汽车雷达系统
第5期
雷达科学与技术
V01.10No.52012年10月
RadarSclenceandTechnoIogy
October2012
特约文章
汽车雷达系统
赫尔曼・罗林
摘
要:100多年以来,无线电探测与测距(RADAR)一直是一项全球众所周知的技术.它曩初基于德
国工程师克里斯蒂安・侯斯芙尔在1904年4月第30届柏林皇家专利会上获得了发明。按照传统习惯。他将其技术创新起拉丁语名为电动镜。雷达的历史始于詹姆斯・克拉克・麦克斯韦的理论工作,随后生于德国汉堡的亨里希・赫兹通过很多试验以弄清电磁波的特性。舰艇间的避撞是这项技术的首次应用。然而,今天讨论的是汽车间的避撞应用,但这仅仅是汽车雷达系统成功应用的开始。
关键词:汽车雷这;调频连续渡;24GHz雷达中圈分类号:TN959
文献标识码:A
文章编号:1672—2337(2012)05—0453・07
AutomotiveRadarSystems
HermannRohling
(Department
ofTelecommunications,HamburgUniversityofTechnology,21073Hamburg,Germany)
Abstract:Radiodetectionandranging(RADAR)is
a
worldwidewell—knowntechniquesincemorethan
100
years,whichisoriginallybased
on
theinventionoftheGermanengineerChristianHOlsmeyer,whoap—
pliedhispatentat
theKaiserlichePatentamtinBerlin
on
Aprilthe30th,1904.Hecalledhisinventiontelem—
obiloskopin
a
goodtraditionofusingLatintermsfortechnicalsubjects.Theradarstorystartedwiththethe—
oreticalworkofJamesClerkMaxwell,followedbyHeinrichHertz,borninHamburg。whodidalItheexper—imentalwork
to
understandthe
nature
ofelectromagneticwaves.Collisionavoidancebetweenshipswasthe
firstapplicationforthisnewtechnique.Todaywecomeback
tO
thecollisionavoidanceapplicationhowever
nowbetweencars.Thisisjustthebeginningoftheautomotiveradarsystems
success
story.
Key
words:
automotivewave(FMCW)l24GHzradar
’
radar;frequencymodulatedcontinuous1Introduction
Figure1showsthethreescientistswho
con—
tributed
to
theunderstandingoftheelectromag-
neticwavenature.Thetheoreticalworkwasde—velopedbyMaxwell,theexperimentalvalidationhasbeencarriedout
by
HeinrichHertzandthe
Figure1JamesMaxwell,HeinrichHertz,Christian
technical
applicationsystem
was
invented
by
HUlsmeyer
ChristianHtilsmeyer.
important
to
noticethatdriving
acar
remains
a
Figure2showsthe
cover
pageoftheworld—
demandingandextremelyriskytask!Humanbe—widefirstradarpatent.The
text
isclearlywrit-
ingsdo
not
recognizethisriskindividually.How—
ten
andwellorganized,SOthateverybodycould
ever
theincreaseofsafety
inroadtrafficand
understandwhatthecomplicated
nature
ofelec-
transportationisstilla
realtechnicalchallenge.
tromagneticwaves,propagationofwavesandtheThere
are
about100.000fatalities
on
European
designof
a
radarsystemisabout.
motorwaysandroadseveryyear[11,whichisfarRegardingautomotiveradarapplicationsitis
beyond
a
realsafetysituation.
收稿日期:2012—03・10
454
■达科学与技术第10卷第5期
CHRHOLSMEYER・“DOSSELOOAF
v-■■-._-■-●■-H■h‘■■■_-t●_ittkeot●h-h■妇w●■一
■_●■●●■●-l●■■_
●●■_■●■■_●__帼-●一糖-
F÷||
Figure3Anearlyradarsystemfortomotiveapplications
au—
progress
isobservedinthe
developmentof24inthelastyears.
we—
GHzautomotiveradarThese
Figure2
HUlsmeyer’spatent
on
sensors
sensors
thetelemobiloskop
measurethepositionandradial
locityofall
Anti-lockbreakingsystem(ABS)andelectronicstabilitycontrol(ESC)belongsistantsystems(DAS),which
to
cars
inthelocalenvironment.The
calculatedtrafficinformationandsituationanaly-sisbased
on
importantdriverbased
on
as—
radar
sensor
measurements
andin
a
ispro—
auto-
are
theinfor—
current
cessedinsidetheradar
sensor
possible
mationofinternalsensors,thatmonitorthe
state
maticreaction
throttlecontr01.
isconsideredthebrakeand
ofthevehiclesuccessfully.Tllismeans。modem
cars
fulfilalreadyveryhigh
can
The24GHzradartechnologyhasbeendevel—
oped
on
a
safetystandards,buttheroadtrafficsafetyitself
professionalbasissince2000.The24
a
beimprovedverymuch.Inthisrespectitisworthy
to
GHztechnologyisduedue
toto
commercial
successnot
only
noticethathumanbeings
to
have
stronglimitationsinsignificantlowerproductioncosts,butalsoradiographyofplasticparts,allweather
theability
ence
m鼬.surethedistanceandspeeddiffer—
cars
betweentwo
whichisquiteoftenthe
re.ason
capability,etc.Nowadaysthe24GHztechniqueprevailsin
for
ugly斟酒融嘎慨Therdore
additionaldriver’//.Sis-
to
automotiveapplicationsandenables
tance
isneededfromthetechnicalpointofview
blindspotdetection,adaptivecruisecontrol,lanechangeassistantandcollisionavoidancesystems,
improvethesafety.
Thefirstradarappliwith
one
measurementsfOUa《tomotive
。蝴g髓;娃e
syst蛳:.:such
which
evencan
distinguishbetweenpedestrians
1970s
as
andvehicles.
Figure4showsthe24GHzradar
sensor
thethat
showninFigure3.Nevertheless,themeas—
at
isplacedbehindthefrontbumper.Themaximum
rangeofthis
urementprinciplesandresultsobtained
that
automotiveradar
sensor
isabout200
timewereimportantstepsfortheprofessionalde—velopmentofautomotiveradar.
Theallweathercapabilityandthe
accurate
mandtheazimuthanglecoverageistypically40。.Inthisobservation
ans
area
severalvehicles,pedestri—
andfixedtargetswillbeobservedanddetec-
reason
measuringschemefortargetrange,radialveloei—
tyandazimuthanglemakestheautomotiveradar
sensor
a
tedbytheradarsensor.Forthat
themulti
targetcapabilityofaradarsensorisanimportant
strong
candidate
forseveral.,,differentadvantagecomparedwithothertypesof
sensors.
driverassistantsystems.
Thereis
a
Thetargetrange,theradialvelocityandtheazi—
sensors,
largepotentialofradar
muthangleshouldbemeasuredprecisely,simul-taneouslyandunambiguously.Thistaskis
a
whichmeasurethetargetrangeandradialveloci—
ty
real
simuhaneously,unambiguouslyand
even
verypre—
technicalchallenge.Therefore,aspecificradarwaveformhas
to
cisely
inmultipletargetsituations.Alarge
bedesigned
to
processtheradar
echosignalspreciselyand
tO
resolvedifferent
tar—
getsinmultitargetsituati。ns.
P(r)
!!
-
i—瓦≯:慕2主=1
I
.
1
_●
Figure5
GHzradar
Pulseradar
Figure4
Z4
sensor
Targetrangeismeasuredbaseddelaybetweentransmitsignalthe
echo
on
on
thetime
Thetremendousprogressinthedevelopmentof24and77GHZradar
sensors
andreceptionof
is
measured
ofthelastyears
fl
signal.
Radial
velocity
recentlyculminatedinthepresentationofsystem
on
a
radar
based
theDopplerfrequency.Butthiswave—
tO
a
singlechipbased
on
SiGeBiCMOS
costs
formleadsistherefore
highcomputationcomplexityandappropriateforautomotiveradar
MMICtechnology.Thisreducesthesystem
not
inmassproductionandthemeasurementaccuracy
applications.
dramatically[2。.Mountingsuch
a
radar
sensor
be—
2.2
hindthefrontbumpermeansthatalltargetsinsidethelocalenvironmentup
to
a
ContinuousContinuous
Wave
Radar
continuously
maximumrangeof200
wave(CW)radars
meterswillbedetected.Additionally,thetargetazi—muthangleisestimatedveryaccuratelyutilizingthewell-knownmonopulsetechnique.
Fromferent
an
transmitandreceivethedesignatedsignalwave-form.Theechosignalisdown-convertedbytheinstantaneous
transmit
a
frequencyfrequency
to
thebase—
applicationpointofviewseveraldif—
band.With
linear
modulated
CW
butcharacteristictasks,e.g.,adaptive
or
(FMCW)--transmit
timedelayresultsinradialvelocity
causes
waveform(seeFigure6)the
a
a
cruisecontrol(ACC),blindspotdetectionlisionavoidancesystems,canbethese
tyto
to
cases
col—
frequencyshift.Alsothefrequencyshiftdue
to
considered。Inhastheabili—
the
the24GHzradar
sensor
Doppler-effect.Thereforetargetrangeandradialvelocity
on
a
controlthedistanceandvelocitydifference
monitors
theblind
cannot
bemeasuredindependentlybased
theprecedingtraffic.It
singlechirpsignal.
spotarea,controls
thebreakandthrottleassis—
a
tant,andinforms
driveraboutpossibleobsta—
clesinsidethelane.
2
Waveform
Design
Thecapabilityofmeasuringtargetrangeandradialvelocitysimultaneouslyistialfeatures,whichmakeradar
oneSO
ofthe
essen—
au—
suitablefor
tomotiveapplications.Theappropriate。,,vaveform
designisthekey
tO
Figure6FMCW—-waveformandFSK—-waveform
thisability.
Afrequencyshiftkeying(FSK)radarwave—
2.1PulseRadar
form,seeFigure6,consistoftWOmonofrequentcarrierfrequencieswhich
are
ThefirstandmostintuitivewaveformiSthepulsewaveform
as
organizedin
an
in—
picturedinFigure5.tertwinedway.TheDopplerfrequencyof
anytar-
get
can
bemeasuredinthis
case
unambiguously
fromthetwomonofrequent
echosignalsby
ap—
plying
an
FFT.Additionally,the
timedelayof
theechosignalresultsina
phaseshiftbetween
theechoes
of
both
alternating
transmitted
sig—
nals.Thereforetarget
range
andradialvelocity
can
bemeasuredindependently
on
a
single
trans—
mitsignalbasis
even
in
multi
target
situations.
Thedrawbackofthistransmitschemeisthefactthatstationarytargets
cannot
beresolved.
With
the
multiplefrequency
shift
keying
(MFSK。seeFigure7)acombinationofboththe
FMCW—and
FSK—transmitsignal
can
be
pro—
eessed.Twostepwiselinearlyfrequencymodula—tedsignalswiththesameslopebut
a
certainfre—quency
shift
are
transmittedin
an
intertwined
way.Themeasuredfrequencydifference
as
well
as
thecorrespondingphaseshiftbetweenthein—
tertwined
signals
depends
on
both
the
target
rangeandradialvelocity.
Figure7
MFSKwaveforrn
Thereforemovingandstationarytargets
can
be
resolvedinrangeandvelocity.Alsothecapabilityto
independentlymeasuretargetrangeandvelocity
even
inmultitargetsituationsisstillgiven[3].Forthese
reasons
andforthesetechnicaladvantagestheMFSK
signalwaveformisappliedinautomotiveradarsys—
tems.With
a
frequencybandwidthof
1
5.9MHz
a
rangeresolutionoflmisobtained.
ItiS
important
to
noticethatthiswaveform
isnotprocessedby
a
classicalmatchedfilterpro—
cedure
or
analyzedby
an
ambiguityfunction.Itisprocessedalternatelyin
a
non—matchedfilterform
to
obtain
an
unambiguousandsimultaneous
tar—
get
range
and
Doppler
frequency
measurement
withhighresolutionandaccuracy
even
inmultiple
targetsituations.
Theechosignalofthe
stepwise
andinter—
twinedwaveformis
down-convertedby
thein—
stantaneous
transmitfrequency
tO
thebaseband
andsampled
at
theendofeachshortfrequency
step.Thistimediscretesignalis
thenFourier—
transformedseparately
forthe
tWO
intertwined
sub-signals.Thisallowsthemeasurementofthebeatfrequencyaf,whichissimultaneouslyin—fluencedbytargetrangeRandradialvelocity口inthefollowing1inearway:
△厂一÷・口一慧・R
㈩
Asingletargetwillbedetected
on
thesamespectralline
at
position
af
forthe
tWOinter—
twinedsignals.Therefore,thephasedifference△。betweentWOcomplexvalued
signals
on
the
spectral1ineA厂willbecalculatedafterthetargetdetectionprocedure.Thisphasedifference△Pde—pendsagain
on
target
rangeRandradial
velocity
剐as
describedinthefollowingequation:
4丁c
47c,。。
卸一一瓦可‘础一
(2)
C
Hence,targetrangeRandradialvelocityu
can
bedeterminedin
an
unambiguouswaybysol—
vingthelinearsystemofequations(1)and(2).Thusghosttargets,whichwill
occur
inclassical
FMCW
signalwaveform,arecompletelyavoided
inmultipletargetsituations
if
an
MFSKtransmit
signaliSapplied.
3
MonopulseTechnique
Acomprehensiveanalysisofthetrafficsitua—
tionrequires
theknowledgeoftheazimuthangleof
a
target
in
addition
to
itsrangeandvelocity
measurement.
Themonopulse
techniqueallows
theazimuth
angle
to
bedeterminedbytherelativereceivepower
oftwostationarymountedreceive
antennas
with
a
defineddirectivityeach(seeFigure8).
2012年第5期
赫尔曼・罗林:汽车雷达系统457
These
antennascan
bedesigned
as
flatpatch
antennas
withverysinaildimensions1ikethe
one
picturedinFigure9.Becausethese
antennas
may
beassembledbehindthebumper,themonopulsetechniqueis
a
veryconvenient
solutionfor
auto-
motiveradar
sensors.
RXI
RX2
TX
Figure9
Patch-antennawithtWOreceive-antennas
(1eft)and
one
transmit-antenna(right)
formonopulsetechnique
4
SignalProcessing
Intheanalogueworldtheradarechosignalwas
directlymapped
to
theplanpositionindicator(PPI),
as
shownin
Figure10.Thesituationanalysisand
targetdetectioneveninstrongclutterandnoise
areas
wasdonemanually.Today,'theechosignal
isdown-
converted
to
thebaseband,sampledandprocessedin
a
timediscretedigitalsignalprocessingscheme.The
mostimportantstep
isthetargetdetection,which
distinguishesbetweenrelevanttargetsandnoise
or
clutter.
4.1
CFAR
Therefore,
constant
falsealarm
rate
(CFAR)--algorithms
are
necessaryineveryradarsystem
to
displayonlyrelevanttargets(seeFigure11)t‘-5].
Figure10Displayof
a
radarmeasurementwithout
cluttersuppression
A∞
/~1}厂
Jtude
-
二勰¥-2htmhold
‘弋∽
.‰M肌M肭A从i^h础川。1.
Figure11
CFARdetectionprocedure
Inautomotiveradarhoweverthehandlingofthefalsealarm
rate
plays
a
particularrole,be-
causean
interventioninthedriver’Sactions
or
in
theperformanceof
the
car
due
toa
falsealarm
wouldhavefatal
consequences.
Thecontinuousobservationof
a
target’Spo—
sitionandvelocityleads
to
thepossibility
tO
pre-
dictthetarget’Sprospectivepositionandmaneu-
vers.
Consequentlydangerous
situations
can
clearlybeanalyzed,whichisofinterestespeciallyunderbadweatherconditions.Theinformationgainedfrom
trackingisrequired’byallDAS
to
avoidaccidents(seeFigure12).
5
AutomotiveApplmaUons
Thefirstapplication,forautomotiveradars
wasautomaticcruisecontrol(ACC).Inthis
case
a
singleradar
sensor
ismountedbehindthefront
bumper.The
radar
sensor
measures
target
range,radialvelocityandazimuth
angleof
all
cars
withintheobservationarea.Thetargetin—
formationiSforwardedto
thebrake-throttle
con—
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!竺I
458
雷达科学与技术
一
第10卷第5期
一
a
一一一一一——————————I!I
humanbeingand
a
car
flectionmodelofwith
sixdifferentreflectionpointseach.Due
ent
tO
differ-
velocitiesoftheassumedreflectionpoints,a
broad
Doppler
characteristic
spectrum
occur
and
a
a
point-shapedrangeprofilewillpedestrian[71.
in
case
of
Figure12
Procedureforautomaticsignalprocessingindriverassistancesystems
trolunit
to
take
a
decisionaboutacceleration
car
or
decelerationofthe
to
keep
a
safedistance.
腐风I■—■■一I
『‘
,
一
瓜_I
a
Thelanechangeassistant(LCA)isanotherimportanttworadar
automotive
sensors
Figure14
Sixpointreflectionmodelofpedestrianand
vehicle
a
application,which
rear
uses
mountedbehindthe
area
bump-
car
AvehicleshowsgenerallyDopplerspectrumin
a
a
muchnarrower
er.Theseradarsobservetheincludingtheblindzone.In
behindthe
a
sinceitisconstantlymoving
a
case
of
targetdetec—
certaindirectionbutprofile.The
car
has
an
extendedandthe
tionthesystemwarnsthedriverbyilluminatingwarningLEDsplacedshowninFigure13.
at
range
Dopplerspectrum
thesidemirror,e.g.,as
rangeprofilearesuccessfullyusedfortargetclas。
sificationprocedures.Figure15showsthechar—acteristicrangeprofileandthetargetextensionAR
as
well
as
the
Dopplerspectrum
a
andthe
DopplerspreadAvfor
a
pedestrian(1eft)andfor
car(right).Itisclearlyindicatedbythepicture
thattherangeprofileandtheDopplerspectrum
are
totallydifferentfor
a
pedestrianandfor
a
car.Thesesignalfeatureswillthereforebeusedin
Figure13
Lanechangeassistant(source:AudiAG)
a
targetrecognitionscheme.
AnupcomingspecifictaskisthepedestrianrecognitionFrom
a
6Conclusions
Startinginthelate1990sthedevelopment
using
automotiveradar
sensors.
puremeasurementpointofviewtheob—
can
ofautomotiveradar
sensors
producedimpressive
servedradartargetsand
in
radial
haveextensionsinrangedirection
respectively.
results.Figure16belowshowsthedevelopmentmilestonesfromtomotive
one
to
velocity
ofthefirstexperimentalcommercial
sensors
an-
Hence,therangeprofileandtheDopplerspec—trum
are
sensors
with
thebasisfor
a
targetrecognitionproce—
verysmalldimensionsandfinallytotheradaron
dure,whichwillbeintegratedintothesoftwareof
an
chipsolutionbysmartmicroBraunschweig,Germany.
This
paper
presents
start
upcompanyin
automotiveradarsensor.Thebackscat-
velocity
are
teredamplitudeandthemeasuredtwoadditionaIfeatures,which
can
the
largepotentialof
sensor
beusedintheshows
a
re—
the24GHzautomotiveradarwith150
technology
on
classificationprocess[引.Figure14
MHz
bandwidthandbasedtheMF一
2012年第5期
赫尔曼・罗林:汽车雷达系统459
×10.
6
5
壹4
营,
《2
I
FFT
Bin[N】
孙TBin[N】
Xto'
名暑一_旨司
矸T
Bin[N1盯rBinlN】
Figure
15
Rangeprofile
and
Dopplerspectrum
for
a
pedestrian(1eft)and
a
vehicle
(right)
Figure16Developmentofautomotiveradarsensors
fromexperimentalsystems
tOsmallandlow
COSt
sensors
andfinally
tO
the
radar
on
a
chipsolutionbysmartmicro
SKwaveform.Theseradar
sensors
makedriv-
ingmorecomfortable,lessexhaustingandmuchsafer.Automotiveapplicationssuch
as
ACC,
lanechangeassistant
or
theblindspotdetection
are
alreadyimplementedinmodern
cars.
Future
additionalapplications,e.g.,pe。
destrian
recognition
based
onan
extended
Dopplerspectrumand
a
point-shapedrangepro—
fileshowtremendousprogress,whichwilllead
toa
furtherincreaseofthesafetyinnormalroad
trafficsituations.
References
[1]Statistics
ofroad
traffic
accidents
in
Europe
and
NorthAmerica[M].New
York:UnitedNations,
2007.
[2]MendeR.MMIC
for
a
24GHzAutomotiveRadarSensor[C]}}International
RadarSymposium2009,
Hamburg:s.n.],2009.
[3]MeineckeMM,Rohling
H.CombinationofFSKand
LFMCWModulation
PrinciplesforAutomotive
Radars[-C]∥GermanRadarSymposiumGRS2000,
Berlin,Germany.[s.n.],2000:155-159.[4]Rohling
H.RadarCFARThresholdingin
Clutter
andMultipleTargetSituations[J].IEEE
Trans
on
AerospaceandElectronicSystems,1983,19(4):608-621.
[5]Rohling
H.25YearsResearchin
Range
CFAR
Techniques[C]∥Proceedings
oftheIRS2003。
Dresden,Germany:[S.n.],2003:363-368.
[6]F61ster
F,RohlingH,MeineckeMM.Pedestrian
Recognition
Based
on
Automotive
RadarSensors
[C]∥5
thEuropeanCongresson
IntelligentTrans—
portation
SystemsandServices2005・Hannover,
Germany:[S.n.],2005.[7]Rohling
H,HeuelS,Ritter
H.PedestrianDetec—
tionProcedureIntegratedinto
an
24GHzAutomo。tire
Radar[C]∥2010
IEEE
RadarConference,
WashingtonDC。USA:[S.n.],2010:1229—
1232.
HermannRohlingiswiththeTech‘nical
University
Hamburg-Har—
burg,Germany.HeisheadoftheTelecommunicationsdepartment
wherehe
hasdeveloped
an
interna-
tionalreputation
for
MobileCorn-
munication(4G)andautomotiveradarsystems.Previ-ouslyProf.RohlingwaswiththeAEGResearchInsti—tute。Ulm
as
a
researcherworkinginthe
area
ofdigital
signalprocessingforradarandcommunicationsapplica—tions.HiSresearchinterestshaveincludedWidebandMo-
bileCommunicationsespeciallybased
on
MulticarrierTram-
mission僦hniques(OFDM)for
futurebroadbandsystems
(4G),signaltheory,digital
radar逝柏l
processing,detec-
tion,estimationanddifferentialGPS[orhighprecisionnavi—
gation.Prof.Rohlingis
a
memberof
Irdormationstechnischa
Gesellschaft(ITG),president
ofthe
GermanInstitute
of
Navigation(DGON)and
a
FellowofIEEE.HeiSa
chairman
oftheSeptemher2010InternationalO同DMWorkshop(In-
oWo2010)andtheInternationalRadarSymposium(1RS
2011)inLeipzig,Germany.Prof.RohlingistheVicepresi—dentof
the1&hnical
UIIiversityHamburg-Harburg.