Received 7 April 2024, accepted 29 April 2024, date of publication 6 May 2024, date of current version 20 May 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3397165

Bandwidth Allocation and Power Control

Optimization for Multi-UAVs
Enabled 6G Network
MOHAMMAD AHMED ALNAKHLI 1 , (Member, IEEE),
EHAB MAHMOUD MOHAMED 1 , (Member, IEEE), AND
MOSTAFA M. FOUDA 2,3 , (Senior Member, IEEE)
1 Department of Electrical Engineering, College of Engineering in Wadi Addwasir, Prince Sattam bin Abdulaziz University, Wadi ad-Dawasir 11991, Saudi Arabia
2 Department of Electrical and Computer Engineering, Idaho State University, Pocatello, ID 83209, USA
3 Center for Advanced Energy Studies (CAES), Idaho Falls, ID 83401, USA

Corresponding author: Mohammad Ahmed Alnakhli ([email protected])

This work was supported by Prince Sattam bin Abdulaziz University under Project PSAU/2024/R/1445.

ABSTRACT In this paper, we undertake an examination of the complicated challenges associated with
bandwidth allocation and power control in multi- unmanned aerial vehicles (UAVs) network. The focal point
of our investigation is the formulation and subsequent proposition of a novel algorithm aimed at optimizing
bandwidth utilization and energy efficiency across multiple UAVs. The intricate joint optimization problem
is cast as a nontrivial nonlinear programming (NLP) challenge; thus, to efficiently address it, we split it into
two subproblems. In the first sub-problem, UAVs bandwidth resources are being optimized while considering
that they are operating at their maximum transmit powers. In the second sub-problem and after optimizing
bandwidth allocations, UAVs transmit powers are optimized based on their inter-links speed. Sequential
quadratic programing (SQP) based on coloring graph representation is proposed to address the first sub-
problem, while M- matrix theory is employed to address the second one. Rigorous numerical simulations
are conducted to prove the effectiveness of the proposed scheme against other benchmarks in maximizing
both data rate and energy efficiency of the proposed multi-UAV network.

INDEX TERMS 6G, multi-UAVs, spectrum efficiency, energy efficiency, SQP, M-matrix.

I. INTRODUCTION natural disasters, emergency response, rescue and search,

Unmanned aerial vehicles (UAVs) are expected to play information dissemination, and data collection from the inter-
a significant role in the beyond fifth generation (B5G) net of things (IoTs) devices, and mounting reconfigurable
and six generation (6G) wireless networks in the next intelligent surface (RISs) boards [4], [5]. However, the incor-
decade [1]. The deployment of UAVs is ever-increasing, poration of UAVs into 6G networks calls for a paradigm
and the number of commercial UAV fleets can reach up shift in the design of both cellular and UAV communi-
to 1.6 million by 2024 [2]. UAVs can be dispatched as cation systems due to the high altitude and mobility of
high-mobility aerial communication platforms to provide UAVs, the unique channel characteristics of UAV-ground
opportunistic line-of-sight (LoS) links and further assist links, the asymmetric quality of downlink and uplink data
terrestrial communications in 6G [3]. The integration of transmission, the stringent constraints imposed by the size,
UAVs into 6G networks has numerous use cases, including weight, power limitations of UAVs, and the intra-system
base station (BS) offloading, swift service recovery after and inter-system interference of the integrated networks [6].
Key challenges associated with the widespread commercial
The associate editor coordinating the review of this manuscript and use of UAVs towards 6G include safety, traffic control, and
approving it for publication was Xijun Wang. security aspects [2]. Moreover, researchers are interested in

2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
VOLUME 12, 2024 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 67405
 M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

developing green UAV communication with low power con- control [14]. Its flexibility allows modeling of diverse
sumption to prolong UAVs’ lifetime [4], [7]. system configurations, optimizing transmit powers of
Multi-UAV is a promising solution for coverage area multi-UAVs while maintaining minimum link qualities.
problems, especially for 6G networks. The scalability of • Numerical analyses are conducted to substantiate the
multi-UAV systems aligns seamlessly with the expansive efficiency of the proposed approach, revealing substan-
nature of 6G applications, making them well-suited for tial improvements in both data rate and energy efficien-
urban environments, remote areas, and critical infrastruc- cies when compared to other benchmark schemes.
ture inspection. Through coordinated multi-UAVs, these The subsequent sections of this paper are organized as
UAVs contribute to optimizing coverage, resource utiliza- follows: Section II provides an overview of related works,
tion, and network reliability, thereby positioning themselves Section III introduces the system model and formulates the
as integral components in the realization of resilient and problem, Section IV outlines the proposed approach for band-
high-performing communication networks in the era of 6G. width allocation and power control optimization, Section V
In this context, researchers have proposed various frame- discusses the results and offers a detailed analysis, and,
works for persistent multi-UAV coverage with global decon- finally, conclusions in the last section, summarizes the key
fliction such as UAV gateway selection given in [8]. findings and implications of the proposed approach.
In this paper, we will tackle the problem of optimal
bandwidth and power allocations in multi-UAV enabled 6G II. RELATED WORKS
network. In this regard, we will make use of graph theory [9] Recently, there was a significant development in multi-UAVs
for bandwidth optimization. Besides bandwidth allocation, research in wireless communication networks, focusing on
efficient power control in multi-UAV scenarios is essential for various applications and key contributions. The highlighted
overall system performance, interference management, and studies include topics such as multi-UAV enabled mobile-
resource utilization, considering the limited battery capacities edge computing for IoT networks [15], cooperative control
of UAVs [7]. Thus, in this paper, M-matrix theory will be strategies for multi-UAV formation keeping [16], federated
utilized as an efficient tool for UAVs power control. Thus, learning-assisted multi-UAV networks in image classification
the main contributions of this paper can be summarized as scenarios [17], multi-agent deep reinforcement learning for
follows: trajectory optimization [18], joint communication and trajec-
• A two-step algorithmic approach will be proposed to tory optimization for mobile internet of vehicles (IoVs) [19],
address the problem of bandwidth allocation and power efficient and secure multi-UAV communication via a secured
control in multi-UAV networks. In the first step, band- UAV (S-UAV) model [20], and a clustering strategy to
width allocations are optimized while assuming UAVs enhance security in multi-UAV flight formations [21]. These
are transmitting at their maximum power. In the second works collectively showcase advancements in optimizing
step, UAVs transmit powers are optimized based on their communication, coordination, security, and computational
inter-link speeds result from their optimized bandwidth efficiency in the evolving field of multi-UAV systems.
allocations done in the first step. For multi-UAV spectrum allocation and power control
• Graph theory in the form of coloring graph will be optimization, which is the main focus of this paper, there
proposed for UAVs bandwidth allocation optimization. are some research works studied them in [17], [18], [19],
In this regard, graph theory will be utilized as an effi- and [20]. The authors in [17] studied multi-UAVs and cellu-
cient tool for modeling and analyzing connectivity and lar base stations (BSs) within IoT framework, emphasizing
spectrum allocation patterns of the multi-UAV network uplink channel modeling. Introducing a binary exponen-
within a given area. tial power control algorithm tailored for 5G networked
• For addressing the optimization intricacies inherent in UAV transmitters, the research considered 3D distance and
the proposed graph coloring, sequential quadratic pro- multi-UAV reflections in the channel model. Simulations
gramming (SQP) [10] will be proposed to find the assessed the algorithm’s impact on coverage probability,
sub-optimal solution iteratively. SQP is chosen for its spectrum efficiency, and energy efficiency under various 3D
iterative optimization approach, which is well-suited for distances. While this work prioritized power control ini-
nonlinear programming problems (NLP), accuracy, and tially and subsequently improved related parameters, our
stability [11], [12], [13]. SQP has been applied in various approach focuses on maximizing spectrum allocation first,
research in wireless communication systems for small- followed by power control for multi-UAV networks. This
cell networks [11], cognitive radio systems [12], V2V strategy ensures UAVs receive the necessary data rate
communication [13], etc. The algorithm’s convergence without unnecessary losses, subsequently contributing to
and efficiency make it a good choice for optimizing spectrum utilization improvement. This step strategy not
spectrum resources among the interconnected UAVs, only ensures efficient data transmission but also mini-
contributing to improved spectrum efficiency. mizes the required power of UAVs, a facet not explored
• For UAVs power control, M-matrix theory, known for in the referenced work. The authors in [18] conducted a
its optimality and computational efficiency, will be uti- study aimed at enhancing resource allocation and addressing
lized as a powerful method for effective UAVs power power control challenges within multi-UAV formation flight

67406 VOLUME 12, 2024

M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

FIGURE 1. Provide an overview of a versatile multi-UAV network system, illustrating its capability to simultaneously engage in
diverse applications. The network is adeptly controlled by ground control systems, showcasing the flexibility and adaptability of
the multi-UAV setup. This integrated system has the potential to serve various purposes concurrently, demonstrating its efficiency
in managing different tasks. The network’s responsiveness to ground control systems ensures seamless coordination, making it a
dynamic and multifunctional solution for a wide range of applications, all within a unified and controllable framework.

system engaged in radar sensing. It formulates a complex spectral efficiency (SE) improvements compared to current
optimization problem to maximize the minimum signal-to- small cell (SC) and cellular massive MIMO systems. How-
interference-plus-noise ratio (SINR) of radar echo signals, ever, this work neither considered bandwidth allocation nor
demonstrating its NP-hardness nature. The proposed iterative power control in multi-UAV networks. In [23], the utilization
channel allocation and power control algorithm (ICAPCA), of cell-free mMIMO systems in mobile communications was
coupled with a reduced-complexity greedy channel alloca- examined, addressing deployment architectures, challenges,
tion algorithm (GCAA), effectively addressed this problem, and innovative solutions such as predictor antennas, hier-
significantly enhancing minimum SINR and radar sensing archical cancellation, rate-splitting, and dynamic clustering.
performance compared to conventional methods. In con- It also outlines future research directions to advance mobile
trast, our proposed approach aims at maximizing bandwidth cell-free massive MIMO communication technologies. Also,
allocations for all multi-UAVs, contributing to the com- the problem of multi-UAV networking was not considered.
prehensive optimization of spectrum utilization. Notably, Consequently, the existing literature on multi-UAV net-
our work operates in 3D space, providing more accurate work exhibit strengths, whereas faces limitations that the
distance considerations, and differs in UAV distribution, proposed method aims to overcome. The fact is that most of
influencing SINR values. The authors in [19] proposed a these studies considered a sequential optimization approach,
hierarchical multi-agent Q-learning framework for optimal addressing energy consumption first and then enhancing its
frequency reuse and transmit power control in multi-UAV related parameters, which leads to suboptimal outcomes.
wireless networks, aiming to improve spectrum efficiency On the other hand, our proposed approach prioritizes max-
and equipment utilization. However, our work distinction imizing spectrum allocations before refining power control.
lies in its comprehensive approach to jointly optimizing Thus, our proposal addresses these gaps by providing a
channel allocation and transmit power for enhanced net- detailed and comprehensive strategy for joint spectrum
work efficiency. The authors in [20] worked on meeting allocation and power control optimization in the multi-UAV
the requirements of high average spectral efficiency and networks.
maximizing the minimum average energy in multi-UAV for-
mation communication, focusing on resource optimization III. PROPOSED SYSTEM MODEL
using Deep Q-network. The paper emphasized the intelligent Figure 1 shows the proposed multi-UAV network archi-
choice of power and frequency by UAVs based on remaining tecture, where multiple UAVs are deployed to provide
time and the number of UAVs in the formation. However, communication services to users in remote and hard-to reach
it does not explicitly highlight specific details or optimization areas. The system model consists of several UAVs hovering
techniques related to power control. In [22], using cell-free at an altitude of h, with a maximum number of M UAVs
massive multiple-input multiple-output (mMIMO) for UAV represented as (v1 , v2 , v3 , . . . , vM ). In this scenario, UAV
communications, wireless power transfer (WPT) was incor- channel allocation and power control are highly needed to
porated to support both uplink data and pilot transmission. maximize both spectrum and energy efficiencies of the over-
The research explores novel closed-form expressions consid- all network based on multi-UAV connection and data traffic in
ering UAV hardware impairments, showing significant uplink the UAV-to-UAV links. In this section, UAV-to-UAV channel

VOLUME 12, 2024 67407

 M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

multi-UAV networks. This SINR, denoted as SINRij , is writ-

ten as:
Pr ij
SINRij = PM (4)
v=1,v̸ =i Pr vj + ρ0
where Pr vj , is Rx powerP at UAV j from  Tx UAV v where
M
v ̸= i. In (4), the term v=1,v̸ =i Pr vj represents the sum
of interference powers, while ρ0 represents the additive white
gaussian noise (AWGN) power. Subsequently, the spectrum
FIGURE 2. Provides a succinct visualization of a fully connected multi-UAV
efficiency Rij in bps/Hz is calculated as follows:
network system represented as a graph G(V , E). Each UAV V is depicted

as a vertex in the graph, and edges E signify direct communication links Rij = log2 1 + SINRij (5)
among UAVs, forming a comprehensive and interconnected network
structure. This illustration offers a quick and clear overview of the
extensive communication possibilities within the multi-UAV system.
B. OPTIMIZATION PROBLEM FORMULATION
In this section, we will formulate the optimization problem of
spectrum and power allocation for multi-UAV networks. The
model and the optimization problem formulation will be aim of the optimization problem is to maximize the overall
given. data rate of the pre-connected multi-UAV network subject
to the limited total bandwidth available for the network and
A. UAV TO UAV CHANNEL MODEL UAVs maximum Tx powers, as follows:
In the proposed model, the locations of UAVs are denoted by XM XM
max ηij Bij Rij (6)
xi , yi and zi ∈ R+ , where each UAV i, 1 ≤ i ≤ M is spatially Bij Pti i=1 j=1
defined by its coordinates within the 3D space, represented s.t. ηij ∈ (0, 1) (6.a)
by Fig. 2. The separation distance dij between UAV i and XM XM
UAV j is determined based on their 3D inter-distances as ηij Bij = Btot (6.b)
i=1 j=1
follows [15]:
r Pti ∈ (0, Ptmax ) (6.c)
 2  2
dij = dijHr + dijVt (1) where ηij is the connection index which is equal to one if UAV
i is connected to UAV j and zero otherwise. In this paper,
where dijHr is the horizontal distance between the projected ηij is assumed to be pre-assigned based on the multi-UAV net-
locations of UAV i and UAV j. Also, dijVt is defined as work architecture. The optimization of ηij based on a certain
the vertical distance between their projected locations. For objective function is out of the scope of this paper. Bij , is the
UAV-to-UAV channel modeling, the Air–to–Air (A2A) chan- bandwidth allocated for UAV i and UAV j communication
nel modeling given in [16] will be utilized in this paper. link, and Btot is the total available bandwidth for the multi-
In this modeling, only line-of-sight (LoS) links, characterized UAV network. The second constraint given in (6.b) means
by an unobstructed path between the transmitter (Tx) and that the sum of all bandwidths for all UAV-UAV links in the
receiver (Rx), are a notable type [16], while non-LoS (NLoS) network should be equal to Btot . The third constraint given
links rarely occurred and were not considered. Thus, the path in (6.c) indicates that the Tx power of UAV i should be
loss of the LoS in the link Lij between UAV i and UAV j is bounded by its maximum Tx power Ptmax .
modeled as follows:
IV. PROPOSED APPROACH FOR BANDWIDTH
 
dij
PLijLoS [dB] = PL (d0 ) + 10αlog10 , (2) ALLOCATION AND POWER CONTROL
d0
The optimization problem given in (6) is an NLP due to
where PL LoS
ij depends on the separation distance dij , its value its non-linear objective function. In this paper, to efficiently
at a reference distance d0 , i.e., PL (d0 ), and the path loss address it within its constraints, we will split it into two
exponent α. The parameter d0 is crucial in characterizing sub-problems. In the first sub-problem, Bij optimization is
how the signal strength attenuates with increasing distance conducted while considering all UAVs transmit at their max-
between the Tx and Rx in the communication link. The Tx imum Tx power, i.e., Pti = Ptmax , while in the second
power of a UAV, denoted as Pti , signifies the intensity of the sub-problem, Pti s are adjusted based on the pre-adjusted Bij .
signal it emits during communication. The Rx signal power,
denoted as Prj in dB, is determined by adding Pti in dB with A. GRAPH BASED BANDWIDTH OPTIMIZATION
the path loss PLijLoS as follows. Herein, graph theory using coloring graph is used to opti-
mize the bandwidth allocations of the multi-UAV network
Pr ij [dB] = Pti [dB] − PLijLoS [dB], (3)
while considering fixed maximum power allocation. In the
SINR between Rx UAV j and Tx UAV i is crucial for context of a multi-UAV network shown in Fig. 2, it can be
evaluating the communication quality in the context of formally represented as a graph G(V , E) where V denotes

67408 VOLUME 12, 2024

M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

The initial step involves defining constraints, particularly

those related to frequency reuse, with quantitative preci-
sion. To do that, we refer to a crucial definition from graph
theory:
Definition 1: An independent set in a graph Ĝ(V̂ , Ê) con-
stitutes a subset of the vertex set V̂ such that no two vertices
in V share an edge between them. A maximal independent
set (MIS) of Ĝ is an independent set that is not a proper subset
of another independent set in Ĝ [24].
The MIS, in the context of Ĝ, identifies sets of vertices
that can effectively reuse the same bandwidth. Suppose that
FIGURE 3. Provides an overview of a multi-UAV network system as
graph Ĝ contains Q such MIS. Let an MIS represents a
corresponding link graph theory Ĝ(V̂ , Ê ). The number of nodes in graph matrix of size (Q × K ) that systematically conveys infor-
Ĝ(V̂ , Ê ) corresponds to the number of links in the original graph G(V , E). mation about each vertex’s membership in various MISs
as follows:
the set of UAVs, and E represents the set of links among 
s11 s12 · · · s1K

them. Correspondingly, the link graph Ĝ(V̂ , Ê) is derived  s21 s22 · · · s2K 
based on the UAV entities and their connections as shown MIS =  .. .. . . ..  (9)
 
in Fig. 3. In this link graph, the edges in G(V , E) given  . . . . 
in Fig. 2, i.e., the channel links among UAVs, become the sQ1 sQ2 · · · sQK
vertices in Ĝ(V̂ , Ê), and the vertices in G(V , E), i.e., UAVs
where, the elements of the MIS matrix, sqk , indicates whether
nodes, become the edges in Ĝ(V̂ , Ê). Thus, the number of
vertex k belong to MIS q or not, as follows:
nodes in the link graph Ĝ(V̂ , Ê) in Fig. 3, matches to the
number of links in the original graph G(V , E) in Fig. 2.
(
1 Vertix k belongs to MIS q
This relationship between the number of nodes and links sqk = (10)
0 otherwise
underscores the significance of the link graph representa-
tion in providing a condensed yet comprehensive view of Let the variable Xq denotes the bandwidth assigned to all ver-
the network connectivity. By leveraging link graph theory, tices within the MIS q. This representation significant is that
we can simplify the visualization and analysis of network all vertices belonging to a specific MIS can collectively uti-
properties and optimization algorithms, ultimately facilitat- lize the same allocated bandwidth. By considering MISs and
ing a deeper understanding of the multi-UAV network system. constant power allocation, the optimization problem given
In the overall link graph representation Ĝ(V̂ , Ê) of the multi- in (6) can be re-written as follows:
UAV network, let ηk(ij) ∈ (0, 1) indicates the connection XK XQ
index of vertex k in Ĝ between UAVs i and j in graph G. Also, max Xq sqk (11)
Xq k=1 q=1
let Rk(ij) represents its associated spectrum efficiency when
s.t. ηk ∈ (0, 1) ∀i, j ∈ k (11.a)
UAV i transmits to UAV j in graph G. For example, vertex one
skq ∈ (0, 1)
 
in Ĝ, i.e., k = 1, is used to connect UAV 1 and UAV 2 in G, see (11.b)
Figs. 2 and 3. Thus, the total spectrum efficiency of vertex k gk ≤ 1, ∀k (11.c)
will be: XK
gk = 1 (11.d)
k=1
Rk = Rk(ij) + Rk(ji) (7) XQ
Xq = Btot (11.e)
Also, let gk be the percentage of the system’s traffic q=1
XQ
injected by vertex k assuming UAVs have full buffer traffic Xq skq ≥ Btot gk , ∀ηk = 1 (11.f)
q=1
loads. Thus, gk is only influenced by the spectrum efficiency
of link k, and can be evaluated as follows: This optimization problem aims to maximize the weighted
Rk sum of MISs bandwidth allocations Xq subject to various
gk = PK (8) constraints. The first constraint given in (11.a) is the same
l=1 Rl as that given in (6.a), which indicates the connection sta-
In this scenario, Btot needs to be collectively shared among all tus of vertex k connecting UAV i and UAV j. The second
vertices in graph Ĝ. In G, edges that do not interfere with each constraint given in (11.b) indicates that the element skq can
other have the capability to reuse the same bandwidth [21]. be either 0 or 1 based on the MIS structure. The third and
Consequently, the corresponding vertices in Ĝ can efficiently the fourth constraints given in (11.c) and (11.d) respectively,
share and reuse the same bandwidth. Based on graph repre- bound the value of gk to be less than or equal to 1, and
sentation, achieving optimal spectrum efficiency necessitates their summation ∀k should be equal 1. The fifth constraint
the formulation of a suitable optimization problem, encom- given in (11.e) indicates that the total bandwidths assigned
passing an appropriate objective function and constraints. to all MISs should be equal to Btot . The sixth constraint

VOLUME 12, 2024 67409

 M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

given in (11.f) indicates that the bandwidth assigned to an Algorithm 1 Proposed SQP for X ∗q Optimization
established vertex k from all its designated MISs should be Output:X ∗q
greater than or equal to its percentage of the total injected Input: sqk , Btot , ηk , gk , Nit , and δ
system traffic. Initialize: X 0q ← initialguessatω = 0
To solve the NLP problem given in (11), nonlinear opti- While ω ≤ Nit      
mization techniques such as SQP method can be utilized. 1. Evaluate: f X ωq , c X ωq , and h X ωq using (12),
SQP is an iterative algorithm that aims to find the optimal (13) and (14) 
solution by iteratively updating the decision variable using a 
Formulate: L X ωq , λω , µω using (16)
sequence of quadratic subproblems. Each of these subprob-  
lems serves as an approximation to the original nonlinear 2. Calculate: ∇ 2 L X ωq , λω , µω using (18)
optimization problem. The objective is to find an optimal 3. Solve:The quadratic subproblem given in (15) and
solution by refining the approximation in each iteration. The then obtain β ω
algorithm begins with an initialization step, by randomly 4. Calculate:X ω+1 q , λω+1 , and µω+1 using (17)
selecting a feasible starting solution of the vector X ωq = 5. Set X ∗q = X ω+1
[X1ω , X2ω , . . . .XQω ], i.e., X 0q at ω = 0, where ω indicates q 2
  6. If Q q=1 Xqω+1 − Xqω+1 ≤ δ
1 PQ
the iteration number. Then, the objective function f X ωq Break While
will be written as given  in (12) to iteratively
  solve
 (11). The End If
ω ω ω
constraints of f X q , i.e., c X q and h X q , which are 7. ω = ω + 1
defined in (13) and (14), respectively, ensure the satisfaction End While
of the main constraints given in (11.e) and (11.f) at each
iteration ω.
  XK XQ function with respect to X ωq , which can be expressed as
f X ωq = X ω sqk (12) follows:
k=1 q=1 q
  XQ  
c X ωq = X ω − Btot = 0 (13) ∇ 2 L X ωq , λω , µω
q=1 q
 ∂2L ∂2L
· · · ∂X∂ω ∂X
  XQ 2L 
h X ωq = Xqω skq − Btot gk ≥ 0 (14) ∂X ω ∂X ω ω
 ∂ (X21 )
ω 2
1 2 1 Q
q=1
 ∂ L ∂ L
2 ∂ L
2

 ∂X2ω ∂X1ω ∂ X ω 2 · · · ∂X2ω ∂XQω

Then, at each iteration ω, the aim is to solve the following ( 2)

= .. .. .. , (18)
 
quadratic equation iteratively: ..

 . . . . 

     ∂2L ∂2L ∂2L
∂X ω ∂X ω ∂X ω ∂X ω · · ·

T 1
min β w ∇f X ωq β ω + β Tω ∇ 2 L(X ωq , λω , µω )β ω (15)
 2
Q 1 Q 1 ∂ XQω
2
 T
Solving this optimization problem yields the optimal values
 
s.t. c X ωq + ∇c X ωq β ω = 0 (15.a)
of β ω , which is used to find the values of X ω+1q , λω+1 and
µω+1 using (17).
   T
h X ωq + ∇h X ωq β ω ≥ 0 (15.b)
Algorithm 1 outlines the proposed SQP approach to
where L(X ωq , λω , µω ) indicates the Lagrangian function at an sequentially optimize the X q . The inputs to the algorithm are
iteration ω, which is defined as follows: the MIS structure sqk , the total available bandwidth Btot , the
network topology including ηk and gk , the total number of
iterations Nit and the convergence threshold δ. For initializa-
       
L X ωq , λω , µω = f X ωq + λω c X ωq + µω h X ωq
tion, an initial guess for X 0q at iteration ω = 0 is randomly
(16) initialized. Thenduring ω ≤ Nit ,
 the
 algorithm
 and
 while

ω ω ω
Thus, β ω denotes the increment vector of the decision the values of f X q , c X q , and h X q are evaluated
variable X ωq and the Lagrangian multiplies λω and µω , using (12), (13) and (14), respectively. Lagrangian
 function

ω
as follows: given in (16) is formulated and the ∇ L X q , λω , µω is
2

T calculated using (18). Then, the quadratic optimization given

β ω = [(X ω+1
q − X ωq ), (λω+1 − λω ), (µω+1 − µω )] , (17) in (15) is solved and β ω is obtained. Afterwards, the values
of X ω+1
q , λω+1 , and µω+1 are evaluated using (17). Then
where T indicates Tthe transpose  operation. In (15)
  T T the optimal values of X ∗q are set equal to X ω+1 q . The con-
ω ω ω
∇f X q , ∇c X q , and ∇h X q are the trans- vergence threshold is tested to decide whether to terminate
poses of the vectors containing the first derivative the algorithm or continue till ω = Nit , where the minimum
of the objective function and the constraints stated square error (MSE) between X ω+1 q and X ωq is calculated and
in (12), (13), and (14), respectively, with respect to X ωq . compared with δ as given in Algorithm 1. Then, the value of
Also, ∇ 2 L(X ωq , λω , µω ) is the hessian of the Lagrangian ω is incremented by one.

67410 VOLUME 12, 2024

M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

   
B. MULTI-UAVS POWER CONTROL USING M-MATRIX Pt1 1
 Pt2 1
After optimizing B∗k of each vertex k based on the optimized

   
Pti =  Pt3 and w = ρ0  1  (23)
   
bandwidths of its related MISs, power allocations of its corre-
 ..  .. 

. . 

sponding UAVs i and j are adjusted. The goal is to minimize 
the total UAVs Tx powers while satisfying the optimized PtM 1
data rates of their associated links.This maximizes the sys-
It is important to mention that the matrix Ai , which has a
tems’ energy efficiency in consequence. Additionally, power
dimension of M × M , is considered a Z-matrix, following
control leads to reduced interference among UAVs. In this
the description given in [23] and [24]. In other words, it is a
context, we will utilize the M-matrix theory to optimize the
matrix where all off-diagonal elements are less than or equal
UAVs’ Tx powers [26]. To do that, we transfer back from the
to zero. In order to find the solution for Pti , it is necessary for
link graph Ĝ given in Fig. 3 to the original node graph G given
Ai to be also an M -matrix. An M -matrix is a special type of
in Fig. 2. This shift is necessitated by the fact that bandwidth
Z-matrix whose inverse consists of all nonnegative elements.
is assigned to links, while power is transmitted by devices.
Thus, to solve equation (21) and obtain A−1 i , we will apply the
Thus, the power optimization sub-problem can be formulated
Sherman-Morrison formula given in [26]. To facilitate this,
as follows:
we will re-express Ai as follows:
XM
min Pti , (19) Ai = Fi − 1M 1Ti (24)
pti i=1
s.t. Pti ∈ (0, Ptmax ) , ∀i ∈ M , (19.a) where
2
  
Pti hij 1
≥ ϕk∗ ∀i, j ∈ k (19.b) Fi = diag hij ∗ + 1 , 1 ≤ j ≤ M
PM 2 ϕk
v=1.v̸=i Ptv hvj + ρ 0    T
  1 h1j
where ϕk∗ = 2ck /Bk − 1 is the adjusted SINR of the link
∗ 1  h2j 
   
1
k connecting UAV i and UAV j. Herein, ck is the minimum 1M =   , and 1Ti =  h3j  (25)
 
 ..   .. 
data rate requirement in bps of link k using its adjusted B∗k .  . 
2
coming from the SQP algorithm. hij is the channel gain 1 hMj
2
between UAV i and UAV j, while hvj is the channel gain The Sherman-Morrison theorem [26] provides a method
between UAV v, v ̸ = i, and UAV j. Both can be calculated to compute the inverse of any off-diagonal singular square
using the path loss equation given in (2). The first constraint matrix in the form of equation (24). This inverse is given by:
given in (19.a) is the same as that given in constraint (6.c) of
i 1M 1i Fi
F−1 T −1
the original problem. The second constraint given in (19.b) A−1 −1
i = Fi + (26)
indicates that the minimum SINRij of the link k connecting 1 − 1Ti F−1
i 1M
UAVs i and j is bounded by its ϕk∗ value. This ensures that the
However, there is a crucial condition for finding a feasible
power control algorithm should satisfy the minimum chan-
non-negative solution for Pti using equation (26), which is as
nel quality level among UAVs. Using convex optimization,
Pti |hij |
2 follows:
constraint (19.b) can be relaxed as PM 2 = ϕk∗ M 
v=1.v̸=i Ptv |hvj | +ρ0 1 −1
X 
and then the energy efficiency is maximized. Thus, the power 1+ ∗ <1 (27)
ϕk
control can be solved as a set of linear equations. i=1

2 M
The proof of this condition directly follows from equa-
Pti hij X 2 tion (26), where Fi , 1M , and 1Ti are all positive diagonal
− Ptv hvj = ρ0 , (20)
ϕk∗ matrices and vectors as depicted in equation (25). Therefore,
v=1.v̸=i
from equation (26), Ai becomes an M-matrix, meaning that
which can be expressed using matrix-vector notation as A−1
i contains only non-negative elements, if and only if the
follows: denominator if 1 − 1Ti F−1 i 1M > 0 or 1i Fi 1M < 1. This
T −1

leads to the condition in (27) using (25). However, it is impor-

Ai Pti = w ⇒ Pti = A−1
i w, (21) tant to note that the condition in (27) solely ensures the lower
bound condition of Pti , which is Pti ≥ 0, without addressing
where
. the upper bound condition, i.e., Pti ≤ Ptmax . Therefore, the
 
h1j
2
ϕk∗ − h2j ···
2
− hMj
2 solution for Pti obtained through (26) should be constrained

2 2
.
2 
 by its maximum value, Ptmax . It is important to note that by
 − h1j h2j ϕk∗ · · · − hMj 

reducing the UAVs Tx powers, we enable the neighboring
Ai =  .. .. ..
 (22)
.. links in the multi-UAV network to reuse the same bandwidth.
. . . .. 
 


2 2 2
 This set influences the enhancement of spectrum efficiency
− h1j − h2j ··· hMj ϕk∗ as discussed in section A.

VOLUME 12, 2024 67411

 M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

TABLE 1. Simulation parameters.

V. NUMERICAL RESULTS AND DISCUSSION FIGURE 4. Data rate against the number of UAVs with network
density = 1 and fixed power allocation.
In this section, the performance of the proposed algo-
rithms including coloring-graph based SQP and M-matrix
theory-based power control is assessed through extensive
Monte-Carlo numerical simulations. The simulation area
encompasses a post disaster area, where random number of
UAVs are deployed in this area with arbitrarily horizontal
distances in the range of [1,100] m, and the UAVs’ vertical
distance is adjusted to 50 m. The reference distance d0 is
set to 1 m, and the maximum UAVs’ Tx powers are set
to 1 watt. The total available bandwidth for the multi-UAV
communication system is set to 10 MHz. Additional simula-
tion parameters are listed in Table 1; unless otherwise stated.
The performance of the proposed approach is evaluated by
comparing it with three different benchmarks, which are
the scheme given in [14], the maximum spectrum packing
(MASPECT) algorithm given in [29], and the random selec- FIGURE 5. Data rate against the number of UAVs with network density =
tion (RS). In the scheme given in [14], a linear optimization 0.8 and fixed power allocation.
problem using graph theory was formulated for bandwidth
allocation in device to devices (D2D) network, accounting for
link heterogeneity in demands and traffic injection patterns.
The objective was to maximize the spectrum efficiency while
efficiently utilizing the available bandwidth. In MASPECT
algorithm given in [29], a distributed opportunistic channel
acquisition mechanism for improving spectrum utilization
was proposed using graph theoretic approach. In RS both
bandwidth and power resources are set randomly in the ranges
of [0,Ptmax ] and [0, Btot ], respectively. In the simulation sce-
narios, we generate random geometric graphs with graph
densities of 0.5, 0.8 and 1, where a graph density of 1 means
that the network is fully connected, where all UAVs are mutu-
ally connected and ηij = 1∀i, j. For graph density of 0.8 (0.5),
the number of UAVs mutual connections is 80% (50%) of that
of the fully connected network using random ηij .
FIGURE 6. Data rate against the number of UAVs with network density =
Figures 4, 5 and 6 represent the data rate performances in 0.5 and fixed power allocation.
Mbps of the schemes involved in the comparisons at fixed
UAVs’ Tx power of Ptmax and network densities of 1, 0.8, and
0.5, respectively. Generally, as the number of UAVs increases, the issue. That is, as the number of UAVs escalates, the
the data rates of all compared schemes are decreasing in bandwidth resources allocated per UAV diminish, resulting
consequence. This is due to the increased interference level in reduced data rates for each individual UAV. Moreover,
among the deployed UAVs, where the limited availability of by comparing the results presented in Figs 4, 5 and 6,
bandwidth resources and fixed power allocation exacerbate we can conclude that network density plays a crucial role

67412 VOLUME 12, 2024

M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

in influencing data rate outcomes. When network density is

low, the data rate tends to be high due to the less interference
among the limited number of connected UAVs. This can be
observed when comparing the data rates of graph density
of 1 presented in Fig. 4, with those results from low density
of 0.5 presented in Fig. 6, where a higher density results in
lower data rate performance, and vice versa. This discrepancy
can be attributed to the increased number of links, leading
to elevated interference levels with a fixed power allocation.
Thus, the juxtaposition of these scenarios underscores the
intricate interplay between network density, interference, and
link dynamics, ultimately influencing the overall data rate
performance of the compared schemes.
Looking closer, the figures highlight that the proposed
FIGURE 7. Energy efficacy against the number of UAVs with network
coloring-graph based SQP approach outperforms other meth- density = 1 and with power control (WPC) and without power
ods in maximizing data rates. Notably, in Figs 4, 5 and 6, control (WoPC).
the proposed coloring-graph based SQP algorithm emerges as
the standout performer, demonstrating a remarkable improve-
ment of over 70% in data rate when the number of UAVs
is set to 3 in all tested network densities. This substantial
enhancement, however, exhibited diminishing returns as the
number of UAVs increased, eventually converging to a dif-
ference of less than 10% compared to alternative methods
due to excessive mutual interference. Also, UAVs’ fixed Tx
power allocation may lead to congestion and heightened
interference, counteracting the potential benefits gained from
the addition of more UAVs.
From Fig. 4 and when using 3 fully connected UAVs,
i.e., network density of 1, RS yields the lowest data rate of
approximately 1.72 Mbps. In contrast, MASPECT improves
performance, achieving nearly 2.46 Mbps, while the scheme
proposed in [14] excels further with a data rate of 4.21 Mbps. FIGURE 8. Energy efficacy against the number of UAVs with network
Notably, the proposed SQP algorithm stands out, surpassing density = 0.8 and with power control (WPC) and without power
control (WoPC).
all alternatives with an impressive data rate of approximately
25.6 Mbps. Shifting to scenarios with network density of 0.8,
in Fig. 5 and using 3 UAVs, MASPECT becomes less favor- performance of the scheme given in [14] is owed to its linear
able, yielding a data rate of 2.12 Mbps, while RS achieves optimization formulation, strategically allocating resources,
a relatively higher data rate of 2.94 Mbps. The scheme and optimizing bandwidth in response to varying network
proposed in [14] maintains commendable performance of conditions. It is interesting to note that at a network den-
approximately 5 Mbps, and the proposed SQP with color- sity of 0.5, the Rand scheme outperforms MASPET. This is
ing graph excels with a remarkable data rate of 36 Mbps, because in less densely connected networks, RS’s efficient
showcasing adaptability to varying network densities. Sim- random allocation allows for more effective bandwidth uti-
ilarly, at a network density of 0.5 given in Fig. 6 and lization. In contrast, MASPECT faces challenges as network
using 3 UAVs, MASPECT remains less favorable with a data density decreases as it impacts its ability to leverage the
rate of 1.5 Mbps, while RS achieves 2.96 Mbps. The schemes network topology for optimal bandwidth utilization.
proposed in [14] reaches around 6.2 Mbps, and the proposed Figures 7, 8, and 9 provide a comprehensive analysis of
SQP with coloring graph excels with a data rate of 41 Mbps. energy efficiency (EE) in (bits/Joule) in multi-UAV networks
The observed lower data rate of RS in a high-density network under different power control scenarios, i.e., with power
of 1 can be attributed to its random allocation mechanism, control - WPC and without power control – WoPC, across
leading to increased interference in the densely connected varying network densities of 1, 0.8, and 0.5, respectively.
UAVs. This unstructured allocation results in contention for The integration of power allocation, particularly employ-
bandwidth among UAVs, thereby diminishing the data rate. ing M-matrix theory, emerges as a standout performer in
On the contrary, MASPECT by leveraging graph-based algo- terms of EE, showcasing its superiority over fixed power
rithms excels in a fully connected network by considering allocation when integrated with the schemes involved in
node degrees and probabilities, mitigating interference, and the comparison. Examining the results in Fig. 7 for a network
achieving a higher data rate compared to RS. The outstanding density of 1 and fully connected 3 UAV, it becomes evident

VOLUME 12, 2024 67413

 M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

of power control on EE improvement. MASPECT-WPC

reaches an EE of approximately 0.528 Gbit/Joule, while
MASPECT-WoPC exhibits the lowest EE among the meth-
ods, at around 0.182 Gbit/joule.
In Fig. 9 and using 3 UAVs, SQP-WPC stands out
with a remarkable EE of nearly 11.4 Gbit/Joule, while
SQP-WoPC maintains a low EE of 2.61 Gbit/Joule, further
highlighting the significance of power control in enhanc-
ing performance. However, other methods exhibit lower EE
values than the proposed SQP scheme, with [14]-WPC at
1.65 Gbit/Joule, RS-WPC at 0.93 Gbit/Joule, MASPECT-
WPC at 0.355 Gbit/Joule. For WoPC, [14]-WoPC achieves
0.692 Gbit/Joule, RS-WoPC achieves 0.535 Gbit/Joule, and
MASPECT-WoPC achieves 0.293 Gbit/Joule. These results
FIGURE 9. Energy efficacy against the number of UAVs with network underscore the pivotal role of power control, particularly in
density = 0.5 and with power control (WPC) and without power scenarios with lower graph density, in optimizing the EE of
control (WoPC).
multi-UAV communication networks.

that the proposed SQP, particularly when complemented with VI. CONCLUSION
the proposed power control strategy, achieves remarkable EE This paper delves into the intricate challenges of bandwidth
of approximately 9 Gbit/Joule. This stands in stark contrast allocation and power control in multi- UAV networks. In this
to the significant 50% reduction observed without employing context, two stage optimization were proposed. In the first
power control. The underlying reason behind this substantial stage, bandwidth allocations were optimized by the means of
improvement lies in the ability of power control to dynam- coloring graph based SQP algorithm while considering fixed
ically adapt power levels in response to evolving network power allocations. In the second stage, M-matrix theory was
conditions, minimizing total UAV transmission powers, and utilized to adjust the Tx powers of the UAVs based on their
subsequently maximizing EE. Comparing EE across differ- minimum data rate requirements using its adjusted band-
ent methods, we observe that the EE of the proposed SQP width allocations done in the first stage. Through rigorous
WPC is notably higher at 9G bit/Joule, surpassing the EE numerical simulations, our proposed approach demonstrated
of the method given in [14]-WPC at 1.13G bit/Joule. The superior performance compared to other benchmarks for data
EEs of MASPECT- WPC and RS-WPC are lower than that rates maximization. Also, the proposed power control scheme
of [14]-WPC, achieving 0.637 Gbit/Joule and 0.5 Gbit/Joule, demonstrated superior performances over fixed power alloca-
respectively. However, the worst-case scenario is observed in tion counterpart. This research contributes valuable insights
RS-WoPC, where EE is significantly low at 0.167 Gbit/Joule. into the realm of multi-UAV network optimization, paving
Moreover, EEs of all schemes involved in the comparisons are the way for enhanced information transmission rates and
decreasing with the increasing of the number of UAVs. This is resource utilization in dynamic and challenging environ-
owed to the decrease in their achievable data rates presented ments. As the current study primarily focused on bandwidth
in Fig. 4 due to excessive interference. For example, EE of allocation and power control, we only considered static sce-
the proposed SQP-WPC falls from 9 Gbit/joule when using narios, where UAVs are hovering at a fixed altitude above the
3 fully connected UAVs to 0.42 Gbit/Joule when using 8 fully earth’s surface. Detailed exploration of network dynamics,
connected UAVs as shown in Fig .4. including factors such as nodes mobility, traffic patterns,
In Fig. 8, the evaluation of EE across the compared and topology changes, falls slightly outside the scope of the
schemes, each involving 3 UAVs and a graph density of 0.8, current paper, and it will be the subject of our future inves-
reveals distinct performances. The proposed SQP algorithm, tigations. Also, optimizing communication delays against
when paired with power control, demonstrates exceptional interference will be explored during our future studies.
EE, achieving around 9.3 Gbit/Joule. This underscores the
SQP’s efficiency in optimizing data rates while consider- REFERENCES
ing power consumption in the specified network scenario. [1] E. M. Mohamed, M. Alnakhli, S. Hashima, and M. Abdel-Nasser, ‘‘Dis-
SQP-WoPC maintains commendable EE at approximately tribution of multi mmWave UAV mounted RIS using budget constraint
4.19 Gbit/Joule, where it is less than half the performance multi-player MAB,’’ Electronics, vol. 12, no. 1, p. 12, Dec. 2022, doi:
10.3390/electronics12010012.
of QSP-WPC. The scheme [14]-WPC, achieves compet-
[2] M. Mozaffari, X. Lin, and S. Hayes, ‘‘Toward 6G with connected sky:
itive EE at about 2.37 Gbit/Joule, while its non-power- UAVs and beyond,’’ IEEE Commun. Mag., vol. 59, no. 12, pp. 74–80,
controlled counterpart [14]-WoPC shows a notable reduction Dec. 2021.
in EE at around 1.02 Gbit/Joule. RS-WPC and RS-WoPC, [3] H. S. Khallaf, S. Hashima, M. Rihan, E. M. Mohamed, and H. M. Kasem,
‘‘Quantifying impact of pointing errors on secrecy performance of UAV-
achieve EE values of approximately 0.619 Gbit/Joule based relay-assisted FSO links,’’ IEEE Internet Things J., vol. 11, no. 2,
and 0.268 Gbit/Joule, respectively, highlighting the impact pp. 2979–2989, Jan. 2024, doi: 10.1109/jiot.2023.3297857.

67414 VOLUME 12, 2024

M. A. Alnakhli et al.: Bandwidth Allocation and Power Control Optimization for Multi-UAVs Enabled 6G Network

[4] X. Jiang, M. Sheng, N. Zhao, C. Xing, W. Lu, and X. Wang, ‘‘Green UAV [28] E. M. Mohamed, K. Sakaguchi, and S. Sampei, ‘‘Wi-Fi coordinated WiGig
communications for 6G: A survey,’’ Chin. J. Aeronaut., vol. 35, no. 9, concurrent transmissions in random access scenarios,’’ IEEE Trans. Veh.
pp. 19–34, Sep. 2022. Technol., vol. 66, no. 11, pp. 10357–10371, Nov. 2017.
[5] E. M. Mohamed, S. Hashima, and K. Hatano, ‘‘Energy aware multiarmed [29] S. Anand, S. Sengupta, and R. Chandramouli, ‘‘MAximum SPECTrum
bandit for millimeter wave-based UAV mounted RIS networks,’’ IEEE packing: A distributed opportunistic channel acquisition mechanism in
Wireless Commun. Lett., vol. 11, no. 6, pp. 1293–1297, Jun. 2022, doi: dynamic spectrum access networks,’’ IET Commun., vol. 6, no. 8,
10.1109/LWC.2022.3164939. pp. 872–882, 2012, doi: 10.1049/iet-com.2010.0607.
[6] E. M. Mohamed, S. Hashima, A. Aldosary, K. Hatano, and
M. A. Abdelghany, ‘‘Gateway selection in millimeter wave UAV
MOHAMMAD AHMED ALNAKHLI (Member,
wireless networks using multi-player multi-armed bandit,’’ Sensors,
vol. 20, no. 14, p. 3947, Jul. 2020, doi: 10.3390/s20143947. IEEE) received the B.Sc. degree in electrical engi-
neering from Umm Al-Qura University, Mecca,
[7] A. Amrallah, E. M. Mohamed, G. K. Tran, and K. Sakaguchi, ‘‘UAV trajec-
tory optimization in a post-disaster area using dual energy-aware bandits,’’ Saudi Arabia, in 2008, the M.Sc. degree in
Sensors, vol. 23, no. 3, p. 1402, Jan. 2023, doi: 10.3390/s23031402. electronic communication and computer engi-
[8] E. M. Mohamed, ‘‘Deployment of mmWave multi-UAV mounted RISs neering from the Electrical and Computer Engi-
using budget constraint Thompson sampling with collision avoid- neering Department, University of Nottingham,
ance,’’ ICT Exp., vol. 10, no. 2, pp. 277–284, Apr. 2024, doi: Nottingham, U.K., in 2011, and the M.Sc. degree
10.1016/j.icte.2023.07.011. in network and security and the Ph.D. degree
[9] W. T. Tutte, Graph Theory, vol. 21. Cambridge, U.K.: Cambridge Univ. in electrical and computer engineering from the
Press, 2001. Stevens Institute of Technology, Hoboken, NJ, USA, in 2017 and 2019,
[10] P. T. Boggs and J. W. Tolle, ‘‘Sequential quadratic programming,’’ Acta respectively. He is currently an Assistant Professor with the Depart-
Numerica, vol. 4, pp. 1–51, Jan. 1995. ment of Electrical Engineering, Prince Sattam bin Abdulaziz University,
[11] M. A. Khan, N. Kumar, S. A. H. Mohsan, W. U. Khan, M. M. Nasralla, Saudi Arabia. His current research interests include spectrum manage-
M. H. Alsharif, J. Zywiolek, and I. Ullah, ‘‘Swarm of UAVs for network ment and energy consumption in next-generation wireless networks, radio
management in 6G: A technical review,’’ IEEE Trans. Netw. Service resource allocation in unmanned aerial vehicles (UAVs) and low earth
Manag., vol. 20, no. 1, pp. 741–761, Mar. 2023. orbit (LEO) satellites, the Internet of Things, energy and spectrum efficient
[12] X. Liu, Y. Liu, Y. Chen, and L. Hanzo, ‘‘Trajectory design and power wireless networks, 5G, 5GB and 6G systems, and cognitive radio networks.
control for multi-UAV assisted wireless networks: A machine learning
approach,’’ IEEE Trans. Veh. Technol., vol. 68, no. 8, pp. 7957–7969,
EHAB MAHMOUD MOHAMED (Member,
Aug. 2019.
IEEE) received the Ph.D. degree in information
[13] Y. Chen and W. S. Wong, ‘‘Power control for non-Gaussian interference,’’
science and electrical engineering from Kyushu
IEEE Trans. Wireless Commun., vol. 10, no. 8, pp. 2660–2669, Aug. 2011.
University, Japan, in 2012. He is currently a
[14] M. Alnakhli, S. Anand, and R. Chandramouli, ‘‘Joint spectrum and energy
efficiency in device to device communication enabled wireless networks,’’
Full Professor with the Department of Electri-
IEEE Trans. Cognit. Commun. Netw., vol. 3, no. 2, pp. 217–225, Jun. 2017. cal Engineering, College of Engineering in Wadi
[15] C. Yan, L. Fu, J. Zhang, and J. Wang, ‘‘A comprehensive survey Addwasir, Prince Sattam bin Abdulaziz Univer-
on UAV communication channel modeling,’’ IEEE Access, vol. 7, sity, Saudi Arabia. He also holds the same position
pp. 107769–107792, 2019. with the Department of Electrical Engineering,
[16] A. A. Khuwaja, Y. Chen, N. Zhao, M.-S. Alouini, and P. Dobbins, ‘‘A sur- Aswan University, Egypt. He was a Specially
vey of channel modeling for UAV communications,’’ IEEE Commun. Appointed Researcher with Osaka University, Japan, from 2013 to 2016.
Surveys Tuts., vol. 20, no. 4, pp. 2804–2821, 4th Quart., 2018. He has more than 150 journals and conference papers. His current research
[17] X. Fu, T. Ding, R. Peng, C. Liu, and M. Cheriet, ‘‘Joint UAV channel interests include 5G/6G networks, cognitive radio networks, millimeter wave
modeling and power control for 5G IoT networks,’’ EURASIP J. Wireless transmissions, Li-Fi technology, MIMO systems, and reconfigurable intelli-
Commun. Netw., vol. 2021, no. 1, p. 106, Dec. 2021. gent surfaces. He is a technical committee member of many international
[18] X. Wang, Z. Fei, J. Huang, J. A. Zhang, and J. Yuan, ‘‘Joint resource allo- conferences and a reviewer of many international conferences, journals, and
cation and power control for radar interference mitigation in multi-UAV transactions, especially highly ranked IEEE TRANSACTIONS/journals. He was
networks,’’ Sci. China Inf. Sci., vol. 64, no. 8, Aug. 2021, Art. no. 182307. the General Chair of the IEEE ITEMS’16 and IEEE ISWC’18 and a guest
[19] S. Lee, S. Lim, S. H. Chae, B. C. Jung, C. Y. Park, and H. Lee, ‘‘Optimal editor in many highly ranked transactions/journals.
frequency reuse and power control in multi-UAV wireless networks: Hier-
archical multi-agent reinforcement learning perspective,’’ IEEE Access, MOSTAFA M. FOUDA (Senior Member, IEEE)
vol. 10, pp. 39555–39565, 2022.
received the B.S. degree (as the Valedictorian)
[20] J. Li, S. Li, and C. Xue, ‘‘Resource optimization for multi-unmanned aerial
and the M.S. degree in electrical engineering
vehicle formation communication based on an improved deep Q-network,’’
from Benha University, Egypt, in 2002 and 2007,
Sensors, vol. 23, no. 5, p. 2667, Feb. 2023.
respectively, and the Ph.D. degree in information
[21] X. Su, S. Chan, and J. H. Manton, ‘‘Bandwidth allocation in wireless ad
sciences from Tohoku University, Japan, in 2011.
hoc networks: Challenges and prospects,’’ IEEE Commun. Mag., vol. 48,
no. 1, pp. 80–85, Jan. 2010. He is currently an Associate Professor with the
[22] F. Harary, Graph Theory. Reading, MA, USA: Addison-Wesley, 1969.
Department of Electrical and Computer Engineer-
ing, Idaho State University, ID, USA. He also
[23] J. Zheng, J. Zhang, and B. Ai, ‘‘UAV communications with WPT-aided
cell-free massive MIMO systems,’’ IEEE J. Sel. Areas Commun., vol. 39, holds the position of a Full Professor with Benha
no. 10, pp. 3114–3128, Oct. 2021, doi: 10.1109/JSAC.2021.3088632. University. He was an Assistant Professor with Tohoku University and a Post-
[24] J. Zheng, J. Zhang, H. Du, D. Niyato, B. Ai, M. Debbah, and K. B. Letaief, doctoral Research Associate with Tennessee Technological University, TN,
‘‘Mobile cell-free massive MIMO: Challenges, solutions, and future USA. He has (co)authored more than 220 technical publications. His current
directions,’’ IEEE Wireless Commun., early access, Feb. 5, 2024, doi: research interests include cybersecurity, communication networks, signal
10.1109/MWC.004.2300043. processing, wireless mobile communications, smart healthcare, smart grids,
[25] G. Poole and T. Boullion, ‘‘A survey on M-matrices,’’ SIAM Rev., vol. 16, AI, and the IoT. He has guest-edited a number of special issues covering var-
no. 4, pp. 419–427, 1974. ious emerging topics in communications, networking, and health analytics.
[26] S. Chen, ‘‘Inequal. For M-matrices and inverse M-matrices,’’ Linear Alge- He currently serves on the editorial board for the IEEE INTERNET OF THINGS
bra Appl., vol. 426, nos. 2–3, pp. 610–618, 2007. JOURNAL (IoT-J), IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY (TVT), and
[27] Q. Zhang, H. Wang, Z. Feng, and Z. Han, ‘‘Many-to-many matching- IEEE ACCESS. He has received several research grants, including NSF Japan–
theory-based dynamic bandwidth allocation for UAVs,’’ IEEE Internet U.S. Network Opportunity 3 (JUNO3).
Things J., vol. 8, no. 12, pp. 9995–10009, Jun. 2021.

VOLUME 12, 2024 67415