Robot systems for rail transit applications 1st edition hui liu

Page 1

Robot Systems for Rail Transit Applications 1st Edition Hui Liu

Visit to download the full and correct content document: https://ebookmass.com/product/robot-systems-for-rail-transit-applications-1st-editionhui-liu/

RobotSystemsforRailTransit Applications

SchoolofTrafficand TransportationEngineering, CentralSouthUniversity, Changsha,China

Elsevier

Radarweg29,POBox211,1000AEAmsterdam,Netherlands

TheBoulevard,LangfordLane,Kidlington,OxfordOX51GB,UnitedKingdom 50HampshireStreet,5thFloor,Cambridge,MA02139,UnitedStates

Copyright © 2020ElsevierInc.Allrightsreserved.

Nopartofthispublicationmaybereproducedortransmittedinanyformorbyanymeans,electronicor mechanical,includingphotocopying,recording,oranyinformationstorageandretrievalsystem,withoutpermission inwritingfromthepublisher.Detailsonhowtoseekpermission,furtherinformationaboutthePublisher’ s permissionspoliciesandourarrangementswithorganizationssuchastheCopyrightClearanceCenterandthe CopyrightLicensingAgency,canbefoundatourwebsite: www.elsevier.com/permissions

ThisbookandtheindividualcontributionscontainedinitareprotectedundercopyrightbythePublisher(otherthan asmaybenotedherein).

Notices

Knowledgeandbestpracticeinthis fieldareconstantlychanging.Asnewresearchandexperiencebroadenour understanding,changesinresearchmethods,professionalpractices,ormedicaltreatmentmaybecomenecessary.

Practitionersandresearchersmustalwaysrelyontheirownexperienceandknowledgeinevaluatingandusingany information,methods,compounds,orexperimentsdescribedherein.Inusingsuchinformationormethodsthey shouldbemindfuloftheirownsafetyandthesafetyofothers,includingpartiesforwhomtheyhaveaprofessional responsibility.

Tothefullestextentofthelaw,neitherthePublishernortheauthors,contributors,oreditors,assumeanyliability foranyinjuryand/ordamagetopersonsorpropertyasamatterofproductsliability,negligenceorotherwise,or fromanyuseoroperationofanymethods,products,instructions,orideascontainedinthematerialherein.

LibraryofCongressCataloging-in-PublicationData

AcatalogrecordforthisbookisavailablefromtheLibraryofCongress

BritishLibraryCataloguing-in-PublicationData

AcataloguerecordforthisbookisavailablefromtheBritishLibrary

ISBN:978-0-12-822968-2

ForinformationonallElsevierpublicationsvisitourwebsiteat https://www.elsevier.com/books-and-journals

Publisher: MatthewDeans

AcquisitionsEditor: GlynJones

EditorialProjectManager: NaomiRobertson

ProductionProjectManager: NirmalaArumugam

CoverDesigner: MatthewLimbert

TypesetbyTNQTechnologies

Listoffiguresandtables

Figure1.1Robotsinthemanufacturing,dispatch,andmaintenanceofrailtransit.

Figure1.2Roleofrobotsinrailtransitmaintenance.

Figure1.3Keyproblemsofrailtransitrobotsystems.

Figure2.1Workingstepsoftheassemblyrobot.

Figure2.2Differenttypesofassemblyrobots.

Figure2.3Overallframediagramofrailtransitassemblyrobotsystem.

Figure2.4Maincomponentsofassemblyrobot.

Figure2.5Mechanicaldiagramofassemblyrobot:(A)basecomponent,(B)rotating joint,(C)armconnectingcomponent,(D)wristjoint,and(E)end effector.

Figure2.6Trajectoryplanningalgorithms.

Figure2.7Processofusingtheartificialneuralnetwork(ANN)forinversedynamics calculation.

Figure3.1Multirobotcollaboration.

Figure3.2Human robotcollaboration.

Figure3.3Theblockdiagramofcollaborativerobotsystem.

Figure3.4Sensorsforcollaborativerobots.

Figure3.5Classificationofendeffectors.

Figure3.6Featureextractionalgorithms.

Figure3.7TheflowchartoftheHOGalgorithm.

Figure3.8TheflowchartoftheSIFTalgorithm.

Figure3.9TheflowchartoftheLBPalgorithm.

Figure3.10Targetdetectionalgorithms.

Figure3.11Targettrackingalgorithms.

Figure4.1Maincontentdiagramofautomaticguidedvehicles(AGVs).

Figure4.2Navigationmethods. AGV,automaticguidedvehicle; LiDAR,lightdetectionandranging; SLAM,simultaneouslocalizationandmapping.

Figure4.3Globalpathplanningalgorithms.

Figure4.4Localpathplanningalgorithms.

Figure4.5Human robotinteractionalgorithms.

Figure4.6Flowchartofthehybridpathplanningmodel. KELM,kernel-basedextreme learningmachine; QPSO,quantumparticleswarmoptimization.

Figure5.1TheadvantagesofAutonomousrailRapidTransit(ART).

Figure5.2TheoverviewdiagramofAutonomousrailRapidTransit(ART).

ix

Listoffiguresandtables

Figure5.3SchematicdiagramofARTsensorfusion.

Figure5.4Thecoreofthepedestriandetectionalgorithm.

Figure5.5TheflowchartofHistogramofOrientedGradient(HOG)feature þ SupportVectorMachine(SVM)classification.

Figure5.6TheflowchartoftheSupportVectorMachine(SVM)classification.

Figure5.7Theflowchartofpedestriancontourextraction.

Figure5.8Posturerecognitionprocess.

Figure6.1Inspectionrobotstructure.

Figure6.2Therailwayapplicationsoftheinspectionrobots.

Figure6.3Maincomponentsoftheinspectionrobots.

Figure6.4Railtransitinspectionrobottechnologies.

Figure7.1Advantagesofdual-armrobots.

Figure7.2ChannelrobotsTroubleofMovingElectricMultipleUnitsDetection Systemdiagram.

Figure7.3Structureofgroundtrack.

Figure7.4Formingdiagramofinfraredimageformation.

Figure7.5Analysisprocessofvisibleimage. HOG,histogramoforientationgradient.

Figure7.6Intelligentanalysismethodofinfraredthermalimage. ANFIS,adaptive network-basedfuzzyinferencesystem; BP,backpropagation.

Figure7.7Bogiefaultdiagnosisstructurediagram.

Figure7.8Flowoffaultdiagnosis. EMD,empiricalmodedecomposition; WPT, waveletpackettransform.

Figure7.9Thesigmoidfunction.

Figure7.10Thetanhfunction.

Figure7.11Therectifiedlinearunitfunction.

Figure8.1Maincomponentsofrailtransitinspectionunmannedaerialvehicles (UAVs). GPS,globalpositioningsystem; IMU,inertialmeasurementunit.

Figure8.2Schedulingsystemofanunmannedaerialvehicle(UAV).

Figure8.3Flowchartofperceptionalgorithm. ROI,regionofinterest.

(A)Thereal-timeimagestabilizationofthecollectedvideoisfinalized.

(B)Thetrackregionofinterestisextractedfromtheperceptionimage accordingtorailwayboundaryregulations,toreduceunnecessary operationsandspeedthesubsequentimageprocessingrate.

(C)Theintrusionisidentifiedbasedonatrackmap.

Table2.1Advantagesanddisadvantagesofthreeassemblymethods.

Table2.2AdvantagesanddisadvantagesofjointspacetrajectoryplanningandCartesianspacetrajectoryplanning.

Table3.1Comparisonoftraditionalindustrialrobotsandcollaborativerobots.

Table3.2Severalmainstreamcollaborativerobotsandmanufacturers.

Table5.1ComparisonofmaintechnicalsolutionsofARTandmoderntramsignal system.

Table5.2Traditionalpedestriandetectionalgorithms.

x

Table5.3Pedestriandetectionalgorithmsbasedondeeplearning.

Table5.4ThepedestrianposturerecognitionaccuracyofBPneuralnetwork.

Table7.1Compositionoflasersensor.

Table7.2Functionsandcharacteristicsoftwo-dimensionallasersensor.

Table7.3Comparisonofchannelrobotwirelessrechargingmethods.

Listoffiguresandtables xi

Preface

Railtransitisthelifebloodofmanynationaleconomiesandthebackboneoftransportation. Safeandefficientrailtransitisbasedonhighlyreliablemanufacturingandhigh-quality maintenance.Railtransitrobotscanreplacemanualrepetitivetasksandimprovetheautomationleveloftherailtransitsystem.Asaresult,workefficiencycanbeimproved,accidentalfailuresduetohumannegligencecanbeavoided,andthesafetyoftherailtransit systemcanbeimproved.

Railtransitrobotsinvolvetheintersectionofautomaticcontrol,artificialintelligence,signal processing,patternrecognition,mechanicalengineering,andtransportationengineering. Whenappliedtotherailtransitsystem,therobotisfacedwiththekeyproblemstobe solvedurgently.Therefore,railtransitrobotsarecurrentlyrecognizedasresearchhotspots amongscientificproblems.Basedonresearchfromthepast10years,theauthorputs forwardaframeworkofrailtransitrobottechnologyandcompletestherelatedwork.

Thisbookcoverssevenmainstreamrailtransitrobots,includingassemblyrobots, collaborativerobots,automatedguidedvehicles,autonomousrailrapidtransit,inspection robots,channelrobots,andinspectionunmannedaerialvehicles.Thekeyproblemsof robotsaredescribedindetail,includingpositioningnavigation,pathplanning, human robotinteraction,andpowermanagement,etc.Forstudentsandmanagersinrelated departments,thisbookcanprovidevaluableinformationaboutrailtransitrobots.For researchersanddoctoralstudents,thisbookcanprovidesomeideasandencouragefuture researchinrailtransitrobots.

Thisbookcontainseightchapters:

Chapter1:Introduction

Thischapterfirstoutlinestherailtransitrobot.Thenthechapterdescribesthreebasicissues ofrobotics,includingnavigation,human robotinteraction,andpowercontrol.

Chapter2:Railtransitassemblyrobotsystems

Thischapterfirstintroducesthedevelopmentprogressandkeytechnologiesofassembly robots,andthenintroducesthemaincomponentsofassemblyrobots.Afterthat,the dynamicsmodelsfortheassemblyrobotsystemsareexplained.Finally,theartificialneural networkalgorithmfortheinversedynamicoptimizationcalculationoftherobotarmis introduced. xiii

Chapter3:Railtransitcollaborativerobotsystems

Thischapterfirstgivesthebasicdefinitionofacollaborativerobot.Thenthechapter summarizesthedevelopmenthistory,applicationfieldandcollaborativemodeof collaborativerobots.Thecomponentsofthecollaborativerobotaresummarized.Finally,the basicconceptsofvisualperceptioninhuman-robotcollaborationareintroduced.

Chapter4:AutomaticGuidedVehicles(AGVs)intherailtransitintelligent manufacturingenvironment

ThischapterfirstintroducesthedevelopmentprogressandtypesoftheAGVinrailtransit intelligentmanufacturingenvironment.ThenthemaincomponentsofAGVsareintroduced. Afterthat,thekeytechnologiesandtheapplicationsoftheAGVareintroduced.

Chapter5:AutonomousrailRapidTransit(ART)

ThischapterfirstlyintroducesthehardwareofART.Then,thetechnologiesofARTare introduced.Finally,thepedestriandetectionalgorithmsofARTareintroducedindetail.

Chapter6:Railtransitinspectionrobots

Thischapterfirstintroducesthedevelopmenthistory,functionandmaincomponentsofthe railtransitinspectionrobot.Then,twokeytechnologiestoensurethattheinspectionrobots normallycompletetheinspectionwork,positioningmethods,pathplanningmethods,and hand eyevisionsystemareintroducedindetail.

Chapter7:Railtransitchannelrobotsystems

Thischapterfirstlyintroducesthedevelopmenthistoryandmaincomponentsoftherail transitchannelrobot,includingthegroundrail,dual-armrobot,infraredthermometer,laser sensor,etc.ThentheTEDSintelligentsensingsystemisdescribedindetail.Finally,fault diagnosisalgorithmsbasedondeeplearningmodelsareintroduced.

Chapter8:RailtransitinspectionUnmannedAerialVehicle(UAV)

ThischapterfirstintroducesthedevelopmenthistoryoftheUAVanditsapplicationsin variousfields.Secondly,itintroducesthebasicstructureoffixed-wingUAVs,unmanned helicoptersandrotary-wingUAVs.Varioussensorsareappliedtorailtransitinspection.

ThentheUAVtechnologiesaredescribedindetail.Finally,theapplicationsoftheUAVin thedetectionofrailtransitintrudingdetectionareintroduced.

xiv
Prof.Dr.-Ing.habil.HuiLiu Changsha,China November2019 Preface

Acknowledgement

ThestudiesinthebookaresupportedbytheNationalNaturalScienceFoundationofChina, theNationalKeyR&DProgramofChina,andtheInnovationDriveofCentralSouth University,China.ThepublicationofthebookisfundedbytheHigh-levelPostgraduate TextBookProjectoftheHunanProvinceofChina.Intheprocessofwritingthebook, Mr.ZhuDuan,Mr.JiahaoHuang,Mr.KairongJin,Mr.YuXia,Mr.RuiYang,Ms.ShiYin, Mr.YeLi,Mr.GuangjiZheng,Ms.JingTan,Mr.HuipengShi,Mr.HaipingWu,Mr.Chao Chen,Mr.ZhihaoLong,andotherteammembershavedonealotofmodelverificationand otherwork.Theseteammembersasmentionedhavethesamecontributiontothisbook.

xv

Nomenclaturelist

ACO AntColonyOptimization

ADC Analog-to-DigitalConverter

ADU AutomaticDrillingUnit

AGV AutomaticGuidedVehicle

AGVS AutomatedGuidedVehicleSystem

AMR AnisotropicMagnetoresistive

ANN ArtificialNeuralNetwork

ANIFS AdaptiveNetwork-BasedFuzzyInferenceSystem

AR AugmentedReality

ARM AdvancedRISCMachine

AP AccessPoint

APF ArtificialPotentialField

API ApplicationProgrammingInterface

ARMA AutoregressiveMovingAverage

ART AutonomousrailRapidTransit

ASK AmplitudeShiftKeying

ATC AutomaticTrainControl

ATO AutomaticTrainOperation

ATP AutomaticTrainProtection

ATS AutomaticTrainSupervision

AUC AreaUnderCurve

# 2D Two-Dimensional 3C Computer,Communication,ConsumerElectronic 3D Three-Dimensional A ABB
AC
ACMS
AseaBrownBoveri
AlternatingCurrent
AircraftConditionMonitoringSystem
xvii

Nomenclaturelist

B

BFS Best-FirstSearch

BP BackPropagation

BPNN BackPropagationNeuralNetwork

BRIEF BinaryRobustIndependentElementaryFeatures

C

CAD Computer-AidedDesign

CCD ChargeCoupledDevice

CCOT ContinuousConvolutionOperatorsTracker

CIMS ComputerIntegratedManufacturingSystem

CNC ComputerizedNumericalControl

CNN ConvolutionalNeuralNetwork

CNR ChinaNorthernLocomotiveRollingStockIndustryGroup

CPU CentralProcessingUnit

CRF ConditionalRandomFields

CSM CorrelationScanMatch

CSR ChinaSouthernLocomotiveRollingStockIndustryGroup

CW ContinuousWave

CWT ContinueWaveletTransform

D

DC DirectCurrent

D-DCOP DynamicDistributedConstraintOptimizationProblem

Dec-MDP DecentralizedMarkovDecisionProcess

DEM DigitalElevationModel

DFT DiscreteFourierTransform

D-H Denavit-Hartenberg

DLT DeepLearningTracker

DMPC DistributedModelPredictiveControl

DNFO DynamicNetworkFlowOptimization

DNN DeepNeuralNetwork

DOF DegreeofFreedom

DSP DigitalSignalProcessor

DTW DynamicTimeWarping

E

ECO EfficientConvolutionOperator

EKF ExtendedKalmanFilter

EKF-SLAM ExtendedKalmanFilterSLAM

ELM ExtremeLearningMachine

EMD EmpiricalModeDecomposition

EMU ElectricMultipleUnits

xviii

F

FAS FlexibleAssemblySystem

FAST FeaturesfromAcceleratedSegmentTest

FasterRCNN FasterRegionalConvolutionalNeuralNetwork

FDD FrequencyDivisionDuplexing

FFT FastFourierTransform

FMS FlexibleManufacturingSystem

FSK FrequencyShiftKeying

FTP FileTransferProtocol

G

GA GeneticAlgorithm

GOA GradeofAutomation

GPRS GeneralPacketRadioService

GPS GlobalPositioningSystem

H

HDT HedgedDeepTracking

HDFS HadoopDistributedFileSystem

HMM HiddenMarkovModel

HOG HistogramofOrientedGradient

HRI Human-RobotInteraction

HSB Hue-Saturation-Brightness

I

ID IdentityDocument

IDIM-LS InverseDynamicIdentificationModelandLinearLeastSquaresTechnique

IFF IdentificationFriendorFoe

IFR InternationalFederationofRobots

IGBT InsulatedGateBipolarTranslator

IL ImitationLearning

IMU InertialMeasurementUnit

IMW IntelligentManufacturingWorkshop

IMF IntrinsicModeFunction

IoU IntersectionoverUnion

INS InertialNavigationSystem

ISM IndustrialScientificMedical

J

JFET JunctionField-EffectTransistor

K

KCF KernelizedCorrelationFilters

KF KalmanFilter

KNN

K-NearestNeighbors

Nomenclaturelist xix

L

LAN LocalAreaNetwork

LBP LocalBinaryPattern

LCD LiquidCrystalDisplay

LiDAR LightDetectionandRanging

LRR Long-RangeRadar

M

MAE MeanAbsoluteError

MANET MobileAdHocNetwork

mAP meanAveragePrecision

MBTA MassachusettsBayTransitAuthority

MDPs MarkovDecisionProcesses

MEEM MultipleExpertsusingEntropyMinimization

MEMS Micro-Electro-MechanicalSystems

MF MorphologicalFilter

MIL MultipleInstanceLearning

MILP MixedIntegerLinearProgramming

MLP MultilayerPerceptron

MSE MeanSquareError

MTSP MultipleTravelingSalesmanProblem

MVB MultifunctionVehicleBus

N

NFS NetworkFileSystems

NMS NonmaximunSuppression

NOMA NonorthogonalMultipleAccess

NP NondeterministicPolynomial

O

OFDM OrthogonalFrequencyDivisionMultiplexing

OGA PSO OptimumGeneticAlgorithm ParticleSwarmOptimizationalgorithm

ORB OrientedFASTandRotatedBRIEF

P

PG PolicyGradient

PLC ProgrammableLogicController

POS PointofSale

PRM ProbabilisticRoadMap

PSK PhaseShiftKeying

PSO ParticleSwarmOptimization

PTZ Pan-Tilt-Zoom

PUMA ProgrammableUniversalMachineforAssembly

Nomenclaturelist xx

QPSO QuantumParticleSwarmOptimization

QR QuickResponse

R

RANSAC RandomSampleConsensus

RBF RadialBasisFunction

RBPF Rao-BlackwellizedParticleFilter

RCC RemoteCenterCompliance

RCNN RegionalConvolutionalNeuralNetwork

R CNN RegionswithCNNfeatures

RDD ResilientDistributedDataset

R FCN Region-basedFullyConvolutionalNetworks

RFID RadioFrequencyIdentification

RGB Red-Green-Blue

RGB-D Red-Green-Blue-Deep

RL ReinforcementLearning

RNN RecurrentNeuralNetwork

ROC ReceiverOperatingCharacteristiccurve

ROI RegionofInterest

ROS RobotOperatingSystem

RPN RegionProposalNetworks

RRT Rapid-ExplorationRandomTree

RSSI ReceivedSignalStrengthIndication

RTP Real-TimeProtocols

S

SARSA StateActionRewardStateAction

SCARA SelectiveCompliantAssemblyRobotArm

SDA StackedDenoisingAutoencoder

SEA SeriesElasticActuator

SfM StructurefromMotion

SIA SwarmIntelligenceAlgorithm

SIFT ScaleInvariantFeatureTransform

SLAM SimultaneousLocalizationandMapping

SMT SurfaceMountTechnology

SPP SpatialPyramidPooling

SRDCF SpatiallyRegularizedDiscriminativeCorrelationFilters

SRR Short-RangeRadar

SSD SingleShotmultiboxDetector

SURF SpeededUpRobustFeatures

SVM SupportVectorMachine

SMT SurfaceMountTechnology

Q
Nomenclaturelist xxi

T

TCN TrainCommunicationNetwork

TCP TransmissionControlProtocol

TCSN TrainControlandServiceNetwork

TDD TimeDivisionDuplexing

TEDS TroubleofmovingEMUDetectionSystem

TOF TimeofFlight

U

UAV UnmannedAerialVehicle

UDP UserDatagramProtocol

UHV UltrahighVoltage

UNECE UnitedNationsEconomicCommissionforEurope

USB UniversalSerialBus

UWB Ultrawideband

V

VR VirtualReality

VRP VehicleRoutingProblem

VSA VariableStiffnessActuator

W

WiFi WirelessFidelity

WPT WaveletPacketTransform

WTB WireTrainBus

Y

YOLO YouOnlyLookOnce

xxii
Nomenclaturelist

Introduction

1.1Overviewofrailtransitrobots

Railwaytransitisvitaltothenationaleconomy.ResearchinJapanandChinaindicates thatdevelopmentoftherailwaycanleadtoeconomicgrowthintherailwayfield[1,2]. Tostimulateeconomicdevelopment,governmentsareactivelydevelopingtherailway transportindustry.Insuchalargeindustrialsystem,thedevelopmentofautomationcan improveefficiencyandreducecosts,sothelevelofautomationinrailwaysshouldbe increased.Theincreaseinautomationrequirementsintherailtransitsystemgeneratesa pursuitforrailtransitrobotsystems.

Inworldwide,manycountrieshaveproposedsimilarstrategicplanstoencouragethe developmentofrobotsintherailwaytransitsystem.TakingChinaasanexample,“A CountryWithaStrongTransportationNetwork”and“SmartRailway”aretwoimportant plansthatencouragethedevelopmentofautomationoftherailtransitsystem.Drivenby theseplans,manyChineserailequipmentmanufacturersandoperatorshavecarriedout extensiveresearchinrobotics.ZhuzhouCRRCTimesElectricCo.,Ltd.appliedan automaticgeneratinglinetoproducehigh-speedtrainconverters[3].CRRCQishuyan InstituteCo.,Ltd.developedanintelligentmanufacturingworkshopforgeartransmission systemsforhigh-speedtrains.CRRCZhuzhouInstituteCo.,Ltd.combinedautomatic guidedvehicle(AGV)technologywithurbantransportationequipmentanddesigned autonomousrailrapidtransit(ART)[4].

Avarietyofrobotsystemsareemployed.Theuseofrobotsystemsintherailtransit systemcanbedividedintothreeaspects:manufacturing,dispatch,andmaintenance.In thesethreepartsofrailtransit,differentkindsofrobotshavecompletelydifferentroles,as shownin Fig.1.1.

1.1.1Railtransitrobotsinmanufacturing

“ACountryWithaStrongTransportationNetwork”pointsoutthatintelligentmanufacturing isrequiredfortherailtransitsystem.Inthemodernproductionline,robotscangreatly improveprocessingefficiency.Takingtheproductionofahigh-speedtraingearboxasan RobotSystemsforRailTransitApplications. https://doi.org/10.1016/B978-0-12-822968-2.00001-2

CHAPTER1
©
1
Copyright
2020ElsevierInc.Allrightsreserved.

example,intheprocessflowofatransmissiongear,robotarmscanresultintheefficient transmissionofgearsbetweendifferentmachinetools;whenweldingthegearbox,the weldingrobotscanimprovemachiningefficiency;whenassemblingthegearbox,assembly robotscanworkwithhigh-precision;whenthetrainisassembled,AGVenablesthegearbox tobetransportedquicklybetweenworkshops.Applyingrobotsinmanufacturingnotonly savesprocessingtimeandlaborcosts,itimprovesmanufacturingquality.

1.1.1.1Assemblyrobots

Asforassemblyrobots,theprimarymissionistoachievehigh-precisionpositioningofthe workpiece.Accordingtopreviousresearch,assemblycostsaccountfor50%oftotal manufacturingcosts[5].Assemblyrobotsystemscanalsobedividedintorigidassembly andflexibleassemblyrobots.Rigidassemblyrobotsarecustomizedprocessingsystemsfor specificworkpiecesinthetraditionalindustrialenvironment.

Rigidassemblyrobotshavepoorgeneralization.Iftheproductionlineisreplacedwith processedparts,theequipmentneedstobecustomized.Replacementofequipmentwill causeagreateconomicburden.Comparedwithrigidassemblyrobots,flexibleassembly robotscandesigncustomizedprocessingprogramsaccordingtotheworkpiece.Flexible assemblyrobotsareprogrammable,whichcanresultindifferentassemblyschemesfor differentworkpieces.Flexibleassemblyrobotsaresignificantforaflexibleassemblysystem. Incurrentindustrialdevelopment,flexibleassemblyrobotsarethefocusofdevelopment[6]. Inthefollowingdiscussion,assemblyrobotsrefertoflexibleassemblyrobots.

Figure1.1
2Chapter1
Robotsinthemanufacturing,dispatch,andmaintenanceofrailtransit.

Anassemblyrobotconsistsoffourcomponents:machinerycomponents,sensors, controllers,andactuators.Tobringaboutacomplexworkpiecetrackintherealassembly environment,assemblyrobotsusuallyhavemorethanfourdegreesoffreedom(DOFs).

Mainstreamassemblyrobotscanbedividedintotwotypes:selectivecompliantassembly robotarms(SCARAs)andsix-DOFrobots.

SCARAshavefourDOFs,whicharecommonlyusedinelectronicassembly,screw assembly,andsoon[7].SCARAsarespeciallydesignedforassemblyapplicationsby YamanashiUniversity.SCARAscontaintwoparalleljoints,whichcanassemblea workpieceinaspecifiedplane.Comparedwithsix-DOFrobots,advantagesofSCARAs areahigherassemblyspeedandprecision;disadvantagesarelimitedworkspace. CommonlyusedcontrolstrategiesforSCARAscontainadaptivecontrol,forcecontrol, robustcontrol,andsoforth[8].Instate-of-the-artresearchonrobotcontrol,intelligent algorithmsareemployedtoimprovecontrolperformance[9].Dulgeretal.applieda neuralnetworktocontroltheSCARA[10].Theneuralnetworkwasoptimizedbyparticle swarmoptimizationtoimproveperformance.Sonetal.adoptedanoptimizedinverse neuralnetworkforfeedbackcontrol[11].Todealwithdisturbancesinrunning,the parametersoftheinverseneuralnetworkareupdatedbyabackpropagationalgorithm. Luanetal.usedtheradialbasisfunction(RBF)neuralnetworktoachievedynamic controloftheSCARA[12].

Six-DOFrobotscanlocatetheworkpieceatalmostanypoint.Thus,six-DOFrobotscan handletheassemblytaskofcomplexthree-dimensional(3D)workpieces.Thedynamics ofsix-DOFrobotsarebasicforoperatingther obots.Zhangetal.consideredthefriction oftherobotsandusedahybridoptimizationmethodtomodelthedynamicsofthe six-DOFrobot[ 13].Afteroptimization,dynamicaccuracyincreasedsignificantly.Yang etal.proposedasimulatorforthedynamicsofthesix-DOFrobot[14 ].Robotswitha largedegreeoffreedomhavelargefeasibility.However,toomuchfreedomis uneconomical.Tohandlethetrade-offbetweeneconomyandfeasibility,theDOFcanbe optimizedforspecifictasks.Yangetal.proposedanoptimizationmethodtominimize theDOF[ 15 ].ThisoptimizationmethodcanreducetheDOFandimprovetheuseofthe DOF.

Assemblyrobotsshouldcooperatewiththeancillaryequipment.Thefixturesarevital equipmenttoensurecooperationinperformance.Thefixturescanfixtherelativeposition betweentheworkpieceandtherobotunderload.Iftheprecisionofthelocationofthe fixturesislow,nomatterhowaccuratethepositioningprecisionoftherobotis,itcannot achievehigh-precisionassembly.Currently,theflexiblefixtureisafuturedevelopment [16].Lowthetal.proposedauniquefixturethatcanadjusttheradialandangular adaptively[17].Althoughauxiliarydevicesareappliedforassemblyrobots,theresultsof assemblyrobotsmaystillbeunsuccessful.Avoidingunsuccessfulassemblyisparticularly

Introduction3

importantinelectricconnectorassembly,becausetheelectricconnectorisnotarigid component.Todetecttheunsuccessfulassemblyoftheelectricconnectorassembly,Di etal.proposedahybriddetectionsystemwithaforcesensorandcamera[18].

Thefaultdiagnosisandprognosissystemofassemblyrobotsguaranteesassembly accuracy.Therearemanystudiesaboutfaultdiagnosisandprognosissystems.Huang etal.designedaclassifierforthewiringharnessrobot[19].Thatstudymodeledthe manufacturingprocesswasandcalculatedthefaultwithafuzzymodel.Baydaretal. introducedadiagnosismodelwitherrorprediction[20].Theproposedmodelintegrated theMonteCarlosimulation,geneticalgorithm,andsoforth.Thefunctionsofthefault diagnosisandprognosissystemforassemblyrobotsshouldcontainthemainaspectsas giveninChooetal.[21]:

(a)Thehealthstatesoftheassemblyrobotsaremonitoredinrealtime.Themonitored dataareloggedintothedataset.Thehealthfeaturesareextractedfromthehealth statesoftheassemblyrobots.Thefaultsandremainingusefullifecanbecalculated accordingtothefeatures.

(b)Accordingtothefaultdiagnosisandprognosisresults,theassemblytasksarereassignedtomakesurethefailedassemblyrobotsarereplacedbythefullyfunctioning robots.Themaintenanceplanscanbemadetorepairthefailedrobots.

1.1.1.2Collaborativerobots

Whenapplyingtheresultingclassicalindustrialrobotsystems,interactionbetweenrobots andhumansislimited.Therearethreereasonsforthisphenomenon:

(a)Traditionalindustrialrobotsdonotconsidermovinghumans.Ifacollisionoccursalong acertaintrajectory,itmaycausegreatdamagetohumans.Therefore,theworkingarea oftheindustrialrobotismostlyseparatedfromtheworkingareaofthehuman.

(b)Theweightandvolumeoftraditionalindustrialrobotsarelarge,anditisdifficultfor humanstooperaterobots.

(c)Reprogrammingofrobotsisdifficultandrequiresspecialprogrammingtoolsfor tuning.

However,human robotcollaborationcancombinehumancreativitywiththeefficiencyof robotsandcanamplifytheflexibilityofrobotsandfurtherimproveworkefficiency.Under thisdemand,collaborativerobotsareborn.Comparedwithtraditionalrobots,collaborative robotshavethreemainadvantages:safety,easeofoperation,andeaseofteaching.Some robotmanufacturershavelaunchedcollaborativeroboticsproducts.Therobotcompany UniversalRobotlaunchedtheUR3collaborativerobot[22].Thisrobotisthefirsttruly collaborativerobot.TheUR3collaborativerobotisbasedonasix-DOFroboticarmandis flexibleenoughtoachievecomplexmotiontrajectories.Intermsofsafety,theUR3

4Chapter1

collaborativerobothasacollisionmonitoringsystemthatprotectshumansafetyby monitoringthejointposition,speed,andpoweroftherobot.

KUKARoboticshaslaunchedacollaborativerobot,theLBRiiwa[22].Therobot’ssevenDOFdesignprovidesgreaterflexibilitythantraditionalsix-DOFrobots,enablingmore complextrajectoriestocopewithcomplexenvironmentsthatworkwithhumans.The shapeoftherobotisdesignedtobeergonomicandeasyforhumanstooperate.Theouter casingismadeofaluminumalloy,whichcanreduceweightandimproveoperability.The robotisequippedwithtorquesensorsateachjointtomonitorcollisionsinrealtime.The teachingmethodoftherobotisdragging,whichreducesthetechnicalthresholdofthe robotoperator.

ABBlaunchedtherobotYuMi[22].Therobotishighlysafeandcanachieve human robotinteractioninasmallspace.Toimprovetheperformanceofcollaborative robots,AABacquiredGomtecRobotics,whichlaunchedthecollaborativerobotRoberta [22].TheRobertacanhandlehigherloadapplicationscomparedwithYuMi.

FrankaEmikalaunchedtheFrankaCollaboration[23].LiketheLBRiiwa,therobothas sevenDOFs.Itisalsoequippedwithatorquesensoroneachjointtoenablecollision monitoring.

RethinkRoboticslaunchedthetwo-armcollaborativerobotBaxterandtheone-arm collaborativerobotSawyer[24].Thetworobotsareexquisitelydesignedwithhigh positioningaccuracyandcanbeassembledwithhighprecision.

Safetyinhuman robotinteractionisessentialforcollaborativerobots.Threatstothe safetyofcollaborativerobotscanbedividedintotwoaspects[25]:

(a)Thefirstkindofthreatisfromrobots.Duringoperation,therobotmaycollidewith workersandcauseinjuries.Toensurethesafetyofemployees,therobotneedsto detectthelocationofworkersinrealtimeanddeterminewhetherthelocationof workersisinasafeposition.Ifworkersintrudeonthesafearea,therobotshould immediatelystoptoavoidacollision.Incaseacollisionbetweenapersonandarobot occurs,therobotneedstodetectthecollisionintimeandchangetorquetominimize damagetoworkers.Mohammedetal.proposedacollisionavoidancesystemfor collaborativerobots[26].Thissystemuseddepthvisionsensorstodetecttheposition oftheworker.Consideringthevirtualmodelofthecollaborativerobot,acollision couldbedetected.Thecollaborativerobotcouldtakemeasurestoavoidcollisions.In additiontocollisiondetection,itisnecessarytoensuretheintegrityofthecollaborativerobotcontrolsystemduringoperation.Failureofanypartofthesensors,controllers,oractuatorswillleadtothefailureofhuman computerinteraction,thus threateningthesafetyofworkers.Inaddition,interactionwithrobotsmaycause mentalstresstoworkers[27],whichwillincreasetheriskforacollision.

Introduction5

(b)Thesecondkindofthreatisfromtheindustrialprocess.Intheprocessofhuman robotinteraction,workersneedclosecontactwiththemanufacturingprocess.The temperatureofthemanufacturedworkpiecescancausedamagetoworkers.Fallen workpiecescanalsoendangerworkers.Therefore,itisnecessarytofullyconsiderthe impactofmachiningpartsonworkersinthedesignprocessofcollaborativerobots.In addition,theunreasonableergonomicdesignofthecollaborativerobotduringmaintenancewillhaveanimpactonsafety.

1.1.1.3Automaticguidedvehicles

Accordingtothelevelofautomation,manufacturingsystemscanbedividedintothree levels[28].Manufacturingsystemsinthefirstlevelaremanual.Thoseinthesecondlevel havesmall-scaleautomatedmanufacturinginwhichtransportationiscarriedoutmanually. Manufacturingsystemsinthethirdlevelhavelarge-scaleautomatedmanufacturingand useautomatedtransportation.Manufacturingsystemsinthethirdlevelarealsocalled flexiblemanufacturingsystems(FMSs).AccordingtotheMaterialHandlingIndustryof America,only20%ofthetimeisspentonprocessingandmanufacturing;theremaining 80%isusedforstorage,handling,waitingforprocessing,andtransportation[29].As factoryautomationincreases,transportationefficiencybetweenworkstationsneedsto improve.IntheFMS,theAGVcanimprovetheuseofspaceinthefactoryandthe efficiencyoftransportationinthematerialhandlingsystem.Therefore,transportationcosts canbereduced.

TherailtransitmanufacturingsystemisatypicalFMS.Transportationisanimportantpart oftherailtransitmanufacturingsystem.Intherailtransitprocessingenvironment,not onlythetransferofworkpiecesbetweenprocesseswithintheplantbutalsothefreeflow ofworkpiecesbetweenplantsisrequired.Inthetraditionalrailtransitmanufacturing environment,transportationinsidethefactoryiscarriedoutbygantrycranesandthe transportationbetweenfactoriesiscarriedoutbytrucks.Thesemodesoftransportation havesomedisadvantages.Thereisasafetyhazardwhenusinggantrycranestolift.Ifthe liftingworkpiecefalls,itmaycauseserioussafetyaccidents.Theuseoftruckscarriesa highercostandislessefficienttotransport.Improvingthesafetyandeconomyofproducts inthetransportationprocessisimportantforthedevelopmentoftherailtransitindustry.In thecurrentmanufacturingenvironment,theAGVisaneffectivemodeoftransportation. TheAGVissafercomparedwithgantrycranesandismoreautomatedandmoreefficient fortransportationthantrucks.

ThetypicalstructureofanAGVconsistsofsensors,chassis,acontrolunit,andsoon [30,31].Duringwork,thesensorcandeterminethepositionoftheAGVandtransmitthe currentpositiontothecontroldecisionsystem,andthecontroldecisionsystemplansan

6Chapter1

optimalpath.Thechassisisdrivenbythecontroller’scontrolcommandtotransportthe workpiecetothedesignatedposition.

Toimprovetransportationperformance,themodelpredictivecontroloftheAGVsshould beachieved.TheremainingpowerandtheworkingstateoftheAGVshouldbepredicted tocalculatetheremaininglifeoftheAGV,andthedispatchplanoftheAGVcanbe optimizedwithconsiderationofthesefactorstoimprovetheoperationalsafetyofthe AGV.Populardata-drivenforecastingmethodscontainstatisticalmethods,intelligent methods,andhybridmethods.Thestatisticalmethodscandiscoverthestatisticalruleof thedataandgenerateanexplicitequationforprediction.Commonlyusedstatistical methodscontainautoregressivemovingaverage,Winnerprocess,Gaussianprocess,andso on.Intelligentmethodscangeneratebetterforecastingperformancethanstatistical methodswiththehelpofthestronglyfittingcapacityoftheneuralnetwork.TheElman neuralnetwork,multilayerperceptron,andextremelearningmachine(ELM)arethethree mostpopularintelligentmethods.However,thetrainingprocessoftheseneuralnetworks dependsontheinitialvaluesinsomeway.Iftheinitialvalueisunsuitable,trainingofthe neuralnetworkmaystopatthelocallyoptimalsolution.Toimprovetheperformanceof intelligentpredictionmethods,theinitialvaluescanbeoptimizedbyoptimization algorithms[32].Hybridpredictionmethodscombinedataprocessingalgorithmswith statisticalorintelligentmethods.Thedecompositionalgorithmsareprovedtobeeffective [33].Thedecompositionalgorithmscandividetherawseriesintoseveralmorestationary subseries.Eachgroupofsubserieshasasimplerfluctuationmodethantherawseries,soit ismorepredictable.

Afteroptimization,thecontrolcommandcanbeassignedtotheAGVsintwodifferent ways:staticcontrolanddynamiccontrol[34]:

(a)Staticcontrol.Thecontrolcommandsareassignedbeforethetask.OncetheAGVs receivethecontrolcommand,thetransportationpathwillnotbechangeduntilthe AGVsreceiveanothercontrolcommand.Thiscontrolschemeissimpleandeasyto operate.However,flexibilityisweak.

(b)Dynamiccontrol.Thiscontrolmethodcanadjustthecontrolcommandsaccordingto thereal-timestateoftheAGVs,sothetaskschedulingstrategyiscomplex.

Themulti-AGVsystemisbeingstudiedworldwide.ComparedwiththesingleAGV,the advantagesofthemulti-AGVsystemare:

(a)Themulti-AGVsystemcancoveralargearea.ThesingleAGVcanachievetransportationonlybetweenpoints.Inthemodernmanufacturingenvironment,the transportationtaskisfarmorecomplexthanthepoint-to-pointtransportation.The multi-AGVsystemcanbuildatransportationnetworkandimprovetransportation efficiency.

Introduction7

(b)Themulti-AGVsystemcanexecutethetransportationtaskinparallel.Amulti-AGV systemcanperformtaskssimultaneouslyforacomplextask.Thus,themulti-AGV systemcangreatlyimprovetheefficiencyoftransportation.

Thedispatchandroutingofthemulti-AGVareimportant.Theconflict-freefunctionofthe multipleAGVsisthebottleneckforthemulti-AGVsystem.Draganjacetal.proposeda controlalgorithmforthemulti-AGVsystem[35].Theproposedalgorithmcandetectthe conflictbetweentheAGVsandguaranteethesafeoperationofthemulti-AGVbythe prioritymechanism.Miyamotoetal.proposedaconflict-freeroutingalgorithmforthe multi-AGVsystem[36].BecauseofthelimitedmemoryspaceofeachAGV,aheuristic algorithmwasadoptedforrouting.Małopolskietal.consideredthetransportationsystem inthefactoryasacombinationofsquares[37].Basedonthesquaretopology,anovel conflict-freeroutingalgorithmforthemulti-AGVsystemwasproposed.

Thedispatchplanofthemulti-AGVshouldbecalculatedtooptimizethetaskwaiting time,collision,loaduse,andsoforth[38].Theoptimizationmethodscanbedividedinto single-andmultiple-objectiveoptimization.Single-objectiveoptimizationcanoptimize onlyoneobjectivefunction.Iftheobjectivefunctionconsistsofseveralobjective subfunctions,thesesubfunctionsshouldbecombinedastheweightedsum[39].However, itisdifficulttodesigntheseweights.Therefore,thegeneratedoptimizationresultsmight notbetheglobaloptimalsolution.Themultiple-objectiveoptimizationcanbalancethe trade-offbetweendifferentsubfunctionsandgenerateaParetofront[40].TheParetofront containsmanysolutions,eachofwhichhasbothadvantagesanddisadvantages.Thefinal optimizationresultsshouldbeselectedaccordingtotheexpert.Inthismanner,the intelligentdecision-makingabilityofhumanscanbeusedtoimprovethedispatch performanceoftheAGV.

Theenvironmentofthefactoryisdynamic.TheAGVsshouldbeflexibleenoughtocope withthedynamicmanufacturingsystem.Therearemanystudiesonthedynamic transportationsystem.Britoetal.proposedadynamicobstacleavoidancealgorithmfor thedynamicunstructuredenvironment[41].Inthisalgorithm,modelpredictivecontrolis appliedtoimprovecontrolperformance.Thisalgorithmwasverifiedintheenvironment withwalkinghumans.Lietal.proposedanintegratedalgorithmforobstacleavoidance [42].ThisalgorithmcangenerateapathfortheAGVbyamodel-predictivealgorithm. TheAGVcantrackthegeneratedpathreliably.

1.1.1.4Manufacturingrobots

Manufacturingrobotscontainmanytypesincludingweldingrobots,drillingrobots, grindingrobots,millingrobots,andsoon.Manufacturingrobotscanproducebetter machiningperformancethanaclassicalcomputerizednumericalcontrol(CNC)machine.

8Chapter1

Forexample,itisproventhataworkpiecepolishedbyamanufacturingrobothasabetter surfacequalitythananyCNCmachine[43].Thebetterperformanceofthemanufacturing robotsisbecausetherobotshavemoreflexibilitytomakesurethetoolisintheright position.

Weldingrobotsareoneofthemostwidelyusedmanufacturingrobots.Difficultiesof weldingrobotsare[44]:(1)itishardtoobservetheweldingseaminacomplex manufacturingtask,(b)itishardtoobtaintheabsoluteandrelativelocationsofthe workpieces,and(c)thetrajectoryishardtotrack.Currentcommonlyusedlocation methodsfortheweldingseamarebasedonopticalsensorssuchasadepthcameraand lasersensor[45,46].Jiaetal.proposedaweldingseamlocationmethodandtrajectory trackingalgorithm[44].Theproposedlocationmethodwasachievedusingalaserscanner, whichcouldobtainthelocationanddirectionofthepipe.Thecubicsplinewasappliedto fittheobtainedweldingseam.Thevelocitycontrolwasusedtotrackthewelding trajectory.Liuetal.proposedatrajectoryplanningalgorithmforweldingrobotstocope withasingleY-grooveweldingtask[47].Thisstudyprovidedtwodifferentvelocity planningalgorithms.

Akeyconcernwithdrillingrobotsispositioningaccuracy.Aninaccurateholecanreduce themechanicalperformanceoftheequipment.Thepositioningcompensationmethodcan bedividedintotwotypes:model-basedandmodel-free[48].Model-basedmethodscan guidetherobottomoveaccordingtothemeasuredpositioningerror.Theessentialsof model-basedmethodsarethedynamicsandkinematicsofthedrillingrobots.Model-based methodsaretime-consuming.Model-freemethodscansolvethisdrawback.Model-free methodsbuildamodeltodescribetherelationshipbetweenthepositioningerrorandthe robot’sjoints’parameters,whichdonotconsiderthedynamicsandkinematicsofthe robots.Commonlyusedmodel-freemethodscontaintheinterpolationmethod[49], cokrigingmethod[50],andsoon.Neuralnetworksareappliedtocompensateintelligently forpositioning.Withthehelpofthestronglyfittingcapacityofneuralnetworks,these intelligentpositioningcompensationmethodsobtaingoodperformance.Yuanetal.used ELMtopredictpositioningerrorandguidetherobottocompensateforit[51].Chenetal. usedtheRBFneuralnetworktoestimatepositioningerror.Thebandwidthoftheadopted RBFneuralwasfine-tuned.Positioningaccuracycanbeimprovedbymorethan80%with theseintelligentpositioningcompensationmethods[52].

Thegrindingmanufacturingishighlyprecise,sothepositioningaccuracyofthegrinding robotsrequiresattention.Inrealapplications,thegrindingrobotsmaydeviatefromthe presettrackbecauseofthedisturbanceandweartothetools.Thus,grindingrobotsshould havetheabilitytoself-adjust,toensuremanufacturingquality.Controlofgrindingrobots ismoredifficultthanforCNCsbecausegrindingrobotshavemoreDOFs.Huangetal. proposedanintelligentgeargrindingsystem[53]inwhichtherobotarmcandetectthe

Introduction9

actualtrajectorybyavisionsensorandadjustthetrajectoryadaptively.Twocamerasare adoptedfortoolcenterpointcalibration.Whencopingwiththecomplexgrindingtask, multiplegrindingrobotsarenecessary,becausemultiplegrindingrobotscancopewiththe complexmanufacturingtaskmoreeasilythanasinglegrindingrobot.Hanetal.[54] proposedamultiplegrindingrobotsystemandintroducedthetrajectoryplanningmethod forthemultiple-robotsystem.Experimentalstudiesindicatedthatthemultiplegrinding robotsystemcangenerateasteadiermanufacturingtrajectory.

TherobotonlyneedsfiveDOFsforthemillingtask;thereservedoneistheDOFofthe spindleofthemillingcutter[55].Intherealapplication,six-DOFrobotsareappliedfor millingtoimproveflexibility.Thestiffnessofmillingrobotsisachallengein manufacturing.Millingrobotsusuallyhavelowstiffness.Robotsmaydeformorvibrate whenmillingworkpieces.Therehavebeenmanystudiesonimprovingmilling performance.Pengetal.optimizedthestiffnessoftherobotsinthefeeddirectionwitha seven-DOFrobot[56].

Commonlyusedmanufacturingrobotsaredevelopedfromsix-DOFrobots.Trajectory planningisacommontechnologyformanufacturingrobots.Thetrajectoryplanning algorithmcangeneratetheoptimaltrajectoryandtherobotcontrolalgorithmcandrivethe robottofollowapredeterminedtrajectory[57].Theoptimizationtargetsofthetrajectory aretheposition,velocity,acceleratedvelocityofeachjoint.Thetrajectoryplanning algorithmisthebasisoftheindustrialrobotcontrolsystem.Theoptimizationfunctionsof thetrajectorycontainmanyaspects,includingreduceexecutiontime,consumedenergy, andimpact.

1.1.1.5Loading unloadingrobots

Intheexistingtransportationmanufacturingenvironment,processingeachworkpiece requiresdifferentprocedures.Theconnectionbetweenprocessesrequirestheworkpieceto bemovedfromonemachinetoanother[58].Thisrepetitiveworkcanbereplacedby loading unloadingrobots.Theapplicationofloading unloadingrobotshasmany benefits.Ontheonehand,loadingandunloadingofrobotsisrepetitiveandconsistent, whichcanavoidadecreaseinworkefficiencycausedbyworkerfatigue.Ontheother hand,usingtherobotscanparameterizetheloadingandunloadingoperationandthefault sourcecanbequicklylocatedwhentroubleshootingproblems,whichisconvenientfor improvingtheprocessingquality.

Loading unloadingrobotshaveawiderangeofapplications.Liuetal.adopteda loading unloadingrobotfortheCNCanddesignedacontrolsystembasedona programmablelogiccontroller[59].Zhangetal.developedarobotsystemfor loading unloading[60].Thedesignedrobotwasabletoseparategoodandbad workpieces.Fanetal.usedtheloading unloadingrobotforelectricequipmentdetection

10Chapter1

[61].Withthehelpoftheloading unloadingrobot,automaticandhigh-precision detectionisachieved.

Thedevelopmentofloading unloadingrobotsshowsatrendinintelligence.Computer visionandneuralnetworktechnologyhavebeenwidelyapplied[62].Theadoptionof computervisioncanincreasetheflexibilityofloading unloadingrobots.

Loading unloadingrobotscanobservethepostureoftheworkpieceandadjusttherobots tograsptheworkpiecefromtheappropriateposition.Theneuralnetworkiswidelyusedin controllerdesignwithitsstronggeneralizationability.Guetal.proposedavisualservo loading unloadingrobot[63].Thecameraonthemanipulatorcandetectthelocationof theworkpieceandhelptherobotcapturetheworkpieceprecisely.Thedesigncontainstwo steps:(1)extractthefeaturesoftheobservedimageandtargetimageandcalculatethe errorbetweenthesefeatures;and(2)inputtheerrorintotheneuralnetworkandoutputthe controlcommandfortherobot.Intheproposedrobotcontrolstructure,thefuzzyneural networkwasadoptedasthecontroller.

1.1.2Railtransitrobotsindispatch

Therunningstageoftherailtransitsystemtakesupthelongestlengthandgeneratesthe highestcostinthewholelifecycleoftherailtransitsystem.Withthehelpofrobot technology,therailtransitsystemcanachieveautomaticoperation,thusimproving operationalsafetyandreducingtransportationcosts.Autonomousdrivingisthemost importantformofautomationinrailtransit.Autonomousdrivingistheapplicationof robotcontroltechnologytotherailwaytrain.Theautomatictraincontroloftherailway traincontainsautomatictrainsupervision(ATS),automatictrainprotection(ATP),and automatictrainoperation(ATO)[64].TheATScansupervisetherunningstatesofthe railwaytrain.TheATPsystemcanmonitorthetrain’srunningpositionandobtainits speedlimittoensureitsrunningintervalandsafety.TheATOcanacceptoutput informationoftheATSandATPandgeneratecontrolinstructions[65].

TheinternationalstandardInternationalElectrotechnicalCommission62,290definesfour gradesofautomation(GOA)[66]:

(a)GOA1.Thetrainisabletomonitorthetrain’soperatingstatuscontinuously.Operation needstobecarriedoutbydrivers.

(b)GOA2.Thislevelisself-drivingwithdriverduty.Thetrainisabletodriveautomaticallythroughthesignalsystem,butthedriverisrequiredtoclosethedoorandissue thecommand.

(c)GOA3.Thereisnoneedtoequipdriversonthislevelofthetrain;thetraincanautomaticallycompletethewholeprocessofoperation,includingoutbound,pitstop,

Introduction11

switchdoor,andsoon.However,thereisstillaneedforonboardpersonneltodeal withemergenciesonthisleveloftrain.

(d)GOA4.Nooperatorisrequiredonthisclassoftrains,andthetrain’scontrolsystem canautomaticallyrespondtounexpectedsituations.

ARTisanimportantexampleoftheautomationofrailtransitoperation;itisdevelopedby CRRCZhuzhouInstituteCo.,Ltd.Thecoretechnologyofthistrainisvirtualorbit trackingcontroltechnology.Thatistosay,theARTtraindoesnotneedatruetrack,but rather,themarksontheroadserveastracks.Becausetheintelligentrailtraindoesnot haveaphysicaltrack,existingpositioningdevicesusedinthesubwaytrainandhigh-speed traincannotbeused.TheARTadoptstheglobalsatellitepositioningsystemasthemain positioningmethod[4].Duringtheoperation,inertialsensorsandanglesensorsinstalled inintelligenttrainsareusedtomonitorthemovementpostureandpositionofthetrain, therebyenablingacomprehensiveperceptionofthestateoftheARTtrain.Themonitored trainstatusinformationisfedbacktothecontrolleroftheintelligentrailtraintocontrol theintelligentrailtraintorunalongthevirtualtrack.Usingthiscontroltechnology,all wheelsoftheARTtraincanbedrivenalongavirtualtrack[4].Thispositionaccuracycan ensurethepassingperformanceoftheARTrailtraininurbantraffic.Toimprove operationalefficiency,theARTtrainisalsoequippedwithautomaticdrivingtechnology [4].TheARTtraincanintelligentlysensetheenvironmentandcontroltheautomatic operationofthetrain.

1.1.3Railtransitrobotsinmaintenance

Robotshavebecomethefocusofmodernrailtransitequipmentmanufacturingandoperation. Withtheincreaseinrailwayoperatingmileageandoperationaldensity,theworkloadof railwaymaintenancesupportisalsoincreasing,andhigherrequirementsareputforwardfor themaintenanceandrepairofrailwayinfrastructureequipment.Currently,themaintenanceof railwayinfrastructureequipmentismainlycarriedoutbyworkers[67].Themainproblems

Figure1.2 Roleofrobotsinrailtransitmaintenance.
12Chapter1

arethat(1)diversifiedtestoperationshavehigherrequirementsforoperators,(2)itis inconvenienttohaveoperatorscarrymeasurementequipment,and(3)manualinspections maintainlaborintensityandqualityisuncontrollable.Asshownin Fig.1.2,robotscanhave animportantroleinthemaintenancestageofrailtransitinthiscase.

1.1.3.1Inspectionrobots

Theobjectivesofrailwaymaintenanceincludetracks,instruments,andequipmentnextto railways,tractionsubstations,pantographs,temperatureofinstruments,bogies,andforeign matterintrusion.Accordingtotheworkscenario,theinspectionareaofrailtransit inspectionrobotsismainlydividedintotractionsubstationsandrailway.

1.1.3.1.1Inspectionintractionsubstation

Thetractionsubstationconvertselectricenergyfromtheregionalpowersysteminto electricenergysuitableforelectrictractionaccordingtothedifferentrequirementsof electrictractiononcurrentandvoltage.Theresultingelectricityisthensenttocatenary linessetupalongtherailwaylinetopowerelectriclocomotives,ortotheurbanrailtransit systemtopowerelectricsubwayvehicles.Theinspectionworkisthetoppriorityforthe dailymaintenanceofthetractionsubstation.However,becauseofinsufficientstaffingand otherreasons,theinspectioncannotmeettheneeds.Inaddition,withtheincreasein substations,operation,andmaintenance,personnelarestrugglingtocomply.Itisalso difficulttokeepsatisfactoryworkingconditionsandastrongsenseofresponsibility.

Currently,thetechnologyofintelligentrobotsisincreasinglydevelopingandtheuseof robotstoreplacemanualinspectionsofequipmentpartiallyorcompletelyhasbecomethe trendoffuturesubstations[68].Substationintelligentinspectionrobotsareequippedwith advancedequipmentsuchasinfraredthermometers,high-definitioncameras,andaudio signalreceivers.Accordingtothepresetinspectiontasks,theentirestationequipmentand partitionsareinspectedintime.Alltaskmodulesofmeterreading,temperature measurement,soundcollection,andfunctionalinspectionareprocessedtogetherto identifyandalertregardingdiscoveredequipmentabnormalities,defects,andhidden dangersinatimelymanner.Inaddition,robotsintuitivelyformchartsandreportweekly onoperationaldataindicatorsofdevicesstoredintheinspectionsothatmaintenance personnelcanperformequipmentmaintenanceandfaultanalysisbetter.

Byusingtheinspectionrobotsinthesubstation,theequipmentinthesubstationcanbe inspectedmorereliablyandthedataanalysisreturnedbythesystemcangivethe maintenancepersonnelabetterunderstandingoftheworkingstateoftheequipment.

Recordingandanalyzinghistoricaldatacanalsosignificantlyimprovethepredictabilityof equipmentinthesubstation,providingmoreeffectivedatasupportfortheconditionof maintenanceandastatusassessmentoftheequipment,whichwilldirectlyreducethe possibilityofequipmentaccidents[69].Usinginspectionrobotstoinspectsubstation

Introduction13

equipmentcannotonlykeepfieldmaintenancepersonnelawayfromthefieldequipment, whichimprovesthesafetyandreliabilityoftheinspectionwork,itprovidesanewmethod forspotinspectionforthepopularizationofanunattendedsubstationandtrulyimproves thesafetyandreliabilityofsubstationequipment.

1.1.3.1.2Inspectionalongtherailway

Withtherapiddevelopmentofrailtransitandthecontinuousimprovementinpassenger high-speedprocesses,railwaytransporthasbeengraduallyheadingtowardfunction integration,informationsharing,commandconcentration,andhighlyautomatedtransition andtransformation.Theurgentdemandforrailwaytransportationsafetyposesnew challengestotherailwaytrafficsafetyassurancesystem.Toadapttothenewneedsof railwaydevelopment,ithasbecomethefocusoftheconstructionoftherailwaytraffic safetyassurancesystemtobuildafullycoveredandhighlyreliablerailwayintelligent monitoringsystembasedonhighinformationsharing[70].Thehigherspeedbrings conveniencetopeople,butitalsoincreasesthedifficultyofpreventingrailwaydisasters.

Althoughahigh-speedtrainisequippedwithamoreadvancedsafetysystem,thelarge brakingdistancecausedbyhigh-speeddrivingreducestheflexibilityofthetrainto respondtoemergencies.Intheeventofanaccident,ahigh-speedtrainoftenhasmore seriousconsequencesthananordinarytrain.Therefore,variousfactorssuchasthe intrusionofforeignmatter,naturaldisasters,andlinefailuresmaycausemajorrailway accidents.Casualtiesandpropertylosscausedbytheintrusionareenormousand frequentlyoccur.Therefore,researchanddevelopmentintorailwayforeignmatter intrusiondetectionareofgreatsignificancetoensurethesafeoperationoftrains[71].

Theintrusionofforeignmatterhasthecharacteristicsofirregularity,suddenness,and unpredictability.Thefixedtransportationmodeoftheoperationroutemakesemergency measuresthatcanbeadoptedlimitedwhenthetraindetectsanintrusionontheroad.If real-timedetectionofinvasiveforeignmattercanbeachievedandnecessarymeasuresare takenintime,propertylosscanbereducedasmuchaspossible.Thetraditionalforeign matterintrusionapproachistoplacesensorsatkeypositionsalongtherailwaytomonitor intrusionconditionsinreal-time.Thismethodhasmanydrawbacks,suchasan unreasonablearrangementofthesensorandfixedmonitoringposition.Withtherapid developmentofrobotics,robotshaveenteredthefieldofrailwayforeignmatterinvasion detection.Therailwayisinspectedbyaninfraredsensor,alasersensor,asmartcamera, andanultrasonicdevice.

1.1.3.2Channelrobots

High-speedtrainsmayexperiencepoorperformanceorevenfailureofkeycomponents duringlong-termhigh-speedservice.Ifasafetyaccidentoccurs,itwillcausesevere economiclossandevencasualties.Therefore,howtomonitoranddiagnoseitsoperating

14Chapter1

statushasbecomeanimportantresearchtopic[72].Amongthem,thebogieisamain componentsofthetrain.Thebasicstructureofthebogiemainlyincludesthewheelset, suspensionsystem,frame,auxiliarysuspensionsystem,andsoon.Therefore,itisvery importanttoresearchfaultdiagnosisforthehigh-speedtrain,especiallythediagnosisof bogiefaults[73].

Withthecomplexdiversificationofhigh-speedrailoperatingconditionsandtheincrease inoperatingmileage,mechanicalwearandagingofvehiclesaregraduallyaccelerating, whichcausessafetyhazardsfortrainoperation.Amongthem,thetrainaxleisan importantcomponenttosupporttherunningofthetrain.Duringthedrivingprocess, almostalloftheweightofthetrainandtheimpactcausedbyvibrationareassumed. Therefore,theaxleisoneofthemostvulnerablepartsofthetrain.Asaresultofimproper assemblyoftheaxlesandexcessiveoperationoftheaxlesduringmaintenance,theweight ofaxleswillbeaggravated.Whentheaxletemperaturerisesabnormally,movement betweentheaxleandbearingswillbeworse,frictionandwearwillbeaggravated,and smoothnesswillbereduced.Inthiscase,ifthevehiclecontinuestorunwithout emergencytreatment,hiddensafetyhazardswilloccurandthetrainwillbedelayedor evenderailed.Therefore,thehealthstatusofthebogieisdirectlyrelatedtothesafetyof thevehicleoperation.Thefaultwarninganddiagnosisoftheaxleofthehigh-speedtrain areanurgentproblemtobesolved.

Withtherapiddevelopmentofsensortechnology,alargeamountofreal-timemonitoring datageneratedduringtheoperationofhigh-speedtrainshasbeencollected.Throughthe analysisandminingofhistoricalmonitoringdata,thepotentialcorrelationandperiodicity canbeexploredtoprovideabasisfortrainfaultdiagnosis.Byanalyzingthemonitoring datacollectedduringthetrain’soperation,itisimportanttoidentifythehealthstatusand thepotentialsafetyhazardsthatmayexistatthetimeoftheearlywarning,whichis greatlysignificantforensuringthesafeoperationofthetrainandimprovingthereliability ofthetrain’soperation.

Bogiefaultdiagnosisincludesfaultvibrationsignalprocessingandfaultidentification. Amongthem,faultvibrationsignalprocessingincludestime-domainanalysis,frequencydomainanalysis,andtime-frequencyanalysis.Faultidentificationusestheexpertsystem, patternrecognition,neuralnetworks,deeplearning,andsoforth[74,75].Becausethe high-speedtrainbogiehasnonlinearcharacteristics,itsvibrationsignalisanonlinear, nonstationarysignal.Therefore,applyingeffectivefaultvibrationsignalprocessing methodstoextractstateinformationandconstructingafastandeffectivebogiefault diagnosismethodiscriticalforaccuratelyevaluatingthesafetystateofthetrain.

Withtherapiddevelopmentofrobotics,moreandmoretrainsareonthelineandrail transitchannelrobotshaveemerged.Comparedwithindustrialrobotsandinspection robots,railtransitchannelrobotshavebeenindevelopmentforashorttime,buttheyform

Introduction15

acertainscale.Channelrobotsaremainlybasedonintelligenttechnologyandintegrate varioussensingequipment.Itisdedicatedtothecollectionanddiagnosisoffault information.Thebiggestcontributionofrailtransitchannelrobotsistoreplaceon-site workerstocheckthetrain’sstatus.Thestaffoftheinspectionstationcancarryoutremote controloperationsthroughnetworktechnologyandcanreliablymonitorthesiteinreal time,whichcanobjectivelymonitortheequipmentandmeettherequirementsof flexibility,highprecision,andantiinterference.Ingeneral,railtransitchannelrobotsneed tomeetthefollowingrequirements:

(a)Thechannelrobotbelongstothemobilerobot.Unlikethefixedindustrialrobot,it needstomoveaccordingtotheactualtask.Therefore,thenavigationcontrolsystem needstohaveacontrolmoduletoperformrobotmotioncontrol.Ofcourse,themovementofthechannelrobotcanbealongtheguiderailorautonomousnavigation.

(b)Therobotneedstocarryavarietyofsensorstoobtaininformationaboutthedetected target,layingthefoundationforintelligentfaultdiagnosis.

(c)Thenavigationsystemneedstobemodularizedtomakeiteasytoreplacemodules andeasytomaintain.Communicationbetweenmodulesmustbecompleted.

(d)Thechannelrobotneedstouploadcollectedinformationandthecorrespondingfault type,faultlocation,andsoontotheprebuilthistoricalfaultdatabase.Withthecontinuousincreaseinfaultinformation,thedatabasehasgraduallygrown.Dataminingcan beusedtopredictthefaultandlifeoftrainequipmentandtoidentifypotential dangersinadvance.

Railtransitchannelrobotsmainlyperformafaultdiagnosisoftrains.Generallyspeaking, theworkflowisasfollows:(1)themaintenancetaskiscarriedoutalongthegroundtrack tothecheckpoint;(2)thevehiclemaintenanceinformationiscollectedbytheonboard sensorincludingthevisiblelightcamera,thelasersensor,andtheinfraredcamera,and somesensorsareinstalledontherobotarm;(3)thecollectedimageforfaultdiagnosis analysisisuploaded;and(4)thefaultdiagnosisresultsarereportedandstored,including thefaulttypes,faultproblems,faultdegrees,andsoon.Inadditiontothegrowing database,thebigdataanalysismodulewillperformdataminingontrainbogieequipment failureinformationtopredictpotentialequipmentfailures.Inaddition,therobot continuouslydetectsitsownpowerandperformsarechargingtaskoncetheminimum batterythresholdisreached.

Channelrobotswithatroubleofmovingelectricmultipleunits(EMU)detectionsystem (TEDS)intelligentsensingsystemhavebeenusedforEMUfaultdiagnoses,aimingtouse therailtransitchannelrobotstocompletetrainstatusdetectionandfaultdiagnosis[76].

Thehardwarestructureofachannelrobotmainlyincludesagroundrail,dual-armrobot, lasersensor,visiblelightcamera,infraredcamera,server,andrechargingdevice.Among them,thefunctionoftherobotguideistodrivetherobotandmakeitmovealongthe

16Chapter1

preplannedroutetoexpandtheworkingradiusoftheindustrialrobot.Thearmofa two-armedrobotissimilartothatofanindustrialrobot.Thedifferenceisthatatwoarmedrobotcanperformmorecomplextasksandcontrolismoredifficult.Lasersensors, visiblelightcameras,infraredcameras,andotherin-vehiclesensorsbelongingtoa channelrobotareusedtocollectdifferenttypesofdata.Thebackgroundserverismainly usedforthechannelrobotcontrol,datastorage,faultdiagnosis,andbigdataprocessing. Theself-rechargingtechnologyoftherechargingdeviceisanimportantmeanstoensure thenormaloperationofthechannelrobot.Currently,themoreadvancedoneisnoncontact recharging.Comparedwithcontactrecharging,noncontactrecharginghastheadvantages ofasmallfootprintandlowcost.

Thesoftwareofarailtransitchannelrobotmainlyincludesintelligentvisiblelightimage analysis,infraredthermalimagingtemperatureanalysis,locationdetection,andbigdata analysis.Amongthem,visibleimageintelligentanalysisandinfraredthermalimage analysisareusedforfaultdiagnosisofthetrain.Visiblelightimageanalysismainly includesimageregistrationandchangedetectionalgorithms,andthetrainfaultisfound bycomparingimages.Infraredthermalimagingtemperatureanalysismainlyincludesfault locationandimagesegmentation,faultlocationimagerecognition,andclassification.The formerusesimageprocessingtechnologyandthelatterreliesonmachinelearning methods.Locationdetectionmainlyuseslaserrangingtechnologytodeterminethe locationoftrainfaults.Bigdataanalysisismainlybasedonthetrainbogieequipment faultinformationdatabase;itusesdataminingandmachinelearningandpredictsthe potentialfaultoftheequipmenttoavoidaccidents.

1.1.3.3Unmannedaerialvehiclesforinspection

Railwayaccidentsaresuddenandrandom,soregularinspectionsofrailwaysmustbe rigorousandcritical.Withthecontinuousdevelopmentofunmannedaerialvehicle(UAV) technology,thedemandforcivilianUAVsinmanycountriesisgrowingandUAVshavean increasinglyimportantroleinmanyfields.Researchscholarshaveamajorinterestin developingnewtypesofUAVsthatcanflyautonomouslyindifferentenvironmentsand locationsandperformavarietyoftasks[77].Therefore,varioustypesoftheUAVswith differentsizesandweightshavebeeninventedanddesignedfordifferentsituationsand scenes.Forexample,UAVsslowlybegantoreplacesomemanualworkofrailway operationsandmaintenanceandhavebeensuccessfullyappliedinthefieldofrailway safety.CombiningexistingtechnicalskillsofUAVswiththedailysafetyrequirementsof railwaytracks,UAVscaninspectfordailymonitoringofrailwaytracks.Railtransit inspectionUAVinspectionreferstotheuseoftheUAVsequippedwithvisiblelight camerasandinfraredsensorstodetecttheoperationstatusofrailwayequipmentandrealtimeweatheranddiscoverthehiddendangersofrailwaytrainoperation.Equipmentalong therailwayincludespantographs,powerlines,andsoon.

Introduction17

1.1.3.3.1Pantograph

Thepantographandcatenarytechnologycanconnectelectricalenergyfromacontact networktoanelectriclocomotive[78].Asthespeedofthetrainincreases,thevibration characteristicsofthecatenaryandthepantographareparticularlyimportantforthe stabilityandsafetyofthetrain.Toobtainastablecurrent,thecontactpressureshouldbe maintainedbetweenthecontactslideofthepantographandthecontactline.Receiving currentfromthetractionnetbythepantographisaprocessinwhichseveralmechanical movementsoccursimultaneouslyduringthehigh-speedrunningofahigh-speedtrain.The pantographandthelocomotivearecloselyconnected.Whenthelocomotiveisdrivingfast, thepantographisalsorubbingquicklyandmovingforwardrelativetothecontactline.

Becauseofthestructuralcharacteristicsofthepantograph,itmaintainstension,soit generatesvibrationintheverticaldirection.Toensurethepressurevalue,thepantograph isalsoincontactwiththecatenary,andhigh-speedlateralvibrationoccurs.Thecatenary hasaninherentvibrationfrequencythatcreatesatravelingwavepropagatingalongthe catenary.Thesekindsofmovementsoccuratthesametimeandarelikelytocause separationbetweenthenets[78].Inthecaseofhigh-speeddriving,thesetypesofmotion aresuperimposed,andtheaccuracyofconstructionofeachlineisdifferent.Thismakesit difficulttomaintainstablecontactpressurebetweenthepantographandthecontactline. Whenthecontactpressureiszero,itislikelythatmechanicaldisengagementwilloccur, whichisthepantograph-catenaryoffline.Whentheelectriclocomotiverunsathighspeed, contactbetweenthepantographandthehigh-voltagecatenaryisgood,whichisequivalent toaclosedcircuitwithaninductiveload,andtheloadoperatesnormally.Whenthenetis separated,itisequivalenttodisconnectingtheinductiveloadcircuit.

Becauseofitscircuitcharacteristicsorovervoltage,overvoltagebetweennetworksishigh enoughtorupturetheairaroundthegap,whichwillcauseasparkdischargeandthenan arcdischarge.Whetheritissparkdischargeorarcdischarge,thelocomotivecurrentatthis stagewillbegreatlyaffected.Atthisstage,notonlywillovervoltagebegenerated, electromagneticnoisewillbedisturbed.Inaddition,highheatreleasedbythearcwill meltthemetalandcausewearonthepantographandcatenary.Whenthereisanoffline situationbetweenthepantographandcatenary,safeandstableoperationofthehigh-speed railwaywillfacegreathazards:unstableoperationoftheelectriclocomotive,severewear ofthepantographslide,deteriorationofthetractionmotorrectificationconditions,and generatedovervoltageinthemaincircuit.

Accordingtostatistics,railwayaccidentscausedbyfailureofthetractionpowersupply systemaccountedfor40%and30%ofaccidentsin2009and2010,respectively[79].

Thequalityofthepantograph-catenarysystemhasbecomeanimportantprerequisitefor speedingandimprovingoperationalreliability[79].Currently,thereal-timedynamic monitoringsystemofthepantograph-catenarysystemhasbeeninstalledonnewlybuilt trains,andthetestsystemhasbeeninstalledintheoperationEMUs.Thesystemis

18Chapter1

installedontopofthetrainandcandetecttheworkingstateofthepantographinreal time.Themainmethodistocollectimageinformationthroughthecameraforanalysis andprocessing,andthepan/tiltcanadjusttheshootingangleofthecamera.Nevertheless, themonitoringeffectofthebowmonitoringsystemwillbelimitedbythefixed installationpositionofthecameraandtherangeofcameraangles,sothatthepantograph canbemonitoredfromonlyonedirectionandthepantographcannotbefullyfaultmonitored.WiththegradualriseofUAVtechnology,atrainpantographmonitoring systembasedontheUAVinspectionhasemerged.Dependingontheflexibilityofthe UAVs,itistheoreticallypossibletoachieveall-arounddetectionofthestateofthe pantographandavoidblindspotsinthepantographfailure.Moreover,considering simultaneousworkindifferentareas,multiplesetsofparallelUAVscanbeusedfor detectiontoimproveefficiency.

1.1.3.3.2Powerlines

Theoverheadtransmissionlineisakindofsteelframestructurethaterectsthewireand keepsitawayfromtheground.Thesafetyofoverheadtransmissionlinesaffectspeople’s normallife,isrelatedtoeconomicproductioninthecountry,andevenposesathreatto nationalsecurity.Long-distancetransmissionlinesarecharacterizedbylonglinedistances andhighvoltagelevels.Transmissionlinesarealsocalledoverheadtransmissionlines. Thetransmissioncapacityofthelinesislarge,thetransmissiondistanceislong,andthey areintheopenairforalongtime.Theyareoftenaffectedbythesurrounding environmentandnaturalchangessuchassnow,lightning,ice,externalforces,birds,and soforth.Anultrahighvoltage(UHV)transmissionlinehashighvoltage,alongline,a highpoletower,alargeconductor,andacomplexgeographicalenvironment.Itputshigh requirementsonoperationandmaintenance[80].Traditionallinemaintenanceand overhauloperationshavecomplicatedprocesses,highrisks,apoormonitoringwork environment,andlowlaborefficiency.Ifthereisanemergencytrip,powerfailure,and natureforcemajeure,onlylow-endinspectionequipmentsuchasatelescopeandnight visiondevicecanbeused.Therefore,themanualoperationandmaintenancemodeare unabletocopewiththesafeoperationandmaintenanceoftheUHVoveralongdistance.

1.1.3.3.3Foreignmatterinvasiondetection

Comparedwiththelandinspectionrobotsintroducedearlier,theUAVscandetectforeign matterintrusionwiththeobviousadvantagesofwidedetectionrangeandemergency response[81].Whentherailwayishitbyamajornaturaldisaster,theadvantagesof inspectionUAVsarehighlighted.Forexample,seriousmudslidedisastersmayoccuralong therailwayline,posingaserioushazardtotherailwaysubgrade.Ifamudslideisnot foundandtakenemergencymeasuresintime,itwillleadtoserioustrafficaccidents.In general,theoccurrenceofmudslidedisastersisaccompaniedbystrongrainfall.Inthis case,thetraindriver’slineofsightisseriouslydegradedanditisdifficulttodetectthe dangerintimeduringheavyrain.Historically,therehavebeentraincrashesinwhichthe

Introduction19

Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.