You are on page 1of 9

24 GHz Interferometric Radar for Road Hump Detections in

Front of a Vehicle

Suzanne Melo*, Emidio Marchetti**, Scott Cassidy**, Edward Hoare**, Antonella


Bogoni*, Marina Gashinova**, Mikhail Cherniakov**
*
TeCIP Institute, Scuola Superiore Sant’Anna
Pisa, ITALY
email: s.assisdesouzamelo@santannapisa.it
**
Microwave Integrated System Laboratory, University of Birmingham
Birmingham, UK

Abstract: This paper presents an interferometric radar system using SFCW to detect small
road obstacles at 24 GHz. Experimental results confirm the effectiveness of the method
employed to estimate the hump’s height, and it can be used in the context of driver
assistance or even driverless cars in a future.

1. Introduction
Obstacle detection is an essential task for driving assistance and autonomous ground vehicles
(AGV). An AGV can provide many advantages to society, including prevention of road
accidents due to human error, optimal fuel usage and convenience [1].
However, when trying to make a vehicle autonomously travel to a predefined destination,
several challenges need to be overcome. The first requirement is to know how the vehicle will
navigate to the predefined destination. The second requirement is to detect surrounding
environments to avoid a collision. The third requirement is to detect signs and obstacles on the
road, such as lanes, crosswalks, and speed humps [1].
Referred to the third task, traditional sensors are mostly based on techniques such as: video
camera [2], LIDAR [1], [3]–[5], mobile sensing technologies/GPS [6]–[8], and radar based
techniques [9]–[11]. Detecting small objects such as a speed-hump is a challenging task due to
its low height. Moreover, it is necessary to detect it in difficult visibility conditions. In this
situation, radar based techniques can provide this kind of obstacle detection even at night or
under adverse weather conditions [12].
One of the commonly used types of radar for object detection is the millimeter-wave linear
frequency modulated continuous wave (FMCW) radar. In the field of automotive applications,
the two standard carrier frequencies in use are 24 GHz and 77 GHz [13]. The main
characteristics that make this system suitable for this kind of application are: high sensitivity,
no range blind area, short wavelength and hence reasonably small antenna dimensions.
Recently, Interferometric Synthetic Aperture Radar (InSAR) systems using a forward-looking
technique have been proposed to tackle the problem of road object detection [14]. It can give
more information of harsh off-road environments than existing AGV based radars, by providing
a scattering image, and a digital terrain model (DTM) of the scene under observation.
In this work, we focus on the use of a 24 GHz real aperture radar combined with an
interferometric technique for detecting small obstacles such as speed humps in road scenes. The
use of this carrier frequency with an interferometric technique for road object height estimation
is explored for the first time, to the best of our knowledge. We take advantage of the high

The 19th International Radar Symposium IRS 2018, June 20-22, 2018, Bonn, Germany
1
978-3-7369-9545-1 ©2018 DGON
sensitivity of phase information to small height variations in order to obtain a high resolution
accuracy. The analysis is based on experimental results obtained through indoor measurements
using a configuration that approximates a road scenario in which the radar is mounted in front
of the vehicle.
This paper is organized as follows: section 2 describes the theory behind the interferometric
techniques used in our algorithms, section 3 presents the experimental setup and the description
of the methodology. Section 4 shows the results, and finally section 5 presents the conclusions.

2. Interferometric Processing for Height Detection


Interferometry in radar systems has been explored for many years especially for geodesy and
remote sensing. With this technique it is possible to obtain elevation information of a terrain
and produce high resolution topographic maps. It makes use of the phase information, which is
obtained through the interaction of two different signals to retrieve spatial information [15].
The basic geometry of the interferometric measurements is depicted in Fig. 1. We use a spatially
separated antennas configuration that uses a fixed transmitter and receiver and an additional
receiver position in above/under the transmitter. The distance between the two receivers is a
critical parameter in interferometry and its influence on system performance is a subject of
study in this paper. Therefore, different baselines will be investigated to find out what influence
the baseline has on the accuracy of the estimated height. Moreover, since the baselines are much
shorter compared to distance to the hump and hence, we can consider the setup as a monostatic
configuration, where transmitter, receiver and target area are completely aligned [16].
With respect to Fig. 1, θr is the incidence angles for the first receiver; Bn is the perpendicular
baseline from master receiver to the slave receiver; R is the distance from the antenna to the
target and h is the target height.

Figure 1. Backscattering geometry for interferometric data acquisition.

The interferometric phase for a point P with respect to a reference height is calculated according
to (1), where ΨP is the interferometric phase, λ is the wavelength, h is the height of the point
and Bn, R1, and θr are geometric parameters, respectively baseline, distance from the target to
the receiver, and incidence angle for the receiver [16].

2
2π Bn
ΨP = ⋅h⋅ (1)
λ R1 ⋅ sin θ r

If the phase varies more than 2π, ambiguities are produced. It is possible to determine if the
residual phase is likely to be wrapped, knowing approximately the height of the objects which
are going to be measured, using the following expression [16]:

R1 .λ
H amb = (2)
B v . sin 2 θ r

where Bv is the vertical baseline. If the height of the objects under measurement is above the
ambiguity height, the phase will have to be unwrapped.
3. Experimental Setup
The experimental setup block diagram is shown in Fig. 2.

Figure 2. Block diagram of the experimental setup.

Automotive radars often transmit an FMCW waveform, which can be approximated by a


Stepped Frequency Continuous-Wave (SFCW) signal. Therefore, a Rohde and Schwarz
ZVA67 vector network analyzer (VNA) is used to directly generate a SFCW signal at carrier
frequency of 24 GHz. Two horn antennas are connected to the VNA, one for transmitting (Tx)
and the other for receiving (Rx) the signals. In our experiments we used a repeat track
configuration, where one receiver is used to acquire data in two different positions, at different
times [17]. The antennas have about 24 dBi gain and about 10 degrees beam width each.
The sweep bandwidth (BW) was chosen to be 2 GHz, which gives a range resolution
ΔR = c / (2 ⋅ BW ), of 0.075 m. In order to be measured by the system, the speed hump must present
more than one scatterer in different range cells, which is assured with this range resolution.
The chosen BW is compatible with the maximum allocated BW at 24 GHz, i.e. 4 GHz. At the
standard automotive carrier frequency 77 GHz only 1 GHz BW is available, however, 4 GHz
BW is also available for 79 GHz sensors [18].
All the parameters used in the measurements are summarized in Table I.
3
Table 1 - Parameters used in the experiments

Parameters 24 GHz
Radar
Output Power 25 dBm
Sweep Bandwidth 2 GHz
No. of points 201
IF Bandwidth 100 Hz
Square Horn 20o
Antennas
Azimuth/Elevation
Beamwidth
Antennas gain 24 dBi

In the setup the Tx antenna was fixed on a platform. The Rx antenna was mounted on a vertical
linear positioner, which features micrometer precision and maximum total excursion of 10 cm.
The optimal baseline regarding the maximum height sensitivity was theoretically calculated
according to [19]:

−3a 2 d .BW + (3a 2 d ⋅ BW ) 2 + (a ⋅ b ⋅ c 2 − 4 ⋅ a3d ) ⋅ 2a ⋅ d ⋅ BW 2


Bopt = (3)
a ⋅ b ⋅ c 2 − 4a 3 ⋅ d

where:
a = c / ( R1 ⋅ λ ⋅ sin θr ),
b = 0.0833 cos 2 θ r + 0.36 ⋅ sin 2 θ r / tan 2 (θ r − α ),
d = ( R1 ⋅ λ ⋅ sin θ r / 4) 2 ,
λ is the transmitted wavelength, c is the speed of light in the air, α is the slope angle (which is
90 degrees in our case) and BW is the signal bandwidth.
Considering the parameters used in our experiments, the optimal baseline is approximately 3
cm, following (3).
Starting from this calculation and in order to investigate the effect of different baselines on the
height estimation accuracy, the linear positioner was moved from 0 to 10 cm, with 1 cm steps.
In this way, it was possible to simulate a virtual baseline between every two positions of the
receiver. For example, when the receiver antenna moved from 1 cm to 2 cm, it was possible to
create a virtual baseline of 1 cm, and so on.
The whole system was placed on a motorized horizontal linear positioner (motor rail) to
simulate an automotive scenario, where a car is approaching the road object under
measurement. The motor rail is 2 meters long, with millimeter positional accuracy. It was
moved from 0 to 1 meter with 10 cm steps. The movements of the linear positioner and of the
motor rail were independent. The vertical linear positioner completes its 10 cm excursion for
every position of the horizontal motor rail. Fig. 3 shows a picture of the experimental setup.
The left picture shows the front view of the experiment with labels indicating each component
in the setup. The right picture is a side view showing the complete horizontal motor rail.

4
Figure 3. Experimental setup: (left) front view of the system, (right) side perspective view of the experiment.

The experiments were carried out in an indoor environment with two different targets: two
spheres with 20 cm height difference and a speed hump of 8 cm height. Fig. 4 depicts the targets
used in the experiments (spheres and speed hump). As shown in the picture we also used
absorbers around the targets to minimize the impact of external sources of reflection. In order
to have the targets completely illuminated by the antennas’ beam, the range to target is 3 meters,
with a vertical offset of 0.5 m.

Figure 4. Targets used in the experiments (left) two spheres, (right) speed bump.

A preliminary experiment was carried out with the two spheres in order to validate the
measurement methodology and find the optimal baseline. The spheres behave as point-like
reflectors and therefore are phase stable targets [20]. They were placed at two different heights
and in two different range cells in order to obtain two clear peaks in the range profile.
After validating the experimental setup with the spheres, a scene with a speed hump was
created. The interferometric algorithm outputs the height of the objects, and in the case of the
spheres, the difference in height between them.

4. Results
The algorithm used to provide the results is as follows: after calculating the IFFT of the stepped
frequency signal to convert the signal from frequency to time domain, a technique to extract
the peaks where the targets are located in the range profile is employed. The phase is then
measured in the positions of the peaks and the height estimated using (1).
The range profile for the two spheres and for the speed hump can be seen in Fig. 5. The graphs
show the measurements of the target (blue curve) plotted against the background, i.e. the scene
without the targets (red dashed curve). It is possible to clearly identify where the target’s

5
scatterers are located. For the two sphere case, we can see one located around 3.7 m and the
other one located at 4.2 m, corresponding to their real separation, which is 50 cm. The speed
hump instead, presents closer peaks, 15 cm apart, one around 4.1 m, and the other one around
4.25 m, which corresponds respectively to the very beginning of the speed hump (lowest point)
and the highest point of the speed hump. In a real scenario, the interferometric algorithm should
take into consideration the whole scene under observation, creating a map of the area. Here, our
only purpose is to identify a single object and evaluate the effectiveness of the height estimation.

Figure 5. Range Profile for (a) two spheres (b) speed bump.

The aim of the first step was to find the optimal baseline, by carrying out a set of measurements
using the metal spheres. The height difference between the two spheres from the center of one
sphere to the center of the other sphere was 0.20 m. 10 different measurements were taken for
each position of the vertical linear position. It was found that the correct height can be measured
using any baseline between 0.02 m to 0.08 m, this result is in line with the expected value
calculated using (3).
Since the reflections from the target can come from different scatterers when observed from
different positions of the antennas, we developed an algorithm that takes into consideration the
heights obtained for all positions assumed by the vertical linear positioner and the horizontal
motor rail and gives the final height estimation by averaging all the results. The horizontal rail,
was moved from 0 cm to 0.9 m, in 10 cm steps, totalling 10 different positions. At each position
the vertical linear positioner assumes 7 different positions (from 2 cm to 8 cm, in 1 cm steps),
6
simulating 70 different baselines. In total there are 700 different measurements, which are used
for computing the height estimation according to (4):

 N − b i −1 
B 

 ΨP 2 (h i ) − ΨP1 (h i ) 

1
H= 
 h i =1
B b =1 N − bi − 1


(4)
i
 
 
 

where B is the total number of baselines, N is the total number of measurements, ΨP1 is the
interferometric phase for first scatterer, and ΨP2 is the interferometric phase for second scatterer.
The height estimation results for both targets, at different positions of the horizontal motor rail,
are depicted in Fig. 6. Fig. 6(a) depicts the graph of the two spheres, while the graph of fig. 6(b)
shows the height estimation for the speed hump. Both graphs show the height measured by the
system (y axis) for every position of the horizontal motor rail (x axis). The mean value across
all measurements is plotted in red.

Figure 6. Height estimation for (a) spheres (b) speed bump.

For comparison purposes, the targets were first measured with a ruler. For the speed hump, the
height measured was 8 cm. As we can see in the results, the height estimation obtained with the
interferometric technique was very close to these values. For the spheres the height measured
with the system was 20.7 cm, while for the speed hump it was 7.9 cm.

7
The experiment was repeated many times, giving very similar values for each of them,
confirming the effectiveness of the method employed.

5. Conclusions
This paper shows that an interferometric radar system operating at 24 GHz could be used for
automotive environment awareness in a context of unmanned ground vehicle applications.
The results show that the system can measure the height of small road objects, which tend to
be some number of centimeters high, with good precision using an interferometric technique
allied with SFCW modulation.
The experiments were performed at 24 GHz using two horn antennas for transmitting and
receiving, respectively. We used a vertical linear positioner in order to create many virtual
baselines in the processing, and a horizontal motor rail to measure the targets from different
range positions. Therefore, a model of an automotive scene was created in an indoor
environment. This technique could be used for autonomously driven car applications, as it could
detect small objects on the road and avoid vehicle damage or more serious accidents.

References
[1] J. Choi et al., “Environment-Detection-and-Mapping Algorithm for Autonomous Driving in
Rural or Off-Road Environment,” IEEE Trans. Intell. Transp. Syst., vol. 13, no. 2, pp. 974–982,
Jun. 2012.
[2] L. Rogozea and WSEAS (Organization), Eds., Advances in mathematical and computational
methods: 12th WSEAS international conference in mathematical and computational methods
in science and engineering (MACMESE ’10): University of Algarve, Faro, Portugal, November
3-5, 2010. Greece: WSEAS, 2011.
[3] C. Fernández et al., “Free space and speed humps detection using lidar and vision for urban
autonomous navigation,” in Intelligent Vehicles Symposium (IV), 2012 IEEE, 2012, pp. 698–
703.
[4] A.-V. Vo, L. Truong-Hong, and D. F. Laefer, “Aerial laser scanning and imagery data fusion
for road detection in city scale,” 2015, pp. 4177–4180.
[5] T. Wu, H. Cui, Y. Li, W. Wang, D. Lui, and E. Shang, “A feature matching and fusion-based
positive obstacle detection algorithm for field autonomous land vehicles,” Int. J. Adv. Robot.
Syst., vol. 14, no. 2, p. 172988141769251, Apr. 2017.
[6] Y.-A. Daraghmi and M. Daadoo, “Intelligent Smartphone based system for detecting speed
bumps and reducing car speed,” in MATEC Web of Conferences, 2016, vol. 77, p. 09006.
[7] Y.-T. Chuang, C.-W. Yi, T.-H. Wu, and C.-Y. Chan, “Model the vibrations of rack-mounted
on-vehicle devices by underdamped oscillations,” in Intelligent Transportation Systems (ITSC),
2014 IEEE 17th International Conference on, 2014, pp. 982–987.
[8] C.-W. Yi, Y.-T. Chuang, and C.-S. Nian, “Toward Crowdsourcing-Based Road Pavement
Monitoring by Mobile Sensing Technologies,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 4,
pp. 1905–1917, Aug. 2015.
[9] J. Wang, Q. Song, Z. Jiang, and Z. Zhou, “A novel InSAR based off-road positive and negative
obstacle detection technique for unmanned ground vehicle,” in Geoscience and Remote Sensing
Symposium (IGARSS), 2016 IEEE International, 2016, pp. 1174–1177.
[10] Z. Jiang, J. Wang, Q. Song, and Z. Zhou, “Negative obstacle sensing based on real data of ultra-
wideband SAR,” 2015.
[11] J. Guo, Y. Luo, Y. F. Sui, D. Q. Yan, and Q. Zhong, “Airport Runway FOD Detection from
LFMCW Radar and Image Data,” Appl. Mech. Mater., vol. 719–720, pp. 1124–1131, Jan. 2015.

8
[12] J. S. Jung, J. M. Bak, H. J. Yang, Y. H. Seo, and Y. K. Kwag, “Obstacle detection radar system
for highway safety,” in Synthetic Aperture Radar (APSAR), 2011 3rd International Asia-Pacific
Conference on, 2011, pp. 1–4.
[13] J. Wenger, “Automotive mm-wave radar: status and trends in system design and technology,”
1998, vol. 1998, pp. 1–1.
[14] Z. Jiang, J. Wang, Q. Song, and Z. Zhou, “Off-road obstacle sensing using synthetic aperture
radar interferometry,” J. Appl. Remote Sens., vol. 11, no. 1, p. 016010, Jan. 2017.
[15] S. N. Madsen and H. A. Zebker, “Imaging radar interferometry,” Princ. Appl. Imaging Radar
Man. Remote Sens., vol. 2, pp. 359–380, 1998.
[16] S. Duque, P. Lopez-Dekker, and J. J. Mallorqui, “Single-Pass Bistatic SAR Interferometry
Using Fixed-Receiver Configurations: Theory and Experimental Validation,” IEEE Trans.
Geosci. Remote Sens., vol. 48, no. 6, pp. 2740–2749, Jun. 2010.
[17] P. A. Rosen et al., “Synthetic aperture radar interferometry,” Proc. IEEE, vol. 88, no. 3, pp.
333–382, Mar. 2000.
[18] J. Hasch, E. Topak, R. Schnabel, T. Zwick, R. Weigel, and C. Waldschmidt, “Millimeter-Wave
Technology for Automotive Radar Sensors in the 77 GHz Frequency Band,” IEEE Trans.
Microw. Theory Tech., vol. 60, no. 3, pp. 845–860, Mar. 2012.
[19] X. Huaping, Z. Yinqing, and L. Chunsheng, “Analysis and simulation of spaceborne SAR
interferometric baseline,” in Radar, 2001 CIE International Conference on, Proceedings, 2001,
pp. 639–643.
[20] C. Werner, U. Wegmuller, T. Strozzi, and A. Wiesmann, “Interferometric point target analysis
for deformation mapping,” in Geoscience and Remote Sensing Symposium, 2003. IGARSS’03.
Proceedings. 2003 IEEE International, 2003, vol. 7, pp. 4362–4364.
[21] X. Mustermann, Y. Wellknown, and Z. Pfiffig, “New research in radar field”, Proc. of 2010
Very Important Radar Conference, Bigcity, Incountry, June 2010, vol. 3, pp. 210-214.
(Times New Roman, 11 pt)
[22] H. R. Hertz, “Ueber sehr schnelle electrische Schwingungen“, Annalen der Physik, May 1887,
vol. 267, no. 7, pp. 421-448.
[23] J. C. Maxwell, A Treatise on Electricity and Magnetism, 3rd ed., vol. 2. Oxford: Clarendon,
1892, pp. 68–73.

You might also like