remote sensing

Article
MIMO Radar Waveform Design for Multipath Exploitation
Using Deep Learning
Zixiang Zheng 1 , Yue Zhang 1, *, Xiangyu Peng 1 , Hanfeng Xie 1 , Jinfan Chen 1 , Junxian Mo 1 and Yunfeng Sui 2

1 School of Electronic and Communication Engineering, Sun Yat-sen University, Shenzhen 518107, China;
[email protected] (Z.Z.)
2 Research Center, Second Research Institute of CAAC, Chengdu 610041, China
* Correspondence: [email protected]

Abstract: This paper investigates the design of waveforms for multiple-input multiple-output (MIMO)
radar systems that can exploit multipath returns to enhance target detection performance. By making
reasonable use of multipath information in the waveform design, MIMO radar can effectively improve
the signal-to-interference and noise ratio (SINR) of the receiver under a constant modulus (CM)
constraint. However, optimizing the waveform design under these constraints is a challenging non-
linear and non-convex problem that cannot be easily solved using traditional methods. To overcome
this challenge, we proposed a novel waveform design method for MIMO radar in multipath scenarios
based on deep learning. Specifically, we leveraged the powerful nonlinear fitting ability of neural
networks to solve the non-convex optimization problem. First, we constructed a deep residual
network and transform the CM constraint into a phase sequence optimization problem. Next, we
used the constructed waveform optimization design problem as the loss function of the network.
Finally, we used the adaptive moment estimation (Adam) optimizer to train the network. Simulation
results demonstrated that our proposed method outperformed existing methods by achieving better
SINR values for the receiver.

Keywords: MIMO radar; multipath exploitation; waveform design; deep learning; SINR

Citation: Zheng, Z.; Zhang, Y.; Peng, 1. Introduction

X.; Xie, H.; Chen, J.; Mo, J.; Sui, Y.
Radar systems often encounter multipath effects when detecting low-altitude targets.
MIMO Radar Waveform Design for
In such scenarios, the received returns consist not only of the backscattered line-of-sight
Multipath Exploitation Using Deep
Learning. Remote Sens. 2023, 15, 2747.
(LOS) component, but also of the multipath returns component [1–3]. However, the
https://doi.org/10.3390/rs15112747
presence of multipath returns can cause the received echo signal to fluctuate and even
cancel, which will reduce the performance of target detection and parameter estimation.
Academic Editor: Stefano Tebaldini The early research has been devoted to suppressing the multipath returns [4–7]. In [4],
Received: 28 March 2023 multipath returns are regarded as clutter and suppressed.
Revised: 5 May 2023 However, the principle of multipath generation suggests that both the direct and
Accepted: 22 May 2023 multipath returns are coherent and contain target energy [8]. If the energy of the multipath
Published: 25 May 2023 returns can be accurately estimated and accumulated, it can improve the detection and
tracking performance of the target [9–11].
Multiple-input multiple-output (MIMO) radar is a new type of radar system in which
the transmitting antennas can transmit mutually independent signals [12–16]. By adaptively
Copyright: © 2023 by the authors. adjusting the MIMO radar transmission waveform for different tactical needs, the radar
Licensee MDPI, Basel, Switzerland. detection performance can be significantly improved in complex environments [17,18].
This article is an open access article MIMO channel models can be broadly categorized into two types based on the meth-
distributed under the terms and
ods used for their establishment. The first type is deterministic channel models [19–21],
conditions of the Creative Commons
where precise information about the channel is obtained, and the wireless propagation is
Attribution (CC BY) license (https://
deemed as a deterministic process. This enables the determination of the spatio-temporal
creativecommons.org/licenses/by/
characteristics of any point in space. One study [19] employed a multilayer artificial neural
4.0/).

Remote Sens. 2023, 15, 2747. https://doi.org/10.3390/rs15112747 https://www.mdpi.com/journal/remotesensing

Remote Sens. 2023, 15, 2747 2 of 20

network method to predict channel characteristics, thereby overcoming the low compu-
tational efficacy of traditional deterministic channel models. In contrast, the other type is
statistical channel models [22–25], wherein the channel is regarded as a stochastic process,
and a probability distribution is used to describe its temporal and spatial changes.
Depending on the specific application requirements of the radar, the waveform
optimization design criteria for MIMO radar can be divided into four types. The first
type of optimization criterion aims to maximize the signal-to-interference and noise ratio
(SINR) [26–29]. Through appropriate waveform design, MIMO radar can maximize the
SINR to improve target detection performance. In [29], an optimization design problem
for maximizing the SINR based on constant modulus, similarity, and spectrum constraints
was presented, and a semi-definite programming method was used to solve the problem.
The second type of optimization criterion aims to maximize mutual information [30,31].
By maximizing the amount of mutual information between the corresponding target and
the received waveform, the echo can exhibit more target characteristics. The third type
of optimization criterion is the pattern matching problem [32,33]. The goal of this type of
problem is to concentrate the transmitted beam energy of the MIMO radar in a specified
airspace while minimizing the transmitted energy of the side lobes. The last type of opti-
mization criterion is the orthogonal waveform design problem [34–36]. By reducing the
correlation between the transmitted waveforms, the performance of matched filtering can
be improved.
Combining high degrees of freedom in the MIMO radar waveform design with multi-
path exploitation [37–39] has shown great potential for further improving target detection
performance. Most existing waveform design methods using multipath returns focus on
slow time domain weight optimization for fixed waveform. In [37], the authors proposed to
improve the detection performance of moving targets in multipath scenarios by optimizing
the weights between different coherent processing intervals (CPIs), which they performed
by using the orthogonal frequency division multiplexing (OFDM) waveform. On the other
hand, in [38], the authors proposed an OFDM MIMO adaptive waveform design algorithm
based on the mutual information criteria, thus aiming to select the best OFDM waveform
by maximizing the mutual information between the state and measurement vectors. This
particular type of slow time domain MIMO radar waveform can present challenges in
scenarios that demand low range domain sidelobes, due to its high range domain sidelobes.
Additionally, complex signal processing algorithms are required for slow time domain
waveforms, which can increase system cost and overall complexity.
In recent years, there has been an increasing interest among radar technicians in explor-
ing fast time domain waveform design methods that make use of multipath information.
In [40], the transmit waveform and receive filter of a MIMO radar system were jointly de-
signed to maximize the SINR of the receiver for multipath exploitation. In [41], the robust
joint design of MIMO radar transmit waveform and receive filter banks was considered un-
der the uncertainty of multipath returns information. This method addressed the limitation
of requiring accurate prior knowledge of the multipath returns information in the method
presented in [40]. The optimization problem discussed in the literature is non-convex and,
therefore, cannot be solved directly. Existing research primarily relies on algorithms such
as semi-definite relaxation algorithms to solve non-convex problems indirectly by relaxing
the objective function or the constraint to a more tractable form. Oftentimes, these methods
experience degradation in performance due to the relaxation process.
The constant modulus (CM) constraint is a common requirement for MIMO radar
waveform design to avoid distortion of the transmit signal in near-saturated operating
modes of the high frequency amplifier [42]. The objective of MIMO radar waveform design
for multipath exploitation is to maximize the SINR of the receiver while adhering to the CM
constraint of the transmit waveform. However, this problem is non-linear and non-convex,
thereby making it difficult to solve with traditional optimization algorithms.
Deep learning, as a natural non-linear system [43], can effectively solve such problems.
In this paper, we proposed a method to design MIMO radar fast time domain waveforms for
Remote Sens. 2023, 15, 2747 3 of 20

multipath exploitation using deep learning. We used a residual neural network to directly
solve the non-convex optimization problem, rather than indirectly solving it by relaxing
the CM constraint or SINR function. Firstly, we transformed the CM constraint problem
into an unconstrained phase optimization problem and took a random phase sequence as
the input of the network. Next, the reciprocal of the SINR function of the received signal
model in the multipath scenario was used as the loss function of the network. After training
the network using an adaptive moment estimation (Adam) optimizer, the output of the
network became the phase sequence of the designed waveform.
The simulation results demonstrated that the proposed algorithm made full use of
the multipath energy, thus resulting in the waveform with higher SINR performance and
detection probability compared to existing methods.
In summary, the main contributions of this article can be summarized as follows:
1. A MIMO radar signal model was constructed for multipath scenarios;
2. The MIMO radar waveform design problem for multipath exploitation was mod-
eled as a maximizing SINR problem with the constant modulus constraint on the
transmit waveform;
3. Our proposed MIMO radar waveform design algorithm employed deep learning
that utilized the non-linear fitting ability of neural networks to directly solve the
non-convex waveform optimization problem.
The remainder of this paper is organized as follows. In Section 2, the multipath
signal model for the MIMO radar is established. In Section 3, the MIMO radar waveform
optimization design problem is formulated, and the waveform design algorithm based
on deep learning is presented. Section 4 provides the simulation results to demonstrate
the effectiveness of the proposed algorithm and the superiority of the designed waveform.
Finally, Section 5 draws the conclusions.
Notations: In this paper, we use italic letters for scalars, bold italic lowercase letters
for vectors, and bold italic uppercase letters for matrices. The superscripts (·) T , (·) H and
(·)∗ denote the transpose, conjugate transpose, and conjugate, respectively. C N × N denotes
the sets of N × N complex matrices. vec(·) denotes stacking the matrix by column. The
symbol ⊗ denotes the Kronecker product. E(·) represents the calculation expectation, and
tr(·) denotes the trace of a square matrix. k·k denotes the Frobeneous norm. Re{ a} and
Im{ a} denote the real part and imaginary part of the vector a, respectively.

2. Signal Model
A colocated MIMO radar consisting of NT transmitting antennas and NR receiving
antennas was considered in the multipath scenario. The transmit waveform of the lth
snapshot can be expressed as

sl = [sl (1), sl (2), . . . , sl ( NT )] T , (1)

where l = 1, 2, . . . , L, L denotes the number of samples in the fast time domain. The
transmit signal matrix can be represented by

S = [s1 , s2 , . . . , s L ] ∈ C NT × L . (2)

It is assumed that the unit spacing in the transmitting antenna array and the receiving
antenna array is set at half wavelength, that is, d = λ/2. The synthetic signal of the lth
snapshot at the azimuth θ can be expressed as

yl = atT (θ )sl , l = 1, 2, . . . , L, (3)

Remote Sens. 2023, 15, 2747 4 of 20

j2π ( NT −1)d sin θ T

 
j2πd sin θ
where at (θ ) = 1, e− λ , . . . , e− λ denotes the transmit steering vector.
The signal received by the MIMO radar receiver of the lth snapshot can be modeled as

zl = αA(θ )sl = αar (θ ) atT (θ )sl , l = 1, 2, . . . , L, (4)

j2π ( NR −1)d sin θ T
 
j2πd sin θ
− −
where α is the complex reflection coefficient, and ar (θ ) = 1, e λ ,...,e λ

denotes the receive steering vector A(θ ) = ar (θ ) atT (θ ).

As shown in Figure 1, it is assumed that there is a reflective surface in the scenario that
results in multipath returns. When the radar system transmits signals to detect a target,
the returns received by the radar array include not only direct returns, but also multipath
returns from the reflector. The transmit–receive path of direct returns is A → D → A.
Multipath returns include the multipath of the transmission process and the multipath of
the reception process, namely, A → B → D → A and A → D → B → A. In addition, there
are returns from the clutter area, namely A → C → A. The received signal model is given
by
y = yd + ym + yc + yn , (5)
where yd , ym , yc , and yn represent the direct returns, multipath returns, clutter returns, and
noise, respectively.

Wood
Clutter area
(C)

Target
Building (D)
Reflection surface $
%
(B)

"
!

Transmit array
(A)
Receive array

Figure 1. Multipath scenario diagram.

2.1. Direct Returns Model

Assuming that the target velocity and the angle between the target motion direction
and LOS are v and θv , respectively, the Doppler frequency of the direct returns can be
expressed as
2v cos θv
Fd = , (6)
λ
where λ is the wavelength of the transmit waveform. The shift matrix is defined as G l ,
which is represented as

1, if m − n + l = 0
G l (m, n) = . (7)
0, if m − n + l 6= 0
Remote Sens. 2023, 15, 2747 5 of 20

The direct returns model of the pth pulse can be expressed as

Y d,p = αar (θd ) atT (θd )SG0 e j2πFd ( p−1) ∈ C NR ×2L , (8)

where α is the complex scattering coefficient of the target, and at (θd ) and ar (θd ) denote the
transmit steering vector and receive steering vector in the target azimuth θd , respectively.
Assuming that the radar transmits P pulses in a CPI, we have
h iT
yd = yd,1 T , yd,2 T , . . . , yd,P T , (9)

where yd,p = vec(Y d,p ). The direct return model vector can be expressed as
  
yd = α f ⊗ G0T ⊗ ar (θd ) atT (θd ) s, (10)

h iT
where s = vec(S), f = 1, e j2πFd , . . . , e j2πFd ( P−1) .

2.2. Multipath Returns Model

The Doppler frequency of the multipath returns can be given by

Fd cos(θi − θv ) Fd
Fm = + , (11)
2 2 cos θv

where θi is the angle between DB and DA.

For the multipath returns, the transmit steering vector and the receive steering vector
will point in different azimuths due to the reflective surface. Taking into account both the
receiving and transmitting multipath, the received signal model of the pth pulse can be
represented as
 
Y m,p = ρα ar (θm ) atT (θd ) + ar (θd ) atT (θm ) SG lm e j2πFm ( p−1) ∈ C NR ×2L , (12)

where ρ and θm are the complex reflection coefficient of the surface and the arrival azimuth
of multipath returns, respectively, and lm is the relative fast time delay of the multipath
path with respect to the direct path.
Considering the CPI, the multipath returns model can finally be expressed as
  
ym = ρα f 0 ⊗ G lTm ⊗ ar (θd ) atT (θm ) + ar (θm ) atT (θd ) s, (13)

h iT
where f 0 = 1, e j2πFm , . . . , e j2πFm ( P−1) .

2.3. Clutter
In radar operations, there will inevitably be reflections from other objects in the scene
that are not desired, which will have a noticeable effect on the received returns. The clutter
returns of the pth pulse can be expressed as

Y c,p = βar (θc ) atT (θc )SG lc ∈ C NR ×2L , (14)

where β is the complex backscattering coefficient of the clutter area, θc is the azimuth of
the clutter area, and lc is the relative fast time delay of the clutter path with respect to the
direct path. The clutter signal model can be further expressed as
  
yc = β 1 P ⊗ G lTc ⊗ ar (θc ) atT (θc ) s, (15)

where 1 P denotes a row unit vector of length P.

Remote Sens. 2023, 15, 2747 6 of 20

2.4. Noise
There is also signal-independent interference in the received data, which is caused by
system noise. Typically, signal-independent noise is modeled as a Gaussian distribution.

3. Methods
3.1. Problem Formulation
The SINR is a commonly used optimization criterion that is closely related to target
detection and parameter estimation performance. In our approach, we maximized the
SINR in multipath scenarios to optimize the waveform design. Unlike common methods
that treat multipath returns as clutter and suppress them, we leveraged the information
from the multipath returns to improve the SINR of the output.
In order to calculate the SINR, the signal power of the received direct returns, multipath
returns, clutter, and noise should be calculated separately and expressed as Pd , Pm , Pc ,
and Pn .
Based on the direct returns model, the signal power of the direct returns can be
expressed as

P L 
2

1
Pd =
L ∑ ∑E αe j2πFd ( p−1)
A(θd )sl
p =1 l =1
P 
2
 h i1 L
= ∑ E e j2πFd ( p−1) E k α k2
L l∑
slH at (θd ) arT (θd ) ar∗ (θd ) atH (θd )sl
p =1 =1
P   h L
! (16)
2 i
= ∑ E e j2πFd ( p−1) E kαk tr at (θd ) ∑ sl sl at (θd )
2 H H
p =1 l =1
h i
 E kαk 2
P 
2 2
= ∑ E e j2πFd ( p−1)
L
atH (θd )S ,
p =1

where A(θd ) = ar (θd ) atT (θd ).

The signal power of the multipath returns consists of two parts, namely, the signal
power of the transmitting multipath returns and the signal power of the receiving multipath
returns, which can be expressed as

P L 
2
 P L 
2

1 1
Pm =
L ∑ ∑E ραe j2πFm ( p−1) B(θmd )sl +
L ∑ ∑E ραe j2πFm ( p−1) B(θdm )sl
p =1 l =1 p =1 l =1
P i  1 L H

hi h 2
= ∑ 2 2
E kαk E kρk E e j2πFm ( p−1)
L l∑
sl at (θd ) arT (θm ) ar∗ (θm ) atH (θd )sl
p =1 =1
P hi h i   L
2 1
+ ∑ 2 2
E kαk E kρk E e j2πFm ( p−1)
L l∑
slH at (θm ) arT (θd ) ar∗ (θd ) atH (θm )sl (17)
p =1 =1
! !
P L P L
= ∑ η tr atH (θd ) ∑ sl slH at (θd ) + ∑ η tr atH (θm ) ∑ sl slH at (θm )
p =1 l =1 p =1 l =1
P 
2 2

η
= ∑ L
atH (θd )S + atH (θm )S ,
p =1

h i h i  2

where η = E kαk2 E kρk2 E e j2πFm ( p−1) , B(θmd ) = ar (θm ) atT (θd ), and
B(θdm ) = ar (θd ) atT (θm ).
Remote Sens. 2023, 15, 2747 7 of 20

The signal power of the clutter is given by

P L  
1
Pc =
L ∑ ∑E k βA(θc )sl k 2

p =1 l =1
P L
2 1
h i
= ∑ E k β k
L ∑ slH at (θc )arT (θc )ar∗ (θc )atH (θc )sl
p =1 l =1
P h i L
! (18)
= ∑E k βk 2
tr atH (θc ) ∑ sl slH at (θc )
p =1 l =1
h i
P E k β k2 2
= ∑ L
atH (θc )S ,
p =1

where A(θc ) = ar (θc ) atT (θc ).

The noise power can be expressed as

Pn = PNT Lσn2 , (19)

where σn2 represents the variance of Gaussian white noise.

To sum up, the output SINR can be expressed as

Pd + Pm
SINR =
Pc + Pn
2 2 (20)
∑ Pp=1 µ1 atH (θd )S + ∑ Pp=1 µ2 atH (θm )S
= 2
,
∑ Pp=1 µ3 atH (θc )S + PNT Lσn2
   
2 h 2
i
E[kαk2 ] E e j2πFd ( p−1) ke j2πFm ( p−1) k E [ k ρ k2 ] 2
h i
+E E[kραk2 ] E ke j2πFm ( p−1) k
where µ1 = L , µ 2 = L ,
E [ k β k2 ]
and µ3 = L .
The aim of this paper was to optimize the waveform design by maximizing the output
SINR with the constant modulus constraint. Therefore, the waveform optimization problem
in this paper can be summarized as
2 2
∑ Pp=1 µ1 atH (θd )S + ∑ Pp=1 µ2 atH (θm )S
max 2
S ∑ Pp=1 µ3 atH (θc )S + PNT Lσn2 (21)
s.t. |s(l )| = 1, l = 1, . . . , MNT .

3.2. The Proposed Design Method

The problem of optimizing MIMO radar waveform to maximize SINR with a CM
constraint is challenging due to the non-convex nature of both the SINR function and the
CM constraint. This is a high-dimensional and non-convex problem that cannot be solved
optimally using conventional methods. However, deep learning models are highly suitable
for solving this problem, since they are high-dimensional and non-linear systems.
As the residual network can effectively tackle the issues of gradient disappearance
and explosion while increasing the number of network layers [44], we constructed a deep
residual network to use its non-linear fitting ability to solve the nonconvex problem of
maximizing SINR with the CM constraint. The optimized training network designed in this
paper is displayed in Figure 2 and comprises five modules: input, forward propagation,
output, loss function, and Adam optimizer.
Remote Sens. 2023, 15, 2747 8 of 20

J "# $

Input Output

! 1 % 1

! 2 % 2
J J J J J J J "# $

瀖瀖
瀖瀖

! "# $ % "# $

Adam optimizer Compute Loss

&'( $)** Function

Figure 2. Constructed residual network.

3.2.1. Input and Output

A randomly generated normalized phase sequence of length NT L was used as the
input of the network and can be represented as

x I = [ x I (1), . . . , x I (i ), . . . , x I ( NT L)] ∈ C NT L×1 , (22)

where x I (i ) ∈ [0, 1], i = 1, 2, . . . , NT L.

The output of the network is a normalized phase sequence and can be represented as

xO = [ xO (1), . . . , xO (i ), . . . , xO ( NT L)] ∈ C NT L×1 , (23)

where xO (i ) ∈ [0, 1], i = 1, 2, . . . , NT L.

3.2.2. Forward Propagation Module

The constructed residual network consists of 10 residual blocks, each comprising two
fully connected neural network layers. The first layer of the first residual block is a neural
network with weight matrix dimension of NT L × J, where the number of neurons is J. The
second layer is a neural network with weight matrix dimension of J × J and J neurons.
The first layer of the last residual block is a neural network with weight matrix
dimension of J × J, with J neurons. The second layer is a neural network with weight
matrix dimension of J × NT L, with J neurons.
To achieve dimension matching of the residual network, there is a network with
weight matrix dimension of NT L × J and J neurons between the input and output of the
first residual block, and there is a network with weight matrix dimension of J × NT L and J
neurons between the input and output of the last residual block.
The other residual blocks in between consist of two layers of neural networks with
weight matrix dimension of J × J and J neurons.

3.2.3. Loss Function

The output of forward propagation is a normalized phase sequence x̂O ∈ C NT L×1 .
Assuming ϕ̂ = 2π × x̂O , the original signal from one forward propagation output can
be expressed as ŝ = e jϕ̂ ∈ C NT L×1 . By the inverse process of vectorizing the waveform
columns, an output signal matrix of dimension NT × L can be obtained, where the lth
column is the signal of the lth snapshot.
As the residual network cannot directly handle complex problems, an algebraic trans-
formation ofthe original objective function is necessary. Assuming ŝ = cos ϕ̂ + j sin ϕ̂ and
atT (θ ) = Re atT (θ ) + j ∗ Im atT (θ ) , the synthetic signal of the lth snapshot of MIMO


radar in azimuth θ can be expressed as

Remote Sens. 2023, 15, 2747 9 of 20

yθ (l ) = atT (θ )ŝl
 
= Re{ atT (θ )} + j ∗ Im{ atT (θ )} (cos ϕ̂l + j ∗ sin ϕ̂l ) (24)
h i
= Re{ atT (θ )} cos ϕ̂l − Im{ atT (θ )} sin ϕ̂l + j ∗ Re{ atT (θ )} sin ϕ̂l + Im{ atT (θ )} cos ϕ̂l .

According to (24), we have

n o n o
Re{yθ (l )} = Re atT (θ ) cos ϕ̂l − Im atT (θ ) sin ϕ̂l , (25)

n o n o
Im{yθ (l )} = Re atT (θ ) sin ϕ̂l + Im atT (θ ) cos ϕ̂l . (26)

The direct returns, multipath returns, and clutter of the lth snapshot of the pth pulse
can be expressed as

yd,p (l ) =αe j2πFd ( p−1) Re{ atT (θd )} cos ϕ̂l − Im{ atT (θd )} sin ϕ̂l
h i (27)
+ j ∗ Re{ atT (θd )} sin ϕ̂l + Im{ atT (θd )} cos ϕ̂l ,


ym,p (l ) = ραe j2πFm ( p−1) Re{ atT (θd )} cos ϕ̂l − Im{ atT (θd )} sin ϕ̂l
h i
+ j ∗ Re{ atT (θd )} sin ϕ̂l + Im{ atT (θd )} cos ϕ̂l
(28)
+ Re{ atT (θm )} cos ϕ̂l − Im{ atT (θm )} sin ϕ̂l
h i
+ j ∗ Re{ atT (θm )} sin ϕ̂l + Im{ atT (θm )} cos ϕ̂l ,


yc,p (l ) = β Re{ atT (θc )} cos ϕ̂l − Im{ atT (θc )} sin ϕ̂l
h i (29)
+ j ∗ Re{ atT (θc )} sin ϕ̂l + Im{ atT (θc )} cos ϕ̂l .

Considering the delay of multipath returns and clutter relative to direct returns, the
direct returns, multipath returns, and clutter of the pth pulse can be expressed as

yd,p (1)
 
 yd,p (2) 
 
yd,p =  ... , (30)
 
 
y ( L)
d,p
0L

0l m
 
 ym,p (1) 
 
 ym,p (2) 
ym,p =  . , (31)
 
 .. 
 
ym,p ( L)
0 L − lm
Remote Sens. 2023, 15, 2747 10 of 20

0l c
 
 yc,p (1) 
 
 yc,p (2) 
yc,p =  . , (32)
 
 .. 
 
yc,p ( L)
0 L − lc
where 0 L denotes a column zero vector of length L. Taking into account the coherent pulse
interval P, the direct returns, multipath returns, and clutter can be finally expressed as
h iT
yd = yd,1 T , yd,2 T , . . . , yd,P T , (33)

h iT
ym = ym,1 T , ym,2 T , . . . , ym,P T , (34)

h iT
yc = yc,1 T , yc,2 T , . . . , yc,P T . (35)

The signal power of the direct returns, multipath returns, clutter, and noise can be
given by

1 H
Pd = y y
L d d
1
= (Re{yd }2 + Im{yd }2 ), (36)
L
1 H
Pm = y y
L m m
1
= (Re{ym }2 + Im{ym }2 ), (37)
L
1 H
Pc = y y
L c c
1
= (Re{yc }2 + Im{yc }2 ), (38)
L
Pn = PNT Lσn2 . (39)

The loss function of the network is set to be the inverse of the objective function in the
above optimization problem and can be expressed as

1 Pc + Pn
loss = = . (40)
SINR Pd + Pm

3.2.4. Adam Optimizer

The optimizer is a tool to guide the neural network for parameter updates. After the
neural network achieves forward propagation once and calculates the loss function, the
optimizer is needed to perform backward propagation to achieve the update of the network
parameters. Adam optimizer is a classical deep learning optimizer based on gradient
descent algorithm, combining the ideas of momentum method and adaptive learning rate,
with the advantage of adaptively adjusting the learning rate to adapt to different data
and models.
The Adam optimizer computes first-order moment estimates and second-order mo-
ment estimates of the gradient in each iteration, thus updating the learning rate and
network parameters.
Remote Sens. 2023, 15, 2747 11 of 20

The gradient parameter of the loss function at the tth iteration with respect to the
network weights w can be expressed as

∂loss
g t = ∇loss = . (41)
∂wt

The first-order moments and second-order moments of the network weights in the tth
iteration can be expressed as

m t = β 1 · m t −1 + (1 − β 1 ) · g t , (42)

r t = β 2 · r t −1 + (1 − β 2 ) · g t gt, (43)
where β 1 and β 2 are two hyperparameters of the Adam optimizer, usually set to β 1 = 0.9
and β 2 = 0.999.
At the beginning of the iteration, there is a deviation of mt and r t to the initial value, so
it is necessary to correct the deviation of the first-order moments and second-order moments.
The corrected first-order moments and second-order moments can be expressed as
mt
m̂t = , (44)
1 − βt1

rt
r̂ t = . (45)
1 − βt2
We iterate over the network weights, and the network weight parameters can be
updated and represented as follows

m̂t
w t +1 = w t − √ , (46)
r̂ t + δ

where δ refers to a small constant used for numerical stabilization, usually set to δ = 10−8 .
We summarize the proposed algorithm in Algorithm 1.

Algorithm 1: Proposed algorithm.

Input: Random normalized phase sequence x I , learning rate of Adam γ, number
of iterations E.
Output: Desired waveform phase sequence xO .
Set e = 0, Adam learning rate set to γ > 0;
1: Construct forward propagation module according to Figure 2;
2: Input x I to the forward propagation module to obtain output xO ;
3: Compute yd , ym and yc according to (27)–(35);
4: Compute Pd , Pm , Pc and Pn according to (36)–(39), and the loss function is
constructed by (40);
5: Optimizing loss function with Adam optimizer;
6: If e = E, stop and output the result. Otherwise, update e, i.e., e = e + 1, and
back to the step 2.

3.2.5. Complexity Analysis

To evaluate the complexity of our proposed model, we measured its time complexity
using floating-point arithmetic, FLOPs, and its spatial complexity using parameter quanti-
ties, Params. Assuming that the input of the network is a sequence with a length of 320, the
complexity of the proposed model is shown in Table 1.
Remote Sens. 2023, 15, 2747 12 of 20

Table 1. The complexity of the proposed model.

Model FLOPs Params

The proposed model 786,432.0 396,416

4. Results and Discussion

In this section, a series of simulations were conducted to verify the effectiveness of the
proposed algorithm and the superiority of the designed waveform. All simulations were
analyzed on a PC with a i7-12700H CPU and a GTX3080 GPU and 16 GB RAM. The neural
network in this article was processed with Python 3.7 and pytorch 1.12 and ran on GPU.
The MIMO radar employed a transceiver co-array that consisted of a uniform line
array with half-wavelength spacing and a wavelength of λ. The simulation parameters
shown in Table 2 were used in the subsequent simulations.

Table 2. Simulation Parameters.

Parameters Value
Number of transmitting antennas NT 20
Number of receiving antennas NR 20
Number of snapshot L 16
Number of neurons J 128
CPI P 16
Carrier frequency f c 3 GHz
Target azimuth θd 20◦
Multipath azimuth θm −10◦
Clutter azimuth θc −5◦
θi 45◦
θv 10◦
Target velocity v 45 m/s
Relative delay for multipath returns lm 5
Relative delay for clutter lc 2
Specular reflection coefficient ρ 0.8e jπ/4
Signal-to-noise ratio SNR 20 dB
Interference-to-noise ratio INR 20 dB

4.1. Convergence
This section analyzes the convergence of the proposed algorithm. The learning rate
of the network was γ = 0.01, and the number of iterations was E = 1000. The number of
optimized signals was NT = 20, and the number of snapshots for each signal was L = 16.
As illustrated in Figure 3, the curve of the loss function gradually decreased as the
number of training iterations increased. Particularly, during the first 10 training sessions,
the loss function decreased rapidly. After 20 training sessions, the loss function started to
converge and eventually approached zero, thus indicating that the proposed algorithm can
converge to the global optimal solution.
Remote Sens. 2023, 15, 2747 13 of 20

Figure 3. Loss function curve.

4.2. SINR Performance

In this section, the learning rate of the network and the number of iterations were
γ = 0.01 and E = 1000. The number of optimized signals was NT = 20, and the number
of snapshots for each signal was L = 16. The interference-to-noise ratio was INR = 20 dB.
The set of SNR were [−20 dB, −15 dB, −10 dB, −5 dB, 0 dB, 5 dB, 10 dB, 15 dB, 20 dB].
To evaluate the SINR performance of the waveform designed by the proposed al-
gorithm, we compared it with the SINR performance of the waveforms designed using
the multipath suppression algorithm [4] and the CAN algorithm [34] at different SNR
conditions, as shown in Figure 4. It can be seen that the proposed algorithm achieved
a higher SINR at both low SNR and high SNR. The SINR performance of the proposed
algorithm was more than 2 dB higher than that of the multipath suppression algorithm
at different SNR conditions. Because the purpose of the CAN algorithm is to generate
orthogonal waveforms, the SINR performance of the generated waveform was much worse
than that of the proposed algorithm and the multipath suppression algorithm.

30

20

10

0
SINR/dB

-10

-20

-30 Proposed algorithm

Multipath suppression
Orthogonal waveform
-40
-20 -15 -10 -5 0 5 10 15 20
SNR/dB
Figure 4. Output SINR versus the SNR.
Remote Sens. 2023, 15, 2747 14 of 20

To further highlight the superiority of the SINR performance of the waveform gener-
ated by the proposed algorithm, we conducted a comparative analysis assuming
SNR = 20 dB and NT = 4. Specifically, we compared the SINR performance of the pro-
posed algorithm against an existing multipath exploitation algorithm [40] and a multipath
suppression algorithm under these conditions. As shown in Table 3, the SINR performance
of the proposed algorithm was higher than that of the multipath exploitation algorithm
and multipath suppression algorithm.

Table 3. SINR performance of the proposed algorithm, existing multipath exploitation algorithm,
and multipath suppression algorithm.

Methods Proposed Algorithm Multipath Exploitation Multipath Suppression

SINR/dB 25.64 23.18 20.64

4.3. Transmit Beampattern

In this section, we set the learning rate of the network to γ = 0.01, and we set the
number of iterations to E = 1000. The number of optimized signals was NT = 20, and the
number of snapshots for each signal was L = 16. The signal-to-noise ratio was SNR = 20 dB,
and the interference-to-noise ratio was INR = 20 dB. The azimuth ofthe target, multipath,
and clutter were θd = 20◦ , θm = −10◦ , and θc = −5◦ . The azimuth ranged from −90◦ to
90◦ with a grid point spacing of 1◦ .
Figure 5 shows a comparison between the transmit beampattern generated by the
proposed algorithm and the multipath suppression algorithm. As can be observed from
the figure, the transmit beampattern generated by the proposed algorithm forms peaked in
the target and multipath azimuths, while creating a deep notch in the clutter azimuth. In
contrast, the multipath suppression algorithm treated the multipath returns as clutter, thus
resulting in deep notches in both the multipath azimuth and clutter azimuth. It is evident
from the transmit beampattern generated by the proposed algorithm and the multipath
suppression algorithm that the multipath suppression algorithm achieved a suppression
level of approximately −40 dB for both multipath returns and clutter. On the other hand,
the proposed algorithm achieved a significantly superior suppression level of −70 dB for
clutter energy.

Figure 5. The transmit beampattern.

Remote Sens. 2023, 15, 2747 15 of 20

4.4. Detection Probability

In this section, we compared the detection performance of the waveforms generated
by the proposed algorithm and multipath suppression algorithm under different specular
reflection coefficients. The learning rate of the network was γ = 0.01, and the number of
iterations was E = 1000. The number of optimized signals was NT = 20, and the number
of snapshots for each signal was L = 16. The set of SNR were [−20 dB, −15 dB, −10 dB,
−5 dB, 0 dB, 5 dB, 10 dB, 15 dB, 20 dB], and the interference-to-noise ratio was INR = 20 dB.
The set of specular reflection coefficients included the values of 0.7e jπ/4 , 0.8e jπ/4 , and
0.9e jπ/4 .
The corresponding detection probabilities can be calculated by
!
√
r  
Pd = MarcQ 2SINR, −2 log Pf a , (47)

where MarcQ(·) denotes the Marcum-Q function [45], and Pf a denotes the false alarm
probability.
Assuming the false alarm probability Pf a = 10−6 , the simulation results are shown in
Figure 6. The proposed algorithm exhibited superior detection probability performance
compared to the multipath suppression algorithm under the same input of SNR, as is
evidenced by its higher Pd value.
Regarding the proposed algorithm for multipath exploitation, the detection probability
increased with the increase in multipath reflection intensity. In contrast, for the multipath
suppression algorithm, the detection probability was not significantly affected by the
multipath reflection intensity, as the multipath returns were treated as clutter and were
suppressed by the algorithm.

Figure 6. Comparisons of the detection probability.

To confirm the impact of the false alarm probability on the detection probability per-
formance of the proposed algorithm, we evaluated and compared the detection probability
performance of the proposed algorithm with existing and multipath exploitation algorithm
and a multipath suppression algorithm under different false alarm probabilities. We set
the signal-to-noise ratio to SNR = 20dB and the number of optimized signals to NT = 4.
As shown in Table 4, the detection probability of the proposed algorithm was higher than
Remote Sens. 2023, 15, 2747 16 of 20

that of the multipath exploitation algorithm and multipath suppression algorithm under
different false alarm probabilities.

Table 4. The detection probability performance under different false alarm probabilities.

Methods Proposed Algorithm Multipath Exploitation Multipath Suppression

Pf a = 10−7 0.9408 0.8872 0.7968
Pf a = 10−6 0.9765 0.9491 0.8952
Pf a = 10−5 0.9928 0.9820 0.9570

4.5. Effect of Initial Input

In this section, we set the learning rate of the network to γ = 0.01 and the number of
iterations to E = 1000. The number of optimized signals was NT = 20, and the number of
snapshots for each signal was L = 16. The signal-to-noise ratio was SNR = 20 dB, and the
interference-to-noise ratio was INR = 20 dB.
We compared the SINR performance of the proposed algorithm using two different
initial inputs: a random normalized phase sequence and a sequence optimized by the CAN
algorithm. As shown in Figure 7, the proposed algorithm was insensitive to the initial input.
For both cases, where the initial input was a random normalized sequence and a sequence
optimized by the CAN algorithm, the SINR performance of the proposed algorithm had
an acceptable error at the beginning of training and tended to become consistent with the
increase in training iterations.

Figure 7. SINR of different initial inputs.

4.6. Effect of Number of Transmitting Antennas

In this section, we compared the SINR performance of the proposed algorithm with
different numbers of transmit antennas. We set the learning rate of the network to γ = 0.01,
and the number of iterations to E = 1000. The number of snapshots for each signal was
L = 16. The signal-to-noise ratio was SNR = 20 dB, and the interference-to-noise ratio was
INR = 20 dB. The set of the number of transmit antennas were [4, 8, 12, 16, 20, 24, 28].
As shown in Figure 8, the SINR performance of the proposed algorithm increased
with the number of transmit array antennas NT . This is because the increase in the number
of transmit antennas provided higher degrees of freedom (DOFs), which enabled the
Remote Sens. 2023, 15, 2747 17 of 20

algorithm to exploit the multipath reflections better and achieve a higher SINR. This
implies that augmenting the number of transmit antennas in the proposed algorithm leads
to a better detection performance of the MIMO radar. Nonetheless, the increase in the
number of transmit antennas may also escalate the cost and computational complexity of
MIMO radar systems, thereby necessitating a comprehensive consideration in choosing an
appropriate number of transmit antennas for practical applications.

33

32

31

30
SINR/dB

29

28

27

26

25
5 10 15 20 25
Number of transmitting antennas

Figure 8. SINR with different numbers of transmitting antennas.

4.7. The Phase of the Waveform

In this section, we plotted the phase of the waveform optimized by the proposed
algorithm under different numbers of transmit antennas. We set the learning rate of the
network to γ = 0.01 and the number of iterations to E = 1000. The number of snapshots for
each signal was L = 16. The signal-to-noise ratio was SNR = 20 dB, and the interference-
to-noise ratio was INR = 20 dB.
As shown in Figure 9, the phase of the waveform optimized by the proposed algo-
rithm varied between 0 and 2π for different numbers of transmit antennas. This result
demonstrates the effectiveness of the algorithm in optimizing the transmission waveform.
7 7

6 6

5 5
Phase/rad

Phase/rad

4 4

3 3

2 2

1 1

0 0
0 10 20 30 40 50 60 70 0 20 40 60 80 100 120 140
Waveform index Waveform index
(a) (b)

Figure 9. Cont.
Remote Sens. 2023, 15, 2747 18 of 20

7
7

6
6

5 5

Phase/rad
4
Phase/rad

3 3

2 2

1 1

0 0
0 50 100 150 200 0 50 100 150 200 250
Waveform index Waveform index
(c) (d)

Figure 9. Phase of the optimized waveform with different numbers of transmit antennas. (a) NT = 4,
(b) NT = 8, (c) NT = 12, (d) NT = 16.

5. Conclusions
In conclusion, this paper presents a novel approach for designing transmit waveforms
for multipath exploitation using deep learning. We established the MIMO radar signal
model for scenarios where a multipath exists and proposes an optimization objective that
maximizes the receiver’s SINR with a CM constraint. To solve this non-convex problem,
we developed a deep residual optimization training network that directly and optimally
solved the problem without relaxation.
Our simulation results demonstrate that the proposed algorithm effectively utilized
the energy of multipath reflections, thus outperforming existing methods in terms of SINR
performance and detection probability. We also showed that the proposed algorithm was
robust to different initial inputs and benefited from the increased degrees of freedom
provided by additional transmit antennas.
Overall, our research has important implications for MIMO radar applications, where
optimizing waveform design can significantly impact the accuracy and reliability of target
detection. Future work could explore the generalization of our approach to more complex
scenarios, such as those with unknown multipath information or multi-target situations.

Author Contributions: Conceptualization, Z.Z. and Y.Z.; methodology, Y.Z.; software, Z.Z.; vali-
dation, Z.Z., Y.Z. and X.P.; formal analysis, Z.Z.; investigation, Z.Z.; resources, Y.Z.; data curation,
Z.Z.; writing—original draft preparation, Z.Z.; writing—review and editing, Z.Z., X.P., H.X. and Y.Z.;
visualization, J.M., J.C. and Y.S.; supervision, Y.Z.; project administration, Y.Z.; funding acquisition,
Y.Z. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China under
Grant U2133216.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Barton, D.K. Low-angle radar tracking. Proc. IEEE 1974, 62, 687–704. [CrossRef]
2. White, W.D. Low-Angle Radar Tracking in the Presence of Multipath. IEEE Trans. Aerosp. Electron. Syst. 1974, 10, 835–852.
[CrossRef]
3. Bar-Shalom, Y.; Kumar, A.; Blair, W.D.; Groves, G.W. Tracking low elevation targets in the presence of multipath propagation.
IEEE Trans. Aerosp. Electron. Syst. 1994, 30, 973–979. [CrossRef]
Remote Sens. 2023, 15, 2747 19 of 20

4. Hickman, G.; Krolik, J.L. MIMO GMTI radar with multipath clutter suppression. In Proceedings of the 2010 IEEE Sensor Array
and Multichannel Signal Processing Workshop, Jerusalem, Israel, 4 October 2010; pp. 65–68.
5. Xin, J.; Sane, A. Linear prediction approach to direction estimation of cyclostationary signals in multipath environment. IEEE
Trans. Signal Process. 2001, 49, 710–720. [CrossRef]
6. Xin, J.; Sane, A. Direction estimation of coherent narrowband signals using spatial signature. In Proceedings of the 2002 IEEE
Sensor Array and Multichannel Signal Processing Workshop, Rosslyn, VA, USA, 6 August 2002; pp. 523–527.
7. Yu, J.; Krolik, J. MIMO adaptive beamforming for nonseparable multipath clutter mitigation. IEEE Trans. Aerosp. Electron. Syst.
2014, 50, 2604–2618. [CrossRef]
8. Aubry, A.; De Maio, A.; Foglia, G.; Orlando, D. Diffuse Multipath Exploitation for Adaptive Radar Detection. IEEE Trans. Signal
Process. 2015, 63, 1268–1281. [CrossRef]
9. Hayvaci, H.T.; De Maio, A.; Erricolo, D. Improved detection probability of a radar target in the presence of multipath with prior
knowledge of the environment. IET Radar Sonar Navig. 2013, 7, 36–46. [CrossRef]
10. Rong, Y.; Aubry, A.; De Maio, A.; Tang, M. Automatically tunable AMF for radar detection in diffuse multipath. In Proceedings of
the 2020 IEEE Sensor Array and Multichannel Signal Processing Workshop, Hangzhou, China, 8 June 2020; pp. 1–5.
11. Yilmaz, S.H.G.; Hayvaci, H.T. Multipath exploitation radar with adaptive detection in partially homogeneous environments. IET
Radar Sonar Navig. 2020, 14, 1475–1482. [CrossRef]
12. Li, J.; Stoica, P. MIMO Radar with Colocated Antennas. IEEE Signal Process. Mag. 2007, 24, 106–114. [CrossRef]
13. Fishler, E.; Haimovich, A.; Blum, R.; Chizhik, D.; Cimini, L.; Valenzuela, R. MIMO radar: An idea whose time has come. In
Proceedings of the 2004 IEEE Radar Conference, Philadelphia, PA, USA, 29 April 2004; pp. 71–78.
14. Sun, H.; Brigui, F.; Lesturgie, M. Analysis and comparison of MIMO radar waveforms. In Proceedings of the 2014 International
Radar Conference, Lille, France, 13 October 2014; pp. 1–6.
15. Stoica, P.; Li, J.; Xie, Y. On Probing Signal Design For MIMO Radar. IEEE Trans. Signal Process. 2007, 55, 4151–4161. [CrossRef]
16. Stoica, P.; Li, J.; Xu, L.; Roberts, W. On Parameter Identifiability of MIMO Radar. IEEE Signal Process. Lett. 2007, 14, 968–971.
17. Yu, X.; Qiu, H.; Yang, J.; Wei, W.; Cui, G.; Kong, L. Multi-spectrally constrained MIMO radar beampattern design via sequential
convex approximation. IEEE Trans. Aerosp. Electron. Syst. 2022, 58, 2935–2949. [CrossRef]
18. Raei, E.; Alaee-Kerahroodi, M.; Shankar, M.B. Spatial-and range-ISLR trade-off in MIMO radar via waveform correlation
optimization. IEEE Trans. Signal Process. 2021, 69, 3283–3298. [CrossRef]
19. Qian, J.Y.; Zheng, G.X.; Saleem, A. Channel modeling based on multilayer artificial neural network in metro tunnel environments.
ETRI J. 2022, 1–13. [CrossRef]
20. Jia, M.H.; Zheng, G.X.; Ji, W.L. A new model for predicting the characteristic of RF propagation in rectangular tunnel. In
Proceedings of the 2008 IEEE International Conference on Personal Wireless Communications Conference Proceedings, Dalian,
China, 10 September 2008; pp. 1–4.
21. Kermani, M.H.; Kamarei, M. A ray-tracing method for predicting delay spread in tunnel environments. In Proceedings of
the 2000 IEEE International Conference on Personal Wireless Communications Conference Proceedings, Hyderabad, India,
17 December 2000; pp. 538–542.
22. Zhao, X.W.; Du, F.; Geng, S.Y.; Sun, N.Y.; Zhang, Y.; Fu, Z.H.; Wang, G.J. Neural network and GBSM based time-varying and
stochastic channel modeling for 5G millimeter wave communications. China Commun. 2019, 16, 80–90. [CrossRef]
23. Sun, N.; Geng, S.; Li, S.; Zhao, X.; Wang, M.; Sun, S. Channel modeling by RBF neural networks for 5G Mm-wave communication.
In Proceedings of the 2018 IEEE/CIC International Conference on Communications in China, Beijing, China, 12 August 2018;
pp. 768–772.
24. Ertel, R.B.; Reed, J.H. Angle and time of arrival statistics for circular and elliptical scattering model. IEEE J. Sel. Areas Commun.
1999, 17, 1829–1840. [CrossRef]
25. Petrus, P.; Rappaport, T.S. Geometrical-based statistical macrocell channel model for mobile environments. IEEE Trans. Commun.
2002, 50, 495–502. [CrossRef]
26. Li, J.; Guerci, J.R.; Xu, L. Signal Waveform’s Optimal Under Restriction Design for Active Sensing. In Proceedings of the 2006
IEEE Sensor Array and Multichannel Signal Processing Workshop, Waltham, MA, USA, 12 July 2006; pp. 382–386.
27. Tang, B.; Li, J.; Liang, J. Alternating direction method of multipliers for radar waveform design in spectrally crowded environments.
Signal Process. 2018, 142, 398–402. [CrossRef]
28. Yu, X.; Cui, G.; Kong, L.; Li, J.; Gui, G. Constrained Waveform Design for Colocated MIMO Radar With Uncertain Steering
Matrices. IEEE Trans. Aerosp. Electron. Syst. 2019, 55, 356–370. [CrossRef]
29. Aubry, A.; De Maio, A.; Piezzo, M.; Farina, A. Radar waveform design in a spectrally crowded environment via nonconvex
quadratic optimization. IEEE Trans. Aerosp. Electron. Syst. 2014, 50, 1138–1152. [CrossRef]
30. Yang, Y.; Blum, R.S. MIMO radar waveform design based on mutual information and minimum mean-square error estimation.
IEEE Trans. Aerosp. Electron. Syst. 2007, 43, 330–343. [CrossRef]
31. Naghsh, M.M.; Modarres-Hashemi, M.; Shahbazpanahi, S. Unified Optimization Framework for Multi-Static Radar Code Design
Using Information-Theoretic Criteria. IEEE Trans. Signal Process. 2013, 61, 5401–5416. [CrossRef]
32. Yu, X.; Cui, G.; Zhang, T. Constrained Transmit Beampattern Design for Colocated MIMO Radar. Signal Process. 2017, 144,
145–154. [CrossRef]
Remote Sens. 2023, 15, 2747 20 of 20

33. Wang, Y.C.; Wang, X.; Liu, H. On the Design of Constant Modulus Probing Signals for MIMO Radar. IEEE Trans. Signal Process.
2012, 60, 4432–4438. [CrossRef]
34. Stoica, P.; He, H.; Li, J. New Algorithms for Designing Unimodular Sequences with Good Correlation Properties. IEEE Trans.
Signal Process. 2009, 57, 1415–1425. [CrossRef]
35. He, H.; Stoica, P.; Li, J. Designing Unimodular Sequence Sets with Good Correlations—Including an Application to MIMO Radar.
IEEE Trans. Signal Process. 2009, 57, 4391–4405. [CrossRef]
36. Li, Y.; Vorobyov, S.A. Fast Algorithms for Designing Unimodular Waveform(s) with Good Correlation Properties. IEEE Trans.
Signal Process. 2018, 66, 1197–1212. [CrossRef]
37. Sen, S.; Hurtado, M.; Nehorai, A. Adaptive OFDM radar for detecting a moving target in urban scenarios. In Proceedings of the
2009 International Waveform Diversity and Design Conference, Kissimmee, FL, USA, 8 February 2009; pp. 268–272.
38. Sen, S.; Nehorai, A. OFDM MIMO radar with mutual-information waveform design for low-grazing angle tracking. IEEE Trans.
Signal Process. 2010, 58, 3152–3162. [CrossRef]
39. Sen, S.; Nehorai, A. Adaptive OFDM radar for target detection in multipath scenarios. IEEE Trans. Signal Process. 2011, 59, 78–90.
[CrossRef]
40. Xu, Z.; Fan, C.; Huang, X. MIMO Radar Waveform Design for Multipath Exploitation. IEEE Trans. Signal Process. 2021, 69,
5359–5371. [CrossRef]
41. Fan,C.; Xie, Z.; Wang, J.; Xu, Z.; Huang, X. Robust MIMO Waveform Design in the Presence of Unknown Mutipath Return. Remote
Sens. 2023, 14, 4356. [CrossRef]
42. Imani, S.; Ghorashi, S.A. Sequential quasi-convex-based algorithm for waveform design in colocated multiple-input multiple-
output radars. IET Signal Process. 2023, 10, 309–317. [CrossRef]
43. Hertz, J.A. Introduction to the Theory of Neural Computation; CRC Press: Boca Raton, FL, USA, 2018.
44. He, K.; Zhang, X.; Ren, S.; Sun, J. Deep Residual Learning for Image Recognition. In Proceedings of the 2016 IEEE Conference on
Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27 June 2016; pp. 770–778.
45. Richards, M.A. Fundamentals of Radar Signal Processing; McGraw Hill: New York, NY, USA, 2005.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.