You are on page 1of 12

International Journal of Advanced Robotic Systems

Mobile Ground-Based Radar


Sensor for Localization and Mapping:
An Evaluation of two Approaches

Regular Paper



Damien Vivet
1,2
, Franck Grossier
1,2
, Paul Checchin
1,2,*
,
Laurent Trassoudaine
1,2
and Roland Chapuis
1,2


1 Institut Pascal, Universit Blaise Pascal, Clermont Universit, Clermont-Ferrand, France
2 Institut Pascal, CNRS, UMR 6602, Aubire, France
* Corresponding author E-mail: paul.checchin@univ-bpclermont.fr

Received 2 Oct 2012; Accepted 10 May 2013



DOI: 10.5772/56636

2013 Vivet et al.; licensee InTech. This is an open access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.


Abstract This paper is concerned with robotic
applications using a groundbased radar sensor for
simultaneous localization and mapping problems. In
mobile robotics, radar technology is interesting because
of its long range and the robustness of radar waves to
atmosphericconditions,makingthesesensorswellsuited
for extended outdoor robotic applications. Two
localizationandmappingapproachesusingdataobtained
from a 360 field of view microwave radar sensor are
presentedandcompared.Thefirstmethodisatrajectory
oriented simultaneous localization and mapping
technique, which makes no landmark assumptions and
avoids the data association problem. The estimation of
the egomotion makes use of the FourierMellin
transform for registering radar images in a sequence,
from which the rotation and translation of the sensor
motion can be estimated. The second approach uses the
consequence of using a rotating range sensor in high
speed robotics. In such a situation, movement
combinations create distortions in the collected data.
Velocimetryisachievedherebyexplicitlyanalysingthese
measurementdistortions.Asaresult,thetrajectoryofthe
vehicleandthentheradarmapofoutdoorenvironments
can be obtained. The evaluation of experimental results
obtained by the two methods is presented on realworld
data from a vehicle moving at 30 km/h over a 2.5 km
course.

Keywords Radar Sensor, Mapping, Localization, Fourier


MellinTransform,Distortion

1.Introduction
The increased autonomy of robots is directly linked to
their capability to perceive their environment.
Simultaneous Localization and Mapping (SLAM)
techniques, which associate perception and movement,
are particularly interesting because they provide
advanced autonomy to vehicles such as robots. In
outdoorenvironments,climaticconstraintsoroperational
context(dust,lightcondition,rain,etc.)showthelimitsof
usualperceptionsystems.Thisworkproposesradarasan
alternative to existing perception systems for ground
1 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com
ARTICLE
www.intechopen.com
Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013
mobile robots. Its ability to work in all weather, day and
night, makes the radar very attractive for outdoor
applications. The combined use of radar with SLAM
techniques in the field of mobile ground robotics is the
maincontributionofthepresentedwork.

SLAM has been studied intensively over the past two


decades. For a broad and quick review of the different
approaches developed to face this high level of
complexityinautonomousmobilerobotapplications,the
reader can consult [14]. Localization and mapping in
large outdoor environments are applications related to
theavailabilityofefficientandrobustperceptionsensors,
particularly with regard to the problem of maximum
rangeandcompliancewiththeenvironmentalconditions.
Even though lasers and cameras are wellsuited sensors
for environment perception [57], their strong sensitivity
to atmospheric conditions has, among other reasons,
given rise to an interest in the development of mapping
andlocalizationmethodsworkingwithdatacomingfrom
othersensorslikeradarorsonar[8].

In this paper, two SLAM techniques are presented using


data from a 360

field of view radar sensor. This radar is


based on the Frequency Modulated Continuous Wave
(FMCW) technology [9]. In section 2, a review of articles
relatedtoourresearchworkiscarriedoutinthefieldsof
outdoor mapping and egomotion estimation. Section 3
briefly presents the microwave radar scanner we used
and that was developed by a team from the CEMAGREF
Institute(IRSTEA)workingonenvironmentalsciencesand
technologies. Section 4 deals with a scanmatching
approachforlocalizationandmapping.Section5presents
an approach based on an analysis of distortion due to
movement in order to extract information about the
robots trajectory and then the radar map of outdoor
environments. Finally, section 6 shows experimental
results.
2.Relatedwork
2.1Outdoormapping
In order to perform outdoor SLAM, laser sensors have
been widely used [10]. To provide localization and map
building, the input range data are processed using
geometric feature extraction and scan correlation
techniques. Less research exists using sensors such as
underwater sonar [8] and Frequency Modulated
Continuous Wave (FMCW) radar. Largearea SLAM has
already been carried out successfully for visionbased
underwater missions over long distances [11] [12].
Interestingly, a radar sensor has already been used by
Clark in [13] at the end of the last century. In an
environment containing a small number of well
separated, highly reflective beacons, experiments were
led with this sensor to provide a solution to the SLAM
problem [2] using an extended Kalman filter framework
and a landmarkbased approach. In [14], a model
dedicated to the radar was developed to build an
occupancy grid map. Finally, in [15], a method for
building a map with sensors returning both range
information and received signal power information was
presented. An outdoor occupancy grid map related to a
30mvehiclestrajectoryisanalysed.

A common method of pose estimation for mobile robots


isscanmatching[16].Bysolvingtherigidtransformation
between consecutive scans from a range sensor, the
robots motion in the time period between the scans can
beinferred.Oneofthemostpopularapproachesforscan
matching is the Iterative Closest Point (ICP) [17]
algorithm. In ICP, the transformation between scans is
founditerativelybyassumingthateverypointinthefirst
scan corresponds to its closest point in the second scan,
and by calculating a closed form solution using these
correspondences. However, sparse and noisy data, such
asthosefromanimagingradar,cancauseanICPfailure.
A single noisy reading can significantly affect the
computed transformation, causing the estimated robot
pose to drift over time. Yet, it is wellknown that the
differentpossiblecombinationsofradarsignalsintroduce
the speckle effect which can lead to ghost detections or
false disappearances. To overcome this problem, the
FourierMellin transform for registering images in a
sequenceisusedinourfirstapproach[18,19]toestimate
the rotation and translation of the radar sensor motion
(seesection4).InthecontextofscanmatchingSLAM,the
use of the FourierMellin transform is original and
provides an accurate and efficient way of computing the
rigid transformation between consecutive scans (see
section6).Itisaglobalmethodthattakesintoaccountthe
contributionsofboththerangeandpowerinformationof
theradarimage.Insomeways,thisglobalapproachis
alsoclosetothehistogramcorrelationtechnique[10].
2.2Egomotionestimation
Theestimationofavehiclesdisplacementoregomotion
is a widely studied problem in mobile robotics. Most
applications are based on proprioceptive data provided
by odometer sensors, gyrometers, IMU or other
positioning systems such as GPS [20]. But in order to
estimate motion, some research works tried to use only
exteroceptivedata.Thus,theauthorsof[2123]proposed
a visual odometry without proprioceptive data. Tipaldi
and Ramos [24] proposed filtering out moving objects
before doing egomotion. In such an approach, the
exteroceptive egomotion sensor is considered as
augmentingratherthanreplacingclassicalproprioceptive
sensors.Sometimes,classicaldisplacementmeasurements
are much more difficult and have limitations: inertial
sensors are prone to drift, and wheel odometry is
unreliableinroughterrain(wheelstendtoslipandsink)
2 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com
and as a consequence, visual odometric approaches are
widelystudied[25,26].Forexample,inanunderwateror
naval environment, classical egomotion techniques are
not suitable. In [27], an egomotion technique was
proposed based on visual SLAM fused with IMU. In
order to find displacement with exteroceptive sensors
such as range finders, the scan matching method is once
more commonly used [8, 28], but each scan is corrected
withproprioceptivesensors,especiallywhenthesensoris
slow. In all scan matching work, distortion is taken into
account but considered as a disturbance and thus
corrected.

The only work dealing with distortion as a source of


informationusedaspecificcamerawitharollingshutter.
In [29], AitAider et al. computed the instantaneous 3 D
pose and velocity of fast moving objects using a single
cameraimage,butintheircontextpriorknowledgeofthe
observed object is required. In mobile robotics, we have
no a priori about the surrounding environment of the
robot. To the best of our knowledge, there is absolutely
no work in the field of mobile robotics literature
considering distortion as a source of information for an
odometric purpose. So, the goal of this second approach
istostudyandusedatadistortionwithamobileground
based panoramic radar sensor in the context of field
robotics and to show its usefulness for localization and
mapping[30].
3.TheK2Piradar
The exploited radar uses the frequency modulation
continuous wave (FMCW) technique [9]. The FMCW
radariscalledK2Pi(2forpanoramicinKband)andis
equipped with a rotating antenna in order to achieve a
complete360

persecondmonitoringaroundthevehicle,
withanangularresolutionof3,inthe3100mrange.A
general view of the radar is presented in Fig. 1 and its
main characteristics are listed in Table 1. The example of
two radar images is presented in Fig. 2. Variations of
shading indicate variations of amplitude in the power
spectra.
4.FMTRSLAM:FourierMellin
transformbasedradarSLAM
Inordertoovercomethecomplexityofradarimageanalysis,
a trajectoryoriented EKFSLAM technique using data from
the360

fieldofviewradarsensorhasbeendeveloped.This
process makes no landmark assumptions and avoids the
data association problem. The method of egomotion
estimation makes use of the FourierMellin transform for
registering radar images in a sequence, from which the
rotation and translation of the sensor motion can be
estimated [31]. In the context of the scanmatching SLAM,
the use of the FourierMellin transform is original and
provides an accurate and efficient way of computing the
rigidtransformationbetweenconsecutivescans.
4.1RadarscanmatchingSLAM
The used formulation of our SLAM approach is to
estimate the vehicle trajectory defined by the estimated
state
0 k 1 k
T
T T T
v v v
x(k) x , , x , x ,

(
=

where
i
T
v i i i
x x , y ,| ( =

is the state vector describing the
location and orientation of the vehicle at time
i
t . There is
no explicit map; rather each pose estimate has an
associated scan of raw sensed data that can be next
alignedtoformaglobalmap.

Scanmatchingistheprocessoftranslatingandrotatinga
radarscansuchthatamaximaloverlapwithanotherscan
emerges.Assumingthedistributionofalignmenterrorsis
approximately Gaussian, a new vehicle pose is added to
the SLAM map by only adding the pose to the SLAM
statevector.

So, observations are associated to each pose. They are


comparedandregisteredtoofferpotentialconstraintson
theglobalmapofvehicleposes.Thisisnotonlyusefulfor
motionbased state augmentation, but it is also an
essentialpointforloopclosing.

The estimator used here is the extended Kalman filter


(EKF). Given a noisy control input (k 1) + u at time
k 1
t
+
measured from a gyrometer and odometers, upon
calculation of the new vehicle pose, x
n 1
v
(k 1|k),
+
+ and a
corresponding covariance matrix, P
n 1
v
(k 1|k),
+
+ the
predicted global state vector, x(k 1|k), + and
corresponding covariance matrix, (k 1|k), + P can be
augmentedasfollows:
( )
( )
( )
( )
( )
( )
( ) ( )
( ) ( )
( ) ( )
( ) ( )
0
1
n
n
n 1
n
n
n
n 1
n
v
v
v
v
v
T
v
v
T v
v
v
x
x
x k k
x k 1|k k 1|k ,
x k 1
x
x
x k 1
k k k k
x
k 1 k .
x k 1
k k k 1 k
x
+
+
(
(
(
(
(
( + = = +
(
+ (
(

(
(

(
c +
(
(
c
(
+ =
(
c +
(
+
(
c
(

u
u
P P
P
u
P P

The operator is the displacement composition operator


(e.g., [32]).
( )
n 1
v
k 1 k
+
+ P is the covariance of the newly
added vehicle state. Let us assume that two scans or
images,
i
I and
j
, I havebeenregistered.So,anobservation
i, j
T of the rigid transformation between poses and in the
state vector exists. Therefore, a predicted transformation
betweenthetwoposescanbefoundfromtheobservation
modelasfollows:
3 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com

Figure1.TheK2PiFMCWradar
Carrierfrequency F0 24GHz
Transmitterpower Pt 20dBm
AntennagainG 20dB
Bandwidth 250MHz
Scanningrate 1Hz
Angularresolution 3
Angularprecision 0.1
RangeMin/Max 3m/100m
Distanceresolution 0.6m
Distanceprecision(canonicaltargetat100m) 0.05m
Size(lengthwidthheight) 2724
30cm
Weight 10kg
Table1.CharacteristicsoftheK2PiFMCWradar

Figure2.Twoconsecutiveradarimages((a)&(b))obtainedwith
theFMCWradarsensor(seeFig.1)
( ) ( ) ( )
( ) ( )
( )
j i
i, j
v v
k 1 k h x k 1 k
x k 1 k x k 1 k
+ = +
= + +
T

wheretheoperator istheinversetransformationoperator.

Observations are assumed to be made according to a


modeloftheform:
( ) ( ) ( ) ( ) ( )
i, j i j
k 1 h x k 1 k 1 , , + = + + + = T I I
in which represents a registration algorithm, ( )
k 1 + a
vector of observation errors, and h the observation
function.Thestateupdateequationsarethentheclassical
EKF update equations. The search for a transformation
i, j
T is achieved by maximizing a cross correlation
function [33]. This process has been adapted in order to
alsoreturntheassociatedcovariance.
4.2FourierMellintransformforegomotionestimation
The problem of registering two scans in order to
determine their relative positions has to be solved. The
choiceofanalgorithmisstronglyinfluencedbytheneed
for realtime operation. A FFTbased algorithm was
chosentoperformscanmatching.

Fourierbased schemes are able to estimate large


rotations, scalings and translations. Let us note that the
scale factor is irrelevant in our case. Most of the DFT
based approaches use the shift property [31] of the
Fourier transform, which enables robust estimation of
translationsusingnormalizedphasecorrelation.

To match two scans which are translated and rotated


withrespecttoeachother,thephasecorrelationmethod
is used, stating that a shift in the coordinate frames of
twofunctionsistransformedintheFourierdomainasa
linear phase difference. To deal with the rotation as a
translational displacement, the images are previously
transformed into a uniform polar Fourier
representation.

It is known that if two images


1
I and
2
I differ only by a
shift, ( x, y), A A (i.e.,
2 1
(x, y) (x x, y y)), = A A I I then
theirFouriertransformsarerelatedby:
( ) x y
1 w x w y
2 x y 2 x y

(w , w ).e (w , w ).
A + A
= I I
Hence,thenormalizedcrosspowerspectrumisgivenby

( )
( ) x y
1 w x w y
2 x y
1 x y

(w , w )
Corr x, y e .

(w , w )
A + A
= =
I
I
(1)
TakingtheinverseFouriertransform:
( )

( ) ( )
( )

= = A A
1
x y
Corr x, y F Corr w , w x x, y y , d
which means that ( )
Corr x, y is nonzero only at
( )
( )
( ) { }
x,y
x, y argmax Corr x, y . A A = If the two images
differ by rotational movement ( )
0
q with translation
( )
x, y , A A then
2 1 0 0
0 0
(x, y) (xcos ysin x,
xsin ycos y).
= + A
+ A
I I q q
q q

Converting from rectangular coordinates to polar
coordinates makes it possible to represent rotation as
shift:theFouriertransforminpolarcoordinatesis
( ) ( )
( ) x y
i w x w y
2 1 0

, , .e .
A + A
= I I r q r q-q
4 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com
Introducing the magnitudes of
1

I and
2 1

I I and
2

I
respectively),theyarerelatedby

1 2 0

. I I , , q
The shift between the two images can now be resolved
using(1).
4.3Scanregistration
In order to perform a scan registration algorithm, the
FourierMellintransform(FMT)hasbeenchosen[34][31].
The FMT is a global method that takes the contributions
from all points in the images into account, in order to
provide a way to recover all rigid transformation
parameters,i.e.,rotation,translation.Itisanefficientand
accurate method to process a couple of images that are
fairlysimilar(seeFig.2).Thestepsofthescanregistration
algorithmarepresentedinAlgorithm1.

Input:Radarimages
k
I and
k 1
I
1.Apply the thresholding filter to eliminate the speckle noise
inbothimages
2.ApplytheFFTtoimages
k k

I I and
k 1 k 1


I I
3.Computethemagnitudes
k k 1

,

I I
4.Transform the resulting values from rectangular to polar
coordinates.

I I ,
5.Apply the FFT to polar images, a bilinear interpolation is
used.


I I , ,
6.Compute


Corr w , w

between

w , w

I , and


k 1

w , w

I , using(1)
7.Compute




1
Corr F Corr w , w

,
8.Findthelocationofthe maximumof

Corr , andobtain
therotationvalue
9.Construct Ir byapplyingreverserotationto
k 1
I
10.ApplytheFFTtoimage
k 1
Ir
11.Computethecorrelation


x y
Corr w , w using(1)
12.TaketheinverseFFT
Corr x, y of


x y
Corr w , w
13.return the values ( x, y) of the shift and the rotation
value
Algorithm1.Scanregistrationalgorithm
5.ROLAM:radaronlylocalizationandmapping
Inaroboticscontext,itisusuallyassumedthatthescanof
a range sensor is a collection of depth measurements
takenfromasinglerobotposition.Thiscanbedonewhen
working with lasers that are much faster than radar
sensors and can be considered instantaneous when
compared with the dynamics of the vehicle. However,
whentherobotismovingathighspeed,mostofthetime,
this assumption is unacceptable. Important distortion
phenomena appear and cannot beignored.In the case of
alaserrangefinderwitha75Hzscanningrate,distortion
exists but is ignored. This assumption is valid for low
speed applications, nevertheless still moving straight
aheadataspeedof5m/s,a7cmdistortioneffectappears.
At classical road vehicle speeds (in cities, on roads or
motorways) more important distortions can be observed.
Of course, the rotation of the vehicle itself during the
measurementacquisitionisanothersourceofdisturbance
which cannot be neglected for high speed displacement
or with slow sensors (cf. Fig. 3). Finally, let us note that
when the sensor is too slow, a stop & scan method is
oftenapplied[35].

For example, in the previous radar mapping application


(presented in section 4), the sensor delivers one
panoramicradarimagepersecond.Whentherobotkeeps
moving straight ahead, at a low speed of 5 m/s, the
panoramic image includes a 5metre distortion. We deal
with this distortion by using proprioceptive sensors in
ordertoestimatethetrajectoryandtobuildavirtualscan
taken from a single position. In fact, distortion is
considered as a noise and filtered out. In our second
approach, we use distortion as a source of information
regarding the displacement of the vehicle. The objective
here is not to remove distortion but to extract movement
informationfromit.
5.1Distortionproblemstatement
Distortionisthealterationofthedataasalsoobservedin
ourradarimages(cf.Fig.3).Witharotatingrangesensor,
thedesiredacquisitionshouldrepresentthesurroundings
oftherobotattime t. Whenthesensorsrotationspeedis
slowcomparedtothevehiclesmotion,thechangesinthe
robots location during the acquisition period lead to
image distortion. This distortion effect is presented on
realandsimulateddatainFig.3&4.

Without any prior knowledge of the environment, a


unique scan acquisition from a single sensor cannot give
information about distortion. This is one of the reasons
why distortion is usually considered as noise and
corrected by proprioceptive sensors. The aim of this
second approach is to measure the distortion of the data
bycomparingsuccessiveradarimagesinordertoextract
information about the robots movement. By using a
rotating sensor without any knowledge of the
environment shape, two successive observations are
required. The stated assumption is the local constant
velocity of the vehicle during two successive
measurements.Theposeofeachmeasurementisdirectly
linked to the observation pose and to the angle of
observation.Thisposecanbeexpressedwiththeconstant
5 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com
velocitymodelofthevehicleandisonlyafunctionofthe
linearandangularspeedoftherobot.
Datadistortionresultsfromcombinedsensorandvehicle
movements. The distortion equation can be represented
bytheparametricequationofatrochoidasshowninFig.
4(a).Indeed,attime t, thepositionofadetectiondoneat
range r is a function of the centre pose
( )
c c c
t t t
x , y ,| and of
thesensorbearing
t
. q

( )
( )

( ) ( )
( ) ( )
( )
( )
c
t
c
t
A)Centerpositionattimet
c c
t t
t
c c
t
t t
B)Centerrotation
C)Detectionpositionattimet
x t x
y t
y
cos sin
cos
.
sin
sin cos
| |
| |
( (
= ( (
( (

(

(
(
+ (
(
(

(

r q
r q

In order to obtain the centre pose at time t, an evolution


model taking into account the linear ( )
V and angular ( )
w
velocities is formulated. Pose
( )
c c c
t t t
x , y ,| is obtained as
follows:

c
t
c
t
c c
t 0
t
cos
x 2 2V t
sin ,
2 t y
sin
2
t. | |
( | |
( |
(
| |
\ . (
= (
|
(
| | ( \ .

(
|

( \ .

= +

w
w
w w
w
(3)
If based on this distortion parametric equation, the
measurements can be distorted or undistorted. The
equation is parameterized with respect to the linear and
angular velocities V and w. With a prior estimate of these
parameters, detection( )
, , q r done at time t in the sensor
frame,canbetransformedintotheworldframebasedon
(3)and(2)(cf.Fig.5).

Theobjectiveistoestimateproprioceptiveinformation,in
fact velocities V

and w,

which best undistort the


measurements.
5.2Proprioceptiveinformationextraction
According to the assumption that the velocity of the
vehicle is locally constant during two successive
measurements,theeffectivevelocitiesoftherobotwillbe
those that, using distortion formulation, allow
superimposingdetectionsofthesameobjectintheworld.

Figure3.Distorteddatafromarealsensor:radardata(a)withdistortioneffect(b)withoutanydistortioneffect

Figure4.(a)Distortionphenomenon:whenthevehicleismovingalongthegreentrajectory,thesensorisscanning.Thesensorbeams
arerepresentedinredandblueduringthefirstandsecondradaracquisitionrespectively.Thefirstandthelastbeamofeachacquisition
do not measure the same thing. Each scan is distorted by movement. (b) represents the desired acquisition from the second position
(orincaseofstop&scanmethod)withcorrectedbeam
i
j
t
z' fromscan j takenattimestamp
i
t and(c)theobtainedmeasurementswhen
movingwithequivalentuncorrecteddetection
i
j
t
z
6 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com

Figure5.(a)Dataobtainedinsensorframewithoutconsideringdistortion.(b)Undistorteddatabasedondistortionformulation
The pose of each measurement is directly linked to the
observation pose and to the angle of observation. This
posecanbeexpressedwiththeconstantvelocitymodelof
the vehicle and is only a function of the linear and
angularspeedsoftherobot,see(3).Let
1
and
2
bethe
landmarks representing the same point in the world
w

in their respective distorted scans. We can transform


1

and
2
into the undistorted world frame by using the
parameters(i.e.,linearvelocity V andangularvelocity w )
anddistortionfunctions f andg (cf.Fig.6).Bycomparing
the different projected poses in each acquisition, velocity
parameterscanbeextracted.Inordertoachievethistask,
thedataassociationbetweenimages1and2isrequired.

Because the radar data are submitted to different


perturbations (such as speckle, Doppler, antenna
aperture, etc.) the appearance of a detected object keeps
changingfromonescantoanother.Asaconsequence,no
descriptorisavailabletoperformdataassociation,soitis
based only on the Mahalanobis distance between
predictions and measurements. The function
1
h g f

= is
unknown because
1
g

cannot be obtained; consequently,


minimization techniques have to be used in order to
estimate
1

. Finally,eachassociationcangivenewvalues
ofthevelocityparameters.

The robot moves from an initial pose


0
T
v 0 0
[x , y ] = x with
aninitialorientation
0
| ataconstantvelocity
T
v
[V, ] e = V
during two successive sensor scans. Each landmark
M M T
i i i
[x , y ] = M observed at time
i
t is distorted by the
robots displacement. At this step, we suppose that
i
M is
the expression of the landmark after Doppler correction.
If
M M T
d d d
[x , y ] = M is the perturbed detection of
landmark
i
M at time
d
t , the correction is obtained as
follows:

( ) ( )
( )
( )
( )
2 2
sensor d M M
i d d
sensor d
sensor d
Vcos t
x y 2
cos t
sin t
| |
= + + |
|
\ .
(
(
(

M
w
a
l
w
w
(4)
where l is the wavelength of the radar signal, a a
coefficient which links frequency and distance, and
sensor
w therotatingrateofthesensor.IfnoDopplereffect
has to be considered, as is the case with laser sensors,
simplynotethat
i d
. = M M

So,
1
and
2
detected in their respective scans can be
propagated in the world frame by their two respective
propagationfunctions f andg.
( ) ( )
w 1 w 2
f , V, g , V, . = = M M and M M w w (5)
For the first radar image, the function f can be expressed
as:

( ) ( )
( ) ( )
0
0 1 0 1
w v 1
0 1 0 1
1
0
1
1
0
cos t sin t
x
sin t cos t
t
cos
2 t 2V
sin
2 t
sin
2
| |
| |
|
|
( + +
= + (
+ +
(

( | |
+
( |
| |
\ . (
+
|
(
| |
\ .
( +
|
(
\ .
M M
w w
w w
w
w
w w
(6)
with
( )
1 1
1
sensor
arctan y , x
t . =
w

Similarly, for the second scan ( )
w 2
g , V, = M M w can be
easily deducted with
( )
2 2
2
sensor
arctan y , x 2
t .
+
=
p
w
The
functionarctanisdefinedon ; . ( +

p p

The entire set of detections in the world frame can be


easily expressed in a matricial form based on (6). Based
on these equations we can conclude that distortion is
linkedtothevelocityparameters ( )
V, , e tothelandmarks
in the two successive scans
1
and
2
, to the initial pose
of the robot
0
v 0
( , ) | x and to the sensor scanning rate
sensor
. w But, in fact, the only parameters that need to be
estimated are the unknown velocities and consequently
thecurrentradarpose.
7 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com
5.3Estimationofvelocities
In order to estimate the velocity parameters
T
V, , the
data association between landmarks from the two
successivescansneedstobedone.
1

hastobepredicted
from
1
(in the first scan) onto the second scan. A
minimization technique is applied in order to calculate
the function
1 1

h( , V, ) M M because h cannot be
calculated directly. The cost function S for one landmark
isgivenby
2
w2 w1
( ) M M or:

2
1 1 1 1

S(x , y ) (g( , V, ) f( , V, )) . M M (7)


A gradient method with adaptive stepsizes is used to
minimizethiscostfunction.Asaresult,thepredictionofthe
firstradarimagelandmarkscanbecomputedinthesecond
image as well as its uncertainty ellipsis. Data association
between prediction
1

and landmark
2
is then calculated,
based on Mahalanobis distance criteria by taking into
accountuncertaintiesofmeasurementsandpredictions.

As radar data are very noisy (cf. Fig. 2), both landmark
extractionanddataassociationcanbefalse.Forexample,
the speckle effect can lead to some ghost detections or
false disappearances due to the different possible
combinationsofradarsignals.Moreover,duetomultiple
reflections, radar data are not as accurate as laser data.
Thus,allpossibledataassociationshavetobeconsidered.

Two assumptions are made at this point. First, the


percentage of detections due to static objects has to be
sufficient.Indeeddetectionscomingfrommovingobjects
can lead to false velocity estimates. Moreover, if all the
detections coming from moving objects are coherent and
give the same velocity estimate, the most pessimistic
requirement is that 50% of the detections in the
environmenthavetocomefromstaticobjects.

Figure 6. Principle of distortion analysis: the detected landmark in


eachscaninred,andthecorrespondinglandmarkintherealworld;
thepredicteddetectionposefromscan1ontoscan2ingreen

Figure7.Trajectoriesobtainedbythetwomethods.Inblack the
DGPS reference trajectory, in red (dotdashed curve) the
trajectory obtained by the FMTRSLAM approach and in blue
(stardashedcurve)theROLAMtrajectory
Second, the vehicle equipped with the radar sensor is
supposed to be moving during two consecutive
acquisitions at a constant velocity ( V and ). Actually,
whenthevehicleacceleratesordecelerates,theestimated
velocityobtainedwillbethemeanspeedofthevehicle.

For each data association allowed by the Mahalanobis


distance, a new estimate of the robots velocity is
computed and sent over to an extended Kalman filter
process. Then, updated speeds are projected into the
velocity space with their respective uncertainties. In this
space, the global coherence of the scene is achieved by
fusing all the consistent estimates. This fusion process is
done after removing the outliers (i.e., wrong detections
and associations) using a RANSAC process. Indeed, we
suppose that most detections are static and well
associated. The fusion uses the covariance intersection
(CI) [36] method in order to be more pessimistic in the
case of the presence of any residual wrong vote during
fusion.

Positioningerror(m) Mean Std


ROLAM 10.4721 11.2428
FMTRSLAM 12.9435 14.1574
Displacementerror(m) Mean Std
ROLAM 0.20261 0.69689
FMTRSLAM 0.092042 0.37894
Rotationalerror(deg) Mean Std
ROLAM 0.002779 0.14556
FMTRSLAM 0.02757 0.11151
Table2.Trajectoryevaluation

8 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com

Figure8.Translationaltransformationnormandestimationofrotationforbothapproaches.(a)Normofthetranslationaldisplacement
ateachstep,(b)normofthetranslationaldisplacementerror.(c)Valueofrotationateachstep,(d)errorinrotation

Figure9.(a)AerialGoogleview.(b)RadarmapobtainedbytheFMTRSLAMprocess.(c)RadarmapobtainedbytheROLAMprocess
9 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com
6.Experimentalresults
This section provides the experimental results of both
radarbased approaches. The radar and the
proprioceptivesensorsweremountedonautilitycar.The
experimental run that is presented was conducted in an
outdoor field, around the Auvergne Zenith car park (cf.
aerial view in Fig. 9(a)), with a semistructured
environment(buildings,trees,roads,roadsigns,etc.).The
K2Pi radar was on top of the vehicle, two metres above
the ground. A 2.5 km trajectory was travelled at a mean
speed of 30 km/h. In order to evaluate the performance
assessmentofeachapproach,aDGPSsystemwasusedto
provide a reference trajectory. Radar images as well as
the data obtained from the other sensors were recorded
andpostprocessedasexplainedpreviously.

The trajectories obtained by the FMTRSLAM and


ROLAM algorithms are presented, respectively in red
(dotdashedcurve)andblue(stardashedcurve),inFig.7
andthereferencetrajectoryisplottedinblack.Duetothe
factthat,withoutaloopclosure,theerroriscumulatedall
alongthetrajectory,thepositioningerrorisgrowingand
representsthedriftoftheprocess.

In order to obtain an efficient and correct evaluation of


the results, relative displacements between robot poses
are studied. Figure 8(a) represents the translational
transformation value for the two approaches and the
reference. Figure 8(b) is the corresponding estimation
error. As a result of the initial acceleration, a peak at the
beginning of the translational transformation error plot
can be observed for the ROLAM approach. This effect is
not so obvious with the FMTRSLAM method. The
angular comparison is described in Fig. 8(a)&(b). It may
be observed that at the given times 90s and 135 s, peaks
appear in the estimation of the rotation angle (and, to a
lesser extent, in translation). This error is due to driving
around the roundabouts at high speed, thus making the
assumption of constant angular velocity erroneous; as a
consequence, the estimates of angular velocity are not as
goodasexpectedatthisparticularstage.

In order to compare both results in a quantitative way,


different criteria have been considered. First, the global
positioning error, directly linked to the visual evaluation
of thetrajectory, is analysed.The cumulative error of the
dead reckoning method makes the final error with
respect to the ground truth trajectory equal to 65 and 48
metres with respectively FMTRSLAM and ROLAM
algorithmsafteratrajectoryof2.5km.Duetothefactthis
metrictakesintoaccountthecumulativeerrorswhichdo
not represent directly the quality of the estimates, two
metrics based on relative displacement are considered.
Successive translational and rotational transformations
areestimated.Allquantitativeresultsaresummarizedin
Table2.
The FMTRSLAM approach provides displacement
results with a mean error of 0.09 m and a standard
deviation of 0.38 m, while the ROLAM algorithm has a
mean error of 0.20 m with a 0.70 m standard deviation.
Considering rotation, the FMTRSLAM algorithm
obtains a mean error of 0.03 deg with a standard
deviation of 0.1 deg, while the ROLAM process gives a
zeromeanerrorwitha0.15degstandarddeviation.

Byanalysingthesevalues,wecanconcludethatboththe
FMTRSLAMandROLAMalgorithmsallowobtaininga
goodestimateofthevehicletrajectory,eveniftheFMTR
SLAM approach seems to provide more accurate
displacement estimates while the ROLAM technique
gives better results in rotation. Both methods produce
interestingtrajectoryresultswhichallowbuildingaradar
map by positioning each radar scan at its corresponding
estimated position. A radarmapping result based on the
FMTRSLAMprocess is given in Fig. 9(b). As illustrated
in Fig 9(c), a map can also be obtained by the ROLAM
algorithmiftheacquisitionposesofeachbeam,thathave
beenpredicted,arerecorded.Inthiscase,eachrawradar
spectrum is positioned and plotted in order to build a
map. In both cases, the maps reveal details of the
environment,butadifferenceappearsduetothefactthat
nofilteringisappliedinthesecondapproach.Ofcourse,
postprocessing could have been added in order to
remove these different perturbations (speckle effect,
Doppler, antenna aperture, etc.) and to improve the
appearanceofthemap.Wechosetopresenttheresultsin
this manner in order to highlight the difference between
both approaches: the first one is based on a scan
matchingtechnique,thesecondonedealswitheachradar
beamaspreviouslyexplained.

Considering time processing, the pose estimation with


MATLAB is done by ROLAM in an average processing
time of 26 ms with an average number of 10 considered
landmarks. One step of the FMTRSLAM process (scan
building, scan registration, prediction and update),
implemented in C/C++, is achieved in less than one
secondwithaQuadCoreIntelXeon(2.8GHz)withsixGo
DDR2 FBDIMM RAM (667 MHz). The processing time
increaseswhentryingtomanagehowtoclosetheloop,
but this case is not considered here. The radar frequency
being 1 Hz, both approaches can be implemented in real
time.
7.Conclusion
Two methods for computing a mobile robots
localization estimates in natural or seminatural
environments are presented using radar technology. By
using such a groundbased microwave radar sensor,
which is most unusual in mobile robotics, a trajectory
oriented SLAM process is first proposed. It is based on
10 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com
the FourierMellin transform and the extended Kalman
filter. The second approach presented is based on the
fact that when the robot moves at high speed while the
sensor is rotating, data are subject to distortion
phenomena. This second method computes the
velocimetry of the vehicle by explicitly analysing these
measurement distortions. Such radarbased odometry
doesnotuseanyproprioceptivesensors.Thispartofthe
presented work can be extended to any kind of non
instantaneous rotating sensor. Our two approaches
work without any knowledge of the vehicles
surroundings; not only do they allow estimating the
robotstrajectory,theyalsopermitbuildingamapofthe
environment.Anevaluationandacomparativestudyof
thetwoproposedmethodsonrealworlddataovera2.5
km course have been conducted, which have
demonstrated their feasibility and reliability for mobile
groundvehiclesathighspeed(30km/h).
8.Acknowledgements
ThisstudywassupportedbytheAgenceNationaledela
Recherche(ANRtheFrenchNationalResearchAgency)
(ANRImpalaPsiRobANR06ROBO0012).Theauthors
would like to thank the members of CEMAGREF (IRSTEA)
fortheirkindloanoftheradarsensorusedinthispaper.
This work has also been funded by the French
government research programme Investissements
davenir through the RobotEx Equipment of Excellence
(ANR10 EQPX44) and the IMobS3 Laboratory of
Excellence(ANR10LABX1601),bytheEuropeanUnion
through the programme Regional Competitiveness and
Employment20072013(ERDFAuvergneregion),bythe
Auvergne region and by the French Institute for
AdvancedMechanics.
9.References
[1] Tim Bailey and Hugh DurrantWhyte. Simultaneous
Localization and Mapping (SLAM): Part II. IEEE
RoboticsandAutomationMagazine,13(3):108117,2006.
[2] G. Dissanayake, P. Newman, H. F. DurrantWhyte, S.
Clark,andM.Csorba.ASolutiontotheSimultaneous
Localization and Map Building (SLAM) Problem.
IEEE Transactions on Robotics and Automation,
17(3):229241,2001.
[3] Hugh DurrantWhyte and Tim Bailey. Simultaneous
Localization and Mapping: Part I. IEEE Robotics and
AutomationMagazine,13(2):99110,2006.
[4] U. Frese. A Discussion of Simultaneous Localization
andMapping.AutonomousRobots,20(1):2542,2006.
[5] D. Hehnel, W. Burgard, D. Fox, and S. Thrun. A
highly efficient FastSLAM algorithm for generating
cyclic maps of largescale environments from raw
laser range measurements. In Proc. of the Inter. Conf.
onIntelligentRobotsandSystems,pages206211.IEEE,
2003.
[6] CharlesBibbyandIanReid.Simultaneouslocalisation
and mapping in dynamic environments (slamide)
withreversibledataassociation.InWolframBurgard,
Oliver Brock, and Cyrill Stachniss, editors, Robotics:
ScienceandSystems.TheMITPress,2007.
[7] J. Leonard, J. How, S. Teller, M. Berger, S. Campbell,
G. Fiore, L. Fletcher, E. Frazzoli, A. Huang, S.
Karaman,O.Koch,Y.Kuwata,D.Moore,E.Olson,S.
Peters, J. Teo, R. Truax, M. Walter, D. Barrett, A.
Epstein, K. Maheloni, K. Moyer, T. Jones, R.Buckley,
M. Antone, R. Galejs, S. Krishnamurthy, and J.
Williams. A Perception Driven Autonomous Urban
Vehicle. Journal of Field Robotics, 25:727774, October
2008. Issue 10, Special Issue on the 2007 DARPA
UrbanChallenge,PartIII.
[8] D. Ribas, P. Ridao, J.D. Tards, and J. Neira.
Underwater SLAM in a Marina Environment. In
ProceedingsofInt.Conf.onIntelligentRobotsandSystems,
pages14551460,SanDiego,USA,October2007.IEEE.
[9] M.O. Monod. Frequency modulated radar: a new
sensor for natural environment and mobile robotics.
Doctoraldissertation,ParisVIUniversity,1995.
[10] M. Bosse and R. Zlot. Map Matching and Data
Association for LargeScale Twodimensional Laser
Scanbased SLAM. Int. Journal of Robotics Research,
27(6):667691,2008.
[11] R. Eustice, H. Singh, J. Leonard, M. Walter, and R.
Ballard. Navigating the RMS Titanic with SLAM
Information Filters. In Proc. of Robotics: Science and
Systems (RSS), pages 5564, Cambridge, MA, USA,
June2005.TheMITPress.
[12] I.Mahon,S.B.Williams,O.Pizarro,andM.Johnson
Roberson. Efficient ViewBased SLAM Using Visual
Loop Closures. IEEE Trans. on Robotics, 24(5):1002
1014,October2008.
[13] S. Clark and G. Dissanayake. Simultaneous
localizationandmapbuildingusingmillimeterwave
radartoextractnaturalfeatures.InProc.ofInter.Conf.
on Robotics and Automation, pages 13161321, Detroit,
Michigan,USA,May1999.IEEE.
[14] A. Foessel, J. Bares, and W.R.L. Whittaker. Three
dimensional map building with MMW RADAR. In
The 3rd International Conference on Field and Service
Robots (FSR), page 6, Helsinki, Finland, 6 2001.
YleisjljennsPainnoprssi.
[15] J.Mullane,E.Jose,M.D.Adams,andW.S.Wijesoma.
Including Probabilistic Target Detection Attributes
Into Map Representations. International Journal of
RoboticsandAutonomousSystems,55(1):7285,2007.
[16] F. Lu and E. Milios. Robot Pose Estimation in
Unknown Environments by Matching 2D Range
Scans. Journal of Intelligent and Robotics Systems,
18:249275,1997.
[17] P.J. Besl and N.D. McKay.A Method for Registration
of3DShapes.InIEEETransactiononPatternAnalysis
andMachineIntelligence,volume14(2),pages239256,
USA,1992.IEEEComputerSociety.
11 Damien Vivet, Franck Grossier, Paul Checchin, Laurent Trassoudaine and Roland Chapuis:
Mobile Ground-Based Radar Sensor for Localization and Mapping: An Evaluation of two Approaches
www.intechopen.com
[18] P.Checchin,F.Grossier,C.Blanc,R.Chapuis,andL.
Trassoudaine. Field and Service Robotics Springer
TractsinAdvancedRobotics,volume62,chapterRadar
Scan Matching SLAM using the FourierMellin
Transform,pages151161.Springer,2010.
[19] F.Grossier,P.Checchin,C.Blanc,R.Chapuis,andL.
Trassoudaine. Trajectoryoriented EKFSLAM using
the FourierMellin Transform applied to Microwave
Radar Images. In IEEE/RSJ Inter. Conf. on Intelligent
RobotsandSystems(IROS),pages49254930,StLouis,
Missouri,USA,102009.IEEE.
[20] J. Borenstein, H. R. Everett, L. Feng, and D. Wehe.
Mobile robot positioning: Sensors and techniques.
JournalofRoboticSystems,14(4):231249,1997.
[21] Andrew Howard. Realtime stereo visual odometry
forautonomousgroundvehicles.InProc.ofInt.Conf.
on Intelligent Robots and Systems, pages 39463952.
IEEE,2008.
[22] Bernd Kitt, Andreas Geiger, and Henning Lategahn.
Visual Odometry based on Stereo Image Sequences
with RANSACbased Outlier Rejection Scheme. In
Intelligent Vehicles Symposium, pages 486492, San
Diego,USA,June2010.IEEE.
[23] David Nistr, Oleg Naroditsky, and James Bergen.
Visual odometry for ground vehicle applications.
JournalofFieldRobotics,23,2006.
[24] Gian Diego Tipaldi and Fabio Ramos. Motion
clustering and estimation with conditional random
fields. In Proceedings of the Inter. Conf. on Intelligent
Robots and Systems, (IROS 2009), pages 872877, St.
Louis,MO,USA,2009.IEEE.
[25] B.WilliamsandI.Reid.OnCombiningVisualSLAM
and Visual Odometry. In Proc. of Inter. Conf. on
Robotics and Automation, pages 34943500,
Anchorage,Alaska,USA,2010.IEEE.
[26] Alberto Pretto, Emanuele Menegatti, Maren
Bennewitz, Wolfram Burgard, and Enrico Pagello. A
VisualOdometryFrameworkRobusttoMotionBlur.
InInter.Conf.onRoboticsandAutomation,pages1685
1692,Kobe,Japan,2009.IEEE.
[27] Michael Jenkin, Bart Verzijlenberg, and Andrew
Hogue. Progress towards underwater 3D scene
recovery. In Proc. of the 3rd Conf. on Computer Science
and Software Engineering, pages 123128, Montral,
Quebec,Canada,2010.ACM.
[28] E. Olson. Realtime correlative scan matching. In
Proc. of Inter. Conf. on Robotics and Automation, pages
43874393.IEEE,2009.
[29] Omar AitAider, Nicolas Andreff, JeanMarc Lavest,
andPhilippeMartinet.SimultaneousObjectPoseand
Velocity Computation Using a Single View from a
Rolling Shutter Camera. In European Conf. on
ComputerVision,volume3952,pages5668.Springer,
2006.
[30] D. Vivet, P. Checchin, and R. Chapuis. Radaronly
Localization and Mapping for Ground Vehicle at
High Speed and for Riverside Boat. In Inter. Conf. on
Robotics and Automation (ICRA), pages 26182624, St
Paul,Minnesota,USA,52012.IEEE.
[31] B.S. Reddy and B.N. Chatterji. An FFTbased
Technique for Translation, Rotation, and Scale
Invariant Image Registration. IEEE Trans. on Image
Processing,3(8):12661270,August1996.
[32] D.M. Cole and P.M. Newman. Using Laser Range
Data for 3D SLAM in Outdoor Environments. In
Proc. of Inter. Conf. on Robotics and Automation, pages
15561563,Orlando,FL,USA,May2006.IEEE.
[33] P. Aschwanden and W. Guggenbuhl. Experimental
resultsfromacomparativestudyoncorrelationtype
registration algorithms. In Robust computer vision:
quality of vision algorithms, pages 268 289. W.
ForstnerandSt.Ruwiedel,editors,Wichmann,1992.
[34] Q. Chen, M. Defrise, and F. Deconinck. Symmetric
PhaseOnly Matched Filtering of FourierMellin
Transforms for Image Registration and Recognition.
IEEE Trans. Pattern Anal. Mach. Intell., 16(12):1156
1168,1994.
[35] Andreas Nchter, Kai Lingemann, Joachim
Hertzberg, and Hartmut Surmann. HeuristicBased
Laser Scan Matching for Outdoor 6D SLAM. In
Advances in Artif. Intellig. 28th German Conf. on AI,
pages304319.Springer,2005.
[36] S. Julier and J. Uhlmann. Using Covariance
Intersection for SLAM. Robotics and Autonomous
Systems,55(1):320,2007.

12 Int. j. adv. robot. syst., 2013, Vol. 10, 307:2013 www.intechopen.com