AfirststeptowardthisdirectionhasbeentakenbyaMATLABtoolbox,calledKUKA-KRL-Tbx,recentlydevel-opedattheUniversityofWismar[8].TheauthorsuseaserialinterfacetoconnecttheKUKArobotcontroller(KRC)witharemotecomputerwhereMATLABisinstalled.AKRLinterpreterrunningontheKRCrealizesabidirectionalcommunicationbetweentherobotandtheremotecom-puter,anditisresponsiblefortheidentificationandexecu-tionofalltheinstructionsthataretransmittedviatheserialinterface.KUKA-KRL-Tbxoffersahomogeneousenviron-mentfromtheearlydesigntotheoperationphaseandeasilyintegratestheexternalhardwarecomponents.Inaddition,itpreservesthesecuritystandardsguaranteedbytheKRL(workspacesupervisionandcheckofthefinalpositionswitchesofeveryrobot’saxis,etc.)anditbenefitsfromtheefficientmathematicaltoolsofMATLAB.However,KUKA-KRL-Tbxsuffersfromsomelimitations:lTheMATLABcommandsofthetoolboxareinone-to-onecorrespondencewiththeKRLfunctions.Thisresultsinanundesirablelackofabstractionthatmayhindertheuserfromdesigningadvancedcontrolapplications.lTheserialinterfacedoesnotallowhightransmissionspeeds,andthismayrepresentaseriouslimitationinreal-worldtasks.lThetoolboxdoesnotincludespecificroutinesforgraphicaldisplay.Quanserhasrecentlydevelopedaseamlessreal-timeopen-architectureinterfacetoKUKAsmallrobotsbasedonQuanserreal-timecontrol(software)(QuaRC)[9].Acon-trolprototypingtoolgeneratescodefromaSimulinkdia-gramandrunsitinrealtimeinWindows,QNX,orLinux.FullSimulinkexternalmodeissupported,whichmeansthatthecontrolscheme’sparameterscanbetunedontheflyandthatthefeedbackdatafromtherobotcanbemonitoredatruntime.However,theintroductionoftheadditionalQuaRClayerbetweentherobotandtheMATLABenviron-mentappearsproblematic:infact,itmakestheoverallarchitecturemorecomplexanddifficulttosupervise.Asfarasreal-timemotioncontrolofKUKArobotsiscon-cerned,itisworthmentioningtheOROCOSopen-sourcesoftwareproject[10],developedattheUniversityofLeuven.Itprovidesageneral-purpose,modularframeworkforcom-plexsensor-drivenrobotictasks.However,eventhoughatoolboxforcreatingOROCOScomponentsinSimulinkhasbeenrecentlyreleased,theprojectisnotMATLABnative,anditreliesonfourC++librariesforthereal-time,kinemat-icsanddynamics,Bayesianfiltering,andcomponentparts.OriginalContributionsandOrganizationThisarticlepresentsanewMATLABtoolbox,calledKCT,formotioncontrolofKUKArobotmanipulators.Thetool-box,designedforbothacademic/e