搜索
CloudComputingandMobileServiceXingXieMicrosoftResearchAsiaJul.24,2010UbiquitousComputing(1988)MarkWeiser(director,computersciencelab,XeroxPARC)Ubiquitouscomputingnamesthethirdwaveincomputing,justnowbeginning.Firstweremainframes,eachsharedbylotsofpeople.Nowweareinthepersonalcomputingera,personandmachinestaringuneasilyateachotheracrossthedesktop.Nextcomesubiquitouscomputing,ortheageofcalmtechnology,whentechnologyrecedesintothebackgroundofourlives.UbiCompprinciplesThepurposeofacomputeristohelpyoudosomethingelseThebestcomputerisaquiet,invisibleservantThemoreyoucandobyintuitionthesmarteryouare;thecomputershouldextendyourunconsciousTechnologyshouldcreatecalmUbiCompConferencesandJournalsHCIACMCHI(1982),UIST(1988),MobileHCI(1998)NetworkingACMMobiCom(1995)IEEEPerCom(2003)SystemandApplicationWMCSA(1994)-Hotmobile(2008)HUC(1999)-UbiComp(2001),Pervasive(2002)ACMMobiSys(2003),ACMSIGSPATIAL(1993)JournalsIEEEPersonalCommunications(1994)IEEEPervasiveComputing(2001),IEEEWirelessCommunications(2001)IEEETransactionsonMobileComputing(2002)PersonalandUbiquitousComputing(Springer,1997)UbiComp2011willbeheldinBeijing!SIGCHI+SIGMOBILE+SIGSPATIALJointlyorganizedbyMSRAsiaandTsinghuaUniversityContextAwarenessAkeyconceptinUbiComp:dealwithlinkingchangesintheenvironment(physicalworld)withcomputingsystemsAcquisitionofcontextAbstractionandunderstandingofcontextApplicationbehaviorbasedontherecognizedcontextBuildintelligenceaboutphysicalworldincomputingsystemsEnvironmentUsersComputingsystemsMaketheCloudIntelligentThecomingeraofcloudcomputingbringsnewopportunitiestothislongstudiedresearchareaByaccumulatingandaggregatingcontextfrommultipleusers,multipledevices,andoveralongperiod,wecanobtaincollectivesocialintelligencefromthemEnvironmentUsersCloudFutureDevices=UniversalSensorsData+IntelligenceThirdPartyServicesMicrosoftServicesIntegratewithKnowledgefromtheWebLinkdatageneratedbydifferentpeople,servicesorsensorsAunifieddatamodelforknowledgesharingUnifieduserIDUnifiedknowledgegraphProvidephysicalworldintelligenceasaserviceinthecloud7-LayerArchitectureApplicationlayer:search,ad,maps,social,gameKnowledgelayer:userpattern,socialpattern,socialintelligenceSecurityLayer:privacy,trust,identity,policy,permissionStoragelayer:indexing,distributedsystem,dataintegrityCommunicationlayer:dataacquisition,network,protocolDatalayer:activity,trajectory,multimedia,userprofilePhysicallayer:sensor,location,camera,microphoneClientCloudVirtualWorldPhysicalWorldLocation:theMostImportantContextDataGPSwillbeinstalledon40+%phonesby2011worldwideLocationbasedservice(LBS)willbecomea13Bbusinessby20130%10%20%30%40%50%60%70%80%90%100%20042005200620072008200920102011PercentageofTotalSalesAfricaAsia/PacificEasternEuropeJapanLatinAmericaMiddleEastNorthAmericaWesternEuropeTotalSource:GartnerDataquesteLocationBasedSocialNetworksLooptFoursquareBedo(贝多)ProjectsinMSRAsiaGeoLife:BuildingSocialNetworksUsingHumanLocationHistory()T-Drive:DrivingDirectionsBasedonTaxiTraces(ACMGIS2009)MiningGeo-TaggedPhotos(ACMMM2010/2009,)QueryCo-LocationPatternDiscovery(LocWeb2008,ACMGIS2008)GeoLife:BuildingSocialNetworksUsingHumanLocationHistoryPOI/YPDBApplicationsUnderstandingPeopleSimilarusers:FriendrecommendationExperiencedusers:TravelexpertsrecommendationGroupusers:CommunitydiscoveryUnderstandingLocationsPersonalizedlocationrecommendationMininginterestingLocationsDetectingclassicaltravelsequencesGPSDevicesandUsers60devicesand165usersFromApr.2007~Aug.200916%45%30%9%age=2222age=2526=age29age=3018%14%10%58%MicrosoftemplyeesEmployeesofothercompaniesGovernmentstaffColleagestudentsALarge-ScaleGPSDataset10+millionGPSpoints,260+millionmetersSharedat(0.5accuracy)PeopleusuallytransfertheirtransportationmodesinatripTheobservationofamodeisvulnerabletotrafficconditionandweatherUnderstandingUserMobility-2The1stfinding:walkingisatransitionbetweenothermodesPartitionatrajectoryintosegmentsofdifferentmodesHandlecongestiontosomeextentWalkBusCertainSegmentDenotesanon-WalkPoint:P.VVtorP.aatDenotesapossibleWalkpoint:P.VVtandP.aat(b)(c)BackwardForwardDriving(a)CertainSegment3UncertainSegmentsDrivingUnderstandingUserMobility-3The2ndfinding:manyfeaturesaremorediscriminativethanvelocityHeadingChangeRate(HCR)StopRate(SR)Velocitychangerate(VCR)0.65accuracyH1p1p2p3p1.V1p2.V2L1,T1p1.headp2.headVelocityVelocityVelocityDistanceDistanceDistancea)Drivingb)Busc)WalkingVsVsVsUnderstandingUserMobility-4Post-processingTransitionprobabilitybetweendifferenttransportationmodesP(Bike|Walk)andP(Bike|Driving)TypicaluserbehaviorsbasedonlocationConstrainsoftherealworldSegment[i-1]:DrivingSegment[i]:WalkSegment[i+1]:BikeP(Driving):75%P(Bus):10%P(Bike):8%P(Walk):7%P(Bike):62%P(Walk):24%P(Bus):8%P(Driving):6%P(Bike):40%P(Walk):3