2019 International Conference on Control, Automation and Information Sciences (ICCAIS)

Generalized Space-Time Adaptive Monopulse

Angle Estimation Approach
Rong Zhou Chenxiao Zhang Guanglei Zhang
AVIC Leihua Electronic Technology AVIC Leihua Electronic Technology AVIC Leihua Electronic Technology
Research Institute Research Institute Research Institute
Wuxi,Chian Wuxi,Chian Wuxi,Chian
[email protected] [email protected] [email protected]

Abstract—For airborne radar, Space-time adaptive also employed because less training data are required and the
processing (STAP) is the key technique to suppress both clutter computation load is minimized. The accuracy of angle
and interference and detect moving target in strong ground estimation are improved by the novel approach. The outline
clutter. Monopulse technique after STAP is used to estimate of this paper is as follows. Signal model and dimension
target angle. However, with this adaptive processing, the reduction are formulated in section II. In section III we
typical monopulse technique is no more effective because the present the complete process of the space-time adaptive
main beams are distorted, that may lead to angle errors. This monopulse procedure. In section IV simulation results are
paper is concerned with the main beam distortion that results shown to validate the effectiveness of the new method.
from clutter suppression by STAP. Generalized monopulse is
Conclusions are given in section V..
extended to a space-time adaptive architecture to improve the
accuracy of angle estimation. Simulation results show that
performance of the proposed method is superior to that of the II. SIGNAL MODEL
typical method. Consider a radar antenna made of N sensors and the inter-
element spacing is d. M pulses compose one coherent
Keywords—STAP, angle estimation, adaptive monopulse,
processing interval, and the observed vector from the range
dimension reduction
gate l ( l  1, ,L ) is denoted as
I. INTRODUCTION
Angle estimation plays a significant role in the signal  X  l   bs t , t   X c  l   X n  l   
processing for modern radar systems. Monopulse is widely
used for precision angle estimation, which inherits all the The complex space-time data X has size of MN  L . Here
flexibility of digital beamforming. It allows the direction X c and X n are the clutter and noise components. We
accuracy of a target to be better than the beamwidth from a
single time sample of data [1][2]. Nevertheless, the assume that the clutter and noise is Gaussian distributed with
performance of typical monopulse is decreased severely zero mean and covariance matrix Q . b contains the
because of the ground clutter faced by airborne radar. To unknown target amplitude and phase with respect to a
suppress the clutter, STAP is employed in newly developed reference. s t , t  denote the space-time steering vector of
airborne phased array radar. Conventional space-time the target, and can be written as
adaptive monopulse is efficacious when the target is located
in side lobe clutter area. However, it decreases seriously
when the target appears in main beam because the  s t , t   st  t   ss t   
monopulse formula is dependent on the constant ratio
between the difference and sum beam gradients at the look
direction. z
A number of papers have been written on correction v
schemes of space-time adaptive monopulse, which account
for the distortion after STAP, to improve the angle x
estimation performance of the target. Adding suitable
constraints for the required adapted difference beam is a
viable method. Different processing structures and
performance studies were conducted in [3]-[5]. However, the
employment of constraints cost some of the system’s degrees y
of freedom. The adaptive degrees of freedom are accordingly 

reduced and the excess power of clutter after difference
weighting is increased. Techniques based on the maximum
likelihood estimator have been developed [6]. Davies derived
the technique for uniform linear array in [7] and Nickel
applied it to arbitrary subarray and arbitrary sum and
difference beam weightings in [8][9].
In this paper, we extend the generalized monopulse to a
Fig. 1. Geometry of airborne radar in down-looking mode
STAP architecture. Reduced dimension STAP techniques are

978-1-7281-2311-0/19/$31.00
Authorized ©2019
licensed use limited to: UNIVERSIDAD IEEE
CARLOS III MADRID. Downloaded on June 13,2020 at 18:46:03 UTC from IEEE Xplore. Restrictions apply.
 2019 International Conference on Control, Automation and Information Sciences (ICCAIS)

Where  denotes the Kronecker product, and Fig. 2. The flow chart of space-time adaptive monopulse

The complete process of angle estimation after space and

j 2 t j 2  M 1 t
T time adaptive processing is illustrated in Fig. 2. The
st  t   1 e  e 
 generalized monopulse estimation scheme is extended to
 T
  STAP for arbitrary sum and difference beams and arbitrary
ss t   1 e j 2t  e j 2  N 1t  subarrays in space and time. Main steps are listed as follows.
(a) Selecting specific training data from space-time data
are the M×1 time steering vector at the normalized Doppler
cube to estimate the covariance matrix Q
frequency  t and the N×1 space steering vector at the
spatial frequency t
1
 Q  X l  X H l  
L l

d
 t  cos  sin   
 Here  denotes the range cell set of the training samples
and L is the length of it. X is the dimensionally reduced
Where  ,  describe the azimuth and elevation respectively. training data. To ensure the performance of estimation,
The clutter signal of the range gate l is given by L  2 * M sub Nsub . M sub and N sub represent the reduced time
and space dimensions respectively.
Nc (b) Calculating the adaptive sum and difference weight
 X c  l    blk s lk , lk    vectors w and w of STAP, that is
k

Where N c denotes the number of independent clutter Q 1 s

sources of each range ring. blk is the amplitude of the kth w 
sH Q 1 s
clutter patch in the lth rang. s lk , lk  is the space-time   
Q 1 s
steering vector of the normalized Doppler frequency  lk and w  H 1
s Q s
the space frequency lk .
Where s and s are the dimensionally reduced sum and
Because of the high dimensionality of the space-time
samples, one is interested to reduce the computational difference space-time steering vectors respectively, and
expense by reducing the dimension. If dimension-reduced
matrix for space and time, characterized by Ts , Tt are used,
s  T H  st  ss   T H  st   wHs ss  
the data’s transformation matrix T can be formed by   
Kronecker product T  Tt  Ts . Then the receiving data and s  T H  st  ss   T H  st   wHs ss  
the steering vector can be written as X  T H X ,
s t , t  = T H s t , t  . ()H is the conjugate transpose Where st is the time steering vector , and ss  wHs ss ,
operator. ss  wHs ss are the sum and difference space steering vectors.
ws , ws are the quiescent sum and difference weights. For
III. GENERALIZED SPACE-TIME ADAPTIVE MONOPULSE
simplicity, the elements are given uniform weights in this
paper, that is

ws  [1, ,1]

 
N
  
ws  [1,

,1,  ,
1, 
1]
N 2 N 2

(c) Weights application. Adaptive Sum and difference

outputs z and z can be formed in this step

z  wH X
  
z  wH X

Then both clutter and interference can be suppressed

effectively and the Signal to Interference-plus-Noise Ratio

Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:46:03 UTC from IEEE Xplore. Restrictions apply.
 2019 International Conference on Control, Automation and Information Sciences (ICCAIS)

(SINR) can be improved before signal detection. Hence, With all variables determined in (12)-(15), the space-time
better estimation accuracy could be obtained. adaptive monopulse can be used to estimate target angle. The
bias between the estimated value and the real one can be
(d) Comparing the sum output z with the threshold reduced compared with the conventional method.
value to determine whether target appears or not. If the target
appears, calculating its monopulse ratio RSTAP , that is IV. EXPERIMENTAL RESULTS
In this section, several computer simulations are
developed to verify the effectiveness of the method
z 
 RSTAP  real     mentioned above. The experimental parameters are listed in
 z  Table 1.

However, the modulation generated by STAP perturbs TABLE I. EXPERIMENTAL PARAMETERS

the angle information of original signal, which results in the
distortion of monopulse pattern. Therefore, with the Target angle (0.5 ,-15 )
Target parameters
traditional monopulse technology, evident angle errors could SNR -5dB
be introduced by the adapted pattern.
CNR 20dB
We extend the generalized monopulse to STAP Number of clutter rings 2600
architecture and the formula for space-time adaptive Clutter parameters
monopulse can be written as Azimuth range [  45 45 ]
Azimuth spacing 0.25
Number of elements 32
 ˆ  0   1  RSTAP     
Look direction (0 ,-15 )
Radar parameters Back lobe level -30dB
Where ̂ is the estimated spatial frequency and 0 is the
Radar platform velocity 100m/s
spatial frequency of the pointing detection.  ,  denote the
Radar platform height 9km
slope and deviation corrections respectively, and
Fig.3 shows the results of conventional PD processing for
received echo data ，the target is submerged by clutter in
 wH s  this scenario. The color bar demonstrates the power of every
   real  H   clutter ring. To make the target visible, the clutter should be
 w s  suppressed and the range-doppler power spectrum shown in
Fig. 4 is employed to assess clutter suppression performance
after STAP.
   p q 
0 0


H
 ds  ds  
real  wH  sH w  wH s    w 
 d  d  
p  
H 2
w s
  
 H ds 
 w 
q   2real  H d 
 w s 
 

ds
Where denotes the derivative of s in the look
d
direction, that is
Fig. 3. The range-doppler power spectrum image after PD processing



ds d Tt st   T  w s  
H
 s
H
 H
s s 
d d

   ds   
  T H  st   wHs  s     
   d   

 T H st   j 2  N  1 wHq ss  

Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:46:03 UTC from IEEE Xplore. Restrictions apply.
 2019 International Conference on Control, Automation and Information Sciences (ICCAIS)

the clean area and is completely submerged by the clutter at

last. 200 Monte Carlo experiments are used for each case and
RMSE result is given in Fig.7. The two methods can achieve
the same RMSE of 0.1 when the target is in the clean area.
With the target approaching the main clutter, RMSE of the
conventional method increases sharply, while relatively high
accuracy can still be achieved by the generalized method
presented in this paper.

V. CONCLUSIONS
STAP technique is employed widely to suppress the
clutter in newly developed phased array radar. The typical
monopulse technique is no more effective after STAP
because the beams are perturbed due to the adaptation, and
this can lead to angle errors. To solve this problem, the
generalized monopulse has now been extended to space-
time adaptive processing. This restores a good detection
Fig. 4. The range-doppler power spectrum image after STAP
performance and the distinction is extremely obvious when
Calculate the target monopulse ratio after signal detection the target of interest is located in main beam. The results of
via (10), then computer simulations indicate that the generalized space-
time adaptive monopulse algorithm performs better than the
conventional algorithm.
 ˆ   0   1 RSTAP  

is the estimated result by conventional adaptive space-time

monopulse method.  0 is the pointing angle, and  is the
slope of static monopulse ratio curve, which is obtained by
array elements.
200 Monte Carlo experiments are carried out by
conventional and generalized adaptive space-time monopulse
techniques, and the results are shown in Fig.5. The blue
marker ‘*’ indicates the angle estimation obtained by the
conventional method, while the red marker ‘o’ indicates the
generalized one. The black line set as reference represents
the real position of the target. Clearly, with the conventional
method, an obvious deviation exists between the estimated
value and the real one. However, the angle estimations
obtained by the generalized method are distributed near the
real value.
Fig. 5. Angle estimation results by conventional and generalized methods
For more intuitively expression, the root mean square
error (RMSE) of angle estimations is employed, i.e.

KM
1 2
  RMSE 
KM
 ˆ
k 1
k  t   

Where ̂ and  t are the estimated value and the real one
respectively. K M is the number of the Monte Carlo trials.
The RMSE of the experiments above is listed in Table 2. The
results verified that the method mentioned above better than
the conventional monopulse method.

TABLE II. RMSE

Conventional space-time adaptive monopulse 0.3324

RMSE
Generalized space-time adaptive monopulse 0.1645
Furthermore, we assume that the target is set on different Fig. 6. The range-doppler power spectrum image after STAP
doppler gates, far to near from the main clutter as shown in
Fig 6. That is, the target gradually enters the clutter area from

Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:46:03 UTC from IEEE Xplore. Restrictions apply.
 2019 International Conference on Control, Automation and Information Sciences (ICCAIS)

[2] Y. Kaibor. “Advanced monopulse processing of phased array radar,”

Radar Conference, pp. 174-179, 2015.
[3] G Chen , “Space-time adaptive monopulse based on space-time
uniform constraint,” IEEE International Conference on Signal
Processing, Comunication and Computing, pp. 215-128, 2014.
[4] W Lu, “Linear-constraint-based space-time adaptive monopulse
processing technique,” Journal of Electronics and Information
Technology, vol.32, no.10, pp.2501-2505, 2010.
[5] R. Wu, “Parameter estimation of moving target based on linearly
constrained space-time adaptive monopulse technique”, Phased Array
Systems and Technology, pp. 107-112, 2010.
[6] J. Ward, “Maximum likehood angle and velocity estimation with
space-time adaptive processing radar,” Signal, Systems and
Computers, vol.2, pp.1265-1267, 1996.
[7] R. C. Davies, L. E. Brennan and L. S. Reed, “Angle estimation with
adaptive arrays in external noise filed,” IEEE Transactions on
Aerospace and Electronic Systems, vol. 12, no. 2, pp. 179-185, 1976.
[8] U. L. Nickel, “Performance analysis of space-time adaptive
monopulse,” Signal Processing, vol. 84, no. 9, pp. 1561-1579, Sep
2004.
Fig. 7. Target angle estimation RMSE versus normalised doppler [9] U. L. Nickel, E. Chaumette, and P. Larzabal, “Statistical performance
prediction of generalized monopulse estimation,” IEEE Transactions
on Aerospace and Electronic Systems, vol. 47, no. 1, pp. 381-405,
REFERENCES 2011.

[1] U. L. Nickel, “Overview of generalized monopulse estimation,” IEEE

Aerospace and Electronic Systems Magazine, vol. 21, no. 6, pp. 27-57,
June 2006.

Authorized licensed use limited to: UNIVERSIDAD CARLOS III MADRID. Downloaded on June 13,2020 at 18:46:03 UTC from IEEE Xplore. Restrictions apply.