NATO/PFP UNCLASSIFIED

Click here to view PowerPoint presentation; Press Esc to exit

Radar Emitter Localization Using TDOA Measurements

from UAVs and Shipborne/Land-Based Platforms

Huai-Jing Du, Ph.D. / Jim Lee, Ph.D.

Defence R&D Canada – Ottawa
3701 Carling Avenue
Ottawa, Ontario K1A 0Z4
CANADA
Email: [email protected] / [email protected]

ABSTRACT
This paper presents passive radar emitter localization using the Time-Difference-Of-Arrival (TDOA)
technique from Unmanned Aerial Vehicles (UAVs) working in conjunction with a fixed/mobile station (such as
a shipborne/land-based station) to geolocate a radar emitter to fulfill maritime and littoral area surveillance.
Instead of using multiple stationary sensors, multiple TDOA measurements are obtained over time intervals
from moving sensors along their flight paths. A simple Least-Squares (LS) solution is derived from TDOA
measurements in combination with known sensor locations to solve three-dimensional (3-D) location
estimation. The proposed location method takes into account of both random and systematic or nonrandom
measurement errors and it allows the removal of measurement errors that are nonrandom. Therefore,
a position-location problem can be solved more accurately when biased measurements are present.
The advantages and effectiveness of the proposed method are demonstrated using simulation.

1.0 INTRODUCTION
Unmanned Aerial Vehicles (UAVs) today are primarily used in reconnaissance, surveillance and intelligence-
gathering missions [1]. For future operations, a UAV will most likely be working in conjunction with other
platforms (such as a ship/land-based station, manned aircraft or a space station) to increase the capability of
an existing system. This paper presents UAV applications in passive localization of a radar emitter to
fulfill Electronic Support Measures (ESM) roles for Electronic Warfare (EW) applications. Various
localization algorithms such as Angle-Of-Arrival (AOA), Frequency-Difference-Of-Arrival (FDOA) [2,3] and
Time-Difference-Of-Arrival (TDOA) [2] have been developed in the past. In this paper, the TDOA technique
is chosen to solve a three-dimensional (3-D) location problem for the purpose of maritime and littoral
surveillance.

The time interval or Time-Of-Arrival (TOA) for signals propagating from an emitter to a sensor is dependent
on the emitter-sensor geometry and medium characteristics. In a constant velocity medium as considered in
the following analysis, differences in arrival times or TDOA of emitter signals observed at different sensor
locations are proportional to differences in emitter-sensor ranges [4]. Fig. 1 illustrates the geometry of TDOA
technique by utilizing multiple sensors for passive emitter localization. Let tm (tn) be the time for the signal
radiating from the emitter to arrive at sensor m (n) and dm ( dn ) be the distance from the emitter to sensor
m (n). The TDOA between sensors m and n is then defined by ∆tm,n as follows:

Paper presented at the RTO SCI Symposium on “Multi-Platform Integration of Sensors and Weapons Systems
for Maritime Applications”, held in Norfolk, USA, 21-23 October 2002, and published in RTO-MP-097(I).

RTO-MP-097(I) 9-1

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

dm dn
∆t m, n = t m − tn = −
c c (1)
m, n = 1,2,..., N (m ≠ n)

where c is the speed of light, N is the number of sensors, and dm is the distance defined by

d m = ( xe − xm ) 2 + ( ye − ym ) 2 + ( ze − xm ) 2 (2)

where (xe ,ye ,ze) and (xm,ym,zm) are the position coordinates of the emitter and sensor m, respectively. Eq.(1)
can be expressed in terms of range difference as

c∆t m , n = d m − d n (3)

Emitter: (xe, ye, ze)

d1
dm dn dN

… …
…

Sensor N: (xN, yN, zN)

Sensor 1: (x1, y1, z1)

Sensor m: (xm, ym, zm) Sensor n : (xn, yn, zn)

Figure 1: Location Based on TDOA Technique Using Multiple Stationary Sensors.

In order to solve three unknowns (i.e., xe, ye and ze), a minimum of three TDOA measurements from four
stationary sensors is required.

An extensive work has been done in the past to solve location problems using the TDOA technique. Fang [5]
derived an exact close-form location solution when the number of TDOA measurements is equal to the
number of unknowns, i.e., the position coordinates of an emitter. This solution, however, cannot make use of
extra measurements when they become available to improve position accuracy. Schau and Robinson [6]
derived a location solution for the more general situation with extra TDOA measurements from extra sensors.
Although a closed-form solution has been developed, the estimation may not be accurate when noisy (random
and/or nonrandom) measurements present. Smith and Abel [4] later extended the method in [6] into a linear
Least-Squares solution using a sensor array to deal with random noise but not to handle nonrandom)
measurement errors. Manolakis [7] presented an explicit solution to a 3-D position problem based on three
range measurements from three stationary stations. However, the method did not take range measurement
uncertainty into account in the location estimation. Chan and Ho [8] proposed an explicit approach using

9-2 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

TDOAs measured at multiple sensor locations. This method is an approximate realization of the maximum-
likelihood estimator for the case when the TDOA estimation errors are small and assumed to be independent
zero-mean stationary Gaussian random noise. To handle nonrandom errors, Poirot and McWilliams [9]
derived a location solution based on bearing measurements, which allows estimate and removal of nonrandom
measurement errors.

This paper presents passive radar emitter localization from one or two UAVs working in conjunction with a
shipborne/land-based station to fulfill maritime and littoral surveillance. Most of the TDOA location solutions
are based on multiple stationary sensors to obtain multiple measurements. However, in this approach multiple
TDOA measurements are obtained from the moving sensors over time intervals along UAV’s pre-determined
flight trajectories. In the past, many mathematical models have been developed based on TDOA
measurements. However, most of them are based on an assumption that measurement noises are random with
zero-mean. In fact measurement noise is a combination of random and systematic (or nonrandom) errors.
In this paper, a mathematical model including random and systematic errors is used to deal with the
measurement errors properly. A Least-Squares (LS) solution is derived from TDOA measurements in
combination with known sensor locations to solve 3-D location estimation. The proposed method is adaptable,
and allows estimation and removal of measurement errors that are not random. Therefore, a position-location
problem can be solved more accurately when noisy measurements are present. In addition, the proposed Least
Squares method is useful in solving multiple measurements and can be easily extended to solve mixed-mode
position-location problems. Finally, simulation results are included to present the effectiveness of the
proposed method.

2.0 GEOLOCATION USING TDOA MEASUREMENTS FROM A SHIPBORNE/

LAND-BASED PLATFORM ACCOMPANIED BY UAVS
As stated above, a UAV will most likely operate in conjunction with other vehicles/stations (such as ships,
land mobile stations, manned aircraft or space stations) to increase the capability of an existing system.
Fig. 2 illustrates emitter localization using TDOA measurements from a UAV and a ship/land-based station,
while Fig. 3 presents two UAVs operating in conjunction with a ship/land-based station to locate an emitter.

Emitter: (xe, ye, ze)

TOA2

TOA1

DF Sensor 2: (x2, y2, z2)

Sensor 1: (x1, y1, z1)

UAV flight path
A shipborne/land-
based station

Figure 2: Location Using TDOAs from a UAV Operating

in Conjunction with a Ship/Land-Based Platform.

RTO-MP-097(I) 9-3

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

Emitter: (xe, ye, ze)

TOA2

DF TOA1
Sensor 2: (x2, y2, z2)

Sensor 1:(x1, y1, z1)

A shipborne/land-
based station
UAV flight path
UAV flight path

Figure 3: Location Using TDOAs from Two UAVs Operating

in Conjunction with a Ship/Land-Based Platform.

To intercept radiated signals, two time-synchronized ESM sensors are utilized to obtain TDOA measurements
with one sensor on a UAV and the other on the base station or another UAV. TDOA measurements can be
obtained at various sensor locations along a UAV pre-determined fight path. Sensor information (such as
emitter information and UAV navigation data) may be processed on-board or transferred to the base station
for analysis and location determination. A Line-Of-Sight (LOS) data link from a moving platform to the base
station is required. If beyond LOS, a relay communication (SATCOM or an additional aircraft) between the
moving platform and the base station may be required. A direction finding (DF) system may be used at the
base station to provide an initial position of an emitter to allow UAV to get close to a target.

2.1 Formulation of the Problem

Figs. 2 and 3 can be simplified as given in Fig. 4 to present the geometry between an emitter and two sensors.
The spatial coordinates of sensor 1 (on a fixed/moving platform) and sensor 2 (on a moving platform) at time
interval i are defined by (xi,1,yi,1,zi,1) and (xi,2,yi,2,zi,2), respectively. Instead of using multiple stationary sensors
(see Fig. 1), multiple TDOA measurements are obtained from various sensor positions over time intervals
along a UAV flight path. It can be seen from Eq.(1) that multiplying the difference of signal arrival time by
the speed of light gives us the range difference as follows

c∆ti = d i ,1 − d i , 2 (4)

where i represents the ith time interval, c the speed of light, ∆ti the time difference of arrival between sensor
1 and sensor 2 at the ith time interval, and di,,1 - di,,2 the range difference between sensor 1 and sensor 2 at the ith
time interval. The above equation can be expanded as

c∆ti = ( xe − xi ,1 ) 2 + ( ye − yi ,1 ) 2 + ( xe − zi ,1 ) 2 − ( xe − xi , 2 ) 2 + ( ye − yi , 2 ) 2 + ( xe − zi , 2 ) 2
(5)
i = 1,2 ,...k

where k is the number of measurements along a UAV flight path. To solve for the three unknowns xe ye and ze,
a minimum of three TDOA measurements are required, which allows to expand Eq. (5) into three equations.
The location solution derived from Eq. (5) is based on the geometry between the sensors and the emitter

9-4 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

without consideration of TDOA measurement noise and bias. However, with noisy measurements, the emitter
location derived from the above geometry will not intersect in a single point. Thus, a statistical model is
needed to obtain the desired position of an emitter.

Emitter: (xe, ye, ze)

Z

d1,1 dk,1 d1,2

… dk,2
…
Sensor 1

(x1,2,y1,2,z1,2)
(x1,1, y1,1, z1,1)
(xk,2,yk,2,zk,2) Sensor 2
(xk,1, yk,1, zk,1)

Y
0

Figure 4: Geometry between an Emitter and Two Sensors.

2.2 Statistical Model

Considering TDOA measurement noise, Eq. (5) can be modified and expressed as

∆
yi = c∆t i = hi ( xe , y e , z e , xi ,1 , yi ,1 , z i ,1 , xi , 2 , yi , 2 , z i , 2 ) + ni (6)
i = 1,2,...k

where ni represents the measurement noise and

hi ( xe , ye , ze , xi ,1 , yi ,1 , zi ,1 , xi , 2 , yi , 2 , zi , 2 ) = ( xe − xi ,1 ) 2 + ( ye − yi ,1 ) 2 + ( ze − zi ,1 ) 2
(7)
− ( xe − xi , 2 ) 2 + ( ye − yi , 2 ) 2 + ( ze − zi , 2 ) 2

The position coordinates of the emitter are defined by (xe, ye, ze). The position coordinates of the two sensors
at the ith time interval are given by (xi,1, yi,1, zi,1) and (xi,2, yi,2, zi,2), respectively.

RTO-MP-097(I) 9-5

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

2.3 Least-Squares Approach for Emitter Localization

Since the function hi(xe,ye,ze,xi,1,yi,1,zi,1,xi,2, yi,2,zi,2) in Eq. (7) is a nonlinear function of positions of emitter and
sensor positions, the function is linearized by a Taylor series expansion about the initial estimate of emitter
location (xe0,ye0,ze0). If retaining only the first-order terms, Eq. (6) at the ith time interval may be represented
by
∂hi ∂h ∂h
yi − hi ( xe0 , ye 0 , ze 0 , xi ,1 , yi ,1 , zi ,1 , xi , 2 , yi , 2 , zi , 2 ) = ∆xe + i ∆ye + i ∆ze + ni (8)
∂xe ∂ye ∂ze

For k measurements of data, the above equation may be written as the linear model in a matrix form as
Y = A X+ N (9)
k ×1 k × 3 3×1 k ×1

It is desired to estimate the emitter location that best fits TDOA measurements. In particular, to find the X̂
that minimizes the sum of squares of difference between the measurements and the estimated functions is a
natural choice for a goodness-of-fit criterion [4]
2
min Y − AXˆ (10)

The formal least-squares solution for X̂ is obtained by minimizing Eq. (10) and given by

Xˆ = ( AΤ A) −1 AΤY (11)

If it is desired to weight the measurements according to a priori confidence in each measurement, then the
weighted least-squares solution is [10]

Xˆ = ( AΤWA) −1 AΤWY (12)

where W is a simply diagonal and positive definite and

 xˆe   xe 0  ∆xe 
 yˆ  =  y  + ∆y  (13)
 e   e0   e 
 zˆe   ze 0  ∆ze 

and

 ∂h1 ∂h1 ∂h1 

 ∂x ∂ye ∂ze 
 y1 − h1 ( xe 0 , ye 0 , ze 0 , x1,1 , y1,1 , z1,1 , x1, 2 , y1, 2 , z1, 2 )   e 
   ∂h2 ∂h2 ∂h2  ∆xe 
y2 − h2 ( xe 0 , ye 0 , ze 0 , x2,1 , y2,1 , z2,1 , x2, 2 , y2, 2 , z2, 2 ) 
Y =  A =  ∂xe ∂ye ∂ze  Xˆ = ∆ye 
k ×1 M  k ×3   3×1  
  M  ∆ze 
 yk − hk ( xe 0 , ye 0 , ze 0 , xk ,1 , yk ,1 , zk ,1 , xk , 2 , yk , 2 , zk , 2 )  
 ∂hk ∂hk ∂hk 
 ∂xe ∂ye ∂ze 

9-6 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

As can be seen from Eq. (9), the k measurements presenting the k equations are used to solve three unknowns
xe,ye, and ze. The redundancy of the complete set of TDOA’s is used to increase noise immunity.

2.4 Measurement Noise and Bias

The estimates obtained using Eq. (11) are the best statistical estimation if the ni are random and normally
distributed with zero mean [9]. However, in most cases, ni is a combination of random and nonrandom
(or systematic) errors. Therefore, the location estimate based on an assumption of zero-mean random noise
will not be accurate. To represent the measurement noise properly, we consider ni as [9]

ni = nonrandom/systematic bias + random noise

If considering systematic errors, for example, as a constant a1, then we have

ni = a1 + ei (14)

i.e.

∂hi ∂h ∂h
yi − hi ( xe 0 , ye 0 , ze 0 , xi ,1 , yi ,1 , zi ,1 , xi , 2 , yi , 2 , zi , 2 ) = ∆xe + i ∆ye + i ∆ze + a1 + ei (15)
∂xe ∂ye ∂ze

where ei is the zero-mean random part of ni and a1 is a constant error. The state estimates X̂ and the
associated matrix A for Eq. (15) would be

 ∂h1 ∂h1 ∂h1 

 ∂x 1
∂ye ∂ze
∆xe   e 
∆y   ∂h2 ∂h2 ∂h2 
1
X = 
ˆ e
A =  ∂xe ∂ye ∂ze (16)
4×1 ∆ze  k ×4  
  M 
a1   
 ∂hk ∂hk ∂hk
1
 ∂xe ∂ye ∂ze 

If considering measurement errors as a combination of constant and linear time-varying components, e.g.,

ni = a1 + a2ti + ei (17)

i.e.

∂hi ∂h ∂h
yi − hi ( xe 0 , ye 0 , ze 0 , xi ,1 , yi ,1 , zi ,1 , xi , 2 , yi , 2 , zi , 2 ) = ∆xe + i ∆ye + i ∆ze + a1 + a2ti + ei (18)
∂xe ∂ye ∂ze

where ti is time of observation. The state estimates X̂ and the associated matrix A for Eq. (18) would be

RTO-MP-097(I) 9-7

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

 ∂h1 ∂h1 ∂h1 

 ∂x 1 t1 
∆xe  ∂ye ∂ze
∆y   e 
 e  ∂h2 ∂h2 ∂h2 
1 t2 
ˆ
X = ∆ze  A =  ∂xe ∂ye ∂ze (19)
5×1   k ×5  
a1  M 
a   
 2   ∂hk ∂hk ∂hk
1 tk 
 ∂xe ∂ye ∂ze 

Although higher order systematic errors may exist, for simplicity, only the two cases of systematic errors
(i.e., constant and linear time-varying systematic errors) are described here. Obviously, choosing the correct
model is important in location estimation. A well-known theorem from linear model theory is provided in [11]
to identify systematic errors.

3.0 NUMERICAL SIMULATION

The objective of numerical simulation is to demonstrate the effectiveness of the proposed method. The unit of
distance used in the following simulation is in kilometers. In Figs. 5, 6 and 7, we assume that the emitter’s
true position is at (xe,ye,ze)=(9.84, 12.08, 0.00) and the initial position of the emitter for location estimation is
at (xe0,ye0,ze0) = (19.84, 22.08, 10.00), i.e., about 17.32 kilometers away from its true position. In order to
illustrate the accuracy of 3-D location estimation, the location estimation error is used and defined by

Location Estimation Error = ( xˆe ,i − xe ) 2 + ( yˆ e ,i − ye ) 2 + ( zˆe ,i − ze ) 2

where (xe, ye, ze) is the emitter’s true position and ( xˆe ,i , yˆ e, i , zˆe , i ) is the location estimates at the ith time
interval.

In the following subsections, two operation scenarios for geolocation of a radar emitter are given to
demonstrate the effectiveness of the proposed method.

3.1 Geolocation from a Shipborne/Land-Based Station Accompanied by a UAV

Figs. 5 and 6 show location estimation using noisy TDOA measurements from two sensors with sensor 1 on a
UAV and sensor 2 on a shipborne/land-based platform. In the figures, the crosses represent the position of the
moving sensor at different time intervals along the UAV fight path, the square represents the sensor on the
shipborne/land-based station, and the circle represents the true position of the emitter. For simplicity, we take
the position of the shipborne/land-based station to be the origin in the figures. Based on the scenarios given in
Figs. 5 and 6, the distance between the emitter and sensor 1 varies from 20.58 to 21.52 kilometers, while the
distance between the emitter and sensor 2 is 15.58 kilometers. This gives range difference variation from
5.00 to 5.94 kilometers, corresponding to TDOA variation from 16.66 to 19.81 microseconds.

9-8 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

20

Initial guess of
15 emitter location

10 Sensor 1 on a UAV
Emitter ground truth
Sensor 2 on a ship/land-based platform
5 Location estimates
Z (km)

0
Estimation converges
to the ground truth
-5
30

-10 20

-15 10
30 25 20 15 10 5 0
0 X (km)

Y (km)

1.8

1.6
Location Estimation Error (km)

1.4

1.2

0.8

0.6

0.4

0.2

0
0 5 10 15 20 25 30 35
Estimation Intervals

Figure 5: Location Estimation against the Ground Truth with Noisy Measurements (Random Noise).

RTO-MP-097(I) 9-9

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

20

Initial guess of
emitter location
15

Sensor 1 on a UAV
10 Emitter ground truth
Sensor 2 on a ship/land-based platform
Z (km)

Location estimates

0
20
Estimation converges
to the ground truth 15

-5 10
25
20 5
15
10
5 0
0
X (km)
Y (km)

5
With nonrandom error estimation
4.5 Without nonrandom error estimation

4
Location Estimation Error (km)

3.5

2.5

1.5

0.5

0
1 2 3 4 5 6 7
Estimation Intervals

Figure 6: Location Estimates against the Ground Truth with

Noisy Measurements (Random and Nonrandom Noises).

9 - 10 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

Fig. 5 demonstrates the effectiveness of the proposed method in handling random measurement noise,
in which noisy TDOAs were produced by adding Gaussian random noise whose standard deviation was about
33.0 nanoseconds. The associated location estimation errors (in distance) over estimation intervals are also
given in Fig. 5. As can be seen, the proposed estimation method converges to the true location after three
estimation intervals although the initial location of the emitter is assumed to be 17.32 kilometers away from
the emitter’s true position. The first estimation in Fig. 5 was computed using first 15 measurements and
subsequent estimates incorporated each new measurement with one old measurement removed from the data
set.

To illustrate the effectiveness of the proposed method in dealing with nonrandom (or systematic)
measurement errors, Fig. 6 shows the estimation results from noisy TDOAs with an additional measurement
bias of 7.0 microseconds to the Gaussian random noise. As can be seen from the figure, the proposed method
converges to the true location although a bias is present in the measurements. This is because the proposed
method can estimate the systematic error and have it removed while computing the location of the emitter.
The estimated systematic error in this simulation was 7.06 microseconds while the true measurement error is
7.0 microseconds. Fig. 6 also illustrates the location estimation errors (shown by the squares in Fig. 6) without
the consideration of systematic error estimation and its removal. As can be seen from the figure, the location
estimate errors without systematic error removal are large and converge to about 2.0 kilometers.

In Fig. 6, the first estimate was computed using the first 40 measurements and subsequent estimations
incorporated each new measurement while eliminating one old measurement from the data set. It has been
found from numerical simulations that to obtain the same estimation accuracy, the estimation of both location
and a systematic error requires more measurement data than does the estimation of location alone. This seems
reasonable because more parameters are being estimated. However, this requirement for additional data is
certainly justified when accurate locations are desired. It has also been found from the numerical simulations
that with noisy measurements, an increase in number of measurements in location estimation will improve
estimation accuracy.

3.2 Geolocation from a Shipborne/Land-Based Station Accompanied by Two UAVs

Fig. 7 shows location estimates using TDOAs from two moving UAV sensors along their flight paths.
The crosses in Fig. 7 represent the positions of the moving sensors at different time intervals along the UAV
paths. According to the scenario given in Fig. 7, the distance between the emitter and sensor 1 varies
from 30.14 to 30.99 kilometers, while the distance between the emitter and sensor 2 varies from 20.33 to
22.44 kilometers. The range difference varies from 7.73 to 10.65 kilometers, corresponding to TDOA
variations from 25.76 to 35.49 microseconds.

RTO-MP-097(I) 9 - 11

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

Sensor 1

30

25

20 Sensor 2

Initial guess of
15 emitter position

10
Z (km)

5
Estimation converges
to the ground truth
0

-5

Sensor 1 along a UAV path

-10 Sensor 2 along a UAV path
Emitter ground truth
Location estimates 20
-15
15
10
-20
25 5
20 15 10 5 0
0
X (km)
Y (km)

1.8

1.6
Location Estimation Error (km)

1.4

1.2

0.8

0.6

0.4

0.2

0
4 5 6 7 8 9 10 11 12
Estimation Intervals

Figure 7: Location Estimates against the Ground Truth with

Noisy Measurements (Random and Nonrandom Noises).

9 - 12 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

To illustrate the effectiveness of the proposed method in handling both random and nonrandom measurement
errors and the ability to remove these errors, the noisy measurements were produced by adding a constant
TDOA bias of 7.0 microseconds to the random noise in the simulation. As can be seen from the simulation
result shown in Fig. 7, the proposed method converges to the true location after four estimation intervals even
though measurements with a bias are introduced. In the simulation, the first estimation is computed using the
first 40 measurements and subsequent estimates are obtained by incorporating each new measurement while
eliminating one old measurement from the data set. As stated previously, more measurements are required by
the proposed method to handle noisy measurements if accurate location estimation is desired. The estimated
systematic error in this simulation is 7.06 microseconds, which is very close to the actual systematic error
introduced, i.e., 7.0 microseconds. Fig. 7 also shows the estimation error (in distance) over estimation
intervals.

Note: The method is an iterative scheme starting with a rough initial estimate and improving the estimate at
each interval by minimizing the local linear-sum-squared errors. With reasonable initial guesses, the method
does converge in most cases although the absolute convergence is not proven. A direction finding (DF) system
may be used at the base station to provide an initial coarse estimate of emitter location to allow a UAV to get
close to a target to perform better measurements and geolocation of an emitter.

4.0 SUMMARY AND CONCLUSIONS

This paper has addressed passive emitter localization using TDOA technique from UAV sensors working in
conjunction with a shipborne/land-based station. Implicit in the development is the assumption that a
mechanism for measuring the TDOA on radar pulses exists. In this approach, multiple TDOA measurements
are obtained from one or two moving sensors over time intervals along a UAV flight path. A mathematical
model including random and nonrandom (or systematic) measurement errors is derived to achieve an accurate
formulation of a position-location problem when noisy measurements are present. A simple Least-Squares
solution is derived based on the model to solve a 3-D location estimation from TDOA measurements in
combination with known sensor locations. Simulation results have demonstrated that accurate estimates on
position can be obtained using this method because it allows both the estimate and the removal of
measurement bias. In addition, the proposed method can be easily extended to solve a mixed-mode position-
location problem, such as a combination of TDOA, AOA and/or FDOA measurements.

5.0 REFERENCES
[1] TTCP JSA AG-8 Unmanned Aerial Vehicle Concepts, Technology Assessment Final Report Volume I,
AG-8 UAV Technology Assessment Workshop, Washington DC, 16-18 May 2000.

[2] Chan, Y.T. and Towers, J.J., Sequential Location of a Radiating Source By Doppler-Shifted Frequency
Measurement, IEEE Transactions on Aerospace and Electronic Systems, Vol. 28, No. 4, October 1992.

[3] Chestnut, P.C., Emitter location Accuracy Using TDOA and Differential Doppler, IEEE Transactions on
Aerospace and Electronic Systems, Vol. AES-18, No. 2, March 1982.

[4] Smith, J.O. and Abel, J.S., Closed-Form Least-Squares Source Location Estimation from
Range-Difference Measurements, IEEE Transactions on Acoustics, Speech, and Signal Processing,
Vol. ASSP-35, No. 12, December 1987.

RTO-MP-097(I) 9 - 13

NATO/PFP UNCLASSIFIED
 NATO/PFP UNCLASSIFIED

Radar Emitter Localization Using TDOA

Measurements from UAVs and Shipborne/Land-Based Platforms

[5] Fang, B.T., Simple Solution for Hyperbolic and Related Position Fixes, IEEE Transactions on
Aerospace and Electronic Systems, Vol. 26, No. 5, September 1990.

[6] Schau, H.C. and Robinson, A.Z., Passive Source location Employing Intersecting Spherical Surfaces
from Time-of-Arrival Difference, IEEE Transactions on Acoustics, Speech, and Signal Processing,
Vol. ASSP-35, No. 5, August 1987.

[7] Manolakis, D.E., Efficient Solution and Performance Analysis of 3-D Position Estimation by
Trilateration, IEEE Transactions on Aerospace and Electronic Systems, Vol. 32, No. 4, October 1996.

[8] Chan, Y.T. and Ho, K.C., A Simple and Efficient Estimator for Hyperbolic Location, IEEE Transactions
on Signal Processing, Vol. 42, No. 8, August 1994.

[9] Poirot, J.L. and McWilliams, G.V., Application of Linear Statistical Models to Radar
Location Techniques, IEEE Transactions on Aerospace and Electronic Systems, Vol. AES-10, No. 6,
November 1974.

[10] Sorenson, H.W., Kalman Filtering: Theory and Application, IEEE Press, 7-12, 1985.

[11] Seavoy, T., Lavelle, M. and Lommen, L., A Demonstration of Achievable Accuracy for Passive location
of Pulsed Emitter Source, 0-7803-4330-1/98/ 1998 IEEE.

9 - 14 RTO-MP-097(I)

NATO/PFP UNCLASSIFIED