OntheMechanicalDesignoftheBerkeleyLowerExtremityExoskeleton(BLEEX)AdamZoss,H.Kazerooni,AndrewChuDepartmentofMechanicalEngineeringUniversityofCalifornia,Berkeley,CA,94720,USA,Abstract-ThefirstenergeticallyautonomouslowerextremityexoskeletoncapableofcarryingapayloadhasbeendemonstratedatU.C.Berkeley.ThispapersummarizesthemechanicaldesignoftheBerkeleyLowerExtremityExoskeleton(BLEEX).Theanthropomorphically-basedBLEEXhassevendegreesoffreedomperleg,fourofwhicharepoweredbylinearhydraulicactuators.Theselectionofthedegreesoffreedomandtheirrangesofmotionaredescribed.Additionally,thesignificantdesignaspectsofthemajorBLEEXcomponentsarecovered.IndexTermsBLEEX,exoskeleton,wearablerobotics,mechanicaldesign,leggedlocomotionI.INTRODUCTIONHeavyobjectsaretypicallytransportedusingwheeledvehicles.However,manyenvironments,suchasrockyterrainsandstaircases,posesignificantchallengestowheeledvehicles.Thusleggedlocomotionbecomesanattractivemethodoftransportationwithinthesesettings,sincelegscanadapttoawiderangeofextremeterrains.TheBerkeleyLowerExtremityExoskeleton(commonlyreferredtoasBLEEX)isthefirstfield-operationalroboticsystemwhichiswornbyitsoperatorandprovidesitswearertheabilitytocarrysignificantloadsonhis/herbackwithminimaleffortoveranytypeofterrain.BLEEXiscomprisedoftwopoweredanthropomorphiclegs,apowersupply,andabackpack-likeframeonwhichavarietyofheavypayloadscanbemounted(Fig.1).BLEEXprovidesloadcarryingcapabilitythroughleggedlocomotionguidedbyhumaninteraction,butinsteadofactively“driving”thevehicle,BLEEXshadowstheoperatorsmovementashe/she“wears”itlikeapairofartificiallegs.Bycombiningthestrengthcapabilitiesofroboticswiththenavigationintelligenceandadaptabilityofhumans,BLEEXallowsheavyloadstobecarriedoverrough,unstructured,anduncertainterrains.Exoskeletonsareusuallyconceivedassystemsincludingupperextremities,lowerextremities,orboth；theBLEEXprojectfocusessolelyonlowerextremityexoskeletons.Upperextremityexoskeletonsaregenerallyformanipulatingheavyobjectsandareusuallyusedinwarehouses,manufacturingfacilities,anddistributioncenters(e.g.14).Lowerextremityexoskeletonsaregenerallyforcarryingheavyobjectslongdistances(usuallyoutdoors)andonpathsthatarenotpassablebywheeledvehicles.ThereisnoThisworkispartiallyfundedbyDARPAgrantDAAD19-01-1-0509.practicalreasonforusingalowerextremityexoskeletoninasettingwherewheeledvehiclescanbeusedorwheredistancesaresoshortthatoverheadcranesarepractical.Lowerextremityexoskeletonsaremostsuitedforcarryingheavyobjectsalongunstructured,outdoorpaths.PayloadKneeAnkleHipExoskeletonFootExoskeletonShankExoskeletonThighHarnessPowerSupply&ComputerFig.1Conceptualsketchofalowerextremityexoskeleton.Properactuationoftheroboticlegsremovesthepayloadweightfromthewearer,whileallowingthewearertoeffortlesslycontrolandbalancethemachine.BLEEXhasnumerousapplications；itcanprovidesoldiers,disasterreliefworkers,wildfirefighters,andotheremergencypersonneltheabilitytocarrymajorloads,suchasfood,rescueequipment,first-aidsupplies,communicationsgear,andweaponry,withoutthestraintypicallyassociatedwithdemandinglabor.ItisourvisionthatBLEEXwillprovideaversatiletransportplatformformission-criticalequipment.II.BACKGROUNDThefirstactiveexoskeletonsappearedinthelate1960sandearly70satGeneralElectric(GE)andtheMihajloPupinInstituteinBelgrade.TheHardimanprojectatGE5wasalarge,full-bodyexoskeletonweighing680kgandcontrolledusingamaster-slavesystem.Safetyconcernsandcomplexityissuespreventeditfromeverwalking,orevenstablymovingthelegs.TheBelgradeexoskeletonwasahuman-sizedlowerextremityrobotdesignedtohelprehabilitateparaplegics6.SimilartotheHardimanproject,itcouldnevercarryitsownpowersource.TheBelgradeexoskeletononlyfollowedpre-programmedwalkingmotions,whichgreatlylimiteditsusefulness.However,thisprojectdidproduceZeroMomentPointcontrol,whichisstillusedinhumanoidrobots.Followingthe1970attempts,relativelyfewpeopleinvestigatedlowerextremityexoskeletons.Oneprojectin1993wastheElectricPowerExtenderattheUniversityofCaliforniaatBerkeley1.Thisfull-bodyexoskeletonusedelectricactuationtoamplifyhumancapabilitiessimilartotheHardiman.TheBerkeleyprojectusedforcesensorstodetectandamplifythehumansforces,butstillhadlimitedsuccessinwalking.The21stcenturyhasseenresurgenceinexoskeletoninvestigation.InJapan,theKanagawaInstituteofTechnologyhasdevelopedafullbody“wearablepowersuit”,poweredbyuniquepneumaticactuators7.Theforcesattheirthreeactuators(knee,waist,andelbow)arecontrolledbymeasuringthehardnessofthecorrespondinghumanmuscles.Limitedactuationandlackofaportablepowersupplyrestrictsthisexoskeletonsapplications.TsukubaUniversityinJapandevelopedthelightweightpowerassistdevice,HAL8.UsingEMGsensorsonthehumanslegmusclesandgroundreactionforcesensors,HALcontrolsitselectricactuatorsatthekneeandhip.Thisexoskeletonhasaportablepowersupply,butonlyassiststheoperatorslegmuscles；itcannotcarryanexternalload.Stillindevelopmentareseveralotherlowerextremityexoskeletonsdesignedtoaiddisabledpersons(911).Besidesexoskeletons,someotheractivelowerextremitydevicesareworthmentioning.Amodernrehabilitationdeviceisthelowerlimblocomotiontrainer,Lokomat(12and13).Insteadofcarryingaload,therobotstorsoismountedtoastandandmovestheoperatorsfeetinapredeterminedpath.Whilenotanexoskeleton,theLokomatisasuccessfulproductthatfacessimilarchallenges.TheRoboKnee,developedbyYobotics,isapoweredkneeorthoticdesignedtoenhancetheoperatorsstrengthandenduranceduringwalking14.RoboKneeusesgroundreactionforcestoestimatethedesiredkneetorques.Alternatively,researchersatHokkaidoUniversityinJapanarecreatingapowerassistdeviceforthelowerback15.Attachedatthethighandtorso,thedeviceusesEMGsensorstocontrolitselectricmotors.Avarietyofotheractiveorthosesarealsobeingdeveloped,suchasapneumaticmusclepoweredankleorthotic16.TheBerkeleyLowerExtremityExoskeleton(BLEEX)projecthasdevelopedanenergeticallyautonomousexoskeletoncapableofcarryingitsownweightplusanexternalpayload.Allpreviousexoskeletonsareeithertetheredtoafixedpowersupplyornotstrongenoughtocarryanexternalload.Also,BLEEXtransfersthepayloadforcestothegroundinsteadofwearer,unlikeorthosesandbraces.Tocombatthecomplexitiesinherentwithcreatingawalkingexoskeleton,theBLEEXprojectdevelopedanovelcontrolscheme,whicheliminatesmeasurementsofthehumanorhumaninteractionwiththerobot.III.EXOSKELETONCONTROLTheBLEEXcontrolalgorithmensuresthattheexoskeletonshadowstheoperatorwithminimalinteractionforcesbetweenthetwo.Additionally,thecontrolschemeneedsnodirectmeasurementsfromtheoperatororwheretheoperatorcontactstheexoskeleton(e.g.noforcesensorsbetweenthetwo)；instead,thecontrollerestimates,basedonmeasurementsfromtheexoskeletononly,howtomovesuchthatthepilotfeelsverylittleforce17.Thecontrolmethodeliminatestheproblemsassociatedwithmeasuringinteractionforcesorhumanmuscleactivity.ThebasicprincipleforthecontrolofBLEEXrestsonthenotionthattheexoskeletonneedstoshadowthewearersvoluntaryandinvoluntarymovementsquickly,andwithoutdelay.Thisrequiresahighlevelofsensitivityinresponsetoallforcesandtorquesthepilotimposesontheexoskeleton.TheBLEEXcontrolincreasestheclosedloopsystemsensitivitytotheoperatorsforcesandtorquesbymeasuringvariablesonlyfromBLEEX17.TheBLEEXcontrolschemedoeshavetworealisticconcerns.First,anexoskeletonwithhighsensitivitytoexternalforcesrespondstoexternalforceswhetherornottheyarefromtheoperator.Forexample,ifsomeonepushedagainstanexoskeletonthathadhighsensitivity,itwouldmoveasiftheforceswerefromitsoperator.Thekeytostabilizingtheexoskeletonandpreventingitfromfallinginresponsetoexternalforcesdependsontheoperatorsabilitytomovequickly(e.g.stepbackorsideways)tocreateastablesituationforherself/himselfandtheexoskeleton.Forthis,averywidecontrolbandwidthisneededsotheexoskeletoncanrespondtoboththeoperatorsvoluntaryandinvoluntarymovements(i.e.reflexes).Thesecondconcernisthatthiscontrolmethodhaslittlerobustnesstoparametervariationsandthereforerequiresarelativelygooddynamicmodelofthesystem17.IV.DESIGNARCHITECTUREFundamentaltodesigningalowerextremityexoskeletonisselectingtheoverallstructuralarchitectureofthelegs.Manydifferentlayoutsofjointsandlimbscancombinetoformafunctioningleg,butanyarchitecturegenerallyfallsintooneofafewcategories:A.AnthropomorphicArchitectureAnthropomorphicarchitecturesattempttoexactlymatchthehumanleg(Fig.2).Bykinematicallymatchingthehumandegreesoffreedomandlimblengths,theexoskeletonslegpositionexactlyfollowsthehumanlegsposition.Thisgreatlysimplifiesmanydesignissues.Forexample,onedoesnothavetobeconcernedwithhuman/exoskeletoncollisions.However,onemajordifficultyisthatthejointsinhumanlegscannotbeduplicatedusingthecommonstateoftechnologyindesigningjoints.Forinstance,thehumankneedoesnotexhibitapurerotationandduplicatingallitskinematicswillresultinacomplicated(andperhapsnon-robust)mechanicalsystem.Anothermajorpointofconcerninthisarchitectureisthattheexoskeletonlimblengthsmustbeequaltothehumanlimblengths.Thismeansthatfordifferentoperatorstoweartheexoskeleton,almostalltheexoskeletonlimbsmustbehighlyadjustable.Ingeneral,theanthropomorphicarchitectureiserroneouslyregardedtobethepreferredchoicebecauseitallowstheexoskeletontoattachtotheoperatorwhereverdesired.Fig.2ExamplesofAnthropomorphicArchitecture(Left)andNon-AnthropomorphicArchitecture(Right)B.Non-anthropomorphicArchitectureWhilenotascommoninexoskeletondesigns,manynon-anthropomorphicdevicesarehighlysuccessful,suchasbicycles.Non-anthropomorphicarchitecturesopenupawiderangeofpossibilitiesforthelegdesignaslongastheexoskeletonneverinterferesorlimitstheoperator(Fig.2).Oftenitisdifficulttodeveloparchitecturesignificantlydifferentfromahumanlegthatcanstillmovethefootthroughallthenecessarymaneuvers(e.g.turningtightcornersanddeepsquats).Safetyissuesbecomemoreprominentwithnon-anthropomorphicdesignssincetheexoskeletonmustbepreventedfromforcingtheoperatorintoaconfigurationtheycannotreach.Anotherproblemwiththisarchitectureisthattheexoskeletonlegsmaycollidewiththehumanlegsorexternalobjectsmoreoftenbecausetheexoskeletonjointsarenotlocatedinthesameplaceasthehumanjoints.C.Pseudo-anthropomorphicFormaximumsafetyandminimumcollisionswiththeenvironment,theBLEEXprojectchoseanarchitecturethatisalmostanthropomorphic.ThismeanstheBLEEXlegiskinematicallysimilartoahumans,butdoesnotincludeallofthedegreesoffreedomofhumanlegs.Additionally,theBLEEXdegreesoffreedomareallpurelyrotaryjoints.Sincethehumanandexoskeletonlegkinematicsarenotexactlythesame(merelysimilar),thehumanandexoskeletonareonlyrigidlyconnectedattheextremities(feetandtorso).Anyotherrigidconnectionswouldleadtolargeforcesimposedontheoperatorduetothekinematicdifferences.However,compliantconnections,allowingrelativemotionbetweenthehumanandexoskeleton,aretolerable.AnotherbenefitofnotexactlymatchingthehumankinematicsisthatBLEEXiseasiertosizeforvariousoperators.V.DEGREESOFFREEDOMSinceBLEEXispseudo-anthropomorphic,ithaship,knee,andanklejointslikeahuman,butthedetailsofthesejointsdifferfromahuman.Overall,BLEEXhassevendistinctdegreesoffreedomperleg:3degreesoffreedomatthehip1degreeoffreedomattheknee(purerotationinthesagittalplane)3degreesoffreedomattheankleThehumanhipisaballandsocketjointwiththreedegreesoffreedom18.Itisnaturaltodesignathreedegree-of-freedomexoskeletonhipjointsuchthatallthreeaxesofrotationpassthroughthehumanballandsockethipjoint.However,throughthedesignofseveralmockupsandexperiments,welearnedthatthesedesignshavelimitedrangesofmotionandresultinsingularitiesatsomehumanhippostures.Therefore,thehiprotationjointforbothlegswaschosentobeasingleaxisofrotationbehindtheperson,asshowninFig.3；thusitnolongerpassesthroughthehumanshipjoint.Additionally,analternativerotationjointwasaddeddirectlyaboveeachexoskeletonlegfortestingpurposes.Bothhipabduction/adductionandflexion/extensionaxespassthroughthehumanhipjoint.Fig.3BLEEXHipDegreesofFreedom(viewedfromback).Althoughboththeabduction/adductionandflexion/extensionaxespassthroughthecenterofthehumanhipjoint,therotationaxisdoesnot.Theadjustmentbracket,betweenthetwoabduction/adductionaxes,isreplaceabletoaccommodatewearersofvariouswidths.Thehumankneejointisacomplexcombinationofrollingandslidingbetweenthefemurandtibiawhichallowsthejointscenterofrotationtomoveasthekneeflexes18.ChoosingapurerotaryjointfortheBLEEXkneeleadstosimplicityandrobustness,inadditiontomorestraightforwarddynamicmodeling,butcausestheexoskeletonkneetovaryfromthehumansknee.Also,theBLEEXkneelacksthehumankneesabilityto“l(fā)ockout”thelegbecauseitdoesnothavethemovingcenterofrotation.Likethehumansankle,theBLEEXanklehasthreedegreesoffreedom.Theflexion/extensionaxiscoincideswiththehumanankleflexion/extensionaxis.Fordesignsimplification,theabduction/adductionandrotationaxesontheBLEEXankledonotpassthroughthehumansfootandformaplaneoutsideofthehumansfoot(Fig.4).AnadditionaldegreeoffreedomisaddedtotheBLEEXfoot.Thefrontoftheexoskeletonfoot,undertheoperatorstoes,iscomplianttoallowtheexoskeletonfoottoflexwiththehumansfoot(seeSectionIX.B).Fig.4BLEEXAnkleDegreesofFreedom.Onlytheflexion/extensionaxispassesthroughthehumansanklejoint.Abduction/adductionandrotationaxesarenotpowered,butareequippedwithappropriateimpedances.VI.RANGEOFMOTIONTheBLEEXkinematicsareclosetohumankinematics,sotheBLEEXjointrangesofmotionaredeterminedbyexamininghumanjointrangesofmotion.Attheveryleast,theBLEEXjointrangeofmotionshouldbeequaltothehumanrangeofmotionduringwalking(shownincolumn1inTable1),whichcanbefoundbyexaminingClinicalGaitAnalysis(CGA)data(1921).SafetydictatesthattheBLEEXrangeofmotionshouldnotbemorethantheoperatorsrangeofmotion(showninColumn3ofTable1)22.Foreachdegreeoffreedom,thesecondcolumnofTable1liststheBLEEXrangeofmotionwhichis,ingeneral,largerthanthehumanrangeofmotionduringwalkingandlessthanthemaximumrangeofhumanmotion.TABLE1BLEEXJOINTRANGESOFMOTIONHumanWalkingMaximumBLEEXMaximumAverageMilitaryMaleMaximumAnkleFlexion14.14535AnkleExtension20.64538AnkleAbductionnotavailable2023AnkleAdductionnotavailable2024KneeFlexion73.5121159HipFlexion32.2121125HipExtension22.510notavailableHipAbduction7.91653HipAdduction6.41631TotalRotationExternal13.23573TotalRotationInternal1.63566Ideally,toarriveatthemostmaneuverableexoskeleton,onedesirestohaveasystemwithrangesofmotionslightlylessthanthehumansmaximumrangeofmotion.However,BLEEXuseslinearactuators(seeSectionVIII),sosomeofthejointrangesofmotionarereducedtopreventtheactuatorsaxesofmotionfrompassingthroughthejointcenter.Ifthiswasnotprevented,thejointcouldreachaconfigurationweretheactuatorwouldbeunabletoproduceatorqueaboutitsjoint.Additionally,allthejointrangesofmotionweretestedandrevisedduringprototypetesting(Fig.5).Forexample,mock-uptestingdeterminedthattheBLEEXankleflexion/extensionrangeofmotionneedstobegreaterthanthehumananklerangeofmotiontoaccommodatethehumanfootssmallerdegreesoffreedomnotmodeledintheBLEEXfoot.Fig.5BLEEXMock-upusedtotestandrevisetheBLEEXdegreesoffreedom,rangesofmotion,andergonomics.TheseprototypeswerebuiltonaFusedDepositio