I.
INTRODUCTION
Benchmark for Radar
Allocation and Tracking in
ECM
W. D. BLAIR
G. A. WATSON
Naval Surface Warfare Center Dahlgren Division
T. KIRUBARAJAN
Y. BAR-SHALOM,Fellow, IEEE
University of Connecticut
A benchmark problem for tracking maneuvering targets
is presented. The benchmark problem involves beam pointing
control of a phased array (Le., agile beam) radar against highly
maneuvering targets in the presence of false alarms (FAs) and
electronic counter measurements (ECM). The testbed simulation
described includes the effects of target amplitude fluctuations,
beamshape, missed detections, FAs, finite resolution, target
maneuvers, and track loss. Multiple waveforms are included in
the benchmark so that the radar energy can be coordinated with
the tracking algorithm. The ECM includes a standoff jammer
(SOJ)broadcasting wideband noise and targets attempting
range gate pull off (RGPO). The limits on the position and
maneuverability of the targets are given along with descriptions
of six target trajectories. The “best” tracking algorithm is the
one that minimizes a weighted average of the radar energy and
radar time, while satisfying a constraint of 4% on the maximum
number of lost tracks. The radar model, the ECM techniques, the
target scenarios, and performance criteria for the benchmark are
presented.
Manuscript received August 24, 1996; revised June 18, 1997.
IEEE Log NO. T-AESl34l4IO7970.
The development of this benchmark problem and paper
were funded in part by the Inhouse Laboratory Independent
Research (ILIR) Program at the Naval Surface Warfare Center
Dahlgren Division in Dahlgren, VA; ONWBMDO Grant
N00014-91-J-1950; ONR Grant N00014-97-J-0502; and AFOSR
Grant F49620-94- 1-0150.
Authors’ addresses: W. D. Blair, Georgia Tech Research Institute,
SEAL, 7220 Richardson Rd., Symna, GA 30080; G. A. Watson,
Code B32, Systems Research and Technology Department,
Naval Surface Warfare Center Dahlgren Division, Dahlgren, VA
22448-5100; T. Kirubarajan and Y. Bar-Shalom, Electrical and
Systems Engineering Department, University of Connecticut, Rm.
312, Eng 111, U157, Storrs, CT 06269-2157.
0018-9251/98/$10.00 @ 1998 IEEE
IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
While the problem of tracking maneuvering targets
has been studied extensively [l-51, no standard
or benchmark problems had been identified in the
literature for comparison and evaluation of proposed
algorithms until [6, 71. While design objectives,
operating conditions, and system constraints differ
significantly between tracking problems, clearly
defined benchmark problems in the areas of target
tracking that include data association, multiple
targets, and maneuvering targets have been found to
be very helpful in the assessment and comparison
of existing algorithms. The lack of benchmark
problems h,as hindered the progress of the target
tracking community with respect to many “real
world” problems, and thus, many researchers and their
solutions have been omitted from consideration during
the development and demonstration of systems with
advanced technology.
The first benchmark: [6] included many of the
features of an actual phased array radar and realistic
scenarios as well as many of the restrictions that occur
in a “real world” environment. The testbed simulation
for the first benchmark problem included the effects
of target amplitude fluctuations, beamshape, missed
detections, finite resolution, target maneuvers, and
track loss. The tracking requirements were specified
in terms of limits on the position and maneuverability
of the targets rather than by a set of scenarios. The
“best” tracking algorithm was determined to be the
one that minimized the average number of radar
dwells while satisfying: a constraint of 4% on the
maximum number of lost tracks. Using a filter
performance criterion based on system performance
helped to focus the efforts of the researchers onto
those issues that are irnportant to the system designers,
and thus, illustrated the benefits of modern tracking
(i.e., data processing) through the benchmark.
However, the benchmark of [6] did not include false
alarms (FAs) (Le., false detections) or electronic
counter measures (ECM), which are two critical
elements of any “real-world” tracking problem [8, 91.
While parameter control and tracking in the presence
of FAs was considered for a phased array radar in
[lo], no standard problem was available to the authors
to assess the performance of their new techniques
relative to other existing techniques.
This paper extends the benchmark of [6] to include
the effects of FAs and ECM.’ The inclusion of FAs
is accompanied with multiple radar waveforms so
that the waveform energy can be coordinated with the
tracking algorithm. The ECM includes range gate pull
off (RGFO) on the target and a standoff jammer
--
‘While having the same objective as [7], this paper and the
corresponding simulation program include many improvements over
that of [7].
1097
 Fig. 1. Illustration of beam pointing control of phased array
radar.

(SOJ) broadcasting wideband noise. Fig. 1 gives an

illustration of the problem, where the helicopter is a
SOJ and the dotted aircraft illustrates the effects of % of lost trukr
a RGPO. If the pointing of the radar beam, denoted
by the straight solid lines, is bad, the target will not
be detected. For this benchmark the “best” tracking
1
algorithm is the one that minimizes a weighted
average of the radar energy and radar time, while
Fig. 2. Block diagram of simulation program.
satisfying a constraint of 4% on the maximum number
of lost tracks.
benchmark is on maintaining tracks with an allowance
Fig. 2 shows a block diagram of the MATLAB2
for track reacquisition within a few radar dwell^.^
simulation program for the benchmark. Each
The radar model simulation includes the effects
benchmark participant codes his tracking algorithm
of target amplitude fluctuations, beamshape, missed
in the block entitled “Tracking Algorithm,” which
detections, finite resolution, FAs, a SOJ, and RGPO.
is given the range, bearing, and elevation of the The radar performs search dwells, monopulse track
initial detection of the target. For each experiment,
dwells, and monopulse passive dwells. Eight radar
the tracking errors, radar energy, and radar time are
waveforms that differ primarily in the pulse length
saved. After the last experiment of the Monte Carlo
are available for control @e., selection) from the
simulation, the average tracking errors, average radar
tracking (and radar management) algorithm. Since
energy per second, and average radar time per second3
a waveform with a longer pulse length provides a
are computed for maintained tracks and the percent
higher signal-to-noise ratio (SNR) at the cost of
of lost tracks is also computed. A constraint of 4%
more radar energy, the proper coordination of the
is to be imposed on the number of lost tracks. A waveform selection with the tracking algorithm is an
track is considered lost if the distance between the
important “real-world” issue to be addressed by each
true target position and the target position estimate
participant. For example, a higher target SNR results
exceeds one beamwidth in angle or 1.5 range gates.
in fewer FAs at the costs of more radar energy, longer
When FAs and ECM are present in an actual radar
range of detection of the radar by electronic support
system, algorithms for reacquiring the target and
measures (ESM) sensors, and a potentially higher
“coasting” the target tracks through jamming signals
clutter-to-noise ratio. RGPO and standoff jamming
are required in order to maintain a track. However,
in the mainlobe and sidelobes are included in the
in this benchmark problem, the SOJ power is limited
benchmark problem. In RGPO, the target under track
so that it can be defeated with one of the higher
repeats with delay the radar pulse to pull the radar
energy waveforms: Since track initiation is very
range gate off the target so that no detection of the
expensive with respect to radar time and energy (i.e.,
true target will occur. The time delay is controlled
track initiation requires multiple high-energy dwells
so that the false target is separated from the target
that are closely spaced in time), the emphasis of this with linear or quadratic motion. For tracking an SOJ,
passive measurements of the jammer position (ie.,
2MATLAB is a trademark of The Mathworks, Inc., Natick, MA. angle only) are provided by the radar at the request
3This is inversely proportional to the average sample period. More
details on this are given in Section VI. 5Reacquisition is allowed if it can be accomplished before the
“his restriction is made to limit the scope of the problem and estimation errors exceed the criteria for a track being considered
reduce the problems associated with declaring a track as lost. lost.

1098 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
of the tracking algorithm with bearing and elevation
angles for pointing. Thus, the tracking algorithm is
required to track one target and one jammer. When the
jammer is in the mainlobe of the radar beam pattern,
the target return is corrupted or hidden by the jammer
signal. When the jammer is in one of the sidelobes,
the effective SNR (Le., signal to jammer noise plus
receiver noise) for the target is reduced. The initial
bearing and elevation of the jammer are given to the
tracking algorithm.
The targets exhibit radar cross section (RCS)
fluctuations according to the Swerling I11 type and
perform as much as seven gs of lateral acceleration
and two gs of longitudinal acceleration. Target ranges
vary from 20 km to 100 km, while the target elevation
angle can vary from 2" to 80". Since only one face
of the phased array radar is used, the bearing (or
azimuth) of the target will be confined to =k60".The
average RCSs of the targets are large enough so
that an average SNR of 18 dB is achievable with
the highest energy waveform. The SOJ remains at
ranges near 150 km and perform less than two gs of
acceleration.
A. Results of Previous Benchmark Efforts
An invited session was organized at the 1994
American Control Conference (ACC) with the
first benchmark as the theme for the session. The
benchmark problem was presented along with four
different approaches to the problem. The results of
the invited session [l l-131 and other studies such
as [ 141 provided relative cornparisons of different
tracking algorithms on the first benchmark. The cy -?!,
filter provided an average sample period of 0.85 s,
while a standard Kalman filter with a nearly constant
velocity motion model provided an average sample
period of about 1 s. Thus, for an order of magnitude
increase in computations, the Kalman filter provided
only a 20% increase in the average sample period.
A two-model interacting multiple model (IMM)
algorithm provided an average sample period of 1.3 s,
while a three-model IMM algorithm provided an
average sample period of 1.5 s. Incorporating adaptive
revisit times with a three-model IMM algorithm
provided an average sample period of 2.3 s for about
1.6 orders of magnitude increase in computations
relative to the Q! - ,B filter. An H-Infinity filter solution
was originally included in the invited session at the
1994 ACC, but the authors withdrew the paper when
they concluded that the H-Infinity filter provided no
significant advantage over the Kalman filter [ 151.
The panel discussion held during the invited
session at the 1994 ACC revealed the importance
of tracking maneuvering targets in the presence
of ECM. Thus, a second invited session was
organized for the 1995 ACC with a new benchmark
[7] that included FAs and ECM. However, the
BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM
difficulty of the new benchmark, discrepancies in
the calculation of the radar energy, and deficiencies
in the simulaton program resulted in no quantitive
results for comparison of the performances of the
different algorithms. Qualitatively, the IMM algorithm
and multiple hypothses tracking (MHT) yielded
comparable results with the MHT algorithm being an
order of magnitude costlier in computations.
Many of the deficiencies associated with the
benchmark problem at the 1995 ACC were corrected
to produce the benchmark problem described in
this work. The companion paper [16] presents an
IMMPDAF solution to this benchmark problem, while
[17] presents an adaptive Kalman filter solution. The
adaptive Kalman filter provided an average sample
period near 1.2 s, while the IMMPDAF provided
an average sample period near 2.4 s. The adaptive
Kalman filter also required about 30% more energy
than the IMMPDAF. While aspects of this benchmark
problem are addressed via MHT techniques in [18],
no specific results for MHT are currently available to
assess its performance relative to these two solutions.
Preview of an anticipated article for the Transactions
indicate that the IMM/MHT techniques give an
average sample period of 2.2 s with about 50% of
the radar energy required by the IMMPDAF. The
IMM/MHT technique is reported to require about
5 times the computations of the IMMPDAF. The
parameter selection for the IMMPDAF focused on
the reduction of radar time, while parameter selection
for the IMM/MHT focused on the reduction of radar
energy. Additional studies are planned to obtain
comparisons of the IMMPDAF and the I M W H T
with emphasis on radar time and then radar energy.
B. Organization of the Paper
Section I1 describes the radar model, while
Section 111 discusses the SOJ and RGPO. The target
trajectories are presented in Section 1V along with
the average RCS for each target and the timing for
the use of the RGPO. Section V discusses the inputs
and outputs of the tracking algorithm. The criteria for
evaluating the performance of the tracking algorithm
are given in Section VI, and concluding remarks are
given in Section VII.
II. RADAR MODEL
The radar is a 4 GHz phased array using
amplitude-comparison monopulse with uniform
illumination across the array. Each radar dwell
consists of one phase/frequency discrete-coded pulse
[19]. The range gate is approximately 1575 m for
a track dwell and 10 km for a search dwell with
the number of range bins varying from 70 to 444
depending upon the waveform and dwell type. The
radar beam is quasi-circular with the beamwidth
1099
increasing as the beam is steered off the broadside where R,, = 30 km. The antenna gain along the
direction. The radar beam has a 3 dB beamwidth of boresight of the beam is computed in terms of the
OBw = 2.4" on broadside (i.e., normal to the face of number of elements assuming a half-wavelength
the array) and OB, = 4.5" at a broadside angle of 60". separation distance between the elements. Note that
The two-way radar beam has a 3 dB beamwidth of the gain becomes wavelength independent in this
OBw = 1.6" on broadside and OB, = 3.2" at a broadside approximation. The effective aperture is given by
angle of 60". The beam is pointed to the commanded
direction at the next available sample period since A, = E ~ A (3)
the target trajectories are stored in the data file at where is the aperture efficiency, and A = L,L, is
20 Hz. The minimum time period between sets of the actual area for array dimensions L, and Ly. The
radar dwells is restricted to 0.1 s (Le., 10 Hz). A set gain is given by
of radar dwells may consist of as many as 5 dwells
that are requested simultaneously by the tracking
algorithm. Each radar dwell requires 0.001 s of radar
time to be accomplished. However, for the purposes
of simulation, all of the dwells in a set are modeled
as occurring at the same instant of time with respect
to the target state. For a given detection threshold =T E ~ N ~ N ~ (4)
selected by the participant's radar management where N, and Ny are the number of array elements in
algorithm, the radar model will return for detection i a particular dimension. In this benchmark problem,
of each dwell of the set the following: the SNR Xi, N, = Ny = 55 and the aperture efficiency is 0.5. Thus,
bearing monopulse ratio rLb, elevation monopulse Gt = Gr = 36.8 dB for perfect beam pointing.
ratio YE,
and range Y;, where k denotes the dwell time The noise power is given by
which is assumed to be equal for all dwells in the set.
The sum and difference channels used to produce P, = koqB,,F,
the angular measurements are corrupted with white where
Gaussian errors. Note that the errors in the monopulse
ratio are not purely Gaussian. k, Boltzmann constant
To reference temperature
B,, noise bandwidth
A. Basic Radar Equation
F,, receiver noise figure.
The basic radar equation for a single pulse is given
The uncompressed (i.e., no phase/frequency coding)
in [20] as
SNR Xu is given by

where
e received power from target where the SNR is defined as the signal energy to
4 transmitter peak power noise energy ratio at the output of the matched filter.
Gt transmitter antenna gain in direction of For a radar with a given 4, the SNR at the output
target of the matched filter can be increased by extending
Gr receiver antenna gain in direction of target the length of the radar pulse. In order to achieve good
x transmit wavelength range resolution with a long pulse, pulse compression
4 transmitter propagation factor is used. In discrete-coded pulse compression, the pulse
F, receiver propagation factor width re is composed of N, subpulses of width r,
00 target RCS (Le., r, = Nsrs), where each subpulse is coded with
total losses for radar system frequency and/or phase. For this benchmark problem,
Ltot
sensitivity time control (STC) gain. the radar waveform is biphase and trifrequency coded
G,,(R) [19]. The subpulse width r, is assumed to be equal to
STC is employed to prevent saturation of the receiver the compressed pulse width rc and thus, 7c1 defines
by close-in scattering, and to prevent short range the receiver bandwidth of the compressed pulse
detection of objects with a small RCS. For this B,. Modeling the compressed pulse as coherently
benchmark problem, the STC gain is defined as integrated N, subpulses, the output SNR for the
compressed pulse is given by
R 2 Rstc
(2)
R Rstc

1100 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 TABLE I The power received from the SOJ is given by
Radar Parameters

pr 1 MW (60 dBW)
Gt,Gr 4752 (36.8 dB)
x 7.5 cm (-11.2 dB)
where
Ft 3, l ( 0 dB)
144.5 (21.6 dB)
Ltot
k 1.38 x J / K (-228.6 dB) 5 jammer power at the radar
(4rl3 1984.4 (33 dB) Grj radar receiver antenna gain in direction of
To 290 K (24.6 dB) jammer
F- 2 f3 dB\ effective bandwidth of radar receive filters
B,
Bj effective bandwidth of jammer emission.
Assuming a RCS of one square meter, no beamshape
loss, and the radar parameters of Table I, the radar
For the benchmark problem, the SOJ power is a 5
multiplicative factor of the noise power with B j >> B,..
equation can be simplified to
Thus.

where "io will be fixed at 8 during the benchmark

where no= 257.6 dB. For each of the eight
waveforms and a one square meter target, the SNRs
at 100 km were calculated using (8). The results are
summarized in Table 11, where Ar = 0 . 5 is~ the
~ ~
range resolution of each pulse with c being the speed
of light.
The RCS of the target is modeled as Swerling 111
type, where the density function for the RCS is given
by
problem, G,,,(Rct,) is the STC gain evaluated at the
center range bin of the active dwells and unity on
passive dwells. While yo = 8 will produce a
jammer-to-signal ratio (JSR) that is much less than the
5-6 dB that is needed to hinder a radar operator,
7o = 8 results in a sufficiently high JSR to require the
tracking algorithm to address the presence of the
jammer in the waveform control.

(9) B. Range Measurements

with the average RCS (i.e., a,,,) varying between
target scenarios. The cumulative distribution function
of the RCS is given by
F(ao) = P ( 0 IfJIgo}
Since F(ao) is uniformly distributed between 0 and 1,
a sample or observation of the RCS can be computed
by solving
where x is a random number that is uniformly
distributed between 0 and 1.
The range is computed as the range bin value that
is closest to the true range in order to eliminate the
need for processing adjacent bins for detection. Also,
since the targets are assumed to be points in space
and no interpolation between bins is performed, the
discrete coding results in range measurements with
errors that are uniformly distributed and defined by
the bin resolution (i.e., independent of SNR).
The center of the range gate is placed at the
predicted range Fk provided by the tracking algorithm.
The range measurement is computed as follows. The
prediction error in range is given by
Yk = RL - Pk (14)
where R: is the true range of the target at time k. If
> 775 m, then a missed detection occurs and none
of the measurements that are returned will include

TABLE I1
Waveform Parameters

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1101
the target. If the target is in the range gate, the range
measurement is given by

-
rk = Ar rd [:r] + ;k
where Ar is specified in Table I1 for each waveform
and rd[.] denotes the nearest integer. The range of where
v
a FA is assigned to that range of the bin in which it
occurs.

C. Antenna Gain Patterns

The array consists of 3025 individual elements

&e., 55 elements in elevation and bearing) with cosine
illumination and half power beamwidth of 140".
The broadside of the array is directed at 0" bearing
and 15" elevation. The normalized voltage patterns
= a = 130.2 dB (28)
[21] are defined by K(E;,i?,), the voltage pattern in E; = q(e:q,e,b,) + *;(-e:q,@q)
elevation pointed at 2, and target at EL, and &(B;,i,),
,.
the voltage pattern in bearing pointed at b, and target
-e:,) +
+ Q;(o:~, -os",) (29)
at Bi, and are given by c,j = x$(e;q,e;q) + q&e;q,e;q)
T sin(Nb,) + $(o;~, -Q+ *&e;q, -Ofq)
X(E,&) = -~ (30)
4(1.43) N sin(b2)
sin@, -I-0 . 5 ~ ) sin(b, - 0 . 5 ~ ) QJ':(X,Y> + x)&(B;,i, + Y)1
= [v,(Ei,2, (31)
[ b, + 0 . 5 ~
+ b, - 0 . 5 ~
q ( x , Y >= [v,(E,,& + X ) & ( B h , i k + Y)1 (32)
(16) = O/,s cos(2, - e b ) (33)
T sin(Na2)
&(Bi,i,)= -~ o;, = e,,/ cos<b (34)
4( 1.43) N sin(a,)
sin(ul + 0 . 5 ~ ) sin(al - 0 . 5 ~ ) and
[ ul + 0 . 5 ~
+ U, - 0 . 5 ~
re pulse length
q5i phase of voltage of target return
where phase of voltage of SOJ
a, = 0.25sin(B;)
6;
Bi bearing of SOJ at time k
,.
u2 = O.S7r[sin(BL)- sin(b,)] (19) E; elevation of SOJ at time k.

b, = 0.25 sin(,!?; - e , ) (20) Both 6; and 4; are independent and uniformly

distributed between 0 and 2n. The and e:, denote
b, = 0.5~[sin(E;- e b ) - sin(& - e,)] (21) the squint angles6 in elevation and bearing,
respectively, where B,po = 0.9" = 0.0157 rad. The
N = 55 (22)
N,(O, 1) denote Gaussian random variables with zero
eb = 15". (23) mean and unity variance. The two independent
Gaussian random variables in (27) are used to
The factor of 1.43 has been included to give a
generate a Rayleigh distributed amplitude for the
normalized sum channel voltage at boresight for a
jammer. Let
broadside angle of 0". Note that the orientation of the (SA), = (s;$ + (S'e,)? (35)
face of the antenna array was selected so that the far
field pattern is approximately separable into sin(&) Then the SNR is computed by Rk = 0.5(~j,)~, where sfc
and sin(;, - e,). is the peak amplitude of the sum signal.
Equations (24)-(34) are used to compute a
D. Channel Voltages measurement for each range bin by setting = 0 for
the bins that do not include the target. The passive
The received voltage is related to the received measurements are obtained by setting r; = 0 and
power as V, = m. In search or track mode, the
sum voltage at time k, sk, is composed of the in-phase 6Note that the squint angles are usually fixed with respect to the
and quadrature portions. The in-phase and quadrature off-broadside angle. The squint angles have been varied with the
components of the sum voltage $6 normalized by the off-broadside angle to increase the minimum monopulse error slope.

1102 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 4.5 1 1
angle, where the gain has been normalized by the
gain at broadside. Thus, antenna gains G, and G, are
decreased by 0.4 dB at a broadside angle of 60". This
antenna gain pattern is optimistic because the mutual
coupling between the elements of a real phased array
antenna results in G, and G, being decreased by about
3 dB at a broadside angle of 60".
The difference voltages in elevation di and bearing
d i have in-phase and quadrature components. The
in-phase and quadrature components of df and d i are
written as
dFk = rixi[Q;(e;q?e:q) + q;(e!q,-e,b,>
-5.1 I
-60 -40 -20 0 20
OFFBROADSIDE ANGLE (DEG)
40 60
Qi(-qq,e,b,) - e;(-e;q,-e:q)l~~~4:
-

Fig. 3. One-way beamwidth versus off-broadside angle. + r,$ui(e;q,e;q)+ Q/(qq, -e;q)

- q&e;q,e:q) -Q&~;~,-~;~)ICOS~{

+ Nf(O,1) (36)
+~
d& = r:c;[e;(e:q,'$q) ~ ( ~ ; q ~ - @ q )

s;(-e;q,e,b,)
- - -e;q)] sin$:
+ rL[i$(e;q,e:q)+ XP;(B;~,
- $(-e;q,e;q) - -e,b,)l sin4{

+ N,(O,1) (37)
d;k = r;x;[*;(qq,e:q)
+ q;(-e;q,e,b,)
-04.-
-60 -40 -20 0 20 40 60 - Q;(qq,
-e;q) - COS 4;
OFF-BROADSIDE ANGLE (DEG)
+ ri[wd(e;q,s:q)
+ xu&-e;q,e:q)
Fig. 4. Normalized antenna gain versus off-broadside angle.
-
-etq) - $(-e;q, -e,",)] COS^^
G,,, = 1 for all range bins. The observed SNR 8; + N7(0,1) (38)
and range I-;are returned for every range bin i with
Ri s F: qh
, where is specified by the tracking d& = r:C;[q;(e;q,e:q) + q;(-e:q,e;q)
algorithm for the dwell set at time k . In other words, @;(e;q,-e;q) - w;(--qq,-e:q)] sin&
-
a measurement is returned for range bin i if Ri > Fk.
Thus, the probability of a FA Pfa is specified by the + r/[u/(qq,etq)+ e/(-efq,e&)
tracking algorithm. The amplitudes of the,FAs are
Rayleigh distributed with parameter ~o(C~)2G,,,(R) qd(qq,-etq) - Qj(-qq,
- sin${
+ 1. The amplitude of the measurements of a passive + N,(O,1) (39)
dwell will be Rayleigh distributed with parameter
-yo(Ch>2+ 1. where the receiver errors in the difference channels
Fig. 3 shows the one-way beamwidth eSw of the have been modeled as independent of the receiver
sum channel versus the off-broadside angle, which errors in the sum channel.
is the angle between the boresight of the beam and
a vector orthogonal (i.e,, normal) to the face of the E. Monopulse Processing
array. The broadside angle is treated independently in
elevation and bearing. The one-way beamwidth is the Monopulse processing is a simultaneous lobing
angle between the two half-power points on the sum technique for determining the angular location of a
channel beam. The one-way beamwidth varies from source of radiation or of a "target" that reflects part
2.4" at broadside to 4.5" at a broadside angle of 60". of the energy incident upon it [22]. The monopulse
The two-way beamwidth varies from 1.6" at broadside ratio is formed by the (complex) division of the
to 3.2" at a broadside angle of 60". Fig. 4 shows the difference-channel voltage phasor by the sum-channel
normalized antenna gain versus the off-broadside voltage phasor. Monopulse ratios are formed

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1103
 10- if the squint angles had been fixed at their boresight
8- values. Since the standard deviations of the angle
measurements are inversely proportional to k,,
maintaining a higher k, provides more accurate
measurements at the off-broadside angles.
The monopulse ratios in elevation and bearing are
8 computed as
Y -21
2
8 -4-

'
0
-6-

8-

-10-
1 -0.5 0 0.5 I When detecting a single target, the directions of
NORMALIZED OFF-BORESIGHT ANGLE (BEAMWIDTH) arrival in elevation and bearing are given by
Fig. 5 . Monopulse error function at broadside.

Using the linear approximation for the monopulse

error function gives the measurements of elevation
and bearing as

where
A"
-60 -40 -20 0 20 40 60 volt
-
k i = 48cOs(zk - eb) (47)
OFF-BROADSIDE ANGLE (DEG) volt. rad
Fig. 6. Monopulse error slopes versus off-broadside angle. volt
kk = 48COS(ik) -
volt. rad
separately for the elevation and bearing directions and zk and i k are the elevation and bearing pointing
of arrival. For every active or passive track dwell, commands from the tracking algorithm. Estimates for
monopulse ratios are returned for every measurement the standard deviations of the measurements are given
with an amplitude that exceeds the specified detection bv
threshold %kh. The monopulse ratios are then used
to estimate the monopulse error function value 7 for
determining the measured direction of arrival. The
processing of monopulse ratios is considered as in
[23] for 1) target only measurements, 2) SOJ only,
and 3) target and SOJ within the mainlobe of the
antenna pattern.
The monopulse error function at broadside is
shown in Fig. 5 along with the commonly used linear Rk > 3 dB. (50)
approximation to the monopulse error function, which Note that (49) and (50) differ from the expression
is given by given in [22] for the standard deviation. The variance
expression in [22] is based on an assumption of 171 <<
7 M k,6, -6sw < 6 < 6,, (40) 1 and 8 t k > 12 dB, while (49) and (50) are valid for
where k, is the average monopulse error slope for 171 < 2 and %k > 3 dB [23]. Fig. 7 gives the standard
-0.58,, < 6 < 0.58,,. Fig. 6 shows the average deviation for various off-broadside steering angles and
monopulse error slopes versus the off-broadside 8 k = 15 and 20 dB with perfect pointing (i.e., d i = 0).
angle. Since the squint angles are increased with Note that %k can vary rather significantly between
the broadside angle, the monopulse error slope k, consecutive measurements since the targets have a
achieves a minimum of 24 rad-' rather than 12 rad-' RCS of Swerling I11 type.

1104 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 6 a passive dwell or an active dwell, where only the
1'
-60 -40 -20 0 20 40
OFF-BROADSIDEANGLE (DEG)
Fig. 7.
When processing the passive measurements of
the SOJ, the jammer-to-noise ratio (JNR) Jk can
be estimated from N observations within a dwell
according to
where 8;is the observed SNR of range bin i at time
k . The direction of arrival estimates for the elevation
and bearing of the SOJ are given by
where r; and rLb denote the monopulse ratio for
range bin i at time k for the elevation and bearing,
respectively. Using ( 5 2 ) and (53) in (45) and (46)
provides the measured elevation and bearing of the
SOJ. Estimates for the standard deviations of the SOJ
measurements are given by
-0.5
(54)
When the target and the SOJ are in the mainlobe
of the antenna pattern, the angular location of the
SOJ can be observed as discussed in (51)-(55) with
BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM
range bins that do not include the target are to be
considered. Thus, the direction of arrival estimates for
the target in the presence of the SOJ in the mainlobe
are given by
Rk > Jk + 3 dB (56)
I Rk > Jk f 3 dB (57)
60
where qf and q t are the true direction of arrival
values for the SOJ and Jk is the JNR at time k.
Equations (56) and (57) indicate that if the SNR Rk
is 12 dB greater than the JNR J k , the effects of the
SOJ can be ignored. Note that while qf and qf can
be estimated from past data by tracking the SOJ, Jk
should be estimated on the active dwell since the SOJ
power can fluctuate between consecutive radar dwells.
Representing the monopulse error function as linear
gives the elevation and bearing measurements for the
target in the presence of the SOJ within the mainlobe
as
Rk > Jk + 3 dB (59)
where EL and B; are the true elevation and bearing of
the SOJ. Since the true elevation and bearing of the
SOJ are not known to the radar system, bearing and
elevation estimates from a track filter are required
unless Rk > Jk + 12 dB. Estimates of the standard
deviations of the measurements are given by
1105
Under the assumption of a linear monopulse error
function, (58) and (61), (59) can be modified with
40r-----7
38

36 -
B
%
w 34-
2
z
n 32 -
Ill. ECM TECHNIQUES 0

The ECM for the benchmark problem involves the

: 30- t=o

SOJ discussed in Section I1 and RGPO. The times

at which the SOJ and RGPO impact the trajectories
are discussed in the next section. The SOJ transmits 261
I I
broadband noise toward the radar and has the effect 120 125 130 135 140 145

of increasing the level of noise. However, since the X COORDINATE (KM)

SOJ noise enters the sum and difference channels with
the same phase, the direction of arrival of the SOJ

I
can be estimated with the monopulse ratios, where
the target amplitude is Rayleigh, since the SOJ noise
is modeled as Gaussian. The SOJ flies an oval (race
course) holding pattern in a clockwise direction at
an altitude of 3.05 km and speed of 168 m / s . The
SOJ is approximately 150 km from the radar and its
trajectory is shown in Fig. 8. The aircraft flies straight
and level for a period of approximately 40 s upon
which a 1.5 g turn is performed through a 180 deg
heading change. Straight and level, nonaccelerating
flight is continued for another 40 s, where a second
1.5 g turn is performed until the aircraft returns to its A , 200
50 100
original position. The cycle starts again with another
TIME ( S )
40 s of level flight. The SOJ transmits broadband
noise that impacts the radar with power no more Fig. 8. Trajectory of SOJ.
than yo times the receiver noise power as discussed
in Section 11. For this benchmark problem, yo = 8. where apo is the acceleration of pull-off of the false
Thus, the SOJ will not completely hide a target and target. The measurements of the RGPO will be
it can be defeated with a higher energy waveform generated as those for the target with (26) modified
in this benchmark problem. While the SOJ energy according to
enters into every radar dwell, the effects of the SOJ
are for the most part negligible when it is not in the (66)
mainlobe of the antenna pattern. Note that this is not
the case for SOJs transmitting a higher level of where y1 is the amplification factor of the RGPO.
power. However, when the times of arrival of the target echo
In RGPO, the target under track repeats with and the false target are less than the resolution of the
delay and amplification the radar pulse so as to pull radar, interference occurs. If the ranges of the two
the radar range gate off the target. The time delay is observations correspond to two different range bins,
controlled so that the false target is separated from the two observations will be received. Otherwise, both
target with linear or quadratic motion. For the linear echoes result in one observation. For this benchmark
case, the range of the false target with respect to the problem, two returns are assumed to be resolved and
radar is expressed as modeled independently when their times of arrival
differ by more than r J 3 . On the other hand, if
Rkf' = Ri f vpo(tk- to) (64)
3/R,ft- R i J
where Ri is the range of the actual target, vpo is the <1
2cre
rate of pull-off, tk is the time of the radar dwell, and to
is the initial reference time of the RGPO false target. the target and RGPO amplitudes are modified
For the quadratic case, the range of the false target according to
with respect to the radar is written as
Rkf' = RL + iapo(tk- to)2
1106 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 ,c#
35

L J
30 ~

where Ek = 4; - 4kf' with$kf' being the phase of the 25 -

false target signal, and c is the speed of light. Note
that the discrete coding of phase has been neglected 20-

in the modeling of the interference of the two returns. z

The RGPO is activated by a radar dwell on the target. 8 15-
8
Thus, RGPO reference times are dependent on the 10 - t=165 s
dwell times of the radar. For the benchmark problem,
RGPO will be activated on the next pulse after a time 5- , , ,

specified in the trajectory file. The reference time to

will be chosen as the dwell time tk minus a uniformly 0
0 10 20 30 40 50 60 70 80
distributed random number between 0.1 and 0.5 ps. X COORDINATE (KM)
Thus, tk - to will be uniformly distributed between

-. 40r
0.1 and 0.5 ps. The amplification factor y1 is chosen
as a uniformly distributed number between 1 and 4. 35-
The rates for both modes have been selected to yield
a 2 km displacement in range of the false target in $ 30-
20 s. Thus, the pull-off rates are vpo = 100 m / s and 0
upo = 10 d S 2 .
5 25-
5
Cover pulse is an RGPO technique that uses tsl 20-
echoes without the waveform coding and attempts 2
15-
to cover the actual target return. When cover pulse
is used, a noncoherent detection occurs. Typically the 10-
range resolution of the noncoherent detection is rather 6
bad. For this benchmark problem, cover pulse will not z" 5-

be considered. : 0 -
While electronic counter counter measures 0 20 40

(ECCM) are part of any modern radar system, the TIME ( S )

benchmark radar model does not include any specific
ECCM techniques other than waveform control and
beam pointing control that can be accomplished from
the tracking algorithm part of the simulation program.
IV. TARGET TRAJECTORIES
The targets exhibit RCS fluctuations according
to the Swerling 111type model and perform as much
as seven gs of lateral acceleration and two gs of
longitudinal acceleration. Target range can vary from
20 km to 120 km, while the target elevation angle
varies from 2" to 80". Since only one radar face is
used, the bearing of the target will be confined to
f60". The average RCSs of the targets are large
enough so that average SNRs of 18 dB are achievable
with the highest energy waveform. The SOJ remains
at ranges greater than 150 km and perform less
than two gs of acceleration. The RGPO can only be
employed when the angle between the target heading
vector and radar range vector is within 180" f60".
While six target trajectories are specified in this
section, the tracking algorithm should be designed to
handle targets satisfying the general specifications.
The first target trajectory is shown in Fig. 9 and
represents a large aircraft, such as a military cargo
aircraft. From an initial range of 80 km, the target
Fig. 9 Trajectory of target 1.
flies at a constant course with a speed of 290 m / s at
an altitude of 1.26 km for the first minute. The aircraft
then performs a mild 2 g turn and continues on the
new course for a period of 30 s, where a 3 g tum is
made and the aircraft flies away from the radar to a
final range of 70 km. The uaveis 4 m2, RGPO starts
after 15 and 40 s, and the SOJ is within 2" of the
target line-of-sight (LOS) from 82 s to 93 s.
The second target trajectory is shown in Fig. 10
and represents a trajectory which would be expected
by a smaller, more maneuverable aircraft, such as a
Learjet or other similar high performance commercial
aircraft. Target 2 initializes at a range of 63 km and
speed of 305 m / s and altitude of 4.57 km. The target
performs a 2.5 g turn through 90 deg of course
change. After the turn, the target then descends
gradually to an altitude of approximately 3.05 km.
A 4 g tum rolling out to straight and level flight is
performed at a constant speed of 305 m / s and the
trajectory ends near a range of 28 km. The gaveis
2 m2, RGPO starts after 12, 50, and 95 s, and the SOJ
does not approach the target LOS.
The trajectories of Targets 3 and 4 are shown in
Figs. 11 and 12 and represent medium bombers flying
at high speeds with good maneuverability. Target 3

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1107
,. -15-
B -20 - k150 s
F
8
0
8
-25-
-30-

-35 -

-40 -

-45
-50 10
i 20 30

X COORDINATE (KM)
40 50 60

X COORDINATE (KM)

7-

6-

5-

4-

3-

i
20 25 30 35 40 45

HORIZONTAL RANGE (KM)

50 55 60 65 70 250
0 20 40 60 80

TIME ( S )
100 120 140

15

10
~

20
ilr 40 60 -

has an initial speed of 457 m / s and flies straight and

level for the first 30 s. A 4 g turn is then performed
through a 45 deg course change. Straight and level,
nonaccelerating flight is continued for the next 30 s.
A second 4 g turn through a 90 deg course change to
80

TIME ( S )

Fig. 10. Trajectory of target 2.

I100 120 ~ 140 160
8
4

0 20
I
LL40 60

TIME (S)
80

Fig. 11. Trajectory of target 3

I00

a speed of 251 m / s and an altitude of 2.29 km for

the first 30 s. A 4 g turn is performed through a
course change of 45 deg. After another 30 s, a 6 g
tum is performed as the throttle is increased to full
afterburner. The aircraft pitches up and climbs to an
120 14(

straight and level flight is performed while the aircraft altitude of 4.57 km. Following the climb, straight
decelerates to a speed of 274 d s . The gaveis 1.5 m2, and level, nonaccelerating flight is maintained for
RGPO starts after 30, 100, and 130 s, and the SOJ the completion of the trajectory. The gaveis 1.7 m2,
does not approach the target LOS. Target 4 maintains RGPO starts after 35 and 65 s, and the SOJ is within

1108 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 40t i

-20
I 10 -

5-

5 10 15 20 25 30 $0 30 40 50 60 70 80 90 100

X COORDINATE (KM) X COORDINATE (KM)

10

9-
91
8-

7-

6-

5- [=I85 s t=182s I
4-

3-
( t=0
2-

I
ii
0
, 7, , , i
1' 3 5 IO 15 20 25 30 35 40 45 40 50 60 70 80 90 100

HORIZONTAL RANGE (KM) HORIZONTAL RANGE (KM)

70

n
j;
l l

30

20

IA
80 100

TIME ( S ) TIME ( S )

Fig. 12. Trajectory of target 4 Fig. 13. Trajectory of target 5 .

2" of the target LOS from 5 s to 12 s and 118 s to
122 s.
Targets 5 and 6 are shown in Figs. 13 and 14 and
represent fightedattack aircraft. Target 5 is initialized
in a thrusting acceleration at an altitude of 1.5 km.
After a period of 30 s, a 5 g turn is performed while
maintaining full throttle. This turn is followed 20 s
later by a 7 g turn. Following the second turn, straight
and level, nonaccelerating flight is performed for 30 s
upon which a 6 g turn is performed concurrently with
a pitch up and a climb. After an altitude of 4.45 km
is reached, straight and level, nonaccelerating flight is
flown for the completion of the trajectory. The gave
is 1.2 m2, RGPO starts after 5, 25, and 52 s, and
the SOJ is within 2" of the target LOS from 75 s to
83 s. Target 6 starts at a speed of 426 d s and an
altitude of 1.55 km. Constant speed and course are
maintained for a period of 30 s upon which a 7 g
turn is performed. The new course is maintained
for another 30 s. A 6 g turn is performed while the

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1109
 40 - V. TRACKING ALGORITHM

:
30 - Fig. 15 shows the input and output of the tracking
algorithm. At any revisit time, as many as five
h

B
6z
n
20 -

10- i_ dwells of any combination of waveforms in Table

11 can be requested by the tracking algorithm. The
collection of dwells requested at a given time by
the tracking algorithm will be referred to as a dwell
0
8* O-

-10 -

-20 -
7 t=188 s 1
set. The tracking algorithm routine will receive for
each dwell in the requested set the Rk, ri, r j , and rk
for each range bin in the range gate with ?Rk above
the commanded SNR threshold Xih. Each quantity
is provided to the tracking algorithm in a matrix
20 30 40 50 60 70 80 where columns 1 through 5 correspond to dwells 1
X COORDINATE (KM) through 5 of the requested set, respectively. Also,
indicators of coherenthoncoherent detections (not
used in this benchmark problem) and the waveforms
used in the previous radar dwells are given to the
tracking algorithm. Since the first detection occurs
on a search dwell, the tracking algorithm will be
required to initialize the track from a measurement
of range and the bearing and elevation angles of the
[
o
250 r beam during the detection @e., no monopulse error
correction in the first measurement). The time for
the next measurement and the pointing commands in
A

range Fk, bearing bk, and elevation z k are computed

by the tracking algorithm. The tracking (and radar
management) algorithm also requests the type of
I
200 1
0 20 40 60 80 100 120 140 160 180
dwell (i.e., search, track, or passive) and selects a
TIME ( S )
waveform (i.e., pulse length) for active dwells from
Table 11.

VI. CRITERIA FOR EVALUATION OF TRACKING

For evaluation of the tracking algorithms,

each algorithm shall maintain tracks on the given
trajectories with a maximum track loss of 4% and
no indication of the target type or number. A track
is declared lost if the error in the estimated position
of the target is greater than 1 two-way beamwidth
in angle or 1.5 range gates (about 2362 m) in range.
Since the phased array radar will operate in different
environments, the primary measure of tracking
0' L 20 40 A
60 performance will be a weighted sum of the average
radar energy per second and the average radar time
TIME ( S )
per second. The average energy per second E,,,
Fig. 14. Trajectory of target 6. is the sum of energy of the track dwells requested
by the tracking algorithm divided by the number
throttle is reduced and the aircraft is nosed over seconds of the trajectory. The radar energy for a
in order to decrease altitude. After a final altitude given radar dwell is computed by multiplying the
of 0.79 km is obtained and a time span of 30 s, peak transmit power pt by the effective pulse width
another 6 g turn and full throttle is commanded. T ~ The
. average radar time per second Faveis the sum
After approximately 30 s, a 7 g turn is performed, of radar time of the track dwells requested by the
and upon completion of the turn, straight and level, tracking algorithm divided by the number seconds
nonaccelerating flight is maintained for the completion of the trajectory. Since the radar time required for a
of the trajectory. The average ,rc is 1.9 m2, RGPO dwell is typically dominated by the signal processing
starts after 15, 75, and 102 s, and the SOJ is within 2" time, each radar dwell is assigned a cost of 0.001 s
of the target LOS from 69 s to 76 s and 98 s to 126 s. of radar time. Note that Fa,, is inversely proportional

1110 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
 Time of previous dwell set
Observed SNIls on dwells at tk

Range measureiuents at tk

hIoiiopulsr ratios in clcvation aiid hcariiig at I ;

SNR detection threshold for dwell set a t t;
Indicators of Iioucohereiit detections (i+ C'n = 1)
Coiiniiaiided time for next chell set
Range, hcaxing. aiid clcwtion for I)cani pointing
control for dwell set at, t;+1
Radar dwell t.ype indicator (ix., search)
M'wefnrm select,ion at. t h + l
SNR detection tlireshold for dwell set a t tk+l

Mean and errnr covariance for target state

estiinate at tinie t; based on measurements
throngh t;
1 .I'k .i:l. ?k i; fk tk 2, 1"'
Fig. 15. Input and output for tracking algorithm.

to the average sample period. The quantities Ea,, and

-
T,, are computed for each Monte Carlo experiment.
Two cost functions for each target are defined as
c, = E,,, + 1037;,,,
c, = +1 0 ~ ~ , ,
where C, corresponds to a period of operation when
radar energy is critical, and C, corresponds to a period
of operation when radar time is critical. Note that the
objective is to minimize one of the two cost functions
upon request. Thus, results should be presented
for minimizing C, and C,, separately. However,
C, and C, should be computed for each target and
presented as the results may indicate the particular
benefits of a given algorithm. For a final assessment
of algorithm performance, each cost in (70) and (71)
is to be averaged over all the targets with the costs
from Target 1 being counted twice, since Target 1
represents a benign target of which many are expected
in the surveillance region of the radar. The algorithm
that provides the minimum of both costs by a simple
change of the design parameters will be considered
the superior algorithm. Note that the energy and radar
time during initialization should be included because
initialization is a portion of the tracking and in all
averages Target 1 should be counted twice.
A secondary measure of algorithm performance is
the computer resources required. Each investigator
will present an assessment of the average number
operations in terms of Kalman filters that their
algorithm will require in final implementation. If
the algorithm has computational requirements that
vary over time, an assessment of the peak number
of operations should also be presented. The average
sample periods and root-mean-square errors (RMSEs)
in the position, velocity, and acceleration estimates
(70) should be plotted for illustration of algorithm
performance. The results should be reported in a
(71) table as shown in Table 111, where the average of the
results are to be reported in the last row. The fourth
column indicates the maneuver density which was
computed as the percent of the total time that the
target acceleration exceeds 5 d s 2 . 'The fifth column is
the average sample period, while the sixth and seventh
columns are the RMSE in position and velocity,
respectively. The eighth and ninth columns are the
costs of (70) and (71), respectively. The tenth column
is the percentage of tracks that are lost during the
Monte Carlo simulation.
VII. CONCLUDING REMARKS
When developing a solution to this benchmark
problem, one common tracking algorithm should
be designed and implemented for all six target
trajectories. The tracking algorithm should be
designed to address the general specifications of the
targets and evaluated with the six trajectories. While
the tracking algorithm can adapt to the trajectory, the
adaptation should be automatic in that no user-defined
inputs denote a target type or specific number.
After a table similar to Table 111 is completed to
summarize the results, the strengths of the solution
should be discussed along with the weaknesses.
Note that while the best algorithm is the one that

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1111
 TABLE I11
Summary of Results
5 182 68 70
6 188 70 59
Ave*
Note: This average i s computed with results of target 1 being counted twice.
minimizes the two cost functions of Section VI, the
computational resources required to achieve a given
performance are very important when selecting a
tracking method for a radar system. Thus, solutions
with various computational requirements are of
interest. Furthermore, this benchmark problem can be
used to illustrate the application of different tracking
methods. Additional documentation of this benchmark
problem is given in [23] along with the simulation
pr~gram.~J
In the development of this benchmark problem,
a number of simplifying assumptions were made
in order to limit the scope of the problem and the
complexity of the simulation program. First, the
targets were modeled as point targets with RCS
fluctuations that were independent of the aspect
angle of the target with respect to the radar. In an
actual radar tracking system, the returns from the
targets can occupy multiple range bins, glint errors
are common for targets at the closer ranges, and the
RCSs of targets can change rapidly when targets
maneuver. Second, the effects of closely spaced
targets have not been considered. If two targets are
separated by less than a beamwidth in angle and their
ranges are not fully resolved, the returns from the
two targets will interfere. In a monopulse tracking
system, this interference can be catastrophic to the
tracking. Third, sea-surface induced multipath has
been neglected by considering only targets with
altitudes above the region where multipath reflections
corrupt the monopulse processing. Fourth, the power
of the SOJ has been limited so that it can be defeated
with the higher energy waveforms. In an actual
system, the SOJ power could require the averaging
of the returns from many dwells or coasting the
track while both the target and the SOJ are in the
mainbeam. Also, if the SOJ power is sufficiently
7The computer programs are also available in electronic format
from the authors.
8The effects of the RGPO and/or the SOJ can easily be removed
from the simulation program by a simple change of the input file.
1112 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998
high, sidelobe jamming can become a serious issue.
Fifth, the potential for a jammer onboard the target
has been ignored. Sixth, the waveform types have
been limited to fixed waveforms with discrete codes.
Pulse Doppler and linear Frequency Modulated (FM)
waveforms are commonly used radar waveforms,
while adapting the discrete coding of the waveforms
may prove beneficial in the presence of unresolved
targets, multipath, or clutter. Seventh, the effects
of background clutter have been neglected. Eighth,
track initiation in a cluttered environment has not
been considered. Many of these issues are open
research problems to be considered in the future.
Other issues for future benchmark problems include
unresolved targets, sea clutter, chaff, track initiation,
and sea-skimming targets.
ACKNOWLEDGMENTS
The authors are grateful to Scott Hoffman for
generating the target trajectories and Greg Gentry for
contributing to the development of the radar model.
REFERENCES
[l] Blaclanan, S. S (1986)
Multiple Target Tracking with Radar Applications.
Dedham, MA: Artech House, 1986.
[2] Bar-Shalom, Y., and Li, X. R. (1995)
Multitarget-Multisenor Trucking: Principles and
Techniques.
Storrs, C T YBS Publishing, 1995.
[3] Bar-Shalom, Y. (Ed.) (1991)
Multitarget-Multisenor Tracking: Advanced Applications,
Vol. I.
Dedham, MA: Artech House, 1991.
[4] Bar-Shalom, Y.(Ed.) (1992)
Multitarget-Multisenor Trucking: Applications and
Advances, Vol. 11.
Dedham, MA: Artech House, 1992.
[5] Bar-Shalom, Y., and Li, X. R. (1993)
Estimation and Tracking: Principles, Techniques, and
Sojhare.
Norwood, MA: Artech House, 1993.
[6] Blair, W. D., Watson, G. A,, and Hoffman, S. A. (1994)
Benchmark problem for beam pointing control of phased
array radar against maneuvering targets.
In Proceedings of I994 American Control Conference,
Baltimore, MD, June 1994, 2071-2075.
 Blair, W. D., Watson, G. A., Gentry, G. L., and Hoffman,
S. A. (1995)
Benchmark problem for beam pointing control of phased
array radar against maneuvering targets in the presence of
false alarms and ECM.
In Proceedings of 1995 American Control Conference,
Seattle, WA, June 1995, 2601-2605.
Daum, E (1992)
A system approach to multiple target tracking.
In Y. Bar-Shalom (Ed.), Multitarget-Multisensor Tracking:
Applications and Advances, Vol. 11.
Dedham, MA: Artech House, 1992.
Daum, F. (1996)
Review of Multitarget-multisensor tracking: Principles
and techniques.
IEEE Aerospace and Electronic Systems Magazine, 12, 7
(July 1996), 39-42.
van Keuk, G., and Blackman, S. S. (1993)
On phased-may radar tracking and parameter control.
IEEE Transactions on Aerospace and Electronic Systems,
29, 1 (Jan. 1993), 186-194.
Daeipour, E., Bar-Shalom, Y., and Li, X. (1994)
Adaptive beam pointing control of a phased array radar
using an IMM estimator.
In Proceedings of 1994 American Control Conference,
Baltimore, MD, June 1994, 2093-2097.
Kalata, P. (1994)
An Q - p target tracking approach to the benchmark
tracking problem.
In Proceedings of I994 American Control Conference,
Baltimore, MD, June 1994,2076-2080.
Sastry, C. R., Slocumb, B. J., West, P. D., Kamen, E. W.,
and Stalford, H. L. (1994)
Tracking a maneuvering target using jump filters.
In Proceedings of I994 American Control Conference,
Baltimore, MD, June 1994, 2081-2087.
Blair, W. D., and Watson, G. A. (1994)
IMM algorithm for solution to benchmark problem for
tracking maneuvering targets.
In M. K. Masten, L. A. Stockum, M. M. Bimbaum, and
G. E. Sevaston (Eds.), Acquisition, Tracking, and Pointing
VlII, Proceedings ifSPIE 2221, 1994, 303-316.
Tsaknakis, H., and Athans, M. (1994)
Tracking maneuvering targets using H, filters.
In Proceedings of 1994 American Control Conference,
Baltimore, MD, June 1994, 1796-1803.
Kirubarajan, T., Bar-Shalom, Y., Blair, W. D., and Watson,
G. A. (1997)
IMMPDAF solution to benchmark for radar resource
allocation and tracking targets in the presence of ECM.
This issue, 1115-1 134.
West, P. D., Slocumb, B. J., and Kamen, E. W. (1996)
Adaptive Kalman filter solution to the second benchmark
problem.
Final report for Contract N00178-95-M-8217, Naval
Surface Warfare Center Dahlgren Division, Dahlgren, VA,
Jan. 1996.
Popoli, R. E, Blackman, S. S., and Busch, M. T. (1996)
Application of multiple hypothesis tracking to agile beam
radar tracking.
In 0. E. Drummond (Ed.), Signal and Data Processing
of Small Targets 1996, Proceedings of SPIE 2759, 1996,
418-428.
Cook, C. E., and Bernfeld, M. (1993)
Radar Signal: An Introduction to Theory a d Applications.
Norwood, MA: Artech House, 1993.
Edde, B. (1993)
Radar: Principles, Technology, Applications.
Englewood Cliffs, NJ: Prentice Hall, 1993.
Skolnik, M. E. (1980)
Introduction to Radar Systems (2nd ed.).
New York: McGraw-Hill, 1980.
Sherman, S. M. (1984)
Monopulse Principles and Techniques.
Dedham, MA: Artech House, 1984.
Blair, W. D., and Watson, G. A. (1996)
Benchmark problem for radar resource allocation and
tracking maneuvering targets in the presence of false
alarms and ECM.
Technical report NSWCDD/TR-96/10, Naval Surface
Warfare Center Dahlgren Division, Dahgren, VA, Feb.
1996.

William Dale Blair (S’83-M’85-SM’96) received the B.S. and M.S. degrees in
electrical engineering from Tennessee Technological University in 1985 and 1987. He
received the Ph.D. degree in electrical engineering from the University of Virginia in
January 1998.
While a graduate research assistant, Dr. Blair performed robitic controls research
for the Center of Excellence of Manufacturing Research and Technology Utilization at
Tennessee Tech. In 1987 he joined the Naval System Division of FMC Corporation in
Dahlgren, VA as an electrical engineer where his work involved the development and
evaluation of new algorithms for weapons control. In 1990, Dr. Blair joined the Naval
Surface Warfare Center, Dahlgren Division (NSWCDD) in Dahlgren, VA.
At NSWCDD he originated two benchmark problems for target tracking and radar
resource allocation, which served as themes for invited sessions at the 1994 and 1995
American Control Conferences. He also led a project that demonstrated through a
real-time tracking experiment that modern tracking algorithms can be utilized to reduce
the radar time and energy required by a phased array radar to support surveillance
tracking. Dr. Blair’s research interests include radar signal processing and control,
resource allocation for multifunction radars, multisensor resource allocation, tracking
maneuvering targets, and multisensor integration and data fusion.
Dr. Blair’s research has been reported in over eighty publications. He has received
technical awards that include an NSWCDD Technical Excellence Award in 1991 for his
contributions to the development of innovative techniques for target trajectory estimation
and prediction and the NSWCDD Independent Exploratory Development Excellence
Award in 1993. He served as a lecturer in 1993, 1994, and 1998 for the UCLA Extension
Program in three short courses related to advanced topics in target tracking. He regularly
serves as a reviewer for the IEEE Transactions on Automatic Control, American Control
Conference, and IEEE Conference on Decision and Control. He was also co-organizer for
invited sessions at the American Control Conferences in 1993, 1994, and 1995.
Mr. Blair is a member of the IEEE Control Systems, Information Theory, and
Education Societies; Tau Beta Pi, Eta Kappa Nu, and Phi Kappa Phi.

BLAIR ET AL.: BENCHMARK FOR RADAR ALLOCATION AND TRACKING IN ECM 1113
G. A. Watson Photograph and Biography not available.

Thiagalingam Kirubarajan (S’95) was born in Sri Lanka in 1969. He received his B.A.
degree in electrical and information engineering from Cambridge University, England, in
1991.
While in England he worked for the Central Electricity Research Laboratories,
Leatherhead, Surrey, as a research assistant. From 1991 to 1993 he was an assistant
lecturer in Electrical and Computer Engineering at the University of Peradeniya, Sri
Lanka. Since 1993 he has been a graduate studenthesearch assistant at the University
of Connecticut, Storrs, in pursuit of a Ph.D. His research interests are in estimation and
target tracking.

Yaakov Bar-Shalom (S’63-M’66-SM’SO-F’84) was born on May 11, 1941. He

received the B.S. and M.S. degrees from the Technion, Israel Institute of Technology, in
1963 and 1967 and the Ph.D. degree from Princeton University in 1970, all in electrical
engineering.
From 1970 to 1976 he was with Systems Control, Inc., Palo Alto, CA. Currently
he is Professor of Electrical and Systems Engineering and Director of the ESP Lab
(Estimation and Signal Processing) at the University of Connecticut. His research interests
are in estimation theory and stochastic adaptive control and has published over 200
papers in these areas. In view of the causdity principle between-the given name of a
person (in this case, “(he) will track”, in the modern version of the original language
of the Bible) and the profession of this person, his interests have focused on tracking.
His other interests are stochastic control of vertical airfoils and of pairs of inclined
foot supports on crystals. He coauthored the monograph Tracking and Data Association
(Academic Press, 1988), the graduate text Estimation and Tracking: Principles, Techniques
and Software (Artech House, 1993), the text Multitarget-Multisensor Tracking: Principles
and Techniques (YBS Publishing, 1995), and edited the books Multitarget-Multisensor
Tracking: Applications and Advances (Artech House, Vol. I, 1990; Vol. 11, 1992). He has
been elected Fellow of LEEE for “contributions to the theory of stochastic systems and
of multitarget tracking”. He has been consulting to numerous companies, and originated
the series of Multitarget-Multisensor Tracking short courses offered via UCLA Extension,
at Government Laboratories, private companies and overseas. He has also developed
the commercially available interactive software packages MULTIDATm for automatic
track formation and tracking of maneuvering or‘ splitting targets in clutt
for data association from multiple passive sensors, BEARDATTMfor t
from bearing and frequency measurements in clutter, 1MDATm for i
and target centroid tracking and FUSEDATm for fusion of possibly heterogeneous
multisensor data for tracking.
During 1976 and 1977 he served as Associate Editor of the IEEE Transactions on
Automatic Control and from 1978 to 1981 as Associate Editor of
Program Chairman of the 1982 American Control Conference, G
1985 ACC, and Co-Chairman of the 1989 IEEE International Conference on Control and
Applications. During 1983-1987 he served as Chairman of the Conference Activities
Board of the IEEE Control Systems Society and during 1987-1989 was a member of the
Board of Governors of the IEEE CSS. In 1987 he received the IEEE CSS Distinguished
Member Award. Since 1995 he is a Distinguished Lecturer of the IEEE AESS. He is
co-recipient of the M. Barry Carlton Award for the best paper in the IEEE Transactions
on Aerospace and Electronic Systems in 1995.

1114 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 34, NO. 4 OCTOBER 1998