A Closed-Form Localization Algorithm Using Angle-

of-Arrival and Difference Time of Scan Time Measure-

ments in Scan-Based Radar
Letter

This letter presents a closed-form localization algorithm using the

angle-of-arrival and difference time of scan time measurements from
the scan-based radar (SBR). With the known radar scanning velocity,
the difference time of the scan time can be converted into a subtended
angle, which has linear relationship with the angle-of-arrival (AOA).
The best linear unbiased estimator is applied for the optimal and reli-
able the AOA estimation. The weighted least square (WLS) method is
used for localization of the SBR system. The presented solution is com-
putationally efficient and statistically consistent in that the measured
mean squared errors achieve the Cramer–Rao lower bounds.

I. INTRODUCTION
The antenna sweeping velocity can be obtained by tak-
ing advantage of the geometric constraints. This is achieved
by the uniform rotating motion of the main beam of antenna
when it sweeps across a number of separate sensors [1]–
[7]. For the positioning of a scan-based radar (SBR), two
measurement methods are commonly used when using the
passive sensor array. One method uses the subtended angle
(SA) [2], another uses the angle-of-arrival (AOA) of the
SBR. The former can be obtained by the difference time of
scan time (DTOST). The latter can be measured directly by
an array sensor.
The SA measurement can be converted into target rang-
ing equations using geometric relations, and the intersec-
tion of multiple circles was applied to locate the target [2].
The SBR localization could be achieved by the AOA and
SA independently, and both results are combined to obtain
the location results [3]. The above two approaches have
suboptimal performance and high computation complexity.
Auxiliary variables were used for the SA localization, but

Manuscript received November 14, 2018; released for publication Decem-

ber 10, 2018. Date of publication January 3, 2019; date of current version
February 7, 2019.
DOI. No. 10.1109/TAES.2018.2890353
Refereeing of this contribution was handled by M. Baum.

This work was supported the Funds for the Meteorological information and
Signal Processing Key Laboratory of Sichuan Higher Education Institutes,
PRC.
Authors’ Addresses: W. Li and P. Wei are with School of Information and
Communication Engineering, University of Electronic Science and Tech-
nology of China, Chengdu 611731, China, E-mail: (liwanchun@uestc.
edu.cn; [email protected]); Y. Li is with School of Communication En-
gineering, Meteorological information and Signal Processing Key Labora-
tory of Sichuan Higher Education Institutes Chengdu, Chengdu University
of Information Technology, Chengdu 610059, China, E-mail: (lyingxiang
@cuit.edu.cn); H.-M. Tai is with the Department of Electrical and Com-
puter Engineering, University of Tulsa, Tulsa, OK 74104 USA, E-mail:
([email protected]). (Corresponding author: WanChun Li.)

0018-9251 
C 2018 IEEE

IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 55, NO. 1 FEBRUARY 2019 511

Authorized licensed use limited to: Hochschule Kempten. Downloaded on January 11,2022 at 03:50:12 UTC from IEEE Xplore. Restrictions apply.
 It can be seen from Fig. 1 that AOA at the kth sensor
can be obtained from the SBR location x
y − yk
αk = fα (x) = tan−1 , −π < αk ≤ π. (3)
x − xk
The estimated SA, obtained by DTOST model [2], can
be expressed as [2]
β̂k1 = βk1 + nβk1 (4)
where nβk1 is the white Gaussian noise with zero mean. The
Fig. 1. Geometry of SBRs.
accurate SA can be expressed as
the performance declines rapidly when the measurement r12 + rk2 − dk1
2
variance is large [4]. Performance of the sensor location βk1 = cos−1 (5)
layout was examined in [5]. The AOA-based algorithm in 2r1 rk
[6] transforms the nonlinear angle measurement equation where rk and dk1 are the distances from the sensor k to
into pseudolinear equation for solution. The passive loca- the radar and to the sensor 1, respectively. It also can be
tion system using array sensors was considered in [7]. Gav- expressed as [2]
ish and Weiss [8] discusses the observability of the location βk1 = ω (tk − t1 )
system based on AOA measurement.
It is a challenging task to find the SBR location using where ω is the antenna scanning speed. Then, we have
the AOA and SA measurement equations, which are highly β̂k1 = ω (tk − t1 ) + nβk1 . (6)
nonlinear. In this paper, we employ the correlation between
AOA and SA measurements for the SBR localization. The The measurement error of the arrival time for sensor k
proposed approach is based on the noticed phenomenon that is t˜k , which is modeled as the white Gaussian noise with
the difference between AOA signals from two adjacent sen- zero mean and variance σt2 . These errors are assumed to be
sors should be equal to SA (i.e., difference of the main lobe uncorrelated
 
time multiplied by the scanning speed ω of SBR). Under the E t˜k t˜m = σt2 δ (k − m) .
circumstance of synchronized sensors and known ω, data
fusion on AOA and SA data using the best linear unbiased δ(k − m) is the Kronecker delta function. Then, the
estimation (BLUE) is performed to obtain better AOA esti- measured SA can be obtained by
 
mation. The combined AOA and SA data are converted into β̂k1 = ω t˜k − t˜1 . (7)
linear equations of the target location. Using the first-order
It has the following statistical characteristics:
Taylor expansion, the residual vector of the linear equations     
is obtained. Using the residual vector variance matrix as the E nβk1 = E ω t˜k − t˜1 = 0
weighted matrix, the weighted least square (WLS) method 
  k = m, ω2 σt2 = σβ2
is extended for AOA and SA positioning systems. E nβk1 nβm1 =
The paper is organized as follows. Section II presents k = m, 2ω2 σt2 = 2σβ2 .
the AOA and DTOST measurement models. Section III
III. LOCALIZATION ALGORITHM
describes the scan-based emitter localization algorithm.
Section IV derives the Cramer-Rao bound for the scan- In this section, we present a closed-form localization
based localization. Section V presents simulation studies. algorithm using the AOA and the time difference of scan
Conclusions are drawn in Section VI. time measurements from the SBR.
It can be seen from Fig. 1 that, in the absence of noise,
II. MEASUREMENTS MODEL the SA is
This section describes the AOA and DTOST measure- βk1 = αk − α1 . (8)
ment models used for the proposed localization algorithm.
Fig. 1 shows that K sensors are located at x k = The DTOST is calculated only by the AOA. However,
[xk , yk ]T , k = 1, . . . , K and an SBR at x = [x, y]T . The due to possibly varying noise level, the noisy measurements
operator (·)T denotes the matrix/vector transpose and the of the AOA and SA can be expressed by the following
operator A−1 denotes the inverse of a square matrix A. vector-matrix equations:
The AOA measurement model is given as [8]–[11] α̂ = α + nα (9)
α̂k = αk + nαk , (1) β̂ = β + nβ (10)
where αk is the AOA, as shown in Fig. 1, and nαk is the where = [α1 , . . . , αK ]T , α̂ = [α̂1 , . . . , α̂K ]T , β = [β21 ,
white Gaussian noise with zero mean and variance σα2 . . . , βK1 ]T , β̂ = [β̂21 , . . . , β̂K1 ]T .
  
σα2 k = m E nα nTα = Q α = σα2 I
E (nαk nαm ) = (2)  
0 k = m. E nβ nTβ = Q β = σβ2 I.

512 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 55, NO. 1 FEBRUARY 2019

Authorized licensed use limited to: Hochschule Kempten. Downloaded on January 11,2022 at 03:50:12 UTC from IEEE Xplore. Restrictions apply.
 The linear relationship between AOA and SA can be TABLE I
expressed as: Localization Algorithm With AOA and SA

β = Hα (11)
 
H = −1K−1 I K−1 .
Here, 1K−1 denotes the all-one vector of size (K − 1)
and I K−1 denotes the identity matrix of size (K − 1).
Under the linear Gaussian assumption [12] and the lin-
ear relationship between AOA and SA, we have the follow-
ing optimization problem:
  T 

−1 −1
min (α̂ − α) Q α (α̂ − α) + β̂ − β Q β β̂ − β
T
α,β
(12)
subject to
β = Hα.
The corresponding BLUE can be obtained as [13] ⎡ ⎤
 −1  x1 sin (α̂ B 1 ) − y1 cos (α̂ B 1 )
−1 −1 −1 ⎢ .. ⎥
α B = Q −1 α + H T
Q β H Q α α̂ + H T
Q β β̂ . b=⎣ . ⎦
(13) xK sin (α̂ B K ) − yK cos (α̂ B K )
The variance of the BLUE estimate is given by ε = CnB k
 −1
D (α B ) = Q −1 −1 C = diag (c1 , . . . , cK )
α + H Qβ H = QB .
T
(14)
ck = (x − xk ) cos (α̂ B k ) + (y − yk ) sin (α̂ B k ) .
The AOA model of (1) can be rewritten as a pseudolinear
equation using the estimated AOA Here, A denotes the data matrix, b denotes the data vec-
tor, ε is the noise disturbance vector, and C is a diagonal
(x − xk ) sin (α̂ B k − nB k ) matrix with diagonals being ci . This allows the transforma-
= (y − yk ) cos (α̂ B k − nB k ) . (15) tion of the problem from the complex nonlinear equations
of measurement angles into the linear equations of the SBR
Using the trigonometric identity position.
The least squares (LS) solution to (19) is [8]
sin(α̂ B k − nB k ) = sin (α̂ B k ) cos (nB k )
 −1
− cos (α̂ B k ) sin (nB k ) xˆLS = AT A AT b. (20)
cos (α̂ B k − nB k ) = cos (α̂ B k ) cos (nB k ) The weighted least squares solution is [8]
+ sin (α̂ B k ) sin (nB k ) (16)  −1
ˆ = AT Q −1 A AT Q −1 b
x WLS (21)
and the assumption that the measurement noise nB k is rel-
atively small, we have where
  
sin (nB k ) ≈ nB k Q = E εε T = C Q B C T . (22)
(17)
cos (nB k ) ≈ 1. Procedure of the proposed localization algorithm with
the AOA and SA is summarized in Table I.
Then, (14) can be simplified as
IV. CRAMER-RAO LOWER BOUND (CRLB) ANALYSIS
sin (α̂ B k − nB k ) ≈ sin (α̂ B k ) − cos (α̂ B k ) nB k
cos (α̂ B k − nB k ) ≈ cos (α̂ B k ) + sin (α̂ B k ) nB k . (18) The Cramer-Rao bound of the proposed localization
algorithm can be expressed as
Taking the first-order Taylor expansion at measured val-  
ues, (15) has the following linear approximation: CRLBx = tr J −1 (23)

Ax = b + ε (19) where tr( A) denotes the trace of A.

The matrix can be derived starting with [3], [14]
where
⎡ ⎤ J = J AOA + J SC (24)
sin (α̂ B 1 ) − cos (α̂ B 1 )
⎢ ⎥ where
A=⎢
⎣ ... ..
.
⎥
⎦ ∂α T −1 ∂α T
sin (α̂ B K ) − cos (α̂ B K ) J AOA = Q x (25)
∂x α ∂

LETTER 513

Authorized licensed use limited to: Hochschule Kempten. Downloaded on January 11,2022 at 03:50:12 UTC from IEEE Xplore. Restrictions apply.
 Monte Carlo simulations were conducted for performance
evaluation.
Assume that SBR is located at (50, 50). The unit of the
distance is km. The sensors are linearly distributed at the lo-
cations (0, 0), (20, 0), (40, 0), (60, 0),(80, 0), and (100, 0).
Two settings were considered as follows.
Set A: The AOA measurement noises are white Gaus-
sian with zero mean and standard deviation varying from
0.1° to 5°. The standard deviation of the SA measurement
noise is fixed at 1◦ .
Set B: The SA measurement noises are white Gaussian
with zero mean and standard deviation varying from 0.1° to
5°. The standard deviation of the AOA measurement noise
is fixed at 1◦ .
Fig. 2 plots the root-mean-square errors (RMSEs) for
the proposed localization algorithm, the CRLB for AOA,
SA, and AOA+SA algorithms. The results were the aver-
age after 10 000 Monte Carlo runs. The errors are measured
against various noise levels. The position the RMSE is de-
fined as

RMSE
 
1 N  
= 10 × log10 (xk − x) + (yk − y)
2 2
.
N k=1

(27)

The horizontal axis in Fig. 2(a) denotes the standard

deviation of the AOA measurement noise in dB, while
that in Fig. 2(b) the standard deviation of the diagonal the
SA measurement noise in dB. The vertical axis in both
graphs represents the RMSE of the positioning, as shown
in (27).
It can be seen from Fig. 2 that the positioning per-
formance of the proposed localization algorithm can ap-
Fig. 2. Localization RMSE errors of joint positioning. (a) Set A proach CRLB under the variations of AOA or SA measure-
measurement: Various AOA measurement noises with SA noise fixed at ment errors. The proposed joint positioning method per-
1◦ . (b) Set B measurement: Various SA measurement noises with AOA forms much better than the AOA or the SA measurement
noise fixed at 1◦ .
alone.

VI. CONCLUSION
∂β T −1 ∂β T
J SC = Q (26) This paper has presented a closed-form localization ap-
∂x β ∂x
 ∂α1   αK  proach using both SA and AOA measurements for the SBR
∂α T ∂x
. . . ∂α K
∂x
− sin α1
r1
. . . − sin
rK
= ∂α = − cos α system. A linear relationship between the SA and AOA is
∂x
∂y
1
. . . ∂αK
∂y r1
1
. . . − cos
rK
αK
utilized and a best linear unbiased estimator is introduced.
 ∂β21  Simulation results show that the presented approach is ef-
∂β T ∂x
. . . ∂β∂xK1 ficient due to its closed-form and is sufficient so that the
= ∂β
∂x 21
. . . ∂β∂yK1 corresponding CRLB can be can attained.
∂y
 sin α1 
r1
− sinr2α2 . . . sinr1α1 − sinrKαK
= cos α1 . ACKNOWLEDGMENT
r1
− cosr2α2 . . . sinr1α1 − cosrKαK
The authors would like to thank the anonymous reviewers
and the editor for their comments that improved the presen-
tation of the paper. This work is supported by Meteorolog-
V. SIMULATION RESULTS
ical information and Signal Processing Key Laboratory of
In this section, we evaluate the proposed localization al- Sichuan Higher Education Institutes of Chengdu University
gorithm using data from SBR under certain circumstances. of Information Technology.

514 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 55, NO. 1 FEBRUARY 2019

Authorized licensed use limited to: Hochschule Kempten. Downloaded on January 11,2022 at 03:50:12 UTC from IEEE Xplore. Restrictions apply.
 WANCHUN LI [6] K. Dogancay
University of Electronic Science and Technology Self-localization from landmark bearings using pseudolinear
of China, Chengdu, China estimation techniques
IEEE Trans. Aerosp. Electron. Syst., vol. 50, no. 3, pp. 2358–
YINGXIANG LI
2365, Nov. 2014.
Chengdu University of Information Technology,
[7] J. Bosse, A. Ferréol, and P. Larzabal
Chengdu, China
A spatio-temporal array processing for passive localization of
PING WEI radio transmitters
University of Electronic Science and Technology IEEE Trans. Signal Process., vol. 61, no. 22, pp. 5485–5494,
of China, Chengdu, China Nov. 2013.
[8] M. Gavish and A. Weiss
HENG-MING TAI , Senior Member, IEEE
Performance analysis of bearing-only target location algo-
University of Tulsa, Tulsa, OK USA
rithms
IEEE Trans. Aerosp. Electron. Syst., vol. 28, no. 3, pp. 817–
828, Jul. 1992.
[9] H. Shao, X. Zhang, and Z. Wang
REFERENCES Efficient closed-form algorithms for AOA based self-
localization of sensor nodes using auxiliary variables
[1] K. Dogancay IEEE Trans. Signal Process., vol. 62, no. 10, pp. 2580–2594,
A closed-form pseudolinear estimator for geolocation of scan- Apr. 2014.
ning emitters [10] K. Doğançay
In Proc. IEEE Int. Conf Acoust., Speech, Signal Process., USA, Relationship between geometric translations and TLS estima-
Jun. 2010, pp. 3546–3549. tion bias in bearings-only target localization
[2] H. Hmam IEEE Trans. Signal Process., vol. 56, no. 3, pp. 1005–1017,
Scan-based emitter passive localization Mar. 2008.
IEEE Trans. Aerosp. Electron. Syst., vol. 43, no. 1, pp. 36–54, [11] Z. Wang, J. Luo, and X. Zhang
Jan. 2007. A novel location penalized maximum likelihood estimator for
[3] K. Dogancay bearing only target localization
UAV path planning for passive emitter localization IEEE Trans. Signal Process., vol. 60, no. 12, pp. 6166–6181,
IEEE Trans. Aerosp. Electron. Syst., vol. 48, no. 2, pp. 1150– Sep. 2012.
1166, Apr. 2012. [12] Y. Shalom, X. Li, and T. Kirubarajan
[4] Y. Zhang, M. Zhang, and F. Guo Estimation With Applications to Tracking and Navigation: Al-
Recursive instrumental variable method for locating a scanning gorithms and Software for Information Extraction. Upper Sad-
emitter by a single observer using time of interception dle River, NJ, USA: Wiley, 2001.
IEE Proc. Radar, Sonar Navigat., vol. 11, no. 12, pp. 1839– [13] S. M. Kay
1844, Dec. 2017. Fundamentals of Statistical Signal Processing. New York, NY,
[5] K. Doganay USA: Prentice-Hall, 1993.
Online optimization of receiver trajectories for scan-based [14] D. Torrieri
emitter localization Statistical theory of passive location systems
IEEE Trans. Aerosp. Electron. Syst., vol. 43, no. 3, pp. 1117– IEEE Trans. Aerosp. Electron. Syst., vol. 20, no. 2, pp. 183–
1125. Nov. 2007. 198, Mar. 1984.

LETTER 515

Authorized licensed use limited to: Hochschule Kempten. Downloaded on January 11,2022 at 03:50:12 UTC from IEEE Xplore. Restrictions apply.