Upload
vantu
View
277
Download
2
Embed Size (px)
Citation preview
第13卷 第4期
2013年8月
交 通 运 输 工 程 学 报
JournalofTrafficandTransportationEngineering
Vol.13 No.4
Aug.2013
犚犲犮犲犻狆狋 犇犪狋犲:20130412
犚犲狊犲犪狉犮犺犘狉狅犼犲犮狋狊:NationalNaturalScienceFoundationofChina(51208054);SpecialFundforBasicScientificResearchofCentralColleges
(2013G1211005,CHD2011JC056)
犃狌狋犺狅狉 犚犲狊狌犿犲:LIYan(1983),Male,Hengshui,Hebei,LecturerofChanganUniversity,PhD,ResearchonTrafficSignalControl,
+862982334636,lyan@chd.edu.cn.
文章编号:16711637(2013)04011611
过饱和状态交通信号控制方法综述
李 岩1,赵志宏2,李鹏飞3,4,黄身森1,陈宽民1
(1.长安大学 公路学院,陕西 西安 710064;2.长安大学 信息工程学院,陕西 西安 710064;3.东南大学 交通学院,
江苏 南京 210096;4.阿尔伯塔大学 智能交通中心,阿尔伯塔 埃德蒙顿 T6G2R3)
摘 要:为向拥堵状态城市路网的交通控制优化提供依据,综述并评价了过饱和状态的交通信号控
制策略;分析了过饱和状态交通流的特征,提出了过饱和状态交通信号控制的对应优化原理;从单
点交叉口、协同控制交叉口及路网3个层次对过饱和状态交通控制方法进行了总结与评价,应用
VISSIM仿真环境评价了已有控制策略的适用性;总结了具备处理过饱和状态能力的自适应交通
信号控制系统的控制策略。仿真结果表明:过饱和状态的交通信号控制应优先优化道路空间的分
配矛盾;排队管理策略对过饱和交通流具有较好的优化效果;建议在进行过饱和交通信号控制优化
时,结合实时交通信息,采用排队管理、关键路径通行能力最大等控制策略,在交叉口群或路网层面
对交通信号控制方案进行优化。
关键词:交通信号控制;过饱和状态;控制策略;排队管理;交通仿真
中图分类号:U491.51 文献标志码:A
犚犲狏犻犲狑狅犳狋狉犪犳犳犻犮狊犻犵狀犪犾犮狅狀狋狉狅犾犿犲狋犺狅犱狊狌狀犱犲狉狅狏犲狉狊犪狋狌狉犪狋犲犱犮狅狀犱犻狋犻狅狀狊
LIYan1,ZHAOZhihong2,LIPengfei
3,4,HUANGShensen1,CHENKuanmin1
(1.SchoolofHighway,ChanganUniversity,Xian710064,Shaanxi,China;2.SchoolofInformation
Engineering,ChanganUniversity,Xian710064,Shaanxi,China;3.SchoolofTransportation,
SoutheastUniversity,Nanjing210096,Jiangsu,China;4.CentreforSmartTransportation,
UniversityofAlberta,EdmontonT6G2R3,Alberta,Canada)
犃犫狊狋狉犪犮狋:Toprovidereferencesfortrafficcontroloptimizationatcongestedurbanroadnetwork,various
trafficsignalcontrolstrategiesunderoversaturatedconditionswerereviewedandevaluated.Basedon
theanalysisofthecharacteristicsofoversaturatedtrafficflow,thecorrespondingoptimizationprinciples
oftrafficsignalcontrolunderoversaturatedconditionswereproposed.Thetrafficcontrolstrategiesfor
isolatedintersection,coordinatedintersectionsandroadnetworkweresummarizedandevaluatedbyusing
VISSIMsimulationenvironment.Thecontrolmeasuresofadaptivetrafficsignalcontrolsystemwiththe
capabilitytodealwithoversaturatedconditionsweresummarized.Simulationresultindicatesthattraffic
signalcontrolstrategiesunderoversaturatedconditionsshouldoptimizetheallocationofroadspacein
thefirstplace.Thequeuemanagementstrategyhasbetterperformanceonoptimizingtrafficsignalunder
oversaturatedconditions.Itissuggestedthatthestrategieslikequeueratiomaintenanceandthroughput
maximizationforcriticalrouteshouldbeutilizedwithrealtimetrafficinformationfortheoptimizationat
theintersectiongrouporroadnetworklevel.4tabs,7figs,56refs.
No.4 LIYan,etal.Reviewoftrafficsignalcontrolmethodsunderoversaturatedconditions
犓犲狔狑狅狉犱狊:trafficsignalcontrol;oversaturatedcondition;controlstrategy;queuemanagement;
trafficsimulation
0 犐狀狋狉狅犱狌犮狋犻狅狀
Trafficsignalcontrolplaysanevidentlyimportant
roleinurbantransportationsystem management.
However,the commonly used traffic control
method,such as TRRL (transport and road
researchlaboratory)method[1],HCM (highway
capacity manual)method[2] oradaptivetraffic
controlsoftwarelikeSCATS(Sydneycoordinated
adaptivetrafficsystem)[3]andSCOOT(splitcycle
offsetoptimizing technique)[4],can not work
efficiently enough when the traffic demand
approachesorexceedsthecapacity[5].Themajor
reasonthatthesematuretrafficcontrolstrategies
havelowefficiencyunderoversaturatedconditions
isthatthey can notrespondtothe primary
restrictionsofoversaturatedtrafficsystem,which
changesfromtheallocationofpassingtimetothe
spatialspacelimitationalongroadsection[6].Thus,
specificsignalcontrolstrategiesareneededforover
saturatedtrafficsystem[7].Duringthepast50years,
manytrafficcontrolstrategiesweredesignedforover
saturatedtrafficsystem.Thepurposeofthispaperis
tosummaryandevaluatetheexistingtrafficcontrol
strategiesforoversaturatedconditions,andprovidesa
referencewaytooptimizetrafficsignalcontrolfor
congestedurbanroadnetwork.
Trafficcontrolstrategiesunderoversaturated
conditionsarefirstlyimplementedforisolated
intersections.Specificcoordinatedtrafficsignal
controlstrategiesforthearterialandroadnetwork
arethendevelopedforthepurposeofdealingwith
congestionstatus.Beginningwithdistinguishing
the conceptions of approaching saturation,
saturationandoversaturation,theoversaturated
trafficcontrolstrategiesforvariousscalesare
summarizedand evaluated.As a part ofthe
advancedtrafficmanagementsystem (ATMS)[8],
adaptivetrafficsignalcontrolsystems withthe
capabilitiesofhandlingoversaturatedflowarealso
discussed.
1 犇犲犳犻狀犻狋犻狅狀犪狀犱犮犺犪狉犪犮狋犲狉犻狊狋犻犮狊狅犳
狅狏犲狉狊犪狋狌狉犪狋犻狅狀
Trafficstatusisgenerallydeterminedbytraffic
intensity,i.e.犞/犆(volume/capacity)ratio.Fora
roadsection,approachingsaturationusuallyrefers
tothesituationof犞/犆ratiolargerthan0.9[9],and
oversaturationisthetrafficstatuswith犞/犆ratio
greaterthan1.0,thesaturationstatushasa犞/犆
ratioexactlyequals1.0.Theconceptionsoftraffic
conditions at intersection are similar to the
conceptionsatroadsection.Ifthetrafficdemand
ofoneapproachoftheintersectionexceedsits
capacity,thisintersectionisunderoversaturated
status.AsshowninFig.1,becauseitisdifficult
tomeasuretheactualtrafficdemandandcapacity
whenthetrafficsystemiscongested,theover
saturationatsignalizedintersectioncanbedefined
asthe condition of having an approach with
residualqueue[6].Thequeueinformationcanbe
directlyestimated byloop detectors[10],mobile
sensors[11]orvideo
[12]databyusingshockwave
theory.
Fig.1 Residualqueuesatintersections
Trafficflow willbecomeunstableunderover
saturationorapproachingsaturationconditions.A
smallfluctuationfromanyvehicleinaplatoonmay
cause adverse consequences and reduce the
efficiency oftraffic system sharply.Thelow
stabilityofsaturatedtrafficflow putsforward
morestringentrequirementstothetrafficcontrol
system.
Generally,oversaturatedconditionwillfirstly
appearatoneorseveralisolatedintersectionswith
relativelyhighsaturationdegrees.Asshownin
711
JournalofTrafficandTransportationEngineering 2013
Fig.2,whenintersectionBcantdischargethe
queue,itturnsintooversaturatedstatus.The
queuewillcontinuallygrowuntilitexceedsthe
roadsectionbetweenintersectionsAandB.Atthe
momenttheresidualqueuereachintersectionA,
bothintersectionsareundertheoversaturated
conditions.Noeffectivetrafficcontrolstrategy
existsatthismoment,buttopreventanyvehicle
fromenteringoversaturatedroadsection.This
detrimentaleffectisnamedspillback,whichshould
beavoidedinthefirstplace.Intersectionsalong
theroutewithlargertrafficvolumesmayspread
thecongestion muchfasterthan otherroutes.
Finally,thewholeroadnetworkwillbecongested,
evenunderalockoutstatus[13].Itisimportantto
implement traffic signal control strategies to
preventoversaturatedconditionsinthefirstplace,
ratherthantoreacttotheissuesafterthefact.
Oncethetrafficdemandgoesbacktothenormal
level,thefrequentlyusedtrafficcontrolstrategy
willmakethepractitionersmanagethetrafficflow
easily.Thecauseofoversaturatedstatusindicates
thatthedeclineofcapacitycausedbydetrimental
effectsisacriticalinterferingfactor.Therefore,
theessenceoftrafficcontrolstrategiesunderover
saturatedstatusistoeliminatethedetrimental
effectsbyadjustinggreentime,cyclelengthand
offset.
Fig.2 Spillbackatintersections
2 犗狏犲狉狊犪狋狌狉犪狋犲犱犮狅狀狋狉狅犾狊狋狉犪狋犲犵犻犲狊
犳狅狉犻狊狅犾犪狋犲犱犻狀狋犲狉狊犲犮狋犻狅狀
2.1 犈狓犻狊狋犻狀犵狋狉犪犳犳犻犮狊犻犵狀犪犾犮狅狀狋狉狅犾狊狋狉犪狋犲犵狔
2.1.1 Optimizationoffixedtiming
SinceWebsterconnectedtheaveragedelaywith
thesignaltimingparametersandtheobserved
variables by using cumulative diagram,these
parameterscanbeoptimizedby minimizingthe
delayofsignalizedintersection(knowingasTRRL
method)[1].Similarto TRRL method, many
optimizationobjectivesarealsoappliedinsignal
timingoptimizationprocess,suchasminimizingcycle
failures[14],minimizingstops
[15]andminimizingaverage
saturationdegree[2].However,becauseofthe
principleofalgorithm,theycanhardlybeutilized
under oversaturated conditions. Although the
coordinate transformation model[1617] and the
deterministicmodel[18]wereestablishedtoestimate
thedelayorqueueinthesignalizedintersection
withrelativelyhighsaturationdegree,thetraffic
signalcontrolcandonothingbuttoallocatemore
greentimetoreducethedelay.
2.1.2 Switchingofgreen
Whenthegreentimebecomesflexible,switching
ofgreenisthefirstandsimplesttrafficcontrol
strategyfortheoversaturatedintersections.This
strategycanallocatetheunusedgreentimefrom
uncongested phaseto oversaturated phaseto
ensurethe elimination ofresidual queue.By
applyingthiscontrolphilosophy,thetotaldelays
can be minimized by maximizing intersection
productivity, while keeping the queues
developmentwithinacceptablelevelsatthesame
time.A maximum greentimeisgiventothe
movementwiththehighersaturationflowuntilthe
queueisdissipated,whileothermovementsreceive
theminimumgreentime[19].
Widelyusedtechnologiesofthisstrategyinclude
semi/fullactuated control,NEMAs (national
electrical manufacturesassociation)ringbarrier
phasemodelinNorthAmerica,andphasestage
model[2023]in Europe.Intheactuated control
system,controllercanextendorterminatethe
currentphasetoallocatetheunused greento
congested phase. NEMAs ringbarrier phase
modelextendstheideaofactuatedcontrol.As
illustratedinFig.3,thedualringcontrolunitcan
givemorepriority(i.e.moregreentime)totraffic
flowofspecificintersectionapproachbyadjusting
thebarriersbetweenphases1,5(3,7)and2,6
(4,8).Phasestage modelcangroupallthe
unconflicted movements and order them by
811
No.4 LIYan,etal.Reviewoftrafficsignalcontrolmethodsunderoversaturatedconditions
priorities. The phase associated with critical
movementgroupcangetmoregreentime.
Fig.3 TypicaldualringcontrolunitofNEMA
Theswitchingofgreenstrategydoesimprove
theefficiencyoftrafficsignalsystem,especially
forthesituationofintersectionwithonlyoneor
twomovementsbeingcongested.However,when
the volumes of all movements exceed their
capacity,thesealgorithmswillextendallphasesto
maximumgreentime.Atthattime,trafficcontrol
systemhasnodifferenceswithfixedtimecontrol.
Ingeneral,minimizingdelayviasplitswitchingis
notanonlinehighlyresponsivepolicy,butan
optimizationpolicybasedonadvancedmathematics
andthedetailedknowledgeofdemand[24].
2.1.3 Maintainingqueueratio
The nature of green adjustment policeisto
reallocatethetemporalresourcesforthepurpose
ofimprovingtheirefficiencies.However,when
theroadspacebecomestherestriction ofthe
system,maintainingqueueratioshouldbeselected
astheoptimizationobject[25].Underthisstrategy,
thegreentimeshouldbeadjustedtobalancethe
queuesatintersectionsapproaches.Theobjective
istominimizethenumberofblockedintersections
by managing queuesofcriticalintersection[26].
Thisstrategyinvolvestheuseofsampledloopdata
ateachintersection[10,27]ormobilesensorsdata
[11].
Recent studies indicated that the signalized
intersection wont maximize its capacity and
minimizeits delay at the same time under
approaching saturated or oversaturated
conditions[28].Inthis way,signaloptimization
objectivesshouldbeswitchedtothemeasuresof
reducingthequeuelengthatthebottleneckand
minimizingthe detrimentaleffects underover
saturated status.These optimization objectives
involvethethroughputmaximizationorthequeue
minimizationofcriticalroute.
2.2 犛犻犿狌犾犪狋犻狅狀犫犪狊犲犱犲狏犪犾狌犪狋犻狅狀
2.2.1 Simulationenvironment
Inordertoevaluatetheperformanceoftraffic
control strategies for oversaturated isolated
intersection, fixed timing method (TRRL
method),Synchro method (HCM method),
actuated control method, NEMAs dual ring
controlunit,phasestagemodelandqueueratio
maintenancemethodareevaluatedintheprevailing
microscopictrafficsimulationenvironmentbased
onVISSIMsimulationsystem.Theadvantagesof
VISSIM overothersimulationpackagesinclude
someaspectsasfollows.
(1)VISSIM providesthelargestflexibilityfor
userstocalibratedriving behaviorsandtraffic
conditions.
(2)VISSIMisdevelopedunder.NETframework,
which brings flexibility for addon program
development.
(3)VISSIM providesthebesttoolsforthe
developmentofsignalcontrolstrategies,suchas
NEMA controller emulator, vehicle actuated
programming (VAP)language,signalcontrol
application programminginterfaces (SCAPIs),
etc.
VISSIMSCAPImethodisusedtodevelopsignal
controlemulatorinthisresearch.SCAPIsare
writteninC+ +/CLRlanguageandtheoriginal
version of SCAPIs controller requires signal
controlalgorithms be embeddedinto a single
dynamiclinklibrary(DLL)file.Tofacilitatethe
development,a middlewareisdeveloped,which
can synchronously collect all the realtime
detectors/phasesstatesfromtheVISSIMnetwork
tothecontrolleremulatorthenreturnthenew
desired phase states back to the VISSIM
network[29].Ateachtimestep,controllerrunsthe
algorithm, makes decisions according to the
currentstate,thenreturnsthenewdesiredphase
statestotheVISSIMnetwork.Theconceptofthis
simulationenvironmentisillustratedasinFig.4.
911
JournalofTrafficandTransportationEngineering 2013
Fig.4 SimulationenvironmentofVISSIM
2.2.2 Experimentaldesign
Fig.5 Intersectiongeometry
A simplefour wayintersectionisselectedto
evaluatetheselectedalgorithms.Thegeometric
characteristicsoftheintersectionareshownin
Fig.5.Theexperimentalconfigurationsarelisted
inTab.1.Alltheneededdatacanbedirectly
obtainedfromtheVISSIMsoftware.Alltheinput
parametersareoptimizedinadvance,andthenare
inputtotheVISSIMenvironment.TheVISSIM
softwarehasabuiltinfixedtimecontrollerto
achieveTRRL methodand HCM method.The
CrossingsoftwareandVisVAPmoduleinVISSIM
are applied to program the actuated control
method,phase stage model and queue ratio
maintenance method. The VISSIM standard
NEMAeditorisutilizedtoevaluatethedualring
controlunit.
犜犪犫.1 犈狓狆犲狉犻犿犲狀狋犪犾狋狉犪犳犳犻犮狏狅犾狌犿犲狊 pcu·h-1
MovementEastbound Westbound Northbound Southbound
Low High Low High Low High Low High
Leftturn 65 90 30 45 30 45 10 60
Throughput620 880 700 990 370 530 510 720
Rightturn 35 50 20 45 20 25 50 70
2.2.3 Resultanalysis
Eachscenariorunsfor50timestoconductstatistical
analysis.Throughputvehiclesandaveragequeueratio
areselectedtocomparetheselectedtrafficcontrol
methods.AsshowninTab.2,itcanbeindicated
thatundernormalconditions,theactuatedcontrol
method,theNEMA methodandthephasestage
methodallhavesimilarperformances.Theflexible
greencontrolstrategies,suchasactuatedcontrol,
NEMAandphasestage,haverelativelybetter
performance than the fixed timing control
strategies(TRRL,HCM).Althoughthequeue
ratiomaintenancemethodhasgoodperformancein
managingthequeueratio,thiscontrolstrategyis
notthemostefficientoneundernormalconditions.
However,underoversaturatedconditions,the
queue ratio maintenance method has better
performancethanallotherstrategies,andthus,it
is suitable to be utilized under congested
conditions.Ifthetrafficdemand continuesto
grow,specifictrafficcontrolstrategiesforlarger
rangeareneededtobeconcerned.
犜犪犫.2 犆狅犿狆犪狉犻狊狅狀狅犳犲狏犪犾狌犪狋犻狅狀狉犲狊狌犾狋狊
AlgorithmsThroughputvehicles/pcu Averagequeueratio
Normal Oversaturated Normal Oversaturated
TRRL 2294 2986 0.46 >1.00
HCM 2298 3018 0.44 >1.00
Actuatedcontrol 2325 2996 0.45 >1.00
NEMA 2322 3041 0.43 >1.00
Phasestagemethod 2318 3008 0.44 >1.00
Queueratiomaintenance 2309 3328 0.41 0.99
3 犛狋狉犪狋犲犵犻犲狊犳狅狉犮狅狅狉犱犻狀犪狋犲犱犻狀狋犲狉狊犲犮狋犻狅狀狊
3.1 犈狓犻狊狋犻狀犵犮狅狅狉犱犻狀犪狋犲犱狋狉犪犳犳犻犮犮狅狀狋狉狅犾狊狋狉犪狋犲犵犻犲狊
3.1.1 Optimizationofprogressionband
Signalcoordinatedcontrolstrategiestrytoseekfor
themaximumprogressionbandbyadjustingthe
021
No.4 LIYan,etal.Reviewoftrafficsignalcontrolmethodsunderoversaturatedconditions
offsetsbetweengreenphasesundersteadytraffic
conditions.Withpredefined maximumlikelihood
offsets,coordinatedtiming plansaregenerally
designedfortrafficflowundersteadystates.As
theintersectionsbecomemoresaturated,residual
queuesstarttodisrupt movementatupstream
intersections.Then,theprogressionbandisnot
applicableanylonger[30].Ifthe oversaturated
conditionlastsforaconsiderabletimeperiod,
fixedtiming signal coordination is likely to
aggravatemovementdisturbance,whichmaycause
spillback[31]intoupstreamintersections
[32].
Fig.6 Dynamicoffsetsinactuatedcoordination
Variablebandwidth progression[3334] is then
establishedtoadjusttheoffsetsaccordingtothe
traveltimeinformationinrealtime.Asshownin
Fig.6, the variablebandwidth progression
improvestheefficiencyofcoordinatedcontrol,and
makesitworkatmostconditions.The Monte
Carlomethod[35]canbeutilizedtooptimizethe
bandwidthtopreventtheearlystartedcoordinated
greenphaseintheactuatedsignalsystem.Underover
saturatedconditions,therealtimeinformationishard
toobtain.Meantime,inordertomakemorevehicles
passthroughtheprogressionband,thebandwidths
tendtobelarge,whichmakessignaltimingtobe
fixed (maximum greenatthecriticalrouteand
minimumgreenatotherroute)[36].
3.1.2 Maximumstoragecapacity
Insteadofsearchingthe maximum progression
band,thestrategyofmaximumstoragecapacities
ofapproachesisutilizedtooptimizequeuesat
intersectionapproaches[37].Thismethodresponds
totherestrictionsofroadspaceundertheover
saturatedconditions,andtriesto makeefficient
useof upstream links storage capacity.The
procedureofusingsimultaneousoffsetsalongthe
arterialandnegativeoffsetswithflaringofgreen
alongcrossroadsisthentestedundertheover
saturatedarterialnetwork.Becauseofmarkedly
decreasedspillbackblockages,theprocedureof
offsetsettinghaslargerthroughputvehiclenumber
thantraditionaloffsetsetting,and witha20%
reductionin overalltraveltime[38]. With the
possibilityofhighresolutiontrafficdata,recent
researchesdescribedsimplecontrollerlogicfor
truncating a phase early when faced to a
downstreamrestrictionofflow[3940]andtoavoid
detrimentaleffectlike “spillback”or “defacto
red”[13].Defactoredassumesthatnoflowona
detectorwhenthelightisgreenindicatingthat
thereisnowhereforthevehicletogo,andthusit
isbetterforoverallintersectionoperationstomove
ontoanotherphase[41].
3.2 犛犻犿狌犾犪狋犻狅狀犲狏犪犾狌犪狋犻狅狀
3.2.1 Evaluationsetup
Anarterialisselectedtoevaluatetheperformance
ofmaximumprogressionbandstrategy,variable
bandwidth progression strategy,and maximum
storagecapacitiesstrategy.Thearterialisabout
8.85kmlong,fourwaysfortwodirections,and
sixsignalizedintersections.Fig.7showsitslayout
andtrafficmovementcounts.Thefiledobserved
121
JournalofTrafficandTransportationEngineering 2013
trafficiscongestedorclosetocongested.Allsix
signalizedintersectionsarecontrolledbytheEconolite
ASC/3signalcontrollersandinductiveloopdetectors
withfourwaypedestrianphases.
TheVISSIMSCAPImethodisalsoutilizedto
evaluatethecoordinatedtrafficcontrolstrategies.
Theoriginaltrafficsignaltimingplanisgenerated
bySynchro,aprevailingsignaltimingoptimization
package in North America, using maximum
progressionbandstrategy.Fig.7alsoshowsthe
optimized fixed timing plans of the selected
intersections.Thevariablebandwidthprogression
strategyandmaximumstoragecapacitiesstrategy
areimplementedbyCrossingsoftwareandVisVAP
module.
Fig.7 Selectedroadnetwork
3.2.2 Resultanalysis
Duringtheprocessofevaluation,eachscenario
runsfor50timesandsummarizestheaverage
value.Wheneverthetimingplansareupdatedand
aplatoonisidentified,only phase splits are
adjustedwhereasthecyclelengthsandoffsetsare
keptunchanged.Thereforethecoordinationwill
notbe destroyed on the mainline.Sincethe
expected advantages ofthe maximum storage
capacitiesstrategyincludeprovidingsignaltimings
to avoid queue spillback and queue clearance
withoutdestroyingcoordination.Thefollowing
MOEs(measuresofeffectiveness)arechosento
evaluatethe performance ofthe ACS control
strategy:linktraveltime,totalstops,theaverage
controldelay,andthemaximumqueueratioonthe
mainline.Tab.3summarizestheMOEsforallthe
threetrafficcontrolstrategies.
犜犪犫.3 犕犗犈狊犮狅犿狆犪狉犻狊狅狀
Performance
Maximum
progression
band
Variable
bandwidth
progression
Maximum
storage
capacities
Southboundlinktraveltime/s 1156 1278 1029
Northboundlinktraveltime/s 1043 1032 970
Southboundtotalstop/veh 6765 7254 5683
Northboundtotalstop/veh 4064 4147 3576
Averagecontroldelay/s 30 33 29
Southboundmaximumqueueratio >1.00 >1.00 0.97
Northboundmaximumqueueratio 0.91 0.90 0.87
From Tab.3,itisclearthatthe maximum
storagecapacitystrategycanreducethelinktravel
timeandvehiclestoponthemainlinebyfrom5%
to 15%. However, the variablebandwidth
progressionstrategyhasissuesonoptimizingthe
signalofsouthboundtraffic,andcauseserious
221
No.4 LIYan,etal.Reviewoftrafficsignalcontrolmethodsunderoversaturatedconditions
congestion during simulation. Although the
performanceofnorthboundisslightlybetterthan
thefixedtimescenario,thevariablebandwidth
progressionstrategyisnotsuggestedtobeutilized
underoversaturatedconditions.The maximum
storagecapacitystrategy doesnotsignificantly
reducethemaximumqueueratioorcontroldelays
atintersections,becausethedistancesbetween
intersectionsarerelativelylongandtheconstraints
regarding the relationship of the estimated
maximumqueuelengthandlinklengthbarelyplay
aroleduringtheoptimization.Itisexpectedthat
morebenefitofqueuespillbackavoidancewillbe
acquiredifmaximumstoragecapacitystrategyis
deployedatcloselyspacedintersections.
4 犛狋狉犪狋犲犵犻犲狊犳狅狉狉狅犪犱狀犲狋狑狅狉犽
Thesignalcontrolstrategiesintheroadnetwork
levelare the extension ofthe strategies for
coordinatedintersections.Similartothestrategies
forcoordinatedintersections,thestrategiesfor
roadnetworkcanbedividedintoadaptivesignal
controlandmetering/gatingpolice.Theadaptive
controlstrategiestrytoseekforarelativelybetter
allocationofthepassagetimeatthenetworklevel
byapplyinghillclimbingmethodorotherheuristic
algorithms. When the road space becomes a
limitationfortrafficsignaloptimization,itis
suggestedtodistributethetrafficdemandoverthe
network,andmakefulluseoftheentireroadspace
withintheselected area.The metering police
extendsthequeue managementstrategy.This
strategyimpedestrafficinputatasuitablepointor
points upstream to preventthe demand from
reaching criticallevels at critical downstream
locations[42].Itcanmanagetherateofflowinto
andwithinhightrafficdensitynetwork[43].One
wellknownrealtimetrafficcontrolpolicy for
congested arterialsis RT/IMPOST (realtime/
internalmeteringpolicytooptimizesignaltiming),
whichisdesignedtocontrolqueuedevelopmenton
everysaturated approach by suitable metering
trafficto maintain stablequeues[44].A traffic
controlstrategyforintersectiongroupunderover
saturatedconditionfurtherdevelopedthe RT/
IMPOSTstrategybyaddingcriticalroutelayer[5].
Thisprocedureensuresthecriticalroute[4546]isthe
primary optimization object for oversaturated
intersectiongroup.Inordertoaccomplishtheroad
networkcontrolstrategies,anumberofmodels
were developed to identify the queue,queue
dischargetime,available downstream storage,
etc[4749].
5 犃犱犪狆狋犻狏犲狋狉犪犳犳犻犮狊犻犵狀犪犾犮狅狀狋狉狅犾狊狔狊狋犲犿
犳狅狉狅狏犲狉狊犪狋狌狉犪狋犲犱犮狅狀犱犻狋犻狅狀狊
Forthereasonthattheoptimizationobjectsunder
oversaturatedconditionsaredifferentfrom the
onesatsteadystates,specifymodulesorfunctions
areneededtobuiltinthetrafficsignalcontrol
systemtodealwiththeoversaturatedconditions.
Generally,anadaptivetrafficsignalcontrolsystem
withthecapabilityofworkingunderoversaturated
conditions should contain traffic status
identification module, congestion cause
identification moduleandseveralsignalcontrol
parametersoptimizationalgorithms.Commonly
usedcommercialadaptivetrafficcontrolsystems
generally have the capability of calculating
saturationdegreeinapproximatelyrealtime,such
asthesaturationpercentageinSCOOT[4],orthe
degree of saturation in SCATS[3]. However,
becauseofthereasonthatnospecifyalgorithmis
designedforoversaturatedconditions,whenthe
saturateddegreeapproaches1,thestabilityof
thesesystemsdropssharply[50].Todealwiththe
traffic congestion management problem,some
adaptivetrafficcontrolsystemsaredesignedto
avoidcongestionoroversaturatedconditions[51],
suchasSCOOT MC3[52],MOVA
[5354],TUC[55],
andMOTION[56].Detailmeasuresappliedinthese
systemsaresummarizedinTab.4.
6 犛狌犿犿犪狉狔狅犳犾犻狋犲狉犪狋狌狉犲狉犲狏犻犲狑
Accordingtotheoptimizationprinciples,the
trafficsignalcontrolstrategiescanbedividedinto
twocategories:the allocation oftemporalor
spatialresources.Understeadystates,theefficiency
321
JournalofTrafficandTransportationEngineering 2013
犜犪犫.4 犆犺犪狉犪犮狋犲狉犻狊狋犻犮狊狅犳狋狉犪犳犳犻犮犮狅狀狋狉狅犾狊狔狊狋犲犿狊犳狅狉犱犲犪犾犻狀犵狑犻狋犺狅狏犲狉狊犪狋狌狉犪狋犲犱犮狅狀犱犻狋犻狅狀狊
Adaptivetrafficcontrolsystems Optimizationobjects Controlstrategies
SCOOTMC3(SpiltCycleOffsetOptimization
Technique,ManagingCongestion,Communi
cationsandControl)
Roadnetwork
1.Trafficstatusidentificationandcongestioncause
identification
2.Congestionimportancefactorconsideration
3.Trafficguidanceoftrafficdemandfromtheoverloaded
junctiontoothers
MOVA(MicroprocessorOptimizedVehicle
Actuation)
Isolatedintersectionextendedfor
SCOOTinlatestversion
1.Trafficstatusidentification
2.Specifyoptimizationobjectiveforoversaturated
condition
TUC(TrafficresponsiveUrbanControl) Roadnetwork
1.Twolevelsstructure
2.Metering/gatingpolice
3.Specifyoptimizationalgorithmforcyclelength,split
andoffsetunderoversaturatedcondition
MOTION(MethodfortheOptimizationof
Traffic Signals in Online Controlled
Networks)
Roadnetwork
1.Trafficstatusidentification
2.Metering/gatingpolice
3.Specifyoptimizationobjectiveforoversaturated
condition
4.Higherpriorityforcriticalroute
canbeimprovedbyassigningmoreunusedgreento
the movement withrelatively highersaturation
degree. However, under oversaturated
conditions,theroadspacebecomesarestriction
thatmustbeconsidered.Iftoomuchgreentimeis
assignedtoonespecifymovement,thedownstream
road or otherintersections will exceed their
capacities,whichmayleadtothelockoutstatus.
Hence,traffic control strategies under over
saturatedconditionsmustconsidertherestriction
ofroadspace.Suggestedcontrolstrategiesat
differentlevelsarelistedasbelow:maintaining
queueratio atisolatedintersectionlevel,the
maximumstoragecapacitiesofapproachesatthe
coordinatedintersectionlevel,andmeteringpolice
at road network level. These measures are
suggestedtobeutilizedatthecriticalrouteatthe
networkorcriticalintersections,whichcanbe
identifiedbyhighresolutiontrafficdata.
Trafficsignalcontrolstrategiesforhandling
oversaturated conditions still haveissues for
improvements,whicharelistedasbelow.
(1)Realtimeinformationisstrictlyneededto
achievetheproposedtrafficcontrolstrategies,
particularly at the road network level. The
implement of the proposed traffic control
strategies,suchas meteringoradaptivetraffic
control,needtrafficinformation,likequeuelength
orcriticalroute,tobeupdatedatleasteveryfive
minutes. Further development of traffic
informationacquisition method,likefusingthe
fixedlocationdataandmobilesensordata,should
beprocessedtosupporttheproposedintegration.
(2)The detrimentaleffectscaused by over
saturationmayspreadtothewholeroadnetwork
in a very shorttime. Traffic signal control
strategiesshouldoptimizethetrafficflowalongthe
routewithcriticalconflictsinthefirstplace.In
thisway,trafficcontrolstrategiesshouldconcern
onintersections with strong trafficrelevance,
whichcanbenamedasintersectiongroup.
(3)TrafficflowinChinaisdifferentfromother
countries.Trafficsignalparameters,suchasgreen
flashing,allred,shouldalsobecalibratedwith
localfiled data before applying the selected
strategies.Specifictrafficsignalcontrolstrategies
particularly for Chinese urban road networks
shouldalsobeconsideredtostudy.
犚犲犳犲狉犲狀犮犲狊:
[1] WEBSTERFV.Trafficsignalsettings[R].London:Road
ResearchLaboratory,1958.
421
No.4 LIYan,etal.Reviewoftrafficsignalcontrolmethodsunderoversaturatedconditions
[2] YINY.Robustoptimaltrafficsignaltiming[J].Transportation
ResearchPartB:Methodological,2008,42(10):911924.
[3] SIMSAG,DOBINSONKW.TheSydneycoordinatedadap
tivetraffic(SCAT)systemphilosophyandbenefits[J].IEEE
TransactionsonVehicularTechnology,1980,29(2):130137.
[4] BRETHERTONRD,BOWENGT.Recentenhancements
toSCOOT—SCOOTversion2.4[C]∥IEEE.Proceedingsof
the3rdConferenceonRoadTrafficControl.London:IEEE,
1990:9598.
[5] LIYan,GUOXiucheng,YANGJie,etal.Mechanismanaly
sisandimplementationframeworkfortrafficsignalcontrolof
oversaturatedintersectiongroup[J].JournalofTransporta
tionSystemsEngineeringandInformationTechnology,2011,
11(4):2834.
[6] DENNEY R W,HEAD L,SPENCER K.Signaltiming
undersaturatedconditions[R].Washington DC:Federal
HighwayAdministration,2008.
[7] PAPAGEORGIOUM,DIAKAKIC,DINOPOULOUV,etal.
Reviewofroadtrafficcontrolstrategies[J].Proceedingsof
theIEEE,2003,91(12):20432067.
[8] WANGFeiyue.Parallelcontrolandmanagementforintelli
genttransportationsystems:concepts,architectures,and
applications[J].IEEETransactionsonIntelligentTranspor
tationSystems,2010,11(3):630638.
[9] DONGChunjiao,SHAOChunfu,MAZhuanglin,etal.
Temporalspacialcharacteristicofurbanexpresswayunder
jamflowcondition[J].JournalofTrafficandTransportation
Engineering,2012,12(3):7379.
[10] WU Xinkai,LIU H X,GETTMAN D.Identificationof
oversaturatedintersectionsusinghighresolutiontrafficsignal
data[J].TransportationResearchPartC:EmergingTech
nologies,2010,18(4):626638.
[11] BANXuegang,HAOPeng,SUNZhanbo.Realtimequeue
lengthestimationforsignalizedintersectionsusingtravel
timesfrommobilesensors[J].TransportationResearchPart
C:EmergingTechnologies,2011,19(6):11331156.
[12] SHIZhongke,QIAO Yu.Detection methodofqueuing
vehiclesonurbanroad[J].JournalofTrafficandTransportation
Engineering,2012,12(5):100109.
[13] ABBASM M,ADAMZ M,GETTMAND.Development
andevaluationofoptimalarterialcontrolstrategiesforover
saturatedconditions[J].TransportationResearchRecord:
JournaloftheTransportationResearchBoard,2011(2259):
242252.
[14] URBANIKT,BEAIRDS,GETTMAND,etal.Trafficsignal
statetransitionlogicusingenhancedsensorinformation[R].
WashingtonDC:TransportationResearchBoard,2003.
[15] HEJiajia,HOUZaien.Antcolonyalgorithmfortraffic
signaltiming optimization[J].Advancesin Engineering
Software,2012,43(1):1418.
[16] AKCELIKR,ROUPHAIL N M.Estimationofdelaysat
trafficsignalsforvariabledemandconditions[J].Transporta
tionResearchPartB:Methodological,1993,27(2):109131.
[17] PARKBB,LIChenyang.Ananalyticalapproachforesti
matingthehighwaycapacitymanualsignalizedintersection
delayvariability[J].ComputerAidedCivilandInfrastructure
Engineering,2011,26(4):327333.
[18] KIMBERRM,HOLLISEM.Peakperiodtrafficdelaysat
roadjunctionsandotherbottlenecks[J].TrafficEngineering
andControl,1978,19(10):442446.
[19] GAZISDC.Optimumcontrolofasystemofoversaturated
intersections[J].OperationsResearch,1964,12(6):815831.
[20] MICHALOPOULOSPG,STEPHANOPOULOSG.Over
saturatedsignalsystemswithqueuelengthconstraints—Ⅰ:
singleintersection[J].TransportationResearch,1977,11(6):
413421.
[21] MICHALOPOULOSPG,STEPHANOPOULOSG.Over
saturatedsignalsystemswithqueuelengthconstraints—Ⅱ:
systemsofintersections[J].TransportationResearch,1977,
11(6):423428.
[22] CHANGTH,SUNGY.Modelingandoptimizationofan
oversaturated signalized network[J]. Transportation
ResearchPartB:Methodological,2004,38(8):687707.
[23] CHANGTH,SUNGT.Optimalsignaltimingforanover
saturatedintersection[J].TransportationResearchPartB:
Methodological,2000,34(6):471491.
[24] LIU Hongchao,BALKE K N,LIN Weihua.Areverse
causaleffectmodelingapproachforsignalcontrolofanover
saturatedintersection[J].TransportationResearchPartC:
EmergingTechnologies,2008,16(6):742754.
[25] LONGLEYD.Acontrolstrategyforacongestedcomputer
controlledtrafficnetwork[J].Transportation Research,
1968,2(4):391408.
[26] ZHANGMengmeng,JIALei,ZOUNan,etal.Robustoptimal
trafficsignaltimingofurbansinglepointintersection[J].Journal
ofHighwayandTransportationResearchandDevelopment,
2011,28(1):107111.
[27] MUCSIK,KHANAM,AHMADIM.Anadaptiveneuro
fuzzyinferencesystemforestimatingthenumberofvehicles
forqueuemanagementatsignalizedintersections[J].Trans
portationResearchPartC:EmergingTechnologies,2011,
19(6):10331047.
[28] DIONF,RAKHAH,KANGYS.Comparisonofdelayesti
matesatundersaturatedandoversaturatedpretimedsignal
izedintersections[J].Transportation Research Part B:
Methodological,2004,38(2):99122.
[29] LIPengfei,ABBASM.Stochasticdilemmahazardmodelat
highspeedsignalizedintersections[J].JournalofTransporta
tionEngineering,2010,136(5):448456.
[30] HEQing,HEAD KL,DINGJun.PAMSCOD:platoon
basedarterialmultimodalsignalcontrolwithonlinedata[J].
TransportationResearchPartC:Emerging Technologies,
2012,20(1):164184.
[31] LIUYue,CHANGGanglen.Anarterialsignaloptimization
521
JournalofTrafficandTransportationEngineering 2013
modelforintersectionsexperiencingqueuespillbackandlane
blockage[J].Transportation ResearchPartC:Emerging
Technologies,2011,19(1):130144.
[32] LIUHX,WUXinkai,MAWenteng,etal.Realtimequeue
lengthestimationforcongestedsignalizedintersections[J].
TransportationResearchPartC:Emerging Technologies,
2009,17(4):412427.
[33] STAMATIADISC,GARTNERNH.MULTIBAND96:a
programforvariablebandwidthprogressionoptimizationof
multiarterialtrafficnetworks[J].TransportationResearch
Record,1996(1554):917.
[34] HUPeifeng,TIANZongzhong,YUANZhenzhou,etal.
Variablebandwidthprogressionoptimizationintrafficopera
tion[J].JournalofTransportationSystemsEngineeringand
InformationTechnology,2011,11(1):6172.
[35] LIPengfei,GUOXiucheng,LIYan.AnATMSdatadriven
methodforsignalizedartrialcoordination[J].Journalof
SoutheastUniversity:EnglishEdition,2012,28(2):229235.
[36] LUKai,ZENGXiaosi,LILin,etal.Twowaybandwidth
maximizationmodelwithprorationimpactfactorforunbal
ancedbandwidthdemands[J].JournalofTransportation
Engineering,2012,138(5):527534.
[37] GORDONRL.Atechniqueforcontroloftrafficatcritical
intersections[J].TransportationScience,1969,3(4):279287.
[38] RATHIA K.A controlschemeforhightrafficdensity
sectors[J].Transportation ResearchPartB:Methodological,
1988,22(2):81101.
[39] BEAIRDS,URBANIKT,BULLOCKDM.Trafficsignal
phasetruncationineventoftrafficflow restriction[J].
TransportationResearchRecord,2006(1978):8794.
[40] CHENJuan,XULihong,YUANChangliang.Hierarchy
controlalgorithmanditsapplicationinurbanarterialcontrol
problem[J].JournalofSystemSimulation,2008,20(15):
41224127,4131.
[41] PUTHA R,QUADRIFOGLIO L,ZECHMAN E.Com
paringantcolonyoptimizationandgeneticalgorithmapproaches
forsolvingtrafficsignalcoordinationunderoversaturation
conditions[J].ComputerAided CivilandInfrastructure
Engineering,2012,27(1):1428.
[42] LIU Qin,XU Jianmin.Coordinated control modelof
regionaltrafficsignals[J].JournalofTrafficandTransporta
tionEngineering,2012,12(3):108112.
[43] RATHIAK.Trafficmetering:aneffectivenessstudy[J].
TransportationQuarterly,1991,45(3):421440.
[44] LIEBERMANEB,CHANGJ,PRASSASES.Formulation
ofrealtimecontrolpolicyforoversaturatedarterials[J].
TransportationResearchRecord,2000(1727):7788.
[45] LIYan,YANGJie,GUOXiucheng,etal.Criticalroute
identificationmethodatrelatedintersectiongroupbasedon
wavelettransform[J].ChinaJournalofHighwayandTransport,
2012,25(1):135140.
[46] LIYan,GUOXiucheng,YANGJie,etal.Routesclassifi
cationmethodatintersectionsgroupusingwavelettransform
andspectrumanalysis[J].JournalofSoutheastUniversity:
NaturalScienceEdition,2012,42(1):168172.
[47] PARKES,RILETTLR,SPIEGELMANCH.AMarkov
chainMonteCarlobasedorigindestinationmatrixestimator
thatisrobusttoimperfectintelligenttransportationsystems
data[J].JournalofIntelligentTransportationSystems:Tech
nology,Planning,andOperations,2008,12(3):139155.
[48] LERTWORAWANICH P,KUWAHARA M,MISKA M.
Anew multiobjectivesignaloptimizationforoversaturated
networks[J].IEEETransactionsonIntelligentTransporta
tionSystems,2011,12(4):967976.
[49] WEIHeng,PERUGU H C.Oversaturationinherenceand
trafficdiversioneffectaturbanintersectionsthroughsimula
tion[J].JournalofTransportationSystemsEngineeringand
InformationTechnology,2009,9(4):7282.
[50] KOUVELASA,ABOUDOLASK,KOSMATOPOULOSEB,
etal.Adaptiveperformanceoptimizationforlargescaletraf
ficcontrolsystems[J].IEEE TransactionsonIntelligent
TransportationSystems,2011,12(4):14341445.
[51] MIRCHANDANIPB,ZOUNing.Queuingmodelsforanalysis
oftrafficadaptivesignalcontrol[J].IEEETransactionson
IntelligentTransportationSystems,2007,8(1):5059.
[52] BRETHERTOND,BODGERM,COWLINGJ.SCOOT—
managingcongestion,communicationsandcontrol[J].Traffic
EngineeringandControl,2006,47(3):8892.
[53] VINCENTRA,YOUNGCP.Selfoptimisingtrafficsignal
controlusingmicroprocessors—theTRRL MOVAstrategy
forisolatedintersections[C]∥IEEE.SecondInternational
ConferenceonRoadTrafficControl.London:IEEE,1986:
102105.
[54] TENEKECIG.A modellingtechniqueforassessinglinked
MOVA[J].ProceedingsoftheInstitutionofCivilEngineers:
Transport,2007,160(3):125138.
[55] DINOPOULOU V,DIAKAKIC,PAPAGEORGIOU M.
ApplicationsoftheurbantrafficcontrolstrategyTUC[J].
EuropeanJournalofOperationalResearch,2006,175(3):
16521665.
[56] BIELEFELDTC,BUSCHF.MOTION—anewonlinetraf
ficsignalnetworkcontrolsystem[C]∥IEEE.SeventhInter
nationalConferenceonRoadTrafficMonitoringandControl.
London:IEEE,1994:5559.
621