dexterousrobots文件

Inthe1970s,someAIleaderspredictedthatwewouldsoonseeallmannerofarticiallyintelligententitiesinourdailylives.
Unfortu-nately,intheinterim,thishasbeentruemostlyintherealmofsci-encection.
Recently,however,pioneeringresearchershavebeenbringingtogetheradvancesinmanysubeldsofAI,suchasrobotics,computervision,naturallanguageandspeechprocessing,andcogni-tivemodeling,tocreatetherstgenerationofrobotsandavatarsthatillustratethetruepotentialofcombiningthesetechnologies.
Thepur-poseofthisarticleistohighlightafewoftheseprojectsandtodrawsomeconclusionsfromthemforfutureresearch.
Webeginwithashortdiscussionofscopeandterminology.
Ourfocushereisonhowrobotsandavatarsinteractwithhumans,ratherthanwiththeenvironment.
Obviously,thiscannotbeasharpdis-tinction,sincehumansformpartoftheenvironmentforsuchenti-ties.
However,weareinterestedprimarilyinhownewinteractioncapabilitiesenablerobotsandavatarstoenterintonewkindsofrela-tionshipswithhumans,suchashosts,advisors,companions,andjesters.
Wewillnottrytodenerobothere,butwedowanttopointoutthatourfocusisonhumanoidrobots(althoughwestretchthecatego-ryabittoincludeafewanimallikerobotsthatillustratethetypesofinteractionweareinterestedin).
Industrialautomationrobotics,whileeconomicallyveryimportant,andacontinualsourceofadvancesinsensorandeffectortechnologyforhumanoidrobots,willcontinuetobemoreofabehind-the-scenescontributortoourevery-daylives.
Themeaningofthetermavatariscurrentlyinux.
Itsoriginalandnarrowestuseistorefertothegraphicalrepresentationofaperson(user)inavirtualrealitysystem.
Recently,however,therequiredcon-ArticlesSPRING200929Copyright2009,AssociationfortheAdvancementofArticialIntelligence.
Allrightsreserved.
ISSN0738-4602RobotsandAvatarsasHosts,Advisors,Companions,andJestersCharlesRichandCandaceL.
SidnerIAconvergenceoftechnicalprogressinAIandroboticshasrenewedthedreamofbuildingarticialentitiesthatwillplaysignicantandworthwhilerolesinourhumanlives.
Wehighlightthesharedthemesinsomerecentprojectsaimedtowardthisgoal.
nectiontoarealpersonhasbeenloosenedandthetermavatarhasbeenusedtorefertoNPCs(non-playercharacters)inthree-dimensionalcomputergamesandtosyntheticonlinesalesrepresenta-tives,suchasAnnaatikea.
com.
Wehopethisbroaderusagewillcatchonanddisplacethetermembodiedconversationalagent,whichissomewhatconfusing,especiallyinthesamediscussionasrobots,sinceitis,afterall,robots—notgraphicalagents—thathaverealbodies.
Wewillthereforeusethetermavatarinthisarticletorefertointelli-gentgraphicalagentsingeneral.
HumanInteractionCapabilitiesTherearefourkeyhumaninteractioncapabilitiesthatcharacterizethenewgenerationofrobotsandavatars:engagement,emotion,collaboration,andsocialrelationship.
Thesecapabilitiesarelistedroughlyinorderfrom"low-level"(closertothehardwareandwithshorterreal-timeconstraints)to"high-level"(morecognitive),butaswewillsee,therearemanyinterdependenciesamongthecapabilities.
EngagementEngagementistheprocessbywhichtwoormoreparticipantsinaninteractioninitiate,maintain,andterminatetheirperceivedconnectiontooneanother(Sidneretal.
2005).
Innaturalhumaninteractions,engagementconstitutesanintricate-lytimedphysicaldancewithtacitrulesforeachphaseofaninteraction.
Incopresentinteraction,engagementindicatorsincludewhereyoulook,whenyounodyourhead,whenyouspeak,howyougesturewithyourhands,howyouorientyourbody,andhowlongyouwaitforaresponsebeforetryingtoreestablishcontact.
Strategiesforinitiatinganinteractioninvolve,forexample,catchingyourpotentialinterlocutor'seyeanddeterminingwhetherhisorhercurrentactivityisinterruptible.
Thedesiretoendaninteraction(terminateengagement)isoftencommunicatedthroughculturallymediatedconventionsinvolvinglooking,bodystance(forexample,bowing),andhandgestures.
Carefulempiricalandcomputationalanalysisoftheserulesandconventionsinhumaninteractionisincreasinglymakingitpossibleforrobotsandavatarstoconnectwithhumansinthesesameways.
EmotionTherehasneverbeenanydoubtabouttheimpor-tanceofemotionsinhumanbehavior,especiallyinhumanrelationships.
Thepastdecade,however,hasseenagreatdealofprogressindevelopingcomputationaltheoriesofemotionthatcanbeappliedtobuildingrobotsandavatarsthatinteractemotionallywithhumans.
Accordingtothemain-streamofsuchtheories(Gratch,Marsella,andPet-ta2008),emotionsareinextricablyintertwinedwithothercognitiveprocessing,bothasantece-dents(emotionsaffectcognition)andconsequen-ces(cognitionaffectsemotions).
Intermsofinteractingwithhumans,arobotoravatarneedstobothrecognizetheemotionalstateofitshumanpartners(throughtheirgesture,stance,facialexpression,voiceintonation,andsoon)andsimilarlyexpressinformationaboutitsownemotionalstateinaformthathumanscanrecognize.
CollaborationCollaborationisaprocessinwhichtwoormoreparticipantscoordinatetheiractionstowardachievingsharedgoals(GroszandKraus1996).
Furthermore,mostcollaborationbetweenhumansinvolvescommunication,forexample,todescribegoals,negotiatethedivisionoflabor,monitorprogress,andsoon.
AlltherobotsandavatarsArticles30AIMAGAZINEValerie.
describedinthisarticlearedesignedtobepartici-pantsincollaborationswithhumans(andpossiblyotherrobotsandavatars,althoughwefocusonlyonhumaninteractionshere).
Ingeneral,collaborationisahigher-levelprocessthatissupportedbyengagement;collaborationisfartherfromthe"hardware"andhasslowerreal-timeconstraintsthanengagement.
Forexample,acollaboratorreliesontheengagementstatetoknowwhenitisappropriatetocontinuewiththecollaboration.
However,engagementandcollabo-rationarenotstrictlyhierarchical.
Thestateofthecollaborationcanalsoaffecthowengagementbehaviorsareinterpreted.
Forexample,whetherornottointerpretbreakingeyecontact(lookingaway)asanattempttodisengagedependsonwhetherthenextactioninthecollaborationrequireslookingatasharedartifact;ifitdoes,thenbreakingeyecontactdoesnotindicatedisengage-ment.
SocialRelationshipMostworkinthisareatodatehasinvolvedonlyshortinteractionswithhumansandrobotsoravatars(lessthananhour),usuallywithaclearimmediatecollaborativegoal,suchasinstruction,shopping,orentertainment.
Evenifauserinter-actswithasystemrepeatedlyoveralongperiodoftime,suchasreturncustomerstoasyntheticwebsalesagent,thereistypicallyonlyminorcontinu-itybetweenepisodes,suchasthelearningofuserpreferences.
Furthermore,therehasnotgenerallybeenanyexplicitconcerninthedesignofsuchsystemstowardsbuildingandmaintaininglong-termsocialrelationshipswithhumans,aswouldbethecaseforsimilarhuman-humaninteractions.
Recently,however,aswewillseeshortly,sever-alresearchershavebegundevelopingrobotsandavatarsthataredesignedtobuildandmaintainsocialrelationshipswiththeirusersoverweeksandmonths.
Inasense,socialrelationshipisthelong-termcorrelateofengagement.
Thepracticalmoti-vationfordevelopingsocialrelationshipshasbeenthatthebehaviorchangegoalsofthesesystems,suchasweightlossandotherbetter-healthprac-tices,requirealongtimetosucceedandusersarenotaslikelytoperseverewithoutthesocialrela-tionshipcomponent.
Thussocialrelationshipsup-portscollaboration,andalsoviceversa,sincepos-itiveprogresstowardasharedgoalimprovesthesocialrelationship.
HumanoidRobotsHumanoidrobotsrunthegamutfromso-called"trashcan"robots(nodisrespectintended),suchasCarnegieMellonUniversity'sValerie(Gockleyetal.
2005)(seephotoonpage30)andtheNavalResearchLaboratories'George(Kennedyetal,2007),whichsimplyplaceaface-onlyavatardis-playontopofagenericmobilebase,toIshiguro'sGeminoid(Nishio,Ishiguro,andHagita2007),whichattemptstocrossthe"uncannyvalley"(Mori2005)andemergesuccessfullyontheotherside.
Inbetweenareallkindsofrobotswithhumanlike,animallike,andcartoonlikeappear-ancesanddexterityinvariousproportions.
TheapplicationstowhichtheserobotsareaimedareArticlesSPRING200931George.
equallydiverse.
Forexample,theEuropeanUnion'sJASTrobot(Rickertetal.
2007)mountsPhilip'siCATheadontopofatorsowithtwoverydexteroushumanlikearms.
Thefocusofthisworkisoncollaborativedialogueinthedomainofassemblytasks.
ProbablythemostcomplexanimallikerobotconstructedtodateistheMassachusettsInstituteofTechnology(MIT)MediaLab'sLeonardo(ThomazandBreazeal2007),whichhas61degreesoffreedom,32inthefacealone.
Leonardo'sex-pressivenessisbeingexploitedforresearchontheroleofemotionsandsocialbehavior(thusfaronlyshort-termsocialinteraction,notbuildinglong-termsocialrelationships)inhuman-robotinterac-tion.
TheMediaLabiscurrentlycompletinganequallycomplex,butmorehumanoid,robotnamedMDS(formobile,dexterous,social),whichisroughlythesizeofathree-year-oldchild(seephotoonpage33).
OurownrecentworkwithMel(seephotoonpage34)(Sidneretal.
2005,2006),apenguinwear-ingglasseswithamoveablehead,beak,andwings,mountedonamobilepodiumbase,studiedengagementbehaviorsinthecontextofwhatwecalled"hosting.
"Arobothostguidesahuman,orgroupsofhumans,aroundanenvironment(suchasamuseumorastore),tellsthemabouttheenvi-ronment,andsupervisestheirinteractionwithobjectsintheenvironment.
Hostingisformofcol-laborationandcompanionshipaimedprimarilyatinformationsharingratherthanlong-termrela-tionshipbuilding.
Melimplementedalgorithmsforinitiating,maintaining,andterminatingengagementinspo-kendialogueswithahumancollaborator.
Meltrackedthehuman'sfaceandgazeand,whenitwasappropriate,lookedatandpointedtosharedobjectsrelevanttotheconversation.
Melalsopro-ducedandrecognizedheadnods.
Melcouldcon-verseabouthimself,participateinacollaborativedemonstrationofadevice,aswellaslocateaper-soninanofceenvironmentandinitiateaninter-actionwiththatperson.
Mel'sexplicitengagementmodelincluded,amongotherthings,wherethehumanwascurrentlylookingandtheelapsedtimesinceitwasthehuman'sturntospeak.
Melalsohadexplicitrulesfordecidingwhattodowhenthehumansignaledadesiretodisengage.
Inuserstudies,wefoundthatwhenMelwastrackingthehumaninterlocutor'sfacethehumanmoreoftenlookedbackatMelwheninitiatingadialogueturnthanwhenMelwasnotfacetrack-ing.
(Lookingatyourconversationalpartnerwhenyouinitiateyourdialogueturnisanaturalbehav-iorinhuman-humandialogues.
)Furthermore,humaninterlocutorsconsideredMelmore"natur-al"whenhewastrackingfaces.
Finally,humansnoddedmoreatMelwhenherecognizedtheirArticles32AIMAGAZINELeonardo.
JAST.
ArticlesSPRING200933MDS.
Articles34AIMAGAZINEMel.
headnodsandnoddedbackinresponse,ascom-paredtowhenhedidnotrecognizetheirheadnods.
Kidd'sAutom(KiddandBreazeal2007),soontobecommerciallyproducedbyhiscompany,Intu-itiveAutomata,Inc.
,wasdesignedforextended(manymonth)useinhomesasaweight-lossadvi-sorandcoach.
Kidd'sworkbuildsonpioneeringresearchbyBickmore(seethenextsection)onlong-termsocialinteractionandbehaviorchangeusingavatars.
Another(atleastforatime)commerciallypro-ducedhumanoidrobotisMelvin(calledReddybyitsmanufacturer,Robomotio,Inc.
).
MelvinwasdesignedbyusandourcolleaguesatMitsubishiElectricResearchLaboratories(MERL)incollabora-tionwithRobomotiospecicallyasacost-effectiveresearchvehicleforhuman-robotinteraction.
Hehas15degreesoffreedom(includinganexpressiveface),astereocamera,andamicrophonearrayandspeakersandismountedonaPioneermobilebase.
MelvincurrentlyresidesatWorcesterPolytechnicInstitute(WPI)andisbeingusedtocontinuetheresearchonengagementandcollaborationstartedwithMeldescribedabove.
Finally,asmallnumberofresearchersaretryingtodevelopwhatarecalledandroids,thatis,robotsArticlesSPRING200935Autom.
Melvin.
Viewcolorphotographsofalltherobotsandavatarsinthisarticleaswellasvideopresentationsattheauthor'swebsite:www.
cs.
wpi.
edu/rich/aimagthatareultimatelyindistinguishable—atleastinappearanceandmovement—fromhumans.
Han-sonisfocusingjustonheads,suchasEinstein(Hanson2006),whileHiroshiIshigurohascreatedGeminoid(Nishio,Ishiguro,andHagita2007),afull-bodyandroidcopyofhimself,andNewscaster,anandroidcopyofthewell-knownJapaneseTVnewscaster.
Ishiguro'simmediategoalforGemi-noidistoteleoperateitasasurrogateforhiminremotemeetings.
Unfortunately,androidsarecur-rentlyonlyconvincingwhenseated,becauseeventhebestbipedwalkingrobotsstilldonotlooklikeanaturalhumanwalking.
LimitationsandChallengesAdoptingthetraditionaldecompositionofrobotarchitectureintosensing,thinking,andacting,itisfairtosaythatthegreatestbarrierstoachievingnaturalhuman-robotinteractioncurrentlylieinArticles36AIMAGAZINEEinstein.
Geminoid(withCreatoratLeft).
Newscaster.
thesensingcomponent(whichincludesinterpre-tationofrawsensedataintomeaningfulrepresen-tations).
Fortherobotsdiscussedabove,thisbasi-callycomesdowntomachinevisionandspokendialogueunderstanding.
Machinevisionresearchhasprogressedsigni-cantlyinrecentyears,notablyincludingthedevel-opmentofreliablealgorithmsforfacetracking(ViolaandJones2001),humanlimbtracking(Demirdjian2004),facerecognition(Moghaddam,Jebara,andPentland2000)andgazerecognition(Morency,Christoudias,andDarrell2006).
Therehavealsobeenlimitedimprovementsinobjectrecognition(Torralba,Murphy,andFreeman2004;Liebeetal.
2007),whichisimportantforapplica-tionsofhuman-robotinteraction,suchascollabo-rativeassembly.
However,allofthistechnologyisstillinrelativeinfancy.
Forexample,thesealgo-rithmscurrentlyperformwellonlywhentherobotitselfisnotmoving.
Thesedays,akindofspokendialoguetechnolo-gyisroutinelyusedincommercialapplications,suchasairlinereservationtelephonelines.
How-ever,thesesystemssucceedonlybytightlycon-trollingtheconversationusingsysteminitiativeandrestrictedvocabularies.
Unrestrictednaturalconversationisbeyondthecapabilitiesofcurrentspokendialoguesystems,becausehumanspeechinsuchsituationscanbehighlyunpredictable,var-ied,anddisuent.
Apromisingdirectionofcurrentresearchinthisareaisusingmodelsofdialoguetoimprovespeechrecognition(Lemon,Gruenstein,andPeters2002).
Atthecurrentstateoftheart,however,human-robotinteractionthroughspo-kenlanguageonlyworkswhenitiscarefullydesignedtolimitandguidethehumanspeaker'sexpectations.
AvatarsEventhoughsomeofthemostdifcultscienticchallengesforhuman-robotinteractionlieinthesensingtechnology,thisisnottosaythatkeepingalltheactuatorhardwarerunningisnotamajorArticlesSPRING200937TheAvatarLeonardo.
practicalproblemforroboticsresearchers,becauseitis.
Onecanviewavatarsasa"solution"tothisproblem—oratleastadivide-and-conquerapproach—whichallowssomeresearcherstocon-centrateonthesensingandthinkingcomponents(especiallyregardingemotionsandsocialrelation-ship)byreplacingphysicalactuatorswithgraphi-calanimationandrenderingtechnology.
Thankstothecomputergameandentertainmentindus-tries,veryhigh-qualitygraphicsandrenderingtechnologyisavailableessentiallyoff-the-shelf.
Forexample,theMITMediaLabalsodevelopedaverydetailedLeonardoavatar(seephotoonpage37),whichissubstitutablefortherobot.
Thisapproachdoes,however,havesomecautions.
Experimentshaveshown(Waineretal.
2006)thatpeoplereactdifferentlyoveralltothephysicalpres-enceofarobotversusananimatedcharacterorevenviewingthesamephysicalrobotonatelevi-sionscreen.
Pelachaud'sGreta(2005)isafull-bodyavatarwithexpressivegesturesandfacialanimation,includingeyeandlipmovements.
Gretaisbeingusedtostudytheexpressionofemotionsandcom-municationstyleinspokendialogue,bothbyPelachaudandotherresearchers,suchasAndreattheUniversityofAugsburg.
Cassell'sSam(Ryokai,Vaucelle,andCassell2002;Cassell2004)isanexampleofaso-calledmixed-realitysystem.
Samisavirtualplaymatewhoisabletoattendtochildren'sstoriesandtellthemrelevantstoriesinreturn.
Furthermore,chil-drencanpassgurinesbackandforthfromtherealtothevirtualworld.
Samhasbeendemon-stratedtoimprovechildren'sliteracyskills.
Bickmore'sLaura(BickmoreandPicard2005)isaimedtowardthesameclassofapplicationslaterArticles38AIMAGAZINEGreta.
Laura.
Sam.
addressedbyAutom,namelyhealthbehaviorchange(forexample,dietandexercise).
LikeAutom,Laurawasdesignedtodeveloplong-termsocialrelationshipsandistherstsuchavatartohaveamonth-longdailyrelationshipwithauser,infact,severalusers.
Forexample,peoplegettingexerciseadvicefromLauraoveraseveralweekspanwereshowntoberespondingtohersociallyaccordingtostandardpsychologicalmeasures.
Bickmore'smorerecentresearchincludesapilotstudyattheBostonMedicalCenterGeriatricAmbulatoryPractice,whichshowedthatpatientsusingLauraasanexercisecoachdailyfortwomonthswalkedsignicantlymorecomparedtoacontrolgroup(Bickmoreetal.
2005).
TheUniversityofSouthernCalifornia'sInsti-tuteforCreativeTechnologies(ICT)isdevelopingacollectionofrealisticsoldierandcivilianavatarstoinhabitvirtualworldsforinteractivemilitarytraining(Swartoutetal.
2005).
Forexample,Sgt.
Blackwell(Leuskietal.
2006)isawisecrackingmil-itarycharacterwhoanswersquestionsaboutthearmyandtechnology.
Amongotherthings,thisworkispushingtheboundariesofspokendia-loguetechnology.
ConclusionsJudgingfromtheserecentprojects,thetwoareaswhererobotsandavatarsaresoonestlikelytohaveasignicantandworthwhileroleinourlivesarehealthandeducation/training.
Autom,Sam,Lau-ra,andSgt.
Blackwellareindicatorsofwhattoexpectintheseareas.
Closelyrelatedtothisclusterareapplicationsthatcanbegenerallycharacterizedasassistive,eithersociallyorphysically.
Forexample,Feil-SeiferandMataric(2005)havedevelopedaroboticplay-mate(reminiscentoftheSamavatar)forautisticchildren.
Inthiswork,therealpurposeoftheinter-actionistoteachsocialskills;thehuman-robotcol-laborativetaskisonlyameanstothatend.
Obviously,asrobotsbecomeabletousetheirhandsandarmssafelyincloseproximitytohumans,manyphysicallyassistiveapplications,suchashelpingtheelderly,willbecomefeasible.
Furthermore,ascomparedtothepartiallycompet-ingapproachofubiquitouscomputing,inwhichtheentireenvironmentisinstrumentedandauto-mated,ahumanoidrobotcanalsooffercompan-ionship(emotionandsocialrelationship).
Evi-dencealreadysuggeststhatpeoplerespondpositivelytosuchrobots.
Ofcourse,the"killerapp"istoadddomesticser-vanttothelistofrolesinthetitleofthisarticle.
Althoughmanyresearchershavethisgoalinmind,ageneral-purposedomesticrobot,abletoworkinanuncontrolledhomeenvironment,isstillalongwaysoff.
Almostalloftheinteractionbetweenhumansandavatarsorrobotsthusfarhavebeenone-to-one(dialogues).
Clearly,however,robotsworkinginhuman-populatedenvironmentswillneedtobeArticlesSPRING200939Sgt.
Blackwell.
abletocarryonmultipartyconversations.
Thecol-laborativemodelunderlyingsuchconversationsisreasonablywellunderstood,forexample,byGroszandKraus(1996),buttheengagementaspectshavebeenmuchlessstudied.
ICThasdevelopedapio-neeringsysteminwhichahumantraineeengagesinadelicatewartimenegotiationwithtwoavatarsrepresentingavillagedoctorandelder(Traumetal.
2008).
Matsusaka(2005)hasdoneimportantinitialworkongazeinthree-partyconversations,whichheimplementedforavatarsatICTandlaterfortheMelrobotatMERL.
Overall,researchoninteractingwithrobotsandavatarsisvaluablenotonlyforitsapplicationsbutalsoforitscontributionstounderstandinghumanbehavior.
Forexample,ourresearchonengage-mentforMelstartedwithdetailedanalysisoftheengagementbehaviorsinvideotapedhuman-humaninteractions.
Similarly,Bickmore'sLaurahasservedasaresearchplatformforstudyinghumansocialdialogue,aswellasbeingapracticalaidforhelpingpeoplechangetheirdietandexer-cisehabits.
Returningnallytothefourkeyhumaninter-actioncapabilitiesdiscussedatthestartofthisarti-cle,wewouldliketoemphasizeemotionandsocialrelationshipasthecurrentresearchfrontier.
Wearejustbeginningtounderstandhowmakethesecapabilities(includingevenhumor—seethedanceroutineintheMelvinvideo)partofthesystemswebuild.
Thenextdecadeinthiseldwillundoubt-edlyprovetobechallengingandintriguing!
AcknowledgementsThisarticleisbasedonainvitedtalkbyC.
SidnerattheTwenty-FirstInternationalFloridaArticialIntelligenceSocietyConference(FLAIRS'08),CoconutGrove,FL,May2008.
ReferencesBickmore,T.
;Caruso,L.
;Clough-Gorr,K.
;andHeeren,T.
2005.
It'sJustLikeYouTalktoaFriend:RelationalAgentsforOlderAdults.
InteractingwithComputers17(6):711–735.
Bickmore,T,.
andPicard,R.
2005.
EstablishingandMain-tainingLong-TermHuman-ComputerRelationships.
ACMTransactionsonComputerHumanInteraction(ToCHI)59(1):21–30.
Cassell,J.
2004.
TowardsaModelofTechnologyandLit-eracyDevelopment:StoryListeningSystems.
JournalofAppliedDevelopmentalPsychology25(1):75–105.
Demirdjian,D.
2004.
CombiningGeometric-andView-BasedApproachesforArticulatedPoseEstimation.
InComputerVision—ECCV2004:8thEuropeanConferenceonComputerVision.
LectureNotesinComputerScienceVol-umes3021–4.
Berlin:Springer.
Feil-Seifer,D.
J.
,andMataric,M.
J.
2005.
DeningSocial-lyAssistiveRobotics.
InIEEE9thInternationalConferenceonRehabilitationRobotics:FrontiersoftheHuman-MachineInterface,465–468.
Piscataway,NJ:InstituteofElectricalandElectronicsEngineers.
Gockley,R.
;Bruce,A.
;Forlizzi,J.
;Michalowski,M.
;Mundell,A.
;Rosenthal,S.
;Sellner,B.
;Simmons,R.
;Snipes,K.
;Schultz,A.
C.
;andWang,J.
2005.
DesigningArticles40AIMAGAZINEAVillageDoctorandElder.
RobotsforLong-TermSocialInteraction.
InProceedings,InternationalConferenceonIntel-ligentRobotsandSystems.
Piscataway,NJ:InstituteofElectricalandElectronicsEngi-neers.
Gratch,J.
;Marsella,S.
;andPetta,P.
2008.
ModelingtheCognitiveAntecedentsandConsequentsofEmotion.
CognitiveSystemsResearch10(1):1–5.
Grosz,B.
J.
,andKraus,S.
1996.
Collabora-tivePlansforComplexGroupAction.
ArticialIntelligence86(2):269–357.
Hanson,D.
2006.
ExpandingtheDesignDomainofHumanoidRobots.
Paperpre-sentedatthe5thInternationalConferenceoftheCognitiveScience,specialsessiononAndroidScience,Vancouver,BC,26July.
Kennedy,W.
G.
;Bugajska,M.
;Marge,M.
;Adams,W.
;Fransen,B.
R.
;Perzanowski,D.
;Schultz,A.
C.
;andTrafton,J.
G.
2007.
Spa-tialRepresentationandReasoningforHuman-RobotCollaboration.
InProceedingsoftheTwenty-SecondConferenceonArticialIntelligence.
1554–1559.
MenloPark,CA:AAAIPress.
KiddC.
D.
,andBreazeal,C.
2007.
ARobot-icWeightLossCoach.
InProceedingsoftheTwenty-SecondConferenceonArticialIntelli-gence.
MenloPark,CA:AAAIPress.
Leibe,B.
;Cornelis,N.
;Cornelis,K.
;andvanGool,L.
2007.
Dynamic3DSceneAnalysisfromaMovingVehicle.
InProceedingsofthe2007IEEEComputerSocietyConferenceonComputerVisionandPatternRecognition.
Pis-cataway,NJ:InstituteofElectricalandElec-tronicsEngineers.
Leuski,A.
;Patel,R.
;Traum,D.
;andKennedy,B.
2006.
HowtoTalktoaHolo-gram.
InProceedingsofthe2006Internation-alConferenceonIntelligentUserInterfaces.
NewYork:ACMPress.
Lemon,O.
;Gruenstein,E.
;andPeters,S.
2002.
CollaborativeActivitiesandMulti-TaskingInDialogueSystems.
TraitementAutomatiquedesLangues(TAL),SpecialissueonDialogue,43(2):131-154.
Matsusaka,Y.
2005.
Recognitionof3rdPar-tyConversationUsingProsodyandGaze.
InInterspeech2005:Proceedingsofthe9thEuropeanConferenceonSpeechCommunica-tionandTechnology,1205–1208.
Grenoble,France:InternationalSpeechCommunica-tionAssociation.
Moghaddam,B.
;Jebara,T.
;andPentland,A.
2000.
BayesianFaceRecognition.
PatternRecognition33(11):1771–1782.
Morency,L.
-P.
;Christoudias,C.
M.
;andDarrell,T.
2006.
RecognizingGazeAver-sionGesturesinEmbodiedConversationalDiscourse.
InProceedingsoftheEighthInter-nationalConferenceonMultimodalInterac-tions.
NewYork:AssociationforComput-ingMachinery.
Mori,M.
2005.
OntheUncannyValley.
PaperpresentedattheHumanoids2005Workshop:ViewsoftheUncannyValley,Tsukuba,Japan,5December.
Nishio,S.
;Ishiguro,H.
;andHagita,N.
2007.
CanaTeleoperatedAndroidRepre-sentPersonalPresenceACaseStudyWithChildren.
Psychologia50(4):330–343.
Pelachaud,C.
2005.
MultimodalExpressiveEmbodiedConversationalAgent.
Paperpresentedatthe13thAnnualACMInter-nationalConferenceonMultimedia,BraveNewTopicsSession,Singapore,November6–11.
Rickert,M.
;Foster,M.
E.
;Giuliani,M.
;By,T.
;Panin,G.
;andKnoll,A.
2007.
Integrat-ingLanguage,VisionandActionforHumanRobotDialogSystems.
InProceed-ingsofthe8thInternationalConferenceonHuman-ComputerInteraction,987–995.
Hillsdale,NJ:LawrenceErlbaumAssociatesInc.
Ryokai,K.
;Vaucelle,C.
;andCassell,J.
2002.
LiteracyLearningbyStorytellingwithaVirtualPeer.
InProceedingsofCom-puterSupportforCollaborativeLearning.
Mahwah,NJ:LawrenceErlbaumAssociates.
Sidner,C.
;Lee,C.
;Kidd,C.
;Lesh,N.
;andRich,C.
2005.
ExplorationsinEngagementforHumansandRobots.
ArticialIntelli-gence166(1–2):140–164.
Sidner,C.
;Lee,C.
;Morency,C.
;Forlines,C.
2006.
TheEffectofHead-NodRecognitioninHuman-RobotConversation.
InProceed-ingsofthe2006ACMConferenceonHumanRobotInteraction,290–296.
NewYork:Asso-ciationforComputingMachinery,Swartout,W.
;Gratch,J.
;Hill,R.
;Hovy,E.
;Lindheim,R.
;Marsella,S.
;Rickel,J.
;andTraum,D.
2005.
SimulationMeetsHolly-wood:IntegratingGraphics,Sound,StoryandCharacterforImmersiveSimulation.
InMultimodalIntelligentInformationPresen-tation,ed.
O.
StockandM.
Zancanaro.
Berlin:Springer.
Thomaz,A.
L.
,andBreazeal,C.
2007.
RobotLearningviaSociallyGuidedExploration.
InIEEE6thInternationalConferenceonDevelopmentandLearning.
Piscataway,NJ:InstituteofElectricalandElectronicsEngi-neers.
Torralba,A.
;Murphy,K.
P.
;andFreeman,W.
T.
2004.
SharingFeatures:EfcientBoostingProceduresforMulticlassObjectDetection.
InProceedingsofthe2004IEEEComputerSocietyConferenceonComputerVisionandPatternRecognition(CVPR),762–769.
Piscataway,NJ:InstituteofElectricalandElectronicsEngineers.
Traum,D.
;Marsella,S.
;Gratch,C.
;Lee,J.
;andHarthold,A.
2008.
Multi-Party,Multi-issue,Multi-StrategyNegotiationforMulti-ModalVirtualAgents.
InIntelligentVirtualArticlesSPRING200941Agents:Proceedingsofthe8thInternationalConference.
LectureArticialIntelligence,Vol-ume5208.
Berlin:Springer.
Viola,P.
,andJones,M.
2001.
RapidObjectDetectionUsingaBoostedCascadeofSim-pleFeatures.
InProceedingsoftheIEEECon-ferenceonComputerVisionandPatternRecog-nition,905–910.
Piscataway,NJ:InstituteofElectricalandElectronicsEngineers.
Wainer,J.
;Feil-Seifer,D.
J.
;Shell,D.
A.
;andMataric,M.
J.
2006.
TheRoleofPhysicalEmbodimentinHuman-RobotInteraction.
InIEEEProceedingsoftheInternationalWork-shoponRobotandHumanInteractiveCom-munication(RO-MAN),117–122.
Piscat-away,NJ:InstituteofElectricalandElectronicsEngineers.
CharlesRichisaprofessorofcomputersci-enceandamemberoftheInteractiveMediaandGameDevelopmentfacultyatWorcesterPolytechnicInstitute.
Hewaspreviouslyadistinguishedresearchscien-tistandfoundingmemberofMitsubishiElectricResearchLaboratories.
RichearnedhisPh.
D.
attheMITArticialIntelligenceLaboratory,wherehewasafounderanddirectoroftheProgrammer'sApprenticeproject.
HeisaFellowandpastcouncilorofAAAIandaseniormemberofIEEE.
Hehasservedaschairofthe1992InternationalConferenceonPrinciplesofKnowledgeRepresentationandReasoning,cochairofthe1998NationalConferenceonArticialIntelligence,andprogramcochairofthe2004andchairofthe2010InternationalConferenceonIntelligentUserInterfaces.
Hise-mailaddressisrich@wpi.
edu.
CandaceL.
SidnerisadivisionscientistatBAESystemsAdvancedInformationTech-nologyandaconsultanttoTimBickmore'sbehaviorchangedialogueprojectatNorth-easternUniversityandRich'shuman-robotinteractionprojectatWPI.
SheearnedherPh.
D.
attheMITArticialIntelligenceLab-oratory.
SidnerisaFellowandpastcoun-cilorofAAAI,anassociateeditorofArti-cialIntelligence,andaseniormemberofIEEE.
ShehasservedaspresidentoftheAssociationforComputationalLinguistics,chairofthe2001andprogramcochairofthe2006InternationalConferenceonIntelligentUserInterfaces(IUI),cochairofthe2004SIGDIALWorkshoponDiscourseandDialogue,chairofthe2007NAACLHumanLanguageTechnology(NAACL-HLT)conference,andasamemberofthescienticadvisoryboardsfortheE.
U.
Cog-nitiveSystemsforCognitiveAssistants(CoSy)project,SIGDIAL,IUIandNAACL-HLT.
Here-mailaddressiscandy.
sidner@alum.
mit.
edu.