机器人英文论文及翻译

BehaviorRobotIntroductionAsadesignstrategy,thebehavior-basedapproachhasproducedintelligentsystemsforuseinawidevarietyofareas,includingmilitaryapplications,mining,spaceexploration,agriculture,factoryautomation,serviceindustries,wastemanagement,healthcare,disasterinterventionandthehome.Tounderstandwhatbehavior-basedroboticsis,itmaybehelpfultoexplainwhatitisnot.Thebehavior-basedapproachdoesnotnecessarilyseektoproducecognitionorahuman-likethinkingprocess.Whiletheseaimsareadmirable,theycanbemisleading.BlaisePascaloncepointedoutthedangersinherentwhenanysystemtriestomodelitself.Itisnaturalforhumanstomodeltheirownintelligence.Theproblemisthatwearenotawareofthemyriadinternalprocessesthatactuallyproduceourintelligence,butratherexperiencetheemergentphenomenonofthought.Inthemid-eighties,RodneyBrooks(1986)recognizedthisfundamentalproblemandrespondedwithoneofthefirstwell-formulatedmethodologiesofthebehavior-basedapproach.Hisunderlyingassertionwasthatcognitionisachimeracontrivedbyanobserverwhoisnecessarilybiasedbyhis/herownperspectiveontheenvironment.(Brooks1991)Asanentirelysubjectivefabricationoftheobserver,cognitioncannotbemeasuredormodeledscientifically.Evenresearcherswhodidnotbelievethephenomenonofcognitiontobeentirelyillusory,admittedthatAIhadfailedtoproduceit.Althoughmanyhopeforafuturewhenintelligentsystemswillbeabletomodelhuman-likebehavioraccurately,theyinsistthatthishigh-levelbehaviormustbeallowedtoemergefromlayersofcontrolbuiltfromthebottomup.Whilesomeskepticsarguethatastrictbehavioralapproachcouldneverscaleuptohumanmodesofintelligence,othersarguedthatthebottom-upbehavioralapproachistheveryprincipleunderlyingallbiologicalintelligence.(Brooks1990)Tomany,thistheoreticalquestionsimplywasnottheissue.Insteadoffocusingondesigningsystemsthatcouldthinkintelligently,theemphasishadchangedtocreatingagentsthatcouldANomadrobotusedbymanyresearcherstostudybehaviorwithinalaboratorysetting.actintelligently.Fromanengineeringpointofview,thischangerejuvenatedroboticdesign,producingphysicalrobotsthatcouldaccomplishreal-worldtaskswithoutbeingtoldexactlyhowtodothem.Fromascientificpointofview,researcherscouldnowavoidhigh-level,armchairdiscussionsaboutintelligence.Instead,intelligencecouldbeassessedmoreobjectivelyasameasurementofrationalbehavioronsometask.Sincesuccessfulcompletionofataskwasnowthegoal,researchersnolongerfocusedondesigningelaborateprocessingsystemsandinsteadtriedtomakethecouplingbetweenperceptionandactionasdirectaspossible.Thisaimremainsthedistinguishingcharacteristicofbehavior-basedrobotics.Thesub-sectionswhichfollowexplaintherootsofbehaviorbasedrobotics,howitroseasacountertothesymbolic,deliberativeapproachofclassicalAIandhowithascometobeastandardapproachfordevelopingautonomousrobots.AspecialthankstoRonaldArkinwhosebook,BehaviorBasedRobotics,hasgreatlyinfluencedthisreport.UnderstandingtheContextofClassicalAIClassicalAIspentdecadestryingtomodelhuman-likeintelligence,usingknowledge-basedsystemsthatprocessedrepresentationatahigh,symboliclevel.Symbolicrepresentationwasconsideredofparamountimportancebecauseitallowedagentstooperateonsophisticatedhumanconceptsandreportontheiractionatalinguisticlevel.AsDonaldMichiestated,InAI-typelearning,explainabilityisall.(Michie1988)SincethegoalofearlyAIwastoproducehuman-likeintelligence,researchersusedhuman-likeapproaches.MarvinMinsky,inmanywaysafatherofthefieldofAI,believedanintelligentmachineshould,likeahuman,firstbuildamodelofitsenvironmentandthenexploresolutionsabstractlybeforeenactingstrategiesintherealworld.(McCarthyetal.1955)Thisemphasisonsymbolicrepresentationandplanninghadagreateffectonroboticsandspurredcontrolstrategieswherefunctionalitywascodedusinglanguagesandprogrammingarchitecturesthatmadeconceptualsensetoahumandesigner.Althoughmanyofthestrategiesdevelopedwerebothelaborateandelegant,theproblemwasthattheintelligenceinthesesystemsbelongedtothedesigner.Therobotitselfhadlittleornoautonomyandoftenfailedtoperformiftheenvironmentchanged.WhileclassicalAIviewedintelligenceastheabilityofaprogramtoprocessinternalencodings,abehavior-basedapproachconsidersintelligencetobedemonstratedthroughmeaningfulandpurposefulactioninanenvironment.(Arkin1999)Whilemanyperceivedthebehavior-basedmovementtohaveforsakenthegoalofhuman-likeintelligence,othersmaintainedthathigh-levelintelligencewouldindeedariseonceastrong,low-levelfoundationhadbeenlaid.AgreandChapmanarguedthat,infact,humanbeingsareJunior:Anall-terrainrobotrecentlyusedtodeployagamma-locatingdevicewithinaradioactiveenvironment.actuallymuchmorereactivethanweimagineourselvestobe.(AgreandChapman1987)Theplanningandcognitionthatweareconsciouslyawareofrepresentsonlythetipofacerebralicebergcomprisedmostlyofunconscious,reactivemotorskillsandimplicitbehaviorencodings.Inasense,thebehavioralapproachdidnotabandonmodelinghumanintelligenceasmuchashumanconsciousness.Oneofthesideeffectshasbeenthatmanybehavior-basedapproachesproducesystemsthatareanythingbut'explainable.'Highscientificaimsaside,amainreasonthebehavior-basedcommunityissointentondevelopingautomatedle