You are on page 1of 4

Proceedings of the 11th European Radar Conference

Resolution of Multiple Unresolved Targets via Dual


Monopulse with Array Radar
Jiani Wu∗ , Zhenhai Xu, Ziyuan Xiong, Bin Rao, Huimei Wang, Yongguang Chen†
∗ State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information System,
National University of Defense Technology, Changsha, China, 410073.
Email: tuotuonini@163.com
† Beijing Institute of Tracking and Telecommunications Technology, Beijing, China, 100094.

Email: ygchen@netease.com

Abstract—In the antimissile context, it is important to detect it should be noted that the conventional monopulse technique
and resolve the unresolved targets in the middle of the ballistic is not applicable straightforwardly to the case when there are
flight. As the targets in group are very close to each other, the two unresolved targets in a single beam. Researchers have
traditional techniques are difficult to resolve them. In this paper,
we study the problem of grouped target resolution in the angle made a lot of efforts on improved monopulse techniques [3-8].
dimension with array radar. We first establish an echo model of In order to resolve the Swerling II targets, Blair and Brandt-
the array radar when there are two unresolved targets in one Pearce made use of the distribution of the real and imaginary
beam. Then we derive an improved monopulse method based parts of the monopulse ratio, and proposed a method based on
on the maximum likelihood estimation principle, which realizes generalized maximum likelihood (ML) ratio testing according
angle resolution based on a dual monopulse system. Finally, the
performance of this method is evaluated by simulations in the to the Neyman-Person rule [3-5]. Sinha and Bar-Shalom then
cases of varying SNR, inter-target angle separation and inter- developed the maximum likelihood angle estimation method
target phase/amplitude differences. The simulation results prove to resolve two Rayleigh targets with monopulse radar [6]. Af-
that the method performs very well in many aspects, including ter that, Zhang exploited two target echo samples to proposed
smaller estimation error and enhanced adaptation to inter-target a maximum likelihood method for target parameter estimation
amplitude difference.
Index Terms—Two Unresolved Targets; Maximum Likelihood
[7], which requires the target echo energy to distribute over
Estimation; Monopulse; Dual Monopulse System the two adjacent samples. Zhou LanFeng researched how the
RCS fluctuation of targets affects monopulse radar system and
I. I NTRODUCTION addresses a method of resolving two unresolved targets based
on the statistical feature of monopulse angular estimation for
The middle of the ballistic flight is the critical stage slow fluctuation targets [8]. However, none of these methods
for the antimissile interceptor to complete its tactical tasks. have referred to and made use of the advantages of the array
However, the interceptor has to cope with the complex en- radar.
vironments composing of real warheads, various light and This paper first establishes an echo model of two tar-
heavy bait (dummy warheads, balloons, etc.), chaff, active gets received by the array radar, and then develops a dual
jammers and the abandoned of boost phase. Moreover, the monopulse method to estimate the direction-of-arrival (DOA)
true and disguised targets are grouped together and they of the targets according to the maximum likelihood estimation
have similar ballistic movement characteristics, which makes principle. The method is expected to gain advantages in DOA
ballistic target recognition even more difficult. In the context estimation accuracy and moderate requirement on the system
of ballistic missile defense, if the targets are not detected and complexity as it is based on the dual monopulse system. The
resolved correctly, undesirable consequences will be produced core idea of the proposed method is that, two monopulse
[1]. For example, for the tracking and guidance radar, unre- systems are formed and each of them points at one of the
solved targets will introduce deviation to the radar observation targets, and iterations are then carried out to reduce the DOA
orientation, which probably results in interception failure and estimation errors. In the Section V, the performance of this
target loss [2]. Therefore, the resolution of ballistic targets is method is demonstrated via Monte Carlo simulations.
a key technical problem that should be solved.
II. T WO - TARGET ECHO MODEL OF ARRAY RADAR
The grouped targets may be resolved in the distance di-
mension, angle dimension or Doppler dimension, and among Consider the uniform linear array consisting of N elements,
them, the angle information is of great importance, especially the measurement of a single echo received by the array in the
when the targets lie in the same rang cell and have similar case of two targets is
velocities. Therefore, we seek to resolve those targets in x=A1 s(u1 )+A2 s(u2 )+n (1)
the angle dimension in this paper based on the monopulse
where
array radar, which has lots of advantages such as high angle (N −1)dπ (N −3)dπ (N −3)dπ (N −1)dπ
measuring accuracy and simple system construction. However, s(ui )=[e−j λ ui
,e−j λ ui
,...,ej λ ui
,ej λ ui
]T

978-2-87487-037-8 © 2014 EuMA 273 8-10 Oct 2014, Rome, Italy


Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:53:34 UTC from IEEE Xplore. Restrictions apply.
is the array steering vector,ui =sinθi , i=1,2, θ1 and θ2 are where P s1 ⊥ =I−s1 s1 H and P s2 ⊥ =I−s2 s2 H are the orthog-
the angles of the two targets with respect to the normal of the onal projection matrices with respect to s1 and s2 , respec-
array.λ is the wavelength, d is the distance between adjacent tively, which satisfy
array elements. A1 ,A2 ∈C1×1 are the complex amplitudes of
kP s1 ⊥ s2 k2 =kP s2 ⊥ s1 k2 =1−|s2 H s1 |2 (11)
the two target echoes, respectively. n∈CN ×1 represents the
thermal noise received by array, which is independently and Moreover,
identically distributed white Gaussian noise with zero mean H H
and a variance of σ 2 . (P s2 ⊥ s1 ) (P s1 ⊥ s2 )
wΣ1 = 2, wΣ2 = 2 (12)
kP s2 ⊥ s1 k kP s1 ⊥ s2 k
III. M AXIMUM LIKELIHOOD DOA ESTIMATION
According to the distribution of the noise in (1), one can It can be easily obtained that wΣ1 H s1 =1,wΣ1 H s2 =0 and
conclude in the following Gaussian density function of the wΣ2 H s1 =0,wΣ2 H s2 =1 .
array measurement with respect to A1 ,A2 ,u1 ,u2 , Thus Â1 ,Â2 can be deemed as the sum beams of the
monopulse system I and II, respectively, which are denoted
H
p(x;u,a)=[1/(πσ 2 )]N exp{−[1/σ 2 ](x−Sa) (x−Sa)} as Σ1 ,Σ2 in the rest of the paper. wΣ1 ,wΣ2 are the weight
(2) vectors of Σ1 ,Σ2 . The formulation of the sum beams helps
The maximum likelihood estimates can then be obtained by to enhance the target signal of interest and suppress the signal
maximizing the above formula, which equals to minimizing of the other target at the same time.
H According Eq.(5) and (6), one can obtain that,
(x−Sa) (x−Sa). For notational convenience, we normal-

ize x as √1N x= √1N (A1 s(u1 )+A2 s(u2 )+n)=A1 s1 +A2 s2 + H H
∆1 =s01 x−A2 s01 s2 =s01 (x−A2 s2 )
H
(13)
h i

n0 , in which s1 s2 = √1N s(u1 ) √1N s(u2 ) =
 
H H H
S and n0 = √1N n . Denote Q=(x−Sa)H (x−Sa) ,then ∆2 =s02 x−A1 s02 s1 =s02 (x−A1 s1 ) (14)
A1 ,A2 ,u1 ,u2 can be estimated by setting the partial differ- Substituting Eq.(9) and (10) into Eq.(13) and (14) respectively
entiations of Q with respect to them to 0, i.e., yields
∂Q/∂A1 =−s1 H x+A1 +A2 s1 H s2 =0 (3)
!
H
H (P s1 ⊥ s2 ) x
∂Q/∂A2 =−s2 H x+A1 s2 H s1 +A2 =0 (4) ∆1 =s01 x− 2 s2 =w∆1 x (15)
kP s1 ⊥ s2 k
0H 0H
∂Q/∂u1 =2Re[−AH H
1 s1 x+A1 A2 s1 s2 ]=0 (5) !
H
H (P s2 ⊥ s1 ) x
0H 0H ∆2 =s02 x− s1 =w∆2 x (16)
∂Q/∂u2 =2Re[−AH H
2 s2 x+A2 A1 s2 s1 ]=0 (6) kP s2 ⊥ s1 k
2

which conclude in the following estimates according to (3) where


and (4), H
!
H s2 (P s1 ⊥ s2 )
s1 H x−s2 H xs1 H s2 w∆1 =s01 I− 2
Â1 = (7) kP s1 ⊥ s2 k
1−||s1 H s2 ||2 H
H s0 1 s2
=s01 − 2 (P s1 ⊥ s2 )
H
(17)
s2 H x−s1 H xs2 H s1 kP s1 ⊥ s2 k
Â2 = (8)
1−||s2 H s1 ||2 !
H
However, it is difficult to obtain the expressions of û1 ,û2 H s1 (P s2 ⊥ s1 )
w∆2 =s02 I− 2
straightforwardly based on these equations [9], so we propose kP s2 ⊥ s1 k
a dual monopulse method for angle estimation following H
H s0 2 s1
the guidelines of the above derivations. In the method, two =s02 − 2 (P s2 ⊥ s1 )
H
(18)
monopulse systems are constructed and the angles of the two kP s2 ⊥ s1 k
unresolved targets are estimated via iterations. It can be easily concluded that w∆1 H s1 =0 , w∆1 H s2 =0 ,
IV. D UAL - MONOPULSE - BASED DOA ESTIMATION METHOD w∆2 H s2 =0 , w∆2 H s1 =0
Thus ∆1 ,∆2 can be deemed as the difference beams of
It can be concluded from Eq. (7) and (8) that,
monopulse system I and II respectively, and w∆1 ,w∆2 are
H
s1 H x−s2 H xs1 H s2 (P s2 ⊥ s1 ) x the weight vectors of ∆1 ,∆2 . The difference beams of the
Â1 = = 2 =w Σ1 x (9)
1−||s1 H s2 ||2 kP s2 ⊥ s1 k two systems help to suppress the jamming signal. The iterative
formulas of monopulse system I and II are
H
s2 H x−s1 H xs2 H s1 (P s1 ⊥ s2 ) x
   
∆1 ∆2
Â2 = = 2 =w Σ2 x (10) u1t =u10 +k −1 , u2t =u20 +k −1 (19)
1−||s2 H s1 ||2 kP s1 ⊥ s2 k Σ1 Σ2

274
Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:53:34 UTC from IEEE Xplore. Restrictions apply.
where k is the slope of the monopulse response curve (MRC) |θ1 −θ2 |/θ3dB , the unitary root mean square error (URMSE)
and its value is derived according to literature [10] when the is URMSE=RMSE/θ3dB .
angle estimation error is very small. The formula for setting The influences of SNR, the azimuth and the correlation of
the value of MRC is the two target echoes on the performance of the proposed
method are studied in the following simulations. The corre-
H H lation of the two target echoes is defined as the amplitude
k=[s00 (u) s(u)+s(u) s00 (u)]/sH (u)s(u)
ratio and the phase difference of two target echo amplitudes,
N −1
X 2 
d
2 namely A2 =ρA1 ej∆φ with ρ being the amplitude ratio and
=8 πn( ) /N (20) ∆φ the phase difference. When the influence of a certain
λ
n=− N −1
2
factor on the DOA estimation performance is studied, the
other parameters are fixed on typical values. In each situation,
According to the above derivations and Eq.(19), the dual
1000 Monte Carlo simulations are carried out to obtain the
monopulse system for angle estimation can be constructed
statistical performances.
as that shown in Fig.1.
A. Relationship between the DOA Estimation Accuracy and
the SNR
In this experiment, we assume the radian of the two target
azimuths to be θ1 =−0.007,θ2 =0.009 , respectively, and the
corresponding unitary azimuths are θ̄1 =−0.4,θ̄2 =0.5. Fig.2
shows the relationship curve between the URMSE of the two
targets and the SNR with SNR=SNR1 =SNR2 .

Fig. 1. The process of the dual-monopulse-based DOA estimation method

According to Eq.(19) , this system estimates the directions


of the two unresolved targets based on the current beam
pointing orientations. Then the beam orientation is updated
according to the refreshed DOA estimates. The iteration is
terminated when a predefined convergence criterion is sat-
isfied. In this paper, the convergence criterion is set when
Fig. 2. Relationship curve between URMSE and SNR
the difference between the new DOA estimates and the beam
pointing orientations is less than a certain iterative threshold Fig.2 shows that the RMSE decreases and the angle esti-
in a single cycle. In Fig.1, u10 ,u20 are the sine values of the mation accuracy improves with increasing SNR.
current beam pointing angles and u1t ,u2t are the sine values
of the DOA estimates of the two targets, γ is the iterative B. Relationship between the DOA Estimation Accuracy and
threshold. Inter-Target Angle
V. S IMULATION In this experiment, we assume the SNR of the two signals
as SNR1 =SNR2 =30dB, Fig.3 shows the relationship curve
Consider a vertical uniform linear array whose inter-
between the URMSE1 of the first target and the unitary inter-
element distance equals half-wavelength. The number of array
target angle ∆θ .
elements is N=100, then the beamwidth is θ3dB =0.886 Nλd ≈
Fig.3 shows that the URMSE increases when the inter-target
1.016◦ . The iterative threshold γ=10−8 . SNR is defined as
angle decreases. The phenomenon accords with the experience
the ratio of the direct signal power to the noise variance
2 of most radar users that, when the inter-targer angle decreases,
of each array element, i.e., SNRi = |Aσi2| .qThe root mean the two targets get closer and become more difficult to be
square error (RMSE) defined as RMSEi = E[(θ̂i −θi )2 ] is resolved, and the precision of DOA estimation decreases ac-
introduced for DOA estimation precision evaluation. In or- cordingly. From the above figure one can conclude that when
der to enhance the generality of the conclusions, the target SNR=30dB and the inter-target angle is bigger than 0.4 times
azimuth, the included angle and the RMSE are normalized the beamwidth, the URMSE is smaller than 0.004. When the
with respect to the beam width, the unitary azimuth of the inter-target angle is smaller than 0.4 times the beamwidth, the
target is θ̄i =θi /θ3dB , the unitary inter-target angle is ∆θ= URMSE increases significantly but upper bounded by 0.02.

275
Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:53:34 UTC from IEEE Xplore. Restrictions apply.
Fig. 3. Relationship curve of the URMSE and the unitary inter-target angle
Fig. 5. Relationship curve between the URMSE and the phase difference

The results indicate that the proposed method has small DOA
estimation errors. Fig.5 shows that the URMSE of the first target increases
and the accuracy of the DOA estimates decreases when the
C. Relationship between the DOA Estimation Accuracy and
phase difference ∆φ improves.
the Amplitude Ratio
In this experiment, we assume the unitary azimuths of the VI. C ONCLUSION
two targets are θ̄1 =−0.4,θ̄2 =0.5 , respectively. The SNR of This paper studied the problem of resolving two unresolved
the first target is SNR1 =10dB , and that of the second one targets in a single beam using an array radar, and a dual-
is SNR2 =ρ2 SNR1 . The variety of the URMSE of the two monopulse-based maximum likelihood method is proposed.
targets while the amplitude ratio increases is showed in Fig.4. The simulation results indicate that, this method works effec-
tively, and it gains advantages in small estimation error and
adaptation to inter-signal power difference. The simulation
results also demonstrate that, the DOA estimation error of the
proposed method decreases with increasing SNRs and inter-
target angles. Moreover, conclusions has also been obtained
that, when the echo amplitude phase of one target increases,
the DOA estimation error of the other one increases.
R EFERENCES
[1] M. Feldmann, F. X, D. Nken, and W. Koch, Tracking of Extended Objects
and Group Targets Using Random Matrices, IEEE Transactions on Signal
Processing, vol. 59, pp. 1409-1420, 2011.
[2] ZHAO Feng, YANG Jianhua, DAN Mei, LIUJiaqi, WANG Xuesong.
Detection of presence of multiple unresolved targets based on range glint
[J]. Acta electronica sinica,2008,36(12): 2290-2298.
[3] W. D. Blair and M. Brandt-Pearce, Statistical description of monopulse
parameters for tracking rayleigh targets, IEEE Trans. Aerosp. Electron.
Fig. 4. Relationship curve between the URMSE and the amplitude ratio Syst., vol. 33, pp.597C611, Apr. 1998.
[4] W. D. Blair and M. Brandt-Pearce, Unresolved rayleigh target detection
Fig.4 shows that when the amplitude ratio improves, the using monopulse measurements, IEEE Trans. Aerosp. Electron. Syst.,vol.
33, pp. 543C552, Apr. 1998 .
URMSE of the first target varies slightly and that of the [5] W. D. Blair and M. Brandt-Pearce, Monopulse DOA estimation of two
second target decreases significantly because of the SNR unresolved rayleigh targets,IEEE Trans. Aerosp. Electron. Syst., vol. 37,
enhancement. The results indicate that the influence of the pp. 452C469, Apr. 2001.
[6] A. Sinha and Y. Bar-Shalom, Maximum likelihood angle extractor of two
SNR of one target on the DOA estimation performance of the closely spaced targets, IEEE Trans. Aerosp. Electron. Syst., vol. 38, pp.
other target is small. Namely, the big target will not envelop 183C203, Jan. 2002.
the small one and affect its DOA estimation performance. [7] Zhang, X., Willett, P., and Bar-Shalom, Y.Detection and localization
of multiple unresloved targets via joint multiple-bin processing. In
D. Relationship between the DOA Estimation Accuracy and Proceedings of IEEE Radar Conference, Huntsville AL, May 2003.
[8] Zhou LanFeng, Zhang Min, Zhao YiNan. Angle Estimation for Two
the Phase Difference Unresolved Swerling I Targets in Monopulse Radar, ICMMT 2010
In this experiment, we assume the unitary azimuths are Proceedings,1622-1625
[9] W.D.White, Low-angle radar tracking in the presence of multipath. IEEE
θ̄1 =−0.35,θ̄2 =0.4 , SNR1 =SNR2 =30dB , the phase of the Transactions on Aerospace and Electronic Systems, 10(6), 1974, 835-852
echo amplitude of the first target is 0. The relationship curve [10] Nickel U., Overview of Generalized Monopulse Estimation, IEEE A&E
between the URMSE and ∆φ is showed in Fig.5. Systems Magazine, 21(6) June 2006,27-36

276
Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:53:34 UTC from IEEE Xplore. Restrictions apply.

You might also like