從邊緣到云有效利用NVAIE構(gòu)建自動(dòng)駕駛的端到端數(shù)據(jù)閉環(huán)-2025-03-自動(dòng)駕駛

FromEdgetoCloud

BuildanEnd-to-endData

Closed-loopforAutonomousDrivingbyEffectivelyUtilizeNVAIE

YooCar2025.3

Empowerglobalautomakers|BecomeaforerunnerintheAIera

IndustryDevelopmentTrends

fromRule-driventoData-driven

Empowerglobalautomakers|BecomeaforerunnerintheAIera

IntelligenceisTheNewWaveinAutomotiveIndustry

Intelligentdrivinghelpssellingmorecars

Intelligentdrivinghasbecomethetopthreefactorsfor

userstochooseacar.Theupgradingandpopularizationofintelligentdrivingsystemsiscontributingtothe

growthofcarsales.

Eg:XPENGandHuaweisoldmorecars(60%-80%)with

RoboTaxiisdevelopingrapidly

InWuhan,Baidureleasedthesixth-generationdriverless

carsequippedwiththeworld'sfirstautonomousdrivinglarge-scalemodelsupportingL4leveldriverless

applications,whichisexpectedtoachievebreakevenin

Wuhanbytheendof2024；

ADASversion.

L3iscoming

ChinahasannouncedalistofthefirstbatchofL3self-drivingroadtrials,includingnineautomakersthatcandevelopmass-producedL3self-drivingproducts.

Tesla'stechnologicalevolution

TheperformanceofTeslaFSD12.5versioninresponseto

complexroadconditionsandsmoothdrivinghasgeneratedheateddiscussion,whichisexpectedtoenableautonomousparking,navigationinparkinglotsanddriverlesstransportofpassengers；

DrivingForceofAutonomousDrivingTechnology

Phase3

DataDriven

Phase1

SensingHardwareDriven

Newtechnology

Datascalegrow

Algorithmoptimization

Data

transmit

Datacollect

Model

training

driven

Rule

DataMining

Data

process

ing

radar/LidarCamera

Modeltraining

Hardware

improvement

First

dispose

Simulati

on

testing

Application

deployment

Datalabel

Emphasisonhardware

Closed-loopdatasystem

Feature:earlystage,mainlydrivenbyhardware,theimportanceofdataisbeginningtoreveal,anddataloopisintheinitialperiod;

ADASFUNCTION:L0/L1;AEB、ACC.

Feature：datascalegrowsrapidly,thedataclosed-loopdrivenbysmallmodelscannothandlelarge-scaledata,andtheemergenceofnewtechnologiessuchasAllargemodelsandcloudsimulationcontinuestoempowerhigh-levelintelligentdriving.

ADASFUNCTION：L2;highway/cityNOA

Phase2

AlgorithmDriven

Software

modularization

Component

Algorithmupdate

Feature：algorithmoptimizationmakesdataloopmoreintelligent，reducesmanualintervention,andimprovestheaccuracyandefficiencyofdataprocessing.Italsoenablesmoreefficientdataacquisition,processingandanalysis

ADASFUNCTION：L2;TJA

Datasecurity

Data

compliance

Environmentsensing

Route

planning

algorithmoptimization

Motioncontrol

DecisionMaking

update

OTA

Empowerglobalautomakers|BecomeaforerunnerintheAIera

Empowerglobalautomakers|BecomeaforerunnerintheAIera

AutonomousDrivingEvolutionRoute

Rules

driven

Datadriven

Multisens

or

Fusion

Sensor

2D

perception

Traditional

Sensor

PlanningControl

Environment

Sensor

Control

Mainstream

BEV+Transformer

Sensor

Partialrulescontrol

Sensor

Leading

EndtoEndModel（perception+PnC）

Sensor

Sensor

Future

Sensor

EndtoEndSelfDriving,Explainable

RequirementofDataClosed-LoopinADEvolution

Leading

iMlainstream

Carcontrol

Datacollect

Datatype

Dataprocess

Labeltype

Training

Datascale

Computing

requirement

Largemodel+rules

Massmarketcars

video+humanoperation

Cloud

Manual+AIlabel

sensing+largemodel

GB/day/car

*millionsofcars

high，thousandsGPU

Traditional

Carcontrol

Datacollect

Datatype

Dataprocess

Labeltype

Training

Datascale

Computing

requirement

Datadrivenrules

Collection/massmarketcars

Pointcloud+video

car+cloud

4Dlabel

prefusion+targetperception

TB/day/car

*hundreds

middle，hundredsGPU

Carcontrol

Datacollect

Datatype

Dataprocess

Labeltype

Training

Datascale

Computing

requirement

Largemodel

Massmarketcars

video+humanoperation

Cloud

AIlabel

Largemodel

GB/day/car

*millionsofcars

Veryhigh，10thousandsGPU

Future

Carcontrol

Datacollect

Datatype

Dataprocess

Labeltype

Training

Datascale

Computing

requirement

rules

Collectioncars

Pointcloud+video+map

Incar

3Dlabel

Target

perception+postfusion

GB/day/car

*hundreds

low，10GPU

IndustryBenchmark：Tesla

AftertheFSDV12update,theunderlyinglogichasbeenexplicitlychanged,

replacingalotoftherulecodewithanend-to-endautomateddriving

algorithm.Learnhumandrivinghabitsthroughthemassivedatacollectedby

shadowmode.ThelatestversionofFSDisbecomingmoreandmorelikeanoldhumandriver

WiththeAutopilotdataengineframeworkasthecore,datacollection,cleaningandAIlabel,theyaccumulatearichscenelibrary,whichisusedforend-to-endmodeltraininganddeploymentthroughasuper-large-scalecluster.

TeslahasanequivalentclusteroftensofthousandsofH100cardsandspendsmorethan$10billionperyearfordataacquisition,storage,andtraining.

ChallengesforAutomobileManufacturers

inBuildingaDataClosed-loop

CostPressureandProductMaturity

DateClosed-LoopforAutonomousDriving

Algorithmupgrade

Hardwareandsoftware

updates

Themostprominentissueinthecommercializationofintelligentdrivingdataclosedloopisthehighcostofresearchanddevelopment,processing,andoperations.Thehighcostsincreasethedifficultyofcommercializationandlimitthewidespreadapplicationofdataclosedloops.AchievingcostreductionandefficiencyimprovementinallaspectssuchasR&D,operations,andprocessingwillhelppromotethecommercializationofdataclosedloops.

Researchand

developmentcosts

datacenter

Deploymentverification

Engineer

Programmer

Operator

Toolchain

development

cloudservices

rateofflow

?Vehicleenddataneedstobetransmittedtothedatacenterorcloudplatforminrealtimeor

periodically,whichwillproducedatatransmissiontrafficcosts

?Alargeamountofstoragespaceisrequiredto

savetheoriginaldata,processeddataand

intermediateresults,whichisaffectedbystoragecapacityandstoragetype

?Thecostofcomputingincludesthecostof

computingresourcesusedfordataprocessing,analysisandmachinelearningmodeltraining

?Theresearchanddevelopmentprocessdrivenby

autonomousdrivingdataismulti-faceted,and

targeteddesignanddevelopmenttoolsarerequired,suchasdataminingtools,modelingtools,simulationscenelibrarybuildingtools,etc.

?Self-builtcomputingpowercentersgenerallyexceed1billionyuan,suchasGeelyStarWisdomComputingCenterwithatotalinvestmentof1billionyuan,andcloudcomputingpowerof8.1billiontimesper

second.

?ThecostofrelatedAImodels,suchasGPT-3trainingcostsmorethan$12million)

computing

storage

operatingcosts

timecost

Operationsand

maintenance

costs

treatmentcost

Deploymentcosts

?Itisgenerallybelievedintheindustrythatthedevelopmentofautonomousdrivingplatformtheoreticallyneeds3

months,andthecollection,labelingandtrainingneedalotoftimecost

?Highfrequencyupdatesandoperationsarerequired,

includingmassivedatastorage,softwaresystemiteration,operationsinfrastructure,andhardwareequipment

operations

?Theannotatedandprocesseddatashouldbeputintothelargemodelfortraining,andthenusedforsimulation

testinganddeploymentatthevehicleend

Empowerglobalautomakers|BecomeaforerunnerintheAIera

Empowerglobalautomakers|BecomeaforerunnerintheAIera

StatusandObjectiveofDataClosed-LoopforAutonomousDrivinginOEM

StatusofDataClosed-LoopforAutonomousDrivinginOEM

HighCostindata

TransmissionandStorage

LackofDataCompliance

Practice

TimeConsuminginModelTraining

DifficultyinDataCollection

?Qualificationsareneededfor

surveyingandmapping；

?Highcostfordatacollection

FewScenariosforSimulation

?simulation

scenariosetupneedslarge

amountofdata；

?Lowefficiencyinbuildingcornercasescenarios

?Highcostfor

datatransmissionandstorage

betweenvehicleandcloud

?Lowefficiencyinmanuallabelingoflargeamountofdata

?Slowgenerationofqualitydata;

?Lackofcomputingpowerresultingintraining

LessEfficiencyinDataLabeling

?Highcostfordatacompliance;

?Lackofpracticesfordata

compliance

in

EfficientDatatransmission

andStorage

AutomaticGenerationofData

DataLabelingEfficientModelTrainingQualitySimulationTest

Automatic

DataTransmissionCompliance

Objective

Cornerandrare

scenarioscanbe

definedin

simulationtoolstogeneratevariousdatawith

multimodallargemodel.

Dataarepre-

processedinvehiclesorlocaledgesitestoeaseoftheburdenoftransmissionand

storage.Onlykey

dataareuploadedtocentralsites.

Distributed

architectureand

parallelcomputingpowerofcloud

simulationenhanceoverallefficiency;

scenariogenerationabilityenriches

DatalabelingisdoneautomaticallywithAIlargemodeltolowerhumancostand

improveefficiency.

Largeamountof

qualitydataareusedformodeltrainingtoensuremodel

accuracyand

generalizationability.

Datacompliancepracticesare

enforcedtoavoiddatabreachandtampering.

simulationscenarios.

DataClosed-loopToolchain

TheYooDriveCloudsolutioncollects,stores,processesanddistributeslargeamountsofin-vehicledatatosupportawiderangeof

applications.Bybuildingacompliancedataframework,wecanrelievecustomers'worriesaboutdatacomplianceandhelpcustomersrealizetheclosed-loopofautonomousdrivingdataandtheefficientiterativedevelopmentandupdateofcrowd-sourcedmaps.Theself-developeddataclosed-looptoolchainprovidesdata-drivenandclosed-loopprocessing,includingacquisition,processing,storage,training,simulationandothermodulestopromotethematurityofalgorithmmodels.

Thedataclosed-loopincludesmodulessuchascollection,generation,processing,

Out-of-boxSolutions

Data

recharge,andanalysis.Graduallyfromtheclosed-loopcollectionofvehicledatatothe

applicationoflarge-scalemodelsofmassproductionvehicles.

AIDC

ProvideAIcomputingcentersolutionsandcomputingpowerleasingbasedonNVIDIAordomesticGPUclusters,andprovidecomputingpowermanagementandbillingservices.

Toolchain

Itprovidesintegratedmultipleannotationservices,automaticannotationofdata,distributedalgorithmtraining,andsupportmulti-taskandqueueresourceschedulingmanagement.

Model

Providemodelapplicationmarketandonlineinferenceservices,one-stopsolutiontolargemodelapplicationandinference,andprovidemorepersonalizedandintelligentservices.

One-stopservice

Providefull-stackandcompletetoolchain

solutions,including:hardware+software+

computingpower+serviceone-stopservice;Connecttoolstoenhancebusinesscontinuity.

Strongcompatibilityandunified

management

Throughdataformatconversionandunification,multipledifferentplatformscanbeconnectedatthesametime,reducingmanagementcomplexity.

CustomizedService

Flexiblesystemarchitecture,with

generalizationabilityandadaptability,toprovidecustomizedservices.

AutonomousDrivingSolution

One-stopServiceforVehicle,Network,CloudandEdgecomputing

AutonomousDrivingSolution

Cloud

Edge

Network

Vehicle

TSP

AutonomousDrivingToolchainPlatform

Empowerglobalautomakers|BecomeaforerunnerintheAIera

Empowerglobalautomakers|BecomeaforerunnerintheAIera

DualSIMCardsNetworkingConnectionManagement-(TBOX+SDK)

MobileOperatorscanswitchflexiblyondemand

DualslMicardssupportcapability

nIncooperationwithleadingautomotivecompaniesintheindustry,YooCarTechnologytakestheleadinrealizingthedual-cardnetworkingsupportcapability,helpingcustomertoflexiblyselectnetworkingoperators,andprovidingservicesincludingscenarioanalysis,switchingmodedesignandswitchingstrategymanagement.

nMNO,thevehiclenetworkingbusinessmanagementplatformofYooCarTechnology,hasbeenfullyequippedwiththedual-cardmanagementcapabilitiesofnetworkedvehicles,suchasdual-cardstatusmaintenance,dual-cardpackageorderingandconsumptioninquiry.

nProvideuserswithhighlyavailablevehiclenetworkingservicesandgreatlyimprovetheuserexperience.Dual-cardnetworkingsupportcapabilitycanflexiblyadapttodual-carddual-standby,dual-carddual-passanddual-cardsingle-standbyoperationmodesaccordingtocustomerneeds.

Thedataqualityishigh

Theintegrityofdatauploadisensuredthroughthedual-cardhigh-qualitytransmissionchannel,andthedataqualityissignificantlyimproved.

Gooduserexperience

Selectthebestoperatorthroughdual-cardswitchingtoensurethattheentertainmentexperiencedoesnotconflictwiththedatacollectionandreturn,andenhancetheuserexperience.

Andthatreal-timeperformanceisgood

Real-timelargebandwidthdatauplinkanddownlinkpowerdataclosed-loopandremotedrivingfunctions.

DedicatedCoreNetworkforOEM

ComparedwithtelecommunicationoperatorsprovidingconnectionservicesbycommonIoTCore,YoocarhasbuilttheDedicatedCoreandassociatedCMPthroughextensiveCAPEXandOPEX.TheDedicatedCoreandCMPprovidesdifferentiatedcapabilitiesforconnected-carservices,featuringrobustconnection,flexiblemanagementandquickresponse.

OEMTSPCloud

Yoocar

ConnectivityMgt.

Platform

Operator

OEM

MNO

Other

Modules

4G/5GDedicatedCore

ServiceResponseTimeCutby80%

Downby

83%

s

10day

s

30min

CommonIoTNWDedicatedCore

SignalingCapture

15day

Downby

80%

Downby

90%

Config.Change

BillingDelay

3days

5min

1day

?FlexibleAPN-basedcontrolandapplication-basedcontrolandanalysiswithinoneAPN

connectedvehicleissuesareresolvedbeforecustomercomplaintsarise.The

inspectionservices.Routineprobingandinspectionsareconductedonbusinessnodestoensurethat

platformhasalreadyprovidedinspectionandO&MservicesforallbrandsunderGeely,aswellasfor

Empowerglobalautomakers|BecomeaforerunnerintheAI

IntelligentOperationandMaintenanceManagementServiceforVehicleNetwork

Carafter-salesservice400seatsYoukaiexpertTripartiteexpertsOperationsand

maintenanceexpert

Dialoguerobot

Manualprocessing-after-sales/agent/expertaccess

clothes

affair

request

ask

come

into

contact

with

enter

reportoutsideknow

know

storeroo

m

joinin

marriage

place

Thetube

texture

transport

goout

Diagnostictemplates

Workorderaccess

Inspectionrobot

Workordermatching

Workorder

Workorder

processingtime

Processtemplates

completed

Tripartitediagnosisaccess

ThethreepartiessupportKPI

Telepresencerobot

Workorderprocessing

manualhandling

Routinediagnosticaccess

SLAestimate

Systemquery

businessdiagnosis

Documentrobot

Theepicenterofevents

Changereleasemanagement

knowledgebase

Interfacerobot

TSP

DataPlatform

CP/SPserve

Automatic/systemprocessing--accesstoexternalsystemcapabilities

carrieroperatorCMP

Intelligentdrivingdataaccess

operationandmaintenanceservices

Lnrt,LiAuto,andZhiJi.

TheYoukaIntelligentOperationsandMaintenanceManagementPlatformisanewgenerationofoperationsmanagementplatformsdevelopedbasedonthelatestAIOpsstandardsfromTMForum.Itoffersone-stopintelligentO&MservicespoweredbyAIrobots,providingpredictivemaintenanceforautomakersthroughself-serviceresponses,probing,and

Empowerglobalautomakers|BecomeaforerunnerintheAIera

OTAandRemoteDriving

5GEdgeCloud

Thankstothe5GDedicatedCoreand5GSAdataoffloadingtechnology,Yoocarprovidescustomizededge-computingbasedservicesfeaturingworry-freebandwidth,telecomgradeandend-to-endmonitoringthatenablescommercialdeploymentofOEMserviceslikeassistanttele-driving,OTA,voiceassistantandcloud-basedin-vehicleinfotainment.

MNOCoreInternetPublicCloud

ADData

MNOCore

CollectionPrivateCloud

VehicledataandimagesuploadingTele-drivinginstructionssending

EdgeCloud

LatencyCutforBetterExperience

Datatransmissionlatencyisreducedforquicktransactionresponseandbetteruserserviceexperience

LoadReductionforCostEffectiveness

Dataprocessingatedgesiteseasesthe

computingandstorageburdenofcentralsitesandcutsutilizationoftransmissionbandwidth

SecurityandPrivacyProtection

Dataencryptionandprivacyprotectionaredoneatlocaledgesitesforenhancedsecurityoftelematicsservices.

EdgeCloudEmpowersOTAandRemoteDriving

Usingtheedgecomputingcapability,thedelayandjitterare

Province-1

I-SMF

AMF

significantlyreduced.Themeasureddelayisreducedfrom55msto25ms,andthejitterisreducedfrom280msto35ms.

I-UPF

TSPOTAserver

Province-2

I-SMF

AMF

Vehiclenetworkdedicatednetwork

APN1

PCF

I-UPF

SMF

AMF

……

RemotedrivingcontrolOTAupgradepatchMapdataupdated

Edgecomputingcloudresourcepool

Province-N

UPF

APN2

(SpecificIPfor

services)

I-SMF

AMF

I-UPF

Internet

APN2

Empowerglobalautomakers

|BecomeaforerunnerintheAIera

Empowerglobalautomakers|BecomeaforerunnerintheAIera

Data

standardization

Datacleansing

datamining

Dataannotation

DQC

datamining

DataMasking

Auditmonitoring

Datacompliance

Complianceflow

Dataprocessing

Dataannotation

Datastorage

Distributed

storage

filestorage

Storage

acceleration

SimulationPlatform

Parallelsimulation

Simulationplayback

Simulation

evaluation

SystemArchitecture

Modeltraining

Modeltraining

Modelevaluation

Queue

management

machinelearning

Bigdata

Trainingcluster

/inferenceframework

AIworkloadmanagement

HDFS

MinIO

Spark

Flink

Prometheus

CICD

MySQL

Redis

Kafka

ES

Databasesandmiddleware

storage

Systemintegration

Hashpools

Computingpowerscheduling

Hashratesplitting

Computingpowermanagement

K8s

storage

Internet

compute

DataManagementPlatform

?Thedatamanagementplatformisoneoftheimportantfoundationsfortheresearchanddevelopmentofhigh-levelassisteddriving/autonomousdriving,and

undertakesservicessuchasdataupload,datastorage,dataprocessing,dataapplicationandmanagementofintelligentdriving.Theintegrateddatamanagementplatformallowsdatatoformabenignclosed-loopnetworkbetweenvariousbusinessplatforms,helpingtoreleasemorevaluefromdata.

?Itprovidesdatasupportforcornercaseproblemanalysis,modeltraining,dataannotation,simulationscenarios,evaluationandverificationintheR&Dprocessofhigh-levelassisteddriving/autonomousdriving.

DeployonBoard

Acquisition

Requirements

AlgorithmEvaluation

ProductionCar

ModelTraining

SimulationTesting

ResourceSharing

AlgorithmModelIsolation

Data

UnifiedAnnotationFormatConversion

Simulation

Library

Manageme

nt

Algorithmi

cModels

Manage

Label

Data

Manage

SceneLibraryManage

Source

Data

Manage

Data

DataSetManage

managementplatform

UnifiedScheduling

andMulti-task

DistributionofData

tobeLabeled

DataCollectionDataUploadsDataProcessingSceneMiningDataAnnotation

Empowerglobalautomakers|BecomeaforerunnerintheAIera

DataClosed-LoopComplianceServices

Compliancecloud

(ClassAqualificationentitycontrol)

Compliancedatamanagement

Datastorageusagearea

Compliancesupervision

CustomerVehicularCloud

ComplianceBusinessCloud

Onlinereturn

(Internetline)

Dataferryarea

ComplianceBusinessCloud

network(non-

GISdata)

Corporate

Customerofficearea

TerminalPC

Otherpublicnetworkfacilities

DataAccommodation

Service

GPSdeflection

Massproduction

mapsupplier

ProductionVehicles

Businessdata

(Testreports,modelfiles,

traininglogfiles,etc.)

VDI

(Qualificatio

nEntity

Control)

Picturevideoafterdesensitizationand

decryptionanddiscontinuoussingle-frame

pointclouddataafterframeextraction

(excludingabsoluteposition)

Dataforwarding

Datausagearea

ComplianceBusinessCloud

Specialline(samecity)Complianceroom

Problem

management

Dataannotation

Dataannotation

reductionarea

CompliancePrivateCloud

Offlinemailing

Datauploadservice

......

VDI

(Managementandcontrolofqualification

subject)

Customer

Datastore

Trafficflow

Dataforwarding

geographicDesensitizationanddeclassificationof

non-geographicdataComplianceWorkstation

Dataflow

AutopilotR&DandDataComplianceUse

ScenelibraryAdministration

Compliance

Workstation

Scenariolibrarymanagement

Modeltraining

Modeltraining

Special

line(samecity)

Simulationtest

Simulationtest

Subjectofqualification

EngineeringVehicles

Uploadroom

DataDesensitizationandDecryptionService

Three-partyannotation

supplier

SpecialLine

(Inthesamecity)

Desensitizedanddeclassifieddata

Datauploadandcomplianceprocessing

Datacollectionandtransfer

Desensitizationanddensity

Problem

management

Datausagearea

ComplianceBusiness

Cloud

Complianceapproval

Empowerglobalautomakers|BecomeaforerunnerintheAIera

Product&Service

?VehiclenetworkservicemeetsthespecialneedsofAIEVnetworkcommunication;

?IntelligentcomputingservicesmeetthetechnicalneedsofautomobileenterprisestobuildAIEV,including:①Autonomousdrivingtoolchain②Specializedlargemodelresearchanddevelopmentsupportforautomobileenterprises③CockpitAIinteractiontechnologysupport;

?One-stopsolutionforartificialintelligenceonbicycles,coveringnetworkconnection,datasecurity,AIlargemodeldevelopment,AIfunctionalloadingandpracticalapplication,AIecologicalconstruction,AIfull-chainproductservicesystem;

Globalconnectivityservices

High-speednetworkingisanecessityforAI

Basedonthedualcardservicecapability,itensureslowlatencyandnobreakpointofnetworkconnection,

layingafoundationforautonomousdrivingandintelligentcockpitAI.

Theprivatenetworkensuresdatasecurity

Basedontheadvantagesofdedicatednetwork

technology,dynamicmonitoringofthewholelinkofdatacirculationisrealizedtoensuredatasecurityandeliminatetheriskofdataprivacyleakagecausedby

largeAImodels.

Fullchainintelligentcomputingproductsandservices

Autonomousdrivingtoolchainplatform

Itincludesdatamanagementplatform,data

annotationplatform,modeltrainingplatformandsimulationtestingplatform,andself-

developedfullchain.

Automobilecompanieslargemodelcomputingpower+R&Dsupport

Computingpower+R&Dtoolssupportcarcompaniestodevelopdedicatedlargemodelsbasedontheirown

needs/cabinoperatingsystem.

CabinAIinteractionfunctiontechnicalsupport

Basedonthemaincirculationlargemodelandthe

applicationplatformofthelargemodel,anAIinteractiveassistantwillbebuilttorealize"AIinteraction"onthebus.

NVAIEEmpowersDataClosed-loop

forAutonomousDriving

YoocarandNVIDIACooperatetoEmpowerCarCompanies

Empowerglobalautomakers|BecomeaforerunnerintheAIera

DataClosed-loopToolChainTechnologyEvolutionandIterationAV1.0->AV2.0

?DataCleaning

?Manual

Annotation

?ModelTraining

?SceneReplay

?Inference

Optimization

AV1.0AV2.0

?AutomaticCleaningandLabeling

?TrainingAcceleration

?BaseModelPool

?Large

Language/MultiModelSupport

?WorldModelReasoning

Optimize

?SyntheticDataGeneration

?BasedonNeural

NetworksScene

ReconstructionEngine

?WorldModelSimulation

?ImageIndexingandRetrieval

AV…

Empowerglobalautomakers|BecomeaforerunnerintheAIera

DataClosed-loopToolChain&NvidiaNVAIE/OmniverseEna