Ad Hoc Networks 158 (2024) 103474

Contents lists available at ScienceDirect

Ad Hoc Networks
journal homepage: www.elsevier.com/locate/adhoc

Survey paper

Metaheuristic Algorithms for 6G wireless communications: Recent advances

and applications
Ammar Kamal Abasi a ,∗, Moayad Aloqaily a , Mohsen Guizani a , Bassem Ouni b
a
Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, United Arab Emirates
b
Technology Innovation Institute (TII), Abu Dhabi, United Arab Emirates

ARTICLE INFO ABSTRACT

Keywords: The widespread distribution of intelligent applications and devices, coupled with the vast array of mobile data,
Sixth-generation (6G) technologies, and architectures within Sixth Generation (6G) networks, highlights the importance of optimiza-
Metaheuristics algorithms (MHAs) tion strategies. Concurrently, state-of-the-art Metaheuristic Algorithms (MHAs) have emerged as a promising
Beyond 5G (B5G)
optimization approach for wireless networks, bridging game theory and convex optimization domains. Falling
Optimization
within the domain of Artificial Intelligence (AI), MHAs draw their inspiration from Evolutionary Algorithms
Artificial intelligence (AI)
(EAs), which are rooted in the principle of natural selection, the principles of Swarm Intelligence (SI), and
Trajectory-based Algorithms (TAs). In recent times, MHAs have been implemented in the 6G network. These
simple solutions with limited capabilities have proven effective in solving complex, challenging and high-
dimensional problems. Nevertheless, the literature has yet to study the MHAs comprehensively, especially
within 6G networks. The main objective of this study is to investigate how to merge the 6G and MHAs
domains. We begin by giving a comprehensive overview of MHAs, from fundamental to popular optimization
methods. Then, we examine the utilization of MHAs to address the challenges in 6G, such as network security,
resource allocation, spectrum management, edge computing, wireless caching, and others. Finally, we identify
the current literature’s limitations and suggest avenues for further research directions.

1. Introduction The advent of Sixth Generation (6G) networks has led to substantial
progress in the telecommunications industry. These advancements are
The growth of data traffic has been phenomenal in recent years, instigating transformative changes in the operation of communication
with an increase of over 50% annually [1]. With the advent of new systems, resulting in the incorporation of groundbreaking technolo-
technologies and the increasing reliance on multimedia content, this gies like drone-based communications, THz communications, and the
trend is expected to continue, with predictions indicating that video succeeding generation of the Internet of Things (IoT) [8]. Compared
will account for 82% of all IP traffic [2]. This burgeoning data traf- to the Fifth Generation (5G) networks, 6G networks are expected to
fic necessitates advancing state-of-the-art communication systems to provide faster speeds, lower latency, connected intelligence, and aug-
meet evolving digital requirements. The future deployment of mobile mented capacity, thus enabling a broad spectrum of new applications
networks will pose challenges owing to this significant surge in data and services in the telecommunications industry [9]. 6G technology is
traffic, which includes the necessity for a broader spectrum [3], in- aimed at providing higher data rates for communication systems by
creased system capacity [4], enhanced mobility [5,6], and escalating
incorporating current technological advancements [10]. Essential tech-
energy consumption [7]. The existing cellular networks are unable to
nologies, such as channel modeling [11], transceiver design [12], and
address these issues efficiently. The pressing need to resolve various
the use of massive Multi-Input-Multi-Output (MIMO) with Millimeter
issues has accelerated the evolution of mobile technologies, prompting
Wave (mmWave) communications [13], are driving the development
researchers to concentrate on the forthcoming generation of mobile
of 5G and 6G mobile networks.
communication. Innovative solutions often demand the amalgamation
In 6G, several issues need to be investigated, including network
and collaboration of multiple technologies to attain the desired results.
This can be achieved by applying advanced techniques across the in- security, resource allocation, task offloading, spectrum management,
volved technologies, creating a holistic, effective, and efficient solution. edge computing, and wireless caching [14]. As a result, advanced

∗ Corresponding author.
E-mail addresses: [email protected] (A.K. Abasi), [email protected] (M. Aloqaily), [email protected] (M. Guizani),
[email protected] (B. Ouni).

https://doi.org/10.1016/j.adhoc.2024.103474
Received 14 September 2023; Received in revised form 22 December 2023; Accepted 6 March 2024
Available online 15 March 2024
1570-8705/© 2024 Elsevier B.V. All rights reserved.
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

optimization methods are necessary to support the implementation and cases. The survey would also identify research gaps and areas for
advancement of 6G networks. Optimization methods involve math- future work, thus aiding future research on optimizing and securing
ematical formulations, algorithms, or specific procedures to identify 6G networks using MHAs. This survey would be a valuable resource
the optimal combination of elements and solutions to achieve desired for the 6G and MHA research communities.
objectives through the minimization and/or maximization of specific
functions [15]. 1.2. Comparison with existing surveys
There are several techniques that have been proven to be effective
in handling these challenges, including standard Machine Learning This survey focuses on using MHAs in addressing the challenges
(ML), game theory and convex optimization. When an optimization associated with 6G networks. Unlike existing studies, such as [21,
problem is convex, it means that its mathematical properties allow 22], which largely overlook the specific applications of MHAs in 6G
for well-known approaches and pre-existing solvers to be used to networks, especially in areas like feature selection [22], this survey
obtain near-optimal solutions. However, this is only sometimes the
fills a critical void in the current literature, as indicated in Table 1.
case, as most optimization problems are not purely convex and must
To provide a more comprehensive perspective on the comparative
be represented as a series of convex sub-problems. In addition, the
analysis with existing surveys, it is essential to consider both the
computational complexity of approximate algorithms tends to increase
positive and negative aspects of these surveys. While previous studies
as the optimization becomes more complex, such as when there are
like [23,24], and [25] offer extensive insights into various dimensions
high levels of channel quantization or a large number of IoT devices.
of IoT networks, they fall short in addressing the distinctive chal-
In these situations, heuristic algorithms are often recommended due
lenges and applications pertinent to 6G networks. For instance, [23]
to their simplicity and low computational complexity. However, these
predominantly emphasizes security and privacy solutions for resource
algorithms cannot guarantee an optimal solution or even convergence
allocation and computation offloading, while [24] centers its focus on
to the optimal solution. Metaheuristic algorithms (MHAs), a sub-field
resource allocation, clustering, and routing within the IoT network
of Artificial Intelligence (AI), have become widely used in research
context. On the other hand, recent surveys, exemplified by [25,26],
for their ability to provide computationally tractable and high-quality
present a broader overview of various topics encompassing security
solutions while ensuring robustness and convergence [16].
and privacy, Multiple-Input Multiple-Output (MIMO) detection, UAV,
MHAs have several advantages over conventional methods. They are
clustering and routing, ML, and DL. However, they need to specifically
easy to implement, offer global optimization, handle various optimiza-
delve into the unique challenges posed by 6G networks, which makes
tion problems, provide superior constraint handling, and have efficient
this survey’s emphasis on these emerging issues particularly relevant.
parallelization capabilities [17–19].
Additionally, surveys like [27] concentrate solely on specific types of
MHAs are often based on the principles of evolutionary-based algo-
rithms (EAs), trajectory-based algorithms (TAs), and swarm intelligence MHAs, particularly SI approaches, for various aspects of IoT networks.
(SI). In general, MHAs are nature-inspired and take inspiration from While they offer valuable insights into these niche applications, they
fields such as ethology, biology, or physics. These algorithms exhibit must address the broader challenges and applications inherent to 6G
stochastic behaviors, involving random variables and have a number of networks, which this survey comprehensively covers. Furthermore, the
parameters that must be adjusted to the specific problem at hand [20]. temporal relevance of this survey must be considered. In contrast to
Evolutionary algorithms (EAs) mimic natural selection by start- most related surveys on MHAs published years ago, this survey offers
ing with a population of solutions and using genetic operations like a fresh and up-to-date examination of MHAs’ applications to emerging
crossover and mutation to produce better solutions step-by-step. Some issues in 6G networks. It explicitly addresses a wide range of critical
examples of EAs include Genetic Algorithm (GA), Biogeography-Based issues, including security and privacy, resource allocation, spectrum
Optimization (BBO), Differential Evolution (DE), Genetic Programming management, energy efficiency, load balancing, user association, spec-
(GP), and Evolution Strategies (ES). Swarm Intelligence (SI) methods, trum sensing, MIMO detection, channel estimation, precoding, antenna
on the other hand, draw inspiration from the collective intelligence designs, intelligent reflecting surfaces, joint computing, caching, com-
of social animals such as ants, birds and fish to solve optimization putation offloading, UAV, smart city network slicing, clustering, and
problems. routing within the IoT network context, as well as ML and DL aspects.
Various algorithms are employed to seek optimal solutions. Some
such algorithms include Particle Swarm Optimization (PSO), which 1.3. Contributions and structure of the paper
is based on bird swarms; the Firefly Algorithm (FA), which draws
inspiration from fireflies; the Salp Swarm Algorithm (SSA); and the Bat The purpose of this survey is to provide a comprehensive overview
Algorithm (BA), modeled after the echolocation of bats. These algo- of the current state of research in the domains of MHAs and 6G. This
rithms utilize interactions among agents to arrive at optimal solutions. survey presents several vital contributions, which are outlined below:
Local search algorithms, such as Tabu Search (TS), make incremental
‘‘neighborhood moves’’ to improve solutions within a nearby search • Overview of MHAs: The survey provides a thorough under-
space. Although these algorithms are efficient for smaller search spaces, standing of the fundamentals of MHAs and evaluates various
they may only occasionally find the global optimum. As a result, representative algorithms such as GA, PSO, ABC, ACO, FA, GWO,
they are often combined with other methods to ensure comprehensive and WOA.
optimization. • MHAs Applications in 6G: The survey explores the use of MHAs
to address technical challenges in 6G, including data privacy and
1.1. Motivation of the survey network security, spectrum management and resource allocation,
wireless caching and edge computing, and other related issues.
The emergence of 6G wireless technology presents opportunities • Exploring the Security and Privacy Issues in 6G Networks:
and challenges that must be addressed. These challenges include in- The survey provides an in-depth examination of the security and
creased speed, connectivity, complexity, and security risks. A compre- privacy aspects of MHAs applications in 6G.
hensive survey on recent MHA advancements for 6G optimization and • Future prospects: The survey presents the current status of re-
security/privacy issues would help researchers and practitioners navi- search on the use of MHAs in addressing optimization problems
gate this complex landscape. Such a survey would provide an overview within the scope of 6G. It identifies significant obstacles and
of the latest MHAs and their applications in 6G communications, ex- outlines potential avenues for enhancing and advancing this field
amining various algorithms, performance metrics, and real-world use in the future.

2
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 1
Comparison with previous surveys.
Refs. [28] [29] [30] [31] [23] [24] [32] [21] [22] [25] [26] [27] [33] This paper
Year 2019 2019 2019 2020 2020 2020 2021 2021 2021 2021 2021 2021 2022 2023
Edge computing ✓ ✓ ✓ ✓ ✓ ✓ ✓
Scope Iot ✓ ✓ ✓ ✓ ✓ ✓ ✓
Cloud ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
EA ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
MHAs
SI ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
techniques
TA ✓ ✓ ✓ ✓ ✓
Security and privacy ✓ ✓ ✓ ✓ ✓ ✓ ✓
Resource allocation ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Spectrum management ✓ ✓ ✓ ✓ ✓
Energy efficiency ✓ ✓ ✓ ✓ ✓ ✓
Load balancing ✓ ✓ ✓ ✓
User association ✓ ✓ ✓
Spectrum sensing ✓ ✓ ✓
MIMO detection ✓ ✓ ✓ ✓ ✓ ✓
Channel estimation ✓ ✓ ✓ ✓
Field Precoding ✓
Antenna designs ✓ ✓
Intelligent reflecting surfaces ✓
Joint computing ✓ ✓
Caching ✓ ✓ ✓ ✓ ✓ ✓
Computation offloading ✓ ✓ ✓ ✓ ✓ ✓
UAV ✓ ✓ ✓ ✓ ✓
Smart city ✓ ✓
Network slicing ✓ ✓
Clustering and routing in Iot networks ✓ ✓ ✓ ✓ ✓ ✓ ✓
ML ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
DL ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓

Fig. 1. Roadmap of MHAs in 6G wireless communications.

The roadmap for the paper is further explained in Fig. 1 presented in Section 6. Section 7 highlights the challenges and future
The survey is structured as follows: Section 2 provides an overview research directions. Finally, the paper concludes in Section 8.
of MHAs. Sections 3, 4, and 5 review the implementation of MHAs In this paper, Table 2 lists the abbreviations that are most frequently
to address 6G network issues. The use of MHAs for 6G applications is used.

3
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 2
List of abbreviations.
Abbreviations Definition Abbreviations Definition
6G Sixth Generation IDS Intrusion Detection System
ABC Artificial Bee Colony IIOT Industrial Internet-Of-Things
ACO Ant Colony Optimization IOT Internet Of Things
AI Artificial Intelligence IRS Intelligent Reflecting Surfaces
AO Aquila Optimizer LO Lemurs Optimizer
AVOA African Vultures Optimization Algorithm LTE Long-Term Evolution
BA Bat Algorithm MCC Mobile Cloud Computing
BBO Biogeography-Based Optimization MEC Mobile Edge Computing
BUO Bottom-Up Optimization MHAS Metaheuristic Algorithms
CHIO Coronavirus Herd Immunity Optimizer CHS Cluster Heads
C-RAN Centralized Radio Access Network MIMO Multi-Input-Multi-Output
CSS Communication Systems MMWAVE Millimeter Wave
D2D Device-To-Device MPA Marine Predators Algorithm
DE Differential Evolution NFV Network Function Virtualization
DL Deep Learning NOMA Non-Orthogonal Multiple Access
DOS Denial-Of-Service NS Network Slicing
DSL Deep Supervised Learning POS Proof Of Space
EAS Evolutionary-Based Algorithms PSO Particle Swarm Optimization
ES Evolution Strategies QOS Quality Of Service
FA Firefly Algorithm RIS Reconfigurable Intelligent Surfaces
FL Federated Learning SA Simulated Annealing
F-RAN Fog Radio Access Network SAC Soft Actor-Critic
GA Genetic Algorithm SBSS Small Cell Bss
GANS Generative Adversarial Networks SCA Sine Cosine Algorithm
GP Genetic Programming SDN Software-Defined Networking
GWO Gray Wolf Optimizer SI Swarm Intelligence
HBA Honey Badger Algorithm IOV Internet Of Vehicles
IWN Industrial Wireless Network THZ Terahertz
IPS Intrusion Detection And Prevention UAV Unmanned Aerial Vehicles
IWN Industrial Wireless Network VLC Visible Light Communication
VNFS Virtual Network Functions – –

2. Metaheuristic algorithms (MHAs): An overview

2.1. Fundamentals of metaheuristics

The term ‘‘metaheuristics’’ is a combination of two words: ‘‘meta’’

and ‘‘heuristics’’. The term ‘‘meta’’ means ‘‘beyond’’, and ‘‘heuristic’’
refers to a problem-solving method that uses experience-based tech-
niques to find a solution. The term ‘‘metaheuristics’’ refers to a class of
algorithms used to solve optimization problems using techniques that
go beyond or are ‘‘meta’’ to traditional heuristics [34]. MHAs are a
general class of problem-solving methods that can be applied to a wide
range of optimization problems [35]. They are used when the problem
is too complex for conventional optimization techniques. If the problem
has multiple solutions, an MHA can be used to find the best solution
from a set of possible solutions.
As shown in Fig. 2, the basic idea behind MHA is to start with
an initial set of candidate solutions and then iteratively improve these
solutions by using heuristics and stochastic methods. The solutions can Fig. 2. The general framework of MHAs.
be manipulated using different strategies such as mutation, crossover,
or random walks. The algorithms often repeat generating and refining
solutions until a satisfactory solution is found or a stopping criterion is function. More specifically, the fitness function generally depends
met. on the specific optimization problem being solved. For example,
The strength of MHA lies in their ability to find high-quality so- in a problem of minimizing a cost function, the fitness function
lutions relatively quickly, even for large and complex problems [36]. would be the negative of the cost function. The fitness function
They are beneficial for problems where traditional optimization meth- would be the profit in a problem of maximizing a profit. The
ods, such as gradient-based or linear programming, are not applicable fitness function must satisfy some fundamental properties, such as
or practical. being computationally efficient, easy to evaluate, and providing
Here is a high-level explanation of how MHAs work: a clear and consistent measure of the quality of the candidate
solution. The choice of the fitness function depends on the specific
• Initialization: The algorithm creates an initial set of candidate solu- optimization problem and the application’s requirements. The fit-
tions, also known as a population, representing different potential ness function must be carefully designed to ensure that the MHA
solutions to the optimization problem. can effectively explore the solution space and find high-quality
• Fitness function evaluation: The algorithm then evaluates each can- solutions.
didate solution’s fitness or quality using a mathematical function • Generation of new solutions: The algorithm then generates new
that maps a candidate solution to a scalar value. The goal is candidate solutions from the current population using various
to find the solution that maximizes (or minimizes) the fitness heuristics and stochastic methods.

4
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Fig. 3. Classification of metaheuristic optimization techniques.

• Selection of solutions: The algorithm then selects a subset of the In other words, optimization algorithms can only solve some prob-
new and existing candidate solutions to form the next popula- lems with equal success [38]. This theorem highlights the importance
tion based on their fitness values. This process, known as the of selecting the appropriate algorithm for a given problem and the
selection, ensures that the next population consists of high-quality need to consider the problem-specific characteristics when designing
candidate solutions. optimization algorithms. This has prompted researchers to work on
• Repeat: The process of generating and refining solutions is re- enhancing these algorithms. One popular method is hybridization,
peated until a satisfactory solution is found or a stopping criterion which can involve using elements from different algorithms, using
is met. multiple algorithms in sequence, or incorporating chaotic maps. While
these methods may increase computation costs, they often yield more
MHAs are characterized by their flexibility and generality, as they accurate results. Fig. 3 illustrates another classification of MHAs.
can be applied to a wide range of optimization problems and do not Real-world problems often involve constraints that limit the feasible
rely on a particular mathematical model of the problem. They are also solutions to a specific region [39]. These constraints must be consid-
capable of finding high-quality solutions in a relatively short time, even ered when evaluating candidate solutions obtained through stochastic
for large and complex problems [37]. methods, such as MHA. Ignoring constraints may result in candidate
MHAs can be classified in several ways: solutions that are not feasible, or that need to meet the requirements
of the problem. In order to handle constraints, many MHAs incorporate
• Nature-inspired vs. nonnature-inspired MHA: This classification is penalty functions or other methods to ensure that the candidate solu-
based on the origins of the MHA, with nature-inspired examples tions stay within the feasible region and meet the problem’s constraints.
being ACO, GA, PSO, and simulated annealing (SA). These methods are crucial to finding solutions that are not only optimal
• Memory-based vs. memory-less MHA: The presence of memory, or but also feasible and acceptable. Solutions must meet these constraints
the use of historical information, is considered a vital component to be considered infeasible and practical. Methods such as penalty
in the design of effective MHA. Memory-based MHA utilizes the functions are commonly used to handle these constraints, penalizing
history of search in their optimization process, while memory-less solutions that violate them so they do not appear more favorable than
MHA does not consider past information in their optimization. others [40].
• Population-based vs. single-point MHA: This classification is based
on the number of solutions used during the MHAs execution. 2.2. Class 1 of MHAs: Genetic Algorithm (GA)
Population-based MHA maintains a set of candidate solutions
and iteratively updates this population to find a better solution. GA is a class of MHAs inspired by the principles of natural evolu-
Examples of population-based MHA include GA, ACO, and PSO. tion [41]. This algorithm operates based on the principle of simulating
On the other hand, single-point MHA focuses on finding the natural selection, a process in which the most vital individuals are
best solution by iteratively modifying a single-candidate solution. selected for reproduction to produce offspring with enhanced genetic
Examples of single-point MHA include hill climbing, SA, and tabu traits. The algorithm uses this concept to generate new candidate
search. solutions and gradually improve the quality of solutions over time.
By mimicking the process of evolution, the algorithm can identify and
Table 3 summarizes various algorithms and their attributes, includ-
preserve the most promising solutions while discarding the less optimal
ing the algorithm’s memory, type of memory, the content of memory,
ones, leading to the discovery of increasingly better solutions as the
whether it is a nature-inspired algorithm, complexity, whether it is
optimization process continues [57]. The GA aims to find solutions to
population-based, and its popularity in literature. With the advance-
complex problems by iteratively improving a population of candidate
ments in population-based stochastic algorithms, one may question
solutions through a series of steps, including initialization, evaluation,
the need for further improvements. The No-Free-Lunch (NFL) theorem
selection, crossover, mutation, and replacement.
states that the average performance of any optimization algorithm over
The basic steps in a GA are as follows:
all possible problems is the same, regardless of the algorithm used.
This means that for every optimization problem for which an algorithm • Initialization: A population of candidate solutions is generated
performs well, another problem exists for which it performs poorly. randomly.

5
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 3
Enhanced characteristics of MHAs in 6G wireless communications.
Algorithm Memory Type of memory Content of memory Nature-inspired Complexity Population-based Popularity in
algorithm algorithm literature
ST MT LT
GA [41] – – – – Yes High Yes High: Proven
optimization

PSO [42] ✓ – ✓ – Global and personal Yes Medium Yes High: Extensively
best particles. studied

GWO [43] ✓ – ✓ Best three solutions. Yes Medium Yes Medium:

Adaptability

ABC [44] ✓ ✓ – ✓ Best food source. Yes Medium Yes Medium:

Bio-inspired

ACO [45] ✓ ✓ – ✓ Pheromone matrix. Yes High Yes High: Ant colony
behavior

BA [46] – – – – – Yes Medium Yes Medium: Social

behavior

BA with ✓ – ✓ – Best position. Yes Medium Yes Low: Memory

memory incorporation
[47]

FA [48] – – – – – Yes Medium Yes Medium:

Bioluminescence

FA with ✓ – ✓ – Best fitness value with Yes Medium Yes Low: Memory
memory Fireflies. utilization
[49]

CSO [50] ✓ ✓ Position of the best cat. Yes Medium Yes Low: Feline-inspired

SA [51] ✓ ✓ – Best solution found so No Low No High:

far. Simulation-based

LO [15] ✓ ✓ – ✓ Position of the best Yes Medium Yes Moderate (2022)

lemur

HBA [52] ✓ ✓ – ✓ Best solution found Yes (Honey badger Medium Yes Moderate (2021)
foraging)

AVOA [53] ✓ ✓ – ✓ Best solutions found by Yes (African Medium Yes Moderate (2021)
the swarm vulture scavenging)

AO [54] ✓ ✓ – ✓ Global best solution Yes (Eagle hunting) Medium Yes Moderate (2021)

MPA [55] ✓ ✓ – ✓ Best prey locations Yes (Marine Medium Yes Moderate (2020)
predator hunting)

CHIO [56] ✓ ✓ – ✓ Best solution Yes Medium Yes Moderate (2021)

• Evaluation: The fitness of each candidate solution is evaluated, which work together to find food or a new home. In PSO, a population
usually by using a fitness function that measures how well the of candidate solutions is represented as a set of ‘‘particles’’ that move
solution solves the problem. through the search space. Each particle has a position and a velocity,
• Selection: The fittest individuals are selected for reproduction and it updates its position based on its own experience and the experi-
based on their fitness scores. ence of other particles in the swarm. The particles are guided towards
• Crossover: The selected individuals are combined to produce off- better solutions by a set of ‘‘personal best’’ and ‘‘global best’’ positions,
spring through a crossover process, where traits from one parent representing the best solution that a particular particle has found so far
are passed on to the offspring. and the best solution found by any particle in the swarm [58].
• Mutation: A small number of offspring are randomly mutated to The basic steps in a PSO algorithm include initialization, evaluation,
introduce new genetic material and prevent the population from update, and termination. During initialization, a population of particles
becoming too homogeneous. is randomly generated, with each particle having a random position and
• Replacement: The new offspring replace the least fit individuals velocity. The fitness of each particle is then evaluated using a fitness
in the population. function. In the update step, the position and velocity of each particle
are updated based on its personal best and global best positions, as
The process is then repeated, with the population evolving towards well as a set of acceleration constants. The algorithm terminates when
better and better solutions with each iteration. GA has been successfully a satisfactory solution is found or when a predetermined number of
applied to many problems, including the optimization of continuous iterations have been reached.
functions, combinatorial optimization, and ML. They can find reason- In PSO, the position and velocity of each particle are updated
able solutions to complex problems even when the search space is iteratively according to the following equations:
enormous, and they are robust to noise and uncertainty. Position update:

2.3. Class 2 of MHAs: Particle Swarm Optimization (PSO) 𝑥(𝑡 + 1) = 𝑥(𝑡) + 𝑣(𝑡 + 1) (1)

Velocity update:
PSO is an MHA developed by Kennedy and Eberhart in 1995 [42].
It is based on the behavior of social animals, such as birds and bees, 𝑣(𝑡 + 1) = 𝑤 × 𝑣(𝑡) + 𝑐1 × 𝑟𝑎𝑛𝑑()

6
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

× (𝑝𝑏𝑒𝑠𝑡 − 𝑥(𝑡)) + 𝑐2 × 𝑟𝑎𝑛𝑑() × (𝑔𝑏𝑒𝑠𝑡 − 𝑥(𝑡)) (2) positions based on three types of bees: the employed bees, the onlooker
bees, and the scout bees. The employed bees represent the solutions
In these equations, 𝑥(𝑡) is the current position of the particle, 𝑣(𝑡) is
currently being explored, the onlooker bees represent the solutions
the current velocity, 𝑝𝑏𝑒𝑠𝑡 is the personal best position of the particle
waiting to be explored, and the scout bees represent the solutions
(i.e., the best solution that the particle has found so far), 𝑔𝑏𝑒𝑠𝑡 is the
randomly generated to explore new areas of the search space [60].
global best position (i.e., the best solution found by any particle in the
The position of each bee is updated using the following equations:
swarm), 𝑤 is the inertia weight, 𝑐1 and 𝑐2 are acceleration constants,
Employed bee:
and 𝑟𝑎𝑛𝑑() is a random number between 0 and 1.
The position update equation adds the velocity to the current po- 𝑥𝑖 (𝑡 + 1) = 𝑥𝑖 (𝑡) + 𝑟𝑖 × (𝑥𝑖 (𝑡) − 𝑥𝑗 (𝑡)) (4)
sition to determine the particle’s new position. The velocity update
equation combines the current velocity, the difference between the per- Onlooker bee:
sonal best and current positions, and the difference between the global 𝑥𝑘 (𝑡 + 1) = 𝑥𝑘 (𝑡) + 𝑟𝑘 × (𝑥𝑖 (𝑡) − 𝑥𝑘 (𝑡)) (5)
best and current positions, weighted by the acceleration constants and
the inertia weight. This equation guides the particles towards better Scout bee:
solutions by incorporating the experience of the particle and the rest of 𝑥𝑠 (𝑡 + 1) = 𝑥𝑚𝑖𝑛 + 𝑟𝑎𝑛𝑑() × (𝑥𝑚𝑎𝑥 − 𝑥𝑚𝑖𝑛 ) (6)
the swarm.
PSO has been applied to a wide range of problems, including the where 𝑥(𝑡) is the position of the bee at time t, 𝑟𝑖 and 𝑟𝑘 are random
optimization of continuous functions, combinatorial optimization, and numbers in the range [0, 1], 𝑥𝑗 (𝑡) is a randomly chosen solution, 𝑥𝑚𝑖𝑛
ML. It has the advantage of being simple to implement and requiring and 𝑥𝑚𝑎𝑥 are the lower and upper bounds of the search space, and 𝑟𝑎𝑛𝑑()
only a few parameters to be tuned. is a random number in the range [0, 1].
The basic steps in an ABC algorithm include initialization, evalu-
2.4. Class 3 of MHAs: Grey Wolf Optimizer (GWO) ation, update, and termination. During initialization, a population of
bees is randomly generated, with each bee having a random position.
Grey Wolf Optimizer (GWO) is an MHA introduced by Mirjalili et al. The fitness of each bee is then evaluated using a fitness function. In the
in 2014 [43]. It is based on the hunting behavior of grey wolves, update step, the positions of the employed bees and onlooker bees are
in which they cooperate to locate and attack their prey. In GWO, a updated based on the equations above, and the scout bees are generated
group of ‘‘wolves’’ explores the search space represented as candidate to explore new areas of the search space. The algorithm terminates
solutions. The wolves update their positions based on three distinct when a satisfactory solution is found or when a predetermined number
leadership roles: the alpha, the beta, and the delta wolves. The alpha of iterations have been reached.
wolf represents the global best solution found so far, the beta wolf The ABC algorithm has widespread usage in various fields due to
represents the second-best solution, and the delta wolf represents the its simplicity and versatility. It has been applied to optimization prob-
worst solution [59]. lems involving continuous functions, combinatorial optimization, and
The position of each wolf is updated using the following equations: ML. The algorithm’s implementation is straightforward and requires
minimal parameter tuning, making it accessible to practitioners with
𝑥⃗𝑖𝑡+1 = 𝑥⃗𝑗 𝑡 − 𝐴⃗ ⋅ 𝐷𝑖,
⃗ 𝑗𝑡 (3) limited computational resources or experience. However, it is essential
where 𝑥⃗𝑖𝑡+1
is the updated position of the 𝑖th wolf at time 𝑡+1, 𝑥⃗𝑗 𝑡
is the to remember that ABC is sensitive to the initial conditions and may not
⃗ 𝑗𝑡
position of the prey (i.e., the alpha, beta, or delta wolf at time 𝑡), 𝐷𝑖, always find the optimal solution to a given problem. Despite this, ABC
is the distance vector between the 𝑖th wolf and the 𝑗th prey at time 𝑡, remains a popular choice for optimization problems due to its ease of
and 𝐴 is a scaling factor that controls the step size of the update. use and effectiveness in many applications.
The basic steps in the GWO algorithm include initialization, eval-
uation, update, and termination. During initialization, a population of 2.6. Class 5 of MHAs: Ant Colony Optimization (ACO)
wolves is randomly generated, with each wolf having a random posi-
tion. The fitness of each wolf is then evaluated using a fitness function. Ant Colony Optimization (ACO) is an optimization algorithm in-
In the update step, the positions of the alpha, beta, and delta wolves troduced by Dorigo et al. in 1992 [45]. It is based on the foraging
are updated based on the equations above. The algorithm terminates behavior of ants, in which ants communicate the locations of food
when a satisfactory solution is found or when a predetermined number sources through pheromones. In ACO, a group of ‘‘ants’’ explores the
of iterations have been reached. search space represented as candidate solutions. The ants update their
The GWO algorithm is a versatile optimization technique applied positions based on the pheromone trails left by other ants and the
to many problems, including continuous functions, combinatorial opti- quality of the solutions at each position. The pheromone trails serve as
mization, and ML. GWO is easy to implement and requires minimal a guide for the ants, with more robust trails indicating more promising
parameter tuning, making it accessible to practitioners with limited solutions [61].
computational resources or experience. The position of each ant is updated using the following equation:

𝑥𝑖 (𝑡 + 1) = 𝑥𝑖 (𝑡) + 𝑟𝑖 × (𝑥𝑏 𝑒𝑠𝑡 − 𝑥𝑖 (𝑡)) + 𝑐𝑖 × 𝑟𝑎𝑛𝑑 (7)

2.5. Class 4 of MHAs: Artificial Bee Colony (ABC)
where 𝑥(𝑡) is the position of the ant at time t, 𝑟𝑖 is a random number in
The artificial Bee Colony (ABC) algorithm is an MHA first proposed the range [0, 1], 𝑥𝑏 𝑒𝑠𝑡 is the best solution found so far, and 𝑐𝑖 and rand
by Karaboga in 2005 [44]. The behavior of honeybees inspires them in are constants that control the behavior of the ants.
their search for food sources. The ABC algorithm treats the candidate The basic steps in an ACO algorithm include initialization, evalu-
solutions as food sources, and the optimization process simulates bees’ ation, update, and termination. During initialization, a population of
foraging behavior. The algorithm uses a population of ‘‘bees’’ to explore ants is randomly generated, with each ant having a random position.
the solution space and exploit the best solutions. The bees communicate The fitness of each ant is then evaluated using a fitness function. In the
their findings to each other, leading to an efficient search for the opti- update step, the positions of the ants are updated based on the equation
mal solution. This algorithm has been applied to various optimization above, and the pheromone trails are updated based on the quality of the
problems and has shown promising convergence and solution-quality solutions at each position. The algorithm terminates when a satisfactory
results. In ABC, the search space is explored by a group of ‘‘bees’’, solution is found or when a predetermined number of iterations have
which are represented as candidate solutions. The bees update their been reached.

7
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

ACO has been applied to many problems, including the optimization • Increased Security and Privacy: MHAs can be used to design se-
of continuous functions, combinatorial optimization, and ML. It can cure and private communication schemes, such as secure routing
find reasonable solutions to complex problems even when the search and jamming, which can protect the network and its users from
space is vast and robust to noise and uncertainty. However, it should malicious attacks.
be noted that ACO can be computationally expensive and may require • Self-organizing and Self-optimizing Networks: MHAs can be used to
careful tuning of the various parameters that control the algorithm’s develop algorithms for self-organizing and self-optimizing net-
behavior. works, which can automatically adapt to changing network con-
ditions and optimize performance without human intervention.
2.7. Class 6 of MHAs: Simulated Annealing (SA) • Network design: MHAs can be used to design and plan the network
infrastructure to provide the best coverage and capacity.
Simulated Annealing (SA) is an MHA introduced by Kirkpatrick • Network Slicing (NS): MHAs can be used to separate the network
et al. in 1983 [51]. It is based on the concept of the annealing process into multiple virtual networks optimized for different use cases
used in metallurgy. In this process, a material is first heated to a high or services, for example, for specific verticals like eHealth or
temperature and then gradually cooled to reduce defects and improve Industrial IoT.
its structural purity. The SA algorithm uses a similar approach to • Power Efficiency: MHAs can minimize energy consumption in the
optimization, exploring the solution space and gradually narrowing in network and increase the energy efficiency (EE) of the devices,
on the optimal solution by mimicking the cooling process. This allows ultimately reducing the network’s carbon footprint.
the algorithm to overcome local optima and converge to the global opti- • Improving Interoperability and seamless connectivity: MHAs can help
mum, making it a powerful optimization technique for solving complex design devices and networks that can seamlessly communicate
problems. In SA, the search space is explored by a ‘‘system’’, which is with each other and interact with other technologies like 5G, IoT,
represented as a candidate solution. The system updates its position and satellite networks.
based on a probabilistic transition rule that allows it to ‘‘accept’’ worse • Edge–Cloud integration: 6G networks will heavily rely on the inte-
solutions with a certain probability. This allows the system to escape gration of edge computing and cloud computing resources.MHAs
local minima and explore a broader range of the search space [62]. can be used to optimize the allocation and management of edge
The position of the system is updated using the following equation: and cloud resources to ensure that the overall system can perform
optimally.
𝑃 (𝑥′ |𝑥) = 𝑒𝑥𝑝[−(𝐸(𝑥′ ) − 𝐸(𝑥))∕𝑇 ] (8) It is important to note that as 6G networks are still in the research
where 𝑥 and 𝑥′ are the current and new positions of the system, and development phase, the optimization techniques that will be in-
𝐸(𝑥) is the energy (fitness) of the system at position 𝑥, and 𝑇 is tegrated with them are still actively being studied and developed by
the temperature, which controls the probability of accepting worse researchers. It is expected that as more research is conducted in this
solutions. area, more optimization techniques will be discovered and developed
The basic steps in an SA algorithm include initialization, evaluation, that can be used to improve the performance of 6G networks.
update, and termination. During initialization, the system is initialized
at a random position. The energy (fitness) of the system is then eval- 3. Security and privacy in 6G networks
uated using a fitness function. In the update step, the position of the
system is updated based on the equation above, and the temperature Network security is an increasingly complex issue for each new
is gradually decreased according to a predetermined schedule. The generation of networks due to the constantly evolving nature of net-
algorithm terminates when a satisfactory solution is found or when the works and the rapid expansion of new technologies and services [63].
temperature reaches a predefined minimum value. MHAs have been shown to manage security-related challenges in 6G
SA has been applied to a wide range of problems, including the networks effectively. In this segment, we will examine the utilization
optimization of continuous functions, combinatorial optimization, and of these algorithms in addressing security challenges in the context of
ML. It has the advantage of finding suitable solutions to complex 6G networks. Fig. 4 presents six common use cases and six critical
problems even when the search space is vast, and it is robust to noise security requirements that are relevant to both 5G and 6G networks,
and uncertainty. However, it should be noted that SA can be computa- as succinctly summarized from sources such as [63–66].
tionally expensive and may require careful tuning of the temperature
schedule and other parameters that control the algorithm’s behavior. 3.1. Access control

2.8. Integrating MHA techniques with 6G networks Attribute-based access control is a reliable method of protecting
resources, such as computational or storage resources, from unau-
6G networks, the next generation of wireless communication net- thorized access or malicious activities. This approach grants access
works, are expected to bring significant advances in performance, only to authorized devices and users and effectively prevents external
security, and capabilities. One key area of research in developing 6G threats, making it a highly secure and adaptable network administra-
networks is the integration of MHAs to improve network performance tion technique. In the context of 6G networks, efficient access control
and efficiency. There are several ways in which MHAs can be integrated can be established through MHAs to address complex optimization
with 6G networks to improve their performance: challenges associated with resource allocation, user verification, and
network security. Abdulqadder et al. [67] employed a cuckoo search to
• Improved resource allocation: MHAs can be used to optimize the improve a PSO-based routing protocol designed for Software-Defined
allocation of various resources, such as bandwidth, power, and Networking (SDN) controllers, optimizing routing for better QoS and
computation, to different devices and services in the network. reduced susceptibility to malicious user infiltration. This work was
This can lead to more efficient use of resources, reducing costs further developed in a subsequent study by Chang et al. who introduced
and increasing overall network capacity. an attribute-based access control mechanism to enhance SDN integrity
• Enhanced Quality of Service (QoS): MHAs can be used to optimize and confidentiality.
various parameters, such as scheduling, routing, and modulation, MHAs have effectively secured data access in 6G cloud networks
to improve the QoS for different types of traffic, such as video and [68]. They use cryptography algorithms to determine the most secure
real-time applications. keys using statistical probability to prevent cryptanalysis. MHAs also

8
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Lin et al. [72] investigate the integration of 6G technology and indus-

trial IoT in the smart industry, focusing on the complexities of ensuring
data security and privacy in highly sensitive contexts. The authors
propose an ACO technique to enhance the security of 6G IoT networks
and address these challenges. This technique involves representing each
ant in the population as a set of potential deletion operations aimed at
concealing confidential information. It also uses a prelarge concept to
streamline multiple database searches and external solutions, resulting
in optimized outcomes while maintaining data privacy.

3.3. Communication technologies

This subsection introduces recent research on the utilization of

MHAs to enhance security performance in various communication tech-
nologies. In one study [73], a Non-Orthogonal Multiple Access (NOMA)
system is evaluated in a UEH network using a MIMO architecture
with MRT and MRC techniques. The study employs untrusted relays
and power-splitting for energy harvesting and an amplify-and-forward
protocol while introducing artificial noise at the source to increase
secrecy. The authors provide formulations for secrecy outage proba-
bility and validate these through simulations. They use the PSO al-
gorithm to maximize the sum secrecy rate, comparing MIMO/NOMA
Fig. 4. A comparison of applications and security requirements between 5G and and MIMO/OMA performances. The study also examines the impact of
6G. The security framework in 6G advances over 5G by incorporating intelligence, system parameters on secrecy efficacy.
automation, and energy efficiency. In another study [74], PSO is applied to enhance physical layer
security (PLS) in visible light communication (VLC) systems that use
intelligent reflecting surfaces (IRS) made up of intelligent mirror arrays.
have various other applications in the cloud, such as time-varying prob- The authors analyze IRS channel gains and calculate a lower bound for
lems, task priority in multi-objectives, optimizing execution time, and the attainable secrecy rate in an IRS-assisted VLC framework, including
data offloading. For example, the work in [69] proposes an I-MCTSO an access point, a legitimate user, and an eavesdropper. Optimization
model to account for uncertain factors like network bandwidth, access of mirror orientations using PSO is employed to maximize the achiev-
reliability level, and MIPS in cloud computing. The proposed InMaOEA able secrecy rate through an optimization problem. The simulation
uses interval credibility and hyper-volume strategies to improve the results demonstrate a significant improvement in PLS for VLC systems
convergence and population diversity of MHA. Simulation results show achieved through this approach.
that the InMaOEA effectively provides efficient allocation plans for A recent research work presented in [75] focuses on enhancing the
cloud task scheduling. security of Internet of Vehicles (IoV) applications, which have been
The study presented in [70] introduces the SliceBlock framework, rapidly growing in popularity. These applications face significant chal-
a comprehensive solution designed to tackle security concerns in Net- lenges due to the complex communication environment and dynamic
work Slicing (NS) for Software-Defined Networks (SDNs) and network attributes of mobile IoV users. To address this issue, the researchers
function virtualization (NFV) within the context of 6G landscapes. By introduce a novel approach that utilizes the GWO to predict and
combining generative adversarial networks (GANs) for network coding evaluate the secrecy performance of IoV applications. The proposed
(NC) with Directed Acyclic Graph (DAG)-based blockchain technology algorithm incorporates a Generalized Regression (GR) neural network
and integrating a Proof of Space (PoS) consensus algorithm, the Slice- that includes pattern and summation layers. The GWO algorithm is
Block model significantly enhances security measures. Furthermore, the leveraged to optimize the spread factor, thereby accelerating the GR
model incorporates advanced security mechanisms, including secure network’s performance convergence.
handover and context-based authentication through Markov decision Another research work presented in [76] introduces an Intrusion
making (MDM) and the weighted product model. To address the strain Detection prevention (IPS) that enhances communication security be-
tween vehicles and Roadside Units (RSUs) in Vehicular Ad Hoc Net-
on SDN controllers and switches, the SliceBlock model employs in-
works (VANETs) against malicious attacks such as black holes and
novative techniques like intruder packet classification and migration
wormholes. The RSU implements the IPS mechanism and utilizes a
facilitated by a hybrid neural decision tree (HyDNT) and a Hybrid Polit-
swarm optimization approach, specifically the PSO algorithm, to iden-
ical Optimizer with a Heap-based Optimizer (HPoHO). Additionally, the
tify and prevent malicious actions targeting vehicles. The proposed
model incorporates environment learning utilizing the SAC algorithm,
security system allows vehicles to transmit and relay traffic informa-
enabling accurate load prediction. Rigorous testing of the SliceBlock
tion while the RSU monitors traffic data exchanges to detect malev-
model’s performance is carried out in a simulated 6G environment.
olent activities. The proposed IPS scheme’s effectiveness is assessed
through simulations, and a comparative analysis against previous In-
3.2. Data privacy trusion Detection Systems (IDS) and attacker models is conducted. The
PSO-equipped IPS scheme demonstrates superior performance against
The emergence of 6G networks presents new opportunities for mo- blackhole and wormhole attackers, as demonstrated by the achieved
bile users and network operators to establish innovative business mod- performance metrics.
els. One of these opportunities is the exchange of large volumes of data In a recent study, researchers investigated the use of a Hybrid
over the networks. However, this also raises concerns about user pri- Reflecting-Intelligent Surface (HR-RIS) to improve secrecy in millimeter
vacy, including the risks of data leakage and information breaches [71]. -wave MIMO systems when dealing with multi-antenna eavesdropping.
To address these concerns, 6G networks need to take a comprehensive They utilized alternating optimization techniques to concurrently op-
approach to data protection. One potential solution is to use secure timize relay-reflecting coefficients and the transmit beamformer at the
information to provide highly effective privacy solutions. In their work, HR-RIS. This involved determining the optimal transmit beamformer

9
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Fig. 5. Flow diagram depicting the utilization of MHAs in secrecy prediction.

at the sender using a closed-form expression and utilizing the PSO to interconnectedness of smart devices and services with seamless inte-
optimize both passive and active elements of the HR-RIS, aiming for gration of humans. Employing multi-agent systems to create industrial
the best solution. agents shows promise in addressing challenges in the evolving smart
Moreover, MHAs can be employed to fine-tune the hyperparameters industry. However, these systems can be vulnerable to data injection
of a secrecy prediction model, as demonstrated in Fig. 5. This approach attacks originating from insiders due to the extensive information
can enhance the precision and efficiency of the secrecy prediction exchange and agent reactions. Researchers must focus on enhancing the
model in 6G networks by iteratively adjusting these hyperparameters resilience of these systems against such threats. In [78], a trust-aware
and assessing the model’s performance on a validation dataset at each approach is proposed to fortify these systems against data injection
iteration until a satisfactory set of hyperparameters is found. attacks, with the initial PSO algorithm serving as an illustrative case,
A recent study in [77] proposed a new approach to optimize spo- potentially extending the contribution to more intricate techniques.
radic spectrum in the Internet of Everything (IOE) and 6G networks. The introduction of 6G satellites has expanded the attack surface,
The approach involved short-packet communication, windowing-based requiring novel security architectures. One proposed solution is to
orthogonal frequency division multiplexing, and filtering with a GA equip satellites with security services such as Virtual Network Functions
to optimize power spectral density while adhering to peak-to-average (VNFs). However, dynamically orchestrating these services remains a
power ratio constraints. The simulation results showed its superiority complex research challenge. In [79], a new approach is introduced
in peak-to-average power ratio and out-of-band emission, making it to dynamically and optimally allocate security VNFs across satellites,
promising for IOE applications. accounting for intermittent connectivity, computational limitations,
and the requirement for secure VNFs. The optimization problem is
3.4. Attack detection tackled using the Tabu Search algorithm, aiming to minimize the
cumulative service provisioning delays. Additionally, in [80], an en-
In the context of 6G networks, safeguarding wireless networks semble ML method based on weight-based techniques is developed to
against threats and attacks is a critical concern due to the increasing detect anomalous signals transmitted over the Controller Area Network
sophistication of attackers. Advanced techniques for detecting attacks, (CAN) bus of vehicles. To support this approach, a Multi-Objective
such as MHAs methods, are essential to ensure network security. These Evolutionary Algorithm with a Balanced Convergence and Diversity
methods use security postures and intelligence information to create (MaOEA-BCD) is introduced. The proposed approach is validated using
decision-making models that effectively counteract malicious intru- diverse tamper attack scenarios and open-source CAN bus message
sions and threats by integrating probability analysis, logical reasoning, datasets. A summary of the reviewed studies, including techniques
mathematical optimization, and full-text search techniques. employed and critical highlights, is presented in Table 4.
Industry 4.0, driven by technological trends like the Industrial IoT
and cyber–physical systems, has enabled transformative applications 3.5. Lessons learned
such as intelligent manufacturing, predictive maintenance, and asset
tracking. Emerging technologies, including 6G mobile networks, ML, This section delves into the current applications of MHAs in enhanc-
AI, and digital twins, will facilitate the widespread deployment and ing security and privacy within 6G networks. We present examples of

10
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 4
Summary of recent MHAs for 6G network security.
Category Refs. Year Techniques Highlights
Access control [67] 2018 PSO The cuckoo search and the PSO are combined to optimize
routing and improve QoS while minimizing the risk of
malicious user access.
[81] 2019 PSO The PSO improves SDN’s confidentiality and integrity
using attribute-based access control.
[70] 2022 Po, HO The Po with HO are utilized in the SliceBlock model.

Data privacy [72] 2020 ACO The ACO is adopted for securing 6G IoT networks.
Communication technology [73] 2020 PSO The PSO is used to maximize the sum secrecy rate.
[74] 2021 PSO The authors analyze the IRS channel gain and use the
PSO algorithm’s optimization problem to maximize the
achievable secrecy rate.
[75] 2022 GWO The GR network’s performance is optimized using the
GWO algorithm.
[76] 2022 PSO The IPS algorithm is implemented by the RSU using the
PSO to detect and prevent malicious actions from
attacking vehicles.
[82] 2022 PSO The HR-RIS’s passive and active elements are optimized
using the PSO.
Malicious behavior [77] 2021 GA An optimization problem using GA to minimize power
spectral density and meet constraints on peak-to-average
power ratio is formulated.
Attack detection [80] 2021 Multiple-objective Researchers have developed an ML algorithm to detect
optimization anomalous signals transmitted over the CAN bus in
algorithm vehicles. They have introduced a multiple-objective
optimization algorithm called MaOEA-BCD that focuses
on balance convergence and diversity.
[78] 2022 PSO A study proposed an efficient trust-aware approach to
secure multi-agent systems from data injection attacks,
using the classic PSO algorithm as a simple use case.
[79] 2022 Tabu Search The Tabu Search algorithm is employed to minimize the
sum of service provisioning delays, and the results from
realistic scenarios demonstrate its feasibility and highlight
potential areas for future research.

their use in this domain. Attribute-based access control is an effective 4. Resource allocation and spectrum management
approach to protect resources from unauthorized access and potential
attacks when considering access control methods for 6G networks. This Resource allocation and spectrum management are essential tech-
approach offers greater network management flexibility and security. niques for optimizing the performance of wireless communication sys-
The use of MHAs such as cuckoo search and PSO can significantly tems, including 6G [83–86]. Resource allocation refers to the process
improve access control efficiency within 6G networks. of assigning resources, such as time, frequency, or power, to different
Ensuring data security within 6G networks requires a comprehen- users or devices in a wireless network. The goal of resource allocation
sive end-to-end strategy to address privacy concerns related to infor- is to efficiently use the available resources to meet the demands of
mation attacks and data leaks. One potential avenue is to use secure different users or applications. Spectrum management refers to the
information to provide streamlined privacy solutions. Researchers sug- process of regulating the use of the electromagnetic spectrum, which
gest that the application of MHAs can improve the security of 6G is the range of frequencies that can be used for communication. The
goal of spectrum management is to ensure that the available spectrum is
networks by modeling each potential solution as a deletion transaction
used efficiently and fairly and to prevent interference between different
aimed at concealing sensitive information.
users or systems.
In communication technology, MHAs have proven to be highly valu-
In 6G, resource allocation and spectrum management will be crucial
able in assessing and predicting the performance of IoV applications
for achieving the goals of high throughput, spectrum efficiency, and
while also enhancing communication security between vehicles and
EE. For example, advanced techniques such as cognitive radio network
RSUs in VANETs to prevent malicious attacks. MHAs also contribute
(CRN), dynamic spectrum access, and mmWave communication may
to maximizing the sum secrecy rate within NOMA systems operating be used for more efficient use of the available spectrum. Additionally,
within a UEH network, integrating a MIMO architecture. MHAs opti- advanced resource allocation algorithms may be used to optimize
mize achievable secrecy rates by manipulating mirror orientations in the allocation of resources in real-time based on the demands and
an IRS-assisted VLC system. Furthermore, MHAs play a pivotal role in conditions of the network. The integration of MHAs with IoT systems
enhancing the secrecy capacity of MIMO systems by jointly refining the can revolutionize resource allocation and spectrum management in 6G
transmit beamformer and the phase shifts of HR-RIS. Algorithms such networks, as shown in Fig. 6. By incorporating these algorithms into
as PSO, GWO, GA, and Tabu Search extend the potential of MHAs to IoT systems, network operators can dynamically allocate resources and
bolster the robustness and security of systems within the evolving smart manage spectrum to meet the changing demands of IoT devices and
industry landscape, including 6G mobile networks and ML frameworks. services in real-time. This can solve many problems, such as scheduling,
These algorithms are highly proficient in addressing challenges such energy consumption, load balancing, etc., and increase system perfor-
as data injection attacks and the dynamic orchestration of security mance for IoT applications in 6G networks. Moreover, by using MHAs,
services. network operators can also reduce the manual intervention required for

11
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Fig. 6. Integration of MHAs in IoT systems.

Fig. 7. An example of a HetNet with three-tier (femto cell, pico cell, macro cell), and the applications of MHAs.

resource allocation and spectrum management, reducing the potential to improve network performance without requiring additional Base
for human error and freeing up staff to focus on other essential tasks. Stations (BS). The simulation results show that the network’s per-
formance has significantly increased thanks to the MWOA-CR relay
4.1. Channel allocation and power control technique. This work is extended to offloading optimization, where the
combinatorial optimization problem is formulated to minimize energy
Channel allocation and Power control are crucial in wireless sys- consumption and the weighted sum of latency under energy constraints
tems. ‘‘Channel allocation’’ means different things in different con- and latency. The correlation between the amount of energy consumed
texts, e.g., heterogeneous networks (HetNets), device-to-device (D2D) and the lag time is investigated. The effectiveness of the method can
be deduced from the simulation results.
communications channel/subcarrier allocation [87], and user group-
Multiple studies have investigated the use of MHAs with non-
ing/clustering in NOMA [88]. For example, in some cases, D2D users
orthogonal multiple access (NOMA) in 6G mobile communications,
can reuse channels occupied by cellular users if the interference caused
proposing various approaches for optimizing performance. The study
by D2D users is lower than a specified limit. In these scenarios, optimiz-
in [89] explores a RIS-assisted mmWave NOMA system and aims to
ing channel allocation and power control for D2D users is crucial. This
maximize the system sum rate by optimizing power allocation, hybrid
section focuses on using optimization techniques to improve channel precoding, passive beamforming, and user grouping. Meanwhile, [90]
allocation and power control in 6G. Fig. 7 shows a diagram of a three- proposes a GA for NOMA downlink systems to determine user groups
tier Heterogeneous Network (femto cell, pico cell, and macro cell) and and allocate optimal transmission power levels.
the application of MHAs. The work in [88] proposes a power allocation algorithm using PSO,
which significantly improves the EE of the NOMA system. In [83],
4.1.1. Optimizing NOMA systems in 6G mobile communications a new MHA with a blockchain-based resource allocation technique
MHAs have solved several problems in D2D communications. For (MWBA-RAT) is designed for 6G networks and the Internet of Every-
example, the modified WOA with the CRN has been utilized in [87] thing (IoE), which incorporates blockchain technology for effective

12
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

resource management. The study in [86] suggests NOMA as a solu- reflective surfaces, RIS, and PA. Optimizing algorithms such as SSA,
tion for 6G networks to enhance IOV communications in small-cell SCA, PSO, and BUO can help find the optimal solution for non-convex
networks. A multi-objective optimization problem was formulated to optimization problems, leading to improved performance [96]. The
maximize capacity and minimize power for the NOMA network. The MHAs utilized in channel allocation and power control are summarized
simulation results demonstrate that this optimization framework out- in Table 5.
performs other benchmark frameworks in terms of system sum capacity
and EE. 4.2. Load balancing and user association
Finally, in [91], a non-convex optimization problem is formulated
to maximize the sum rate of a downlink transmission by optimizing the Load balancing involves distributing the workload across multiple
decoding sequence, phase shift, and power allocation. This problem was BS, while user association refers to linking users with their preferred
solved using a low-complexity iterative algorithm. BS [97]. According to [98], five key metrics are used for user asso-
ciation: fairness, QoS, EE, spectrum efficiency, coverage probability,
4.1.2. Optimizing spectrum efficiency and EE in wireless networks and outage. Tractable methods for optimizing these issues can be
Spectrum efficiency and EE are critical concepts in the deployment obtained using frameworks such as stochastic geometry, game theory,
and design of 6G networks. Spectrum efficiency refers to the amount and convex optimization. This paper will focus on load-balancing and
of data that can be transmitted over a given frequency band per unit user association mechanisms that utilize MHAs.
of bandwidth. EE, on the other hand, refers to the amount of energy
consumed by the network per unit of data transmitted. In 6G, it is 4.2.1. Spectrum sensing in CRN
expected that the networks will be designed to support high data rates, Spectrum sensing is one of the most critical enablers for CRN. The
low latency, and increased capacity while reducing energy consump- studies in [99–102] all focus on the optimization of spectrum sensing in
tion [92]. This will be achieved through innovative technologies like CRN. The first study in [99] proposes a modified version of GWO called
AI, ML, and mmWave communications. As the world moves towards a MOMGWO, which finds a balance between exploration and exploitation
more connected and data-driven future, spectrum efficiency and EE will to attain global optimums for optimizing multi-objective problems in
play a vital role in enabling 6G networks to meet the growing demand CRN. The second study in [100] integrates the Internet of Connected
for high-speed, low-latency communications. Vehicles (IoCV) with a CRN and uses an improved version of the
Combining MHAs with 6G networks is expected to improve spec- DBSCAN method and a multi-agent-based categorical Deep-Q Network
trum efficiency and EE. By incorporating these algorithms into 6G (categorical DQN) for hybrid beamforming to increase the efficiency of
networks, the system can dynamically allocate spectrum and power the spectrum. The third study in [101] proposes a new RSMA scheme
resources to meet the changing demands of the network, resulting in for CRN, using PSO to find the best beamforming vectors and common
improved spectrum utilization and reduced energy consumption. This rate vectors to achieve maximum EE while maintaining a minimum
will help 6G networks support the increasing demand for high-speed, signal-to-interference-plus-noise ratio (SINR) for primary users (PUs).
low-latency communications while reducing the environmental impact The proposed system design allows an arbitrary number of PUs and
of communication systems. Integrating MHAs with 6G networks will secondary users (SUs) to be present in the system and ensures that the
also enable the development of new use cases and applications, such SINR at any PU is maintained above a specified minimum value, which
as IOT, smart cities, and autonomous vehicles, which will require the guarantees a higher priority for the PUs than for the SUs. The fourth
efficient use of limited spectrum and energy resources. study in [102] presents a novel approach for spectrum sensing in 6G
The studies in [84,85] both focus on optimizing spectrum effi- networks, called EDHO-SSA, which is based on the hunting behavior
ciency and resource management in wireless networks. The former pro- of deer and optimizes the use of available spectrums in 6G networks
poses various optimization algorithms, including GWO, PSO, quantum- by considering factors such as energy and throughput to evaluate
inspired termite colony optimization, and continuous quantum-inspired spectrum sensing performance. By combining the techniques proposed
termite colony optimization for improving communication reliability in these studies, it may be possible to further improve the spectrum
and energy conservation in ultra-dense THz networks, full-duplex mas- sensing performance in CRN using optimization algorithms such as
sive MIMO systems, and B5G wireless networks. At the same time, the GWO, DBSCAN, PSO, and EDHO-SSA. Additionally, the integration of
latter presents a study on the optimization of spectrum allocation for CRN and cooperative spectrum sensing is emphasized as essential for
the IoT using MHAs, such as a binary PSO algorithm, a GA, and an ABC improving the user experience in 6G networks, which face the challenge
algorithm. The performance of wireless networks in terms of spectrum of limited spectrum availability.
efficiency and resource management could be enhanced further by
integrating the methods suggested in these studies. 4.2.2. Optimizing user association, hyperparameter selection, traffic man-
The studies in [93–95] all focus on the optimization of EE in agement, and UAV-based solutions in wireless communication systems
different contexts. The first study in [93] presents research on the opti- Regarding ultra-dense THz networks, [103] presents a throughput-
mization of EE in IRS-assisted multi-user RSMA downlink systems under aware user association system. The scheme considers THz channels’
THz propagation by proposing and comparing a novel optimization characteristics, the antennas’ directivity, the devices’ positions, and
algorithm called the SSA with a traditional algorithm called successive the users’ minimum rate requirements. The authors present a solution
convex approximation. The second study in [94] outlines a sine cosine framework based on the GWO that returns the optimal user association
algorithm to optimize EE in a multiple-input-single-output system for table. They demonstrate that their proposed framework outperforms
RIS systems. The method jointly optimizes passive beamforming at the the commonly used PSO approach through simulation.
RIS and active beamforming at the BS using PSO and SCA algorithms. The work in [104] presents research on determining optimal hy-
The third study in [95] presents an automated method for design- perparameters for using deep neural networks (DNN) to model the
ing and optimizing power amplifiers (PA) using a numerical analyzer behavior of antennas for next-generation mobile communication net-
and an electronic design automation tool by applying the bottom- works, specifically 6G systems. It is acknowledged that DNNs have
up optimization (BUO) method and PSO algorithm with an external proven to be a valuable modeling technique in recent years, but deter-
neural network. The proposed optimization process is fully automatic mining optimal hyperparameters for the networks can be challenging.
and efficient, making it a helpful tool for reducing the involvement of The authors investigate various stochastic methods for determining the
designers during the design process. optimal hyperparameters of a DNN, including trial error, grid search,
Combining the techniques proposed in these studies, it may be pos- random search, Bayesian optimization, GA, and PSO to determine the
sible to further improve the EE in various systems such as intelligent, best-suited method for the specific task of antenna designs. They also

13
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 5
Comparison of various MHAs used in channel allocation and power control.
Refs. Year Optimization Application Pros Cons
techniques
[90] 2017 GA Resource allocation in NOMA downlink The proposed method is effective in The proposed method has yet to be
systems, with the aim of maximizing achieving a balance between total tested in real-world scenarios, so it is
system throughput while maintaining system throughput and fairness among unclear how well it will perform in
fairness. users. practice.
[88] 2018 PSO Power allocation in NOMA systems for The algorithm improves the NOMA The paper only focuses on the
6G mobile communications, with the system’s energy efficiency and solves the simulation results of a single algorithm,
aim of improving EE of the system. EE optimization problem of the NOMA and the results may not be generalizable
system as long as the minimum user to other algorithms
speed is met.
[83] 2021 QO-SRO Resource allocation in 6G networks and The technique offers minimal system Compare the capabilities and prospects
the Internet of Everything using a average cost and minimum power of the proposed method with
combination of blockchain technology consumption compared to existing commercial nodes in 6G in terms of
and the QO-SRO algorithm to improve methods. power consumption and cost.
resource allocation.
[86] 2021 SQP Optimizing wireless communication The proposed optimization framework The paper only presents the proposed
systems. The objectives are to maximize maximizes the sum capacity and EE of NOMA SVNet framework and its
system capacity and minimize the total the SVNet simultaneously through joint performance compared to other
transmit power while considering the power optimization. benchmark approaches, but it does not
quality of service requirements, and mention any specific limitations or
maximum transmit power. drawbacks of the proposed technique.
[84] 2022 GWO, PSO, colony Improving spectrum efficiency and The proposed CQTCO algorithm is The method needs to be enhanced for
optimization resource management in wireless superior to traditional algorithms and widespread usage.
communication systems, including strategies in simulation results.
ultra-dense THz networks, full-duplex
massive MIMO systems, and B5G
wireless networks.
[85] 2022 PSO, GA, and ABC Optimizing spectrum allocation for the A promising tool to tackle the spectrum They provide little detail on the specific
IoT in order to maximize allocation problem in the context of IoT. algorithms or methods being discussed.
signal-to-interference ratio and increase
active devices.
[93] 2022 SSA Optimizing EE in IRS-assisted multi-user Improve the system’s energy efficiency Still, the limitations or drawbacks
RSMA downlink systems under THz and consume less time using MHAs. needed to be fulfilled.
propagation.
[95] 2022 BUO, PSO Automated and efficient design of PA by Presents numerical simulation results A detailed explanation of how the
generating the PA configuration using a that indicate adding reflection features improved SCA and PSO algorithms are
BUO method and optimizing the values can increase system EE performance used to optimize the phase shift matrix
of components using a PSO algorithm, without increasing energy consumption. and switch needs to be provided.
reducing the involvement of designers
during the design process.

propose to use a regression DNN with LSTM layers for performing in The work in [107] discusses the challenges facing satellite IoT
a broad frequency range and to forecast the extended frequency band networks (S-IoT-N) in the context of the digital twins and metaverse.
responses in terms of H-plane radiation patterns (RPs), E-plane, gain These networks, which offer high-performance sensing and monitoring
(G), and S11. capabilities, will encounter increased traffic loads. The authors focus
The work in [105] presents research on the challenges of managing on satellite internet networks (SIN) as a promising solution to these
traffic demand in wireless communication systems, particularly in light challenges, highlighting their characteristics such as high data rates
of the growing complexity of HetNets and the rising expectations and low earth orbit (LEO). They argue that existing load-balancing
of mobile users for high-quality service. The authors point out that schemes for satellite networks are ineffective in dealing with non-
traditional association mechanisms based on maximum power received uniform user distributions and dynamic density variations. To address
can lead to imbalanced HetNets, making it challenging to meet the this problem, the researchers propose a new load-balancing method
traffic requirements of mobile users. The authors propose a solution to that uses advanced GA and models the spatial–temporal distribution of
this problem by combining the CRE method with a PSO algorithm to users. They first model a PDF of user distribution in the direction of the
improve user downlink requirements. The new approach considers both satellite’s trajectory, which provides a multidirectional constraint for
BS load and user SINR to determine CRE bias values for each small BS. non-uniform user distribution in S-IoT-N. They also propose an adaptive
The authors present simulation results that demonstrate their proposed inheritance iteration that optimizes the crossover and mutation factors
method effectively reduces network imbalance and meets mobile user for GA. Using this improved GA, they obtain the optimal scheme for
data traffic needs compared to traditional PSO approaches. load balancing that adapts from a local to a global scheme and selects
The study in [106] suggests a cost-effective approach to cater to the beams for accessing.
the changing demands of mobile users by using Flying Base Stations
(FlyBSs) instead of conventional static BSs. The paper introduces a 4.2.3. Optimizing EE in 6G mobile networks: Techniques and algorithms
method to determine the best placement for FlyBSs and maximize Edge devices (EDs) carry minimal energy, but 6G mobile networks
user satisfaction through optimized data rates. The performance of two require more incredible communication energy [108]. Due to the dif-
methods – one based on GA and the other based on PSO – are evaluated ficulties in collecting dynamic UAV coordinates, EDs cannot adjust
and compared. The results indicate that both solutions enhance user emission energy. A DL-based Energy Optimization (DEO) technique,
satisfaction, with PSO having lower complexity but needing fewer presented in [109], enables EDs to communicate with mobile relay
active FlyBSs. UAVs with reduced energy consumption. Energy harvesting has the

14
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

potential to provide unlimited energy in the rapidly evolving 6G net- slicing. It leverages an enhanced PSO strategy, ensuring a balance be-
work, but its effectiveness may be limited by technology. The EH tween local and global solutions. Simulations highlight the approach’s
structures presented in [110] show better performance than traditional advantages in terms of faster convergence speed and better network
continuous transmission systems, as indicated by numerical results. resource utilization compared to other PSO variants.
The work in [111] presents a hybrid computational intelligence al- Another recent study [120] delves into the efficient allocation of
gorithm called MPGND for reducing transmission power in 6G networks limited network resources for 6G NOMA technologies. Their work
with IBN architecture. This is crucial for a sustainable green 6G-IBN, as addresses a Quality of Experience (QoE) optimization problem involv-
these networks gather data from various devices through clusters. The ing power allocation, subcarrier (SC) assignment, and User Equipment
proposed algorithm outperforms other recently proposed algorithms, (UE)-Base Station (BS) association. GA and ABC algorithms are de-
including SPBO, AOA, STSA, and AGWO, as demonstrated by the ployed for UE-BS association, SC assignment, and power allocation.
experimental and statistical analysis results. The article emphasizes the Simulations underscore the rapid convergence of the proposed power
importance of AI techniques and IBN architecture in enabling seamless allocation technique and the enhanced Mean Opinion Score (MOS) as
connectivity in 6G technology. Another optimization algorithm, HW- the number of users and SCs increases.
SHO, developed in [112] by combining WOA and SHO, has been shown Furthermore, a distinct optimization framework for 6G mobile net-
to reduce energy consumption for 6G-enabled IoT devices. works has been introduced recently [121]. This study hones in on radio
The work in [113] presents the design of AIDO-EM for 6G networks, resource allocation and introduces a pioneering bi-objective selection
which is aimed at reducing energy consumption and extending the hyper-heuristic (SHH) technique. This technique optimizes transmis-
lifespan of 6G-enabled IoT devices. This is achieved through cluster- sion power levels to enhance both user QoE and energy efficiency.
based routing, using a DOA for effective data distribution and selecting The framework is structured with an initial heuristic-based allocation
the best GWH. The performance of AIDO-EM is demonstrated through phase followed by the core implementation of the bi-objective SHH.
simulations, which show improved results compared to current meth- Evaluation results underscore the framework’s efficacy in minimizing
ods. The effective management of spectrum resources is emphasized as system failures, improving energy efficiency, and boosting QoE. This
essential for meeting the needs of emerging services and applications underscores the crucial role of the bi-objective SHH in fine-tuning trans-
in 6G networks, including high transmission rates, high reliability, low mission power optimization within 6G networks. The MHAs utilized in
latency, and ubiquitous connections for the IoE. load balancing and user association are summarized in Table 6.
Unmanned Aerial Vehicles (UAVs), which are enabled by 6G, can
provide cost- and time-effective solutions for a wide range of appli- 4.3. MIMO detection, channel estimation, and precoding
cations. EE and data collection are regarded as the two most critical
operations in UAV networks concerning providing high-quality network MIMO is a communication system that uses multiple antennas at the
connections. The work in [114] proposes the AECCDC-UAV technique transmitter and receiver to improve the signal quality and increase the
to reduce the amount of energy used by partitioning the UAV network data transmission rate. MIMO technology is used in many wireless com-
into separate clusters and allocating a cluster head (CH) to each of munication systems, such as Wi-Fi, LTE, 5G, and 6G. In MIMO systems,
those clusters. The QOSSO algorithm is adapted for selecting the CHs the receiver needs to estimate the transmitted symbols from multiple
and constructing clusters. The delivery of personalized QoS will be antennas in the presence of interference and noise. The receiver can use
challenging for future cellular networks due to the limited availability various detection methods, such as MLD, ZF, and MMSE, to estimate the
of radio resources. transmitted symbols. The choice of detection method depends on the
A passive RIS and decode-and-forward relaying are both compo- system requirements, such as the desired trade-off between complexity
nents of the hybrid transmission systems proposed in [115]. Next- and performance. MLD works by finding the transmitted signal vector
generation wireless networks represent a coverage extension option most likely to have been sent, given the received signals and knowledge
that is adaptable, economical, and efficient in terms of energy con- of the channel. This method is computationally complex, but it provides
sumption. Tellurium fiber is used here as a fiber gain medium, and a reliable and efficient way of detecting the transmitted signals in
an updated PSO technique is presented as a solution. It has the capa- MIMO systems. For example, with eight transmitter antennas and using
bility of fast and precisely optimizing the pump power as well as the 64-QAM modulation, the number of possible signal combinations can
wavelength of a Raman fiber amplifier. be large, making it computationally intensive to obtain the optimal
The work in [116] presents the HIOA algorithm, an optimization signal. In this example, to support 16-QAM modulation, the number
solution for reducing energy consumption in sensor-enabled IoT net- of enumerations required to obtain the optimal signal would be ap-
works. It combines fuzzy logic, CSO, and GA to form clusters and proximately 168 , which is around 4.295 billion. This highlights the
minimize energy expenditure. Simulation results showed that the HIOA importance of considering the complexity of the detection method
outperformed traditional solutions in terms of network lifetime, aver- when designing MIMO systems. In practice, more efficient algorithms
age residual energy, and node death percentage, with reductions of and hardware optimization can be used to reduce the computational
23%, 19%, and 12%, respectively. complexity, but the number of enumerations can still be quite large,
The work in [117] proposes two algorithms for reducing the trans- especially for high-dimensional MIMO systems.
mit power of IoT cellular networks while meeting the minimum rate
requirements for all users. The algorithms use a two-step process in- 4.3.1. Optimization of MIMO detection and communication performance in
volving sub-band assignment via GA and unsupervised learning via 6G wireless networks using ML and hybrid algorithms
fully connected deep neural networks. Simulation results demonstrate MHAs can be considered as alternative approaches for solving the
that this two-step algorithm has near-optimal performance and lower MIMO detection problem. The Orbital Angular Momentum (OAM)
computational complexity than exhaustive search methods. In [118], a technique is a promising contender for 6G wireless communication.
study was conducted on the problem of EE resource allocation in wire- In order to raise the maximum capacity of the transmission system,
less communication networks using CoMP technology. A GA algorithm a brand-new method for improving the OAM state interval has been
was proposed to maximize EE and was tested using a multi-objective devised in [122]. An OAM-MIMO transmission system has the potential
optimization problem (MOOP). Results showed that the proposed GA- to enhance its capacity by up to 28.7% and 3203.3%, respectively,
based solution was more effective than the benchmark solution. when the angle of deflection between the transmitting and receiving
Recently, various studies have contributed innovative approaches antennas is −24 dB and −5 dB, respectively. The spatial phase profiles
to optimizing 6G network resource utilization. One such study [119] necessary to generate arbitrary radiation patterns in real-time can be
focuses on accommodating diverse services and customized network predicted using deep reinforcement learning. In [123], the artificial

15
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 6
Comparison of various MHAs used in load balancing and user association.
Refs. Year Optimization Application Pros Cons
techniques
[103] 2018 GWO User association scheme for ultra-dense Proposed solution framework is based The paper only compares the proposed
THz networks that takes into account on a powerful optimizer, the GWO, that solution framework with the PSO
various factors such as THz channels, returns an optimal user association table. approach and does not consider other
antenna directivity, device positions, and possible solutions or optimizers.
user’s minimum rate requirements.
[106] 2019 GA, PSO Comparing the performance of GA and Improves user satisfaction with provided Further research is needed to develop
PSO with flying BS (FlyBSs). data rates and increases total throughput solutions for high mobility and efficient
of all users compared to k-means. interference mitigation among base
stations.
[105] 2020 PSO Increasing the number of users that The proposed scheme offers a better The paper does not consider other
receive satisfactory downlink balance in the network without possible solutions for user association
requirements by combining the cell resolving complex optimization problems and load balancing in HetNets, such as
range expansion technique. or using excessive signaling. game theory or ML algorithms.
[99] 2021 MOMGWO Optimizing spectrum sensing parameters The MOMGWO algorithm is designed to The paper only compares the proposed
in CRN. balance exploration and exploitation, MOMGWO algorithm with a limited
leading to global optima. number of existing algorithms and does
not consider other possible solutions.
[111] 2021 MPA Minimizing transmission power in 6G Improves the power efficiency of user The difference in results between MPA
networks with Intent-Based Network information collection in IBN. and MPGND may appear small but
(IBN) architecture. becomes significant with an increase in
cluster numbers.
[115] 2021 PSO Optimizing the pump power and The proposed hybrid transmission No comparison is made with other
wavelength of a Raman fiber amplifier schemes combine RIS and existing solutions for coverage extension
using the PSO technique. decode-and-forward relaying in a in future-generation wireless networks.
synergistic manner.
[100] 2022 DBSCAN Improving spectrum efficiency in Improved security with the use of The impact of increased complexity from
wireless communication systems by blockchain for authentication of SUs, the integration of multiple technologies
using improved DBSCAN method for PUs, and pedestrians and spectrum on the overall system performance and
clustering authenticated SUs and using utilization is improved by using efficiency is not evaluated.
multi-agent-based categorical DQN in Lite-CNN for spectrum sensing, and
hybrid beamforming. encrypted communication of sensing
reports to the FC
[104] 2022 GA, PSO Determining the optimal The proposed concepts in the letter can The paper only focuses on the 2 × 2
hyperparameters for using DNN to help antenna designers to construct a antenna array, so the validity of the
model the behavior of antennas by using reliable DNN for antenna modeling. proposed concepts for larger antenna
GA and PSO algorithm. arrays or different types of antennas is
not known, and it only focuses on the
prediction of three performance metrics
(S11, gain, and RPs), so the ability to
predict other specifications is not
discussed.
[107] 2022 Improved GA Improving the load-balancing schemes The proposed load balancing scheme has The paper does not mention any
for satellite networks using the proposed many potential applications in S-IoT-N, limitations or weaknesses of the
novel load-balancing scheme. for example, in disaster recovery or proposed load balancing scheme.
timely transmission of critical
information.
[114] 2022 QOSSO Selecting the CHs (Cluster Head) and Achieves energy-efficient data collection Still, the limitations or drawbacks
constructing clusters using the adapted for UAVs in a 6G environment. needed to be fulfilled.
QOSSO.
[101] 2022 PSO Optimizing beamforming and common The model is robust to different CSO Still, the limitations or drawbacks
rate vectors in CRN using a new RSMA parameters and produces a near-optimal needed to be fulfilled.
(Resource Allocation in CRN) scheme solution.
that employs PSO.
[116] 2022 HIOA, CSO, Minimizing energy consumption in The model is robust to different CSO The proposed model has no prediction
GA sensor-enabled IoT networks using an parameters and produces a near-optimal ability, so it cannot be used in a
optimization algorithm that combines solution. sustainable network where electric
fuzzy logic, CSO, and GA. vehicles are used as nodes, and it
requires the addition of a neural
network or DL approach to include
prediction ability for the formation of
clusters based on past experience.
[117] 2022 GA Assigning sub-bands and controlling The proposed algorithm minimizes the The paper does not mention any
transmit power by combining GA and transmit power in NOMA-based uplink limitations or drawbacks of the proposed
unsupervised learning. IoT cellular networks while meeting the algorithm.
minimum rate constraint.
(continued on next page)

16
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 6 (continued).
Refs. Year Optimization Application Pros Cons
techniques
[118] 2022 GA Minimizing power spectral density in the The proposed approach solves the EE The improvement of NSGA-II is needed
stopband while meeting constraints on resource allocation problem for to solve the two matrices simultaneously
peak-to-average power ratio using the downlink OFDMA HetNets, which is an to obtain better results in future work.
proposed GA-based scheme. NP-hard issue.
[120] 2022 GA, ABC 1. UE-BS association: The GA is used to The proposed approach decouples the The optimization problem is complex,
determine the association of user optimization problem into two and the proposed solution may fall into
equipment (UE) to BS in a cellular sub-problems, which makes it easier to sub-optimal values.
network. tackle and accelerates the convergence.
[102] 2023 EDHO Optimizing the use of available The paper presents a novel approach for The results are only evaluated in
spectrums in 6G network using EDHO spectrum sensing in 6G networks that comparison to other techniques and not
algorithm. aims to manage the available spectrum in a real-world scenario, so it is
to enhance network efficiency uncertain how well the proposed
effectively. technique would perform in practical
implementation.
[113] 2023 DOA Minimize energy usage and maximize The proposed technique provides an The paper does not provide any
the lifetime of IoT devices through effective solution for energy comparison with other techniques in
cluster-based routing. management in 6G networks. terms of other performance metrics,
such as reliability and delay.

neural network is trained, which does so in a flexible manner in real- is supported by Monte-Carlo simulations, which show that an active RIS
time. It offers a technique for intelligent and self-adaptive beamforming scheme can improve the secrecy performance of the system, especially
underpinned by deep reinforcement learning. It has the potential to when the users have limited transmission power.
provide multi-user large MIMO systems for wireless communications at The work in [128] focuses on utilizing RIS to enhance the per-
6G THz. The revolutionary technology known as RIS uses metasurface formance of THz communication systems. The authors first use mea-
to reflect the signal from the source node to the destination node. The surements to establish that the fluctuating two-ray distribution can
RIS can be conceptualized as an inactive relay that connects the source model the small-scale amplitude fading of THz signals. Then, a novel
node and the destination node. In contrast, in typical networks, relay approach based on SI is proposed to optimize the phase shifts of the
nodes must remain active and waste energy while transmitting signals RIS elements under discrete constraints. This method aims to minimize
or data. Comparative analysis of RIS and active relay performance in a the end-to-end SNDR to improve the system’s performance. To assess
general MIMO system is presented in [124]. the system’s performance, closed-form expressions for the outage prob-
The work in [125] proposes a GR-GA for use in THz MIMO systems ability are derived, which is the probability that the SNDR falls below
with an IRS. The algorithm improves communication in THz systems by a certain threshold. Monte Carlo simulations are performed to validate
reserving good genes and eliminating inferior genes during the iterative the analysis as well as explain how the various system parameters affect
process. Simulation results show that the AGR-GA has slightly lower the performance. The results show that the effect of path loss on the
performance than an exhaustive search scheme but with significantly system’s performance is more significant compared to other factors.
reduced complexity. This algorithm is intended for use in 6G wireless
communication systems, where THz waves with strong directionality
4.3.2. Optimizing antenna designs
can be easily blocked by indoor obstacles, leading to communication
An antenna design is a critical aspect in the development of 6G
interruption.
networks. Antennas play a key role in determining a communication
In [125], a GR-GA was proposed for THz MIMO systems with an IRS.
system’s coverage, capacity, and data rates. In 6G, innovative an-
The AGR-GA improves communication in THz systems by preserving
tenna designs will be required to support the expected increase in
advantageous genes and discarding inferior genes during iteration.
data rates and the use of higher frequency bands, such as mmWave
The simulation results indicated that AGR-GA slightly underperformed
communications. New antenna designs will also need to address the
compared to an exhaustive search scheme but had significantly lower
challenges posed by the growing demand for high-speed, low-latency
complexity. The algorithm was designed for use in 6G wireless commu-
nication systems where THz waves with high directionality can easily communications and the need for more efficient use of limited spectrum
obstruct indoor obstacles, causing communication disruptions. resources.
The study in [126] investigates the integration of an IRS into Innovative antenna designs, such as massive MIMO, beamforming,
mmWave MIMO systems to boost communication performance and and active antennas, will play a vital role in enabling 6G networks
wireless propagation flexibility. The authors put forth two algorithms to meet the performance requirements of next-generation applications,
to optimize the passive beamforming, combiner, precoder, and power such as autonomous vehicles, virtual and augmented reality, and the
allocation in these systems with finite resolution: aISU algorithm with IOT. The integration of these technologies will allow 6G networks
lower complexity and a GA with enhanced performance. The sim- to support higher data rates, increase capacity, and reduce latency
ulation showed that IRS-aided systems could significantly enhance while also improving energy efficiency and reducing the environmental
communication performance and coverage compared to traditional sys- impact of communication systems. The antenna designs in 6G networks
tems. However, adding more IRS elements only guarantees continuous will also need to support integrating multiple technologies, such as 5G,
improvement when the total IRS area is fixed. Wi-Fi, and satellite communications, to ensure seamless and efficient
The work in [127] examines the effect of multiple passive eaves- connectivity for end-users.
droppers on the security of an uplink MU-MISO communication system Several types of antennas have been designed with the assistance
assisted by an RIS. The focus is on the influence of phase shift errors of artificial neural networks (ANN), which have also been used to op-
on the system’s security, and three different RIS configurations are an- timize the designs. Optimizing the radiation patterns of uniform linear
alyzed: passive RIS, active RIS, and active RIS with EH-RIS. To improve and nonlinear array (ULA) smart antennas is the focus in [129], which
the system performance, the phase shifts at the RIS are optimized using presents a novel artificial neural network (ANN) technique known as
GA. The paper also considers the best RIS configuration selection for a the single neuron weight optimization model (SNWOM). One of the
given target secrecy rate and available power to the users. The analysis benefits of SNWOM is that it allows for quick convergences while

17
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

also requiring less technology to achieve excellent performance. Fu- aims to maximize the asymptotic capacity of the system. They propose
ture 6G systems may implement Federated Learning (FL) as a viable a low-complexity, near-optimal algorithm to solve this problem, which
model for enabling network edge intelligence. Due to the fact that the involves a convex relaxation technique. The paper includes numerical
training data obtained are not independent and identically distributed results that support the effectiveness of their proposed design of RIS
(non-IID), the system’s performance suffered significantly. In wireless and demonstrate that it leads to a significant improvement in the
networks, data sharing and FL are combined in order to bring about asymptotic capacity of the system.
a convergence of the distribution differences. The work in [130] An A new method to improve the performance of IRS in MIMO wireless
algorithm for joint optimization is developed so that a sophisticated communication systems is proposed in [136]. It uses a two-timescale
balance may be maintained between the cost of the model and its beamforming approach to overcome the high cost of acquiring chan-
accuracy. The results of the simulation are presented in order to analyze nel state information (CSI). This approach maximizes the system’s
the effectiveness of the proposed strategies. The work in [131] offers average achievable rate (AAR) by infrequently configuring the IRS
a novel architecture for antenna arrays that can be used in smart 6G with statistical CSI and frequently updating the base station precoder
networks with unequal node distribution. The rapid electromagnetic and power allocation with quickly outdated I-CSI. Two algorithms are
code of the multilevel fast multipole is utilized to simulate the radi- presented, one using mini-batch sampling and PSO and another using
ation properties of unevenly dispersed antenna arrays. The technique model-driven PSO. The simulation results show that this two-timescale
of SA, sometimes known as SA, has seen widespread use in manag- beamforming strategy achieves higher AAR and efficiency than other
ing multiscale nonlinear problems. The results of the simulations and methods.
the tests reveal that the performance of the antenna array that was The work in [137] explores ways to increase transmission reliability
recommended is adequate and accurate. and energy efficiency in satellite communication systems located in
The paper [132] outlines a design methodology for single-feed low earth orbit. The proposed solution is a combination of incremental
multi-beam transmit array antennas using quad-beam designs as case redundancy, hybrid automatic repeat request (IR-HARQ) and variable-
studies. The authors use a quadruple layer of double square loops for power allocation. A numerical integration method based on a fast
the transmit array elements and PSO for aperture phase synthesis. They Fourier transform (NI-FFT), and an EE variable-power allocation algo-
also evaluate different far-field pattern masks and fitness functions. The rithm (VPA-GA) are also proposed. Simulations show that the technique
prototype, operating in the Ku-band, generates four symmetric beams meets transmission reliability requirements and is more energy efficient
with 50◦ elevation separation and a gain of around 23 dB. The impact than existing methods like IR-HARQ with equal power (IR-HARQ-EP)
of unit-cell analysis approximations is also examined. and variable-power allocation. The MHAs utilized in MIMO detection,
The work in [133] presents a design methodology for MIMO an- channel estimation, and precoding are summarized in Table 7.
tennas using GA optimization. MIMO technology is widely used in
5G mobile networks, and recent research has aimed to design MIMO 4.4. Security and privacy issues
antenna systems that offer benefits such as low bit errors, high through-
put, and low latency. The design process begins with MIMO antenna Several security and privacy issues need to be considered in imple-
configuration and uses GA to optimize the parameters. The GA method menting channel allocation and power control in 6G networks. These
leads to a valid electromagnetic-verified post-layout and reduces the include:
designer’s effort. The proposed method is demonstrated by designing
and optimizing a 400 MHz bandwidth MIMO antenna in the 4.1 GHz • Location tracking [138]: Resource allocation algorithms in 6G
to 4.5 GHz frequency range. networks could be used to track the location of users, which could
be a privacy concern if the information is used without the user’s
4.3.3. Optimizing performance of IRS in MIMO wireless communication consent or shared with unauthorized parties.
systems • User profiling [139]: Resource allocation algorithms could also
The work in [134] presents a technique for mmWave massive be used to build profiles of users based on their communication
MIMO systems that utilizes a RIS and angular-based hybrid beam- patterns and behaviors, which could also be a privacy concern if
forming (AB-HBF). The proposed architecture has three stages: RF the information is used without the user’s consent or shared with
beamformer, baseband (BB) precoder/combiner, and RIS phase shift unauthorized parties.
design. In the first stage, RF beamformers are designed to reduce • Eavesdropping [73,140]: Channel allocation techniques may be
the number of RF chains and channel estimation overhead. They are used to increase the security of communications, but malicious
using a 3D geometry-based mmWave channel model with slow, time- actors can also use them to intercept and decode sensitive infor-
varying angular parameters of the channel. Then, the second stage, BB mation.
precoder/combiner, is designed by exploiting the reduced-size effective • Denial of Service (DoS) attacks [141]: Channel allocation tech-
channel seen from the BB stages. Lastly, the phase shifts of the RIS are niques may be vulnerable to DoS attacks, which can disrupt
adjusted to maximize the achievable rate of the system via the nature- communication and cause the network to fail.
inspired PSO algorithm. Simulation results demonstrate that using RIS • Data leakage: Channel allocation techniques may be used to
in AB-HBF systems provides more promising advantages regarding protect data from unauthorized access, but they can also leak data
reliability and flexibility in system design. to unauthorized parties.
A technique for mmWave massive MIMO systems using a RIS and • Man-in-the-middle attacks [142]: Channel allocation techniques
AB-HBF was proposed in [134]. It had three stages: RF beamformer, BB may be vulnerable to man-in-the-middle attacks, in which a ma-
precoder/combiner, and RIS phase shift design. The RIS phase shifts licious actor intercepts and alters communications between two
were optimized to maximize the achievable rate of the system using parties.
PSO. Simulations showed that using RIS in AB-HBF systems improved • Privacy concerns [143]: Channel allocation techniques may be
reliability and design flexibility. used to protect users’ privacy, but they can also be used to track
The work in [135] presents research on using RIS in a specific and monitor users’ activities.
type of wireless communication system called parallel MIMO VLC. The • Interference [144]: Unauthorized access to the channel alloca-
authors propose using RIS to configure the wireless propagation envi- tion process can lead to radio interference, which can cause a
ronment in VLC systems, which they argue is a promising trend for the reduction in communication quality and network performance.
future. They develop a mathematical model of the VLC channel in the • Authentication and authorization attacks by posing as a legitimate
presence of RIS and use it to formulate an optimization problem that device or network [145].

18
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 7
Comparison of various MHAs used in MIMO detection, channel estimation, and precoding.
Refs. Year Optimization Application Pros Cons
techniques
[125] 2020 GA Improving communication in THz The proposed AGR-GA algorithm The proposed algorithm only slightly
MIMO systems with an IRS. decreases computational burden endures 0.98 bps/Hz performance
significantly. compromise.
[126] 2022 GA Optimizing Beamforming and The study focuses on a promising The system performance may not
Power Allocation in mmWave and technology, IRS, for improving increase continuously with the number
MIMO systems. communication performance in of IRS elements when the total area of
6G wireless networks. the IRS is fixed.
[127] 2022 GA Optimizing the phase shifts of the The study optimizes the phase The study only considers the up-link
RIS to enhance the system shifts at the RIS to enhance the communication scenario and does not
performance. system performance, which is a explore the downlink scenario, which
key contribution. may limit the applicability of the results.
In addition, it only considers the
RIS-aided MU-MISO communication
systems and the results may not apply
to other communication systems.
[134] 2022 PSO Optimizing the phase shifts The proposed RIS-AB-HBF The proposed technique is only a
technique reduces the number of simulation study, and there is no
RF chains and the channel experimental validation of its
estimation overhead. performance.
[136] 2022 PSO Optimizing the IRS configuration Reduces the channel training and Still, the limitations or drawbacks
with reduced channel samples. estimation overhead and needed to be fulfilled.
optimizes the IRS configuration
with the reduced channel
samples.
[137] 2022 GA Optimizing the energy-efficient Reliable and energy-efficient The paper only focuses on energy
variable-power transmission is achieved in LEO efficiency improvement and does not
satellite communication systems consider other factors such as cost,
through the IR-HARQ-VPA complexity, or compatibility with
scheme. existing systems.
[133] 2023 GA Designing MIMO antennas. The paper presents an The paper only focuses on optimizing
optimization-based methodology MIMO antenna design and does not
for designing MIMO antennas, discuss other aspects such as
which can lead to a valid implementation, testing, or performance
electromagnetic (EM) verified evaluation. In addition, the methodology
post-layout generation. is limited to a specific frequency band,
so it is unclear if it can be applied to
other frequency ranges.

• Jamming attacks that disrupt communication and power con- 4.5. Lessons learned
trol [146].
• Insider threats caused by malicious insiders with access to power The main lessons learned from the studies of optimization algo-
control resources [147]. rithms in the context of wireless communication and 6G networks
are:
It is important to note that these security and privacy issues can
be mitigated by using secure channel allocation techniques and power • Optimization algorithms can be used to improve various aspects
control by implementing robust security protocols in 6G networks. of wireless systems, such as spectrum efficiency, resource man-
In order to implement channel allocation and power control in agement, communication reliability, energy conservation, and
6G networks, it is crucial to consider security and privacy as critical system capacity.
factors. There are various issues that need to be taken into account, • Different optimization algorithms such as MWOA, GWO, PSO,
such as location tracking, user profiling, eavesdropping, DoS attacks, quantum-inspired termite colony optimization, GA, ABC algo-
data leakage, man-in-the-middle attacks, privacy concerns, interfer- rithm, and salp swarm algorithm have been proposed and eval-
ence, authentication and authorization attacks, jamming attacks, and uated through simulations for various wireless systems such as
insider threats. When it comes to load balancing and user association,
ultra-dense THz networks, full-duplex massive MIMO systems,
data integrity and confidentiality must be ensured through encryption
B5G wireless networks, IoT, non-orthogonal multiple access
and secure protocols, potential DoS attacks should be mitigated, user
(NOMA) in small-cell networks, and RIS systems.
privacy protected, and the complexity of handling a large number of
• Load-balancing and user association mechanisms can be designed
IoT devices taken into consideration.
using frameworks from stochastic geometry, game theory, and
Furthermore, MIMO detection in 6G networks also poses security
and privacy concerns, including eavesdropping, jamming, spoofing, convex optimization.
privacy breaches, security of MIMO detection devices, and trust man- • Spectrum sensing is a critical enabler for CRN, and modified
agement. Eavesdropping can compromise the confidentiality of com- versions of algorithms such as GWO can be used to attain global
munication, jamming can hinder signal detection, spoofing can lead optimums by balancing exploration and exploitation.
to false signals being detected, unauthorized access to sensitive data • In 6G networks, the use of CRN and cooperative spectrum sensing
can compromise privacy, MIMO detection devices may be suscepti- is essential for improving the user experience and addressing the
ble to hacking, and a trust management mechanism is necessary to challenge of limited spectrum availability.
authenticate devices and users accessing the MIMO detection system. • For ultra-dense THz networks, a throughput-aware user associa-
Addressing these issues is vital to guarantee the security and privacy of tion scheme can be proposed using GWO, which outperforms the
6G networks. commonly used PSO approach.

19
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

power, and scheduling decisions, such as where to execute tasks and

what to cache. The objectives of these optimizations include energy
consumption minimization and delay minimization.

5.1.1. Communication in vehicular networks, caching, and joint computing

In the context of vehicular networks, joint computing, caching, and
communication, a set of solutions have been proposed in [150,151]
in the context of vehicular networks, joint computing, caching, and
communication. Both studies adopt a two-time scale strategy for opti-
mization and employ different optimization algorithms - [150] utilizes
PSO and deep reinforcement learning, while [151] uses a hybrid GA
with a mountain-climbing operator. The aim of both studies is to
decrease the delay and energy consumption of computing tasks by
utilizing the idle computing resources of parked vehicles along the
roadsides as the task offloading platform. The simulation results of
both studies show the success of the proposed solutions in reducing
the overall system cost during task offloading.

5.1.2. Joint container placement and local CPU frequencies, and task
scheduling
Joint Container Placement and Local CPU Frequencies are key as-
pects of resource management in the 6G network. Container placement
Fig. 8. Ubiquitous participants. involves efficiently allocating containers, which are isolated environ-
ments for running applications, across different nodes in the network.
This helps to ensure that the resources are utilized optimally and the
• Determining optimal hyperparameters for DNN used in modeling applications are deployed in a scalable manner. Local CPU frequencies,
the behavior of antennas for next-generation mobile communica- on the other hand, refer to the speed at which a CPU operates, and
tion networks can be challenging; various stochastic methods such it plays a crucial role in determining the performance of applications
as trial-error, grid search, random search, Bayesian optimization, running on the network. Task scheduling is scheduling tasks to be
GA, and PSO can be used for optimal hyperparameter selection. executed by the network resources. In 6G, this is achieved by allocating
• Managing traffic demand in wireless communication systems, tasks to different containers and managing the local CPU frequencies
particularly in light of the growing complexity of HetNets, can to ensure optimal performance. The objective of task scheduling is to
be challenging. Proposed solutions, such as coupling the CRE ensure that the tasks are executed efficiently and in a timely manner
technique with a PSO algorithm, can effectively reduce network while also considering the network’s resource constraints.
imbalance and meet the data traffic requirements of mobile users. Integrating MHAs with 6G makes it possible to achieve optimal
• To increase the density of BS to meet the time-varying require- container placement and effective management of local CPU frequen-
ments of users, flying BS (FlyBSs) mounted on UAVs can be cies and task scheduling. The algorithms can dynamically adjust the
used as a cost-effective substitute for the dense deployment of container placement and CPU frequencies based on the changing net-
conventional static BSs. work conditions, ensuring that the resources are utilized optimally.
• MHAs can serve as substitute solutions for the MIMO detec- Additionally, they can make real-time decisions on task scheduling,
tion issues, as they have lower complexity and provide good taking into account factors such as task priority, resource availability,
performance. and execution time.
• MHAs such as PSO, GA, and SA have also been proposed for The integration of MHAs with 6G can significantly improve the
antenna design in 6G systems. network’s performance and efficiency, ensuring that applications run
smoothly and effectively; for example, in [152], and [153] concentrate
5. Edge computing and wireless caching on scheduling tasks in MEC systems. The former study aims to max-
imize the number of completed tasks while complying with resource
The increasing demand for high-bandwidth and low-latency appli- utilization and request load constraints. The latter focuses on mini-
cations has led to the emergence of caching at the wireless edge as mizing task completion time and mobile device energy consumption,
a potential solution in 6G networks [148]. Caching involves storing particularly for security-sensitive tasks that entail additional runtime
frequently requested data at the edge of the network, closer to the end and energy costs from data encryption. Both papers offer solutions to
user, in order to reduce the time required to access the data and im- address the mobility and unpredictability of edge nodes. In [152], the
prove the performance of latency-sensitive applications. Furthermore, authors propose a combined problem of container placement and task
advancements in edge computing, such as fog computing and MEC, scheduling using an optimization algorithm, while [153] presents two
have been made to shift cloud resources from central clouds to net- slow-movement PSO algorithms to tackle the scheduling problem. Both
work edges, thereby latency-sensitive applications, compute-intensive, studies demonstrate that their proposed solutions surpass traditional
facilitating heterogeneous devices (see Fig. 8). methods in terms of effectiveness and efficiency.
In [154], the WOA is applied to handle the combined local CPU
5.1. Joint resource optimization frequency and task scheduling problem. The results reveal that WOA
results in a shorter task completion time compared to PSO, while both
6G networks support not only communication services but also algorithms have similar energy usage. Several studies have investigated
caching, computing, and control, as highlighted in [149]. This poses the use of optimization algorithms for task scheduling in cloud com-
new optimization challenges, as the variables to be optimized include puting systems. In [155], a hybrid optimization algorithm is proposed
computing resources, such as edge computing capabilities, local CPU for scheduling jobs with minimal waiting time while also considering
frequency, and radio resources, such as subcarrier allocation, transmit other parameters such as production time, efficiency and utilization.

20
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Simulation results show that this method outperforms conventional Overall, the studies demonstrate the effectiveness of using PSO
ACO and PSO algorithms. algorithms in optimizing resource allocation and computation offload-
In [156], a two-step hybrid method is proposed that uses a GA ing in various systems, including hierarchical edge computing, MEC,
and Energy-Conscious Scheduling Heuristic to address challenges with and VANETs, resulting in reduced computation overhead, increased
scheduling and energy consumption in cloud computing systems. Simu- resource utilization, and improved efficiency.
lation results demonstrate that this proposed algorithm performs better
than other methods. 5.1.4. MHAs for joint optimization in NOMA, D2D, UAV, and blockchain
In [157], the Monarch Butterfly Optimization Algorithm (MBOA) is With the emergence of 6G, there is a growing need for effec-
implemented to solve the cloudlet scheduling problem. The hybridized tive optimization strategies to address the challenges posed by var-
version of MBOA is tested on benchmark functions and artificial/real ious new communication paradigms such as OMA, D2D, UAV, and
data sets and is shown to have the potential for solving task scheduling Blockchain [162]. Integrating MHAs with 6G for joint optimization
in cloud environments. The results demonstrate that the hybridized in these domains can help improve the efficiency and performance
version is superior to the original one, which has positive implications of communication systems and enhance the security and privacy of
for cloud management. data transmission. By jointly optimizing the resources in NOMA, D2D,
UAV, and Blockchain systems, MHAs can help ensure that 6G net-
5.1.3. Hierarchical edge computing systems works can effectively support the increasing demand for data-intensive
Hierarchical Edge Computing Systems refer to a distributed comput- applications and services. For example, in [163], PSO is used to max-
ing architecture that places computation and data storage capabilities imize network energy efficiency in a dense HetNet environment that
at the network’s edge, closer to the data source. This hierarchy of combines NOMA and MEC. The optimization involves optimizing the
computing resources enables efficient processing of massive amounts total utility of all BSs through a user association mechanism followed
of data generated by IoT devices and other edge devices, reducing the by solving a resource allocation problem with PSO. The PSO-based
amount of data transmitted to the cloud and improving response times. algorithm proposed outperforms existing minimum-distance and OMA
The hierarchical structure of the system allows for flexible and scalable algorithms. The work in [164] employs a binary version of WOA to
deployment of resources while preserving the benefits of centralized solve the mixed-integer non-linear programming problem of resource
management and control provided by cloud computing. This architec- allocation and computation offloading in NOMA-MEC systems with
ture is designed to address the challenges of managing the growing self-interference.
number of devices and the data they generate, providing a solution that The work in [165] combines PSO and SA to minimize energy con-
is both efficient and cost-effective. sumption by computing resources and to optimize radio in UAV-assisted
Hierarchical Edge Computing Systems play a crucial role in devel- edge computing systems while considering spectrum and latency re-
oping and deploying 6G networks. The hierarchical edge computing strictions. The work in [166] explores a learning-based CPSO approach
architecture in 6G networks enhances the processing of data generated to handle resource allocation and task priority in MEC-powered UAV
at the network’s edge, enabling real-time decision-making and reducing forest fire detection systems. The results indicate that the proposed
the latency and bandwidth constraints that can limit the performance approach outperforms several benchmarks regarding best and average
of traditional cloud-based systems. The hierarchical approach in 6G response time. The work in [167] explores a computing resource al-
networks also enables the integration of edge devices, such as sensors, location and joint server selection problem in blockchain-based edge
cameras, and robotics, into a unified system, supporting new use cases computing systems and resolves it using various MHAs, such as GA,
and applications that were not possible with previous generations of SA, ABC, and PSO.
wireless technology. Overall, the hierarchical edge computing systems
in 6G networks provide a foundation for developing a more intelligent, 5.2. Wireless caching
responsive, and secure edge computing environment, enabling new and
innovative solutions in the industry, healthcare, and transportation. Caching has been used for decades in various areas, such as memory
With the advent of 6G, the next generation of wireless communica- caching in operating systems and web caching. However, centralized
tion technology, Hierarchical Edge Computing Systems (HECS), has caching cannot meet user needs with increased traffic, higher service
emerged as a promising solution to address the challenges posed by the demands, and network density in 6G. To address this, the concept of
increasing demand for low-latency, high-bandwidth communication. wireless/edge caching has been recently explored. The main issues in
Integrating MHAs with 6G in HECS can help optimize the allocation wireless caching are content delivery and content placement. Content
of resources in these systems, improve communication efficiency, and delivery concerns getting the content to the users, while content place-
enhance the quality of services provided to end-users. This integration ment involves determining the location and size of each content chunk.
is expected to play a crucial role in addressing the complex chal- MHAs have been applied to improve performance in wireless caching
lenges posed by the growing demand for data-intensive applications systems.
and services in 6G networks. The authors of [168] propose a multi-objective random caching
The authors of [158] use a PSO-based algorithm to reduce com- scheme for an F-RAN to balance STP and delay in the wireless back-
putation overhead in a hierarchical edge computing system, showing hauled F-RAN. Using MHAs, they derive expressions for association
a significant reduction compared to baseline schemes. In [159], an probability, STP, and average delivery delay. They then present a novel
integrated APSO algorithm with dynamic programming is developed to algorithm, PMOCSA, to optimize cache placement and find the optimal
minimize computing time and service cost in MEC systems, reducing Pareto front. Results show that PMOCSA outperforms other algorithms,
task makespan by 30% and increasing resource utilization by 29% and the proposed caching scheme increases STP by up to 40% and
compared to standard algorithms. The work in [160] proposes a gene- decreases average delay by 85%. In [169], a new approach to edge-
ant colony fusion algorithm to improve the efficiency of heuristic based wireless caching that combines Network NFV and content-centric
algorithms in computing offloading, showing better performance in networking is discussed. The optimization problem of reducing the
terms of time delay, energy consumption, and objective function value total number of weighted hops in a HetNet is solved using ACO and
compared to other algorithms. Finally, the MARIA mechanism, a bio- compared with other caching methods.
inspired algorithm for optimizing computational resources in VANETs, Three papers [170–172] suggest implementing the ACO algorithm
is proposed in [161]. The mechanism is efficient compared to com- for edge computing and vehicular applications. The AVE framework
monly used techniques and is able to increase accepted services, and in [170] organizes vehicular edges through a workflow and assigns jobs
decrease refused services. with an ACO scheduling algorithm. The cloud computing collaborative

21
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

in a study to determine the sleeping status of Small Base Stations (SBSs)

by considering the energy level and amount of traffic offloaded from
the macro BS.
In summary, caching in HetNets can significantly improve cache-
based wireless system performance. Several caching mechanisms have
been proposed and studied, including utilizing full-duplex relays, mod-
eling user mobility patterns, and energy management, with the aim of
optimizing network throughput and EE.

5.3. Computation offloading

5.3.1. Computation offloading in massive connectivity and diverse QoS

environments
With the increasing demands for massive connectivity and diverse
QoS requirements, computation offloading has emerged as a critical
solution for many compute-intensive and latency-critical applications
and services, as highlighted in [180–182].
The study in [180,181] highlight that computation offloading has
become a crucial solution for various latency-critical applications and
compute-intensive services due to the growing demand for diverse QoS
requirements and massive connectivity.
According to [183], MHAs can effectively solve computation of-
Fig. 9. Diagram of a cache-enabled three-tier HetNet, displaying the potential locations floading problems with low complexity and competitive performance,
for content caching, including macro BS, relays, nearby users, and the local device. in addition to network optimization, AI, and game theory. In [184],
the use of bee swarm algorithms for computation offloading in fog
computing is investigated, resulting in significant improvement over
vehicle health monitoring system in [171] utilizes ACO for data trans- GA and PSO algorithms. Meanwhile, [185] optimizes service allocation
mission path allocation. Lastly, a modified ACO approach incorporating for industrial applications through the application of BA and binary
a neural network is proposed in [172] for multi-objective schedul- PSO.
ing. The simulation results demonstrate that the proposed ACO-based
solutions outperform other methods in various scenarios. 5.3.2. Task Allocation in Vehicular Edge Computing Systems
In two studies [173,174], the authors explore the use of UAVs to Task Allocation in Vehicular Edge Computing Systems (VECS) refers
enhance terrestrial vehicular networks in delivering high-bandwidth to assigning computing tasks to available resources in a vehicular net-
content. The goal is to increase overall network throughput by opti- work. In VECS, vehicles with computing resources and communication
mizing content placement, delivery, and UAV trajectory. The proposed capabilities act as mobile edge nodes, providing computing and storage
solution involves a two-step algorithm that uses a clustering-based services to other vehicles and nearby devices. Task allocation in VECS
two-layered method or a time-based graph decomposition method to plays a crucial role in ensuring the efficient utilization of resources,
optimize content delivery and UAV trajectory and then employs PSO reducing latency, and improving the system’s overall performance. The
to optimize content placement. The first study adds deep supervised task allocation process involves deciding which vehicle should perform
learning with a CNN model for real-time online decision-making. The a given task, taking into account various factors such as the task
second study demonstrates the efficacy of the proposed approach in requirements, vehicle capabilities, and network conditions. The increas-
adapting to dynamic network conditions through numerical results. ing demand for real-time and data-intensive applications in VECS has
Very recently, the work in [175] proposes the use of a Quantum led to the need for effective and efficient task allocation algorithms.
PSO (QPSO) algorithm and a Proportional Integral (PI) model in the The development of advanced algorithms for task allocation in VECS
context of IoV (Internet of Vehicles) to achieve the optimal solution is an active area of research aimed at improving the performance and
for active queue control and caching decision. Meanwhile, the study reliability of vehicular networks.
by Yang et al. [176] examines the impact of caching on HetNet (Het- VECS in 6G represents the next generation of intelligent and effi-
erogeneous Network) and shows that HetNet with caching can increase cient task allocation strategies in vehicular networks. With the emer-
the overall throughput by 57.3% compared to HetNet without caching. gence of 6G technology, VECS is expected to provide high-speed,
The study highlights the importance of caching as a crucial technology low-latency, and ultra-reliable communication and computing services.
in improving the performance of wireless systems. The task allocation process in VECS is critical in ensuring that these
In [177], a study was conducted on a caching mechanism that services are delivered effectively and efficiently. Integrating 6G tech-
boosts throughput by modeling user mobility patterns and full-duplex nology with task allocation algorithms in VECS offers several benefits,
relays. Fig. 9 illustrates a cache-enabled three-tier HetNet, in which such as improved communication and computation capabilities, en-
there are four possible scenarios for a content request: (1) the content hanced network reliability, and reduced latency. 6G-enabled VECS can
is available in the local cache, (2) it is available at nearby users, (3) it provide a more comprehensive and dynamic task allocation solution,
is available at nearby relays, or (4) it is available at the macro BS. considering various factors such as network conditions, user require-
In order to optimize network performance, a PSO-based approach ments, and vehicle capabilities. Integrating 6G networks with task
was proposed to maximize throughput and resolve the joint caching allocation in VECS is an active area of research aimed at developing
placement problem in [177]. This approach has low complexity and efficient and effective algorithms for optimizing resource utilization in
near-optimal performance compared to the optimal exhaustive search vehicular networks.
algorithm. In [178], user mobility patterns were also utilized to im- Integrating MHAs with 6G for Task Allocation in VECS is a crucial
prove offloading probability, but it was shown that using an adaptive research area in wireless communication and networking. This integra-
BA outperforms the PSO-based approach. A caching problem was con- tion aims to improve the performance of task allocation in VECS by
sidered in [179] to minimize total energy consumption by backhaul utilizing the capabilities of both MHAs and 6G networks. For example,
links and small cell BSs (SBSs). A binary PSO approach was investigated in [186], a vehicular edge computing system is optimized for task

22
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 8
Comparison of various MHAs used in edge computing and wireless caching.
Category Refs. Year Optimization Delay Energy Problem(s) Topic
techniques consumption
Joint resource [150] 2019 PSO ✓ ✓ Delay and energy consumption of Joint computing, Caching, and
optimization computing tasks Communication in vehicular networks
[151] 2022 HHGA ✓ ✓ Delay and energy consumption of
computing tasks
[152] 2019 ✗ ✓ Scheduling tasks Joint container placement and Task
scheduling and Local CPU frequencies
[154] 2019 WOA ✗ ✓ Task scheduling and local CPU
frequencies
[157] 2019 MBOA ✗ ✗ Cloudlet scheduling
[153] 2020 PSO ✗ ✓ Scheduling tasks
[156] 2021 GA ✗ ✓ Scheduling and energy consumption
[155] 2022 Hybrid ✓ ✗ Scheduling jobs
optimization
algorithm
[158] 2019 PSO ✗ ✗ Computation overhead Hierarchical edge computing systems
[159] 2021 APSO ✓ ✗ Computing time and service cost in MEC
(MEC) systems
[160] 2022 Gene-ant colony ✓ ✓ Computing offloading
fusion algorithm
[161] 2022 Bio-inspired ✗ ✗ Computational resources in VANETs
algorithm
[163] 2018 PSO ✗ ✓ Network energy NOMA and MEC in densely HetNets
[165] 2018 PSO and SA ✗ ✓ Radio and computing resources UAV-assisted edge computing
[164] 2020 WOA ✗ ✗ The mixed-integer non-linear NOMA-MEC systems with SI
programming
[166] 2020 CPSO ✗ ✗ The task priority and resource allocation MEC-enabled UAV forest fire detection
systems
[167] 2020 GA, SA ✗ ✗ Server selection and computing resource Blockchain-based edge computing
algorithm, ABC, allocation systems
and PSO
✗ ✗
Wireless caching [170] 2017 ACO ✗ ✗ Job assignment Wireless caching
[168] 2021 PMOCSA ✓ ✗ Cache placement
[171] 2022 ACO ✗ ✗ Data transmission paths
[172] 2022 ACO ✗ ✗ Schedule multiple activities
[174] 2019 PSO ✗ ✗ Content placement UAV-assisted mobile edge caching
[177] 2019 PSO ✗ ✗ Caching placement Mobility patterns
[177] 2019 BAT ✗ ✗ Offloading probability Mobility patterns
[175] 2020 QPSO ✗ ✗ Caching decision Wireless caching
[179] 2020 Binary PSO ✗ ✓ Energy consumption Energy management and QoS
[173] 2022 PSO ✗ ✗ Content placement UAV-assisted mobile edge caching
✗ ✗
Computation [184] 2018 Bees swarm ✗ ✗ Computation offloading Computation offloading in fog computing
offloading
[185] 2018 Binary PSO, and ✗ ✗ Service allocation Service allocation for industrial
BA applications
[186] 2018 PSO ✗ ✗ Task allocation Task allocation in vehicular edge
computing systems
[187] 2020 PSO ✗ ✗ Computation offloading Latency-sensitive IoV
[188] 2020 PSO ✓ ✗ Computation offloading Task allocation among vehicles
[189] 2020 PSO ✗ ✗ Task allocation Wireless caching
[153] 2020 PSO ✗ ✗ Task allocation Wireless caching
[190] 2020 ACO and binary ✓ ✗ Load balancing IoT offloading applications
PSO
[191] 2020 Fast artificial ✗ ✓ Task allocation Task allocation in data deduplication for
fish swarm industrial IoT applications
optimization
[192] 2019 PSO ✗ ✓ Computation offloading Offloading decisions and resource
allocation
[193] 2019 ACO ✗ ✗ Computation offloading

23
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

allocation to balance latency and quality. A new method for solving 5.5. Lessons learned
the mixed-integer task allocation problem, a binary PSO algorithm,
is introduced in the study. This approach shows less latency than The lessons learned from these studies are:
linear programming, especially when the quality requirements are less
stringent. • MHAs can effectively optimize computation offloading and re-
In recent years, the application of SI techniques in computation source allocation, leading to improved efficiency and resource
offloading has gained significant attention, as highlighted in [153,187– utilization.
189]. A reliable computation offloading scheme was proposed in [187] • MHAs have been successfully applied to optimize various technol-
by combining SDN and edge computing to handle latency-sensitive IoV ogy integrations, such as NOMA, D2D, UAV, and blockchain.
applications. The proposed fault-tolerant algorithm, which uses PSO, • MHAs can improve network EE in dense HetNet environments.
addresses the difficulties of non-convexity and NP-hardness. In [188], • MHAs can be used to resolve computing resource allocation and
the aim was to reduce the cost of task offloading by optimizing the allo- joint server selection problems in blockchain-based edge comput-
cation of tasks among vehicles. A PSO-based algorithm was proposed to ing systems.
address the challenge of idle vehicles and optimize the max–min com- • MHAs have been widely used to solve task allocation problems in
pletion time. [189] proposed an improved discrete PSO algorithm for vehicular edge computing systems and computation offloading in
task allocation, and [153] proposed two slow-movement PSO schemes. IoT applications.
In [190], the performance of two algorithms, one based on binary • There are several variants of MHAs that have been proposed and
PSO and the other on ACO, was compared to satisfy the latency shown to have better performance compared to linear program-
requirements and balance the load among fog nodes in delay-sensitive ming or other algorithms.
IoT applications. The results showed that the ACO-based approach • Multi-objective optimization algorithms, such as WOA and
outperformed the round-robin and PSO algorithms in terms of average FAFSO, have been proposed and used for data deduplication and
completion time and load balancing. resource allocation in industrial IoT applications.
Two studies [192,193] were conducted to reduce costs and en- • MHAs have been proposed to minimize energy consumption for
ergy consumption, respectively. The latter uses a bilevel optimization latency-sensitive IoT offloading and demonstrated to outperform
approach that eliminates infeasible solutions and combines ACO and several baseline schemes.
Monotonic Optimization to make offloading decisions. Simulation re-
sults showed that this method is comparable to the optimal solution Various caching mechanisms have been proposed to enhance net-
and outperforms other baseline schemes. The MHAs utilized in edge work performance, as revealed by recent studies. These approaches
computing and wireless caching are summarized in Table 8. encompass using sophisticated algorithms like PMOCSA, combining
Network NFV with content-centric networking, and adopting MHAs.
5.4. Security and privacy aspects Furthermore, the studies suggest other ways to boost network perfor-
mance, such as implementing full-duplex relays, modeling user mobility
Wireless caching and edge computing in 6G networks can raise patterns, and managing energy consumption efficiently. In addition, the
several security and privacy issues. These include data privacy, security studies demonstrate that computation offloading has gained popularity
of cached data and EDs, the authenticity of cached content and edge- as a solution for compute-intensive and latency-critical applications and
computed data, the confidentiality of communication, denial of service services, owing to the increasing demand for diverse QoS requirements
attacks, resource allocation, trust management, and data governance. and massive connectivity. MHAs, like bee swarm algorithms, have
proved effective in solving computation offloading problems, yielding
• Data privacy: Cached content and edge computing may both optimal performance and low complexity, and have been utilized in
contain and process large amounts of sensitive data, unauthorized fog computing for network optimization, game theory, and AI. These
access to which could lead to privacy breaches. algorithms have also been leveraged to optimize the allocation of
• Security of EDs: EDs, such as routers and gateways, may be services for industrial applications.
vulnerable to hacking and other types of malicious attacks, which
could compromise the security of the entire edge computing 6. MHAs for 6G applications
system.
• Authenticity of cached content and edge-computed data: Without 6.1. UAV placement and path planning
proper authentication mechanisms, it may be challenging to en-
sure that the cached content and data processed and stored at the Unmanned Aerial Vehicles (UAVs) have become increasingly popu-
edge is legitimate. lar in recent years for their various applications in agriculture, surveil-
• Confidentiality of communication: Wireless caching and edge lance, delivery, and search and rescue fields [194,195]. UAV placement
computing could potentially enable eavesdropping on the com- and path planning refer to the problem of determining the optimal
munication between the users or devices and the cache or net- location for the UAV to start and the optimal path for it to follow to
work, compromising the confidentiality of the communication. complete its mission effectively and efficiently. The UAV’s placement
• Denial of service (DoS) attacks: Caching and Edge computing and path planning impact its ability to perform the task, so it is
systems may be vulnerable to DoS attacks, where an attacker crucial to choose them wisely. The process involves considering various
floods the system with requests, causing it to become unavailable. constraints, such as communication, fuel consumption, weather, and
• Resource allocation: Wireless caching and Edge computing can safety, and finding a solution that balances these factors’ trade-offs.
lead to uneven resource distribution, leading to congestion and With advances in technology and the increasing demand for UAVs,
reduced system performance. research in UAV placement and path planning is ongoing and actively
• Trust management: A trust management mechanism is needed developing.
to identify and authenticate the devices and users accessing the Integrating MHAs with 6G technology in Unmanned Aerial Vehicle
cache and EDs and to ensure that they are authorized to access (UAV) placement and path planning is a rapidly growing field that
the content stored in the cache and processed at the edge. holds the potential for significant advancements in the performance
• Data Governance: Edge computing will require strict policies to of UAVs. As shown in Fig. 10, 6G networks offer improved commu-
ensure data is handled according to regulations and standards, nication, navigation, and control capabilities, which can enhance the
such as GDPR, HIPAA, and others. accuracy and efficiency of UAV placement and path planning. On

24
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Fig. 10. UAV path planning systems with MHAs.

the other hand, MHAs provide powerful optimization techniques for studies showed that this approach could achieve over 98% cover-
solving complex problems by mimicking natural processes such as evo- age with a minimum cell capacity outage rate of 2%. The results
lution and swarm behavior. By integrating these two technologies, the demonstrate the effectiveness of EA in solving this NP-hard problem.
UAV placement and path planning problem can be approached more The work in [201] introduces a framework for enhancing secure
intelligently and efficiently, leading to improved UAV performance and transmission in IoT networks, utilizing NOMA and an eavesdropping
reliability. UAV. The K-means cluster pair algorithm and the Block Coordinate
The work in [196] examines the difficulties of 6G mobile networks Descent (BCD) algorithm optimize channel allocation, user power, and
in supporting NG-IoT applications. To address these challenges, the UAV position. Simulation results show that this network architec-
authors suggest using DBSs for network expansion and NG-IoT require- ture has improved security, faster convergence, and lower complexity
ments. However, finding the ideal placement of DBSs is difficult due compared to previous methods.
to limited energy and signal degradation. The authors propose using
MHAs such as Cuckoo Search, EHO, GWO, MBO, SSA, and PSO to 6.2. Localization
find the optimal position of DBSs in 3D space. The performance of
these techniques is evaluated and compared in three scenarios using Localization refers to the process of determining the location of an
non-parametric statistical tests. object or a device in a given space [202]. It is a critical aspect of modern
The work in [197] proposes an enhanced CI algorithm to optimize technology and has numerous applications in areas such as navigation,
UAV-BS coverage in a 6G communications network with RIS in the 3D mapping, and the IoT. With the increasing demand for location-based
space. The proposed algorithm is tested and compared to other hybrid services and the widespread use of mobile devices, the importance
CI algorithms, showing its ability to provide complete coverage with of accurate and efficient localization has only increased. There are
fewer UAV-BSs while meeting constraints. various methods for localization, including GPS, radio-based, and com-
The work in [198] investigates the use of IWN and IIoT as part of In- puter vision-based. Each method has its advantages and disadvantages,
dustry 4.0 and proposes using EI for secure AI integration. The authors and the choice of method depends on the specific requirements and
suggest a massive twinning approach to transfer data and decision- constraints of the application. With technological advancements, new
making logic to a MEC server. Using a UAV case study, they demon- and innovative solutions for the localization problem are emerging,
strate their idea with the PSO algorithm, which results in reduced traffic and it is a dynamic and exciting field of research and development.
load and better reliability. 6G has the potential to revolutionize the way localization is carried
A new energy-efficient clustering approach for 6G UAV networks, out. With its advanced features and capabilities, 6G promises to bring
MMFO-EEC, has been proposed in [199]. It uses the MFA to select the about significant improvements in terms of accuracy, efficiency, and
optimal UAVs as cluster heads and a fitness function based on input speed of localization. The widespread adoption of the IoT and the in-
parameters for improved performance. Simulation results showed that creasing demand for high-precision location-based services have made
MMFO-EEC outperforms previous techniques. localization a crucial aspect of modern communication technology.
The work in [200] presents a two-step approach to optimize the The integration of 6G technology with other cutting-edge technologies,
number and location of MRRUs and mRRUs for next-generation net- such as AI and MHAs, is expected to bring about new and improved
works. This approach involves using low-complexity MHAs and an solutions for the localization problem. Thus, 6G technology is poised
iterative elimination method to reduce redundant cells. Simulation to play a critical role in shaping the future of localization. Integration

25
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

of MHAs with 6G technology has become a crucial area of research in 6.4. Smart city applications
the field of localization. Combining these two technologies promises to
bring about new levels of accuracy and efficiency in the localization The integration of MHAs with 6G networks in smart city appli-
process. MHAs are optimization techniques that can be applied to the cations is a rapidly growing field that holds enormous potential for
problem of localization by considering various parameters, such as the future. By combining the strengths of both MHAs and 6G tech-
radio signal strength, satellite signals, and cell tower signals. This in- nology, smart cities can benefit from innovative solutions that can
tegration is expected to provide more precise and efficient localization optimize various aspects of urban life, including traffic management,
results compared to traditional methods, making it a promising area of energy distribution as shown in Fig. 11, environmental Monitoring,
research. For example, The paper in [203] focuses on improving indoor smart buildings, and public services. Integrating these two technolo-
location estimation accuracy through using RIS. The authors explain gies will create new and innovative applications and services to help
how the RIS can adjust phase shifts to enhance localization accuracy cities become more efficient, sustainable, and livable. For example, the
based on RSS values. The paper presents an optimization problem and work in [207] presents a technique for optimizing the hyperparameter
a phase shift optimization algorithm to find the optimal phase shifts settings of ML models in an FL for smart city services using PSO.
and validate the proposed technique through analysis and simulation The effectiveness of this technique is evaluated in two case studies,
results.
one involving traffic prediction in a smart city setting and the other
Similarly, the work in [204] addresses the challenge of localiza-
involving predicting machine failures in an Industrial IoT (IIoT) setting.
tion in wireless sensor networks without GPS. The authors propose a
The results show that PSO is more efficient than the grid search
hybrid model combining fuzzy logic and a powerful ML optimization
method for tuning the hyperparameters of deep Long short-term mem-
technique to improve traditional range-free localization methods. This
ory (LSTM) models, and it does not negatively impact the accuracy of
method utilizes adaptive weights and the concept of resultant force
the models. The number of client–server communication rounds needed
vectors for improved efficiency in heterogeneous network topologies.
to find near-optimal parameters is significantly reduced when using this
The proposed method is evaluated through simulations and demon-
strates improved performance compared to state-of-the-art range-free PSO-based technique.
localization schemes based on soft computing. Several security issues must be considered for smart city applica-
As UAVs become more integrated with 6G networks for improved tions in a 6G network [208]. These include:
communication and navigation capabilities, it is essential to consider
• Privacy concerns: Smart city applications collect and process
the security risks associated with UAV placement and path planning.
large amounts of personal data, which raises concerns about data
With increased reliance on technology, UAVs are vulnerable to various
privacy and security.
security threats such as interception, jamming, hacking, and spoofing.
• Cybersecurity: Smart city systems are vulnerable to cyber attacks,
To ensure the safe and secure operation of UAVs, it is crucial to address
these security issues and develop practical solutions to mitigate the such as DoS attacks, hacking, and malware.
risks. The following are examples of security issues regarding UAV • Network security: 6G networks will likely rely on advanced tech-
placement and path planning in 6G networks [205]: nologies such as AI and ML, which malicious actors could target.
• Interoperability: Smart city systems will likely involve a wide
• Vulnerabilities in the communication link between the UAV and range of devices and systems from different vendors, which could
the ground control station, which could allow a malicious actor lead to compatibility and interoperability issues.
to intercept or disrupt the control signals. • Physical security: Smart city infrastructure, such as sensor net-
• Lack of secure authentication and authorization mechanisms for works and control systems, could be vulnerable to physical tam-
UAVs, which could allow unauthorized access to the network and pering and destruction.
sensitive data • Reliability: Smart city systems must be highly reliable to en-
• Limited computing power and storage capacity on UAVs, which sure the safety and well-being of citizens and to prevent disrup-
could make them vulnerable to attacks such as denial-of-service tions to essential services such as transportation, energy, e-health
(DoS) or malware infections services [209], and public safety.
• Interference and jamming of the UAV’s radio signals by malicious
actors, which could disrupt communication and navigation 6.5. Network slicing
• Physical attacks on the UAV itself, such as tampering or destruc-
tion of the device. Network slicing (NS) is a key enabler in 6G, providing the ability to
create multiple virtual networks over a common physical infrastructure,
6.3. Precoding optimization
each with distinct characteristics and optimized for specific use cases.
The integration of MHAs can further improve the performance and
MHAs can be combined with precoding techniques to improve the
efficiency of NS by providing intelligent solutions for resource alloca-
performance of wireless communication systems. Precoding refers to
tion, network management, and quality of service optimization. This
the process of shaping the transmission signal to enhance the quality
integration can also address some of the challenges faced by NS, such
of communication by reducing interference and increasing the signal-
to-noise ratio. By using MHAs to optimize the precoding matrix, the as network congestion, security, and reliability. This integration aims to
overall performance of the wireless communication system can be im- create a highly dynamic, flexible, and adaptable network environment
proved, leading to enhanced data rates, reduced latency, and improved that can meet the demands of a diverse range of applications and ser-
reliability. This makes MHAs and precoding a promising approach for vices. For example, the work in [75] presents a solution for allocating
next-generation wireless communication systems, including 6G. For physical resources in 5G/6G networks that utilize NS and Dynamic
example, The study in [206] presents a PS method for FTN signaling Spectrum Access. The solution allows multiple virtual networks, known
to attain shaping gain. The authors also designed a precoding MHA as slices, to coexist on the same wireless infrastructure and for spectrum
using the Bare Bones PSO (BB-PSO) approach to maximize AMI and resources to be shared to support the expected amount of mobile
achieve precoding gain. The resulting precoded FTN PS scheme out- traffic. The key challenge addressed in the paper is the Opportunistic
performs a traditional FTN system, as shown through AMI analysis and Slicing Mapping Problem, which is an NP-hard problem. To address
simulation results. This scheme demonstrates FTN gain, PS gain, and this challenge, the paper proposes a Fuzzy-Genetic solution, which uses
precoding gain, with significant improvements in AMI and system error a GA and a fuzzy system to evaluate the individuals of the GA. The
performance. The study suggests that this scheme can be used in 6G paper also defines a metric called Effective Opportunistic Use (EOU)
communication systems. to capture the degree to which opportunistic slices (OSs) can take

26
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Fig. 11. MHAs for energy resource management.

advantage of transmission opportunities. It also analyzes the proposed • Secure management of the network slices and resources to pre-
scheme in terms of interference, blocking and dropping probabilities vent unauthorized access and manipulation.
and effective opportunistic use. The proposed solution is also compared • Secure inter-slice communication and slicing management.
to the previous GA approach. Similarly, the authors in [210] focused • Handling the device and user identification and authentication.
on the utilization of the Harmony Search-based Hybrid Optimization • Handling the malicious or compromised devices in an NS envi-
(HHO) algorithm for enhancing NS in 5G/6G networks. The primary ronment.
objective of their research was to develop an effective network-slicing
method utilizing the HHO algorithm. Their proposed methodology 6.6. Clustering and routing in IoT networks
consisted of three main phases: dataset loading, HHO optimization
for hyperparameter tuning, and slicing classification using a hybrid Clustering and routing in IoT networks are essential for efficient
DL model. To evaluate the effectiveness of their proposed approach, communication and resource allocation. Clustering is the process of
the authors compared the results obtained from the HHO-based model grouping IoT devices into clusters based on specific criteria, such as
with various existing optimization, ML, and DL algorithms. The find- proximity or device capabilities [211]. The goal of clustering is to
ings demonstrated that the HHO-based model outperformed the other improve the network’s scalability, reliability, and energy efficiency.
algorithms, showcasing its potential for improving NS and delivering Routing, on the other hand, involves the selection of the best path for
high-quality services in 5G and 6G networks. As mentioned before, NS data transmission from one device to another within the network. In
is a promising technology for enabling the efficient and flexible use IoT networks, effective clustering and routing are crucial for ensuring
of network resources for different applications and services. However, a smooth and efficient data flow between devices. The efficient use of
with the increased use of NS comes the need to address its security network resources, such as energy and bandwidth, is a primary concern,
risks. The following are examples of security issues regarding NS in 6G as many IoT devices have limited resources and cannot afford to waste
networks: them. By implementing efficient clustering and routing algorithms, IoT
networks can handle large amounts of data, support a large number of
• Confidentiality and privacy concerns, as sensitive data may be devices, and maintain high levels of performance, even in challenging
transmitted over the network. network environments. Integration of MHAs with 6G in clustering and
• Authenticity and integrity of the data transmitted over the net- routing in IoT networks has become a popular research area in recent
work to prevent unauthorized access and tampering. times. The researchers in [212] present a solution to improve the
• Availability of the network to prevent denial of service attacks performance of the Internet of Energy for bus-based Vehicular Ad-Hoc
and ensure continued operation. Networks (VANETs) by introducing a street-centric routing algorithm.
• Compliance with regulations and standards to maintain the secu- The proposed algorithm employs multipath routing and ACO for im-
rity and privacy of user data. proved end-to-end delay, packet delivery ratio, computation cost, and

27
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

reduced beacon messages. Simulation results suggest that clustering Integration with SDN and NFV technologies can also aid in opti-
and multipath can enhance the algorithm’s performance. mization. These technologies allow for the gathering and maintaining
Integrating MHAs with 6G technology in IoT networks can sig- of global network state information, which can be used to optimize the
nificantly enhance network efficiency by facilitating effective cluster- system.
ing and routing of devices. By utilizing these algorithms in 6G IoT In addition, the MHA community has developed scalable techniques
networks, scalability, adaptability, and reliability can be improved, for large-scale optimization, such as the two-phase optimizer proposed
allowing for efficient resource allocation and reduced energy consump- by Lan et al. [215]. These techniques can improve the performance of
tion. Optimizing device clustering and routing can improve overall large-scale systems compared to conventional approaches.
network performance and functionality, enabling a wide range of IoT Large-scale IoT systems present unique challenges for optimization
applications. A recent study [213] investigated ways to improve call using traditional MHAs. Dividing the network into smaller regions and
quality during handover in mobile networks. Handover involves trans- integrating SDN and NFV technologies are promising solutions. The
ferring a call from one evolved node B (eNB) to another. The authors SI community has also developed scalable techniques for large-scale
proposed optimizing handover parameters such as time to trigger, HO optimization, which can improve system performance.
offset, and HO margin with a PSO algorithm and integrating a self-
organized network for mobility load balancing and PSO. Simulation 7.2. Hybrid algorithms and integration of MHAs and DL techniques
results suggest that their approach can reduce handover failure and
ping-pong rates compared to other techniques. The use of hybrid algorithms, combining two or more basic MHAs,
In IoT networks, clustering and routing are used to organize and has produced better results than using original algorithms alone. This
direct data flow between IOT devices [214]. Some potential security paper explores the potential advantages of hybrid algorithms and recent
issues for these processes in 6G networks include: research in integrating MHAs with DL techniques [216].
Studies have indicated that hybrid algorithms can have improved
• Cluster head compromise: If a malicious attacker can gain control convergence speed and reduced computational complexity compared
of a cluster head (a device that serves as a central point of to original algorithms. For instance, [217] demonstrated that a hybrid
communication within a cluster), they may be able to intercept algorithm outperformed the original ACO and PSO algorithms.
or manipulate data being transmitted within that cluster. Design choices, such as combining multiple basic MHAs, adding
• Routing attacks: Routing protocols can be vulnerable to attacks chaotic operations, and balancing exploration and exploitation, can
such as route hijacking, in which an attacker redirects data to impact the performance of hybrid algorithms. Choosing the right com-
a different destination. This can be used to intercept or disrupt bination of techniques is crucial to obtain optimal results.
communication in the network. Integrating MHAs with DL techniques is gaining popularity to im-
• Eavesdropping: An attacker may be able to intercept and listen in prove and speed up solutions. As demonstrated in a study by [218],
on wireless communications in the network, potentially accessing an MHA solution was approximated using a deep neural network and
sensitive information. showed similar results, reducing computational complexity by up to
• DoS: An attacker may be able to overwhelm a device or network 85%.
with a flood of traffic, preventing legitimate users from accessing
the network. 7.3. Federated Learning (FL) and swarm federated learning
• Privacy and Confidentiality: IoT devices are typically equipped
with sensors and cameras that collect sensitive data, which can FL is a cutting-edge solution that addresses the needs of modern
be used to track and monitor individuals. applications requiring distributed implementation of ML [219]. This
• Misconfigured security settings: Many IoT devices have weak approach allows for model training while preserving data privacy
default security settings that attackers can exploit easily if not locally, making it ideal for use cases that demand privacy protection in
correctly configured. edge networks, such as medical services and autonomous driving [220].
• Physical security: IoT devices are often small and portable, mak- However, deploying FL has challenges, including resource allocation,
ing them vulnerable to theft or tampering. secure communication, and ensuring convergence with non-convex loss
functions [221,222]. Swarm optimization schemes can be combined
7. Future research directions and challenges with FL [223] to overcome these obstacles and improve performance.
Integrating MHAs with FL can further enhance the effectiveness of FL
This section focuses on the main research challenges and potential deployment.
avenues for using MHAs in 6G.
7.4. Network authentication
7.1. Optimizing large-scale IoT systems
Network authentication is an essential process for establishing se-
The IoT has seen a surge in connected devices and systems, making cure communication between networks and devices. Its purpose is to
the network increasingly complex and heterogeneous. The complexity confirm the identity of a user or device seeking network access and
of these systems presents challenges for traditional MHAs, making op- permit entry only if authorized. Within the context of 6G networks,
timization difficult. This paper explores these challenges and presents network authentication takes on heightened significance due to the
potential solutions for optimizing large-scale IoT systems. surge in device count and elevated data rates. With 6G networks
The increasing complexity and heterogeneity of large-scale IoT sys- expected to accommodate an extensive array of connected devices and
tems pose challenges for optimization using traditional MHAs. A large support data rates of up to 1 Tbps, safeguarding these networks poses
number of variables and constraints can make optimization easier. an even greater challenge.
Additionally, traditional MHAs may need to converge more quickly to Leveraging MHAs to optimize the authentication procedure is a
be practical for real-time applications. promising approach to address the intricacies of network authenti-
One potential solution is to divide the network into smaller re- cation in 6G. This encompasses activities such as attack verification
gions and assign regional heads (such as edge servers and macro and authentication classification, streamlining resource utilization and
base stations) to coordinate the optimization process. This approach time consumption, and augmenting network security. One advantage
reduces the overall complexity of the system and improves optimization of integrating MHAs in network authentication is their adaptability
performance. to changing network dynamics. MHAs can dynamically modify the

28
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

Table 9
Security and privacy considerations in emerging technologies with MHAs.
Emerging Applications Security and privacy with MHAs Examples of security and privacy concerns
technologies
Federated Learning Privacy-sensitive applications, Optimal selection of nodes for collaborative Secure communication, participation incentives,
(FL) edge networks (e.g., autonomous ML training without compromising data model inversion attacks, data poisoning attacks,
driving, medical services) privacy differential privacy
Quantum Financial transactions, critical Key distribution, threat detection and Quantum Key Distribution (QKD) attacks,
Communication infrastructure protection mitigation, network topology optimization, privacy concerns, device vulnerability, error
resource allocation correction, quantum hacking
Molecular Parameters for Optimal key management, authentication, Configuring molecular communication network
Communication encrypting/decrypting molecular encryption for secure and private molecular for reduced unauthorized access and signal
signals communication interception
Blockchain Secure data sharing, smart Optimal security and privacy parameters, Developing secure, privacy-preserving data
contracts, decentralized control distributed consensus, trust in blockchain exchange protocols in distributed ledgers
networks
Wireless Medical care, smart control, Unauthorized data access, privacy concerns, Designing secure WBCI protocols, access
Brain–Computer gaming, VR/AR data breaches, malicious software, signal control for wireless brain signal transmission,
Interactions interference authentication, optimizing encryption
Connected Robotics Smart factories, autonomous Anomaly detection, access control, Man-in-the-middle attacks, unauthorized
and Autonomous transportation, remote surgery privacy-preserving data processing, secure hardware access, lack of security updates,
Systems communication, data protection insider threats, unsecured APIs
Multi-sensory Placement of virtual objects Secure communication, privacy-preserving Privacy-preserving ML models, balancing
Extended Reality optimization data processing, and data protection privacy and performance, configuring security
(XR) protocols.

authentication process to accommodate variations in device prolifer- like nanorobots to communicate through exchanging molecular sig-
ation and the spectrum of threats. Additionally, MHAs can optimize nals. This kind of communication is useful in various fields, such as
the deployment of network resources by identifying the most efficient biomedicine, environmental monitoring, and industrial control, due to
authentication methods and parameters, leveraging available resources its low power consumption, high data density, and ability to function
such as processing capability and bandwidth. This optimization can cur- in harsh environments. The information is encoded in the properties
tail the time and resources needed for authentication, thereby elevating of the molecules released, like their type, concentration, and timing,
overall network performance [224]. and is decoded at the receiver using chemical and biological sensing
While no explicit studies or publications have addressed network techniques. Molecular communication is a promising technology for
authentication in 6G and the integration of MHAs, there is a substantial
developing future communication networks, particularly in the context
opportunity for research and advancements in this domain. MHAs have
of 6G communication systems, because of its potential for high speed,
already showcased promising outcomes in refining diverse facets of
low latency, and reliability.
network security. Exploring their potential benefits in 6G network
authentication presents a compelling avenue for further exploration, MHAs can address security and privacy issues in molecular com-
analogous to the establishment of secure and reliable communica- munication as 6G applications by optimizing the encryption and de-
tion in cognitive radio networks [224,225]. In Table 9, the security cryption parameters of the information transmitted through molecular
and privacy considerations in emerging technologies with MHAs are signals. These algorithms can be used to determine the best manage-
summarized. ment of keys, authentication, and encryption schemes, thereby making
the molecular communication system more secure and private. MHAs
7.5. Quantum communication can also help determine the best configuration for the molecular com-
munication network, such as the optimal placement of receptors, to
The study of quantum communication focuses on using quantum reduce the likelihood of unauthorized access and signal interception.
mechanics to transmit information securely and accurately. This in- However, it is crucial to note that the application of MHAs alone is
cludes technologies like quantum cryptography, quantum teleportation, insufficient to address security and privacy issues in molecular com-
and quantum repeaters. Quantum communication uses unique quantum munication fully and should be combined with other security measures
features like quantum entanglement and the Heisenberg uncertainty (see Table 9).
principle to provide unparalleled security and reliability compared to
classical communication methods [226].
By integrating quantum communication, 6G networks can achieve 7.7. Blockchain
even greater security and reliability, making them ideal for applica-
tions such as safeguarding financial transactions and protecting critical Blockchain is a digital ledger technology that enables the secure and
infrastructure. Additionally, 6G networks can use quantum communi-
transparent recording of transactions. Its cryptographic foundations
cation to enhance the precision of positioning and navigation services.
ensure data integrity and immutability, while its operation relies on
Regarding security and privacy in 6G quantum communication net-
a distributed network of computers instead of a centralized authority.
works, MHAs can address various concerns, such as key distribution,
Blockchain and distributed ledger technology have potential use cases
resource allocation, network topology optimization, and the detection
in 6G networks, including secure and transparent data sharing, smart
and mitigation of threats.
contracts, and decentralized control [227].
7.6. Molecular communications The decentralized nature of these technologies can enhance the se-
curity and privacy of 6G transactions, fortifying their resilience against
Molecular communication is the transmission of information us- attacks and safeguarding data integrity. Additionally, the application
ing molecules. It is a small-scale nanonetwork, allowing nanodevices of blockchain can streamline the secure and efficient exchange and

29
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

allocation of resources, such as computational power and spectrum, solutions for a range of tasks, from optimizing virtual object placement
among various stakeholders. MHAs can contribute to tackling secu- to securing user data privacy.
rity and privacy challenges in blockchain by integrating optimization
techniques to enhance the security and privacy facets of these systems. 8. Conclusion
MHAs can search for the optimal blend of security and privacy pa-
rameters within a blockchain network. They can also address concerns
The use of MHAs has proven to be a promising method for wireless
surrounding trust and consensus within distributed systems, including
communications in the 6G era. This survey has provided a comprehen-
the detection and mitigation of malicious actors in the network [76].
sive overview of the current advancements and potential applications
These algorithms can also be utilized to devise and implement secure
in this area, addressing key challenges such as security, privacy, re-
and privacy-preserving protocols for data exchange and storage within
source allocation, spectrum management, edge computing, and wireless
distributed ledger systems. However, the complexities of security and
caching. One of the notable strengths of MHAs lies in their ability to
privacy in blockchain and distributed ledger systems remain an ongoing
optimize complex wireless communication systems by leveraging meta-
challenge, and no solitary algorithm can provide an all-encompassing
heuristic search algorithms. These algorithms excel in exploring large
resolution.
solution spaces and finding near-optimal solutions for critical tasks
7.8. Wireless Brain–Computer Interactions (WBCIs) such as resource allocation, channel assignment, and power control.
By harnessing the power of MHAs, 6G networks can achieve higher
WBCIs involve establishing direct connections between the hu- efficiency, improved spectrum utilization, and enhanced overall system
man brain and computers or other devices using wireless communica- performance. Furthermore, MHAs offer robustness and adaptability in
tion [228]. In the context of 6G networks, it is expected that wireless dynamic and uncertain wireless environments. The ability of MHAs to
WBCIs will enable instantaneous communication between humans and adapt to changing network conditions and varying user demands is
machines with high bandwidth and minimal latency. precious in the heterogeneous landscape of 6G. MHAs enable dynamic
WBCIs have a wide range of applications, including medical health- adjustments of communication parameters, resource allocation, and
care, gaming, virtual and augmented reality, and the governance of network configurations to accommodate diverse requirements and en-
smart devices. In medical contexts, wireless WBCI can monitor and sure seamless connectivity. Another significant contribution of MHAs is
address various neurological conditions, such as Parkinson’s disease enhancing the security and privacy aspects of 6G wireless communica-
and epilepsy. In gaming and virtual reality, wireless WBCI can enhance tions. By incorporating intelligent optimization algorithms, MHAs can
immersive and interactive experiences, allowing users to manipulate strengthen encryption schemes, authentication protocols, and access
virtual elements and environments directly through their thoughts. control mechanisms. This bolstering of security measures protects 6G
To ensure the security and privacy of WBCI systems, MHAs can networks against unauthorized access, data breaches, and other security
optimize the formulation and implementation of security protocols threats. Additionally, MHAs can facilitate privacy-preserving opera-
and mechanisms intrinsic to these systems. Algorithms can explore tions, such as secure data sharing and confidentiality preservation,
the array of possible solutions for intricate or analytically intractable thereby ensuring the protection of user privacy in the 6G era. In the
problems. MHAs can identify the optimal combination of encryption future, the continued development and application of MHAs will play
techniques, authentication methods, and access control mechanisms to an essential role in securing 6G wireless communications. Finally, the
guarantee the confidentiality, integrity, and privacy of wirelessly trans- paper presents research directions to encourage more researchers to
mitted brain signals. However, the effectiveness of these algorithms is work in this vital field.
influenced by the specific issue at hand and the quality of the employed
heuristics. They do not present a universally applicable solution to CRediT authorship contribution statement
security and privacy challenges in WBCI.
Ammar Kamal Abasi: Conceptualization, Data curation, Formal
7.9. Connected robotics and autonomous systems
analysis, Investigation, Methodology, Writing – original draft, Writing –
review & editing. Moayad Aloqaily: Conceptualization, Formal analy-
Connected robotics and autonomous systems are also set to benefit
sis, Funding acquisition, Investigation, Methodology, Writing – original
from 6G networks, which will enable real-time, seamless communica-
draft, Writing – review & editing. Mohsen Guizani: Conceptualiza-
tion between systems and with the network. This will allow for more
tion, Funding acquisition, Investigation, Methodology. Bassem Ouni:
efficient and intelligent automation across diverse industries, including
Funding acquisition, Investigation, Resources.
applications such as remote surgery, autonomous transportation, and
smart factories [229]. MHAs are well-suited to address a range of
security and privacy challenges surrounding connected robotics and Declaration of competing interest
autonomous systems, including data protection, secure communication,
privacy-preserving data processing, access control, and anomaly detec- The authors declare the following financial interests/personal rela-
tion. The advanced technologies offered by 6G networks, such as edge tionships which may be considered as potential competing interests:
computing and AI, will enable more intelligent and efficient control of Moayad Aloqaily reports financial support was provided by Mohamed
these systems, elevating their overall performance and reliability. bin Zayed University of Artificial Intelligence.

7.10. Multi-sensory Extended Reality (XR) Data availability

XR is a technology that offers immersive digital environments with

Data will be made available on request.
haptic, olfactory, and gustatory sensations to enhance user experi-
ence. This technology is commonly experienced through head-mounted
displays, gloves, or bodysuits equipped with sensors and advanced Acknowledgment
interfaces [230]. With the evolution of 6G networks, multi-sensory
XR is expected to gain even greater prominence, with MHAs playing This research was supported by the Technology Innovation Institute
a critical role in enhancing and optimizing user experience. MHAs (TII), Abu Dhabi, UAE, under the CyberAI project (grant number:
can process extensive data from sensory inputs to determine optimal TII/DSRC/2022/3036).

30
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

References [26] E. Osaba, E. Villar-Rodriguez, J. Del Ser, A.J. Nebro, D. Molina, A. LaTorre,
P.N. Suganthan, C.A.C. Coello, F. Herrera, A tutorial on the design, experimen-
[1] A. Maskooki, G. Sabatino, N. Mitton, Analysis and performance evaluation of tation and application of metaheuristic algorithms to real-world optimization
the next generation wireless networks, in: Modeling and Simulation of Computer problems, Swarm Evol. Comput. 64 (2021) 100888.
Networks and Systems, Elsevier, 2015, pp. 601–627. [27] Q.-V. Pham, D.C. Nguyen, S. Mirjalili, D.T. Hoang, D.N. Nguyen, P.N. Pathirana,
[2] T. Cokyasar, Optimization of battery swapping infrastructure for e-commerce W.-J. Hwang, Swarm intelligence for next-generation networks: Recent advances
drone delivery, Comput. Commun. 168 (2021) 146–154. and applications, J. Netw. Comput. Appl. 191 (2021) 103141.
[3] M. Banafaa, I. Shayea, J. Din, M.H. Azmi, A. Alashbi, Y.I. Daradkeh, A. [28] T. Dokeroglu, E. Sevinc, T. Kucukyilmaz, A. Cosar, A survey on new generation
Alhammadi, 6G mobile communication technology: Requirements, targets, metaheuristic algorithms, Comput. Ind. Eng. 137 (2019) 106040.
applications, challenges, advantages, and opportunities, Alexandria Eng. J. 64 [29] Y. Sun, M. Peng, Y. Zhou, Y. Huang, S. Mao, Application of machine learning in
(2023) 245–274. wireless networks: Key techniques and open issues, IEEE Commun. Surv. Tutor.
21 (4) (2019) 3072–3108.
[4] M.Z. Chowdhury, M. Shahjalal, S. Ahmed, Y.M. Jang, 6G wireless communica-
[30] C. Zhang, P. Patras, H. Haddadi, Deep learning in mobile and wireless
tion systems: Applications, requirements, technologies, challenges, and research
networking: A survey, IEEE Commun. Surv. Tutorials 21 (3) (2019) 2224–2287.
directions, IEEE Open J. Commun. Soc. 1 (2020) 957–975.
[31] W.Y.B. Lim, N.C. Luong, D.T. Hoang, Y. Jiao, Y.-C. Liang, Q. Yang, D. Niyato,
[5] I. Shayea, M. Ergen, M.H. Azmi, S.A. Çolak, R. Nordin, Y.I. Daradkeh, Key
C. Miao, Federated learning in mobile edge networks: A comprehensive survey,
challenges, drivers and solutions for mobility management in 5g networks: A
IEEE Commun. Surv. Tutor. 22 (3) (2020) 2031–2063.
survey, IEEE Access 8 (2020) 172534–172552.
[32] C.T. Nguyen, N. Van Huynh, N.H. Chu, Y.M. Saputra, D.T. Hoang, D.N. Nguyen,
[6] I. Shayea, M. Ergen, A. Azizan, M. Ismail, Y.I. Daradkeh, Individualistic dynamic
Q.-V. Pham, D. Niyato, E. Dutkiewicz, W.-J. Hwang, Transfer learning for
handover parameter self-optimization algorithm for 5G networks based on
future wireless networks: A comprehensive survey, 2021, arXiv preprint arXiv:
automatic weight function, IEEE Access 8 (2020) 214392–214412.
2102.07572.
[7] C. Dong, X. Xiong, Q. Xue, Z. Zhang, K. Niu, P. Zhang, A survey on the network
[33] S.B. Joseph, E.G. Dada, A. Abidemi, D.O. Oyewola, B.M. Khammas, Metaheuris-
models applied in the industrial network optimization, 2022, arXiv preprint
tic algorithms for PID controller parameters tuning: Review, approaches and
arXiv:2209.08294.
open problems, Heliyon (2022) e09399.
[8] O.A. Amodu, C. Jarray, S.A. Busari, M. Othman, THz-enabled UAV communica-
[34] F. Glover, Future paths for integer programming and links to artificial
tions: Motivations, results, applications, challenges, and future considerations,
intelligence, Comput. Oper. Res. 13 (5) (1986) 533–549.
Ad Hoc Netw. 140 (2023) 103073.
[35] S.N. Makhadmeh, M.A. Al-Betar, A.K. Abasi, M.A. Awadallah, I.A. Doush, Z.A.A.
[9] U. Ghafoor, H.Z. Khan, A.M. Siddiqui, M. Ali, A. Rauf, A. Wahla, M. Naeem,
Alyasseri, O.A. Alomari, Recent advances in butterfly optimization algorithm,
Cluster based resource management using H-NOMA in heterogeneous networks
its versions and applications, Arch. Comput. Methods Eng. 30 (2) (2023)
beyond 5G, Ad Hoc Netw. 149 (2023) 103252.
1399–1420.
[10] H.H.H. Mahmoud, A.A. Amer, T. Ismail, 6G: A comprehensive survey on
[36] A.K. Abasi, M. Aloqaily, M. Guizani, Optimization of CNN using modified honey
technologies, applications, challenges, and research problems, Trans. Emerg.
badger algorithm for sleep apnea detection, Expert Syst. Appl. 229 (2023)
Telecommun. Technol. 32 (4) (2021) e4233.
120484.
[11] C. Han, A.O. Bicen, I.F. Akyildiz, Multi-ray channel modeling and wideband
[37] A.K. Abasi, A.T. Khader, M.A. Al-Betar, Z.A.A. Alyasseri, S.N. Makhadmeh,
characterization for wireless communications in the terahertz band, IEEE Trans.
M. Al-laham, S. Naim, A hybrid salp swarm algorithm with 𝛽-hill climbing
Wireless Commun. 14 (5) (2014) 2402–2412.
algorithm for text documents clustering, in: Evolutionary Data Clustering:
[12] S. Moghadami, F. Hajilou, P. Agrawal, S. Ardalan, A 210 GHz fully-integrated
Algorithms and Applications, Springer, 2021, pp. 129–161.
OOK transceiver for short-range wireless chip-to-chip communication in 40 nm
[38] M.A. Al-Betar, A.K. Abasi, G. Al-Naymat, K. Arshad, S.N. Makhadmeh, Bare-
CMOS technology, IEEE Trans. Terahertz Sci. Technol. 5 (5) (2015) 737–741.
bones based salp swarm algorithm for text document clustering, IEEE Access
[13] M. Ikram, K. Sultan, M.F. Lateef, A.S. Alqadami, A road towards 6G (2023).
communication—A review of 5G antennas, arrays, and wearable devices, [39] R. Kuo, S.-S. Li, Applying particle swarm optimization algorithm-based collab-
Electronics 11 (1) (2022) 169. orative filtering recommender system considering rating and review, Appl. Soft
[14] M. Wang, T. Zhu, T. Zhang, J. Zhang, S. Yu, W. Zhou, Security and privacy Comput. (2023) 110038.
in 6G networks: New areas and new challenges, Digit. Commun. Netw. 6 (3) [40] R. Kommadath, D. Maharana, P. Kotecha, A metaheuristic-based efficient
(2020) 281–291. strategy for multi-unit production planning with unique process constraints,
[15] A.K. Abasi, S.N. Makhadmeh, M.A. Al-Betar, O.A. Alomari, M.A. Awadallah, Appl. Soft Comput. 134 (2023) 109871.
Z.A.A. Alyasseri, I.A. Doush, A. Elnagar, E.H. Alkhammash, M. Hadjouni, [41] J.H. Holland, Genetic algorithms and adaptation, in: Adaptive Control of
Lemurs optimizer: A new metaheuristic algorithm for global optimization, Appl. Ill-Defined Systems, Springer, 1984, pp. 317–333.
Sci. 12 (19) (2022) 10057. [42] R. Eberhart, J. Kennedy, Particle swarm optimization, in: Proceedings of the
[16] G. Kwon, J. Park, D. Kim, K.C. Hwang, Optimization of a shared-aperture dual- IEEE International Conference on Neural Networks, Vol. 4, Citeseer, 1995, pp.
band transmitting/receiving array antenna for radar applications, IEEE Trans. 1942–1948.
Antennas and Propagation 65 (12) (2017) 7038–7051. [43] S. Mirjalili, S.M. Mirjalili, A. Lewis, Grey wolf optimizer, Adv. Eng. Softw. 69
[17] Z.A.A. Alyasseri, M.A. Al-Betar, M.A. Awadallah, S.N. Makhadmeh, A.K. Abasi, (2014) 46–61.
I.A. Doush, O.A. Alomari, A hybrid flower pollination with 𝛽-hill climbing [44] D. Karaboga, et al., An Idea Based on Honey Bee Swarm for Numerical
algorithm for global optimization, J. King Saud Univ. Comput. Inf. Sci. 34 (8) Optimization, Technical report-tr06, Tech. Rep., Erciyes University, Engineering
(2022) 4821–4835. Faculty, Computer . . . , 2005.
[18] O.M. Alyasiri, Y.-N. Cheah, A.K. Abasi, O.M. Al-Janabi, Wrapper and hybrid [45] M. Dorigo, G. Di Caro, Ant colony optimization: a new meta-heuristic, in:
feature selection methods using metaheuristic algorithms for english text Proceedings of the 1999 Congress on Evolutionary Computation-CEC99 (Cat.
classification: A systematic review, IEEE Access (2022). No. 99TH8406), Vol. 2, IEEE, 1999, pp. 1470–1477.
[19] M.A. Al-Betar, M.A. Awadallah, I.A. Doush, O.A. Alomari, A.K. Abasi, S.N. [46] X.-S. Yang, A new metaheuristic bat-inspired algorithm, in: Nature Inspired
Makhadmeh, Z.A.A. Alyasseri, Boosting the training of neural networks through Cooperative Strategies for Optimization, NICSO 2010, Springer, 2010, pp.
hybrid metaheuristics, Cluster Comput. (2022) 1–23. 65–74.
[20] A.K. Abasi, A.T. Khader, M.A. Al-Betar, S. Naim, S.N. Makhadmeh, Z.A.A. [47] K. Kiełkowicz, D. Grela, Modified bat algorithm for nonlinear optimization, Int.
Alyasseri, Link-based multi-verse optimizer for text documents clustering, Appl. J. Comput. Sci. Netw. Secur. (IJCSNS) (2016) 46–50.
Soft Comput. 87 (2020) 106002. [48] X.-S. Yang, Firefly algorithms for multimodal optimization, in: Stochastic
[21] N. Khanduja, B. Bhushan, Recent advances and application of metaheuris- Algorithms: Foundations and Applications: 5th International Symposium, SAGA
tic algorithms: A survey (2014–2020), in: Metaheuristic and Evolutionary 2009, Sapporo, Japan, October 26-28, 2009. Proceedings 5, Springer, 2009, pp.
Computation: Algorithms and Applications, Springer, 2021, pp. 207–228. 169–178.
[22] P. Agrawal, H.F. Abutarboush, T. Ganesh, A.W. Mohamed, Metaheuristic algo- [49] W.A. Khan, N.N. Hamadneh, S.L. Tilahun, J. Ngnotchouye, A review and
rithms on feature selection: A survey of one decade of research (2009–2019), comparative study of firefly algorithm and its modified versions, in: Optimiza-
IEEE Access 9 (2021) 26766–26791. tion Algorithms-Methods and Applications, vol. 45, IntechOpen Press, London,
[23] M. Asim, Y. Wang, K. Wang, P.-Q. Huang, A review on computational United Kingdom, 2016, pp. 281–313.
intelligence techniques in cloud and edge computing, IEEE Trans. Emerg. Top. [50] S.-C. Chu, P.-W. Tsai, J.-S. Pan, Cat swarm optimization, in: PRICAI 2006:
Comput. Intell. 4 (6) (2020) 742–763. Trends in Artificial Intelligence: 9th Pacific Rim International Conference on
[24] R. Elshaer, H. Awad, A taxonomic review of metaheuristic algorithms for solving Artificial Intelligence Guilin, China, August 7-11, 2006 Proceedings 9, Springer,
the vehicle routing problem and its variants, Comput. Ind. Eng. 140 (2020) 2006, pp. 854–858.
106242. [51] D. Bertsimas, J. Tsitsiklis, Simulated annealing, Statist. Sci. 8 (1) (1993) 10–15.
[25] B. Ji, Y. Wang, K. Song, C. Li, H. Wen, V.G. Menon, S. Mumtaz, A survey of [52] F.A. Hashim, E.H. Houssein, K. Hussain, M.S. Mabrouk, W. Al-Atabany,
computational intelligence for 6G: Key technologies, applications and trends, Honey badger algorithm: New metaheuristic algorithm for solving optimization
IEEE Trans. Ind. Inform. 17 (10) (2021) 7145–7154. problems, Math. Comput. Simulation 192 (2022) 84–110.

31
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

[53] B. Abdollahzadeh, F.S. Gharehchopogh, S. Mirjalili, African vultures opti- [79] A. Petrosino, G. Piro, L.A. Grieco, G. Boggia, An optimal allocation framework
mization algorithm: A new nature-inspired metaheuristic algorithm for global of security virtual network functions in 6G satellite deployments, in: 2022 IEEE
optimization problems, Comput. Ind. Eng. 158 (2021) 107408. 19th Annual Consumer Communications & Networking Conference, CCNC, IEEE,
[54] L. Abualigah, D. Yousri, M. Abd Elaziz, A.A. Ewees, M.A. Al-Qaness, A.H. 2022, pp. 917–920.
Gandomi, Aquila optimizer: a novel meta-heuristic optimization algorithm, [80] Z. Zhang, Y. Cao, Z. Cui, W. Zhang, J. Chen, A many-objective optimiza-
Comput. Ind. Eng. 157 (2021) 107250. tion based intelligent intrusion detection algorithm for enhancing security of
[55] A. Faramarzi, M. Heidarinejad, S. Mirjalili, A.H. Gandomi, Marine predators vehicular networks in 6G, IEEE Trans. Veh. Technol. 70 (6) (2021) 5234–5243.
algorithm: A nature-inspired metaheuristic, Expert Syst. Appl. 152 (2020) [81] D. Chang, W. Sun, Y. Yang, T. Wang, An E-ABAC-based SDN access control
113377. method, in: 2019 6th International Conference on Information Science and
[56] M.A. Al-Betar, Z.A.A. Alyasseri, M.A. Awadallah, I. Abu Doush, Coronavirus Control Engineering, ICISCE, IEEE, 2019, pp. 668–672.
herd immunity optimizer (CHIO), Neural Comput. Appl. 33 (2021) 5011–5042. [82] E.N. Egashira, D.P.M. Osorio, N.T. Nguyen, M. Juntti, Secrecy capacity max-
[57] G. Zhou, W. Tian, R. Buyya, K. Wu, Growable genetic algorithm with imization for a hybrid relay-RIS scheme in mmwave MIMO networks, 2022,
heuristic-based local search for multi-dimensional resources scheduling of cloud arXiv preprint arXiv:2205.13904.
computing, Appl. Soft Comput. (2023) 110027. [83] D.K. Jain, S.K.S. Tyagi, S. Neelakandan, M. Prakash, L. Natrayan, Metaheuristic
[58] D. Li, L. Wang, W. Guo, M. Zhang, B. Hu, Q. Wu, A particle swarm optimizer optimization-based resource allocation technique for cybertwin-driven 6G on
with dynamic balance of convergence and diversity for large-scale optimization, IoE environment, IEEE Trans. Ind. Inform. 18 (7) (2021) 4884–4892.
Appl. Soft Comput. 132 (2023) 109852. [84] Y. Su, H. Gao, S. Zhang, Energy-efficient resource management for CCFD
[59] H. Pan, S. Chen, H. Xiong, A high-dimensional feature selection method based massive MIMO systems in 6G networks, J. Syst. Eng. Electron. 33 (4) (2022)
on modified gray wolf optimization, Appl. Soft Comput. (2023) 110031. 877–886.
[60] S. Zhu, C.-M. Pun, H. Zhu, S. Li, X. Huang, H. Gao, An artificial bee colony [85] J.A. Gómez-Avilés, Á.G. Andrade, A. Martínez-Vargas, A bio-inspired approach
algorithm with a balance strategy for wireless sensor network, Appl. Soft for the spectrum allocation problem in IoT networks, in: Proceedings of the
Comput. (2023) 110083. 2020 Genetic and Evolutionary Computation Conference Companion, 2020, pp.
[61] W. Li, C. Wang, Y. Huang, Y.-m. Cheung, Heuristic smoothing ant colony 37–38.
optimization with differential information for the traveling salesman problem, [86] W.U. Khan, X. Li, A. Ihsan, M.A. Khan, V.G. Menon, M. Ahmed, NOMA-
Appl. Soft Comput. 133 (2023) 109943. enabled optimization framework for next-generation small-cell IoV networks
[62] A. Tabak, İ. İlhan, An effective method based on simulated annealing for under imperfect SIC decoding, IEEE Trans. Intell. Transp. Syst. (2021).
automatic generation control of power systems, Appl. Soft Comput. 126 (2022) [87] G. Eappen, S. Shankar, R. Nilavalan, Efficient spectrum sensing for the relay
109277. based cognitive radio network for enhancing the network coverage for wireless
[63] N. Charef, A.B. Mnaouer, M. Aloqaily, O. Bouachir, M. Guizani, Artificial patient monitoring system, in: 2020 5th International Conference on Smart and
intelligence implication on energy sustainability in internet of things: A survey, Sustainable Technologies, SpliTech 2020, 2020.
Inf. Process. Manage. 60 (2) (2023) 103212. [88] Y.-X. Guo, H. Li, A power allocation method based on particle swarm algorithm
[64] E. Baccour, N. Mhaisen, A.A. Abdellatif, A. Erbad, A. Mohamed, M. Hamdi, for NOMA downlink networks, in: Journal of Physics: Conference Series, Vol.
M. Guizani, Pervasive AI for IoT applications: A survey on resource-efficient 1087, IOP Publishing, 2018, 022033.
distributed artificial intelligence, IEEE Commun. Surv. Tutor. (2022). [89] Y. Zhang, H. Huang, C. Zhang, Z. Zhao, J. Peng, S. Fu, K. Qiu, Joint precoding
[65] J. Singh, M. Wazid, A.K. Das, V. Chamola, M. Guizani, Machine learning secu- and user grouping for RIS-aided mmwave NOMA system, in: 2022 IEEE 14th
rity attacks and defense approaches for emerging cyber physical applications: International Conference on Advanced Infocomm Technology, ICAIT 2022,
A comprehensive survey, Comput. Commun. (2022). 2022, pp. 36–41.
[66] A.K. Abasi, M. Aloqaily, B. Ouni, M. Guizani, M. Debbah, F. Karray, A survey on [90] Ö.F. Gemici, F. Kara, I. Hokelek, G.K. Kurt, H.A. Cırpan, Resource allocation
securing 6G wireless communications based optimization techniques, in: 2023 for NOMA downlink systems: Genetic algorithm approach, in: 2017 40th
International Wireless Communications and Mobile Computing, IWCMC, IEEE, International Conference on Telecommunications and Signal Processing, TSP,
2023, pp. 216–223. IEEE, 2017, pp. 114–118.
[67] I.H. Abdulqadder, D. Zou, I.T. Aziz, B. Yuan, W. Li, SecSDN-cloud: defeating [91] Q. Wang, H. Wang, On sum rate maximization for IRS-aided downlink NOMA
vulnerable attacks through secure software-defined networks, IEEE Access 6 systems, Nanjing Youdian Daxue Xuebao (Ziran Kexue Ban)/J. Nanjing Univ.
(2018) 8292–8301. Posts Telecommun. (Natural Sci.) 42 (1) (2022) 23–29.
[68] A. Rehman, T. Saba, K. Haseeb, T. Alam, J. Lloret, Sustainability model for the [92] Z. Abou El Houda, B. Brik, Next-power: Next-generation framework for secure
internet of health things (IoHT) using reinforcement learning with mobile edge and sustainable energy trading in the metaverse, Ad Hoc Netw. 149 (2023)
secured services, Sustainability 14 (19) (2022) 12185. 103243.
[69] Z. Zhang, M. Zhao, H. Wang, Z. Cui, W. Zhang, An efficient interval many- [93] X. Chen, F. Yan, M. Hu, Z. Lin, Energy efficiency optimization of intelligent
objective evolutionary algorithm for cloud task scheduling problem under reflective surface-assisted terahertz-RSMA system, 2022, arXiv preprint arXiv:
uncertainty, Inform. Sci. 583 (2022) 56–72. 2211.11271.
[70] I.H. Abdulqadder, S. Zhou, SliceBlock: Context-aware authentication handover [94] F. Tan, X. Xu, H. Chen, S. Li, Energy-efficient beamforming optimization
and secure network slicing using DAG-blockchain in edge-assisted SDN/NFV-6G for MISO communication based on reconfigurable intelligent surface, Phys.
environment, IEEE Internet Things J. (2022). Commun. (2023) 101996.
[71] A.K. Abasi, N.M. Hijazi, M. Aloqaily, M. Guizani, Securing federated learning [95] L. Kouhalvandi, L. Matekovits, Automated power amplifier design through mul-
against FGSM attacks with adaptive trust scores and blockchain updates, in: tiobjective bottom-up and particle swarm optimizations using neural network,
2023 Fifth International Conference on Blockchain Computing and Applications, in: International Youth Conference on Electronics, Telecommunications and
BCCA, IEEE, 2023, pp. 194–199. Information Technologies, Springer, 2022, pp. 29–36.
[72] J.C.-W. Lin, G. Srivastava, Y. Zhang, Y. Djenouri, M. Aloqaily, Privacy- [96] J. Chen, S. Wang, J. Jia, Q. Wang, L. Yang, X. Wang, Multi-objective oriented
preserving multiobjective sanitization model in 6G IoT environments, IEEE resource allocation in reconfigurable intelligent surface assisted HCNs, Ad Hoc
Internet Things J. 8 (7) (2020) 5340–5349. Netw. 140 (2023) 103066.
[73] T. Le Anh, I.P. Hong, Secrecy performance of a multi-NOMA-MIMO system [97] X. Jian, L. Wu, K. Yu, M. Aloqaily, J. Ben-Othman, Energy-efficient user
in the UEH relaying network using the PSO algorithm, IEEE Access 9 (2020) association with load-balancing for cooperative iIoT network within B5G era,
2317–2331. J. Netw. Comput. Appl. 189 (2021) 103110.
[74] L. Qian, X. Chi, L. Zhao, A. Chaaban, Secure visible light communications via [98] D. Liu, L. Wang, Y. Chen, M. Elkashlan, K.-K. Wong, R. Schober, L. Hanzo,
intelligent reflecting surfaces, in: ICC 2021-IEEE International Conference on User association in 5G networks: A survey and an outlook, IEEE Commun.
Communications, IEEE, 2021, pp. 1–6. Surv. Tutor. 18 (2) (2016) 1018–1044.
[75] L. Xu, X. Zhou, Y. Fu, G. Jiang, X. Yu, M. Yu, N. Kumar, M. Guizani, Accurate [99] G. Eappen, T. Shankar, Multi-objective modified grey wolf optimization algo-
and efficient performance prediction for mobile IoV networks using GWO-GR rithm for efficient spectrum sensing in the cognitive radio network, Arab. J.
neural network, IEEE Internet Things J. (2022). Sci. Eng. 46 (4) (2021) 3115–3145.
[76] G. Soni, K. Chandravanshi, M.K. Jhariya, A. Rajput, An IPS approach to [100] D. Pari, J. Natarajan, Secure spectrum access, routing, and hybrid beamforming
secure V-RSU communication from blackhole and wormhole attacks in VANET, in an edge-enabled mmwave massive MIMO CRN-based internet of connected
in: Contemporary Issues in Communication, Cloud and Big Data Analytics, vehicle (IoCV) environments, Sensors 22 (15) (2022).
Springer, 2022, pp. 57–65. [101] K. Hussein, M. Rihan, S. Hussin, A. Zaghloul, et al., Rate splitting multiple
[77] S. Lv, S. Wang, J. Jin, Q. Wang, Y. Yuan, G. Liu, An enhanced multi- access scheme for cognitive radio network, Egypt. Int. J. Eng. Sci. Technol. 38
carrier waveform for downlink short-packet communication, in: 2021 IEEE 94th (2) (2022) 47–53.
Vehicular Technology Conference, VTC2021-Fall, IEEE, 2021, pp. 1–6. [102] R. Pandi Selvam, K. Narayanasamy, M. Ilayaraja, Efficient deer hunting op-
[78] B. Han, D. Krummmacher, Q. Zhou, H.D. Schotten, Trust-awareness to secure timization algorithm based spectrum sensing approach for 6G communication
swarm intelligence from data injection attack, 2022, arXiv preprint arXiv: networks, in: AI-Enabled 6G Networks and Applications, Wiley Online Library,
2211.08407. 2023, pp. 111–129.

32
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

[103] A.-A.A. Boulogeorgos, S.K. Goudos, A. Alexiou, Users association in ultra dense [128] H. Xiao, H. Du, J. Zhang, K. Guan, B. Ai, Outage probability of reconfigurable
THz networks, in: 2018 IEEE 19th International Workshop on Signal Processing intelligent surface aided thz communications, in: 2021 IEEE/CIC International
Advances in Wireless Communications, SPAWC, IEEE, 2018, pp. 1–5. Conference on Communications in China, ICCC, IEEE, 2021, pp. 415–420.
[104] L. Kouhalvandi, L. Matekovits, Hyperparameter optimization of long short- [129] K. Senthilkumar, K. Pirapaharan, H. Kunsei, S. Hoole, P. Hoole, Real- and
term memory-based forecasting DNN for antenna modeling through stochastic complex-valued artificial intelligence weight optimization algorithms for smart
methods, IEEE Antennas Wirel. Propag. Lett. 21 (4) (2022) 725–729. antennas in 5/6G wireless systems: Linear and nonlinear arrays, in: Smart An-
[105] H.P. Kuribayashi, M.A. De Souza, D.D.A. Gomes, K.D.C. Silva, M.S. Da Silva, tennas and Electromagnetic Signal Processing in Advanced Wireless Technology:
J.C.W.A. Costa, C.R.L. Francês, Particle swarm-based cell range expansion for with Artificial Intelligence Application and Coding, 2020, pp. 263–302.
heterogeneous mobile networks, IEEE Access 8 (2020) 37021–37034. [130] Z. Zhao, C. Feng, W. Hong, J. Jiang, C. Jia, T.Q.S. Quek, M. Peng, Federated
[106] J. Plachy, Z. Becvar, P. Mach, R. Marik, M. Vondra, Joint positioning of flying learning with non-IID data in wireless networks, IEEE Trans. Wireless Commun.
base stations and association of users: Evolutionary-based approach, IEEE Access 21 (3) (2022) 1927–1942.
7 (2019) 11454–11463. [131] K. Wang, Z. Ma, Y. Zhao, Y. Deng, Y. Luo, M. He, H. Xu, Design of irregularly
[107] W. Lin, Z. Dong, K. Wang, D. Wang, Y. Deng, Y. Liao, Y. Liu, D. Wan, B. distributed antenna array towards smart 6g networks, Electronics (Switzerland)
Xu, G. Wu, A novel load balancing scheme for satellite IoT networks based on 11 (19) (2022).
spatial–temporal distribution of users and advanced genetic algorithms, Sensors [132] A.H. Abdelrahman, P. Nayeri, A.Z. Elsherbeni, F. Yang, Single-feed quad-beam
22 (20) (2022) 7930. transmitarray antenna design, IEEE Trans. Antennas and Propagation 64 (3)
[108] D. Chen, Q. Wang, J. Wang, Z. Li, S. Wu, R. Hao, K. Fan, H. Lu, J. Jin, Energy (2016) 953–959.
efficiency optimization for SLIPT-enabled NOMA system, in: Photonics, Vol. 10,
[133] A.T. Sikan, L. Kouhalvandi, L. Matekovits, I. Peter, MIMO antenna design
MDPI, 2023, p. 791.
through genetic algorithm, in: International Conference Interdisciplinarity in
[109] C. Chen, J. Xiang, Z. Ye, W. Yan, S. Wang, Z. Wang, P. Chen, M.
Engineering, Springer, 2023, pp. 862–869.
Xiao, Deep learning-based energy optimization for edge device in UAV-aided
[134] I. Yildirim, A. Koc, E. Basar, T. Le-Ngoc, RIS-aided angular-based hybrid
communications, Drones 6 (6) (2022).
beamforming design in mmwave massive MIMO systems, 2022, arXiv preprint
[110] T. Wang, X. Yang, F. Deng, L. Gao, Y. Jiang, Z. Yang, Joint power and duty-
arXiv:2208.06622.
cycle design using alternating optimization algorithm under energy harvesting
[135] Q. Wu, J. Zhang, Y. Zhang, G. Xin, J. Guo, Configuring reconfigurable intel-
architectures, China Commun. 17 (12) (2020) 139–155.
ligent surface for parallel MIMO visible light communications with asymptotic
[111] K.A. Eldrandaly, L. Abdel-Fatah, M. Abdel-Basset, M. El-Hoseny, N.M. Abdel-
capacity maximization, Appl. Sci. 13 (1) (2023) 563.
Aziz, Green communication for sixth-generation intent-based networks: An
architecture based on hybrid computational intelligence algorithm, Wirel. [136] Y. Cao, T. Lv, W. Ni, Two-timescale optimization for intelligent reflecting
Commun. Mob. Comput. 2021 (2021). surface-assisted MIMO transmission in fast-changing channels, IEEE Trans.
[112] S. Verma, S. Kaur, M.A. Khan, P.S. Sehdev, Toward green communication in Wireless Commun. 21 (12) (2022) 10424–10437.
6G-enabled massive internet of things, IEEE Internet Things J. 8 (7) (2020) [137] H. Bian, R. Liu, Reliable and energy-efficient LEO satellite communication with
5408–5415. IR-HARQ via power allocation, Sensors 22 (8) (2022) 3035.
[113] P. Singh, M.C. dos Reis, V.H.C. de Albuquerque, Design of artificial intelli- [138] S. Verma, T. Raj, K. Joshi, P. Raturi, H. Anandaram, A. Gupta, Indoor real-
gence enabled dingo optimizer for energy management in 6G communication time location system for efficient location tracking using IoT, in: 2022 IEEE
networks, in: AI-Enabled 6G Networks and Applications, Wiley Online Library, World Conference on Applied Intelligence and Computing, AIC, IEEE, 2022,
2023, pp. 75–92. pp. 517–523.
[114] R. Rajender, C. Anupama, G.J. Moses, E.L. Lydia, S. Kadry, S. Lim, Artificial [139] M. Chen, A.A. Ghorbani, et al., A survey on user profiling model for anomaly
intelligence-enabled cooperative cluster-based data collection for unmanned detection in cyberspace, J. Cyber Secur. Mobil. 8 (1) (2019) 75–112.
aerial vehicles, Comput. Mater. Contin. 73 (2) (2022) 3351–3365. [140] R. Gupta, J. Gupta, Combating full-duplex active eavesdropping using friendly
[115] I. Yildirim, F. Kilinc, E. Basar, G.C. Alexandropoulos, Hybrid RIS-empowered jammer with intelligent optimization algorithms, in: 2021 IEEE Interna-
reflection and decode-and-forward relaying for coverage extension, IEEE tional Conference on Electronics, Computing and Communication Technologies,
Commun. Lett. 25 (5) (2021) 1692–1696. CONECCT, IEEE, 2021, pp. 1–6.
[116] M. Kumar, S. Kumar, P.K. Kashyap, G. Aggarwal, R.S. Rathore, O. Kaiwartya, [141] S.X. Pushpa, S. Raja, Elliptic curve cryptography based authentication protocol
J. Lloret, Green communication in internet of things: A hybrid bio-inspired enabled with optimized neural network based DoS mitigation, Wirel. Pers.
intelligent approach, Sensors 22 (10) (2022) 3910. Commun. 124 (1) (2022) 1–25.
[117] H.J. Park, H.W. Kim, S.H. Chae, Resource allocation for minimizing the transmit [142] K. Das, R. Basu, R. Karmakar, Man-in-the-middle attack detection using
power in uplink NOMA IoT cellular networks, in: 2022 13th International ensemble learning, in: 2022 13th International Conference on Computing
Conference on Information and Communication Technology Convergence, ICTC, Communication and Networking Technologies, ICCCNT, IEEE, 2022, pp. 1–6.
IEEE, 2022, pp. 1387–1392. [143] A. Kumar, S. Kumar, V. Dutt, A.K. Dubey, V. García-Díaz, IoT-based ECG
[118] X. Qi, S. Khattak, S.S. Mustafa, I. Khan, GA based energy efficiency resource monitoring for arrhythmia classification using coyote grey wolf optimization-
allocation in heterogeneous networks with MRC, IET Commun. 16 (4) (2022) based deep learning CNN classifier, Biomed. Signal Process. Control 76 (2022)
369–377. 103638.
[119] S. Zou, J. Wu, H. Yu, W. Wang, L. Huang, W. Ni, Y. Liu, Efficiency-optimized [144] J. Xu, C. Guo, H. Zhang, Joint channel allocation and power control based
6G: A virtual network resource orchestration strategy by enhanced particle on PSO for cellular networks with D2D communications, Comput. Netw. 133
swarm optimization, Digit. Commun. Netw. (2023). (2018) 104–119.
[120] J. Jia, Y. Xu, Z. Du, J. Chen, Q. Wang, X. Wang, Joint resource alloca-
[145] M.A. Rashid, H.H. Pajooh, A security framework for IoT authentication and
tion for QoE optimization in large-scale NOMA-enabled multi-cell networks,
authorization based on blockchain technology, in: 2019 18th IEEE International
Peer-to-Peer Netw. Appl. 15 (1) (2022) 689–702.
Conference on Trust, Security and Privacy in Computing and Communica-
[121] F.d.O. Torres, V.A.S. Júnior, D. da Costa, D.L. Cardoso, R.C. Oliveira, Radio
tions/13th IEEE International Conference on Big Data Science and Engineering,
resource allocation in a 6G D-OMA network with imperfect SIC: A framework
TrustCom/BigDataSE, IEEE, 2019, pp. 264–271.
aided by a bi-objective hyper-heuristic, Eng. Appl. Artif. Intell. 119 (2023)
[146] Y. Lyu, Y. Mo, S. Yue, W. Liu, Improved beetle antennae algorithm based on
105830.
localization for jamming attack in wireless sensor networks, IEEE Access 10
[122] X. Xiong, H. Lou, X. Ge, Modeling and optimization of OAM-MIMO communi-
(2022) 13071–13088.
cation systems with unaligned antennas, IEEE Trans. Commun. 70 (6) (2022)
[147] J. Ye, A. Giani, A. Elasser, S.K. Mazumder, C. Farnell, H.A. Mantooth, T.
3682–3694.
Kim, J. Liu, B. Chen, G.-S. Seo, et al., A review of cyber–physical security for
[123] Y.J. Tan, C. Zhu, T.C. Tan, A. Kumar, L.J. Wong, Y. Chong, R. Singh,
photovoltaic systems, IEEE J. Emerg. Sel. Top. Power Electron. 10 (4) (2021)
Self-adaptive deep reinforcement learning for THz beamforming with silicon
4879–4901.
metasurfaces in 6G communications, Opt. Express 30 (15) (2022) 27763–27779.
[124] Q. Gu, D. Wu, X. Su, J. Jin, Y. Yuan, J. Wang, Performance compar- [148] D. Bepari, S. Mondal, A. Chandra, R. Shukla, Y. Liu, M. Guizani, A. Nallanathan,
isons between reconfigurable intelligent surface and full/half-duplex relays, A survey on cache-aided NOMA for 6G networks, 2022, arXiv preprint arXiv:
2021-September, 2021, 2205.05321.
[125] Z. Li, Z. Chen, X. Ma, W. Chen, C. Han, Artificial gene reservation aided genetic [149] S. Aljubayrin, Computational energy efficient trajectory planning for
algorithm for IRS enabled terahertz systems, in: 2020 13th UK-Europe-China UAV-enabled 6G MEC communication network, Phys. Commun. (2023) 102000.
Workshop on Millimetre-Waves and Terahertz Technologies, UCMMT, IEEE, [150] R.Q. Hu, L. Hanzo, et al., Twin-timescale artificial intelligence aided mobility-
2020, pp. 1–3. aware edge caching and computing in vehicular networks, IEEE Trans. Veh.
[126] G.-H. Li, D.-W. Yue, F. Qi, Joint beamforming and power allocation for Technol. 68 (4) (2019) 3086–3099.
intelligent reflecting surface-aided millimeter wave MIMO systems, Wirel. Netw. [151] X. Shen, Z. Chang, S. Niu, Mobile edge computing task offloading strategy based
28 (5) (2022) 1935–1947. on parking cooperation in the internet of vehicles, Sensors 22 (13) (2022) 4959.
[127] A. Salem, K.-K. Wong, C.-B. Chae, Impact of phase-shift error on the secrecy [152] A. Mseddi, W. Jaafar, H. Elbiaze, W. Ajib, Joint container placement and task
performance of uplink RIS communication systems, 2022, arXiv preprint arXiv: provisioning in dynamic fog computing, IEEE Internet Things J. 6 (6) (2019)
2301.00276. 10028–10040.

33
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

[153] Y. Zhang, Y. Liu, J. Zhou, J. Sun, K. Li, Slow-movement particle swarm [178] A.S. Ali, K.R. Mahmoud, K.M. Naguib, Optimal caching policy for wireless
optimization algorithms for scheduling security-critical tasks in resource-limited content delivery in D2D networks, J. Netw. Comput. Appl. 150 (2020) 102467.
mobile edge computing, Future Gener. Comput. Syst. 112 (2020) 148–161. [179] H. Zhang, Y. Wang, H. Ji, X. Li, A sleeping mechanism for cache-enabled small
[154] H. Peng, W.-S. Wen, M.-L. Tseng, L.-L. Li, Joint optimization method for cell networks with energy harvesting function, IEEE Trans. Green Commun.
task scheduling time and energy consumption in mobile cloud computing Netw. 4 (2) (2020) 497–505.
environment, Appl. Soft Comput. 80 (2019) 534–545. [180] Q.-V. Pham, F. Fang, V.N. Ha, M.J. Piran, M. Le, L.B. Le, W.-J. Hwang, Z.
[155] M.S.A. Khan, R. Santhosh, Task scheduling in cloud computing using hybrid Ding, A survey of multi-access edge computing in 5G and beyond: Funda-
optimization algorithm, Soft Comput. 26 (23) (2022) 13069–13079.
mentals, technology integration, and state-of-the-art, IEEE Access 8 (2020)
[156] P. Pirozmand, A.A.R. Hosseinabadi, M. Farrokhzad, M. Sadeghilalimi, S.
116974–117017.
Mirkamali, A. Slowik, Multi-objective hybrid genetic algorithm for task schedul-
[181] S. Guan, A. Boukerche, A novel mobility-aware offloading management scheme
ing problem in cloud computing, Neural Comput. Appl. 33 (19) (2021)
in sustainable multi-access edge computing, IEEE Trans. Sustain. Comput. 7 (1)
13075–13088.
(2021) 1–13.
[157] I. Strumberger, M. Tuba, N. Bacanin, E. Tuba, Cloudlet scheduling by hybridized
[182] X. Yang, D. Qin, J. Liu, Y. Li, Y. Zhu, L. Ma, Deep reinforcement learning in
monarch butterfly optimization algorithm, J. Sensor Actuator Netw. 8 (3)
NOMA-assisted UAV networks for path selection and resource offloading, Ad
(2019) 44.
Hoc Netw. (2023) 103285.
[158] L.N. Huynh, Q.-V. Pham, X.-Q. Pham, T.D. Nguyen, M.D. Hossain, E.-N. Huh,
[183] H.-G.T. Pham, Q.-V. Pham, A.T. Pham, C.T. Nguyen, Joint task offloading
Efficient computation offloading in multi-tier multi-access edge computing
and resource management in NOMA-based MEC systems: A swarm intelligence
systems: A particle swarm optimization approach, Appl. Sci. 10 (1) (2019) 203.
approach, IEEE Access 8 (2020) 190463–190474.
[159] T. Alfakih, M.M. Hassan, M. Al-Razgan, Multi-objective accelerated particle
swarm optimization with dynamic programing technique for resource allocation [184] S. Bitam, S. Zeadally, A. Mellouk, Fog computing job scheduling optimization
in mobile edge computing, IEEE Access 9 (2021) 167503–167520. based on bees swarm, Enterpr. Inf. Syst. 12 (4) (2018) 373–397.
[160] F. Xu, Z. Qin, L. Ning, Z. Zhang, Research on computing offloading strategy [185] S.K. Mishra, D. Puthal, J.J. Rodrigues, B. Sahoo, E. Dutkiewicz, Sustainable
based on genetic ant colony fusion algorithm, Simul. Model. Pract. Theory 118 service allocation using a metaheuristic technique in a fog server for industrial
(2022) 102523. applications, IEEE Trans. Ind. Inform. 14 (10) (2018) 4497–4506.
[161] D.D. Lieira, M.S. Quessada, A.L. Cristiani, E. Robson, R.I. Meneguette, Mech- [186] C. Zhu, J. Tao, G. Pastor, Y. Xiao, Y. Ji, Q. Zhou, Y. Li, A. Ylä-Jääski, Folo:
anism for optimizing resource allocation in VANETs based on the PSO Latency and quality optimized task allocation in vehicular fog computing, IEEE
bio-inspired algorithm, in: 2022 18th International Conference on Distributed Internet Things J. 6 (3) (2018) 4150–4161.
Computing in Sensor Systems, DCOSS, IEEE, 2022, pp. 283–290. [187] X. Hou, Z. Ren, J. Wang, W. Cheng, Y. Ren, K.-C. Chen, H. Zhang, Reliable
[162] D. He, R. Wu, X. Li, S. Chan, M. Guizani, Detection of vulnerabilities of computation offloading for edge-computing-enabled software-defined IoV, IEEE
blockchain smart contracts, IEEE Internet Things J. (2023). Internet Things J. 7 (8) (2020) 7097–7111.
[163] S. Seng, X. Li, H. Ji, H. Zhang, Joint access selection and heterogeneous [188] C. Chen, L. Chen, L. Liu, S. He, X. Yuan, D. Lan, Z. Chen, Delay-optimized v2v-
resources allocation in UDNs with MEC based on non-orthogonal multiple based computation offloading in urban vehicular edge computing and networks,
access, in: 2018 IEEE International Conference on Communications Workshops, IEEE Access 8 (2020) 18863–18873.
ICC Workshops, IEEE, 2018, pp. 1–6. [189] S. Shao, Q. Zhang, S. Guo, F. Qi, Task allocation mechanism for cable real-
[164] Q.-V. Pham, S. Mirjalili, N. Kumar, M. Alazab, W.-J. Hwang, Whale optimization time online monitoring business based on edge computing, IEEE Syst. J. 15 (1)
algorithm with applications to resource allocation in wireless networks, IEEE (2020) 1344–1355.
Trans. Veh. Technol. 69 (4) (2020) 4285–4297. [190] M.K. Hussein, M.H. Mousa, Efficient task offloading for IoT-based applica-
[165] S. Zhu, L. Gui, J. Chen, Q. Zhang, N. Zhang, Cooperative computation offloading tions in fog computing using ant colony optimization, IEEE Access 8 (2020)
for uavs: A joint radio and computing resource allocation approach, in: 2018 37191–37201.
IEEE International Conference on Edge Computing, EDGE, IEEE, 2018, pp. [191] S. Sharma, H. Saini, Fog assisted task allocation and secure deduplication using
74–79. 2FBO2 and MoWo in cluster-based industrial IoT (IIoT), Comput. Commun. 152
[166] L. Sun, L. Wan, X. Wang, Learning-based resource allocation strategy for (2020) 187–199.
industrial IoT in UAV-enabled MEC systems, IEEE Trans. Ind. Inform. 17 (7) [192] M. Adhikari, S.N. Srirama, T. Amgoth, Application offloading strategy for
(2020) 5031–5040. hierarchical fog environment through swarm optimization, IEEE Internet Things
[167] S. Deng, G. Cheng, H. Zhao, H. Gao, J. Yin, Incentive-driven computation J. 7 (5) (2019) 4317–4328.
offloading in blockchain-enabled E-commerce, ACM Trans. Internet Technol. [193] P.-Q. Huang, Y. Wang, K. Wang, Z.-Z. Liu, A bilevel optimization approach for
(TOIT) 21 (1) (2020) 1–19. joint offloading decision and resource allocation in cooperative mobile edge
[168] A. Bani-Bakr, M.N. Hindia, K. Dimyati, E. Hanafi, T.F. Tengku Mohmed computing, IEEE Trans. Cybern. 50 (10) (2019) 4228–4241.
Noor Izam, Multi-objective caching optimization for wireless backhauled fog
[194] H. Kurdi, M.F. AlDaood, S. Al-Megren, E. Aloboud, A.S. Aldawood, K. Youcef-
radio access network, Symmetry 13 (4) (2021) 708.
Toumi, Adaptive task allocation for multi-UAV systems based on bacteria
[169] H. Jin, H. Lu, Y. Jin, C. Zhao, IVCN: Information-centric network slicing
foraging behaviour, Appl. Soft Comput. 83 (2019) 105643.
optimization based on NFV in fog-enabled RAN, IEEE Access 7 (2019)
[195] H. Alsolai, W. Mtouaa, M.S. Maashi, M. Othman, I. Yaseen, A.A. Alneil, A.E.
69667–69686.
Osman, M.I. Alsaid, Optimization of drone base station location for the next-
[170] J. Feng, Z. Liu, C. Wu, Y. Ji, AVE: Autonomous vehicular edge computing
generation internet-of-things using a pre-trained deep learning algorithm and
framework with ACO-based scheduling, IEEE Trans. Veh. Technol. 66 (12)
NOMA, Mathematics 11 (8) (2023) 1947.
(2017) 10660–10675.
[196] D. Pliatsios, S.K. Goudos, T. Lagkas, V. Argyriou, A.-A.A. Boulogeorgos, P.
[171] N. Jabeen, R. Hao, A. Niaz, M.U. Shoukat, F. Niaz, M.A. Khan, Autonomous
Sarigiannidis, Drone-base-station for next-generation internet-of-things: A com-
vehicle health monitoring based on cloud-fog computing, in: 2022 International
parison of swarm intelligence approaches, IEEE Open J. Antennas Propag. 3
Conference on Emerging Trends in Electrical, Control, and Telecommunication
(2021) 32–47.
Engineering, ETECTE, IEEE, 2022, pp. 1–6.
[197] M. Abdel-Basset, L. Abdel-Fatah, K.A. Eldrandaly, N.M. Abdel-Aziz, En-
[172] N. Sharma, P. Garg, et al., Ant colony based optimization model for qos-
hanced computational intelligence algorithm for coverage optimization of 6G
based task scheduling in cloud computing environment, Measurement 24 (2022)
non-terrestrial networks in 3D space, IEEE Access 9 (2021) 70419–70429.
100531.
[173] H. Wu, F. Lyu, X. Shen, H. Wu, F. Lyu, X. Shen, Optimal uav caching and [198] S. Yuan, B. Han, D. Krummacker, H.D. Schotten, Massive twinning to enhance
trajectory design in the agvn, in: Mobile Edge Caching in Heterogeneous emergent intelligence, 2022, arXiv preprint arXiv:2204.09316.
Vehicular Networks, Springer, 2022, pp. 61–88. [199] A. El Amraoui, Metaheuristic moth flame optimization based energy efficient
[174] H. Wu, J. Chen, F. Lyu, L. Wang, X. Shen, Joint caching and trajectory design clustering protocol for 6G enabled unmanned aerial vehicle networks, in:
for cache-enabled UAV in vehicular networks, in: 2019 11th International AI-Enabled 6G Networks and Applications, Wiley Online Library, 2023, pp.
Conference on Wireless Communications and Signal Processing, WCSP, IEEE, 1–15.
2019, pp. 1–6. [200] H. Ganame, L. Yingzhuang, A. Hamrouni, H. Ghazzai, H. Chen, Evolutionary
[175] C. Huang, J. Cao, S. Wang, Y. Zhang, Dynamic resource scheduling optimization algorithms for 5g multi-tier radio access network planning, IEEE Access 9
with network coding for multi-user services in the internet of vehicles, IEEE (2021) 30386–30403.
Access 8 (2020) 126988–127003. [201] D. Wang, M. Wu, Y. He, L. Pang, Q. Xu, R. Zhang, An HAP and UAVs
[176] C. Yang, Y. Yao, Z. Chen, B. Xia, Analysis on cache-enabled wireless collaboration framework for uplink secure rate maximization in NOMA-enabled
heterogeneous networks, IEEE Trans. Wireless Commun. 15 (1) (2015) 131–145. IoT networks, Remote Sens. 14 (18) (2022) 4501.
[177] R.Q. Hu, L. Hanzo, et al., Heterogeneous networks relying on full-duplex relays [202] O. Alamu, B. Iyaomolere, A. Abdulrahman, An overview of massive MIMO
and mobility-aware probabilistic caching, IEEE Trans. Commun. 67 (7) (2019) localization techniques in wireless cellular networks: Recent advances and
5037–5052. outlook, Ad Hoc Netw. 111 (2021) 102353.

34
A.K. Abasi et al. Ad Hoc Networks 158 (2024) 103474

[203] H. Zhang, H. Zhang, B. Di, K. Bian, Z. Han, L. Song, Metalocalization: Ammar K. Abasi received the B.Sc. degree in computer in-
Reconfigurable intelligent surface aided multi-user wireless indoor localization, formation system from the Jordan University of Science and
IEEE Trans. Wireless Commun. 20 (12) (2021) 7743–7757. Technology, the M.Sc. degree in international business from
[204] S. Phoemphon, C. So-In, D.T. Niyato, A hybrid model using fuzzy logic and an The University of Jordan, and the Ph.D. degree in artificial
extreme learning machine with vector particle swarm optimization for wireless intelligence and software engineering from Universiti Sains
sensor network localization, Appl. Soft Comput. 65 (2018) 101–120. Malaysia. His research interests include evolutionary algo-
[205] J. He, Y.J. Chun, H.C. So, M. EURASIP, Modeling and performance analysis rithms, nature inspired computation, artificial intelligence,
of blockchain-aided secure TDOA localization under random internet-of-vehicle and their applications to optimization problems. Currently,
networks, Signal Process. 206 (2023) 108904. he is a Postdoctoral Fellow at MBZUAI.
[206] W. Kang, H. Che, K. Zhu, Precoding optimization for faster-than-nyquist
signaling with probabilistic shaping, IEEE Trans. Veh. Technol. 69 (10) (2020)
Moayad Aloqaily received a Ph.D. degree in Computer
11470–11478.
[207] B. Qolomany, K. Ahmad, A. Al-Fuqaha, J. Qadir, Particle swarm optimized Engineering from the University of Otawa, Canada, in 2016.
He is currently with the Machine Learning Department,
federated learning for industrial IoT and smart city services, in: GLOBECOM
Mohamed Bin Zayed University of Artificial Intelligence
2020-2020 IEEE Global Communications Conference, IEEE, 2020, pp. 1–6.
(MBZUAI), UAE. He was the recipient of many honors
[208] G. Verhulsdonck, J.L. Weible, S. Helser, N. Hajduk, Smart cities, playable cities,
and awards, such as best paper awards of 2020 Ad Hoc
and cybersecurity: A systematic review, Int. J. Hum. Comput. Interact. 39 (2)
Networks Journal, 2021 Computer Networks, 2022 IEEE
(2023) 378–390.
IWCMC, 2022 IEEE MeditCom, 2022 IEEE GlobeCom, and
[209] M. Letafati, S. Otoum, On the privacy and security for e-health services in the
2023 IEEE Metaverse. His current research interests include
metaverse: An overview, Ad Hoc Netw. (2023) 103262.
the applications of AI and ML, blockchain solutions, and
[210] R. Dangi, P. Lalwani, Harris hawks optimization based hybrid deep learning
sustainable energy and data management. Dr. Aloqaily
model for efficient network slicing in 5G network, Cluster Comput. (2023) 1–15.
[211] Y. Su, Z. Gao, X. Du, M. Guizani, User-centric base station clustering and has chaired and co-chaired many IEEE conferences and
workshops. He is serving on the Editorial Board of many
resource allocation for cell-edge users in 6G ultra-dense networks, Future Gener.
prestigious journals and guest editor in many journals.
Comput. Syst. 141 (2023) 173–185.
He has also been appointed as the Co-Editor-in-Chief of
[212] R. Dhaya, R. Kanthavel, Bus-based VANET using ACO multipath routing
IEEE CommSoft TC eLeter in 2020. He started his Special
algorithm, J. Trends Comput. Sci. Smart Technol. (TCSST) 3 (01) (2021) 40–48.
[213] G.G. Lema, Handover performance evaluation under dynamic user characteris- Interest Groups on Blockchain and Application as well as the
Internet of Unmanned Aerial Networks. He is an IEEE Senior
tics, Wirel. Pers. Commun. 120 (4) (2021) 2991–3004.
[214] N. Chandnani, C.N. Khairnar, Bio-inspired multilevel security protocol for Member, ACM Member, and a Professional Engineer Ontario
(P.Eng.). He is the founder and director of artifitialleap.ai
data aggregation and routing in IoT WSNs, Mob. Netw. Appl. 27 (3) (2022)
laboratory.
1030–1049.
[215] R. Lan, Y. Zhu, H. Lu, Z. Liu, X. Luo, A two-phase learning-based swarm
optimizer for large-scale optimization, IEEE Trans. Cybern. 51 (12) (2020)
Mohsen Guizani is currently a Professor of Machine
6284–6293.
Learning at Mohamed Bin Zayed University of Artificial
[216] M.A. Al-Betar, M.A. Awadallah, I.A. Doush, O.A. Alomari, A.K. Abasi, S.N.
Intelligence (MBZUAI), Abu Dhabi, UAE. His research inter-
Makhadmeh, Z.A.A. Alyasseri, Boosting the training of neural networks through
ests include applied machine learning, artificial intelligence,
hybrid metaheuristics, Cluster Comput. 26 (3) (2023) 1821–1843.
Internet of Things, smart city, and cybersecurity. He was
[217] M. Mandloi, V. Bhatia, A low-complexity hybrid algorithm based on particle
elevated to the IEEE Fellow in 2009 and was listed as a
swarm and ant colony optimization for large-MIMO detection, Expert Syst. Appl.
Clarivate Analytics Highly Cited Researcher in Computer
50 (2016) 66–74.
Science in 2019, 2020 and 2021. Before joining MBZUAI,
[218] N.T. Nguyen, K. Lee, Deep learning-aided tabu search detection for large MIMO
Guizani served in multiple administrative positions in the
systems, IEEE Trans. Wireless Commun. 19 (6) (2020) 4262–4275.
USA and the Gulf region, such as the Founding Associate
[219] N.M. Hijazi, M. Aloqaily, M. Guizani, B. Ouni, F. Karray, Secure federated
Vice President for Graduate Studies at QU, Chair of the ECE
learning with fully homomorphic encryption for IoT communications, IEEE
Department at the University of Idaho, Chair of the Com-
Internet Things J. (2023).
puter Science Department at Western Michigan University
[220] Y. He, M. Yang, Z. He, M. Guizani, Computation offloading and resource
and Professor at the University of Missouri.
allocation based on DT-MEC-assisted federated learning framework, IEEE Trans.
Cognit. Commun. Netw. (2023).
[221] A.K. Abasi, M. Aloqaily, M. Guizani, F. Karray, Sine cosine algorithm for Bassem Ouni is currently a Lead Researcher at Technology
reducing communication costs of federated learning, in: 2022 IEEE International Innovation Institute, Abu Dhabi, United Arab Emirates. He
Mediterranean Conference on Communications and Networking, MeditCom, received his Ph.D. degree in Computer Science from the
IEEE, 2022, pp. 55–60. University of Nice-Sophia Antipolis, Nice, France, in July
[222] A.K. Abasi, M. Aloqaily, M. Guizani, Grey wolf optimizer for reducing com- 2013. Between October 2018 and January 2022, he held
munication cost of federated learning, in: GLOBECOM 2022-2022 IEEE Global a Lead Researcher position in the French Atomic Energy
Communications Conference, IEEE, 2022, pp. 1049–1154. Commission (CEA) within the LIST Institute, Paris, France,
[223] H. Zhang, J. Xue, Q. Wang, Y. Li, A security optimization scheme for data and an Associate Professor/Lecturer position at the Univer-
security transmission in UAV-assisted edge networks based on federal learning, sity of Paris Saclay and ESME Sudria Engineering school,
Ad Hoc Netw. (2023) 103277. Paris, France. Prior to that, he worked as a Lead Researcher
[224] S.A. Balamurugan, S.S. Kumar, IDsMA: An integrated digital signature and between 2017 and 2018 within the department of Elec-
mutual authentication mechanism for securing the cognitive radio networks, tronics and Computer Science, University of Southampton,
Int. J. Commun. Syst. 33 (6) (2020) e4289. Southampton, United Kingdom. Before that, he occupied the
[225] A.K. Abasi, M. Aloqaily, B. Ouni, M. Hamdi, Optimization of CNN-based position of a Research Scientist, between 2015 and 2016,
federated learning for cyber-physical detection, in: 2023 IEEE 20th Consumer at the Institute of Technology in Aeronautics, Space and
Communications & Networking Conference, CCNC, IEEE, 2023, pp. 1–6. Embedded Systems (IRT-AESE) located in Toulouse, France.
[226] S. Kiran, M.R. Babu, B. Kiranmai, K. Gurucharan, Advanced wireless communi- From September 2013 to the end of 2014, he held a post-
cations for future technologies-6G and beyond 6G, in: Wireless Communication doctoral fellow position in Eurecom, Telecom ParisTech
with Artificial Intelligence, CRC Press, 2023, pp. 1–26. institute, Sophia Antipolis, France. Furthermore, he worked
[227] I. Erol, I.M. Ar, I. Peker, Scrutinizing blockchain applicability in sustainable between 2009 and 2013 as a lecturer at the University of
supply chains through an integrated fuzzy multi-criteria decision making Nice Sophia Antipolis (Polytech Nice Engineering School
framework, Appl. Soft Comput. 116 (2022) 108331. and Faculty of Sciences of Nice). Also, he was managing
[228] Q. Ma, W. Gao, Q. Xiao, L. Ding, T. Gao, Y. Zhou, X. Gao, T. Yan, C. Liu, Z. several industrial collaborations with ARM, Airbus Group
Gu, et al., Directly wireless communication of human minds via non-invasive Innovation, Rolls Royce, Thales Group, Continental, Actia
brain-computer-metasurface platform, elight 2 (1) (2022) 1–11. Automotive Group, etc. He co-authored many publications
[229] T.B. Ahammed, R. Patgiri, S. Nayak, A vision on the artificial intelligence for (Book Chapters, Journals, and international conferences.).
6G communication, ICT Express 9 (2) (2023) 197–210. He is an IEEE senior member.
[230] L.U. Khan, M. Guizani, D. Niyato, A. Al-Fuqaha, M. Debbah, Metaverse for
wireless systems: Architecture, advances, standardization, and open challenges,
2023, arXiv preprint arXiv:2301.11441.

35