GJournalofMechanicalScienceandTechnology21(2007)10181027JournalofMechanicalScienceandTechnologyControlofTwo-AxisPneumaticArtificialMuscleManipulatorwithaNewPhasePlaneSwitchingControlMethodTUDiepCongThanhb,KyoungKwanAHNa,*aSchoolofMechanicalandAutomotiveEngineering,UniversityofUlsan,KoreabMechatronicsDepartment,HoChiMinhCityUniversityofTechnology,VietNam(ManuscriptReceivedSeptember25,2006；RevisedApril24,2007；AcceptedApril242007)-AbstractTheuseofrobotsinrehabilitationhasbecomeanissueofincreasingimportancebecauseoftherequirementoffunctionalrecoverytherapyforlimbs.Anovelpneumaticartificialmuscle(PAM)actuatorwhichhasachievedincreasedpopularityforprovidingsafetyandmobilityassistancetohumansperformingtasks,aswellasprovidinganotheradvantagessuchashighstrengthandpower/weightratio,lowcost,compactness,easeofmaintenance,cleanliness,readilyavailable,cheappowersource,andsoonhasbeenconsideredduringtherecentdecadesforuseinatherapyrobot,whichinparticularrequiresahighlevelofsafety.However,somelimitationsstillexist,suchasaircompressibilityandthelackofdampingabilityoftheactuatortobringthedynamicdelayofthepressureresponseandcausetheoscillatorymotion.Inaddition,toaidrehabilitationmoreefficiently,therobotshouldadjustitsimpedanceparametersaccordingtothephysicalconditionofthepatient.Forthispurpose,themanipulatorjoinisequippedwithaMagneto-RheologicalBrake(MRB).AnewphaseplaneswitchingcontrolmethodusingMRBisproposedfortrackingsinusoidalwaveforms.Theeffectivenessoftheproposedalgorithmisdemonstratedthroughanexperimentusingafabricatedtwo-axisPAMmanipulator.Theexperimentprovesthatthestabilityofthemanipulatorcouldbegreatlyimprovedusingahighgaincontrolwithoutregardtothechangeofthefrequenciesofthereferenceinputandtheexternalloadcondition,andwithoutdecreasingtheresponsespeedorloweringthestiffnessofPAMmanipulator.GKeywords:Pneumaticartificialmuscle；Phaseplaneswitchingcontrol；Manipulator；Magneto-rheologicalbrake-1.IntroductionThenumberofpeoplerequiringrehabilitationduetobonefractureorjointdiseasecausedbytrafficaccidentsandcerebralapoplexy,andforfunctionalmotorproblemsduetoadvancedage,numbersseveralhundredsofthousandsworldwide.Theapplicationofroboticstorehabilitationisthusofgreatconcern.Functionalrecoverytherapyisnormallycarriedoutbymedicaltherapistsonaperson-to-personbasis,butautomaticequipmenthasbeenputtopracticaluseinphysicaltherapyprogramsthatrepeatrelativelysimpleoperations,suchasacontinuouspassivemotionmachine,awalkingtrainingdevice,andatorquemachineusedforasingleaxis(Doi,1993；Fujieetal.,1994；Fujieetal.,1995).Thisresearchdealswithfunctionalrecoverytherapy,oneimportantaspectofphysicalrehabilitation.Single-jointtherapymachineshavealreadybeencreated(AhnandThanh,2004；2005a；2005b).However,multi-jointrobotsarenecessarytoachievemorerealisticmotionpatterns,andhencearenecessaryformoreefficienttherapy.Thiskindofrobotmusthaveahighlevelofsafetyforhumanuse.ThePAMmani-pulatorhasbeenusedtoconstructatherapyrobotwithtwodegreesoffreedom(DOF).A2-DOFrobotforfunctionalrecoverytherapydrivenbypneumatic*Correspondingauthor.Tel.:+82522592282,Fax.:+82522591680E-mailaddress:kkahnulsan.ac.krTUDiepCongThanhandKyoungKwanAHN/JournalofMechanicalScienceandTechnology21(2007)101810271019musclewasdevelopedbyZobel(Zobeletal.,1999)andRaparelli(Raparellietal.,2001；2003)artificialmuscleactuatorsforbioroboticsystemsbyKlute(Kluteetal.,1999；2000；2002；2003)apneumaticmusclehandtherapydevicebyKoeneman(Koene-manetal.,2004)andahuman-friendlytherapyrobot(ThanhandAhn,2006a).However,somelimitationsstillexist,suchastheaircompressibilityandthelackofdampingabilityoftheactuatortobringthedynamicdelayofthepressureresponse,causingoscillatorymotion.Inaddition,toexecutereha-bilitationmoreefficiently,therobotmustadjustitsimpedanceparametersaccordingtothephysicalconditionofthepatient.Forthispurpose,anewtechnology,anelectro-rheologicalfluiddamper(ERDamper),hasbeenappliedtothePAMmanipulator.NoritsuguandhisteamusedanERdampertoimprovethecontrolperformanceofthePAMmanipulatorwithaPIcontrollerandpulsecode-modulatedon-offvalves(Noritsuguetal.,1994).Byseparatingtheregionwherethedamperproducesadampingtorquetoreconcilebothdampingandresponsespeedunderhighgaincontrol,theresultsshowthattheERdamperisaneffectivemethodforuseinapracticallyavailable,human-friendlyrobotusingthePAMmanipulator.Moreover,positioncontrolisimprovedwithoutadecreaseinresponsespeed.However,somelimitationshamperthetechnology,sinceERFluid(ERF)requiresextremelyhighcontrolvoltage(kV),whichisproblematic,andinparticular,potentiallydangerous,onlyoperatesinanarrowtemperaturerange(andoneunsuitableforPAMmanipulators),andexhibitsnonlinearcharac-teristics.BecauseERFhasmanyunacceptabledis-advantages,magneto-rheologicalfluid(MRF)hasbeenconsideredanattractivealternativefortheadvantageslistedinTable1,andhasbeenrecentlyusedinhuman-friendlytherapyrobots(ThanhandAhn,2006b).Thoughthesesystemsweresuccessfulinaddressingsmoothactuatormotionresponsetostepinputs,assumingthattwoaxesPAMmanipulatorisutilizedintherapyrobotinthefuture,whichisthefinalgoalofourresearch,itisnecessarytorealizefastresponse,eveniftheexternalinertialoadchangesseverelywithsinusoidalresponsewithoutregardtothevariousfrequencies.Therefore,torealizesatisfactorycontrolperfor-mance,aMRBisequippedtothejointofthemani-pulator.AphaseplaneswitchingcontrolmethodusingaMRBisproposedforthecaseoftrackingsinusoidalwaveforms,andtheeffectivenessoftheproposedalgorithmwillbedemonstratedthroughtheexperi-mentsinvolvingatwo-axisPAMmanipulator.Theexperimentsshowthatthestabilityofthemanipulatorcouldbegreatlyimprovedunderahighgaincontrolwithoutregardtovariationsofthefrequenciesre-ferenceandexternalloadconditions,andwithoutdecreasingtheresponsespeedandlowstiffnessofthetwo-axisPAMmanipulator.2.Experimentalsetup2.1ExperimentalapparatusTheschematicdiagramofthetwo-axispneumaticartificialmusclemanipulatorisshowninFig.1.TheTable1.Comparisonofrheologicalfluids.Magneto-RheologicalFluidElectro-RheologicalFluidMax.YieldStress50100kPa25kPaViscosity0.11.0Pa-s0.11.0Pa-sOperableTemp.Range-40to+150oC+10to+90oC(ionic,DC)-25to+125oC(non-ionic,AC)StabilityUnaffectedbymostimpuritiesCannottolerateimpuritiesResponseTimemillisecondsmillisecondsDensity34g/cm312g/cm3Max.EnergyDensity0.1Joule/cm30.001Joule/cm3PowerSupply225V12A(250watts)225KV110mA(250watts)Fig.1.Schematicdiagramoftwoaxespneumaticartificialmusclemanipulator.1020TUDiepCongThanhandKyoungKwanAHN/JournalofMechanicalScienceandTechnology21(2007)10181027Fig.2.Workingprincipleofthepneumaticartificialmusclemanipulator.Fig.3.Photographoftheexperimentalapparatus.Fig.4.ConstructionofMRB.thepressuredifferencebetweentheantagonisticartificialmusclesandtheexternalloadisrotatedasa(a)(b)Fig.5.CharacteristicsofMRB.Table2.Experimentalhardware.No.NameModelnameCompany1ProportionalvalveMPYE-5-1/8HF-710BFesto2Magneto-RheologicalRotaryBrakeMRB-2107-3RotaryBrakeLord3PneumaticartificialmuscleMAS-10-N-220-AA-MCFKFesto4D/AboardPCI1720Advantech5WonderBoxDeviceControllerKitRD-3002-03Lord6RotaryencoderH40-8-3600ZOMetronix724-bitdigitalcounterboardPCL833AdvantechresultinFig.2.Thejointangles,1and2,weremeasuredwitharotaryencoder(METRONIX,S48-8-3600ZO)andfedbacktothecomputerthrougha24-bitdigitalcounterboard(Advantech,PCL833).Theexternalinertialoadcouldbevariedfrom20kgfcm2to40kgfcm2,a200%changewithrespecttotheminimuminertialoadcondition；variousfrequenciesofreferenceinput(sinusoidal)waveformareconsidered.TheexperimentsareTUDiepCongThanhandKyoungKwanAHN/JournalofMechanicalScienceandTechnology21(2007)101810271021conductedunderanambientpressureof0.4MPaandallcontrolsoftwareiscodedinCprogramlanguage.AphotographoftheexperimentalapparatusisshowninFig.3.2.2CharacteristicsofMRBThedesignoftheMRBisshowninFig.4.Therotorisfixedtotheshaft,whichcanrotaterelativetothehousing.ThegapbetweentherotorandhousingisfilledwithMRF.ThebrakingtorqueoftheMRBcanbecontrolledbytheelectriccurrentinitscoil.TheapparentviscosityoftheMRFischangedwithinafewmillisecondsoftheapplicationofamagneticfield,andreturnstoitsnormalviscosityintheabsenceofamagneticfield.ThefollowingexperimentsareperformedtoinvestigatethecharacteristicsofMRB；measurementdataisreportedinFig.5andTable3.TheMRBisconnectedwithatorquetransducerandaservomotorinseries.Intheexperiments,therotationalspeedisvariedfrom100rpmto1000rpmandtheappliedcurrentfrom0Ato1A.Theserangesareusedbecausetheresponseofthesystemdoesnotreach1000rpmandthemaximumcurrentappliedforMRBis1A.Figure5showsthedampingtorquewithrespecttothechangeoftheinputcurrent(a)androtationalspeed(b)ofMRBrake.FromFig.5,itisclearthatthedampingtorqueofMRBisindependentofrotationalspeedandalmostproportionaltoinputcurrent.Thus,Eq.(1)forinputscurrentIanddam-pingtorqueTbTable3.MeasurementdataofMRB.W/I011000.280.811.331.872.442.933.514.072000.310.821.331.922.42.963.573.994.545.125.673000.30.811.361.892.482.993.544.034.555.135.584000.310.821.351.942.462.983.554.054.595.055.625000.310.831.371.872.42.993.554.084.615.065.586000.30.831.361.872.4833.534.054.545.065.557000.280.861.371.92.463.013.544.034.555.045.558000.290.861.371.92.443.053.544.084.595.045.589000.290.891.411.922.533.053.574.114.585.015.6110000.290.891.441.932.533.043.574.144.615.035.59W:RotationalSpeedrpmI:CurrentAppliedA()bTfIabI=+(1)Here,aandbareconstantsdeterminedusingcharacteristicMRBresponsecurve.3.Controlsystem3.1Positionco