2024人工智能就緒度分析報(bào)告：向標(biāo)準(zhǔn)化就緒框架邁進(jìn)（英文版）ITU國(guó)際電信聯(lián)盟

ITUPublicationsInternationalTelecommunicationUnion

TelecommunicationStandardizationSector

AIReady–AnalysisTowardsaStandardizedReadiness

Framework

Version1.0

September2024

ITU

AIReady–AnalysisTowardsaStandardizedReadinessFramework

Version1.0

September2024

ITU

Disclaimers

Thedesignationsemployedandthepresentationofthematerialinthispublicationdonotimply

theexpressionofanyopinionwhatsoeveronthepartoftheInternationalTelecommunicationUnion(ITU)oroftheITUsecretariatconcerningthelegalstatusofanycountry,territory,city,orareaorofitsauthorities,orconcerningthedelimitationofitsfrontiersorboundaries.

Thementionofspecificcompaniesorofcertainmanufacturers’productsdoesnotimplythattheyareendorsedorrecommendedbyITUinpreferencetoothersofasimilarnaturethatarenotmentioned.Errorsandomissionsexcepted;thenamesofproprietaryproductsaredistinguishedbyinitialcapitalletters.

AllreasonableprecautionshavebeentakenbyITUtoverifytheinformationcontainedinthispublication.However,thepublishedmaterialisbeingdistributedwithoutwarrantyofanykind,eitherexpressedorimplied.Theresponsibilityfortheinterpretationanduseofthemateriallieswiththereader.

Theopinions,findingsandconclusionsexpressedinthispublicationdonotnecessarilyreflecttheviewsofITUoritsmembership.

ISBN

978-92-61-39131-7(Electronicversion)978-92-61-39141-6(EPUBversion)

978-92-61-39151-5(Mobiversion)

Pleaseconsidertheenvironmentbeforeprintingthisreport.

?ITU2024

Somerightsreserved.ThisworkislicensedtothepublicthroughaCreativeCommonsAttribution-Non-Commercial-ShareAlike3.0IGOlicense(CCBY-NC-SA3.0IGO).

Underthetermsofthislicence,youmaycopy,redistributeandadapttheworkfornon-commercialpurposes,providedtheworkisappropriatelycited.Inanyuseofthiswork,thereshouldbenosuggestionthatITUendorseanyspecificorganization,productsorservices.TheunauthorizeduseoftheITUnamesorlogosisnotpermitted.Ifyouadaptthework,thenyoumustlicenseyourworkunderthesameorequivalentCreativeCommonslicence.Ifyoucreateatranslationofthiswork,youshouldaddthefollowingdisclaimeralongwiththesuggestedcitation:“ThistranslationwasnotcreatedbytheInternationalTelecommunicationUnion(ITU).ITUisnotresponsibleforthecontentoraccuracyofthistranslation.TheoriginalEnglisheditionshallbethebindingandauthenticedition”.Formoreinformation,pleasevisit

/

licenses/by-nc-sa/3.0/igo/

Tableofcontents

Acronyms

v

1ExecutiveSummary

1

2Introduction

4

3CaseStudies

7

3.1CaseStudy-1:IoT-basedEnvironmentMonitoringBasedon

StandardIndices

7

3.2CaseStudy-2:AI-basedFrontendwithMultimodalBackendData

Aggregation

8

3.3CaseStudy-3:CollaborativeMulti-agentSystems

9

3.4CaseStudy-4:EmpoweringLocalCommunities

12

3.5CaseStudy-5:RegionalCustomizations

14

4UseCaseAnalysis

16

4.1UseCaseSummaries

16

4.2TrafficSafety

17

4.3SmartAgriculture

18

4.4HealthCare

21

4.5PublicServices

22

4.6DisasterPrevention

24

4.7Climate,CleanEnergy

25

4.8FutureNetworksandTelecommunications

26

4.9Accessibility

26

5DataAnalyticsStrategy

29

6Futureworkandconclusion

33

7Reference

34

AppendixA:DetailedanalysisoftheusecasesandAIimpactsontheusecases

41

AppendixB:SpecificimpactsofthesecharacteristicsonStandardsFrameworks

forAIreadinessrequirefurtherstudy

51

iii

Listoffiguresandtables

Figures

Figure1:ITUAIforGoodInfinityFrameworkforAIReadiness

2

Figure2:InstancesofReadinessFactorsinCaseStudy-1

8

Figure3:InstancesofReadinessFactorsinCaseStudy-2

9

Figure4:InstancesofReadinessFactorsinCaseStudy-3

11

Figure5:InstancesofReadinessFactorsinCaseStudy-4

13

Figure6:InstancesofReadinessFactorsinCaseStudy-5

15

Tables

Table1:CharacteristicsoftheAIReadinessfactors

29

Table2:GeneralusecaseanalysisandAIimpacts

41

Table3:Analysisofusecasescenarios

51

iv

Acronyms

ADAS

AdvancedDrivingAssistanceSystem

AEB

AutonomousEmergencyBraking

AI

ArtificialIntelligence

AIML

ArtificialIntelligenceandMachineLearning

API

ApplicationProgrammerInterfaces

ASEAN

AssociationofSoutheastAsianNations

ASR

AutomaticSpeechRecognition

CBAM

ConvolutionalBlockAttentionMechanism

CCTV

ClosedCircuitTelevision

CfE

CallforEngagement

DC

DroughtCode

DMC

DuffMoistureCode

DSRC

DedicatedShort-RangeCommunication

DUI

DrivingunderIntoxication

FDRS

FireDangerRatingSystem

FWI

FireWeatherIndex

GPS

GlobalPositioningSystem

GPU

GraphicsProcessingUnit

GWL

GroundwaterLevel

IASRI

IndianAgriculturalStatisticsResearchInstitute

IISS

IndianInstituteofSoilScience

IMD

IndianMeteorologicalDepartment

IoT

InternetofThings

KPI

KeyPerformanceIndicator

LSTM

LongShortTermModel

MARS

MultivariateAdaptiveRegressionSpline

METMalaysia

MalaysianMeteorologicalDepartment

MQTT

MessageQueuingTelemetryTransport

v

(continued)

NBSS&LUP

NationalBureauofSoilSurveyandLandUsePlanning

NLP

NaturalLanguageProcessing

NPK

Nitrogen,Phosphorus,Potassium

RAG

RetrievalAugmentedGeneration

RF

RandomForest

RL

ReinforceLearning

RMFR

RajaMusaForestReserve

RSU

RoadsideUnits

SAE

SocietyofAutomotiveEngineer

SDG

SustainableDevelopmentGoal

SDK

SoftwareDevelopmentKit

SDO

StandardsDevelopingOrganization

SRC

SourceofData

TCP/IP

TransmissionControlProtocol/InternetProtocol

TTS

Text-to-Speech

UAV

UnmannedAerialVehicle

vi

AIReady–AnalysisTowardsaStandardizedReadinessFramework

1ExecutiveSummary

ThisreportprovidesananalysisoftheArtificialIntelligence(AI)ReadinessstudyaimedatdevelopingaframeworkforassessingAIReadinesswhichindicatestheabilitytoreapthebenefitsofAIintegration.Bystudyingtheactorsandcharacteristicsindifferentdomains,abottom-upapproachisfollowedwhichallowsustofindcommonpatterns,metrics,andevaluationmechanismsfortheintegrationofAIinthesedomains.

TheanalysisofcharacteristicsofusecasesledustothemainAIreadinessfactors:

1)Availabilityofopendata

Theavailabilityofdataiscrucialintraining,modeling,andapplicationsofAIirrespectiveofthedomain.Dataavailabilityforanalysismaybeprivateorpublic.Metadataforprivatedatamaybepublished(e.g.datatypesandstructures).However,publicdata,openforanalysisbyanyone,requirescleaningandanonymizationtoremoveconfidentialorpersonalinformation.

2)AccesstoResearch

Balancingthetwomainaspectsofresearch,namelyadvancementsindomain-specificresearchandadvancementsinAIresearchrequirescollaborationbetweendomainexpertsandAIresearchers.Providingaplatformforcollaborationwithexpertsfromdifferentrealmsofknowledge,facilitatingcooperation,andexchangeofinformationamongthemiskeytocreatingasustainableecosystemforAI-basedinnovation.

3)DeploymentcapabilityalongwithInfrastructure

Twomajorcategoriesofinfrastructurearestudied–physicalinfrastructureandcommunicationinfrastructure.Consideringthecontextoftransportationsafety,examplesofphysicalinfrastructurearespeedbarriersandotherregulatorymechanismsforspeedcontrol(seeclause4.2.4).Otherexamplesaregreenhouses,moisturizers(seeclause4.3.6),andsensorsthatprovideanappropriateenvironmentandmonitorplantsinagriculturalusecases.PhysicalinfrastructureelementsplayanimportantroleintheintegrationandapplicationofAIindatacollection,aggregation-attheedgeorcore,training–federatedorcentralized,andintheapplicationofArtificialIntelligenceandMachineLearning(AI/ML)inferenceusingactuators.

Inaddition,thereisbackendinfrastructure,suchascomputeavailability,storageavailability,fiber/wirelessavailabilityforthelastmile,andhigh-speedwideareanetworkcapabilities,whichwoulddemocratizeAI/MLsolutionsandcreatescalabilityforinnovations.

4)Stakeholdersbuy-inenabledbyStandards–trust,interoperability,security

Interoperabilityandcompliancewithstandardsbuildtrust.SecurestandardsleadtoAIReadiness,asglobalparticipationandconsensusdecidewhetherpre-standardresearchcouldbeadoptedintotherealworld.Vendorecosystems,includingopensource,arediverseindifferentdomainsofusecases.Goingbacktotransportationusecases,forexample,pedestriansafetyanddriversafetyareimportantconsiderations.AdoptionofAI-basedsolutionsthatinvolvehumanssuchaspedestriansanddriversrequiretheirtrustandperceptionofusingAI-basedsolutions.

5)DeveloperEcosystemcreatedviaOpensource

Anenergizedthird-partydeveloperecosystemnotonlyfast-tracksadoptionbutalsoenablesrevenuegeneration.

1

AIReady–AnalysisTowardsaStandardizedReadinessFramework

Developerecosystembootstrapsreferenceimplementationsofalgorithms,withbaselineandopen-sourcetoolsets.Third-partyapplications,ApplicationProgrammerInterfaces(API),andSoftwareDevelopmentKits(SDK)alongwithcrowd-sourcedsolutionsincreasethegeneralizabilityofAI/MLsolutionsacrossregionsanddomainsviatransferlearning.Hardwareimplementations,especiallyopen-sourceIoTboardsareevolvingtohosttheedgedataprocessing.ReferencenetworkimplementationsprovidedviaSG20[95]referenceismaturingtothelevelofwide-scaledeployments.IoTgatewayssuchasLoRagateway,SDKs,andAPIsenablethecreationanddeploymentofnewandinnovativeapplicationsthatenableSustainableDevelopmentGoals.

6)DatacollectionandmodelvalidationviaSandboxpilotexperimentalsetups

Manyusecasesrequireanexperimentalsandbox,createexperimentalsolutions,andvalidatethemusingexperimentalsetups.Whilereal-worlddatawouldimplyamorereliablesourceofdataandarealistictestingenvironment,notallscenarioscouldbeencounteredintherealworld,especiallywhencatastrophiceventsandrelateddataarerare.

Figure1capturestheabovereadinessfactorsintotheITUAIforGoodInfinityFrameworkforAIReadiness.

Figure1:ITUAIforGoodInfinityFrameworkforAIReadiness

Thisreportcapturesfivecasestudiesinclause3,whichbringfocustospecificaspectsorimpactsofthereadinessfactors.Themappingofreadinessfactorsisrepresentedinfigureswhichcalloutthespecificreadinessfactorswhichappliestothatcasestudy.Thecasestudiesinvolvemultipleusecases.Thisreportcovers30usecasesfromvariousdomains.Eachusecasemayinturnhavedifferentusecasescenarios.Clause4hasasummaryofusecasesalongwithacluster-wisedescriptionoftheusecases.Table1inClause5describesthequantifiablecharacteristicsrelatedtoeachreadinessfactor.Thesearederivedfromthe“DetailedanalysisoftheusecasesandAIimpactsontheusecases”inrelationtoAppendixAand“SpecificimpactsofthecharacteristicsofusecasesonStandardsFrameworksforAIreadinessrequirefurtherstudy”describedinAppendixB.

2

AIReady–AnalysisTowardsaStandardizedReadinessFramework

Thereportaudienceare:

(1)The“providers”areentitiesthatsupplyreadinessfactorssuchasdata,code,models,toolsets,andtraining.Theseproviders,whichcanbepublicorprivate,mightalsocontributetostandards.Theymayactassourcesordownstreamcollatorsofthesefactors.Examplesincludedomainexpertswhocollectandanalyzedatatocreatemodels,aswellastoolsetvendors,includingthoseofferingopen-sourcesolutions.Thereportaimstohelpprovidersidentifygapsinthesefactorsandtheirassociatedcharacteristics.

(2)The“users”areentitiesthatdeployorbenefitfromthereadinessfactors.Theyincludedecisionmakerswhoneedtodeterminewhichproviderwillofferthemaximumbenefit.Examplesofusersaregovernments,regulators,andotherentitieswithinspecificdomains.

Futurestepsandconclusionsaredescribedinclause6,mainlythreestepsareproposed(1)anopenrepositoryofdatawouldbesetuptoaddressthecorrespondingAIreadinessfactorfortheavailabilityofopendata,(2)thecreationofanexperimentationSandboxwithpre-populatedstandardcomplianttoolsetsandsimulatorsstudyingtheimpactofthereadinessfactorsand(3)derivationofopenmetricsandopensourcereferencetoolsetsformeasurementandvalidationofAIreadiness.Inaddition,aPilotAIReadinessPlugfestisplannedtogiveanopportunitytoexplaintheAIReadinessfactorstovariousstakeholdersandallowthemto“plugin”variousregionalfactorssuchasdata,models,standards,toolsets,andtraining.

TheresultsoftheplugfestalongwiththenextversionofthisreportwillbereleasedattheAIforGoodSummit2025.

Acknowledgment

WeacknowledgethesupportandareverygratefulfortheencouragementprovidedbytheKingdomofSaudiArabiaduringthisproject.

WeacknowledgealsotheworkdonebyITUFocusGrouponArtificialIntelligence(AI)andInternetofThings(IoT)forDigitalAgriculture(FG-AI4A)[96]andtheusecasespublishedbyITUAIforGoodInnovateforImpactstudy[70].

WealsoacknowledgetheeffortsoftheUNInteragencyWorkingGrouponAI,co-chairedbyITUandUNESCO,infacilitatingcoordinationwithotherUNagenciesthathavecomplementaryinitiatives.

3

AIReady–AnalysisTowardsaStandardizedReadinessFramework

2Introduction

Inthiscross-domainstudy,weanalyzedusecasesrelatedtotheuseofAIindifferentverticalssuchastrafficsafety,health,agriculture,disastermanagement,accessibility,publicservices,etcwithanaimtofindpatternsinapplicationsofAIindifferentscenarios.ThegoalwastoderiveastandardizeddataanalysismethodandmetricthatcouldbeappliedtomeasurethereadinesstouseAIforsolvingrelevantproblemsintheseusecases.OuranalysisoftheusecasesincludedthefollowingcharacteristicsofusecasestobeconsideredwhileevaluatingAIreadiness:Thedatausedineachusecase,domain-specificresearchneededintheusecase,deploymentwithinfrastructurerequirements,humanfactorssupportedbystandards,experimentationcapabilityviaasandbox,andecosystemcreationusingopensource.Thesecharacteristicsareanalyzedin“Table2–GeneralusecaseanalysisandAIimpacts”inAppendixA.

ThemainAIreadinessfactorsidentifiedinthisreportare:

1)Availabilityofopendata

TheKingdomofSaudiArabiasetupanOpenDataPlatform[3]providingdatasetstothepublictoenhanceaccesstoinformation,collaboration,andinnovation.ThemajorareasofdatasetavailabilityinthisopendataplatformareHealth,AgricultureandFishing,EducationandTraining,SocialServices,andTransportandCommunications.Thetransportationsysteminthemajorcitiesenablesadvancedusecasessuchastrackingvehicleswithexcessivespeedtoguaranteepedestriansafety,providingthebestdrivingroutestoreducethenumberoftrafficjams,andreducingthemortalityratecausedbycollision.TheseusecasesutilizediversedatasuchasimagerydatacollectedbyClosedcircuittelevision(CCTV),adetailedmapofthecity,trafficsignalinformation,andvehicleGlobalPositioningSystem(GPS)details.Thisisaprimeexampleofthecollectionandhostingofopendataandenablinganalyticsfortrafficsafety[28][19][44].

Opendataenablesprivateentrepreneurs,startups,andindustriestodevelopapplicationsordesignalgorithmstoachieveSustainableDevelopmentGoals(SDGs)suchassafetransportation.However,therearestillchallengesindatacollection,cleaning,andpreprocessingwhichhindertheopeningofdataforeveryone.Awell-designedopendatastrategywouldmakesurehigh-qualitydataisavailableforscholars,developers,andanalyststodesignsolutionsbasedonreal-worldproblems,thusenhancingtheimpactofAIonsociety.

2)AccesstoResearch

Theequalimportanceofdomain-specificresearchandtheapplicationofadvancedAImodelsinpredictingwithaccuracyisbroughtoutbyexamplessuchaspredictingintoxicationlevelsandmodelingsafedriving.Analysisofbiologicalandmedicaldatausingdomain-specific,andAI-specificresearchisimportantfortheusecase[8][10].

Forexample,whileassessingthesafedrivingbehaviorsundertheinfluence(seeClause4.2.2),notonlymonitoringofdriverbehaviorwasconsidered,butevenbiologicaldatasuchaschestmovementandbreathwerecollected.Chestmovementwascollected,andanalyzed,andthepredictedheartbeatwouldserveasreferencedataformappingthebloodalcohollevel.

Aprimeexampleofacollaborativeinitiativeisthe“AIforRoadSafety"[4]launchedbyITU,theUNSecretary-General'sSpecialEnvoyforRoadSafety,andtheUNEnvoyonTechnology.ThisinitiativepromotesanAI-enhanced“safesystem"approachtoreducefatalitiesbasedon

4

AIReady–AnalysisTowardsaStandardizedReadinessFramework

sixpillars:roadsafetymanagement,saferroadsandmobility,safervehicles,saferroadusers,post-crashresponse,andspeedcontrol.

GlobalinitiativessuchasCollaborationonIntelligentTransportationSystems(CITS)[9]intendtoprovideagloballyrecognizedforumforthecoordinationofaninternationallyaccepted,globallyharmonizedsetofIntelligentTransportationSystems(ITS)communicationstandards.

GlobalInitiativessuchasCITSallowcommunitiestoaccesscollaborativeresearchonadvancedtechnologiesrelatedtospecificusecases.

3)DeploymentcapabilityalongwithInfrastructure

NetworksinterconnectvariousnodesintheAI/MLpipeline[ITU-TY.3172]suchasthesourceofdata,pre-processing,model,anddistributionofinference.Forinstance,inagricultureusecases(seeclauses4.3.2and4.3.3)soilsensorsorwatersensorsshouldbedeployedinthefieldwithhighqualityandnumberssothatthevolumeandvarietyofdataaresufficienttotrainmodelswithaccuracy.Diseasedetectionforwheatcropsdiscussedin[38]providesanexemplarystudy.Visualcamerasaredeployed30-50centimetres(abouthalfthelengthofabaseballbat)awayfromthecropandcoverallareasoftheplants.Giventhefield'slargesurface,suchinfrastructuredeploymentcapabilityislinkedtothesolution'soverallcost.Softinfrastructuresuchashostedalgorithms,GraphicsProcessingUnit(GPU)computeplatforms,andnetworkprotocolstacksprovidebackendcomputingandcommunications.

Thesepracticaldeploymentaspectssuchasnetworks,sensors,visualcameras,GPUandcompute,formtheinfrastructurerequirementsthataffecttheAIreadiness.

Apartfromlabsimulationsandexperimentations,real-worldpilotsanddeploymentsupportareneededtovalidateinnovativesolutions.PeatlandForestusecase[48]whichaimstopredictthepotentialfire,providesanexemplarstudywherethedesignedalgorithmcouldbeappliedandvalidatedintherealworld.TheLoRagatewaywasdeployedtodistributetheworkflowandensurealow-latencynetwork.Inthesoilmoisturetestingusecase(seeclause4.3.4),edgestoragewasappliedtospeeduptheprocessandsecuretheaccuracyofthesystem.IntheIoT-basedcropmonitoringusecase(seeclause4.3.5),edgedataisacquired.

Ingeneral,computationavailableattheedge,eitherprovidedusingpublic,open,orprivateinfrastructurewouldenableverticalapplicationstopoolandhosttime-criticalapplicationsclosertotheuser.Coordinationofsatellitedata[51]andtheadditionofgeospatialcapabilitiesandinfrastructurewouldcreatevalueandstimulatetheeconomyaroundgeospatialdata.Cloudhostingofopendata,availabilityofschemes,policiesinmachine-readableformat[49],openportals,andreal-timeupdatesfromagencies[50]includingvisualizationdashboardsandmobileappshelpsinbetterintegrationofAIinusecases.

4)Stakeholdersbuy-inenabledbyStandards

Interoperabilityamongdifferentsolutionprovidersbringsthechoiceofdifferentvendors,irrespectiveofopenorproprietarysolutions,tosuchprimaryactors.Standardsplayanimportantroleinensuringcomplianceandinteroperability.

Forexample,primaryactorsintheagriculturedomainarethefarmers[14][35]whotaketheinitiativeinadoptingInternetofThings(IoT)-basedsensorsfordatacollection,edgedevicesforanalytics,andlow-powercommunicationsystems,whichimpliesthattheirtrustandwillingnesstoonboardareimportant.

5

AIReady–AnalysisTowardsaStandardizedReadinessFramework

Asanexample,anadvanceddrivingassistancesystem(seeclause4.2.3)involvesdifferentcarmanufacturerswithdifferentimplementationswhomightadoptdifferentparameters,thedivergenceinimplementationmightcreatelock-insituationsforuserspreventingflexibilityandchoiceofvendors.Additionally,issuesconcerningdataprivacy,dataprotection,andresponsibilitiesaretobestudiedcollaborativelyinopenstandardssuchasthosedevelopedbyITU,whichwillensuresecure,trustable,andinteroperableend-to-endsolutions.

5)DeveloperEcosystemcreatedviaOpensource

Cloud-hostedsolutionswithexposedAPIsforsubscribing/publishingdatafromportals[49]wouldcreatevaluefortheoverallindustryandleadtoinnovativeapplicationsthatsolvereal-worldproblemsusingAI/ML.Aprimeexampleisresearchsolutionsforsatellitedatausageinthefirepropagationmodel[51].

Referencesolutions,openmodels,andtoolsetscreatedinopensourcehelpinmobilizingresearchandinnovation,actingasabaselineforAIintegration,whichcouldbeextended,enhancedoroptimizedbasedonspecificusecaserequirements.SolutionspublishedasaresultofITUAI/MLChallengessuchastheTinyMLChallenge[66]aregoodexamplesofopen,published,anddeveloper-drivensolutions.

6)DatacollectionandmodelvalidationviaSandboxpilotexperimentalsetups

ITUdefinedMLSandboxin[ITU-TY.3172]anddescribedthedetailsofSandboxarchitecturesin[ITU-TY.3181].Inessence,Sandboxisanenvironmentinwhichmachinelearningmodelscanbetrainedandtheireffectstestedandevaluatedbeforedeployingintherealworld.Thishassinceseenwiderapplicationsinvarioususecases.

ImplementingcontinuousimprovementofmodelsusingfeedbackandoptimizationsintheSandboxhelpstooptimizeessentialtaskswithindisaster-strickenareas[52].Unmannedaerialvehicles(UAVs)canlearnandadjusttheiroperations(includingroutenavigation,returningtochargingstations,anddatadetectionandtransmission)basedonfeedbackfromtheenvironment.

Forexample,trafficregulationscenariosusingvisualcameras[36]andothersensorsuseAI/MLfeedbackloops,whichcollectdata,produceinferences,createactionrecommendationsandpolicyapplications,andaretestedandvalidatedusingpre-builttrafficplansforspecificoccasions.

PilotsetupsviaSandboxescanhelpinassimilatinglocalcommunitiesandutilitiesintothesolution.Forexample,in[51],firedetectionandpropagationmodelsaretestedandvalidated,andalarmsareusedtoprovideadvancedinformationtolocalcommunitiesandutilities.

6

AIReady–AnalysisTowardsaStandardizedReadinessFramework

3CaseStudies

Aspartofourstudiesonusecases,andourdetaileddiscussionswiththeusecaseauthors,wehaveselectedcertaincasestudieswhichbringoutthebenefits(orlackofit)forincreasing/measuringAIreadiness.EspeciallywefocusonthosecasestudiesthatutilizethereadinessfactorsmentionedinSection1above.Inaddition,welookforclearmetadata,supportingreferences,andpublishedresearchpapers,withexperimentationthatcanpracticallyshowcasethebenefitsofAIreadinessontheseterms.

Eachcasestudyismappedtothe6readinessfactorslistedinclause2aboveandtheinstancesofthereadinessfactorsareexplainedforeachcasestudy.

3.1CaseStudy-1:IoT-basedEnvironmentMonitoringBasedon

StandardIndices

Thiscasestudyinvolvesasetofusecaseswhichmonitorenvironmentparameterssuchassoilsensor,piezometers,andwaterlevelsensorsetc.andinferstandardizedindicesforspecificusecasese.g.groundwaterlevel(GWL)mappedtodroughtcodes(DC).Theareaofcoveragemaybequitelarge,forexample,multiplehectorsofforestland.Verificationofsenseddataandinferreddatawithgroundtruthincollaborationwithexpertsisanessentialcharacteristicofsuchusecases.Communicationnetworks,includingdataformatconversionsareimportantstandardrequirementsforsuchusecases.

Net-Peat-Zero[48]:NetworkedAssociationofSoutheastAsianNations(ASEAN)PeatlandForestforNet-ZerodeliveredbyUniversityPutraMalaysiaisanexcellentexampleofausecasewithreal-worlddeploymentanditsapplicationofopendata,whichisaccessibletoeveryone.

ThisusecasepresentsthepossibilitytoleverageAIinpredictingForestFireinpeatlandareasinSouthAsia.Animprovedtropicalpeatlandfireweatherindex(FWI)systemisproposed,bycombiningthegroundwaterlevel(GWL)withthedroughtcode(DC).Tomonitorthepeatland,aLoRa-basedIoTsystemisused,andsensorssuchassoilsensors,piezometersensors,waterlevelsensors,andweathersensorsareused,withtheexpectationthatintegralmeteorologicalinformationcouldbedetected.Allthedatamentionedabovecouldbecross-checkedwiththeonesusedbytheMalaysianMeteorologicalDepartment(METMalaysia),whichmeansthatthedatacollectedbytheIoTsystemisauthenticandreadytobeprocessed.

Inaddition,animprovedmodeltoapplytheGWLisproposedfortheFWIformulationintheFireDangerRatingSystem(FDRS).Specifically,DCisformulatedusingGWL,insteadoftemperatureandrainintheexistingmodel.FromtheGWLaggregatedfromtheIoTsystem,theparameterispredictedusingmachinelearningbasedonaneuralnetwork.TheresultsshowthattheDCcalculatedfromtheIoTsystemhasahighcorre