Anda di halaman 1dari 4

IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL.

10, 2011 179

Signal Modeling and Analysis of a Planar


Phased-Array FMCW Radar With Antenna Switching
Moon-Sik Lee, Member, IEEE

Abstract—A frequency-modulated continuous-wave (FMCW) of unknown parameters [15], [16]. Recently, some works re-
radar system with a planar phased array based on antenna garding CRLB expressions for estimating unknown parameters
switching is considered. The radar system consists of 1 linear using antenna arrays have been studied in [14] and [17]–[19].
antenna arrays where each linear antenna array consists of 2
In [17], analytic expressions of the CRLB of azimuth and ele-
receiving antennas. In each cycle, the receiving antennas on each
linear antenna array are switched to the single receiving channel vation angles using a planar antenna array of identical isotropic
in turn from the first antenna to the 2 th antenna. An accurate sensors are developed. The general CRLB expressions for the
and detailed system model for the radar system is addressed. The angle, velocity, and range using an antenna array are derived in
Cramér–Rao lower bound (CRLB) analysis on the accuracy of [18]. In [14], the expressions of the CRLB for estimating the un-
estimating the azimuth angle, elevation angle, velocity, and range known movement parameters using a switching-type linear an-
using the radar system is presented. tenna array are derived. The CRLB for the general case of mul-
Index Terms—Antenna switching, Cramér–Rao lower bound tiple constant-amplitude polynomial-phase sources is derived
(CRLB), frequency-modulated continuous-wave (FMCW) radar, in [19].
planar phased array. Theoretical work for estimating the angle (azimuth angle),
velocity, and range using a linear (one-dimensional) phased-
array FMCW radar with antenna switching has been proposed in
I. INTRODUCTION [14]. In [20], a 24-GHz linear phased-array FMCW radar system
with antenna switching has been developed to evaluate in prac-
tice the theoretical work addressed in [14].
I N RECENT years, frequency-modulated continuous-wave
(FMCW) radar systems have been used in a variety of appli-
cations such as automotive and airborne ones [1]–[3]. Compared
In this letter, we first suggest a planar (two-dimensional)
phased-array FMCW radar structure with antenna switching
to a pulse waveform implementation, an FMCW waveform ap- and present an accurate and detailed signal model for this radar
proach enables a simpler hardware architecture and a lower peak structure. Then, we derive the explicit CRLB expressions for
power output [4], [5]. Information on the range and relative ve- the accuracy analysis of estimating the azimuth angle, elevation
locity of multiple targets can be obtained from the range-depen- angle, velocity, and range using the switching-type planar
dent beat and relative velocity-dependent Doppler frequencies phased-array FMCW radar. We consider a uniform planar
acquired using an FMCW radar, respectively. array (UPA) with a rectangular configuration of antennas, as
Recent advanced radar systems use planar antenna arrays as shown in Fig. 1. The system consists of linear antenna ar-
an essential component of the system [6]–[8]. A planar array rays, where each linear antenna array consists of receiving
of receiving antennas enables the estimation of azimuth and antennas. The receiving antennas on each linear antenna array
elevation angles of targets. Since conventional multiple-channel are switched to the single receiving channel periodically once
antenna array systems use an independent receiver for each every cycle. Some numerical results of the CRLB for the target
antenna element, the hardware expense, complexity, and power parameters are given.
consumption of these systems are considerably high [9]. To
solve this problem, several simple structures with fewer re- II. SIGNAL MODEL
ceivers than antennas have been proposed for linear antenna We consider a UPA FMCW radar system with a rectangular
array systems in the past [10]–[14]. In particular, linear an- configuration of receiving antennas where all antennas
tenna array radar systems with antenna switching have been share a single transmitter radiating a linear frequency-modu-
addressed in [12]–[14]. lated (FM) signal, as shown in Fig. 1. Each antenna element is
The Cramér–Rao lower bound (CRLB) is frequently used to denoted by , where and .
evaluate the performance of an estimation algorithm since it Assume that the transmitted linear FM signal is of the form
gives a lower bound on the variance of any unbiased estimator
(1)
Manuscript received January 31, 2011; accepted February 15, 2011. Date of (2)
publication March 07, 2011; date of current version March 17, 2011. This work
was supported by the Korean Government (MOEHRD) under the Korea Re-
search Foundation Grant KRF-2007-611-D00018. where is the pulse period and is the linear time-varying
The author was with the Information Systems Laboratory, Department of phase with an initial frequency and a chirp rate . The chirp
Electrical Engineering, Stanford University, Stanford, CA 94305, USA. He is rate is given by , where is the bandwidth. The
now with the Wireless Telecommunications Research Department, Electronics
and Telecommunications Research Institute (ETRI), Daejeon 305-700, Korea pulses are transmitted starting at the time instants
(e-mail: moonsiklee@etri.re.kr). , , , where is
Digital Object Identifier 10.1109/LAWP.2011.2123074 the number of cycles. In each cycle, the antennas on each linear
1536-1225/$26.00 © 2011 IEEE
180 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 10, 2011

The echo signal is sampled and numerically


mixed with the complex conjugate transmitted signal ,
and then a frequency down-converted signal is given by

(6)

(7)

In order to represent (6) in the standard form of array pro-


cessing, let us denote

(8)

and rewrite (6) as

(9)

(10)
Fig. 1. System block diagram of the planar phased-array FMCW radar with
antenna switching. The system consists of N linear antenna arrays, where each
linear antenna array consists ofN receiving antennas that are switched to the Then, the system model can be expressed as
single receiving channel periodically once every cycle.
(11)

where is the noisy component of the output signal


of the antenna element denoted by at the th cycle and
the signal is given by

Fig. 2. Transmitted linear FM signals. The pulses are transmitted starting at the
u0 N
time instants (( 1) + n 0 Tu ; ; ;U n
1) , = 1 2 . . . , ; ; ;N
= 1 2 ... .
(12)
In each cycle, the antennas on each linear antenna array receive the signal in
turn from 1 to N . and assuming we find the steering vector
as

antenna array receive the signal in turn from 1 to , as shown


in Fig. 2.
We assume that the reference receiving antenna point is lo-
cated at which is also used as a transmitting
antenna point, as shown in Fig. 1. An echo signal from a single
target received by the antenna element denoted by at
the th cycle is of the form (13)

(3) The system model (11) in vector notation is rewritten as

(4) (14)

(5) where is the


1 array observation vector,
is the 1 steering
where is the attenuated amplitude of the received vector, and is the 1
signal, is the velocity of propagation, is the propa- vector of the sensor noise.
gation time of the return signal from the target, is the For periodical sampling, the sample time-instants inside of
time-delay of the wave propagation from the reference antenna each pulse are given by the variable , and its values are
to the antenna element denoted by , is the interelement
spacing, is the initial radial range to the target at time for (15)
the first transmitted pulse, is the radial velocity of the target,
is the azimuth angle of the target, and is the elevation angle where is the sampling period and is the total number of
of the target. observations in the pulse.
LEE: PLANAR PHASED-ARRAY FMCW RADAR WITH ANTENNA SWITCHING 181

There is a main difference between linear phased arrays taking into consideration only diagonal elements of this matrix
addressed in [14] and planar phased arrays developed in this inequality and defining the corresponding CRLBs as the diag-
letter in the used steering vector model. The steering vector onal elements of the matrix , we obtain
model (13) considered in this letter for the switching-type , , ,
planar phased array carries information about the azimuth and .
angle , the elevation angle , the velocity , and the range . The vector of unknown parameters is generally ,
We wish to note that in multidimensional problems of target and the corresponding FIM is , where
parameters (four parameters in our case), the accuracy and is the number of targets and
resolution performance depends on the difference between
targets measured in 4-D space . For example, one
is able to resolve very tight targets provided that there is a
sufficient difference between the targets at least in one of four
parameters. Therefore, 4-D approach enables better accuracy For the case where multiple targets exist, we consider the esti-
and resolution performance than any lower dimensional cases. mation accuracy for only the desired target and treat all other
targets as interferences. Therefore, since the expressions for the
III. CRAMÉR–RAO LOWER BOUND ANALYSIS expectation (18) and the covariance matrix (19) of the obser-
vations have the additive structure, the formulas for the single
In this section, we derive explicit CRLB expressions on the
target case can be used for the multiple-target calculations.
accuracy of estimating the target parameters using
Finally, we wish to note that the CRLB expressions ad-
the considered radar system. We assume that the attenuation
dressed in this letter are applicable for quite a general case of
parameter in (12) is replaced by , modeling
the switching-type planar configuration of receiving antennas
the time-varying echo-signal attenuation effects by random
provided that the corresponding steering vector model can be
. Then, the observation vector (14) is given by
derived.
IV. NUMERICAL EXAMPLES
(16)
In this section, we give some numerical results of the CRLB
It is assumed that the randoms and are complex
for the target parameters . We assume a UPA FMCW
zero-mean circular Gaussian processes, i.e., ,
radar of receiving antennas with a half-wavelength
. Then, the observation vector (16) is
interelement spacing, and a linear FM signal with GHz,
the complex multivariable with the expectation and the cir-
the bandwidth MHz, and the pulse period s.
cular white random noise component with the covariance matrix
A total of 100 observations are taken during the pulse period
as follows:
with the sampling frequency of 1 MHz.
(17) We generate the observations according to the attenu-
ated model (16), where we assume the random errors and
(18)
to be a white complex-valued circular Gaussian noise with
identically distributed independent real and imaginary parts:
(19) , , and ,
. We use the amplitude
Here, we wish to note that the vectors for different are of the signal equal to 2.
independent. Two moving targets with , ,
The vector of unknown real parameters for the probability km/h, m and , , km/h,
density (17) can be defined as m are considered. Therefore, quite tight targets are
used having very close values for all movement parameters.
(20)
In Fig. 3(a)–(d), the CRLBs are shown for azimuth angle ,
The elements of the Fisher information matrix (FIM) of a elevation angle , velocity , and range as a function of the
complex circular Gaussian process are number of antennas and signal-to-noise ratio (SNR)
given by [19], [21] in multiple-target environments, respectively. The results pre-
sented in Fig. 3 are given for cycles. The values of the
SNR are defined as . Fig. 3
demonstrates a clear improvement in the accuracy of the target
parameters with an increase in the number of antennas
(21) and SNR.
In Fig. 4(a)–(d), we plot the CRLBs for azimuth angle ,
elevation angle , velocity , and range as a function of the
The subindexes in (21) are replaced by the notations of the cor-
number of cycles and SNR in multiple-target environments,
responding parameters. Thus, the CRLB takes on the form of
respectively. We use and antennas. A similar
the matrix inequality
improvement to the results presented in Fig. 3 is shown in Fig. 4
(22) with respect to the number of cycles .
In conclusion of this section, we wish to note that it is hard
Since we are interested in the estimation accuracy of the target to describe effects about antenna switching or other practically
parameters , we can treat all others as nuisance. By related parameters in the signal modeling. However, indirectly,
182 IEEE ANTENNAS AND WIRELESS PROPAGATION LETTERS, VOL. 10, 2011

FMCW radar with a rectangular configuration of antennas and


an antenna switching has been considered as a system model.
Some numerical results of the CRLB for the target parameters
have been demonstrated. The CRLB formulas for the single-
target case can be extended into the multiple-target calculations.

ACKNOWLEDGMENT
The author would like to thank Prof. A. Paulraj at Stanford
University, Stanford, CA, for his valuable comments.

REFERENCES
[1] L. Giubbolini, “A multistatic microwave radar sensor for short range
anticollision warning,” IEEE Trans. Veh. Technol., vol. 49, no. 6, pp.
2270–2275, Nov. 2000.
[2] S. H. Jeong, J. N. Oh, and K. H. Lee, “Design of 24 GHz radar with
subspace-based digital beam forming for ACC stop-and-go system,”
ETRI J., vol. 32, no. 5, pp. 827–830, Oct. 2010.
[3] K. Yamamoto, K. Yamada, N. Yonemoto, H. Yasui, H. Nebiya, and
C. Migliaccio, “Millimeter wave radar for the obstacle detection and
warning system for helicopters,” in Proc. RADAR, Edinburgh, U.K.,
Oct. 2002, pp. 94–98.
Fig. 3. Cramér–Rao lower bound on (a) azimuth angle, (b) elevation angle, [4] M. E. Russell, A. Crain, A. Curran, R. A. Campbell, C. A. Drubin, and
(c) velocity, and (d) range for different numbers of antennasN 2N as a W. F. Miccioli, “Millimeter-wave radar sensor for automotive intelli-
function of SNR. gent cruise control (ICC),” IEEE Trans. Microw. Theory Tech., vol. 45,
no. 12, pp. 2444–2453, Dec. 1997.
[5] N. C. Currie and C. E. Brown, Principles and Applications of Mil-
limeter-Wave Radar. Norwood, MA: Artech House, 1987.
[6] H. L. Van Trees, Optimum Array Processing. New York: Wiley, 2002.
[7] H. Cao, L. Yang, X. Tan, and S. Yang, “Computationally efficient 2-D
DOA estimation using two parallel uniform linear arrays,” ETRI J., vol.
31, no. 6, pp. 806–808, Dec. 2009.
[8] L. Giubbolini, “A microwave imaging radar in the near field for anti-
collision (MIRANDA),” IEEE Trans. Microw. Theory Tech., vol. 47,
no. 9, pp. 1891–1900, Sep. 1999.
[9] J. D. Fredrick, Y. Wang, and T. Itoh, “A smart antenna receiver array
using a single RF channel and digital beamforming,” IEEE Trans. Mi-
crow. Theory Tech., vol. 50, no. 12, pp. 3052–3058, Dec. 2002.
[10] J. Sheinvald and M. Wax, “Direction finding with fewer receivers via
time-varying preprocessing,” IEEE Trans. Signal Process., vol. 47, no.
1, pp. 2–9, Jan. 1999.
[11] J. G. Worms, “RF direction finding with a reduced number of receivers
by sequential sampling,” in Proc. IEEE PAST, Dana Point, CA, May
2000, pp. 165–168.
[12] L. Yang, L. Liwan, P. Weifeng, C. Yaqin, and F. Zhenghe, “Signal pro-
cessing method for switch antenna array of the FMCW radar,” in Proc.
IEEE Radar Conf., Atlanta, GA, May 2001, pp. 289–293.
[13] Y. Asano, S. Ohshima, T. Harada, M. Ogawa, and K. Nishikawa, “Pro-
posal of millimeter-wave holographic radar with antenna switching,” in
IEEE MTT-S Dig., Phoenix, AZ, May 2001, pp. 1111–1114.
[14] M. S. Lee, V. Katkovnik, and Y. H. Kim, “System modeling and
Fig. 4. Cramér–Rao lower bound on (a) azimuth angle, (b) elevation angle,
U
signal processing for a switch antenna array radar,” IEEE Trans.
(c) velocity, and (d) range for different numbers of cycles as a function of Signal Process., vol. 52, no. 6, pp. 1513–1523, Jun. 2004.
SNR. [15] Q. Zou, Z. Lin, and R. J. Ober, “The Cramer–Rao lower bound for
bilinear systems,” IEEE Trans. Signal Process., vol. 54, no. 5, pp.
1666–1680, May 2006.
we can refer to the effects of practical parameters from the re- [16] A. N. Mirkin and L. H. Sibul, “Cramér–Rao bounds on angle estimation
sults in [20]. As shown in [20], measurement results demon- with a two-dimensional array,” IEEE Trans. Signal Process., vol. 39,
strate slightly lower accuracy performance than the numerical no. 2, pp. 515–517, Feb. 1991.
[17] H. Gazzah and S. Marcos, “Cramer–Rao bounds for antenna array de-
results. There are many reasons for this performance degrada- sign,” IEEE Trans. Signal Process., vol. 54, no. 1, pp. 336–345, Jan.
tion: switching speed of switch, mutual coupling between an- 2006.
tenna elements, a nonideal measurement environment, and am- [18] A. Dogandzic and A. Nehorai, “Cramér–Rao bounds for estimating
range, velocity, and direction with an active array,” IEEE Trans. Signal
plitude and phase differences between and outputs [20]. Process., vol. 49, no. 6, pp. 1122–1137, Jun. 2001.
However, this degradation is not very large, and therefore we [19] A. B. Gershman, M. Pesavento, and M. G. Amin, “Estimating parame-
can say that it might be possible to achieve in practice the per- ters of multiple wideband polynomial-phase sources in sensor arrays,”
formance of the theoretical work developed in this letter. IEEE Trans. Signal Process., vol. 49, no. 12, pp. 2924–2934, Dec.
2001.
[20] M. S. Lee and Y. H. Kim, “Design and performance of a 24-GHz
V. CONCLUSION switch-antenna array FMCW radar system for automotive applica-
In this letter, the explicit expressions for the CRLB on the tions,” IEEE Trans. Veh. Technol., vol. 59, no. 5, pp. 2290–2297, Jun.
2010.
accuracy of estimating the azimuth angle, elevation angle, ve- [21] P. Stoica and R. L. Moses, Introduction to Spectral Analysis. Upper
locity, and range of the single target have been derived. A UPA Saddle River, NJ: Prentice-Hall, 1997.