sensors

Article
TMA from Cosines of Conical Angles Acquired by a
Towed Array
Antoine Lebon 1 , Annie-Claude Perez 2 , Claude Jauffret 2, * and Dann Laneuville 3

1 Naval Group, 83190 Ollioules, France; [email protected]

2 Université de Toulon, Aix Marseille Univ, CNRS, IM2NP, CS 60584, 83041 Toulon, CEDEX 9, France;
[email protected]
3 Naval Group—Technocampus Océan, 44340 Bouguenais, France; [email protected]
* Correspondence: [email protected]; Tel.: +33-494-142-414

Abstract: This paper deals with the estimation of the trajectory of a target in constant velocity motion
at an unknown constant depth, from measurements of conical angles supplied by a linear array.
Sound emitted by the target does not necessarily navigate along a direct path toward the antenna,
but can bounce off the sea bottom and/or off the surface. Observability is thoroughly analyzed to
identify the ghost targets before proposing an efficient way to estimate the trajectory of the target of
interest and of the ghost targets when they exist.

Keywords: target motion analysis; observability; fisher information matrix; Cramér–Rao lower
bound; conical angles; nonlinear estimation







Citation: Lebon, A.; Perez, A.-C.; 1. Introduction

Jauffret, C.; Laneuville, D. TMA from Bearings-only target motion analysis (BOTMA) is a problem that has been widely
Cosines of Conical Angles Acquired studied and various solutions have been proposed in the literature: batch [1–5] or recursive
by a Towed Array. Sensors 2021, 21, filter (such as extended Kalman filter [6–8], unscented Kalman filter [9], particle filter [10],
4797. https://doi.org/10.3390/ modified instrumental variable [11–13]), or a mix of recursive and batch methods [14].
s21144797 Citing all the papers dealing with this topic is now a hard task. Among the abundant
literature, most papers share the same assumption: the target is moving in a straight line
Academic Editors: Mahendra Mallick
with a constant speed, while the passive observer is maneuvering adequately in order to
and Ratnasingham Tharmarasa
ensure the observability of the target [15–17]. The bearings are the measurements.
In this paper, we are concerned with the same problem, except that the available
Received: 18 June 2021
measurements are the cosine of the relative bearings, also called conical angles because the
Accepted: 12 July 2021
target belongs to the cone of ambiguity whose revolution axis is the line along which the
Published: 14 July 2021
towed array is moving (see [18] p. 39). Implicitly, we consider a target moving in 3D at a
constant and unknown depth in near field; in this case, the two more energetic rays are the
Publisher’s Note: MDPI stays neutral
direct and the reflected paths (bottom or surface). In most cases, the sound bounces off the
with regard to jurisdictional claims in
published maps and institutional affil-
sea bottom. Therefore, we extend our analysis to surface- and sea bottom-bounced rays.
iations.
Indeed, the array detects the cosine of the relative angle of the direction of arrival by
means of a suitable spatial filtering method such as beamforming, or more sophisticated
techniques (see [19]). In the near field, sound can propagate to the sensor array along the
direct path and/or the bottom-reflected path, and/or the surface-reflected path. Most of
the time, at most, two rays coming from the same target are detected [18,20].
Copyright: © 2021 by the authors.
Unlike Gong [21] and Blanc-Benon [22], who addressed the three-dimensional target
Licensee MDPI, Basel, Switzerland.
motion analysis (TMA) from a sequence of time differences of arrival (TDOA) of a signal
This article is an open access article
distributed under the terms and
traveling by two different paths coupled with a sequence of azimuths, we assume in this
conditions of the Creative Commons
paper that the available measurements are the cosines of the conical angles only. In [23],
Attribution (CC BY) license (https:// a similar problem was addressed, but observability was not studied. We will consider two
creativecommons.org/licenses/by/ situations: the first case is devoted to TMA when sound propagates along a non-direct path
4.0/). at each sampling time. This will be the topic of Section 3: we will conduct observability

Sensors 2021, 21, 4797. https://doi.org/10.3390/s21144797 https://www.mdpi.com/journal/sensors

Sensors 2021, 21, 4797 2 of 20

analysis and identify all the ghost targets, given a set of noise-free measurements. We will
prove that an assumption of the target’s depth makes the target’s trajectory observable,
but not estimable (in the sense that the asymptotic performance given by the Cramér–Rao
lower bound—CRLB—of the estimator of the depth is out of the physical constraints, that
is, the source is navigating between the surface and the sea bottom).
In the fourth section, we will consider scenarios in which the antenna changes its own
route. We will prove that the trajectory of the target is almost certainly observable.
In the fifth section, we will assume that sound will propagate along the direct path and
the bottom-reflected path. The two rays will be assumed as being detected. Observability
analysis will reveal that only three ghost targets at most exist without maneuvering the
antenna. We will check that, in this case, the depth is not “estimable”. We will give a
palliative, allowing us to propose an estimator which is operationally acceptable, the price
being a small bias. Convincing simulations will be given at the end of this section, proving
that, even when the duration of the scenario is short, the estimated trajectory is very close
to the true one. A conclusion ends the paper.

2. Notation and Problem Formulation

We consider two underwater vehicles moving at their own constant depth. The
first mobile is a surface vessel or a submarine towing a horizontal sensor array, and the
second one is the target of interest. Given a Cartesian coordinate system, the acoustic

T
center of the array is located at time t at xO (t) yO (t) zO . At the same time, the

T
target is at x T (t) y T (t) z T . The respective horizontal positions of the target of

T
interest and of the center of the array at time t are denoted by PT (t) = x T (t) y T (t)

T
and PO (t) = xO (t) yO (t) . The sea bottom depth (assumed to be a constant) is de-
noted as D. The source is said to be endfire to the line array if its trajectory is in the
same line as the array (which implies that the array and the source are at the same
depth, and share the same route). It is broadside to the antenna if it navigates in the
vertical plane orthogonal to the line array and passing by the acoustic center of the ar-
ray. The sensor array detects the line of sight of the target; more precisely, ad hoc array
processing (or spatial filtering) delivers at time t the cosine of the conical angle c a (t)
given by cos(c a (t)) = cos(θ (t) − h(t)) cos(φ(t)) , m(t), where θ (t) and φ(t) are, re-
spectively, the azimuth (or bearing) and the elevation of the path along which the sound
emitted by the source propagates. The angle h(t) is the heading of the sensor array.
Denoting the relative position coordinates of the source with reference to the acoustic
center of the array by xOT (t) = x T (t) − xO (t) and yOT (t) = y T (t) − yO (t), we have
θ (t) = arctan( xOT (t), yOT (t)). Figure 1 displays the different angles and the two actors
(the observer reduced to the linear array, and the target).
The ray of the sound (or signal) emitted by the source can be reflected by the bottom
and/or the surface or travels in the surface or deep channel. The sound–speed profile
makes the paths curve. In this paper, we will consider that the target is in the near field
(the distance between the source and the array is less than 20 km), and the bottom depth
is in the range 2000–5000 m. Due to the large curvature of the ray (about 80 km), we
will approximate the path of the sound as a set of zigzags defined by the reflections on
the bottom or on the surface. So, we implicitly use the Snell law widely employed in
geometrical optics. An image-source is created whose depth ζ T will be called “image-
depth”. A path is then defined by the triplet (δ, n B , nS ), where
• δ indicates the direction of the path of the sound emitted by the source: if the path is
toward the surface, δ = −1, otherwise δ = +1,
• n B is the number of bottom reflections, and
• nS is the number of surface reflections.
Sensors 2021,21,
Sensors2021, 21,4797
x FOR PEER REVIEW 3 3ofof2020

Figure 1. A typical scenario, viewed from the sky.

The ray of the sound (or signal) emitted by the source can be reflected by the bottom
and/or the surface or travels in the surface or deep channel. The sound–speed profile
makes the paths curve. In this paper, we will consider that the target is in the near field
(the distance between the source and the array is less than 20 km), and the bottom depth
is in the range 2000–5000 m. Due to the large curvature of the ray (about 80 km), we will
approximate the path of the sound as a set of zigzags defined by the reflections on the
bottom or on the surface. So, we implicitly use the Snell law widely employed in geomet-
rical optics. An image-source is created whose depth 𝜁𝑇 will be called “image-depth”. A
path is then defined by the triplet (𝛿, 𝑛𝐵 , 𝑛𝑆 ), where
 𝛿 indicates the direction of the path of the sound emitted by the source: if the path is
toward the surface, 𝛿 = −1, otherwise 𝛿 = +1,

Figure𝑛1.
Figure 𝐵1.Ais the
Atypicalnumber
typical of viewed
scenario,
scenario, bottom
viewedfromreflections,
from sky.and
thesky.
the
 𝑛𝑆 is the number of surface reflections.
The ray
Figure
Figure of the sound
22illustrates
illustrates three(ordifferent
three signal)
differentemitted
paths by the source can be reflected by the bottom
paths
and/or the surface or travels in the surface or deep channel. The sound–speed profile
makes the paths curve. In this paper, we will consider that the target is in the near field
(the distance between the source and the array is less than 20 km), and the bottom depth
is in the range 2000–5000 m. Due to the large curvature of the ray (about 80 km), we will
approximate the path of the sound as a set of zigzags defined by the reflections on the
bottom or on the surface. So, we implicitly use the Snell law widely employed in geomet-
rical optics. An image-source is created whose depth 𝜁𝑇 will be called “image-depth”. A
path is then defined by the triplet (𝛿, 𝑛𝐵 , 𝑛𝑆 ), where

Figure
Figure 𝛿2.2. Three
indicates
Three examples
examples of paths:
the direction
of ray ray
of paths:
the
thepaththeof
solid solid
the
line line represents
sound
represents emitted
the directby the
path direct
the(δ, n Bpath
source: (the
, nS )if=𝛿,(+ , 𝑛0,𝑆 )0)=
𝑛𝐵1,
path is
,
(+1,0
the ,0), the the
toward
dashed-dotted dashed-dotted
surface,
line 𝛿 line
= −1,
represents represents
the otherwise
bottom the bottom
reflected +1,reflected
𝛿 = path (δ, n B , npath
S ) = ( 𝛿,
(+ 𝑛
1,
𝐵 ,
1, 𝑛0
𝑆 )
) , =
and(+ 1,1,0)
the ,
dashed and
the dashed
𝑛 is line
 represents represents
thethe
number ofthe bottom-surface-bottom
bottom reflections, pathreflected
reflectedand (δ, n B , npath (𝛿, 𝑛 , 𝑛 ) = (+1,2,1).
line 𝐵 bottom-surface-bottom S ) = (+1, 𝐵2, 1𝑆).
 𝑛𝑆 is the number of surface reflections.
WeWehavehavetotoconsider
considerthe thedepthdepthdifference
differencebetween
betweenthe thearray
arrayand andthe theimage-source
image-source
Figure 2 illustrates three different paths
definedby
defined byζ OT𝜁𝑂𝑇,≜ζ 𝜁T𝑇−−z𝑧O𝑂ififthe theray rayhas has
beenbeen reflected
reflected (by(by the the
sea sea
bottombottomor byorthe bysurface),
the sur-
face),
or ζ OT or, z𝜁T𝑂𝑇−≜zO𝑧𝑇if−the 𝑧𝑂 sound
if the sound
wave uses wavethe uses the path.
direct direct path.
AAgeneral
general expression
expression of ζ OTof 𝜁𝑂𝑇onbased
based on (δ,
the triplet then B , triplet (𝛿, 𝑛by
nS ) is given ) (is
𝐵 , 𝑛ζ𝑆OT δ, ngiven
B , nS ) = by
nS +n B 𝑛 +𝑛 n𝐵 +n B 𝑛 +𝑛
−𝜁𝑂𝑇
2δn(𝛿, 𝑛𝐵1, 𝑛) 𝑆 ) = −2𝛿𝑛
B (− D − z𝐵O(−1) + (−1) 𝐷 −z𝑧T𝑂. Note
𝑆 S
+ (−1)that, given𝑧the
𝑆 𝐵 path, the
𝑇 . Note link
that, between
given theζpath, n B ,link
OT ( δ,the nS )
and
between 𝜁𝑂𝑇 (𝛿,
z T is linear: ζ OT , 𝑛n𝑆 )B , and
𝑛𝐵(δ, nS ) =𝑧𝑇azisT + linear: 𝜁𝑂𝑇 (𝛿, 𝑛𝐵being
b, the constants , 𝑛𝑆 ) a=function
𝑎𝑧𝑇 + 𝑏, of the
thetriplet (δ, n B , being
constants nS ), D,a
and zO. Moreover,
function ζ OT (δ,(𝛿,
of the triplet n B ,𝑛n𝐵S, )𝑛𝑆is),null
𝐷, ifand
and 𝑧only if the antenna
𝑂 . Moreover, (𝛿,the
𝜁𝑂𝑇and 𝑛𝑆 ) isare
𝑛𝐵 ,target navigating
null at theif
if and only
same depth (z T = zO), and sound is traveling in the direct path. In this case, cos(φ(t)) = 1. For the
sake of simplicity of the notations, we will simply subsequently denote ζ OT instead of ζ OT (δ, n B , nS ).
For the above examples, we have ζ OT (1, 0, 0) = z T − zO (direct path), ζ OT (+1, 1, 0) =
Figure 2. Three examples of ray paths: the solid line represents the direct path (𝛿, 𝑛𝐵 , 𝑛𝑆 ) =
2D − (z T + zO ) (bottom-reflected path), and ζ OT (+1, 2, 1) = 4D − (z T + zO ) (bottom-
(+1,0,0), the dashed-dotted line represents the bottom reflected path (𝛿, 𝑛𝐵 , 𝑛𝑆 ) = (+1,1,0), and
surface-bottom reflected path). Note that ζ OT (δ, n B , nS ) can be negative (the image-
the dashed line represents the bottom-surface-bottom reflected path (𝛿, 𝑛𝐵 , 𝑛𝑆 ) = (+1,2,1).
source√is above the surface). Consequently, the cosine of the elevation is cos(φ(t)) =
2 2
√ 2 WexOT2have(t)+yOT (t)
to consider. the depth difference between the array and the image-source
2 ( δ,n ,n )
xOT (t)+yOT (t)+ζ OT B S
defined by 𝜁 ≜
Figure 3a𝑂𝑇displays 𝜁𝑇 − 𝑧the
𝑂 ifcone
the ray has been reflected
of ambiguity, (bythe
defined by theset
seaofbottom
sourcesorsharing
by the the
sur-
face),cos
same 𝜁𝑂𝑇
or(φ ≜. 𝑧In𝑇 −
(t)) 𝑧𝑂 if 3b,
Figure thewesound
plot wave uses
a direct raythe
anddirect path.
a bottom-bounced ray, which allows
A general expression of 𝜁 based
us to figure out the various angles with which we will work.
𝑂𝑇 on the triplet (𝛿, 𝑛𝐵 , 𝑛𝑆 ) is given by
𝜁𝑂𝑇 (𝛿, 𝑛𝐵 , 𝑛𝑆 ) = −2𝛿𝑛𝐵 (−1)𝑛𝑆+𝑛𝐵 𝐷 − 𝑧𝑂 + (−1)𝑛𝑆+𝑛𝐵 𝑧𝑇 . Note that, given the path, the link
between 𝜁𝑂𝑇 (𝛿, 𝑛𝐵 , 𝑛𝑆 ) and 𝑧𝑇 is linear: 𝜁𝑂𝑇 (𝛿, 𝑛𝐵 , 𝑛𝑆 ) = 𝑎𝑧𝑇 + 𝑏, the constants being a
function of the triplet (𝛿, 𝑛𝐵 , 𝑛𝑆 ), 𝐷, and 𝑧𝑂 . Moreover, 𝜁𝑂𝑇 (𝛿, 𝑛𝐵 , 𝑛𝑆 ) is null if and only if
 face-bottom reflected path). Note that 𝜁𝑂𝑇 (𝛿, 𝑛𝐵 , 𝑛𝑆 ) can be negative (the image-source is
above the surface). Consequently, the cosine of the elevation is cos(𝜙(𝑡)) =
2 (𝑡)+𝑦 2 (𝑡)
√𝑥𝑂𝑇 𝑂𝑇
.
2 (𝑡)+𝑦 2 (𝑡)+𝜁 2 (𝛿,𝑛 ,𝑛 )
√𝑥𝑂𝑇 𝑂𝑇 𝑂𝑇 𝐵 𝑆
Sensors 2021, 21, 4797 4 of 20
Figure 3a displays the cone of ambiguity, defined by the set of sources sharing the
same cos(𝜙(𝑡)). In Figure 3b, we plot a direct ray and a bottom-bounced ray, which al-
lows us to figure out the various angles with which we will work.

(a) (b)
Figure3.3. Cones
Figure Cones of
of ambiguity.
ambiguity. (a) (a) The
Thecones
conesthat
thatthe
thetarget
targetbelongs
belongsto,to,and the
and one
the that
one thethe
that image-target belongs
image-target to. to.
belongs (b)
𝐷
Example
(b) Exampleof of
conical angles
conical of of
angles thethe
target and
target ofof
and the image-target,
the image-target,for
fora abottom-reflected ray:𝜙φ D and
bottom-reflectedray: 𝜙B𝐵 are
and φ arethe
theelevations
elevations
ofofthe
thedirect
directpath
pathand
andofofthe
thebottom-reflected
bottom-reflectedpath,
path,respectively.
respectively.

Weassume
We assumethat thatthethesource
sourceisismovingmovingininconstant
constantvelocity
velocity(CV)
(CV)motion
motionduring
duringthe the
scenario. Our
scenario. Ourchallenge
challengeisistotoestimateestimate itsits
trajectory,
trajectory,i.e.,i.e.,
thethe
state vector
state defining
vector definingit, 𝑋it,≜
∗ ∗ ). 𝑇 , for ∗
X(𝑥𝑇,(𝑡 ) x T𝑦(𝑇t(𝑡
∗ ) ) y 𝑧(𝑇t∗ )𝑥̇ 𝑇 z 𝑦̇ 𝑇x . a

, for a𝑡chosen
chosen , from tnoisy measurements.
T ∗ , from
T T T yT noisy measurements.
We consider two
We consider two situations:situations:
1.1. Only
Only oneone rayray is
is detected
detected by by thethe array
array during
during thethescenario;
scenario; in in this
thiscase,
case,we
wehavehaveatat
each time t a measurement 𝑚(𝑡), given the path along
each time t a measurement m(t), given the path along which the wave propagates. which the wave propagates.
2.2. Two
Tworaysrays(traveling
(travelingon ontwo
twodifferent
differentpaths)
paths)arrive
arriveatatthe
thesensor’s
sensor’santenna.
antenna.InInthis
thiscase,
case,
the available measurement at time t is a couple of measurements, say (m1 (t), m2 ((𝑡)),
the available measurement at time t is a couple of measurements, say (𝑚 1 (𝑡), 𝑚 2 t)),
given the
given the two
two paths
paths alongalong which
which thethe wave
wave propagates.
propagates.
After the
After the spatial
spatial filtering,
filtering, thethe antenna
antenna supplies
supplies aa noisy
noisy measurement
measurement of 𝑚(𝑡)
of m oraa
(t) or
noisy measurement of (𝑚 (𝑡), 𝑚 (𝑡)).
noisy measurement of (m1 (1t), m2 (2t)). The noisy measurements are regularly acquired atat
The noisy measurements are regularly acquired
t𝑡k𝑘==((𝑘 {1,. .…. , 𝑁},
)∆t, k 𝑘∈∈{1,
k −−11)Δ𝑡, N },for
foraafixed
fixed sampling time ∆t.
sampling time Δ𝑡.
Beforeattempting
Before attemptingto estimateX,𝑋,we
toestimate wemust
mustanswer
answerseveral
severalquestions:
questions:
1.1. Is the
Is vectorX 𝑋observable
thevector observable fromfrom
the the setmeasurements
set of {m(t){𝑚(𝑡),
of measurements , t ∈ [0,𝑡 T∈]}[0, 𝑇]}? that,
? Note Note
that,
in TMA in problems,
TMA problems, observability
observability is often analyzed
is often analyzed in continuous
in continuous time (see
time (see [15,17], for
[15,17], foreven
example), example),
thougheven
the though the noisy measurements
noisy measurements are given inare given in
discrete discrete time.
time.
2.2. If not,
If not, what
what are
are the
the ghost
ghost targets
targets (those
(those which
which could
could be
bedetected
detectedat atthe
thesame
sameset
setof
of
measurements {{𝑚(𝑡),
measurements m(t), t𝑡∈∈ [[0,
0, T𝑇]})?
]})?
3.3. How
How do do we make X𝑋 observable
we make observableororwith withwhich
whichnew
newinformation?
information?
4. Is the vector X observable from the set of couples {(m1 (t), m2 (t)), t ∈ [0, T ]}?
For the cases where X is observable, we have then to compute the asymptotical
performance of an unbiased estimator (given by the CRLB [24]), and the performance of
our estimators in terms of bias and the covariance matrix. It is worth noting that using
the FIM to prove observability can lead to a wrong conclusion [25]. This why we use an
analytic approach.

3. TMA from One Ray

In this section, we consider the case where the array collects the cosine of a coni-
cal angle, the path of the ray being known by the operator. We start by analyzing the
observability of the trajectory of the source of interest.
Sensors 2021, 21, 4797 5 of 20

3.1. Observability Analysis

Theorem 1. Let a linear antenna measure the cosine of a conical angle in the direction of a source,
both in CV motion. The path of the sound emitted by the source is known, as is also the sea
bottom depth.
1. If the target is broadside to the antenna, then the set of ghost targets is composed of virtual
sources broadside to the antenna.
2. If the target is endfire to the antenna, the set of ghost targets is composed of virtual sources
endfire to the antenna.
3. If the target has the same heading as the array (but is not endfire to it), then the set of ghost
targets is composed of virtual targets with the same heading as the antenna. More precisely,
the ghost image of each ghost target is moving on a cylinder whose axis is the antenna axis,
and whose radius is a positive scalar β. The relative ghost target velocity is equal to β times
the target’s velocity. The initial distance between the ghost image and the center of the antenna
is equal to β times the initial distance between the target-image and the center of the antenna.
4. In any other case, for a chosen image-depth ζ G , the set of ghost targets is composed of virtual
targets whose motion relative to the array is defined by POG (t) = βPOT (t) or POG (t) =
βSPOT (t), where S is the 2D axial symmetry around the line of the array, and β is a positive
|ζ |
scalar. The scalar β is equal to |ζOG | if ζ OT 6= 0. If ζ OT = 0 (which can happen with a direct
OT
path only), β can have any positive value.

Preamble: In the following proof, we choose t∗ = 0. Instead of working with the state
. .
T
vector X = x T (0) y T (0) z T x T y T , we will use the relative state vector of the
. .
T
image source, which is Y , x0T (0) yOT (0) ζ OT xOT yOT . The reason for this is
that we are able to recover X from Y without ambiguity.
We will prove this theorem in the special case where the heading of the antenna is
equal to 0◦ , and the value yOT (t) is positive. This can be easily obtained with an ad hoc
rotation of the whole scenario. This will simplify the expression of the measurement,
without loss of generality.

Proof of Theorem 1. We are seeking the ghost target whose horizontal position at time t is

T
xG (t) yG (t) , detected in the same cosine of the conical angle, that is
yOT (t)
√ 2 2 2
= √ 2 yOG2(t) 2
, with xOG (t) = xG (t) − xO (t), yOG (t) =
xOT (t)+yOT (t)+ζ OT xOG (t)+yOG (t)+ζ OG
yG (t) − yO (t), and ζ OG is the image-depth of the ghost target. This equality is equivalent to
2 (t)
yOT 2 (t)
yOG
2 2 (t) + ζ 2
= 2 2 (t) + ζ 2
(1)
xOT (t) + yOT OT xOG (t) + yOG OG

Note that because the target is moving (as is the ghost target also), the denominators of the
left term and of the right term of (1) are two polynomial functions of degree 2.
Case 1: yOT (t) is a zero function, i.e., ∀t yOT (t) = 0.
.
T
This means the source is broadside to the antenna: YT = x0T (0) 0 ζ OT xOT 0 .
In this case, yOT (t) = 0, ∀t ∈ [0, T ]. Hence, the set of ghost targets is composed of the
.
T
virtual targets broadside to the antenna: YG = x0G (0) 0 ζ OG xOG 0 .
Case 2: yOT (t) is not a zero function.
.
If yOT = 0, then yOT (t) is a constant. To respect the degrees of the terms of (1), yOG (t)
is a constant too.
.

If yOT 6= 0, then there is a root, say et, such

as yOT et = 0, since yOT (t) is a polynomial
function of degree 1. Consequently, yOG et = 0, and ∀t 6= et, yOG (t) 6= 0.
Sensors 2021, 21, 4797 6 of 20

. .
We deduce that, in both cases (yOT = 0, and yOT 6= 0), there exists a positive value β
such that yOG (t) = βyOT (t).
 2 2 (t) + ζ 2
 2 x 2 ( t ) + y2 ( t ) + ζ 2
  2
(1) ⇔ xOG (t) + yOG OG − β OT OT OT yOT ( t ) = 0
2 ( t ) + ζ 2 = β2 x 2 ( t ) + ζ 2
  (2)
⇔ xOG OG OT OT

2 ( t ) + ζ 2 can be equal to zero at any time, or at one time or never.

The quantity xOT OT
Subcase 1: ∀t, xOT2 ( t ) + ζ 2 = 0.
OT
Then, xOT (t) = 0, ∀t and ζ OT = 0. Note that this case is the one when the target is
traveling in the endfire to the array and at the same depth as the antenna and the path is
the direct one. For the same reason, xOG (t) = 0, ∀t and ζ OG = 0. The set of ghost targets
is hence composed of virtual targets traveling in the endfire to the array and at the same
depth as the antenna.
2 2 6 = 0.


Subcase 2: ∃t̆ such that xOT t̆ + ζ OT
We deduce from (2) that
2
xOG (0) = β2 xOT
2
(0) + β2 ζ OT
2 2
− ζ OG (3)
. .
xOG (0) xOG = β2 xOT (0) xOT (4)
.2 .2
xOG = β2 xOT (5)
.
If xOT = 0, then
 q .
T
YG = 2 (0) + β2 ζ 2 − ζ 2
± β2 xOT βyOT (0) ζ OG 0 βyOT , for any posi-
OT OG
q
tive constant β and any positive constant ζ OG 2 (0) + β2 ζ 2 . Note that,
less than β2 xOT OT
.
when yOT = 0, the target is motionless relative to the center of the array (both have the
.
same velocity); and when yOT 6= 0, the target has the same heading as the array.
.
If xOT 6= 0, then squaring the elements of (4), and using (5), we draw from (3) that
β2 ζ OT
2 2 . If ζ
= ζ OG OT = 0, then ζ OG = 0, and the scalar β can take any positive value;
|ζ OG |
else β = .
In both cases, the trajectory of a ghost target is defined by the state vector
|ζ OT |
. .
T
YG = ± βxOT (0) βyOT (0) βζ OT ± β xOT βyOT . 

Remark 1.
1. When the source and the observer are at the same depth, and the path is direct, Theorem 1
recovers the conclusions given in [26].
2. The cases (1), (2) and (3) of Theorem 1 are “rare events”, since the events of dealing with a
source in endfire, broadside or having the same heading as the antenna during the scenario
occur with a probability equal to 0. However, when the target has a trajectory close to one of
these special cases, the estimates will have a poor behavior.
3. For case (4), when the detected ray is not a direct path, for example, when the ray is bottom-
reflected, a hypothesis about the source is sufficient to obtain one solution, corresponding to a
ghost target. Indeed, if we suppose that the depth of the target is z As (whereas the true value
2D −(z +z ) 2D −zO
is z T ), then we have β = 2D−(zAs+z O) , whose biggest value β Max = 2D−( z +z )
, and the
T O T O
2D −(z +z )
minimum value is β Min = 2D−(Max O
z T + zO )
, where z Max is the largest depth of a submarine
vehicle. Typically, in deep water, D ≥ 4000 m. hA reasonable choice of z Max couldi be 400 m.
− zO
We can then have a range of β: [ β Min , β Max ] = 80007600 , 8000−zO
−(z T +zO ) 8000−(z T +zO )
. For instance,
when the depths of the antenna and the target are, respectively, 200 and 100 m, we have
[ β Min , β Max ] = [0.974, 1.013]. In this way, we bound the set of ghost targets, and we can
expect that the bias induced by a wrong choice of z As is very low.
4. For case (4) again, with a direct path, if the target is not at the same depth as the antenna,
− zO
β = zzAs T − zO
. Because β is a positive number, z As − zO has the same sign as z T − zO : if
h i
− zO
z T > zO , then zO < z As ≤ z Max , and [ β Min , β Max ] = 0, z Max z T −z ; if z T < zO , then
O
Sensors 2021, 21, 4797 7 of 20

h i
0 ≤ z As < zO , and [ β Min , β Max ] = 0, zTz−Oz . In both cases, the range [ β Min , β Max ] is too
O
wide to be useful. If the target and the antenna are at the same depth, β can take any positive
value.

3.2. Estimation of the Trajectory

We run 500 Monte Carlo simulations for a typical scenario described as follows:

T
The observer starts from 0 0 at the depth zO = 200 m. Its speed and heading
◦

T
are, respectively, 5 m/s and 0 . The initial position of the target is 5000 7000 and its
depth is z T = 100 m. Its route is 45◦ and its speed is 4 m/s.
• The measurements are collected every 4 s (∆t = 4 s). The scenario lasts 20 min.
• The sea bottom depth is 4000 m. The detected ray is a bottom-reflected ray.
• The assumed target depth is z As = 200 m (whereas the true one is 100 m).
• First, the measurements have been corrupted with an additive Gaussian noise whose
standard deviation is σ = 1.7 × 10−2 .
Then, we choose the least squares estimator, which is identical to the maximum
likelihood estimator with these assumptions. Note that, in open literature about TMA,
the confidence regions are given by the confidence ellipsoid obtained with the covariance
matrix of the estimate. Since the maximum likelihood estimate is asymptotically efficient
under nonrestrictive conditions, we use here the Cramér–Rao lower bound to compute
such confidence regions.
The result of the simulation is presented in Table 1 and illustrated in Figure 4. Obvi-
ously, even if the assumption made on the target’s depth makes the state vector observable,
it remains inestimable: the hugeness of the diagonal elements of the CRLB does not allow
this kind of TMA to be employed. We note in Figure 4 that the cloud of horizontal estimates
is hyperbola-shaped. This is because the state vector is “weakly” estimable. The parametric
equation of this hyperbola is
(
x (ω ) = ζ As sinh(ω ) yOT (0)
, with ζ As = 2D − (z As + zO ), and m = √ 2
y(ω ) = √ζ As m 2 cosh(ω ) 2 (0)+ ζ 2
xOT (0)+yOT OT
1− m

Table 1. Performance of the estimator of the reduced state vector when σ = 1.7 × 10−2 , in terms of
bias, sample standard deviation and the one given by the square root of the diagonal of the CRLB.

Xr Bias σsamp σCRLB

5000 m −3525 6962 13,356
7000 m −2367 4052 5599
2.83 m/s −1.37 1.81 4.35
2.83 m/s 0.53 1.62 2.75
 Sensors 2021, 21, x4797
FOR PEER REVIEW 8 of 20

Sensors 2021, 21, x FOR PEER REVIEW 8 of 20

Figure
Figure 4.
4. The cloud of
The cloud ofestimated
estimatedposition
position(in(in green),
green), a piece
a piece of the
of the hyperbola
hyperbola (intersection
(intersection of theofcone
the
(=m −2
cone of ambiguity
of ambiguity andplane
and the the plane 𝑧=
z = z As (=𝑧𝐴𝑠
200 200 m),σ for
), for 𝜎 =×1.7
= 1.7 10−×2 . 10 .
Figure 4. The cloud of estimated position (in green), a piece of the hyperbola (intersection of the
cone of We
Table ambiguity
furtherand
1. Performance theofplane
reduced the
the = 𝑧𝐴𝑠 (= of
𝑧standard
estimator 200 them),reduced
deviation for 𝜎to=σ1.7=×
state 10×−210
vector
1.7 . − 4 in order
when 𝜎 = 1.7 × 10−2 , in
to appreciate
terms of bias, of
the behavior sample standard
the MLE. Withdeviation and the value,
this (unrealistic) one given by theissquare
the MLE rootasofshown
efficient, the di-in
Table
agonal
Table Performance
1. 2 of
andtheinCRLB.of 5the
Figure estimator
(which of theour
validates reduced state vector
observability when 𝜎 = 1.7 × 10−2 , in
analysis).
terms of bias, sample standard deviation and the one given by the square root of the di-
the 𝑿 Bias 𝝈𝒔𝒂𝒎𝒑 𝝈 −4
agonal
Table of CRLB.
𝒓
2. Performance of the estimator of the reduced state vector with σ = 1.7 × 𝑪𝑹𝑳𝑩
10 .
5000 m −3525 6962 13,356
𝑿X𝒓r m
7000 Bias
Bias
−2367 𝝈𝒔𝒂𝒎𝒑
4052
σsamp 𝝈𝑪𝑹𝑳𝑩
5599σCRLB
5000
2.83
5000 mm
m/s −3525
−1.37
60.40 6962
1.81
138.67 13,356
4.35133.56
7000
7000 m
m
2.83 m/s −2367
88.20
0.53 405258.42
1.62 5599
2.7555.99
2.83
2.83 m/s −1.37
0.043 1.81
0.044 4.35 0.044
2.83 m/s
2.83 m/s 0.037
0.53 0.028
1.62 2.75 0.028

Figure 5. The cloud of estimated position (in green) for 𝜎 = 1.7 × 10−4 . The cloud is no longer
hyperbola-shaped. The small black segment is the 90%-confidence ellipsoid.
−4−4
Figure 5. 5.
Figure The cloud
The of of
cloud estimated
estimatedposition (in(in
position green) 𝜎σ
forfor
green) ==1.71.7
×× 1010 . The cloud
. The is is
cloud nono
longer
longer
hyperbola-shaped.
Table
hyperbola-shaped. The
2. PerformanceThesmall
of black
the
small segment
estimator
black ofis the
segment the 90%-confidence
reduced
is the state vector
90%-confidence with 𝜎 = 1.7 × 10−4 .
ellipsoid.
ellipsoid.

Table 2.Our 𝑿𝒓
Performance ofisthe
conclusion that Bias
estimator ofvector
the state the reduced 𝝈𝒔𝒂𝒎𝒑
state vector
is not estimable, with
even 𝜎 =𝝈1.7
though𝑪𝑹𝑳𝑩
it × 10−4 .
is observable
5000 m
with an assumption on the target’s60.40
depth. 138.67 133.56
𝑿 𝒓 Bias 𝝈 𝒔𝒂𝒎𝒑 𝝈𝑪𝑹𝑳𝑩
This7000 m we propose to maneuver
is why 88.20 the antenna 58.42
in order to render55.99
the state vector
5000
observable m
2.83with 60.40
m/s no assumption 0.043 on the target’s depth,138.67 133.56
0.044and to augment0.044 the information
about it.7000
2.83m m/s 88.20
0.037 58.42
0.028 55.99
0.028
2.83 m/s 0.043 0.044 0.044
4. TMA
4. TMA with
2.83
withm/sOne Ray
One Ray When
When the the Array Maneuvers
0.037
Array Maneuvers 0.028 0.028
In this
In this section,
section, the
the antenna
antenna maneuvers,
maneuvers, i.e.,
i.e., itit changes
changes its
its own
own heading.
heading. WeWe start
start by
by
4. proving
TMA with thatOne Ray When
the state vector the Array Maneuvers
is observable (without any assumption on the target’s depth).
proving that the state vector is observable (without any assumption on the target’s depth).
Then, we have
In this recourse
section, to perform
the antenna Monte Carlo
maneuvers, simulations
i.e., itsimulations
changes its to to
own evaluate the performance
heading.
Then, we have recourse to perform Monte Carlo evaluate theWe start by
performance
of the
proving MLE.
that the state vector is observable (without any assumption on the target’s depth).
of the MLE.
Then, we have recourse to perform Monte Carlo simulations to evaluate the performance
of the MLE.
Sensors 2021, 21, x FOR PEER REVIEW 9 of 20

Sensors 2021, 21, 4797 9 of 20

4.1. Observability Analysis

Theorem 2. Suppose
4.1. Observability the antenna’s trajectory is composed of two successive legs at constant ve-
Analysis
locity (however with the same speed). Let the target be in CV motion. The linear array acquires the
Theorem
conical 2. Suppose
angles the antenna’s
of the wave trajectory
emitted from is composed
the target, of two
the path successive
of the ray beinglegs at constant
known as wellvelocity
as the
(however
sea bottom depth. If the target is broadside or endfire to the antenna during a leg, then the
with the same speed). Let the target be in CV motion. The linear array acquires thereconical
is at
anglesa of
most the wave
ghost target.emitted from the
Otherwise, target,
there is nothe pathtarget.
ghost of the ray being known as well as the sea bottom
depth. If the target is broadside or endfire to the antenna during a leg, then there is at most a ghost
target. Otherwise,
Due there is
to its length, thenoproof
ghost oftarget.
this theorem is given in the Appendix A.

Due to its length, the proof of this theorem is given in the Appendix A.
4.2. Estimation
In this subsection, we present the result of 500 Monte Carlo simulations that are run
4.2. Estimation
to illustrate the behavior of the proposed estimators. First, we give the scenario used here.
In this subsection, we present the result of 500 Monte Carlo simulations that are run
The center of the array and the initial position of the source are, respectively, at
to illustrate the behavior of the proposed estimators. First, we give the scenario used here.
(0 0 200)𝑇 and (5000 7000 100)𝑇 at the very beginning of the scenario. The speed
The center of the array and the initial position of the source are, respectively, at
of the array is
Ta constant along the scenario
T and is equal to 5 m/s. The trajectory of the
array 0is composed
0 200 andof two
5000legs7000
linked by anatarc
100 theof
very beginning
a circle. The of theleg
first scenario.
lasts 1 The
minspeed
40 s,
of the array is a constant along the scenario and is equal to 5 m/s.
during which the array’s heading is 135°. Then, the array turns to the right with a turn The trajectory of the
array is composed of two legs linked
rate equal to 20°/min to adopt a new heading by an arc of a circle. The first leg lasts 1
equal to 270°. The duration of the maneuvermin 40 s,
during which the array’s heading is 135 ◦ . Then, the array turns to the right with a turn rate
is hence equal to 6 min 44 s. The second leg lasts 5 min, so the total duration of the scenario
◦ /min to adopt a new heading equal to 270◦ . The duration of the maneuver is
equal
is to 20
13 min and 20 s. Meanwhile, the target is navigating with a heading equal to 45° and a
hence equal to 6 min 44 s. The second leg lasts 5 min, so the total duration of the scenario
speed of 4 m/s. The bottom depth is 𝐷 = 4000 m.
is 13 min and 20 s. Meanwhile, the target is navigating( with a heading equal to 45◦ and)𝑇a
The state vector we have to estimate is hence 𝑋 = 5000 7000 100 2.83 2.83 .
speed of 4 m/s. The bottom depth is D = 4000 m.
The array is assumed to measure the cosines of the conical angles of the bottom-re-
T
The
flected state
path vector
given bywe have to estimate is hence X = 5000 7000 100 2.83 2.83 .
The array is assumed to measure the cosines of the conical angles of the bottom-
reflected path given𝑚(𝑡 by ) = 𝑦𝑂𝑇 (𝑡𝑘 )
𝑘 + 𝜀𝑘
2 (𝑡 ) 2 2
√𝑥𝑂𝑇 𝑘 + 𝑦𝑂𝑇 (𝑡𝑘 ) + [2𝐷 − (𝑧𝑇 + 𝑧𝑂 )]
yOT (tk )
Measurements tk ) =
m(are acquired
q every ∆𝑡 = 4s, with 𝑡𝑘 = (𝑘 − 1)∆𝑡. + εk
2 ( t ) + y2 ( t ) + [2D − ( z + z )]2
xOT
The noise vector 𝜀 is assumed
𝑘
k to OTbe Gaussian,
k T
0-mean O
and its standard deviation
equal to 𝜎 = 1.7 10−2 . The vectors 𝜀𝑘 are also assumed to be temporally independent.
Measurements
Again, we choose arethe
acquired every ∆t
least squares = 4 s, with tk = (k − 1)∆t.
estimator.
The noise vector ε k is assumed to be Gaussian, 0-mean and its standard deviation
to σ = 1.7 ×
equalEstimation −2
4.2.1. of10𝑋 . The vectors ε k are also assumed to be temporally independent.
Again, we choose the least squares estimator.
The 500 obtained estimates of the initial horizontal position are plotted in Figure 6,
together with theof
4.2.1. Estimation trajectory
X of the target, the 90%-confidence ellipse and the trajectory of
the array. Again, the view is
The 500 obtained estimates fromofthe
thesky.
initial horizontal position are plotted in Figure 6,
The with
together performance of theofestimator
the trajectory (bias
the target, the and standard deviation
90%-confidence of each
ellipse and component)
the trajectory of
is presented in Table 3.
the array. Again, the view is from the sky.

Figure 6. The
The cloud
cloud of the 500 initial positions estimates and the 90%−confidence ellipse.
90%-confidence ellipse.

The performance of the estimator (bias and standard deviation of each component) is
presented in Table 3.
Sensors 2021, 21, x FOR PEER REVIEW 10 of 20

Sensors 2021, 21, 4797 10 of 20

Table 3. Performance of the estimator of the plain state vector

𝑿 Bias 𝝈𝒔𝒂𝒎𝒑 𝝈𝑪𝑹𝑳𝑩

5000 m of the estimator
Table 3. Performance −44.77 854.72
of the plain state vector. 868.12
7000 m −68.16 1162.1 1173.60
X Bias σsamp σ
100 m 7.14 558.55 545.99CRLB
5000
2.83mm/s −44.77
0.092 854.72
1.67 1.65868.12
7000 m −68.16 1162.1 1173.60
2.83mm/s
100 0.194
7.14 2.72
558.55 2.68545.99
2.83 m/s 0.092 1.67 1.65
A convenient
2.83 m/s way to evaluate 0.194the behavior of an estimator
2.72 is to compute2.68the so-called
𝑇
normalized estimation error squared (NEES) [27], defined as 𝑁𝑙 = (𝑋̂𝑙 − 𝑋) 𝐹(𝑋̂𝑙 − 𝑋),
where 𝐹 is the FIM,
A convenient wayand 𝑋̂𝑙 is the
to evaluate theestimate
behaviorcomputed at theis 𝑙-th
of an estimator simulation.
to compute If 𝑋̂𝑙 is
the so-called
Gaussian-distributed with 𝑋 squared
as the mathematical
(NEES) [27],expectation and=theX̂CRLB
as
T the covar-

normalized estimation error defined as N l l − X 2 F X̂l − X ,
iance matrix,
where FIM, 𝑁
F is thethen 𝑙 isX̂chi-square
and distributed with d degrees of freedom (𝜒𝑑 ), where d is
l is the estimate computed at the l-th simulation. If X̂l is Gaussian-
the dimension
distributed of 𝑋 (here 5). From
with X as the mathematical theexpectation
central limit
andtheorem,
the CRLBthe as averaged NEES
the covariance 𝑁𝑆 ≜
matrix,
1 𝑁𝑆𝑖𝑚
then∑N𝑙=1
𝑁𝑆𝑖𝑚
𝑁𝑙 is approximately
l is chi-square distributedGaussian; its mathematical
with d degrees of freedom expectation
(χ2 ), where disisd,the
d
anddimension
its stand-
1 N
NSim ∑l =1 Nl is
Sim
ardXdeviation
of (here 5). is
From 2𝑑 limit theorem, the averaged NEES NS ,
thetocentral
equal √𝑁 .
approximately mathematical expectation is d, and its standard deviation is
Gaussian; its𝑆𝑖𝑚
From our2d simulations, we obtain 𝑁𝑆 = 5.34.
q
equal to N .
In conclusion,
Sim the estimator can be declared efficient. However, the minimum stand-
From our simulations, we obtain NS = 5.34.
ard deviation of the target’s depth is not compatible with the physical constraints: with
In conclusion, the estimator can be declared efficient. However, the minimum standard
the standard deviation given in Table 3, the target could be up above the sea surface!
deviation of the target’s depth is not compatible with the physical constraints: with the
Therefore, a palliative of this is to impose a depth on the target. Indeed, we saw in Section
standard deviation given in Table 3, the target could be up above the sea surface! Therefore,
3.1 that a supposed depth creates a small bias in estimation of the horizontal position of
a palliative of this is to impose a depth on the target. Indeed, we saw in Section 3.1 that a
the target.
supposed depth creates a small bias in estimation of the horizontal position of the target.
4.2.2. Estimation
4.2.2. Estimation of 𝑋 Reduced
of X ReducedWhen
Whenthe theDepth
Depthof ofthe
theTarget
TargetIsIsFixed
Fixed
Now, the
Now, the third
third component
component of of 𝑋X does
does not not have
havetotobe
beestimated.
estimated.TheThenew
newstate vector
state vec-
(𝑥 (0) 𝑦 (0) 𝑥̇ 𝑦̇ )𝑇 .
is the denoted as 𝑋𝑟 = 𝑇 𝑇 𝑇 𝑇 . We impose that 𝑧𝐴𝑠 = 200 m (whereas
.
T
tor is the denoted as Xr = x T (0) y T (0) x T y T . We impose that z As = 200 m
the true depth is still 100 m). Hence, we introduce a bias.
(whereas the true depth is still 100 m). Hence, we introduce a bias.
Figure 77 displays
Figure displays the
the position’s
position’s estimates
estimates in in the
the same
same manner
manner asas Figure
Figure 6.
6. The
The bias
bias is
is
not visible to the naked eye. However, Table 4 reveals this bias, which may
not visible to the naked eye. However, Table 4 reveals this bias, which may be acceptable be acceptable
in aa real
in real situation.
situation. Even
Even though
though the
the averaged
averaged NEES NEES (=7.31)
(=7.31) isis out
out of
of its
its 90%
90% confidence
confidence
interval, its
interval, its value
value remains
remains acceptable.
acceptable.

7. The
Figure 7. The cloud
cloud of the 500 initial positions estimates with the reduced state vector
vector and
and the
the
90%−confidenceellipse.
90%-confidence ellipse.
 Sensors 2021, 21, x FOR PEER REVIEW 11 of 20
Sensors 2021, 21, 4797 11 of 20

Table 4. Performance of the estimator of the reduced state vector.

Table 4. Performance of the estimator of the reduced state vector.
𝑿𝒓 Bias 𝝈𝒔𝒂𝒎𝒑 𝝈𝑪𝑹𝑳𝑩
Xr m
5000 Bias
65.40 σsamp
655.61 σCRLB
606.41
5000 mm
7000 65.40
93.05 655.61
831.75 606.41
762.56
7000
2.83mm/s 93.05
0.034 831.75
1.71 762.56
1.58
2.83 m/s 0.034 1.71 1.58
2.83 m/s
2.83 m/s
0.063
0.063
2.76
2.76
2.522.52

The main interest of assuming the depth to be known is to economize on the CPU
The main interest of assuming the depth to be known is to economize on the CPU
time, and reduce the standard deviation of the remaining components to estimate. We are
time, and reduce the standard deviation of the remaining components to estimate. We are
in the presence of the well-known bias–variance tradeoff.
in the presence of the well-known bias–variance tradeoff.
4.2.3. Estimation
4.2.3. Estimation of
of the
the Reduced
Reduced State
State Vector
Vectorbybythe
theConventional
ConventionalBOTMA
BOTMA
Insuch
In suchaascenario,
scenario,the
theconventional
conventionalBOTMA
BOTMA can
can bebe run
run byby neglecting
neglecting thethe
sitesite effect,
effect, so
so by imposing that cos(𝜙(𝑡)) = 1, ∀𝑡.
by imposing that cos(φ(t)) = 1, ∀t. The (incorrect) noise-free measurement model is thenis
The (incorrect) noise-free measurement model
then
cos(αcos(𝛼(𝑡)) (t) − h(t−
(t)) = cos=(θcos(𝜃(𝑡) .
))ℎ(𝑡)).
The results are plotted in Figure 8. Obviously, a huge bias appears, leading to an
The results are plotted in Figure 8. Obviously, a huge bias appears, leading to an
averaged NEES equal to 1960. More precisely, the bias on the components of the reduced
averaged NEES equal to 1960. More precisely, the 𝑇bias on the components of the reduced
state vector is (−3062.8 −2319.9 14.4 15.8)
T, rendering the BOTMA inoperative.
state vector
Clearly, is −3062.8 BOTMA
the conventional −2319.9cannot14.4be15.8 , rendering
recommended thenear
for the BOTMA inoperative.
field. This justifies
Clearly, the conventional BOTMA cannot be recommended for the near field. This justifies
a posteriori the interest in taking the site effect and the nature of the wave ray into account,
aasposteriori
previouslythepointed
interest out
in taking
in the the site effect and
introduction the nature of the wave ray into account,
of [23].
as previously pointed out in the introduction of [23].

Figure 8.8. The

Figure The cloud
cloud (in
(in green)
green) of
of the
the 500
500 initial
initial positions
positions estimates
estimates given
given by
by the
the classic
classic BOTMA
BOTMA
together with
together withthe
the90%-confidence
90%−confidenceellipse.
ellipse.

5.
5. TMA
TMA from
from the
the Direct
Direct Path
Path and
and the
the Bottom-Reflected
Bottom-ReflectedPath
Path
We
We assume in this section that the sound wave emitted by
assume in this section that the sound wave emitted by the
the target
targettravels
travels on
on the
the
direct path and the bottom-reflected path.
direct path and the bottom-reflected path.
5.1. Observability
5.1. Observability

Theorem
Theorem3. 3. Let
Let aa linear
linear antenna
antenna and
and aa source
sourceboth
bothbe
bein
inCV
CV motion.
motion.
The antenna acquires the cosines of the conical angles of thepath
The antenna acquires the cosines of the conical angles of the direct direct
andpath andbottom-reflected
of the of the bottom-
reflected
path. path.
1.
1. If the
If the target
targetisisbroadside
broadsidetotothethe
array, then
array, thethe
then set set
of ghost targets
of ghost is uncountable:
targets it is composed
is uncountable: it is com-
of all the
posed (virtual)
of all targetstargets
the (virtual) at broadside to the array.
at broadside to the array.
2. If the target is endfire to the antenna, the set of ghost targets is composed of virtual sources at
2. If the target is endfire to the antenna, the set of ghost targets is composed of virtual sources at
endfire to the antenna.
endfire to the antenna.
3. If the route of the antenna and the route of the target are parallel, then the set of ghost targets
is uncountable: at each depth zG , there are two ghost targets moving on a cylinder whose axis
Sensors 2021, 21, 4797 12 of 20

q
−zG
is the antenna axis, and the radius is a positive scalar β = D D −z T . The relative ghost target
velocity is equal to β times the target’s velocity. The initial distance between the ghost image
and the center of the antenna is equal to β times the initial distance between the ghost image
and the center of the antenna.
4. If the route of the antenna and the route of the target are not parallel, then there are three ghost
targets whose motion relative to the antenna is POG (t) = SPOT (t), POG (t) = βPOT (t), and
POG (t) = βSPOT (t), where S is the matrix of the axial symmetry around the line of the
− zO
antenna, and β , D D −z T . If the depth of the antenna is equal to the depth of the source, then
there is one single ghost target given by POG (t) = SPOT (t).

Proof of Theorem 3. With no loss of generality, we will again assume that the axis of
the sensor array is pointed toward north and that the target is in the half-space where
the second component y of any vector is positive. A convenient rotation helps us in
yOT (t)
this case. So the noise-free measurements at time t are m1 (t) = √ 2 2 2
, and
xOT (t)+yOT (t)+zOT
yOT (t)
m2 ( t ) = q .
2 ( t )+ y2 ( t )+[2D −( z + z )]2
xOT OT T O
. .
T
We have to seek a five-dimensional state vector XG = xG (0) yG (0) zG x G yG
defining the trajectory of a ghost target, i.e., producing the same noise-free measurement as X,
yOG (t) yOG (t)
that is m1 (t) = √ 2 2 2
, and m2 (t) = q 2
.
xOG (t)+yOG (t)+zOG 2 ( t )+ y2 ( t )+[2D −( z + z )]
xOG G O
OG
hence satisfying the two following equalities (in time):

yOT (t) yOG (t)

q = q (6)
2 ( t ) + y2 ( t ) + z2
xOT 2 ( t ) + y2 ( t ) + z2
xOG
OT OT OG OG

yOT (t) yOG (t)

q = q (7)
2 ( t ) + y2 ( t ) + [2D − ( z + z )] 2 2 ( t ) + y2 ( t ) + [2D − ( z + z )]2
xOT OT T O xOG OG G O

under the constraint that zG is in [0, D ].

Case 1: yOT (t) is a zero function, i.e., ∀t yOT (t) = 0.
The target is broadside to the antenna, so any ghost targets will be too (see Case 1 in
the proof of theorem 1).
Case 2: yOT (t) is not a zero function.
From Case 2 of the proof of theorem 1, there is a positive scalar β such that yOG (t) =
βyOT (t).
q q
(6) ⇔ xOG 2 ( t ) + y2 ( t ) + z2 = β x 2 ( t ) + y2 ( t ) + z2
 2 OG  OG 2  2 OT OT  OT (8)
⇔ xOG 2 ( t ) + z2
(t) + yOG = β x ( t ) + y 2 ( t ) + z2
OG OT OT OT
h i h i
2 ( t ) + y2 ( t ) + [2D − ( z + z )]2 = β2 x2 ( t ) + y2 ( t ) + [2D − ( z + z )]2
(7) ⇔ xOG (9)
OG G O OT OT T O
i h
2 − [2D − ( z + z )]2 = β2 z2 − [2D − ( z + z )]2 .
Subtracting (9) from (8), we get zOG G O OT T O
Now, we simplify the expressions of these two terms:

2
zOG − [2D − (zG + zO )]2 = −4( D − zG )( D − zO )
2
zOT − [2D − (z T + zO )]2 = −4( D − z T )( D − zO )
We deduce from this that s
D − zG
β= (10)
D − zT
Sensors 2021, 21, 4797 13 of 20

Note that β = 1 iif zG = z T .

2 ( t ) + y2 ( t ) + z2 = β2 x 2 ( t ) + y2 ( t ) + z2
 
(8) ⇔ xOG OG OG OT OT OT
2 ( t ) − β2 x 2 ( t ) = β2 z2 − z2 (11)
⇔ xOG OT OT OG

2 ( t ) − β2 x2 ( t ) is a polynomial function of degree 2, (11) is equivalent to

Since xOG OT

2
xOG (0) = β2 xOT
2
(0) + β2 zOT
2 2
− zOG (12)
. .
xOG (0) xOG = β2 xOT (0) xOT (13)
.2 .2
xOG = β2 xOT (14)
.
First case xOT = 0
.
Equation (14) implies that xOG = 0.
Consequently, for any zG in [0, D ], the vector X =
T q OG
.
 q
2 (0) + β2 z2 − z2 D − z
± β2 xOT OT OG βyOT (0) zOG 0 β yOT
(with β = G
D −z T ) defines
the trajectory of a ghost target.
.
Second case xOT 6= 0
Using (14), and squaring the terms of (13), we get xOG 2 (0) = β2 x 2 (0).
OT
2 2 2
Reporting this in (12), we obtain finally β zOT = zOG , i.e.,
 2
zOG
β2 = (15)
zOT

If z T = zO , then zG = zO . In this case, β = 1, and consequently yOG (t) = yOT (t) and
2 ( t ) = x2 ( t ) from (11). The source’s trajectory is observable up to the axial symmetry
xOG OT
around the (Oy)-axis.
zOG 2
 
−zG
Equations (10) and (15) give us D D −z T = zOT .
The unknown zG is hence a root of the following equation of degree 2:
2
z T − zO )
( z G − zO ) 2 − ( D −z T ( D − z G ) = 0 which can be expanded as follows:

2 2
( z T − zO ) z T − zO )
z2G + zG −2zO + D −z T − D (D −z T
2 = 0.
+ zO
Of course, z T is a root of this equation. For this value, zG = z T , hence β = 1.
( z T − zO ) 2
The second root (zG itself) is hence 2zO − z T − D −z T , zG . We can check readily
( z T − zO ) 2 (zO −z T )( D −z T )−(z T −zO )2
that z G − zO = zO − z T − D −z T = D −z T .
z G − zO zO − D
Hence, zT −z = D−zT (which is negative).
O
We deduce from this that:
1. when the target’s depth is larger than the array’s depth, there is a ghost whose depth
is smaller than the array’s depth, and vice versa.
− zO
2. β, which is a positive coefficient, is equal to D
D −z T , or 1.
Therefore, we have identified three ghost targets:
. .
T
the first one is defined by XOG = − xOT (0) yOT (0) zOT − xOT yOT ,
 . T
.
the second is defined by XOG = βxOT (0) βyOT (0) − βzOT β xOT βyOT ,
 . T
.
and the third by XOG = − βxOT (0) βyOT (0) − βzOT − β xOT βyOT . 

Remark 2. Most of the time, the depth of a submarine vehicle is under the operational constraint:
values of z T are in [0, z Max ] and z Max  D. For example, z Max = 400 m, while D = 4000 m.
The proof of the previous theorem must be adapted to this new constraint.
( u − zO ) 2
First, we use the fact that the function u 7→ f (u) , 2zO − u − D −u is an involution, i.e.,
f ( f (u)) = u.
Sensors 2021, 21, 4797 14 of 20

 
z20 z20
Since f (0) = 2zO − D, f 2zO − D = 0.

q 2zO −
Now the question is: what are the values of zO for which the following inequality holds:
z20 z2
D ≤ z Max , the greatest value of zO guaranteeing that 2zO − D−0zT ≤ z Max is D − D 1 − z Max D
(which is less than z Max
q ).
If zO > D − D 1 − z Max D , then z G > z Max . In this case, there is a unique ghost target given
. .
T
by XG = − x T (0) y T (0) z T − x T y T .
q
z Max
If zO ≤ D − D 1 − D , then zG ≤ z Max . In this case, there are three ghost targets:
. .
T
one is defined by XG = − x T (0) y T (0) z T − x T y T ,
. .
T
the second is defined by XG = βx T (0) βy T (0) f (z T ) β x T βy T ,
. .
T
and the third by XG = − βx T (0) βy T (0) f (z T ) − β x T βy T .
Note that the operational constraint allows us to benefit from the following range:
D −z Max
D ≤ β ≤ D−Dz Max . For example, when z Max = 10 D
, 0.9 ≤ β ≤ 1.11. Consequently, the
ghost target is very close to the target of interest.

5.2. Estimation of the Trajectory

This section is devoted to the estimation of the target’s trajectory, or in other words,
the estimation of X with t∗ = 0 (the first time). Before going into detail, we compute
the so-called Cramér–Rao lower bound to evaluate the asymptotical performance of any
unbiased estimator.
We have considered two typical scenarios. In both, the array is assumed motionless (or,
more realistically, all the mobiles are referenced to it) at the depth zO = 200 m, and the state vec-
tor defining the target’s trajectory is given by the state vector

T
X = 5000 7000 100 2.83 2.83 . The standard deviation of the measurement
is σ = 1.7 × 10−2 . The total duration of the scenario is 5 min, and the sampling time is ∆t = 4
s; consequently, the number of measurement couples is N = 75.
In the first scenario, the bottom depth is D = 2000 m, while in the second, D = 4000 m.
Note that in the first scenario, β = 0.89, and in the second one, β = 0.97. The ghost
target is hence very close to the target of interest.

5.2.1. Estimability
As pointed out in Section 1, the state vector X is “estimable” if its asymptotical
performance given by the CRLB is compatible with the physical constraints. Typically, if
the minimum standard deviation defined by the square root of the third diagonal element
of the CRLB (hence of the depth) is much larger than the depth, then X is declared non-
estimable.
1. First scenario
For this scenario, the square root of the diagonal of the CRLB σCRLB

T
= 1.16 × 106 1.59 × 106 8.22 × 105 637.9 646.1 .
2. Second scenario
With the bottom depth, things are not much better, since σCRLB

T
= 6.59 × 105 8.96 × 105 9.73 × 105 352.6 362.2 .
In both cases, the minimum standard deviations are huge. We can conclude that the
state vector is not estimable. Computations of minimum standard deviations were made
for various scenarios; in all, the state vector is not estimable.
A palliative of this is to fix the depth of the source at an arbitrary and realistic value, say
. .
T
z As , and compute the CRLB of the reduced state vector Xr , x T (0) y T (0) x T y T
Sensors 2021, 21, x FOR PEER REVIEW 15 of 20

In both cases, the minimum standard deviations are huge. We can conclude that the
Sensors 2021, 21, 4797 state vector is not estimable. Computations of minimum standard deviations were 15 made
of 20
for various scenarios; in all, the state vector is not estimable.
A palliative of this is to fix the depth of the source at an arbitrary and realistic value,
say 𝑧𝐴𝑠 , and compute the CRLB of the reduced state vector 𝑋𝑟 ≜
when
(𝑥 we assume that
𝑇 (0) 𝑦𝑇 (0) 𝑥̇ 𝑇 𝑦̇ 𝑇z)T𝑇 = z As . we
when Forassume
example, = 𝑧=
thatfor𝑧𝑇z As 300 m, the minimum standard
𝐴𝑠 . For example, for 𝑧𝐴𝑠 = 300m,
deviations
the minimumarestandard deviations are

𝜎𝐶𝑅𝐿𝐵== (281.17
281.17 319.37
319.37 1.78 2.02)T𝑇 for
1.78 2.02 forthe
thefirst
firstscenario,
scenario,and
and


σCRLB
𝜎𝐶𝑅𝐿𝐵 = (130.1 115.3 0.80 0.71
)T𝑇 for the second one.
σCRLB = 130.1 115.3 0.80 0.71 for the sec ond one.
Therefore, we propose to estimate the state vector with this hypothesis (𝑧𝐴𝑠 = 300
m). InTherefore,
so doing,wewepropose to estimate
introduce the state
a bias. The next vector with gives
subsection this hypothesis (z Asof=the
us the result 300500
m).
In so doing, we introduce
Monte Carlo simulations. a bias. The next subsection gives us the result of the 500 Monte
Carlo simulations.
5.2.2. Monte Carlo simulations
5.2.2. Monte Carlo simulations
The computation of the maximum likelihood estimator (MLE) is made with the
The computation of the maximum likelihood estimator (MLE) is made with the Gauss-
Gauss‒Newton
Newton routine.routine. No numerical
No numerical issue
issue was was encountered.
encountered.
1.
1. First
Firstscenario
scenario
The performance of the MLE is summarized in Table 5. We have numerically com-
The performance of the MLE is summarized in Table 5. We have numerically computed
puted the bias and the empirical standard deviation (given, respectively in the second and
the bias and the empirical standard deviation (given, respectively in the second and third
third
column column
of theof the table).
table). We can We can
see seethe
that that the empirical
empirical standard
standard deviation
deviation is veryisclose
verytoclose
that
to that given by the CRLB. However, as expected, the MLE is biased (of
given by the CRLB. However, as expected, the MLE is biased (of course, there is no bias course, there is no
if
bias
we choose z As = 𝑧z𝐴𝑠
if we choose
T ).=In𝑧𝑇Figure
). In Figure
9, the 9,90%
the confidence
90% confidence ellipse
ellipse is is drawn,
drawn, together
together withwith
the
the cloud
cloud of the
of the 500 500 estimates
estimates (in pink).
(in pink).

Table 5. Performance
Table 5. Performance ofof the
the estimator
estimator of the
of the reduced
reduced state state
vector.vector.
𝑿𝒓 Bias 𝝈𝒔𝒂𝒎𝒑 𝝈𝑪𝑹𝑳𝑩
Xr Bias σsamp σCRLB
5000 m 401.12 281.85 281.17
5000 m 401.12 281.85 281.17
7000mm
7000 557.24
557.24 330.87
330.87 319.37
319.37
2.83
2.83 m/s
m/s 0.13
0.13 1.58
1.58 1.78 1.78
2.83 m/s
2.83 m/s 0.12
0.12 1.81
1.81 2.02 2.02

Figure
Figure 9.
9. The
Thelocation
locationofofthe sensor
the array
sensor (in(in
array black), thethe
black), cloud of the
cloud of 500
the estimates and the
500 estimates and90%−con-
the 90%-
fidence ellipse when 𝐷 = 2000
confidence ellipse when D = 2000 m, m, 𝑧 = 300 m, and 𝑧
𝐴𝑠 z As = 300 m, and 𝑇 = 100
z T = 100 m. The symmetricaliscloud
m. The symmetrical cloud plotted
is
too.
plotted too.

2.
2. Second
Secondscenario:
scenario:Bottom depth 𝐷
Bottomdepth D==4000 m.
4000 m.
Again,
Again, the
theperformance
performanceisispresented
presentedininTable
Table6.6.The
Thebias ofof
bias the estimator
the is similar
estimator to
is similar
the one obtained for the first scenario. Only the empirical standard deviations
to the one obtained for the first scenario. Only the empirical standard deviations of of

T
x T (0) y T (0) are larger than that computed from the CRLB. However, Figure 10
shows us that the cloud of estimates is close to the true value and not spread.
Sensors 2021, 21, x FOR PEER REVIEW 16 of 20

Sensors 2021, 21, 4797 16 of 20

(𝑥𝑇 (0) 𝑦𝑇 (0))𝑇 are larger than that computed from the CRLB. However, Figure 10
shows us that the cloud of estimates is close to the true value and not spread.

Table 6.
Table Performanceofof
6. Performance thethe estimator.
estimator.

Xr𝑿𝒓 Bias
Bias 𝝈σ𝒔𝒂𝒎𝒑 𝝈𝑪𝑹𝑳𝑩
σCRLB
samp
5000 m 306.81 219.28 130.08
5000 m 306.81 219.28 130.08
7000
7000 mm 432.46
432.46 276.61
276.61 115.26
115.26
2.83
2.83 m/s
m/s 0.18
0.18 0.74
0.74 0.800.80
2.83 m/s
2.83 m/s 0.18
0.18 0.66
0.66 0.710.71

10. The location of the sensor array (in black), the cloud of the 500
Figure 10. 500 estimates
estimates ofof the
the initial
initial
positions, and
positions, and the
the90%−confidence
90%-confidenceellipse when𝐷D= =
ellipsewhen 4000
4000m,m,𝑧𝐴𝑠
z As= =
300
300m,m,
and 𝑧𝐴𝑠z As
and = 100
= 100 m, to-
m,
gether with the symmetrical cloud.
together with the symmetrical cloud.

What isisremarkable
remarkableisisthethe short
short duration
duration andand
still still the very
the very good good performance
performance (in
(in terms
terms
of of accuracy)
accuracy) of the result.
of the result. NumerousNumerous simulations
simulations (not reported
(not reported here) werehere)performed;
were per-
formed;
all all the
confirm confirm theperformance
correct correct performance of the
of the MLE. TheMLE. The shortness
shortness of the scenario
of the scenario is crucial,is
crucial, because
because everything
everything that we that we propose
propose hereproperly
here works works properly
under theunder the condition
condition that thethat
sea
the sea bottom
bottom is aDuring
is a plane. plane. During a short scenario,
a short scenario, this assumption
this assumption is likely.is likely.

6. Conclusions
6. Conclusions
In
In this
thispaper,
paper,conical-angle
conical-angle TMATMA hashas
beenbeen
addressed, and various
addressed, multipaths
and various of sound
multipaths of
have been taken into account. The sensor is a line array. Observability was
sound have been taken into account. The sensor is a line array. Observability was analyzed analyzed deeply,
allowing all the existing
deeply, allowing all the ghost targets
existing ghosttotargets
be identified. The mainThe
to be identified. results
mainareresults
that, ifare
thethat,
arrayif
detects one ray (corresponding to one path), the trajectory is not observable:
the array detects one ray (corresponding to one path), the trajectory is not observable: the the set of
ghost targetstargets
set of ghost is composed of trajectories
is composed that are
of trajectories homothetic
that to the trajectory
are homothetic of the target
to the trajectory of the
of interest,
target and their
of interest, andsymmetrical images images
their symmetrical by the axial
by the symmetry around the
axial symmetry line array.
around If
the line
the array detects two rays (corresponding to two different paths), the
array. If the array detects two rays (corresponding to two different paths), the number of number of ghost
targets is reduced
ghost targets to threeto(except
is reduced when the
three (except target
when theis target
endfireisorendfire
broadside to the antenna).
or broadside to the
When the antenna maneuvers, the target’s trajectory is observable
antenna). When the antenna maneuvers, the target’s trajectory is observable (apart (apart from the special
from
scenario
the specialwhere therewhere
scenario is one there
singleisghost target).ghost
one single Eventarget).
for “observable” scenarios, thescenar-
Even for “observable” depth
of the target is not estimable (its asymptotical standard deviation is huge). In these cases,
ios, the depth of the target is not estimable (its asymptotical standard deviation is huge).
we give a non-restrictive palliative that allows us to provide estimates close to the truth.
In these cases, we give a non-restrictive palliative that allows us to provide estimates close
In the future, in this context, many problems remain to be faced: identification of
to the truth.
the paths, maneuvering targets, and fusion of data collected by other sensors, as in [28].
In the future, in this context, many problems remain to be faced: identification of the
The problem of seeking a “good” maneuver of the observer, as it was solved in a 2D
paths, maneuvering targets, and fusion of data collected by other sensors, as in [28]. The
environment [29–31], will be addressed in the future. Some of these problems are already
problem of seeking a “good” maneuver of the observer, as it was solved in a 2D environ-
under investigation.
ment [29–31], will be addressed in the future. Some of these problems are already under
investigation.
Author Contributions: Conceptualization, A.L. and A.-C.P.; methodology, C.J.; software, A.L.;
validation, A.L., A.-C.P., C.J. and D.L.; formal analysis, A.L.; investigation, A.L.; resources, D.L.; data
curation, A.L.; writing—original draft preparation, C.J.; writing—review and editing, A.L. and C.J.;
visualization, A.-C.P.; supervision, A.-C.P.; project administration, not applicable; funding acquisition,
not applicable. All authors have read and agreed to the published version of the manuscript.
Sensors 2021, 21, 4797 17 of 20

Funding: This research received no external funding.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

Appendix A

Proof of Theorem 2. The proof

 is  made when the first leg is towards
 North (as previously).
 
0 v sin(α) −1 0
Hence, V1 = , V2 = , S1 = ,
v v cos(α) 0 1
 
− cos(2α) sin(2α)
and S2 = .
sin(2α) cos(2α)
Moreover, we assume α 6= kπ.
From Theorem 1, we have to consider the four following cases for each leg:
• the target is broadside to the antenna,
• the target is endfire to the antenna,
• the target has the same heading as the array (but is not endfire to it),
• the other cases.
Note that if the target is in case (1) during the first leg, then in case (2) during the
second one (provided that this situation is possible), the conclusion about observability
will be the same as if the target is in case (2) during the first leg, then in case (1) during the
second leg. To be convinced of this, we just have to reverse the time in the equation. This
remark allows us to shorten the proof.
Case 1: the targetis broadside  to the antenna  during
 the first leg.
x T (0) cT
Hence, PT (0) = , and VOT = during the first leg, which implies
0 0
 
x T (0) + tc T
POT (t) = for t ≤ τ. The ghost targets are also in the broadside, hence
0
   
xG (0) + tcG cG
POG (t) = and VOG = for t ≤ τ.
0 0
Can the target be endfire to the antenna? If so, the target has the same heading as the
antenna during the second leg or, in other words, VT − V2 = λV2 , and POT (t), which is
equal to POT (t) = POT (τ ) + (t − τ )(VT − V2 ), is collinear
 withV2 , whenever
 t ≥ τ. The
cT ±v
first condition cannot be satisfied since VT = , and V2 = . There is no ghost.
v 0
We skip the case where the target is in case (3) during the second leg. This will be
treated later. Therefore, we now have to consider the other cases during the second leg.
There are two possibilities for the ghost targets: those whose trajectories are defined by (i)
POG (t) = βPOT (t), and those whose trajectories are given by (ii) POG (t) = βS2 POT (t), both
for t ≥ τ.
Thederivative
  VG − V2 = βVT − βV2 , hence VG = βVT + (1 − β)V2 .
of (i) is
cG cT sin(α)
⇔ =β + (1 − β ) v ,
v v cos(α)
which implies that (1 − β) cos(α) = 1 − β. Since cos(α) 6= 1, β = 1. There is no ghost
given by (i).
Thederivative
 of (ii) is VG − V2 = βS2 VT − βV2 , hence VG = βS2 V T + (1 − β)V2 .
cG −c T cos(2α) + v sin(2α) sin(α)
⇔ =β + (1 − β ) v .
v c T sin(2α) + v cos(2α) cos(α)
1−cos(α)
We deduce that β = v c sin(2α)+v cos(2α)−v cos(α) .
T
One ghost exists if c T sin(2α) + v cos(2α) − v cos(α) is a positive quantity. If so, we
then compute cG . There is one ghost at most.
Case 2: the target is endfire to the antenna during the first leg.
Sensors 2021, 21, 4797 18 of 20

   
0 0
Hence, PT (0) = , and VOT = , which implies that
y T (0) cT
 
0
POT (t) = for t ≤ τ. During this first leg, the ghost targets are also
y T (0) + tc T
   
0 0
endfire to the antenna, so POG (t) = for t ≤ τ, and VG − V1 = .
yG (0) + tcG cG
Again, we skip the case where the target is in case (3) during the second leg. This will
be treated later. So, we now have to consider the other cases during the second leg.
There are two possibilities for the ghost targets: those whose trajectories are defined by (i)
POG (t) = βPOT (t) and those whose trajectories are given by (ii) POG (t) = βS2 POT (t), both
for t ≥ τ.
Thederivative  of (i) is VG − V 2 = βVT − βV  2 , henceVG = βVT + (1 − β)V2 .
0 0 sin(α)
⇔ =β + (1 − β ) v .
cG + v cT + v cos(α)
We deduce that β = 1. There is no ghost.
Now, differentiating (ii) gives us VG − V2 = βS2 VT − βV2 , hence VG = βS2 VT +
(1 − β)V2 .     
0 sin(2α) sin(α)
⇔ = β(c T + v) + (1 − β ) v .
cG + v cos(2α) cos(α)
⇒ β(c T + v) sin(2α) + (1 − β)v sin(α) = 0 .
sin(α)
We deduce that β = −v c sin(2α)+v sin(2α)−v sin(α) . One ghost exists if c T sin(2α) +
T
v sin(2α) − v sin(α) is a negative quantity. If so, we then compute cG . There is one ghost at
most.
Case 3: the target has the  same  heading as the array (but is not endfire to it)
0
As in case (2), VOT = , but here, the first component of POT (t) is not zero.
cT
 
0
Hence, VT = , and the target cannot be endfire to the antenna during the second
cT + v
   
0 0
leg. In this case, VOG = , hence VG = .
βc T βc T + v
Can the target be broadside to the antenna? The answer is positive if V2 ⊥VT and  VOT
±v
is collinear to POT (t), when t ≥ τ. The first condition implies that V2 = . Since
0
POT (t) = POT (τ ) + (t − τ )(VT − V2 ), the second condition is satisfied if POT (τ ) is collinear
to VT − V2 . This is not the case when the first component of POT (τ ) is zero, while the first
component of VT − V2 is ±v.
So, we now have to consider the other cases during the second leg. There are two
possibilities for the ghost targets: those whose trajectories are defined by (i) POG (t) =
βPOT (t) and those whose trajectories are given by (ii) POG (t) = βS2 POT (t), both for t ≥ τ.
The derivative of (i) is VG − V2 = βVT − βV2 , hence VG = βVT + (1 − β)V2 or, in other
words,     
0 0 sin(α)
= + (1 − β ) v . We conclude that β = 1, i.e.,
βc T + v β(c T + v) cos(α)
there is no ghost.
 POG (t) =βS2 POT (t), then
If  VG = βS2 VT + (1 − β )V2 
0 sin(2α) sin(α)
= β(c T + v) + (1 − β ) v .
βc T + v cos(2α) cos(α)
This implies that α = 0, which must be rejected by assumption. There is no ghost.
The other cases:
In the other cases, the motion of ghost targets is defined during the first leg by
when t ≤ τ,
POG (t) = β 1 POT (t) (A1)
or POG (t) = γ1 S1 POT (t) (A2)
Sensors 2021, 21, 4797 19 of 20

and during the second leg by

when t ≥ τ,
POG (t) = β 2 POT (t) (A3)
or POG (t) = γ2 S2 POT (t) (A4)
Hence, at time τ, the position of a ghost target is

POG (τ ) = β 1 POT (τ ) (A5)

or POG (τ ) = γ1 S1 POT (τ ) (A6)

and POG (τ ) = β 2 POT (τ ) (A7)
or POG (τ ) = γ2 S2 POT (τ ) (A8)
Of course, (A5) and (A6) are not compatible, and neither are (A7) and (A8).
Now, let us show that (A5) is not compatible with (A8):
Indeed, if POG (τ ) = β 1 POT (τ ) = γ2 S2 POT (τ ), then

β1
P (τ ) = S2 POT (τ ) (A9)
γ2 OT
β β
Equation (A9) implies that POT (τ ) is an eigenvector of S2 , with the eigenvalue γ12 . Since γ12
is positive, this eigenvalue is equal to 1, i.e., γ2 = β 1 . Hence, POT (τ ) is in the second leg.
Hence, the set of ghost targets is reduced to those whose positions at time τ are given
by (A5) or (A6), and (A7). Now suppose that a ghost target satisfies (A6) and (A7). By the
same computation, we conclude that POT (τ ) is in the first leg, which is impossible since
POT (τ ) is in the second leg.
We have proven that (A5) and (A7) only are compatible. It follows that a ghost target
verifies these two equalities (given by (A1) and (A3)):
POG (τ ) = β 1 POT (τ ) = β 2 POT (τ ).
Hence, β 1 = β 2 .
Now taking the derivative of the two members of (A1) and of (A3), we obtain
VG = β 1 (VT − V1 ) + V1 = β 1 (VT − V2 ) + V2 , which is equivalent to
( β 1 − 1)(V2 − V1 ) = 0.
Since V2 6= V1 , β 1 = 1.
Putting this value into (A1) or (A3), we finally get PG (t) = PT (t). The “ghost” is the
target of interest. In conclusion, there is no ghost target. 

References
1. Nardone, S.; Lindgren, A.; Gong, K. Fundamental properties and performance of conventional bearings-only target motion
analysis. IEEE Trans. Autom. Control 1984, 29, 775–787. [CrossRef]
2. Song, T.L.; Um, T.Y. Practical guidance for homing missiles with bearings-only measurements. IEEE Trans. Aerosp. Electron. Syst.
1996, 32, 434–443. [CrossRef]
3. Clavard, J.; Pillon, D.; Pignol, A.-C.; Jauffret, C. Target Motion Analysis of a Source in a Constant Turn from a Nonmaneuvering
Observer. IEEE Trans. Aerosp. Electron. Syst. 2013, 49, 1760–1780. [CrossRef]
4. Farina, A. Target tracking with bearingsOnly measurements. Signal. Process 1999, 78, 61–78. [CrossRef]
5. Jauffret, C.; Pillon, D.; Pignol, A.-C. Bearings-Only Maneuvering Target Motion Analysis from a Nonmaneuvering Platform. IEEE
Trans. Aerosp. Electron. Syst. 2010, 46, 1934–1949. [CrossRef]
6. Zhang, Y.; Lan, J.; Mallick, M.; Li, X.R. Bearings-Only Filtering Using Uncorrelated Conversion Based Filters. IEEE Trans. Aerosp.
Electron. Syst. 2021, 57, 882–896. [CrossRef]
7. Aidala, V.; Hammel, S. Utilization of modified polar coordinates for bearings-only tracking. IEEE Trans. Autom. Control 1983, 28,
283–294. [CrossRef]
8. Arulampalam, S.; Clark, M.; Vinter, R. Performance of the shifted Rayleigh filter in single-sensor bearings-only tracking. In
Proceedings of the 2007 10th International Conference on Information Fusion, Quebec, QC, Canada, 9–12 July 2007; pp. 1–6.
9. Laneuville, D.; Jauffret, C. Recursive Bearings-Only TMA via Unscented Kalman Filter: Cartesian vs. Modified Polar Coordinates.
In Proceedings of the 2008 IEEE Aerospace Conference, Big Sky, MT, USA, 1–8 March 2008; pp. 1–11.
Sensors 2021, 21, 4797 20 of 20

10. Arulampalam, M.S.; Ristic, B.; Gordon, N.; Mansell, T. Bearings-Only Tracking of Maneuvring Targets Using Particle Filters.
Eurasip J. App. Sig. Process. 2004, 15, 2351–2365.
11. Zhang, Y.J.; Xu, G.Z. Bearings-Only Target Motion Analysis via Instrumental Variable Estimation. IEEE Trans. Signal. Process.
2010, 58, 5523–5533. [CrossRef]
12. Doğançay, K. On the efficiency of a bearings-only instrumental variable estimator for target motion analysis. Signal. Process. 2005,
85, 481–490. [CrossRef]
13. Chan, Y.T.; Rea, T.A. Bearings-only tracking using data fusion and instrumental variables. In Proceedings of the Proceedings of
the Third International Conference on Information Fusion, Paris, France, 10–13 July 2000.
14. Kirubarajan, T.; Bar-Shalom, Y.; Lerro, D. Bearings-only tracking of maneuvering targets using a batch-recursive estimator. IEEE
Trans. Aerosp. Electron. Syst. 2001, 37, 770–780. [CrossRef]
15. Jauffret, C.D.; Pillon, D. Observability in Passive Target Motion Analysis. IEEE Trans. Aerosp. Electro. Syst. 1996, AES-32,
1290–1300. [CrossRef]
16. Le Cadre, J.; Jauffret, C. Discrete-time observability and estimability analysis for bearings-only target motion analysis. IEEE Trans.
Aerosp. Electron. Syst. 1997, 33, 178–201. [CrossRef]
17. Nardone, S.C.; Aidala, V.J. Observability Criteria for Bearings-Only Target Motion Analysis. IEEE Trans. Aerosp. Electron. Syst.
1981, AES-17, 162–166. [CrossRef]
18. Urick, R.J. Principles of Underwater Sound; McGraw-Hill: New York, NY, USA, 1983.
19. Van Trees, H.L. Optimum Array Processing: Part IV of Detection, Estimation, and Modulation Theory; Wiley Interscience: New York,
NY, USA, 2002.
20. Abraham, D.A. Underwater Acoustic Signal Processing: Modeling, Detection, and Estimation; Modern Acoustic and Signal Processing;
Springer: Berlin/Heidelberg, Germany, 2019.
21. Gong, K.F. Multipath Target Motion Analysis: Properties and Implication of the Multipath Process; in Technical Report 6687; NUSC:
Newport, RI, USA, 1982.
22. Blanc-Benon, P.; Jauffret, C. Target Motion Analysis from Bearing and Time-delay Measurements: The Use of Multipath. IEEE
Trans. Aerosp. Electron. Syst. 1997, AES-33, 813–824. [CrossRef]
23. Oh, R.; Song, T.L.; Choi, J.W. Batch Processing through Particle Swarm Optimization for Target Motion Analysis with Bottom
Bounce Underwater Acoustic Signals. Sensors 2020, 20, 1234. [CrossRef] [PubMed]
24. Steven, S.M. Fundamentals if Statistical Signal Processing—Estimation Theory; Prentice Hall: Englehood Cliffs, NJ, USA, 1993.
25. Jauffret, C. Observability and fisher information matrix in nonlinear regression. IEEE Trans. Aerosp. Electron. Syst. 2007, 43,
756–759. [CrossRef]
26. Pignol, A.-C.; Jauffret, C.; Pillon, D. A statistical fusion for a leg-by-leg bearings-only TMA without observer maneuver. In
Proceedings of the 13th International Conference on Information Fusion, Edinburgh, UK, 26–29 July 2010; Volume 6, pp. 1–8.
[CrossRef]
27. Bar-Shalom, Y.; Rong Li, X.; Kirubarajan, T. Estimation and Tracking: Principles, Techniques and Software; Artech House: Boston, MA,
USA, 1993.
28. Payan, J.; Lebon, A.; Perez, A.-C.; Jauffret, C.; Laneuville, D. Passive Target Motion Analysis by Fusion of Linear Arrays and
Sonobuoys in a Cluttered Environment. IEEE Trans. Aerosp. Electron. Syst. 2021, 1. [CrossRef]
29. Le Cadre, J.-P.; Gauvrit, H. Optimization of the observer motion for bearings-only target motion analysis. In Proceedings of the
1st Australian Data Fusion Symposium, Adelaide, SA, Australia, 21–22 November 1996; pp. 190–195.
30. Fawcett, J. Effect of course maneuvers on bearings-only range estimation. IEEE Trans. Acoust. Speech, Signal. Process 1988, 36,
1193–1199. [CrossRef]
31. Passerieux, J.M.; Van Cappel, D. Optimal observer maneuver for bearings-only tracking. IEEE Trans. Aerosp. Electron. Syst. 1998,
34, 777–788. [CrossRef]