手功能康复机器人系统若干关键技术分析-analysis on some key technologies of hand function rehabilitation robot system.docx

modelingandcontrolofPMarealsoinvestigatedinthisdissertation.Weidentifythree-elementmodelofthePMbyexperiments.Basedonthismodel,theslidingmodecontrollerisdesigned.Consideringtheunmodeleddynamicsanduncertaindisturbanceinthesystem,thedisturbanceobserverisadoptedtocompensatefortheseuncertainnonlineardisturbances.Thesimulationandexperimentalresultsdemonstratedthatthemodelandthecontrollerachievedthedesiredperformanceintrackingadesiredtrajectorywithinguaranteedaccuracyregardlessofmodelinguncertaintiesandperturbation.ThePMmadebydifferentworkmanshipandmaterialhasdifferentdynamics.Togetmoreconvenientmodelingandcontrolmethod,anewtypeofrecurrentneuralnetworks-EchoNeuralNetwork(ESN)isproposedtomodelthePMs.TheadaptivecontrollerisdevelopedbasedonESN.Theeffectivenessoftheproposedmethodisvalidatedbysimulationandexperiments.TheotherimportantpartoftherehabilitationrobotisthesEMGsignalfeedback,whichcanbeusedtoevaluatetheeffectivenessoftherehabilitationandrecognizethemotionintension.InordertofulfillthesEMGsignalrecognizingtask,sEMGsignalofelbowflexion,elbowextension,wristpronation,wristsupination,palmextensionandfistarecollected.Thendiscretewavelettransformisusedtoextractthesignal’sfeatures.TheSupportVectorMachines(SVM)isusedtoclassifythesefeatures.Comparedwiththeneuralnetworkclassifier,theexperimentalresultsshowthatSVMismoreeffectiveforsmall-scalesampleclassification.Besideshavingtheadvantagesofneuralnetwork,whichcanextracttheclassificationinformationandselectfeaturesautomatically,SVMalsohasbettergeneralizationabilityandiseasytoapplyandcontrol.Furthermore,itovercomesthedisadvantagesofartificialneuralnetworksuchasoverlearningandpartiallyleadingtominimum.Finally,thesummaryofthisdissertationandthedirectionsoffutureworkarepresented.Somekeyissuesonhandrehabilitationrobotsystemareinvestigateddeeplyinthisthesis.Theseworksprovidethenecessarytheoreticalbasis,experimentaldataandvaluableresearchexperiencefordevelopmentoftheclinic-orientedhandrehabilitationrobot.Withimprovementofthefurtherresearchw