Received: 19 September 2023 Revised: 27 December 2023 Accepted: 13 January 2024 The Journal of Engineering
DOI: 10.1049/tje2.12353
ORIGINAL RESEARCH
Joint delay and energy aware dragonfly optimization-based uplink
resource allocation scheme for LTE-A networks in a cross-layer
environment
Leeban Moses1 Perarasi Sambantham1 Muhammad Faheem2 Shoukath Ali K3
Arfat Ahmad Khan4
1Department of Electronics and Communication
Engineering, Bannari Amman Institute of
Technology, Sathyamangalam, Tamil Nadu, India
2Department of Computing Science, School of
Technology and Innovations, University of Vaasa,
Vaasa, Finland
3Department of Electronics and Communication
Engineering, Presidency University, Bengaluru,
Karnataka, India
4Department of Computer Science, College of
Computing, Khon Kaen University, Khon Kaen,
Thailand
Correspondence
Muhammad Faheem, School of Technology and
Innovations, University of Vaasa, Vaasa 65200,
Finland.
Email: muhammad.faheem@uwasa.fi
Abstract
The exponential growth in data traffic from smart devices has led to a need for highly capa-
ble wireless networks with faster data transmission rates and improved spectral efficiency.
Allocating resources efficiently in a 5G communication system with a huge number of
machine type communication (MTC) devices is essential to ensure optimal performance
and meet the diverse requirements of different applications. The LTE-A network offers
high-speed mobile data services and caters to MTC devices and has relatively low data
service requirements compared to human-to-human (H2H) communications. LTE-A net-
works require advanced scheduling schemes to manage the limited spectrum and ensure
efficient transmissions. This necessitates effective resource allocation schemes to minimize
interference between cells in future networks. To address this issue, a joint delay and energy
aware Levy flight Brownian movement-based dragonfly optimization (DELFBDO)-based
uplink resource allocation scheme for LTE-A Networks is proposed in this work to opti-
mize energy efficiency, maximize the throughput and reduce the latency. The DELFDO
algorithm efficiently organizes packets in both time and frequency domains for H2H and
MTC devices, resulting in improved quality of service while minimizing energy consump-
tion. The Simulation results demonstrate that the proposed method increases the energy
efficiency by producing the appropriate channel and power assignment for UEs and MTC
devices.
1 INTRODUCTION
Advancements in wireless communication networks have made
it possible for network operators to meet the growing demand
for services. In the next generation of networks, it is expected
that each user device will be able to use multiple applications
simultaneously, each with its own specific quality of service
(QoS) requirements [1]. For instance, the applications like aug-
mented or virtual reality require high data rates, low delay, and
reliable connections. The rapid progress in wireless commu-
nication technologies has led to the development of 5G (fifth
generation) networks. The 5G network needs to cater to diverse
types of applications and devices that operate without wires. It
must provide different levels of service for various scenarios
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is
properly cited.
© 2024 The Authors. The Journal of Engineering published by John Wiley & Sons Ltd on behalf of The Institution of Engineering and Technology.
[2]. These scenarios include faster internet access for mobile
devices, connecting a large number of machines, and ensur-
ing reliable and fast communication with minimal delays. In
Internet of Things (IoT) networks based on cellular technol-
ogy, numerous IoT devices initially connect to the cellular base
station for network access directly or through IoT gateways.
This is done through a random-access procedure [3]. How-
ever, when a large number of IoT devices simultaneously access
and transmit small data packets or frequently send real-time
status updates, collisions in the initial signals are unavoidable.
Improving the access mechanism of current cellular systems
to enhance the QoS and reduce the power consumption of
IoT devices is a significant challenge for cellular-based IoT
networks [4].
J. Eng. 2024;2024:e12353. wileyonlinelibrary.com/iet-joe 1 of 16
https://doi.org/10.1049/tje2.12353
2 of 16 MOSES M ET AL.
FIGURE 1 Single cell scenario with one Human Type Communication Device and N Machine Type Communication devices with k-sensor data types.
Different IoT services have varying QoS requirements based
on traffic data and functionality. Providing access mechanisms
with satisfactory network performance for a massive number of
IoT devices and diverse network services is a key challenge. Het-
erogeneous networks (HetNets) are considered a useful method
to improve the efficiency of wireless networks and expand their
capacity [5]. This is achieved by transferring some of the work-
load from large base stations to smaller ones that cover smaller
areas. By reducing the size of the cells and using different trans-
mission power levels, HetNets offer benefits such as better
coverage, improved connectivity indoors, balanced distribution
of network traffic, and flexible deployment options [6]. How-
ever, the coexistence of large and small base stations in HetNets
can present challenges in managing network resources, han-
dling interference, ensuring security, maintaining QoS, scaling
the network, and optimizing energy usage. M2M communi-
cation for Human Type Communication (HTC) interactions
can lead to efficiency and congestion challenges due to the
varying data rates between them [7]. As shown in Figure 1,
an eNB serves both HTC devices for voice calls and multi-
media communications and N MTC devices with K different
types of sensor data, each with a unique data size and sam-
pling period. The differences in signalling packet sizes at higher
layers result in lower efficiency. In certain areas, there can be
a large number of MTC devices that communicate with each
other. These devices are often triggered by external events, and
if deployed quickly in large numbers, they can cause network
congestion. Unlike devices used for person-to-person commu-
nication, MTC devices send much more data from the device
to the network (uplink) than they receive from the network [8].
This difference in data flow can impact the overall efficiency
of the network. MTC devices also have specific needs, such as
long battery life, short delays, and reliable connections. These
needs aren’t usually important for other types of devices, like
smartphones, and the current cellular network technology isn’t
designed for them, so they do not work as well as they could.
Additionally, MTC devices send data regularly from specific
locations and do not move much, which is different from how
people use their phones [9].
LTE-A is a cellular communication standard that provides
high bandwidth, mobility, and flexibility to meet the different
needs of UEs, but MTC involves low-bandwidth data bursts
and has different QoS requirements compared to H2H com-
munication. MTC data transmission primarily depends on the
uplink direction and competes with uplink H2H traffic [10].
There are various critical applications of MTC communication
that require prioritization over H2H communication, and man-
aging this competition is a challenge for the uplink scheduler.
Since M2M systems involve a large number of devices with
diverse QoS requirements, designing scheduling schemes that
support MTC over the LTE-A network is a difficult task. MTC
devices often send infrequent data packets of varying sizes, and
the LTE-A bandwidth has limited physical resources optimized
for H2H communication. To ensure efficient MTC communi-
cation, it is important to develop a solution that optimizes the
use of available resources while meeting the specific quality of
QoS requirements of MTC communication. Several optimiza-
tion algorithms used for resource allocation (RA) can also be
performed and tested in the transmission path using antenna
like falcate patch [11], leaky-wave microstrip patch [12], and
isolation microstrip patch [13].
Interference management is a major challenge in MTC com-
munication, as it can affect the quality of communication or
even lead to complete loss of communication when multiple
devices use the same resources [14]. Techniques like power con-
trol, frequency selection, and resource allocation can be used
to address this challenge and minimize interference. Energy
efficiency is another significant challenge in MTC communi-
cation, especially when there are a large number of devices in
the network. It is important to minimize power consumption
while ensuring effective communication. Techniques such as
power control, duty cycling (turning devices on and off period-
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 3 of 16
ically) and implementing sleep modes can help enhance energy
efficiency [15]. Additionally, optimal power allocation plays a
significant role in enhancing network performance, mitigating
interference, and improving energy efficiency. However, it is
important to note that resource allocation schemes have a crit-
ical impact on both QoS and energy efficiency, and optimizing
one parameter may have trade-offs with other parameters [16].
Therefore, joint optimization of user association and power
allocation is essential to enhance energy efficiency.
To address these challenges, we propose a novel cross-layer
joint delay and energy aware Levy flight dragonfly optimization
algorithm for the LTE-A Network with MTC infrastructure.
Our focus is on the coexistence of crowded MTC users and
traditional cellular users within a network environment. In this
scenario, the base station (BS) exercises centralized control over
all users within a cell. It has the capability to collect channel
state information from the system’s links and allocate spectrum
resources to individual users based on their specific require-
ments. In our proposed approach, we assume that all spectrum
resources are shared among both traditional cellular users and
newly introduced MTC users. By considering the transmission
opportunities for traditional cellular users and analyzing the
interference levels in the MTC communication mode, we assign
communication modes to all users. This algorithm aims to allo-
cate resources to users, ensuring better QoS and fairness index.
The rest of the Paper is organized as follows: Section 2 illustrates
the related work; Section 3 presents the resource allocation
problem and establishes the system model. In Section 4, a novel
resource allocation algorithm is introduced, considering the
multi-objective function. Section 5 evaluates the performance
parameters of the proposed algorithm.
2 RELATEDWORK
Resource allocation is the process of assigning frequency com-
ponents to different types of services. It determines how much
time and frequency is shared between the base station and
mobile users. During each scheduling phase in the uplink
(UL) channel, each HTC-UE/MTCD sends its buffer status
information and buffer size to the eNB using a dedicated
channel called the physical uplink control channel (PUCCH).
Scheduling algorithms, which manage traffic flows at a base sta-
tion while considering QoS, can be broadly categorized into
two types: channel independent and channel dependent. In
channel independent scheduling, resource allocation doesn’t
consider channel state information or buffer size. On the
other hand, channel dependent scheduling allocates resources
to UEs based on channel state information. These algorithms
are beneficial for wireless cellular networks aiming to provide
uninterrupted services to UEs. Channel dependent scheduling
algorithms are specifically designed for applications with guar-
anteed bit rate (GBR) and those sensitive to delays [10]. The
eNodeB’s packet scheduler employs specific strategies to allo-
cate the shared spectrum among users. To achieve maximum
spectral efficiency, an optimal resource allocation technique is
employed.
Heuristic algorithms are already in use to significantly reduce
computational complexity. Multiobjective resource allocation
schemes have been developed to achieve desired outcomes,
such as minimizing interference, improving fairness, and
enhancing resource utilization efficiency. The combination of
firefly algorithm and particle swarm optimization (FFA-PSO)
algorithm [17] are employed to obtain a Pareto-optimal solution.
A two-sided matching RA algorithm for optimizing and allo-
cating the resources between HTC-UEs is implemented in [18]
to ensure stability and manage complexity. This approach uses
multipoint transmission and reception, along with carrier aggre-
gation, to send multiple components to users in different cell
corners. The aim is to make efficient use of bandwidth, improve
received signals, and reduce interference. The research article in
[19] enumerate a mode selection algorithm to effectively allocate
the resources available in a mixed traffic environment by consid-
ering the MTC and HTC traffic. In this approach, we determine
the communication mode for users based on traditional cellular
transmission opportunities and analyze interference from dif-
ferent MTC communication modes. Users are then categorized
into three tiers based on their distances from the base station
to address interference problems. Spectrum resource allocation
mechanisms are developed for different communication modes
using the Hungarian algorithm.
The chunk-based and heuristic-based RA algorithm (CBRA
& HORA) developed in [20] solves the tradeoff that exists
between the performance of QoS parameter and computational
complexity of the algorithm. The HORA algorithm allocates
resource blocks to users, taking into account their QoS require-
ments, while CRBA assigns RB chunks to user groups based
on channel conditions. The HORA algorithm optimizes data
rate through 3-tuplet constraints, while the CRBA algorithm
groups users by signal to noise ratio (SNR) values, assembles RB
chunks, and allocates them based on channel conditions and RB
requirements. The cross-layer resource allocation formulated
by [21] enhances the RA mechanism of HTC-UEs in a single
cell by considering the HTC traffic. Statistical QoS assurance
is applied to fulfil critical MTC demands in the cell, restricting
the probability of delay-bound violation for each critical MTC
device. This approach provides an accurate statistical guarantee
for latency and packet losses due to queuing delays. The author
in [22] has proposed an optimized bidirectional RA algorithm
for HTC UEs based on QoS requirements. Users are priori-
tized based on a probability function, and scheduling occurs
by selecting each user from the higher-ranked positions in the
priority queue. These weights are transformed into a probabil-
ity function during each transmission time interval, based on
QoS requirements. In [23], the author tackled the complex joint
problem of allocating resources and power. They used a mixed
integer programming model to handle the stochastic optimiza-
tion issue associated with the drift-plus-penalty (DPP) approach
for Lyapunov opportunistic optimization.
The amended barnacles mating optimizer (ABMO)- [24]
and modified whale optimizer-based [25] RA algorithms, aims
to improve the performance of 5G networks by enhanc-
ing the number of connected users and reducing transmit
power consumption. This algorithm increases its accuracy and
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 of 16 MOSES M ET AL.
effectiveness by optimizing the best locations for base stations,
considering the principles of green communications. A three-
step genetic algorithm (GA) for coexistence of MTC and HTC
UE proposed in [26] includes new operations like initialization,
crossover, mutation, and a QoS-aware fitness function. GA with
a single block approach leads to slow convergence as the algo-
rithm spends time evaluating solutions, even when users could
delay their transmission without exceeding the delay budget.
Sliding window-based RA algorithm for coexistence of MTC
devices with HTC UEs is proposed in [27] takes advantage of
frequency and time diversities by capturing channel coefficients
that vary over time and frequency. To improve uplink through-
put in network resource allocation and to enhance energy
efficiency, particle swarm optimization (PSO) with fuzzy logic
control (FLC) and Taguchi method is proposed in [28]. The pro-
posed method utilizes priority weighted rate control to increase
uplink throughput. PSO is employed to determine the optimal
position for the cluster head, while FLC combined with the
Taguchi method adjusts the PSO fitness parameter value.
The RSMA technique involves splitting a message into two
parts for transmission. Two resource allocation modes are cre-
ated to maximize throughput while adhering to interference
constraints. The switch between modes is determined by deriv-
ing the coverage radius expression. The optimization of D2D
transmit power and reduction of overall network computational
complexity are achieved using Lagrange’s dual-optimization
method. The author in [3] has proposed a scalable type of
priority-based RA scheme when MTC devices coexist with
HTCUEs in a single cell. This allocation scheme considers both
channel quality and application priority as metrics for distribu-
tion. This is beneficial to minimize the balance between making
efficient use of the channel and ensuring satisfaction with QoS.
In [2], the author uses a combination of the fixed variable
method and the Lagrangian method to address the issues faced
in resource allocation for a mixed environment. They suggest an
enhanced particle swarmmethod for resolving the power alloca-
tion problem and introduce a bilateral stable matching algorithm
to find a solution for energy cooperation. To enhance energy
efficiency, the author puts forth a convergent iterative algorithm
that integrates the three mentioned methods to jointly solve the
original optimization problem.
Table 1 summarizes the objectives of the existing work. Here
EE denotes energy efficiency of the HTC/MTC nodes, SE
represents the spectrum efficiency of the resource allocation
strategy, TE denotes the efficiency of throughput, QoS denotes
the QoS parameters, IA denotes the signal to noise interfer-
ence parameter, DA denotes the delay parameter, FI denotes the
Fairness index parameter and CLO denotes whether the model
represented represents cross layer scenarios by considering two
objective function or not.
3 SYSTEMMODEL AND PROBLEM
FORMULATION
We consider the single-cell architecture illustrated in Figure 1. A
cell consists of n UE terminals and mM2M devices to form sets
M = {1, 2, 3, …. m} and N = {1, 2, 3, …. n} of UE terminals
and M2M devices, respectively. The frequency division duplex
scheme is adapted in M2M devices since the channel occu-
pies half of the spectrum. One uplink and downlink orthogonal
channels are allocated to UE to prevent interference between
two UEs. To perform allocation and mitigate interference, we
assume that the uplink and downlink channels can be reused
once by the M2M spectrum, and each M2M device cannot reuse
the channels. The channel gain between nth UE and the base
station of a single cell can be denoted as:
GnUE = 𝜆𝜓nUE 𝜌nUE 𝓁
−𝛼
nUE (1)
where λ represents the path loss constant, 𝜓nUE represents the
fast-fading gain, 𝜌nUE represents the slow fading gain. 𝓁nUE
represents the length between the nth UE and BS, and α rep-
resents the path loss factor. To define the system capacity, the
status parameter of the UE in the uplink channel can be denoted
by a binary variable and is denoted as:
Γ
uplink
mn =
{
1 whenM2M device reuses the uplinkchannel o f nth UE
0 otherwise
(2)
When the nth UE is allocated in the uplink channel it can use
its entire spectral width and its signal to interference and noise
ratio for nth UE and base station shall be denoted as:
Γin,UL =
Ekn GnUE
𝛽2 +
∑
m∈M E
l
m𝛼
u
mnymn
(3)
where Ekn ,E
l
m denotes the transmission energy of nth UE and
mth M2M device pair, GnUE and 𝛼
u
mn represents uplink channel
gain of UE and BS, M2M and BS respectively and 𝛽2 denotes
the Energy of the Noise Gaussian on both D2D Pair and UE
channels. In uplink channel the UE causes interference with the
M2M pairs and Signal to Interference Noise Ratio (SINR) of the
mth M2M pair under the channel of the nth UE can be denoted
as:
Γ
j
m,UL =
ElmGm,M2M
𝛽2 +
∑
n∈N E
k
n 𝛼
u
mnymn
(4)
where Gm,M2M is the channel gain for uplink of the M2M
devices and BS, ymn denotes gain of the interference channel
link between nth UE and mth M2M devices. Considering of both
SINR of M2M device pairs and UE with base station the overall
system capacity can be written as:
C
Uplink
UE , M2M = B log2
(
1 + Γin,UL
)
+ B log2
(
1 + Γ jm,UL
)
(5)
An optimal resource allocation scheme is proposed here
for uplink communication by considering both UE and M2M
device pairs in a single cell structure. We consider the optimiza-
tion problem as the maximization of capacity system and to
assure the SINR requirements. Overall optimization related to
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 5 of 16
TABLE 1 Comparison of multi-objective hybrid optimization-based resource allocation models.
Reference Model EE SE TE QoS IA DA FI CLO Priority
[19] Energy-aware mode
selection for D2D
resource allocation
✓ - ✓ ✓ - - ✓ Resource allocation based
on MTC devices with
single cell
Statistical
QoS
[17] FFA-PSO multiobjective
cooperative swarm
intelligence algorithm
- - ✓ ✓ - - ✓ Resource allocation based
on HTC devices with
macro/femto cell
GBR &
Max
weight
[20] HORA and CRBA
algorithms
- - ✓ ✓ ✓ - ✓ HTC devices with three
eNBs and based on MCS
CQI
[21] Two sided cross-layer
matching-based
scheduler
- ✓ ✓ ✓ - ✓ - single eNB that serves a set
of HTC users and critical
MTCD
CQI,
Delay
[22] Bidirectional allocation of
radio blocks using
MUJCS algorithm based
on weight factors
- ✓ ✓ ✓ - - ✓ Both uplink and downlink,
RA based on HTC
devices with
macro/femto cell
GBR &
Max
weight
[23] Joint stochastic Lyapunov
DPP optimization to
solve MILP problem for
the multiple resource
allocation
✓ ✓ ✓ - ✓ - - Multi-tier and multi-cell
heterogeneous Network
for micro and small base
station UEs
QCI and
GBR
[25] Whale optimization
algorithm-based
resource allocation
scheme
- - ✓ ✓ - - ✓ Resource allocation based
on MTC devices with
single cell
CQI
[28] Energy efficient RA
method by combination
of PSO, FLC with
Taguchi method
✓ ✓ ✓ - ✓ - - Allocation of resources for
IoT devices
Statistical
QoS
[29] RSMA technique with
Han–Kobayashi scheme
for RA in multicell
mm-Wave environment
✓ ✓ ✓ - ✓ - - D2D communication in a
multicell environment for
an uplink channel
QCI and
Max
weight
[30] Red–black tree building
TDPS classifier by
combining channel
quality and application
priority as an allocation
metric
✓ ✓ ✓ - ✓ ✓ ✓ Resource allocation based
on HTC devices with
macro/femto cell
QCI and
GBR
[26] Three-step genetic-based
metaheuristic RA in
mixed traffic
environments
- ✓ ✓ - - - ✓ Resource allocation based
on MTC devices and UE
with single cell
QCI and
GBR
[18] Matching-based cross-layer
resource allocation
- ✓ ✓ - - - ✓ Resource allocation based
on MTC devices and UE
with single cell
QCI and
GBR
[27] Sliding window-based RA
algorithm for Hybrid
eMBB and URLLC
Services
✓ ✓ ✓ - ✓ - - Resource allocation based
on MTC devices and UE
with single cell
QCI and
GBR
[24] Amended barnacles
mating optimizer-based
RA algorithm
- ✓ ✓ ✓ - ✓ ✓ Resource allocation based
on HTC devices with
macro/femto cell
GBR and
Max
weight
[31] FVI based RA algorithm
by combining Lagrange
dual method and
improved PSO with
matching.
✓ ✓ ✓ - ✓ - ✓ Resource allocation based
on HTC devices with
macro/femto cell
GBR and
Max
weight
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 of 16 MOSES M ET AL.
system capacity can be denoted as:
Max
Elm ,𝛼
u
mn, E
k
n
∑
n∈N
∑
m∈m
C
Uplink
UE ,M2M (6)
Such that:
Γin,UL ≥ Γ
i
min&Γ
j
m,UL = Γ
j
min, ∀n ∈ N &m ∈ M (7)
0 ≤ Ekn ≤ E
k
max&E
l
m ≤ E
l
max, ∀n ∈ N &m ∈ M (8)∑
n∈N
𝛼umn ≤ 1&𝛼
u
mn ∈ {0, 1}; ∀n ∈ N &m ∈ M (9)(∑
n∈N
𝛼umn
)(∑
m∈M
𝛼umn
)
= 0; ∀n ∈ N &m ∈ M (10)
where Γimin&Γ
j
min denotes the minimum value of SINR of UE
and M2M users respectively. Ekmax&E
l
max denotes the maxi-
mum transmitted energy of the UE and M2M users respectively.
According to the Constraints developed, the Equation (6) rep-
resents the QoS requirements of the UEs and M2M Pairs.
The constraint developed in (7) assures maximum energy trans-
mitted in the uplink channel of Cross-layer environment. The
constraint developed in (8) and (9) assures the assumed con-
dition, that each uplink channel of UE can be shared only by
one spectrum of M2M users in the same channel. The con-
straint developed in (10) assures that the M2M pair can share
only one spectrum in the uplink channel of UE. The next sec-
tion describes the energy-efficient resource allocation algorithm
according to the developed constraints.
To reduce the computational complexity of the proposed
algorithm, the objective function is decomposed into two terms.
So, the first objective for the constraint developed in (5) is to
maximize the energy transmission between the M2M allocated
pairs and its shared resource by the UE in uplink channel. The
second objective in the constraint developed in (5) is allocating
resources to UE and M2M users using the dragonfly optimiza-
tion algorithm. When the mth M2M device pair reuses the
uplink channel of the nth UE, the energy optimization problem
can be remodelled as:
Max
Elm ,E
k
n
{
log2
(
1 +
Ekn GnUE
𝛽2 +
∑
m∈M E
l
m𝛼
u
mnymn
)
+ log2
(
1 +
ElmGm,M2M
𝛽2 +
∑
n∈N E
k
n 𝛼
u
mnymn
)}
(11)
Such that, Γin,UL ≥ Γ
i
min&Γ
j
m,UL ≥ Γ
j
min; ∀n ∈ N &m ∈
M 0 ≤ Ekn ≤ E
k
max & 0 ≤ E
l
m ≤ E
l
max; ∀ n ∈ N &m ∈
M .
To provide stringent QoS requirements, the value of SINR
for M2M pairs and U∈’s should be higher than that of minimum
designed value. Hence energy transmitted by M2M devices and
U∈’s will not exceed the minimum value. Therefore, the energy
of M2M pair, Elm should lie between the lower limit of energy
defined as EULLower and the upper limit energy E
UL
Upper and can be
denoted as:
EULLower = max
{
0 ,
Γ
j
min
(
𝛽2 + Ekn ymn
)
GM2M
}
(12)
EULUpper = min
{
0 ,
Ekn GnUE − Γ
i
min 𝛽
2
Γimin ymn
}
(13)
When there are huge number of resources available, to deter-
mine the resource allocation between M2M devices and UE, the
optimal M2M and UE pairing solution can be given as:
Max
𝛼 UEmn , 𝛼
M2M
mn
{∑
n ∈N
∑
m∈m
(
𝛼 UEmn C
Uplink
UE ,M2M
)
+
(
𝛼 M2Mmn C
Uplink
UE ,M2M
)}
(14)
Such that, ∑
n∈N
𝛼UEmn ≤ 1&
∑
m∈m
𝛼M2Mmn ≤ 1
𝛼 UEmn 𝛼
M2M
mn ∈ { 0 , 1 } ∀n ∈ N &m ∈ M( ∑
n ∈N
𝛼 UEmn
)(∑
m∈m
𝛼 M2Mmn
)
= 0 ; ∀ n ∈ N & m ∈ M
4 RESOURCE ALLOCATION
ALGORITHM
4.1 Stage I: Modelling of the
environment—MTCDs with HTC UEs
In a cross-layer network, the parameters specified for resource
allocation energy efficiency should be defined in the net-
work modelling stage. The scheduling algorithm proposed
here addresses channel quality Indicators, Buffer status, aver-
age delay, and energy consumption. As this algorithm provides
cross-layer scheduling with m number of UE and n number of
M2M devices, the M2M devices transmit K×DTm times to the
BS, and we require a P number of modulation coding schemes
to transmit, which are identical to each other. Considering this
scenario, it can be treated as a multiobjective knapsack problem.
Therefore, the total number of resource blocks generated by the
UE and M2M devices are considered knapsack, and we define
all the data transmitted as items. The priority factor for each UE
can be denoted as:
PFn =
⎧⎪⎨⎪⎩
DHoL 𝜃n
(
𝛾GBR ∗ 𝛾n (t )
𝛾kn
)𝛼
+ Sn; WhenUE ∈ GBR
𝜂n (𝛾NGBR + Sn ) ;WhenUE ∈ Non − GBR
⎫⎪⎬⎪⎭
(15)
where DHoL denotes the head of line delay of n
th user, θn
denotes the factor to reduce the power consumption of the
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 7 of 16
device, 𝛾n (t ) is the rate of the n
th channel measured at t for
UE, 𝛾kn represents the average data rate of the nth UE at time t
for kth sub-channel, α is exponential factor and Sn denotes the
buffer status of each UE for ith service. The nth user equipment
with the best channel quality can be defined as:
UEn = arg maxn
⎧⎪⎪⎨⎪⎪⎩
Pn. exp
⎛⎜⎜⎜⎜⎝
qn. pd (t )
dn +
√(
1
R Tm
)∑
m pm (t )
⎞⎟⎟⎟⎟⎠
𝛿n
⎫⎪⎪⎬⎪⎪⎭
(16)
Pn, qn, and dn are the adjustable parameters which can be
modified in order to attain the best channel quality among vari-
ous UE and Pd(t) represents the packet delay for each UE which
can be represented as:
Pd (t ) =
∑
m
⎧⎪⎨⎪⎩
−
(
log
Δm
𝛽m
)
RTm
⎫⎪⎬⎪⎭ (17)
where Δm denotes the probability packet delay threshold level,
𝛿n denotes the spectral efficiency of a channel for user equip-
ment, RTm denotes the amount of real time flows, 𝛽m denotes
the delay target of UEn. Each data has some contribution
related to the energy efficiency and if the transmitted resource
block data is Xm, M2M is transmitted by pth MCS for given slot
t, where of knapsack capacity can be given by:(
1,
⌈ size o f (Xm,n,i, j )
Rm,n,i, j
⌉)
(18)
And its contribution to the energy efficiency can be given by,
size o f data(Xm,n,i, j )⌈
size o f data (Xm,n,i, j )
Rm,n,i, j
⌉
× fm,n,i, j
(19)
From the above relations, a variable Zm,n,i, j is defined and it’s
clear that dataXm,n,i, j is transmitted in t with modulation coding
scheme. When Zm,n,i, j= 1 otherwise Zm,n,i, j= 0. Therefore, the
energy optimization be transformed into a 2D knapsack prob-
lem and are defined as NP-hand problem. To arrive solution
for this NP-hand problem we first reformulate the above con-
dition as the constraint in which the resource blocks are owed
in one-time slot, such that it cannot exceed a function S.
N∑
n=1
M∑
m=1
DTi∑
i=1
DTj∑
j=1
(
Zm,n,i, j ∗
⌈ size o f data(Xm,n,i, j )
Rm,n,i, j
⌉)
≤ S
(20)
And the energy distributed for any single resource block
(both M2M and UE) must attain the MCS constraint such that,
⎧⎪⎨⎪⎩
Em,n,l , j + 10 log
⌈
size o f data
(
Xm,n,i, j
)
Rm,n,i, j
⌉
+
(𝛾 − 1) × pathlosscomponent
⎫⎪⎬⎪⎭ ≥ (SINR)i, j ,m,n
(21)
The scheduler changes the buffer state depending on the size
and information that can be transmitted. If there is any prob-
ability that the data from the UE can be associated with the
current transit time interval, then the data from UE will be allo-
cated, otherwise any data from M2M devices can be shared in
that spectrum. The status of the buffer should be updated on
every transmit time interval and it can be defined as:
BS (k) =
{
0; i f k = 0
BS (k − 1) − 𝜂m (k − 1) + dsm (k); i f k > 0
(22)
where dsm (k) denotes the data size of the m
th UE and kth TTI;
𝜂m (k − 1) represents the total throughput of the m
th user equip-
ment at the past TTI. Let Wm (k) denotes the weighted mean
delay of mth UE. wbdm denotes the buffered data and w
bd
m denotes
the transmitted data then:
Wm (k) =
⎧⎪⎨⎪⎩
wbdm BS (k) ≥ 𝜙(
1 −
wkm
𝜙
)
× wtdm +
wm (k)
𝜙
× wbdm BS (k) < 𝜙
(23)
where 𝜑 denotes the size of the window. In order to determine
the mean delay, wbdm (k) can be determined as:
wbdm (k) =
⎧⎪⎨⎪⎩
0 for k = 0[
wbdm (k − 1) + t
]
× BS (k − 1)
BS (k)
−
wm (k − 1) 𝜂m (k − 1)
BS (k)
for k > 0
(24)
wtdm (k) =
⎧⎪⎨⎪⎩
0 for k = 0[
1 −
𝜂m (k − 1)
𝜙 − BS (k − 1)
]
× wtdm (k − 1) +
𝜂m (k − 1)
𝜙 − BS (k − 1)
× wm (k − 1) for k > 0
(25)
where 𝜂m represents the throughput of the m
th UE device
and wm represents the weighted mean delay of the mth UE
device.
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 of 16 MOSES M ET AL.
FIGURE 2 Behaviour of dragonfly swarm.
4.2 Stage II: Levy flight Brownian-based
dragonfly optimization
The nature of a dragonfly is that they make small pairs among
themselves to pursue and kill other flying insects for their food
and separate among themselves to distract their enemies. These
two behaviours of dragonflies are identical to two states of opti-
mization procedure using meta-heuristics named exploration
and exploitation. To make this dragonfly operation artificial
and to create mathematical expressions for solving these opti-
mization problems, there are five weights: separation weight
SLD, alignment weight ALD, cohesion weight CLD, food fac-
tor weight FLD, enemy factor weight ELD, and inertia weight
WLD. These five stages of operation are expressed in Figure 2,
in which the objective would be the dragonfly gets attracted to
the food and distracted from their enemies.
To obtain the results in the search space, we define vari-
ous possible dragonfly behaviours. The first step denotes the
separation weight, and this behaviour represents the Collison
avoidance with the nearest dragonfly. So, if we consider X as the
data to be transmitted, then this behaviour infers the avoidance
of Collison from the nearest UE and M2M devices. This state is
marked as:
SLD = −
M+N∑
g=1
XLD − X
(m+n)
g (26)
where M+N represents the maximum devices available in the
entire single cell scenario and g represents the number of
neighbours around the devices. The second step refers to the
alignment in which all the neighbouring dragonflies will be
matched with the other flies. This refers to the alignment of UE
and M2M devices according to the traffic in the network. The
alignment can be represented as:
ALD =
M+N∑
g=1
−
(
Vmn
MN
)
(27)
where mn represents the product of number of UE and M2M
devices and Vmn denotes the velocity of transmission of UE
and M2M devices. The cohesion factor of the optimization
algorithm be defined as:
CLD =
∑M+N
g=1 X
m+n
g
MN
− XLD (28)
The cohesion factor CLD represents the dragonfly sticking
together depending upon different scenarios. In cross layer
optimization, CLD represents UE and M2M devices sticking
together in a giving bandwidth by sharing the same spectral
resource. The dragonfly also tends to attract themselves towards
their identified food factor and this can be represented as:
FLD = XFo − XLD (29)
where XFo represents the location of the food availability and
available of the dragonflies and this compares with the avail-
ability of the resource block at the user end for allocating the
availability resources according to their priority. XLD represents
the amount of resource blocks which are to be transmitted by
each UE and M2M devices. The dragonflies tend to repel away
from their enemies, and this can be represented as enemy factor
weight,
ELD = Xen + XLD (30)
The factors described by food Xfo and enemy Xen describe
the best and worst positions of the dragonflies. The position of
UE and M2M devices can be updated according to the results
obtained from the food factor and enemy factor,
XLD (t + 1) = XLD (t ) + ΔXLD (t + 1) (31)
where ΔXLD be movement of dragonfly or it is compared with
the movement of UE and M2M devices. It is very important
to note the moving direction of the flies or the direction of
movement of UE and M2M devices and it can be represented
as,
ΔXLD (t + 1) =
(
a.SLD + b.ALD + c.CLD
+ d .FLD + e.ELD
)
+ 𝜃 × ΔX (t + 1)
(32)
where a represents a factor related to the separation, b rep-
resents a factor related to the alignment, c represents a factor
related to the cohesion, d represents a factor related to the food,
and e represents a factor related to the enemies, θ represents a
factor related to the variation in data resources and t represents
the number of iterations. When there are no neighbour dragon-
flies, they search the pairs around the space by using Levy flight
and the position can be updated as,
X LDt+1 = X
LD
t + Levy( f ).X
LD
t (33)
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 9 of 16
where f denotes the frequency factor of the updated dragonfly
position and this can be represented as,
Levy( f ) = 0.28 ×
g1 × 𝛽||g2|| 1𝛼 (34)
where g1 and g2 represents the random value which varies as
(0,1) and α represents a constant value which can be 1.46 here
and β is represented as,
𝛽 =
⎛⎜⎜⎜⎝
Γ (1 + 𝛼) ⋅ sin
(
𝜋
2
)
𝛼
Γ
(
1+𝛼
2
)
⋅
√
𝛼
2
× 2(𝛼−1)
⎞⎟⎟⎟⎠
1
𝛼
(35)
where Γ(𝛼) = (𝛼 − 1)!. The DELFDO algorithm proposed in
this work to rank the UE and M2M devices not only based
upon their QCI values but also based upon their urgency fac-
tor and priority factor with lower energy devices available in
cell. To provide energy efficiency and allocate resource for UE
and M2M devices according to the DELFDO algorithm. A
Levy flight factor β is proposed to adjust the priority factors
among the users in base sectors. The value of β should be care-
fully selected to rank with the UE and M2M devices according
to the priority factor, low delay, and low power consumption.
This situation can be elaborated by defining utility function of
DELFDO algorithm,
Lm (k) = 𝛽 ×Wm (k) + (1 − 𝛽)Em (k) ; 0 ≤ 𝛽 ≤ 1 (36)
where Em(k) denotes the energy efficiency, and the value of F
varies from 0 to 1. When the value of β is set to 0 and 1, then the
algorithm does not provide the energy Efficiency andQOS con-
straints are not resolved. The Levy flight dragonfly algorithm is
proposed to optimize the value of β and at each step the location
of the UE and M2M devices are updated. The periodic motion
of the dragonflies is spread over time with a normal distribu-
tion, involving sudden jumps and random motions resembling
vibrations. The Brownian motion steps are determined using
a Gaussian distribution, and the equation is adjusted based on
the algorithm’s agent size and number, thereby finalizing the
Brownian motion. The updated position can be given as,
X LDt+1 = X
LD
t +
√
T
M
rand () ∗
1
h
√
2Π
e
(
VD−AG
2h2
)
(37)
The symbol T represents the time period for the motion of a
dragonfly, set at 0.01 in this work. In Equation (37), the termM
indicates the proportion of sudden motions for the same agent
over time.VD stands for dimensions, andAG represents agents.
To validate the β in our proposed DELFDO algorithm, we
define a fitness function F is given by,
F = max
M∑
m=1
N∑
n=1
(Lm (k) × 𝜂m ) + (Ln (k) × 𝜂n ) (38)
FIGURE 3 Flowchart representing the operation of Levy flight
Brownian-based dragonfly optimization.
where η depends on the throughput of the UE and M2M
devices and the condition can be described as,
𝜂 (k) ≤
RB∑
m=1
RB∑
n=1
Wm,n (k) ; i, j ∈ RB (39)
M∑
m=1
N∑
n=1
Bs (k) ≤ size o f bu f fer in resource block (40)
Buffer status reports depend upon the size of buffer. When
there is no delay in transmitting data then the size of buffer will
also be decreased, and this describes the optimal value of β by
evaluating the fitness function. Once the value of β is evalu-
ated then according to the estimated utility function and priority
value, ranks will be allocated to the UE and M2M devices.
Figure 3 represents the step-by-step optimization process
using Levy flight Brownian-based dragonfly optimization. In
multiobjective optimization, we start by placing dragonflies ran-
domly in the search space. We set the maximum archive size
and define the number of segments. Using their current posi-
tions, we generate solutions that do not dominate each other. If
the archive is full, we remove some solutions and add new ones.
The hypersphere is updated to cover all solutions if needed. For
dragonflies with neighbours, we calculate the velocity vector and
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 of 16 MOSES M ET AL.
update the position. If there are no neighbours, we use Brown-
ian motion to find a solution and adjust the position. We then
check if the dragonflies are within the search space. The best and
worst solutions in the archive are chosen as nutrient sources and
enemies, respectively.
4.3 Stage III: Resource allocation
This stage represents the allocation of resources to UE and
M2M devices based upon the proposed DELFDO algorithm
by considering their bandwidth requirements. The algorithm
begins by examining the availability of resource blocks from
the UE and M2M devices and scheduler. When the resource
blocks are found at the input, the algorithm explores whether
there are starving UE and M2M devices by considering the pri-
ority factor of the available resources. The second step in the
resource allocation algorithm examines the buffer status report.
ẟf the buffer status is empty and continuous to check whether
there are striving UE and M2M devices. When the buffer sta-
tus is not empty the algorithm checks the quantity of resources
available. When the resource availability is for only one UE,
then the radio resource blocks are all allocated to the speci-
fied UE and M2M devices. When there are a greater number of
resource blocks for multiple UE and M2M devices the sched-
uler proceeds with the next step. In the third step the scheduler
calculates resource blocks and the availability of UE and M2M
devices. For the number of UE and M2M devices less than that
number of resources, the resources are allocated to the users
based upon round robin fashion. When the number of UE and
M2M devices are greater than the number of resources, the
resources are allocated to the users based upon the calculated
utility function in phase 2. The remaining unallocated resources
to the UE and M2M devices are again fed to the first round of
the scheduler.
The complexity of the proposed DELFDO algorithm is ana-
lyzed based upon the resource allocation time per transit time
interval. The parameter to calculate the time complexity of
DELFDO algorithm includes the size of dragonfly (Ds) and
the maximum number of redundant bits (rmax) which does not
vary by the increase of resource blocks. The time Complexity of
DELFDO algorithm can be given as O (PRB log UE) +O (Ds
× rmax) for every transit time interval. From this analysis, it is
predicted that the time complexity of DELFDO algorithm has
minimum effects on the entire resource allocation process.
5 RESULTS AND DISCUSSION
The performance evaluation of the proposed optimal algorithm
in terms of throughput, average PDBV of MTCDs, Energy
efficiency, Resource block utilization, fairness index, packet
loss ratio and percentage of unserved resources in delay were
discussed in this section. Discrete event simulation without
inter-cell interference is conducted where M2M communica-
tion consists of one eNodeB and N MTCDs. The theoretical
and practical values are validated by just substituting the arrival
rate of MTCDs/UEs in the derived equations. The distribu-
TABLE 2 Simulation parameters.
Parameter Value
Cell radius 500m
Number of eNB 1
Simulation period 1000 TTIs
Distances between MTCDs and eNB 20–300 m
Transmitter power 23 dBm
Noise figure 18 dB
H2H arrival rate 64, 128, 256 kbps
M2M arrival rate 10, 20, 30, 40 kbps
Distribution of MTCDs/UEs Fixed and uniform
Delay bound 0.2 ms
Bandwidth 20 MHz
Number of PRBs 180
Number of sensor types on each MTCD 5–12
Maximum PDBV 10%
Propagation loss Urban macro cell model
Placement of HTC UES and MTC devices Random
tion of the arrival rate of users is uniform with the values as in
Table 2. Outcomes of the proposed scheme is compared with
HORA and CRBA algorithm [20]; whale optimization algo-
rithm (WOA)-based resource allocation scheme [25]; RSMA
technique with Han–Kobayashi scheme for RA in multi-
cell mm-wave environment [29]; amended barnacles mating
optimizer-based RA algorithm(ABMO) [24] and energy-aware
mode selection for D2D resource allocation using Hungarian
algorithm [19]. Unit cell that are uniformly distributed MTCDs
and H2H UEs are considered. Due to the presence of addi-
tive white Gaussian noise (AWGN) path loss and shadowing
effect takes place. SNR at receiver projects that value of CQI
is determined by schedulers. Voice-over-IP (VoIP) and video
applications are respectively sources of H2H traffic and M2M
traffic as shown in Table 2.
Figure 5 shows the comparison of energy efficiency for the
proposed DELFBDO technique by varying the HTC UEs from
10 to 70 with fixed MTC devices as 10 and by varying the HTC
UEs from 10 to 70 by varying the MTC devices from 10 to 70.
When the value of MTCDs increases, communication is scat-
tered, which consumes higher energy. Because of the increase
in MTCDs the available RBs are in scarceness to convey all the
sensory data and hence few may be lost. By proper schedul-
ing of data packets to MTCDs, the proposed scheme helps to
achieve higher energy efficiency. The average energy efficiency
for a fixed number of MTC device as 10 and number of HTC
UEs as 10 of the proposed the proposed DELFBDO technique
is 2.8% higher than the Hungarian algorithm, 6.7% higher than
the HORA and CBRA algorithm, 12.2% higher than the WOA,
17% higher than the ABMO, 24.8% higher than the RSMA tech-
nique. When the number of HTC UEs increase to 70 with a
fixed number of MTC devices as 10, the proposed DELFBDO
technique is 2.3% higher than the Hungarian algorithm, 4.9%
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 11 of 16
FIGURE 4 Resource allocation process in a cross-layer environment by using Levy flight Brownian-based dragonfly optimization algorithm for HTC-UEs and
MTC devices.
higher than the HORA and CBRA algorithm, 8.1% higher than
the WOA, 10.1% higher than the ABMO, 12.5% higher than
the RSMA technique. When the number of HTC UEs increase
to 70 with an increase in number of MTC devices as 70, the pro-
posed DELFBDO technique is 2.2% higher than the Hungarian
Algorithm, 5.9% higher than the HORA and CBRA algorithm,
10.5% higher than the WOA, 12.7% higher than the ABMO,
20.9% higher than the RSMA technique. The practical results
demonstrate that the proposed DELFBDO technique outper-
forms alternative scheduling algorithms by achieving the lowest
energy consumption.
Figure 6 shows the comparison of Average system through-
put for the Proposed DELFBDO technique by varying the
HTC UEs from 50 to 300 with fixed MTC devices as 50 and
by varying the HTC UEs from 50 to 300 by varying the MTC
devices from 50 to 300. Mean throughput variation refers to
the statistical measure of how the achieved data throughput
fluctuates or changes over time. It is commonly expressed
as the standard deviation or variance of the data throughput
values obtained during a certain period. The average system
throughput for a fixed number of MTC devices as 50 and num-
ber of HTC UEs as 50 of the proposed DELFBDO technique
is 30.4% higher than the Hungarian algorithm, 16.2% higher
than the HORA and CBRA algorithm, 8.8% higher than the
WOA, 20.7% higher than the ABMO, 2.6% higher than the
RSMA technique. When the number of HTC UEs increase to
300 with a fixed number of MTC devices as 50, the proposed
DELFBDO technique is 29.3% higher than the Hungarian
algorithm,41.7% higher than the HORA and CBRA algorithm,
27.5% higher than the WOA, 13.9% higher than the ABMO,
6.9% higher than the RSMA technique. When the number of
HTC UEs increase to 300 with an increase in the number of
MTC devices as 300, the proposed DELFBDO technique is
25% higher than the Hungarian algorithm, 16.2% higher than
the HORA and CBRA algorithm, 30.2% higher than the WOA,
5% higher than the ABMO, 13.4% higher than the RSMA
technique. With a small number of users, all scheduling algo-
rithms exhibit consistent throughput, utilizing sufficient RBs
to transmit queued data. However, as the user count increases,
the proposed DELFBDO technique stands out by achieving
maximum throughput, selecting the most efficient channel for
transmission.
Figure 7 shows the comparison of Fairness Index for the Pro-
posed DELFBDO technique by varying the HTC UEs from
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 of 16 MOSES M ET AL.
FIGURE 5 Comparison of energy efficiency for the proposed
DELFBDO technique. (a) When number of HTC UEs are varied by fixing the
MTC devices to 10. (b) When number of HTC UEs are varied by varying the
MTC devices.
10 to 100 with fixed MTC devices as 10 and by varying the
HTC UEs from 10 to 100 by varying the MTC devices from
10 to 100. Jain’s fairness index is used to calculate the fair-
ness index of the user. It quantifies how equitably the available
radio resources are distributed among users, ensuring that no
user experiences significant disadvantage in terms of data rate
or QoS. Index varies from 0 to 1 with 1 to be the highest. The
average Fairness Index for a fixed number of MTC devices as
10 and number of HTC UEs as 10 of the proposed the pro-
posed DELFBDO technique is 3.1% higher than the Hungarian
algorithm, 8.7% higher than the HORA and CBRA algorithm,
16.3% higher than the WOA, 23.5% higher than the ABMO,
31.6% higher than the RSMA technique. When the number of
HTC UEs increase to 100 with a fixed number of MTC devices
as 10, the proposed DELFBDO technique is 10% higher than
the Hungarian algorithm, 22.2% higher than the HORA and
CBRA algorithm, 37.5% higher than the WOA, 46.7% higher
than the ABMO, 69.2% higher than the RSMA technique. When
FIGURE 6 Comparison of average system throughput for the proposed
DELFBDO technique. (a) When number of HTC UEs are varied by fixing the
MTC devices. (b) when number of HTC UEs are varied by varying the MTC
devices.
the number of HTC UEs increase to 100 with an increase in
the number of MTC devices as 100, the proposed DELFBDO
technique is 14.3% higher than the Hungarian algorithm, 27%
higher than the HORA & CBRA algorithm, 33.3% higher than
the WOA, 45.5% higher than the ABMO, 77.8% higher than
the RSMA technique. The proposed scheduler achieves a bal-
ance among various QCI, enabling improved response times for
individual applications, with simulation results demonstrating
superior fairness compared to conventional schedulers.
Figure 8 shows the comparison of resource block utilization
for the proposed DELFBDO technique by varying the HTC
UEs and MTC devices from 100 to 700 with an increment of
50 users. The resource block utilization for when the number
of both HTC UEs and MTC devices is 100 for the proposed
the proposed DELFBDO technique is 4.7% higher than the
Hungarian algorithm, 8.7% higher than the HORA and CBRA
algorithm, 17% higher than the WOA, 23.5% higher than the
ABMO, 29.9% higher than the RSMA technique. When the
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 13 of 16
FIGURE 7 Comparison of fairness index for the proposed DELFBDO
technique. (a) When number of HTC UEs are varied by fixing the MTC
devices. (b) When number of HTC UEs are varied by varying the MTC devices.
FIGURE 8 Comparison of Resource block utilization for the proposed
DELFBDO technique when number of HTC UEs and MTC devices increases
from 100 to 700.
FIGURE 9 Comparison of average PDBV for the proposed DELFBDO
technique. (a) When number of HTC UEs are varied by fixing the MTC
devices. (b) When number of HTC UEs are fixed by varying the MTC devices.
number of HTC UEs and MTC devices increases to 700, the
proposed DELFBDO technique is 7% higher than the Hun-
garian algorithm, 12.2% higher than the HORA and CBRA
algorithm, 21.1% higher than the WOA, 29.6% higher than the
ABMO, 46% higher than the RSMA technique.
Figure 9 shows the comparison of average probability of
delay bound violation (PDBV) for the proposed DELFBDO
technique by varying the HTC UEs from 50 to 500 with fixed
MTC devices as 100 and by fixing the HTC UEs as a constant
of 100 users with the variation in MTC devices from 100 to
1000. This metric represents the average potential benefit in
terms of delay reduction or network performance improvement
that can be achieved by scheduling critical MTCDs at specific
times. In the context of LTE-A, critical MTCDs could be high-
priority data packets, real-time communication, or tasks with
strict latency requirements. The average PDBV for a fixed num-
ber of MTC devices as 100 and number of HTC UEs as 50
for the proposed the proposed DELFBDO technique is 25%
higher than the Hungarian algorithm, 46.7% higher than the
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 of 16 MOSES M ET AL.
FIGURE 10 Comparison of percentage of unserved resources in delay
for the proposed DELFBDO technique when number of HTC UEs and MTC
devices increases from 100 to 800.
HORA and CBRA algorithm, 53.8% higher than the WOA,
60.7% higher than the ABMO, 62.5% higher than the RSMA
technique. When the number of HTC UEs increase to 500 with
a fixed number of MTC devices as 100, the proposed DELF-
BDO technique is 4.7% higher than the Hungarian algorithm,
8.9% higher than the HORA and CBRA algorithm, 13.7%
higher than the WOA, 18% higher than the ABMO, 20.4%
higher than the RSMA technique. On the other hand, when the
number of HTCUEs is fixed to 100 with an increase in the num-
ber ofMTC devices as 100, the proposedDELFBDO technique
is 5% higher than the Hungarian algorithm, 9.5% higher than
the HORA and CBRA algorithm, 13.6% higher than the WOA,
18.3% higher than the ABMO, 22.4% higher than the RSMA
technique. When the number of HTC UEs is fixed to 100 with
an increase in the number of MTC devices as 1000, the pro-
posed DELFBDO technique is 8.9% higher than the Hungarian
algorithm, 16.3% higher than the HORA and CBRA algorithm,
24.8% higher than the WOA, 30.5% higher than the ABMO,
31.7% higher than the RSMA technique. Higher availability of
users allows for better optimization of resource allocation and
scheduling of critical MTCDs, leading to a higher potential for
delay reduction and improved network performance. On the
other hand, low availability of users may limit the potential
benefits of scheduling techniques, impacting overall network
efficiency and latency.
Figure 10 shows the comparison of percentage of unserved
Resources in delay for the proposed DELFBDO technique by
varying the HTC UEs and MTC devices from 100 to 800 with
an increment of 50 users. The percentage of unserved nodes
in delay refers to the proportion of user equipment (UEs) or
nodes that experience delays beyond a certain threshold or fail
to receive service within an acceptable time frame. Mathemat-
ically, the formula for calculating the percentage of unserved
nodes in delay is percentage of unserved nodes in delay= (num-
ber of unserved nodes/total number of nodes) × 100. The
percentage of unserved nodes in delay when the number when
FIGURE 11 Comparison of Percentage of packet loss for the Proposed
DELFBDO technique when number of HTC UEs and MTC devices increases
from 100 to 800.
the number of both HTC UEs and MTC devices is 300 for
the proposed DELFBDO technique is 33.3% higher than the
Hungarian algorithm, 55.6% higher than the HORA and CBRA
algorithm, 63.6% higher than the WOA, 71.4% higher than the
ABMO, 75% higher than the RSMA technique. When the num-
ber of HTC UEs combined with MTC devices increases to 800,
the proposed DELFBDO technique is 23.5% higher than the
Hungarian algorithm, 41.6% higher than the HORA and CBRA
algorithm, 46.9% higher than the WOA, 53.6% higher than
the ABMO, 61.5% higher than the RSMA technique. Proposed
algorithm provides the best performance in serving LTE-M
devices due to their urgency consideration in scheduling.
Figure 11 shows the comparison of percentage of packet Loss
for the proposed DELFBDO technique by varying the HTC
UEs and MTC devices from 100 to 800 with an increment
of 50 users. Monitoring and managing the packet loss rate is
essential for LTE-A network operators to ensure reliable data
transmission and a satisfactory QoS for users. It refers to the
statistical measure of the percentage of data packets that are lost
or discarded during transmission between the user equipment
(UE) and the eNodeB. The percentage of packet loss when
the number of both HTC UEs and MTC devices is 300 for
the proposed DELFBDO technique is 23.8% higher than the
Hungarian algorithm, 42.9% higher than the HORA and CBRA
algorithm, 52.9% higher than the WOA, 59% higher than the
ABMO, 61% higher than the RSMA technique. When the num-
ber of HTC UEs combined with MTC devices increases to 800,
the proposed DELFBDO technique is 26.9% higher than the
Hungarian algorithm, 53.7% higher than the HORA and CBRA
algorithm, 59.6% higher than the WOA, 68.9% higher than the
ABMO, 72.1% higher than the RSMA technique.
6 CONCLUSION
Joint delay and energy aware Levy flight Brownian movement-
based dragonfly optimization scheduling algorithm for uplink
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MOSES M ET AL. 15 of 16
LTE channel is presented. The proposed algorithm targets
energy consumption in eNB at trade off with the delay require-
ments of UEs. This is achieved in three processes. First, the
scheduling parameters are obtained and checked. In the next
phase, the algorithm optimizes α which is used in the calcula-
tion of utility function to rank UEs in an order that maximizes
the fitness function. Finally, the scheduler allocates the avail-
able RBs based on a ranked list. These make the DELFBDO
algorithm to dynamically improvise overall system performance.
The proposed algorithm is outperformed by other techniques
when considering individual performance parameters. But it is
essential to consider multiple evaluation metrics at once for
all proposed schemes. In a generic way, the simulated results
demonstrate proposed algorithm shows a balanced perfor-
mance among other scheduling algorithms by sustaining least
power consumption. By intelligently allocating resources, these
algorithms can enhance network performance, improve user
experience, and efficiently utilize available network resources.
AUTHOR CONTRIBUTIONS
The authors of this paper contributed to the research and
preparation of the manuscript in the following ways: Leeban
Moses: Conceptualization; methodology; software; data cura-
tion; writing—original draft. Perarasi Sambantham: Method-
ology; formal analysis; supervision. Muhammad Faheem:
Software; data curation; data validation; review and editing.
Shoukath Ali: Investigation; visualization; data validation;
project administration.Arfat AhmadKhan:Conceptualization;
data curation; data validation; review & editing.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
FUNDING INFORMATION
The authors would like to thank their affiliated institutes for
supporting this study.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available
upon request from the corresponding author. Due to pri-
vacy and ethical considerations, the datasets generated and/or
analyzed during the current study are not publicly available.
However, the authors are committed to facilitating access to the
data for replication, validation, and further research purposes.
ORCID
LeebanMoses https://orcid.org/0000-0001-7468-1660
MuhammadFaheem https://orcid.org/0000-0003-4628-4486
ShoukathAli K https://orcid.org/0000-0001-9256-373X
REFERENCES
1. González, E.E., Estrada, Y., Garcia-Roger, D., Monserrat, J.F.: Perfor-
mance evaluation and optimal management of mode 2 V2X communi-
cations in 5G networks. IEEE Access 11, 128810–128825 (2023). https://
doi.org/10.1109/access.2023.3333680
2. Mai Do, H., Tran, T.P., Yoo, M.: Deep reinforcement learning-based task
offloading and resource allocation for industrial IoT in MEC federation
system. IEEE Access 11, 83150–83170 (2023). https://doi.org/10.1109/
ACCESS.2023.3302518
3. Li, J., Guan, W., Tang, Z.: A Resource allocation scheme for packet
delay minimization in multi-tier cellular-based IoT networks. Mathematics
11(21), 4538 (2023). https://doi.org/10.3390/math11214538
4. Camargo, J.S., et al.: Dynamic slicing reconfiguration for virtualized 5G
networks using ML forecasting of computing capacity. Comput. Networks
236, 110001 (2023). https://doi.org/10.1016/j.comnet.2023.110001
5. Mughees, A., et al.: Energy-efficient joint resource allocation in 5G Het-
Net using multi-agent parameterized deep reinforcement learning. Phys.
Commun. 61, 102206 (2023). https://doi.org/10.1016/j.phycom.2023.
102206
6. Guo, Q., Tang, F., Kato, N.: Resource allocation for aerial assisted digital
twin edge mobile network. IEEE J. Sel. Areas Commun. 41(10), 3070–
3079 (2023). https://doi.org/10.1109/JSAC.2023.3310065
7. Elsayem, M., Abou-Zeid, H., Afana, A., Givigi, S.: Intelligent resource
allocation for grant-free access: a reinforcement learning approach. IEEE
Networking Lett. 5(3), 154–158 (2023). https://doi.org/10.1109/lnet.
2023.3299182
8. Hegde, A., Song, R., Festag, A.: Radio resource allocation in 5G-NR V2X:
a multi-agent actor-critic based approach. IEEE Access 11, 87225–87244
(2023). https://doi.org/10.1109/ACCESS.2023.3305267
9. Bruun, R.L., Santiago Garcia, C.M, Sorensen, T.B., Pratas, N.K., Madsen,
T.K., Mogensen, P.: Signaling design for cooperative resource allocation
and its impact to message reliability. IEEE Access 11, 103569–103584
(2023). https://doi.org/10.1109/ACCESS.2023.3317269
10. Eswaramoorthi, R., Leeban Moses, M., Sahul Hameed, J.B., Ghanti, B.:
Deep graph neural network optimized with fertile field algorithm based
detection model for uplink multiuser massive multiple-input and multiple-
output system. Trans. Emerg. Telecommun. Technol. 33(12), e4614 (2022).
https://doi.org/10.1002/ett.4614
11. Shirkolaei, M.M., Jafari, M.: A new class of wideband microstrip fal-
cate patch antennas with reconfigurable capability at circular-polarization.
Microwave Opt. Technol. Lett. 62(12), 3922–3927 (2020). https://doi.
org/10.1002/mop.32529
12. Shirkolaei, M.M., Ghalibafan, J.: Scannable leaky-wave antenna based on
ferrite-blade waveguide operated below the cutoff frequency. IEEE Trans.
Magn. 57(4), 1–10 (2021). https://doi.org/10.1109/TMAG.2021.3060683
13. Maleki, A., Oskouei, H.D., Mohammadi Shirkolaei, M.: Miniaturized
microstrip patch antenna with high inter-port isolation for full duplex
communication system. Int. J. RF Microwave Comput. Aided Eng. 31(9),
e22760 (2021). https://doi.org/10.1002/mmce.22760
14. Aziz, W.A., Ioannou, I.I., Lestas, M., Qureshi, H.K., Iqbal, A., Vassiliou, V.:
Content-aware network traffic prediction framework for quality of service-
aware dynamic network resource management. IEEE Access 11, 99716–
99733 (2023). https://doi.org/10.1109/ACCESS.2023.3309002
15. Paymard, P., Amiri, A., Kolding, T.E., Pedersen, K.I.: Optimizing mixed
capacity of extended reality and mobile broadband services in 5G-
advanced networks. IEEE Access 11, 113324–113338 (2023). https://doi.
org/10.1109/ACCESS.2023.3323815
16. Sultan, M.T., El Sayed, H.: QoE-aware analysis and management of
multimedia services in 5G and beyond heterogeneous networks. IEEE
Access 11, 77679–77688 (2023). https://doi.org/10.1109/ACCESS.2023.
3298556
17. Moses, M.L., Kaarthick, B.: Multiobjective cooperative swarm intelli-
gence algorithm for uplink resource allocation in LTE-A networks.
Trans. Emerging Telecommun. Technol. 30(12) (2019). https://doi.org/
10.1002/ett.3748
18. Abdelsadek, M.Y., Gadallah, Y., Ahmed, M.H.: Matching-based resource
allocation for critical MTC in massive MIMO LTE networks. IEEE
Access 7, 127141–127153 (2019). https://doi.org/10.1109/ACCESS.
2019.2939120
19. Tsai, H.C., Kao, S.J., Huang, Y.L., Chang, F.M.: Energy-aware mode selec-
tion for D2D resource allocation in 5G networks. Electronics 12(19), 4054
(2023). https://doi.org/10.3390/electronics12194054
20. Kukade, S., Sutaone, M., Patil, R.: Optimal performance of resource allo-
cation in LTE-A for heterogeneous cellular network. Wireless Networks
27(5), 3329–3344 (2021). https://doi.org/10.1007/s11276-021-02635-w
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 of 16 MOSES M ET AL.
21. Abdelsadek, M.Y., Ahmed, M.H., Gadallah, Y.: Cross-layer resource allo-
cation for critical MTC coexistent with human-type communications in
LTE: A two-sided matching approach. IET Commun. 14(18), 3223–3230
(2020). https://doi.org/10.1049/iet-com.2019.0944
22. Gatti, R., Shankar, S.: Bidirectional resource scheduling algorithm for
advanced long term evolution system. Eng. Rep. 2(7), e12192 (2020).
https://doi.org/10.1002/eng2.12192
23. Liu, J.S., Lin, C.H.R., Hu, Y.C.: Joint resource allocation, user association,
and power control for 5G LTE-based heterogeneous networks. IEEE
Access 8, 122654–122672 (2020). https://doi.org/10.1109/ACCESS.
2020.3007193
24. Mohammadnejad, Z., Al-Khafaji, H.M.R., Mohammed, A.S., Alatba, S.R.:
Energy optimization for optimal location in 5G networks using improved
barnacles mating optimizer. Phys. Commun. 59, 102068 (2023). https://
doi.org/10.1016/j.phycom.2023.102068
25. Sing, R., Bhoi, S.K., Panigrahi, N., Sahoo, K.S., Jhanjhi, N., AlZain, M.A.: A
whale optimization algorithm based resource allocation scheme for cloud-
fog based IoT applications. Electronics 11(19), 3207 (2022). https://doi.
org/10.3390/electronics11193207
26. da Mata, S.H., Guardieiro, P.R.: Resource allocation for the LTE uplink
based on Genetic Algorithms in mixed traffic environments. Comput.
Commun. 107, 125–137 (2017). https://doi.org/10.1016/j.comcom.2017.
04.004
27. Sui, W., Chen, X., Zhang, S., Jiang, Z., Xu, S.: Energy-efficient resource
allocation with flexible frame structure for hybrid eMBB and URLLC
services. IEEE Trans. Green Commun. Networking 5(1), 72–83 (2021).
https://doi.org/10.1109/TGCN.2020.3028202
28. Chen, Y.-L., Lin, Y.-H., Chen, H.-Y., Hung, K.-C.: Energy efficient
resource allocation algorithms combining PSO with FLC and Taguchi
method in hybrid opportunistic networks. Appl. Soft Comput. J. 148(129),
110717 (2023). https://doi.org/10.24433/CO.0
29. Sarma, S.S., Hazra, R., Goswami, P.: Power optimization in a multicell D2D
communication for smart city in an mm-wave cellular network: an mIoT
perspective. IEEE Internet Things J. 10(21), 18686–18694 (2023). https://
doi.org/10.1109/JIOT.2023.3262577
30. Singh, U., et al.: Scalable priority-based resource allocation scheme
for M2M communication in LTE/LTE-A network. Comput. Electr.
Eng. 103, 108321 (2022). https://doi.org/10.1016/j.compeleceng.2022.
108321
31. Cao, Y., Zhong, Y., Peng, X., Pan, S.: Energy efficiency maximization for
hybrid-powered 5G networks with energy cooperation. Electronics 11(10),
1605 (2022). https://doi.org/10.3390/electronics11101605
How to cite this article: Moses, L., Sambantham, P.,
Faheem, M., Ali K, S., Khan, A.A.: Joint delay and
energy aware dragonfly optimization-based uplink
resource allocation scheme for LTE-A networks in a
cross-layer environment. J. Eng. 2024, e12353 (2024).
https://doi.org/10.1049/tje2.12353
 20513305, 2024, 2, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/tje2.12353 by University Of Vaasa, Wiley Online Library on [24/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License