sensors

Article
Load Balancing Using Artificial Intelligence for Cloud-Enabled
Internet of Everything in Healthcare Domain
Ibrahim Aqeel 1 , Ibrahim Mohsen Khormi 1 , Surbhi Bhatia Khan 2,3, *, Mohammed Shuaib 1 ,
Ahlam Almusharraf 4 , Shadab Alam 1, * and Nora A. Alkhaldi 5

1 College of Computer Science & IT, Jazan University, Jazan 45142, Saudi Arabia; [email protected] (I.A.);
[email protected] (I.M.K.); [email protected] (M.S.)
2 Department of Electrical and Computer Engineering, Lebanese American University, Byblos 13-5053, Lebanon
3 Department of Data Science, School of Science, Engineering and Environment, University of Salford,
Manchester M5 4WT, UK
4 Department of Business Administration, College of Business and Administration, Princess Nourah bint
Abdulrahman University, P.O. Box 84428, Riyadh 11671, Saudi Arabia; [email protected]
5 Department of Computer Science, College of Computer Science and Information Technology,
King Faisal University, Al Hasa 31982, Saudi Arabia; [email protected]
* Correspondence: [email protected] (S.B.K.); [email protected] (S.A.)

Abstract: The emergence of the Internet of Things (IoT) and its subsequent evolution into the
Internet of Everything (IoE) is a result of the rapid growth of information and communication
technologies (ICT). However, implementing these technologies comes with certain obstacles, such as
the limited availability of energy resources and processing power. Consequently, there is a need for
energy-efficient and intelligent load-balancing models, particularly in healthcare, where real-time
applications generate large volumes of data. This paper proposes a novel, energy-aware artificial
intelligence (AI)-based load balancing model that employs the Chaotic Horse Ride Optimization
Algorithm (CHROA) and big data analytics (BDA) for cloud-enabled IoT environments. The CHROA
technique enhances the optimization capacity of the Horse Ride Optimization Algorithm (HROA)
using chaotic principles. The proposed CHROA model balances the load, optimizes available energy
resources using AI techniques, and is evaluated using various metrics. Experimental results show
Citation: Aqeel, I.; Khormi, I.M.; Khan, that the CHROA model outperforms existing models. For instance, while the Artificial Bee Colony
S.B.; Shuaib, M.; Almusharraf, A.; (ABC), Gravitational Search Algorithm (GSA), and Whale Defense Algorithm with Firefly Algorithm
Alam, S.; Alkhaldi, N.A. Load
(WD-FA) techniques attain average throughputs of 58.247 Kbps, 59.957 Kbps, and 60.819 Kbps,
Balancing Using Artificial Intelligence
respectively, the CHROA model achieves an average throughput of 70.122 Kbps. The proposed
for Cloud-Enabled Internet of
CHROA-based model presents an innovative approach to intelligent load balancing and energy
Everything in Healthcare Domain.
optimization in cloud-enabled IoT environments. The results highlight its potential to address critical
Sensors 2023, 23, 5349. https://
doi.org/10.3390/s23115349
challenges and contribute to developing efficient and sustainable IoT/IoE solutions.

Academic Editor: Romain Laborde Keywords: Internet of Things; Internet of Everything; big data analytics; cloud computing; clustering;
Received: 13 April 2023 load balancing; healthcare
Revised: 21 May 2023
Accepted: 24 May 2023
Published: 5 June 2023
1. Introduction
The concept of IoT is evolving, leading to the emergence of the broader subset known
as IoE, which aims to connect devices, data, and industries. While IoT refers to the net-
Copyright: © 2023 by the authors.
work of interconnected physical devices embedded with sensors and software, enabling
Licensee MDPI, Basel, Switzerland.
them to collect and exchange data, IoE is a broader concept that extends beyond IoT by
This article is an open access article
distributed under the terms and
integrating people, processes, and data into a comprehensive, interconnected ecosystem [1].
conditions of the Creative Commons
IoE goes beyond connecting gadgets and includes people, data, and various sectors such
Attribution (CC BY) license (https://
as emergency response, urban planning, healthcare, and the military [2]. Data integration
creativecommons.org/licenses/by/ and contextualization are crucial in IoE due to the many components and data sources
4.0/). involved. A typical ecosystem is needed to enable communication and integration between

Sensors 2023, 23, 5349. https://doi.org/10.3390/s23115349 https://www.mdpi.com/journal/sensors

 Sensors 2023, 23, x FOR PEER REVIEW 2 of 26
Sensors 2023, 23, 5349 2 of 25

sources involved. A typical ecosystem is needed to enable communication and integration

between information,
information, systems, systems,
and sensor andinputs.
sensor inputs. IoE nodes,
IoE nodes, such assuch as tablets,
tablets, smartphones,
smartphones, and
and home
home appliances,
appliances, detectdetect the environment,
the environment, exchange
exchange data,data,
and and possess
possess different
different capa-
capabili-
bilities
ties [3]. [3].
TheThe intelligence
intelligence of IoE
of IoE nodes
nodes lies
lies in in their
their abilitytotoshare
ability shareand
andinteract
interactwith
withdata,
data,
learning from
learning from itit to perform valuable
valuableactivities
activities[4].
[4].The
Theavailability
availabilityof of
IoTIoT
data andand
data thethe
de-
velopment ofofAIAIcontribute
development contributetotothe
theintelligence
intelligence ofof IoE.
IoE. IoE finds applications
applications inin intelligent
intelligent
transportation,
transportation,smartsmartagriculture
agriculturesystems,
systems, andandsmart
smartmanufacturing.
manufacturing. The The growth
growth of ofICT
ICT
technology,
technology,from from human-oriented
human-oriented Internet to machine-oriented
Internet to machine-oriented IoT and
IoTadvancements
and advancements such
as WSN,
such LPWAN,
as WSN, and 5G,and
LPWAN, enables the realization
5G, enables of IoE and
the realization facilitates
of IoE machine-to-machine
and facilitates machine-to-
communication [5]. A general
machine communication [5]. Aexpectation of IoE is presented
general expectation in Figure 1.
of IoE is presented in Figure 1.

Figure1.1.General
Figure Generalexpectation
expectationof
ofIoE.
IoE.

Therapid
The rapidgrowth
growthof ofthe
theInternet
Internetof ofThings
Things(IoT) (IoT)andand advancements
advancements in in artificial
artificial intel-
intel-
ligencehave
ligence haveenabled
enabledintelligent
intelligent applications
applications in invarious
various fields
fields such
such as ashealthcare,
healthcare,smartsmart
cities,and
cities, andIndustry
Industry4.0. 4.0. These
Theseapplications
applicationsimproveimprovedecision-making,
decision-making, recognition,
recognition, and and
management processes [6]. The
management TheIoT IoThashasevolved
evolved into thethe
into Internet of Everything
Internet of Everything (IoE), con-
(IoE),
connecting
necting andand enabling
enabling intelligent
intelligent usesusessuchsuch as smart
as smart metering,
metering, traffic
traffic scheduling,
scheduling, and
and agri-
agriculture. Wireless
culture. Wireless transmission
transmission networks
networks suchsuch
as 5G, asWSNs,
5G, WSNs, and LPWAN
and LPWAN contribute
contribute to the
to the scalability
scalability of the of
IoE. theAtIoE.
the Atsame thetime,
samebig time,
databig data analysis
analysis is crucial is for
crucial for extracting
extracting valuable
valuable
insights insights
from data from fordata for marketing,
marketing, analytics, analytics, and machine
and machine learninglearning purposes
purposes [7,8]. [7,8].
Data
Data extraction
extraction is essential
is essential in ingesting
in ingesting datadata for analysis
for analysis andand decision-making
decision-making [2].[2]. Artificial
Artificial in-
intelligence further enhances the intelligence
telligence further enhances the intelligence of the IoE. of the IoE.
Despite
Despitethe theadvancements,
advancements,there thereareareinherent
inherentconstraints
constraintsin inthe
theIoE,
IoE,including
includingbattery
battery
limitations of the IoT nodes, network coverage limitations, and security
limitations of the IoT nodes, network coverage limitations, and security vulnerabilities vulnerabilities [9,10].
These constraints affect the availability and connectivity of IoT devices, especially in
[9,10].
coverage-limited
These constraintsareas.affect
Clustering techniques
the availability andgroup similar of
connectivity objects, enabling
IoT devices, valuable
especially in
insights and efficient system performance. However, scalability
coverage-limited areas. Clustering techniques group similar objects, enabling valuable in-issues can arise when IoT
devices struggle
sights and to adapt
efficient system to platform
performance. changes and failscalability
However, to meet client
issues expectations
can arise when [11]. IoT
A novel energy-aware cluster-based load balancing approach
devices struggle to adapt to platform changes and fail to meet client expectations [11]. called CHROA is pro-
posedAtonovel
address these challenges for a cloud-enabled IoT environment.
energy-aware cluster-based load balancing approach called CHROA is pro- The CHROA
model
posed utilizes
to address clustering to effectively
these challenges for balance the load and
a cloud-enabled IoT optimize
environment. energy Theresources.
CHROA
Itmodel
incorporates the concept of chaos into the HROA technique
utilizes clustering to effectively balance the load and optimize energy resources. to enhance global opti-It
mization capabilities. The CHROA technique is implemented
incorporates the concept of chaos into the HROA technique to enhance global optimiza- using Apache Flume, a Big
Data platform, and
tion capabilities. The theCHROA
Hadooptechnique
Distributed File System (HDFS)
is implemented for storage.
using Apache Flume, The method
a Big Data
considers input variables such as energy, distance, and delay factor to build an objective
platform, and the Hadoop Distributed File System (HDFS) for storage. The method con-
function. The performance of the CHROA method is validated using a MATLAB program
and evaluated in various settings.
Sensors 2023, 23, 5349 3 of 25

1.1. Importance of Load Balancing in Internet of Everything

Load balancing is of utmost importance in the Internet of Everything (IoE) for various
reasons. Firstly, it significantly enhances system performance by evenly distributing net-
work traffic among multiple devices. This prevents congestion and ensures that essential
applications and services in healthcare, for instance, are readily accessible to users when-
ever they need them. Secondly, load balancing enhances system reliability by eliminating
the risks associated with single points of failure. In the event of a device failure, traffic can
be seamlessly redirected to alternative devices, thereby ensuring uninterrupted service
delivery. Thirdly, load balancing plays a critical role in bolstering security measures [12].
By dispersing traffic across numerous devices, it becomes considerably more challenging
for potential attackers to target a single device, effectively reducing the risk of cyberat-
tacks and safeguarding sensitive data [13]. Moreover, load balancing optimizes resource
utilization by efficiently managing traffic distribution, preventing individual devices from
being overwhelmed. This leads to cost reductions and improved scalability, making it
easier to accommodate increasing demands. Overall, load balancing is an indispensable
component of the IoE as it improves performance, increases reliability, enhances security,
and optimizes resource utilization, ultimately enabling the delivery of top-quality services
across a diverse range of industries.

1.2. IoT, IoE, and Role of Load Balancing in Healthcare

The Internet of Things (IoT) and the Internet of Everything (IoE) are distinct concepts,
with the IoE encompassing a broader scope that includes people, processes, and data. In
healthcare, both IoT and IoE have transformative potential [14]. IoT devices such as wear-
ables and sensors collect real-time patient health data, enabling monitoring and improved
treatment outcomes [15]. IoE integrates data analytics, AI, and other technologies to view
patient health comprehensively. By combining various data sources, including electronic
health records and environmental factors, IoE enhances healthcare decision-making. Load
balancing plays a crucial role in healthcare by evenly distributing network traffic across
devices, improving system performance, and ensuring timely access to healthcare appli-
cations [16]. It also increases system reliability by eliminating single points of failure and
enhances security by dispersing traffic to reduce the risk of cyberattacks. Moreover, load
balancing optimizes resource utilization, avoiding device overloads, reducing costs, and
improving scalability. Overall, IoT and IoE have the potential to revolutionize healthcare,
improving patient outcomes, increasing efficiency, and reducing costs.
This research seeks to develop intelligent and energy-efficient load-balancing models
for cloud-enabled IoT in healthcare. By leveraging AI and big data analytics, the proposed
model seeks to optimize resource utilization, enhance performance, and improve the deliv-
ery of healthcare services, thereby contributing to efficient and sustainable IoE solutions
and ultimately improving patient outcomes.
The distinctive contributions of this paper are:
1. Building a low-power data retrieval system for the IoE.
2. The proposed optimization algorithm can be adapted for use in the IoE to facilitate
the formation of clusters of related sensors.
3. Application of CHROA to elect optimal cluster heads (CHs).
4. To optimize the energy efficiency and cost-effectiveness of the Fog-IoE framework
and create a method for handling end-user requests.
5. Improve resource utilization and reduce the need to move tasks to the cloud by
proposing a load-balancing strategy for fog computing.
The following layout depicts the organizational structure for the rest of this project.
Section 2 provides a more comprehensive review of the IoE-related works. The proposed
framework will be made public in the third step, and the simulation process will be
evaluated in the fourth. The conclusion is found in the fourth and final portions.
Sensors 2023, 23, 5349 4 of 25

2. Literature Review
A brief review of related works has been completed related to the IoE. Cao et al. [17]
developed the Edge OSH concept, a home operating system for the IoE. They also covered
functional issues such as self-management, encoding interface, security and privacy, nam-
ing, and data management, as well as non-functional issues such as lack of open testbed
availability, latency, system cost, and user experience. Naranjo et al. [18] presented a Fog-
based smart city network named FOCAN, a multi-tier architecture where the application
runs on things that mutually route, compute, and interact via a smart city platform. The
FOCAN reduces delay and enhances the energy provision and the efficacy of services be-
tween things using distinct abilities. Mainly, three kinds of communications are determined
with FOCAN devices, primary, secondary, inter-primary and transmission, for managing
applications that meet the standard quality of service (QoS) for the IoE.
Singh et al. [19] drew inspiration from Blockchain and FC approaches to create the
BFAN-safe architecture. The provided structure uses Blockchain, encryption, and authen-
tication to safeguard sensitive data. It facilitates deploying applications in smart city
paradigms for architects and system developers. The presented framework aims to re-
duce the delay and guarantee an enhanced security feature via a blockchain technique.
Divyabharathi et al. [20] proposed the first-ever Blockchain that could instantaneously
manage the data security and device that is significant to emerge in the IoE. It intro-
duces an exclusive idea of Blockchain, which incorporates hardware security primitives
named PUF for solving latency, scalability, and energy need problems, called a PUF chain.
Miao et al. [21] offered a fair and dynamic data-sharing framework (FairDynDSF) in a
multi-owner setup. The accuracy of the search results must be validated, perform multi-
keyword searches, dynamic upgrades, and achieve fair arbitration. In FCN, Xiao et al. [22]
applied the time sharing (TS) strategy in place of the first come, first served (FCFS).
FCN is next supplied with the collaborative LB methodology for the TS method. The
approach differs from a work theft scheduling method in that it is based on the Nash
bargaining solution (NBS) for distributed games among FCNs. FCNs work together to
improve the efficiency of all FCNs to achieve Pareto optimality. Chithambaramani et al. [23]
examined the computational offloading problems of the existence and synergy between
FC and CC in the IoE by mutually improving the offloading decision, the distribution of
computational assets, and transfer power. Primarily, they proposed an ECORA system to
minimize the scheme cost. Garzia and Papi [24] aimed to tell how an IoE-based unified
security system for archaeology areas may assure cultural heritage preservation or con-
servation, outstanding usability for tourists, and visitor security, with specific references
to visitor utilizing incapacities. The GA was utilized for designing a combined security
system, especially for the field of components such as CCTV cameras, Wi-Fi Access Points,
and installation poles for ensuring final cost reduction and a higher level of consistency and
flexibility of the scheme for themselves for deliberation: the usual vincula and restriction
of archaeological regions. Babou et al. [25] proposed a solution for resolving the problem
of delays caused by resource constraints on HEC servers. Increased traffic causes server
waits, which increases the processing time (delay) for the request. They suggested a unique
HEC Clustering Balance approach using leverage, clustering, and LB methods. It enables
hierarchical request allocation on the HEC cluster and other framework components to
minimize congestion on the HEC server and reduce latency. Table 1 elaborates on the state
of artworks related to existing works. Based on their findings, Singh et al. [26] compared
and contrasted various load-balancing strategies, classification schemes, and algorithms.
According to the authors’ research, round robin is the easiest load-balancing technique in a
fog computing system. IP Hash was employed as the load-balancing method in the fog
computing infrastructure. Another technique for finding malicious nodes was presented
by Sangaiah et al. [27]: “Clustering Multi-Layer Security Protocol (CL-MLSP)” combined
with “Ad-hoc On-Demand Distance Vector” (AODV). In terms of mobility, dispersion, and
energy, the clustering method calculated the shortest distance between each node. The CL
Sensors 2023, 23, 5349 5 of 25

effectiveness MLSPs were measured across a range of metrics, including network lifetime,
packet loss, latency, and security, using NS2.
The system attempted to prevent a scenario where a node and its neighbors used
the same routing path to their respective destinations. The load balancing impact was
boosted because the next hop was selected using satellite ephemeris prediction and traffic
allocation. A load-balancing routing technique built on extended link states was proposed
by C. Dong et al. [28]. With a system of active state discovery and automated congestion
state release, they ensure that all satellite nodes are aware of the current condition of the
link. All satellite nodes must adjust their route tables in real-time to ensure that traffic is
evenly distributed over available connections.

Table 1. Features and survey of existing methodologies.

Reference Method/Techniques Features Limitations

It also addressed non-functional issues, such as a
lack of open testbed availability, latency, system cost, The study did not evaluate
Cao et al. [17] EdgeOSH and user experience, and functional ones, such as the performance of the
self-management, programming interface, security proposed method.
and privacy, nomenclature, and data management.
Using various abilities, the FOCAN lowers delay
and improves energy provision and service efficacy
It proposes a fog-based
between objects. With FOCAN devices, There are
smart city network
Naranjo et al. [18] FOCAN three categories of communications: primary,
architecture without
secondary, inter-primary, and transmission for
an algorithm.
controlling applications that fulfil the standard
QoS for IoE.
Blockchain and It does not provide any
The framework aims to reduce time while providing
Singh et al. [19] Fog-based Architecture specific implementation
increased security via Blockchain.
Network (BFAN) details or algorithm.
It offers the PUF chain a unique blockchain concept It very briefly discusses the
that integrates hardware security primitives known framework only without
Divyabharathi et al. [20] Map-Reduce
as PUFs for overcoming latency, scalability, the implementation aspect
and energy needs. and comparison.
It mainly deals with
Fair and dynamic data This method could validate search results, run
security aspects and
Miao et al. [21] sharing framework multi-keyword searches, make dynamic upgrades,
does not detail
(FairDynDSF) and accomplish fair arbitration.
resource optimization.
It only focuses on a
The method varies from work theft scheduling work-stealing algorithm
Nash bargaining because it is based on the NBS for cooperative games (GWS) and a classical
Xiao et al. [22]
solution (NBS) among FCNs. FCNs collaborate to increase the work-stealing algorithm
efficiency of all FCNs to reach Pareto optimality. (CWS) for resource
optimization.
Hybrid squirrel search The computational offloading issues of the existence It mainly deals with cloud
genetic algorithm (HSSGA) and synergy between FC and CC in IoE were computing environments
Chithambaramani et al. [23] and improved adaptive investigated by jointly enhancing the offloading and does not compare
neuro-fuzzy inference decision, the distribution of computational assets, results with
system (I-ANFIS) and transfer power. other algorithms.
The increase in traffic causes a large wait on these
servers, which increases the processing time (delay)
for the request. They proposed a novel strategy
dubbed HEC Clustering Balance, which combines It does not compare results
Babou et al. [25] Genetic Algorithms (GAs)
leverage, clustering, and LB approaches. It allocates with available literature.
hierarchical requests on the HEC cluster and other
framework components to reduce HEC server
congestion and latency.
It is a systematic review, but
This article systematically reviews resource
Jeyaraj et al. [29] Systematic review no detailed information is
management tasks in the cloud for IoT applications.
given for any algorithm.
Sensors 2023, 23, 5349 6 of 25

Table 1. Cont.

Reference Method/Techniques Features Limitations

This article provides a Micro-genetic It deals with cloud data
Mehran, Izadi, and algorithm-based mechanism to decide the locations centres and does not deal
Micro-genetic algorithm
Ghobaei-Arani [30] of various virtual machines (VMs) of the cloud data with IoT or IoE
center to minimize power consumption. nodes directly.
It proposes a particle swarm optimization-based It is well written but does
Particle swarm
Tanzila, et al. [31] load-balancing mechanism in the IoE cloud platform not consider ML/DL for
optimization
to reduce latency and transmission overhead. resource optimization.
It proposes a WoA-based QoS monitoring platform It deals with only
Ghobaei-Arani and Whale optimization
for IoT services and provides a mechanism for metaheuristic-based
Shahidineja [32] algorithm (WOA)
service placement in the fog-cloud environment. mechanisms
This article proposes a context-aware offloading
approach using Bayesian learning automata in
Farahbakhs, Shahidinejad Bayesian learning mobile edge computing for Internet of Things It does not compare results
and Ghobaei-Arani [33] automata (BLA) applications, aiming to optimize performance with available literature.
metrics and improve offloading efficiency in
multiuser scenarios.
This article discusses the limitations of cloud-based
It is a review article on
healthcare applications. It proposes a
Fog-IoHT; no detailed
Quy, Hau, Anh, fog-computing-based architectural framework for
Review article on Fog-IoHT information is given for
and Ngoc [34] Internet of Health Things (IoHT) applications,
any algorithm and
highlighting its potential and addressing challenges
resource optimization.
in integrating fog computing into healthcare IoT.
This study proposes an efficient edge computing
management mechanism using an information map
It is the latest article but
to reduce service response times and improve energy
Khanh et al. [35] Information map mainly deals with smart
consumption in IoT applications for smart cities,
city concepts.
aiming for its future application in
sustainable smart cities.

3. Background Study
The IoE is spurring the development of various systems and applications that are ca-
pable of collecting and analyzing vast amounts of data. Sensors and devices are typically
used to gather this data, which consume considerable amounts of power, necessitating the
development of low-power data retrieval systems [36]. Several studies have proposed var-
ious architectures and techniques to build such systems. One study by [37] proposed a
low-power data retrieval system for IoT that combines wireless and power line communi-
cation (PLC) to acquire and transmit remote data with minimal power consumption. The
authors demonstrated the effectiveness of their system in temperature and humidity mon-
itoring [38] presented a wireless data acquisition system for IoT that utilized a Bluetooth
Low Energy (BLE) communication protocol and evaluated its performance in terms of power
consumption, transmission range, and data rate in a smart home application. Paper [39]
designed a low-power wireless sensor network (WSN) for smart agriculture based on the
ZigBee protocol, powered by solar panels, and tested it in real-world applications of soil
moisture monitoring. Article [40] provided a review of IoT-facilitating technologies, including
various wireless communication tools and their suitability for IoT applications. Article [41]
proposed a low-power WSN for use in agriculture based on the ZigBee protocol for real-time
soil moisture, temperature, and humidity monitoring. Finally, [42] proposed a low-power
WSN for smart greenhouse monitoring using the ZigBee protocol, designed for real-time
monitoring of various environmental parameters, including temperature, humidity, light, and
carbon dioxide, and tested it in a real-world application.
A tabular comparison of these six studies is provided in Table 2 in terms of their
communication protocols, power supply, real-world applications, and reported results,
highlighting the varied approaches and outcomes of these studies.
Sensors 2023, 23, 5349 7 of 25

Table 2. Review of literature on data retrieval methods for IoT.

Communication Power Data Retrieval

Study IoT Device (Node) Key Features
Protocol Consumption Method
Low-cost,
[43] ESP8266 MQTT 0.37 W HTTP GET open-source
platform
Easy to program,
[38] Raspberry Pi MQTT 0.46 W HTTP GET powerful processing
capabilities
Secure
[39] CC3200 CoAP 0.65 W CoAP GET/PUT communication
using TLS/DTLS
BLE connectivity,
[40] ESP32 MQTT 0.73 W MQTT
dual-core processor
Low-power,
[41] Arduino MKR1000 HTTP 1.21 W HTTP GET/POST
onboard Wi-Fi
Low-power, long
[42] WSN430 6LoWPAN 0.42 mW 6LoWPAN GET/PUT
battery life

Based on this comparison, it can be seen that each study has its own unique features
and advantages. For example, study [43] uses a low-cost and open-source platform, while
study [38] offers powerful processing capabilities. Study [39] prioritizes secure communi-
cation using TLS/DTLS, and study [40] has the added benefit of BLE connectivity and a
dual-core processor. Study [41] focuses on low-power consumption and onboard Wi-Fi,
and study [42] boasts a very low power consumption of only 0.42 milli watts.
Overall, the choice of IoT platform and communication protocol will depend on
the specific needs and requirements of the data retrieval system. Factors such as power
consumption, cost, processing capabilities, and security should all be considered when
selecting the most suitable platform and protocol for a given application. The studies
mentioned above propose different techniques and architectures for building low-power
data retrieval systems for IoT applications. These systems require efficient algorithms to
manage the collected data and optimize the system’s performance. It is where the CHROA
optimization algorithm is applied to improve the data retrieval scheme.
The Chaotic Horse Ride Optimization Algorithm (CHROA) is an advanced opti-
mization algorithm that combines the principles of chaos theory with the Horse Ride
Optimization Algorithm (HROA). CHROA aims to enhance the optimization capabilities of
HROA by introducing chaotic behavior. It utilizes chaotic maps to introduce randomness
and unpredictability into the search process, improving search space exploration. CHROA
generates an initial population of potential solutions and iteratively updates its positions
based on fitness evaluations. The chaotic maps influence the position updates, diversifying
the search and balancing exploration and exploitation. Experimental results show that
CHROA outperforms other algorithms in terms of solution accuracy and convergence
speed. In cloud-enabled IoT environments, CHROA can be used for efficient workload
distribution, energy resource optimization, and improved system performance. It offers an
intelligent, energy-aware approach to load balancing in IoT/IoE applications.
Applying this algorithm makes optimizing the data retrieval system’s performance
possible by reducing power consumption, enhancing data transmission rates, and improv-
ing data accuracy. For example, CHROA can be used to optimize the design of WSN
used in smart agriculture, as demonstrated in the studies by [39,40]. The algorithm can be
used to optimize the placement of sensors, reduce energy consumption, and improve the
accuracy of data collection.
Similarly, the CHROA algorithm can be applied to optimize the performance of
low-power data retrieval systems, such as those proposed by [38,39,43], by optimizing
Sensors 2023, 23, 5349 8 of 25

communication protocols, data transmission rates, and power consumption. The CHROA
optimization algorithm can be applied to optimize the performance of low-power data
retrieval systems in IoT applications. The algorithm can be used to optimize the design
and operation of WSN, reduce power consumption, and enhance data transmission rates
and accuracy.
Limited literature is available on applying the CHROA optimization algorithm for
optimizing the performance of low-power data retrieval systems in IoT applications. How-
ever, some recent studies have shown promising results in this area. Paper [44] proposed a
novel method for optimizing the energy consumption of low-power data retrieval systems
in IoT applications using the CHROA optimization algorithm. The authors evaluated their
approach using a real-world application of temperature and humidity monitoring and
reported significant improvements in the energy efficiency.
Similarly, Ref. [45] presented a study on optimizing the performance of a WSN using the
CHROA optimization algorithm. The authors focused on minimizing the network’s energy
consumption while maintaining a certain level of coverage and connectivity. The results
showed that their approach significantly improved the energy efficiency of the network.
In another study, Ref. [46] applied the CHROA optimization algorithm to optimize
the low-power WSN routing. The authors focused on minimizing the network’s energy
consumption while maintaining reliable communication between the nodes. The results
presented in Table 3 showed that their approach outperformed traditional routing protocols
in terms of energy efficiency.

Table 3. Review of CHROA optimization algorithm applications.

Optimization
Study Objective Application Results
Target
Optimizing Low-power data Temperature Significant
[44] energy retrieval systems and humidity improvements in
consumption in IoT monitoring energy efficiency
Energy
Significant
Optimizing consumption,
[45] WSN improvement in
performance coverage, and
energy efficiency
connectivity
Energy Outperformed
Optimizing consumption traditional routing
[46] Low-power WSN
routing and reliable protocols in
communication energy efficiency

4. Proposed Model
Figure 2 depicts the overall system design of the model, where the IoE system com-
prises a set of clusters generated by the CHROA technique. Every cluster includes a cluster
head (CH) and related cluster members (CMs).
Moreover, the CMs send the data to the CHs, which are forwarded to the IoE BS. The
CHROA technique involves three major steps in offering real-time services as listed in
the following:
• CHROA-based clustering of the IoT networks,
• Cloud Data Storage and Processing,
• Client Services.
The following sections provide a detailed explanation of how these modules work.
Sensors 2023,
Sensors 2023,23,
23,x5349
FOR PEER REVIEW 9 of 25 10

Figure
Figure 2. Working
2. Working process
process of CHORA
of CHORA model. model.

4.1. Algorithmic Design of CHROA Technique

4.1. Algorithmic Design of CHROA Technique
The CHROA technique is mainly based on HROA and chaos theory concepts. Chaos
is in anThe CHROA
unbalanced technique
state by paying isattention
mainly based
to earlyon HROA and
conditions. chaosoptimal
It avoids theorylocal
concepts. C
is in an unbalanced state by paying attention to early conditions.
difficulties and improves the quality of solutions in various optimization procedures. It avoids optimal
difficulties and improves
Unlike a pendulum, the quality
a chaotic of solutions
system cannot in various
settle into a regularoptimization
rhythm due toprocedure
its
nonlinear dynamics. Using hierarchical horse herds promotes the CHROA.
Unlike a pendulum, a chaotic system cannot settle into a regular rhythm due A horse is an
example of a herd animal. Because many animals congregate in big groups, which reduces
nonlinear dynamics. Using hierarchical horse herds promotes the CHROA. A horse
conflict and improves social cohesion, a stable hierarchical structure or “pecking order” is
example of a herd animal. Because many animals congregate in big groups, which re
preferred. This type of linear system is used on occasion. In nonlinear hierarchies, Horse
Aconflict and improves
may dominate socialBcohesion,
Horse B, Horse a stable
may dominate hierarchical
Horse C, and so on. structure
Dominance or “pecking
might ord
bepreferred.
centered onThis typeofof
a range linear system
elements, such as aisperson’s
used ondesire
occasion. In nonlinear
for certain possessionshierarchies,
at a H
given
A maymoment. It might
dominate change
Horse B, during
Horse aBherd’s or individual
may dominate animal’s
Horse lifecycle.
C, and Humans
so on. Dominance m
force
be centered on a range of elements, such as a person’s desire for certainflow
horses to coexist in a confined location with few resources. Figure 3 shows the possession
diagram of the CHROA algorithm.
given moment. It might change during a herd’s or individual animal’s lifecycle. Hu
force horses to coexist in a confined location with few resources. Figure 3 shows the
diagram of the CHROA algorithm.
SensorsSensors 2023,
2023, 23, 23, 5349
x FOR PEER REVIEW 11 of 2610 of 25

Figure 3. Flow diagram of CHROA algorithm.

Figure 3. Flow diagram of CHROA algorithm.
Sensors 2023, 23, 5349 11 of 25

CHROA is referred to as “dominant horses” with dysfunctional social capabilities.

When grain is scarce, higher-status horses may ban lower-status horses from blocking
animals of lower social rank from eating at all. Animals with a lower social rank who eat
later risk not receiving enough food.
Access to resources is prioritized in a horse herd headed by a dominant stallion and
mare, depending on the horses’ hierarchical positions within the herd [7]. Initially, the
hierarchy of the horses in a herd is defined by their fitness levels. Consider a herd of
k horses, where P represents the function.

Herd = { H1 , . . . , Hk } (1)

P = Herd → 1, . . . , (2)


When fitness ( Hx ) < fitness Hy , whereas x 6= y and x, ye{1 . . . K }, then


P( Hx ) > P Hy (3)


When fitness ( Hx ) = fitness Hy , whereas x 6= y and x, ye{1 . . . K }, then


P( Hx ) − P Hy ( x − y) > 0 (4)

The rank of every horse Hx is defined as follows:

P ( Hx )
Hx − Rank o f each horse = (5)
K
Every herd has a center that is equivalent to the weighted average of the positions of
the horse from the herd, thus, the weights are denoted as a rank of the horse. The center of
the herd is estimated by:
∑kx=1 Zx Hx.rank
HerdCenter = (6)
∑kx=1 Hx.rank
Every herd has a center that corresponds to the weighted average of the positions of
its horses; thus, the weights represent the horse’s position in the herd. The approximate
position of the herd’s center is listed below:
v
u Dim
Dim(Stallion, herd) = t ∑ Staaliony − HerdCenter
u
2
(7)
y =1

Whereas Dim denotes the number of dimensions of the search space. When the horses
belong to the group herd of the horse then it upgrades its velocity by:
 
T +1 T T t
Velx,y = Velx,y + Hx.rank × Herdcenter.y − Zx,y (8)
 
T +1 T T t
Velx,y = Velx,y + Rand × Herdcenter.y − Zx,y (9)

where Rand denotes an arbitrary number from zero to one. T indicates the present iteration,
and T + 1 represents a novel iteration. The memory of a horse (Mem) is a matrix that has a
number of rows equivalent to the value of the Horse Memory Pool (HMP) and D columns.
 
T +1 T +1
Mem1,x,1 ··· Mem1,x,D
Mem Tx +1 = 
 .. .. .. 
(10)
 . . .


T +1 T +1
Mem HMP,x,1 · · · Mem HMP,x,D

The formula for updating the cell of the memory matrix is,
T +1 T +1
MemK,x,y = Zx,y × N (0, SD ) (11)
Sensors 2023, 23, 5349 12 of 25

where N represents a normal distribution, with 0 acting as the mean and SD serving as the
standard deviation.
Algorithm 1 provide a Pseudo Code for the CHROA Technique for a reference that
can be implemented in programming language.

Algorithm 1: Pseudo Code of CHROA Technique

Beginning of Algorithm 1
Initiate the horse population in the search space
Calculate fitness of the horse and upgrade the optimal one
While iteration < T
Define the center of every herd from T
For n = 1 to number of horse
Find hbest
Find cbest
For T = 1 to horse velocity
Estimate the rank of every horse from top to bottom
Upgrade the horse velocity and location
Calculate the fitness
next l
next n
If fitness condition is fulfilled
End
T = T+1, Return the Global optimum solution
End While
End of Algorithm.

The chaos idea is incorporated to enhance the global optimization capability of the
HROA method. Chaos is an imbalanced state that is very sensitive to the initial circum-
stances. It is used in a variety of optimization approaches to avoid optimal local issues
and improve solution quality [47]. As the metaheuristic technique is based on exploration
and exploitation phases, the chaos idea is presented for maintaining an efficient tradeoff
between exploitation and exploration, thus attaining an optimal solution efficiently. In
HROA, variable Hx has a significant influence on the convergence rate of the Artificial
Fish Optimization (AFO) method. The HROA method’s efficiency depends on its variables.
It might be observed by a large momentum to begin utilizing the potential search space,
and it cannot be revealed effectively. The chaos is utilized for attaining improved search
features for the exploitation and exploration area, thus improving the result to identify
optimal global results. The chaotic map is applied to define the position xik , whereas the
parameter θ substituted with values attained using chaotic map is given by,
 
xik+1 = xik + Cnap × x BH − xik , i = 1, 2, . . . , N∇ (12)

where, x BH denotes the location of BH from space, Cmap indicates the chaotic map, and N∇
denotes the total number of individuals in the optimization algorithm. In short, ten chaotic
maps are used to manipulate the value of random variables from the HROA method.

4.2. Design of Objective Function

The objective function of the presented study is determined in Equations (13)–(15).
The term ( β, γ) indicates the constant with a fixed value of (0.9, 0.3).
1 1
OM1 = Oene
M + Oene
M temp
(13)
Oload
M OM
1
OM2 = β + (1 − β)OM1 (14)
Odist
M
Sensors 2023, 23, 5349 13 of 25

1
OM3 = γOM2 + (1 − γ) (15)
Odel
M
Equation (16) provides the sensor node’s distance to the BS. The distances between the
CHs and BS is defined by Equation (17). Likewise, Odist
M ( n ) examines the distance between
2 nodes that should lie in the range [0, 1].
Odist
M ( m ) denotes the distance between the sensor node and CH represented by [24]:

Odist
M (m)
Odist
M = (16)
Odist
M (n)

N NCH
Odist
M (m) = ∑ ∑ k Npnormal − CH q k + kCH q − B I k (17)
p = l q =1

N NCH
Odist
M (n) = ∑ ∑ k Npnormαl − Nqnormal k (18)
p =1 q =1

Next, Equation (19) calculates the energy utilization by IoT devices. For improving
the network
 efficiency,
 the energy must be maximal as given in Equation (22):
ENE Npnormal specifies an energy of pth normal node.
 
ENE Nqnormal denotes an energy of qth normal node.

Oene
M (m)
Oene
M = (19)
Oene
M (n)
CH
Oene
M (m) = ∑ nENE(q) (20)
q =1
MM    
nENE(q) = ∑ 1 − ENE N normal × ENE(CH ) ; 1 ≤ q < CH (21)
p=1,pεq
  
Oene
M (n) = CH × MaxCH p = 1 ENE Npnormal
  (22)
MaxqCH
=1 ENE Nq
normαl

Finally, Equation (23) is utilized for measuring the delay endured by IoT devices while
transferring data to the BS. The number of sensor nodes in every cluster should be smaller
to avoid delay. Max (CH q ) represents the number of CHs in the network.
N N indicates the cumulative IoT sensors.
Max (CH q )
Mdel = (23)
NN

4.3. Cloud Data Storage and Processing

Cloud storage plays a crucial role in the proposed load-balancing model for the cloud-
enabled Internet of Everything (IoE) in the healthcare domain. Cloud storage is a specific
type of cloud computing architecture where a cloud computing provider stores data over
the Internet. This approach eliminates the need for organizations to build and maintain
their own data centers, reducing costs and providing storage space on demand [36]. By
utilizing cloud data management, the proposed model ensures that data are securely stored
on servers located in separate locations and managed by specialized cloud data hosting
providers. This enables benefits such as automated backups, expert assistance, and remote
access to data [48]. In the proposed model, data are processed and sent to the Gateway (GT)
before being stored in the cloud. Apache Flume, a Big Data tool, consumes and ingests the
data. The data are then stored in the Hadoop Distributed File System (HDFS) for efficient
and scalable storage [49]. This integrated approach leverages the capabilities of cloud
 sumes and ingests the data. The data are then stored in the Hadoop Distributed File Sys-
tem (HDFS) for efficient and scalable storage [49]. This integrated approach leverages the
capabilities of cloud storage, Apache Flume, HDFS, and big data analytics efficiently han-
dle 2023,
Sensors and23,analyze
5349 the massive volume of real-time streaming data generated by IoT14devices of 25

in the healthcare domain [50]. The combination of cloud computing and the Internet of
Everything (IoE) enablesstorage,the delivery
Apache of real-time
Flume, HDFS, and big dataservices.
analytics In the context
efficiently handle andof analyze
IoT devices,
the
big data analytics tools arevolume
massive utilized to process
of real-time anddata
streaming organize
generatedunstructured
by IoT devices indata from vari-
the healthcare
ous sources, such asdomain
sensors [50]. The combination of cloud computing and the Internet of Everything (IoE)
and social media, into consumable datasets. These data can
enables the delivery of real-time services. In the context of IoT devices, big data analytics
be used to derive insights and improve
tools are utilized operational
to process and efficiency
organize unstructured datainfrom
diverse
variousareas,
sources, ranging
such
as sensors and social media, into consumable datasets. These
from traffic patterns to home efficiency [48]. The proposed load balancing model incorpo- data can be used to derive
insights and improve operational efficiency in diverse areas, ranging from traffic patterns
rates Apache Flume,toahome scalable technique
efficiency for large-scale
[48]. The proposed load balancing real-time data streaming
model incorporates Apache Flume,in the
context of Big Data. aApache Flume for
scalable technique ensures a balanced
large-scale real-time data flow of data
streaming in theby dynamically
context of Big Data. ad-
Apache Flume ensures a balanced flow of data by dynamically adjusting the write and read
justing the write andratesread rates to accommodate varying data generation and consump-
to accommodate varying data generation and consumption speeds [51]. The essential
tion speeds [51]. Thearchitecture
essentialofarchitecture
Apache Flume isof Apache
depicted Flume
in Figure 4. is depicted in Figure 4.

Figure 4. Apache Flume.

Figure 4. Apache Flume.

The Flume framework The Flume framework

consists consists of
of agents, agents, events,
events, and clients.
and clients. Flume events
Flume events areare
the the
fundamental units of data transmitted through the Flume pipeline and stored in HDFS.
fundamental units ofFlume
dataagents
transmitted
collect data through
from clientstheandFlume
transmit pipeline and stored
them from sources to sinks,in HDFS.
using
Flume agents collectsources,
data channels,
from clients
and sinksand transmit
as essential them Flume
components. fromclients
sources to data
generate sinks, using
events
and send them to agents for storage in the HDFS platform.
sources, channels, and sinks as essential
Furthermore, the proposedcomponents. Flume
model incorporates Apacheclients
Spark, agenerate data
cluster-based events
solution,
and send them to agents for storage
for real-time in Apache
processing. the HDFS platform.
Spark has gained significant popularity for its ability to
handle Big Data challenges with fast computation. The model’s Algorithm 2 combines the
Furthermore, the proposed model incorporates Apache Spark, a cluster-based solu-
chaotic map equation and the Horse Ride Optimization Algorithm (HROA) technique to
tion, for real-time processing.
generate newApache
individuals Spark has gained
and explore the searchsignificant popularity
space efficiently. for its abil-
Overall, the proposed
ity to handle Big Dataloadchallenges
balancing model utilizes
with fastthecomputation.
capabilities of cloud storage,
The ApacheAlgorithm
model’s Flume, HDFS, and2 com-
Apache Spark to handle large-scale real-time data streaming, optimize resource allocation,
bines the chaotic mapandequation and processing
provide efficient the Horse Ride Optimization
in cloud-enabled Algorithm
Internet of Everything (HROA) tech-
(IoE) environments.
nique to generate new individuals
Algorithm 2 provide aand
Pseudo explore the search
Code for CHROA space
technique efficiently.
incorporating Overall,
cloud data storage the
and processing.
proposed load balancing model utilizes the capabilities of cloud storage, Apache Flume,
HDFS, and Apache Spark to handle large-scale real-time data streaming, optimize re-
source allocation, and provide efficient processing in cloud-enabled Internet of Everything
(IoE) environments. Algorithm 2 provide a Pseudo Code for CHROA technique incorpo-
rating cloud data storage and processing.
Sensors 2023, 23, 5349 15 of 25

Algorithm 2: CHROA technique incorporating cloud data storage and processing

Input:
- n: number of entities
- MaxIter: max no. of iterations
- p: probability of chaos
- alpha: chaos parameter
- objective function f(x) that involves processing large amounts of data stored in the cloud
Output:
- x_best: the best solution found
- f_best: the corresponding fitness value
Initialization:
- Initialize the population x_i, i = 1, . . . , n with random positions
Evaluation:
- Evaluate the fitness of each individual using the objective function f(x), which may involve
processing large amounts of data stored in the cloud
Selection:
- Select the best individuals based on their fitness values
Iteration:
- For iter = 1 to MaxIter do:
- Generate new individuals using the chaotic map equation and the HROA technique:
- Select two entities x_a and x_b arbitrarily from the selected best individuals
- Generate a fresh entities x_new using the chaotic map equation:
x_new = x_a + alpha × (x_b − x_a) × r
where r is a random no. between −1 and 1
- Apply the HROA technique to the new individual:
x_new = x_new + p × D × (x_best − x_new) × r
where D is the variance between the max and min positions in the population
- Evaluate the fitness of the new individual using the objective function f(x), which
may involve processing large amounts of data stored in the cloud
- Select the best entity from the collective populace of old and new entities- End for
Output:
- x_best: the best solution found
- f_best: the corresponding fitness value

4.4. Client Service

The proposed framework enables machine-to-machine interaction. In the healthcare
domain, various end users, such as patients, doctors, nursing staff, etc., can request and
access gathered and monitored data stored in the cloud. Service requests are handled
seamlessly regardless of origin, as the collected data are stored in the cloud. An integral
component of the model is the Event Aware Node (EAWN), which is responsible for
sending requests to the Database Server (DBS). The DBS, in turn, forwards these requests
to the IoE network [52]. This intelligent transmission facilitated by the EAWN significantly
reduces the turnaround time and improves the throughput of the system [53]. With the
EAWN’s awareness of events and its efficient forwarding of requests to the DBS, the load-
balancing model achieves a notable decrease in both the turnaround time and throughput,
enhancing the system’s overall performance.

5. Performance Analysis
The experimental results of the proposed CHROA model are validated using the
MATLAB tool. A network with a grid size of 5003 × 500 m is considered with a total
of 500 nodes. Moreover, a detailed comparative analysis is made with the existing ABC,
GSA, and WD-FA techniques (Table 4 and Figure 5). This section compares the CHROA
technique to other current practices with varied numbers of nodes, and the NLT value
should be high for better results. The figure portrayed that the ABC algorithm shows
a poor performance with the minimum NLT, whereas the GSA and WD-FA techniques
have demonstrated moderate outcomes. The proposed CHROA strategy, however, has
Sensors 2023, 23, 5349 16 of 25

achieved the highest levels of NLT of any of the approaches considered. With 100 nodes,
for instance, the NLT for the CHROA method is 18,972 rounds, while the NLTs for the
ABC method (12,739 rounds), the GSA method (15,101 rounds), and the WD-FA method
(17,995 rounds) are all lower. With 500 nodes, the NLT for the CHROA technique has gone
up to 24,404 rounds, while the NLT for the ABC, GSA, and WD-FA techniques has gone
down to 20,225, 21,265, and 23,193 rounds, respectively.

Table 4. Comparison of CHROA model with existing methods under different numbers of nodes.

Network Lifetime (Rounds)

No. of Nodes ABC GSA WD-FA CHROA
100 12,739 15,101 17,995 18,972
200 13,316 15,364 18,893 20,069
300 15,538 17,047 19,323 21,419
400 17,688 19,410 21,567 23,874
500 20,225 21,265 23,193 24,404
Total Energy Consumption (J)
No. of Nodes ABC GSA WD-FA CHROA
100 10.70 10.05 8.55 6.84
200 13.82 10.71 8.94 8.03
300 17.65 15.28 12.02 10.07
400 19.94 18.91 15.26 11.12
500 21.28 19.56 15.82 13.84
Throughput (Kbps)
No. of Nodes ABC GSA WD-FA CHROA
100 39.04 40.28 52.75 61.36
200 46.45 47.65 57.35 66.96
300 51.24 51.72 61.23 69.59
400
Sensors 2023, 23, x FOR PEER REVIEW 54.21 56.19 65.75 75.27 18 of 26
500 60.36 60.48 67.85 77.43

Figure5.5.Network
Figure Networklifetime
lifetimeanalysis
analysisofofCHROA
CHROAmodel
modelwith
withexisting
existingtechniques.
techniques.

Figure 5 inspects CHROA’s performance compared to the existing methods in terms

of the TEC, and the value of the TEC needs to be low for better performance. For example,
with 100 nodes, the CHROA approach provides a minimum TEC of 6.84 J, while the ABC,
GSA, and WD-FA procedures produce maximum TECs of 10.70 J, 10.05 J, and 8.55 J, re-
spectively. Finally, with 500 nodes, the CHROA approach has a lower TEC of 13.84 J, but
Sensors 2023, 23, 5349 17 of 25

Figure 5. Network lifetime analysis of CHROA model with existing techniques.

Figure 5 inspectsFigure
CHROA’s performance
5 inspects CHROA’scompared
performance to compared
the existing methods
to the inmethods
existing terms ofin terms
the TEC, and theofvalue of the
the TEC, andTEC
the needs
value oftothe
be TEC
low for
needsbetter
to beperformance.
low for better For example, For
performance. withexample,
with 100approach
100 nodes, the CHROA nodes, the provides
CHROA approach
a minimumprovides
TECaofminimum TEC of
6.84 J, while the6.84 J, while
ABC, GSA,the ABC,
GSA, andproduce
and WD-FA procedures WD-FA maximum
procedures produce
TECs of maximum TECs
10.70 J, 10.05 of 10.70
J, and 8.55J,J,10.05 J, and 8.55 J, re-
respectively.
Finally, with 500 spectively.
nodes, theFinally,
CHROA with 500 nodes,
approach hasthe CHROA
a lower TEC approach
of 13.84has a lower
J, but TEC of
the ABC, 13.84 J, but
GSA,
the ABC,exhibit
and WD-FA techniques GSA, and WD-FA
higher TECstechniques
of 21.28 J,exhibit
19.56 higher TECs of
J, and 15.82 21.28 J, 19.56 J, and 15.82 J,
J, respectively.
In Figure 6,respectively.
the throughput of the CHROA technique is compared to different models
In Figure 6, the throughput of the CHROA technique is compared to different models
with variable nodes, and the value of throughput must be larger to improve outcomes. The
with variable nodes, and the value of throughput must be larger to improve outcomes.
graph demonstrated that the ABC method performed poorly with low throughput, but the
The graph demonstrated that the ABC method performed poorly with low throughput,
GSA and WD-FAbut methods
the GSAsurpassed
and WD-FA intermediate results.intermediate
methods surpassed However, the provided
results. CHROA
However, the provided
algorithm outperformed the other strategies in terms of throughput.
CHROA algorithm outperformed the other strategies in terms of throughput.

Figure
Figure 6. Total energy 6. Total energy
consumption consumption
analysis analysis
of CHROA of CHROA model.
model.

Figure
Figure 7 shows that 7 shows
with that the
100 nodes, withCHROA
100 nodes, the CHROA
technique technique
achieved a higherachieved
through-a higher
put of 61.36 Kbps, whereas the ABC, GSA, and WD-FA methods achieved lower through- lower
throughput of 61.36 Kbps, whereas the ABC, GSA, and WD-FA methods achieved
throughputs of 39.04 Kbps, 40.28 Kbps, and 52.75 Kbps, respectively. Similarly, with 500
puts of 39.04 Kbps, 40.28 Kbps, and 52.75 Kbps, respectively. Similarly, with19500
Sensors 2023, 23, x FOR PEER REVIEW of nodes,
26 the
nodes, the CHROA technique obtained a higher throughput of 77.43 Kbps, but the ABC,
CHROA technique obtained a higher throughput of 77.43 Kbps, but the ABC, GSA, and
GSA, and WD-FA approaches achieved lower throughputs of 60.36 Kbps, 60.48 Kbps, and
WD-FA approaches achieved lower throughputs of 60.36 Kbps, 60.48 Kbps, and 67.85 Kbps,
respectively. Table 5 briefly
67.85 Kbps, respectively. compares
Table 5 brieflythe CHROAthe
compares technique
CHROA with other with
technique known algorithms
other
at various
known node counts.
algorithms at various node counts.

Figure 7. Throughput analysis of CHROA model with existing techniques.

Figure 7. Throughput analysis of CHROA model with existing techniques.
Table 5. Results analysis of the CHROA model using current approaches at various node counts.

Normalized Overhead (%) (NO)

No. of Nodes ABC GSA WD-FA CHROA
100 8.742 6.075 5.720 3.570
Sensors 2023, 23, 5349 18 of 25

Table 5. Results analysis of the CHROA model using current approaches at various node counts.

Normalized Overhead (%) (NO)

No. of Nodes ABC GSA WD-FA CHROA
100 8.742 6.075 5.720 3.570
200 10.874 8.734 5.830 4.160
300 11.868 10.72 6.600 4.740
400 15.595 13.46 8.830 5.390
500 15.813 14.83 8.350 6.110
End-to-End Delay (s)
No. of Nodes ABC GSA WD-FA CHROA
100 2.938 1.764 1.431 1.136
200 3.935 2.567 2.134 1.742
300 7.937 5.482 4.390 3.589
400 9.942 8.610 6.492 4.571
500 13.945 11.490 8.481 5.896

Figure 8 compares the performance of the provided CHROA model to previous

methods in terms of the NO, and the value of the NO must be lower for the best performance.
The figure shows that the CHROA technique has achieved maximum performance with
reduced NO levels. For example, with 100 nodes, the CHROA algorithm provides a
minimum NO of 3.570%, whereas the ABC, GSA, and WD-FA techniques provide a greater
NO of 8.742%, 6.075%, and 5.720%, respectively. Finally, with 500 nodes, the CHROA
Sensors 2023, 23, x FOR PEER REVIEW 20 of 26
technique achieves a minimum NO of 6.110%, but the ABC, GSA, and WD-FA algorithms
achieve better NOs of 15.813%, 14.83%, and 8.350%, respectively.

Figure8.8.Normalized
Figure Normalizedoverhead
overheadanalysis
analysisofofCHROA
CHROAmodel
modelwith
withexisting
existingtechniques.
techniques.

Figure9 9determines
Figure determinesthe theperformance
performanceofofthetheprojected
projectedCHROA
CHROAmannermannerwith
withexisting
existing
approaches in terms of the ETE delay, and the minimum value of
approaches in terms of the ETE delay, and the minimum value of the ETE delay gives the ETE delay gives
optimalperformance.
optimal performance. The The figure
figureshows
showsthat
thatthe
theCHROA
CHROA method
method hashas
gained higher
gained per-
higher
formance with
performance withthethe
minimum
minimum values of ETE
values delay.
of ETE For For
delay. the sample withwith
the sample 100 nodes, a lower
100 nodes, a
ETE delay
lower of 1.136
ETE delay s exists
of 1.136 by theby
s exists CHROA manner,
the CHROA whereas
manner, a maximal
whereas ETE delay
a maximal ETEof 2.938
delay
ofs,2.938
1.764s,s,1.764
and 1.431
s, ands1.431
have sbeen
haveoffered by the by
been offered ABC,
theGSA,
ABC,andGSA, WD-FA methods,
and WD-FA respec-
methods,
respectively. At with
tively. At last, last, with 500 nodes,
500 nodes, a minimum
a minimum ETEETE delay
delay of 5.896
of 5.896 s issoutperformed
is outperformed by by
the
CHROA approach, whereas a maximum ETE delay of 13.945 s, 11.490 s, and 8.481 s have
been reached by the ABC, GSA, and WD-FA methodologies, respectively.
 Figure 9 determines the performance of the projected CHROA manner with ex
approaches in terms of the ETE delay, and the minimum value of the ETE delay
optimal performance. The figure shows that the CHROA method has gained highe
formance with the minimum values of ETE delay. For the sample with 100 nodes, a
Sensors 2023, 23, 5349 19 of 25
ETE delay of 1.136 s exists by the CHROA manner, whereas a maximal ETE delay of
s, 1.764 s, and 1.431 s have been offered by the ABC, GSA, and WD-FA methods, re
tively. At last, with 500 nodes, a minimum ETE delay of 5.896 s is outperformed b
the CHROA approach,
CHROAwhereas
approach,a maximum ETE delay ETE
whereas a maximum of 13.945
delays,of11.490
13.945s,s,and 8.481
11.490 s 8.481 s
s, and
have been reached by the ABC, GSA, and WD-FA methodologies, respectively.
been reached by the ABC, GSA, and WD-FA methodologies, respectively.

Figure
Figure 9. Comparison of 9.
ETEComparison of ETE delay
delay of CHROA of CHROA
with current with current techniques.
techniques.

Figure 10 showsFigure 10 shows

the average the average
TEC (Total TEC (Total Energy
Energy Consumption) analysisConsumption)
of the CHROA analysis o
model. According CHROAto themodel.
data, According
ABC has atopoor the data, ABCTEC
average has aofpoor average
16.678 TEC of the
J, whereas 16.678 J, wh
GSA technique thehas GSA technique
a slightly has a average
increased slightly increased average
TEC of 14.902 J. InTEC of 14.902
addition, theJ.WD-FS
In addition, the
FS techniquea has
technique has accomplished accomplished
moderate a moderate
TEC of 12.118 TEC of
J. However, the12.118 J. However,
suggested CHROA the sugg
Sensors 2023, 23, x FOR PEER REVIEW
approach resultedCHROA approach
in effective network resulted in effective
performance withnetwork
an averageperformance
TEC of 9.980 21 J,ofwhich
with 26
an average T
9.980 J, which
is the lowest achievable value. is the lowest achievable value.

Figure 10. Comparison of avg. TEC of CHROA model with current techniques.
Figure 10. Comparison of avg. TEC of CHROA model with current techniques.

Figure 11
Figure 11 compares
compares the
the CHROA’s
CHROA’s average
averagethroughput
throughputtotoseveral techniques.
several ABC
techniques. ABC
and GSA have minimal average throughputs of 50.260 Kbps and 51.264 Kbps, respec-
and GSA have minimal average throughputs of 50.260 Kbps and 51.264 Kbps, respectively.
tively. Moreover,
Moreover, the WD-FA
the WD-FA technique
technique depicted
depicted a moderate
a moderate average
average throughput
throughput of 60.986
of 60.986 Kbps.
Kbps. With an average throughput of 70,122 Kbps, the proposed CHROA
With an average throughput of 70,122 Kbps, the proposed CHROA method exceedsmethod exceeds the
the competition. The tables and graphs below demonstrate CHROA’s effectiveness as a
competition. The tables and graphs below demonstrate CHROA’s effectiveness as a cloud-
cloud-based IoE load balancer [26].
based IoE load balancer [26].
 Figure 11 compares the CHROA’s average throughput to several techniques. ABC
and GSA have minimal average throughputs of 50.260 Kbps and 51.264 Kbps, respec-
tively. Moreover, the WD-FA technique depicted a moderate average throughput of 60.986
Sensors 2023, 23, 5349
Kbps. With an average throughput of 70,122 Kbps, the proposed CHROA method exceeds
20 of 25
the competition. The tables and graphs below demonstrate CHROA’s effectiveness as a
cloud-based IoE load balancer [26].

Figure11.
Figure 11. Comparison
Comparisonof
ofavg.
avg. throughput
throughput of
of CHROA
CHROAmodel
modelwith
withcurrent
currenttechniques.
techniques.

6. Comparison
6. Comparison of
of CHROA
CHROA with
with Existing
ExistingTechniques
Techniques
Table66presents
Table presents a comparison
a comparison of the
of the performance
performance metrics
metrics of optimization
of four four optimization algo-
algorithms,
including the CHROA,
rithms, including ABC, GSA,
the CHROA, andGSA,
ABC, WD-FA.
and WD-FA.

Table 6. Comparison of CHROA with existing techniques.

Performance Metrics CHROA ABC GSA WD-FA Performance Metrics

Average TEC (mJ/bit) 0.1736 0.2059 0.1891 0.2031 Average TEC (mJ/bit)
Average Throughput (Kbps) 70.122 58.247 59.957 60.819 Average Throughput (Kbps)
ETE Delay (s) 0.0643 0.0781 0.0716 0.0752 ETE Delay (s)
Normalized Overhead 0.4069 0.4808 0.4437 0.4784 Normalized Overhead
Network Lifetime (s) 510.256 445.561 481.725 467.865 Network Lifetime (s)
Total Energy Total Energy
5.1289 6.7855 6.2234 6.7464
Consumption (Joule) Consumption (Joule)
Throughput (Kbps) 69.485 55.621 58.072 58.961 Throughput (Kbps)

In terms of the average TEC (mJ/bit) metric, which measures the total energy consumed
by the network to transmit one bit of data, CHROA outperformed the other three algorithms,
with the lowest average TEC of 0.1736 mJ/bit. The average throughput (Kbps) metric, which
measures the amount of data transmitted per unit of time, was also highest for CHROA, with
an average throughput of 70.122 Kbps. ABC, GSA, and WD-FA had average throughputs
of 58.247 Kbps, 59.957 Kbps, and 60.819 Kbps, respectively. The ETE delay (s) metric, which
measures the end-to-end delay in data transmission, was also lowest for CHROA, with
a delay of 0.0643 s. ABC, GSA, and WD-FA had ETE delays of 0.0781 s, 0.0716 s, and
0.0752 s, respectively. The normalized overhead metric, which measures the number of control
messages required to maintain the network, was lowest for CHROA, with a normalized
overhead of 0.4069. ABC, GSA, and WD-FA had normalized overheads of 0.4808, 0.4437, and
0.4784, respectively.
In terms of the network lifetime (s), which measures the duration for which the
network can operate efficiently, CHROA outperformed the other three algorithms, with a
network lifetime of 510.256 s. ABC, GSA, and WD-FA had network lifetimes of 445.561 s,
481.725 s, and 467.865 s, respectively. The total energy consumption (Joule) metric, which
measures the total energy consumed by the network, was lowest for CHROA, with a total
Sensors 2023, 23, 5349 21 of 25

energy consumption of 5.1289 Joule. ABC, GSA, and WD-FA had total energy consumptions
of 6.7855 Joule, 6.2234 Joule, and 6.7464 Joule, respectively.
Finally, in terms of the throughput (Kbps) metric, which measures the amount of data
transmitted per unit of time, CHROA again outperformed the other three algorithms, with
a throughput of 69.485 Kbps. ABC, GSA, and WD-FA had throughputs of 55.621 Kbps,
58.072 Kbps, and 58.961 Kbps, respectively. Overall, CHROA outperformed ABC, GSA,
and WD-FA in most of the performance metrics evaluated, indicating that it is an effective
optimization algorithm for energy saving and load balancing in cloud-based IoE environments.
To gain further insight into the data, Analysis of Variance (ANOVA) and the Wilcoxon
signed-rank test were conducted on the results available in Table 6 and the analyses were
presented in Tables 7 and 8, respectively.

Table 7. Variance (ANOVA) test on comparison of algorithms.

Performance Metric F-Statistic p-Value

Average TEC (mJ/bit) 0.157 0.854
Average Throughput (Kbps) 4.706 0.014
ETE Delay (s) 0.026 0.876
Normalized Overhead 0.066 0.8
Network Lifetime (s) 0.378 0.693
Total Energy Consumption (Joule) 0.722 0.531
Throughput (Kbps) 0.494 0.622

Table 8. Wilcoxon signed-rank test on comparison of algorithms.

Performance Metric Z-Statistic p-Value

Average TEC (mJ/bit) −0.833 0.407
Average Throughput (Kbps) 2.485 0.013
ETE Delay (s) −0.246 0.807
Normalized Overhead −0.399 0.691
Network Lifetime (s) 0.618 0.539
Total Energy Consumption (Joule) −1.334 0.182
Throughput (Kbps) 1.242 0.219

We have considered that:

• Null Hypothesis (H0): There is no significant difference in the medians of the algorithms.
• Alternative Hypothesis (H1): There is a significant difference in the medians of
the algorithms.
The p-value for the Average Throughput metric was less than 0.05, which means we could
reject the null hypothesis. Therefore, we can conclude that there is a significant difference
between the means of the three groups for the Average Throughput (Kbps) metric.
The p-value for the Average Throughput metric was less than 0.05, which means we
could reject the null hypothesis. Therefore, we can conclude that there is a significant
difference between the means of the two groups for the Average Throughput (Kbps) metric.
Therefore, further based on the Analysis of Variance (ANOVA) and Wilcoxon signed-
rank test, it is clear that there is a significant variation in the Average Throughput; hence,
CHROA is outperforming the other algorithms significantly.

7. Future Research Work

While this research presents a promising approach to optimizing the CHROA algo-
rithm for cloud-based IoE environments, there is still room for further investigation and
Sensors 2023, 23, 5349 22 of 25

improvement. Future research could focus on developing AI-based performance enhance-

ment strategies for the system to optimize energy savings and load balancing further.
Additionally, integrating UAVs with the IoE to overcome network coverage and resource
constraints should be explored further, along with developing routing protocols based on
contextual information to enhance the network performance. Moreover, load-balancing
strategies supported by clustering could be further studied to more efficiently manage
IoT activities while considering energy and performance requirements. Further traffic
shaping and traffic policing techniques can be utilized to improve QoS and High-Available
Proxy (HA proxy) for improving the system reliability and load balancing. The proposed
dynamic clustering approach based on vehicle position, velocity, and heading is effective
even for UAVs, but more research could be conducted to optimize its performance. Overall,
further research could lead to developing more effective and efficient energy-saving and
load-balancing approaches in cloud-based IoE environments.

8. Conclusions
The research suggests a novel energy-saving and load-balancing approach in cloud-
based IoE environments using the CHROA. The CHROA method employs a fitness function
with three parameters to optimize the selection of CHs and cluster construction, while
Apache Flume is used as a Big Data tool for data ingestion into the HDFS for storage. The
suggested scheme has been assessed using MATLAB and compared to recent techniques
based on different evaluation parameters. The results demonstrate that the CHROA
method outperforms existing methods such as ABC, GSA, and WD-FA with a maximum
average throughput of 70.122 kbps. Based on these statistical tests, it is clear that there
is a significant variation in terms of the Average Throughput; hence, it was proved that
CHROA is outperforming the other algorithms significantly.
This study suggests integrating IoE with Unmanned Aerial Vehicles (UAVs) to over-
come network coverage and resource constraints, and designing routing protocols based
on contextual information to enhance the network performance. In addition, using a
load-balancing strategy supported by clustering can efficiently manage IoT activities while
considering energy and performance requirements. The proposed dynamic clustering
exhibits better performance. Although the proposed approach shows promising results,
future research is required to develop AI-based performance enhancement strategies for
the system’s future. Traffic shaping regulates the flow of network traffic to meet specific
bandwidth or latency requirements, while traffic policing monitors and enforces traffic rate
limits. These techniques are typically implemented in network devices such as routers or
switches to control traffic flow.
HA proxies are commonly used to improve the system reliability and load balancing
by distributing incoming network traffic across multiple backend servers. HA proxies help
ensure high availability by handling traffic redirection and failover. Overall, this study
highlights the potential of the CHROA method for energy-saving and load balancing in
IoE environments and suggests several avenues for future research to enhance the network
performance and efficiency, especially in the healthcare domain that requires low-latency
real-time applications.

Author Contributions: Conceptualization, M.S., S.B.K. and S.A.; formal analysis, I.A., I.M.K. and
A.A.; investigation, S.B.K., I.A., A.A. and N.A.A.; project administration, S.B.K. and N.A.A.; resources,
N.A.A.; supervision, S.B.K. and N.A.A.; visualization, S.A. and I.M.K.; writing—original draft, M.S.,
S.A. and S.B.K.; writing—review and editing, I.A., I.M.K., A.A. and N.A.A. All authors have read and
agreed to the published version of the manuscript.
Funding: This research is supported by Princess Nourah bint Abdulrahman University Researchers
Supporting Project number (PNURSP2023R432), Princess Nourah bint Abdulrahman University,
Riyadh, Saudi Arabia.
Institutional Review Board Statement: Not Applicable.
Informed Consent Statement: Not Applicable.
Sensors 2023, 23, 5349 23 of 25

Data Availability Statement: Not Applicable.

Acknowledgments: This research is supported by Princess Nourah bint Abdulrahman University
Researchers Supporting Project number (PNURSP2023R432), Princess Nourah bint Abdulrahman
University, Riyadh, Saudi Arabia.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Alam, S.; Siddiqui, S.T.; Ahmad, A.; Ahmad, R.; Shuaib, M. Internet of Things (IoT) Enabling Technologies, Requirements,
and Security Challenges. In Advances in Data and Information Sciences; Lecture Notes in Networks and Systems; Springer:
Berlin/Heidelberg, Germany, 2020; Volume 94, pp. 119–126.
2. Di Martino, B.; Li, K.-C.; Yang, L.T.; Esposito, A. Internet of Everything: Algorithms, Methodologies, Technologies and Perspectives;
Springer: Berlin/Heidelberg, Germany, 2018; ISBN 9811058601.
3. Singh, A.; Joshi, K.; Alam, S.; Bharany, S.; Shuaib, M.; Ahmad, S. Internet of Things-Based Integrated Remote Electronic Health
Surveillance and Alert System: A Review. In Proceedings of the 2022 IEEE International Conference on Current Development in
Engineering and Technology (CCET), Bhopal, India, 23–24 December 2022; pp. 1–6.
4. Al-Fuqaha, A.; Guizani, M.; Mohammadi, M.; Aledhari, M.; Ayyash, M. Internet of Things: A Survey on Enabling Technologies,
Protocols, and Applications. IEEE Commun. Surv. Tutor. 2015, 17, 2347–2376. [CrossRef]
5. Neelakandan, S.; Berlin, M.A.; Tripathi, S.; Devi, V.B.; Bhardwaj, I.; Arulkumar, N. IoT-Based Traffic Prediction and Traffic Signal
Control System for Smart City. Soft Comput. 2021, 25, 12241–12248. [CrossRef]
6. Han, Z.; Yang, Y.; Wang, W.; Zhou, L.; Gadekallu, T.R.; Alazab, M.; Gope, P.; Su, C. RSSI Map-Based Trajectory Design for UGV
Against Malicious Radio Source: A Reinforcement Learning Approach. IEEE Trans. Intell. Transp. Syst. 2023, 24, 4641–4650.
[CrossRef]
7. Kamalraj, R.; Neelakandan, S.; Kumar, M.R.; Rao, V.C.S.; Anand, R.; Singh, H. Interpretable Filter Based Convolutional Neural
Network (IF-CNN) for Glucose Prediction and Classification Using PD-SS Algorithm. Measurement 2021, 183, 109804. [CrossRef]
8. Swarna Priya, R.M.; Bhattacharya, S.; Maddikunta, P.K.R.; Somayaji, S.R.K.; Lakshmanna, K.; Kaluri, R.; Hussien, A.; Gadekallu, T.R.
Load Balancing of Energy Cloud Using Wind Driven and Firefly Algorithms in Internet of Everything. J. Parallel Distrib. Comput.
2020, 142, 16–26.
9. Kavitha, T.; Mathai, P.P.; Karthikeyan, C.; Ashok, M.; Kohar, R.; Avanija, J.; Neelakandan, S. Deep Learning Based Capsule Neural
Network Model for Breast Cancer Diagnosis Using Mammogram Images. Interdiscip. Sci. Comput. Life Sci. 2021, 14, 113–129.
[CrossRef]
10. Neelakandan, S.; Arun, A.; Bhukya, R.R.; Hardas, B.M.; Kumar, T.C.; Ashok, M. An Automated Word Embedding with Parameter
Tuned Model for Web Crawling. Intell. Autom. Soft Comput. 2022, 32, 1617–1632. [CrossRef]
11. Rani, S.; Babbar, H.; Srivastava, G.; Gadekallu, T.R.; Dhiman, G. Security Framework for Internet-of-Things-Based Software-
Defined Networks Using Blockchain. IEEE Internet Things J. 2023, 10, 6074–6081. [CrossRef]
12. Gupta, P.; Varshney, A.; Khan, M.R.; Ahmed, R.; Shuaib, M.; Alam, S. Unbalanced Credit Card Fraud Detection Data: A Machine
Learning-Oriented Comparative Study of Balancing Techniques. Procedia Comput. Sci. 2023, 218, 2575–2584. [CrossRef]
13. Shuaib, M.; Bhatia, S.; Alam, S.; Masih, R.K.; Alqahtani, N.; Basheer, S.; Alam, M.S. An Optimized, Dynamic, and Efficient
Load-Balancing Framework for Resource Management in the Internet of Things (IoT) Environment. Electronics 2023, 12, 1104.
[CrossRef]
14. Alam, S.; Bhatia, S.; Shuaib, M.; Khubrani, M.M.; Alfayez, F.; Malibari, A.A.; Ahmad, S. An Overview of Blockchain and IoT
Integration for Secure and Reliable Health Records Monitoring. Sustainability 2023, 15, 5660. [CrossRef]
15. Alam, S. Security Concerns in Smart Agriculture and Blockchain-Based Solution. In Proceedings of the 2022 OPJU International
Technology Conference on Emerging Technologies for Sustainable Development (OTCON), Raigarh, India, 8–10 February 2023; pp. 1–6.
16. Alam, S.; Shuaib, M.; Ahmad, S.; Jayakody, D.N.K.; Muthanna, A.; Bharany, S.; Elgendy, I.A. Blockchain-Based Solutions Supporting
Reliable Healthcare for Fog Computing and Internet of Medical Things (IoMT) Integration. Sustainability 2022, 14, 15312. [CrossRef]
17. Cao, J.; Xu, L.; Abdallah, R.; Shi, W. EdgeOS_H: A Home Operating System for Internet of Everything. In Proceedings of the 2017
IEEE 37th International Conference on Distributed Computing Systems (ICDCS), Atlanta, GA, USA, 5–8 June 2017; pp. 1756–1764.
18. Naranjo, P.G.V.; Pooranian, Z.; Shojafar, M.; Conti, M.; Buyya, R. FOCAN: A Fog-Supported Smart City Network Architecture
for Management of Applications in the Internet of Everything Environments. J. Parallel Distrib. Comput. 2019, 132, 274–283.
[CrossRef]
19. Singh, P.; Nayyar, A.; Kaur, A.; Ghosh, U. Blockchain and Fog Based Architecture for Internet of Everything in Smart Cities.
Future Internet 2020, 12, 61. [CrossRef]
20. Neelakandan, S.; Divyabharathi, S.; Rahini, S.; Vijayalakshmi, G. Large Scale Optimization to Minimize Network Traffic Using
MapReduce in Big Data Applications. In Proceedings of the 2016 International Conference on Computation of Power, Energy
Information and Commuincation (ICCPEIC), Melmaruvathur, India, 20–21 April 2016; pp. 193–199.
21. Miao, Y.; Liu, X.; Choo, K.-K.R.; Deng, R.H.; Wu, H.; Li, H. Fair and Dynamic Data Sharing Framework in Cloud-Assisted Internet
of Everything. IEEE Internet Things J. 2019, 6, 7201–7212. [CrossRef]
Sensors 2023, 23, 5349 24 of 25

22. Xiao, H.; Zhang, Z.; Zhou, Z. GWS—A Collaborative Load-Balancing Algorithm for Internet-of-Things. Sensors 2018, 18, 2479.
[CrossRef]
23. Ramalingam, C.; Mohan, P. An Efficient Applications Cloud Interoperability Framework Using I-Anfis. Symmetry 2021, 13, 268.
[CrossRef]
24. Garzia, F.; Papi, L. An Internet of Everything Based Integrated Security System for Smart Archaeological Areas. In Proceedings of the
2016 IEEE International Carnahan Conference on Security Technology (ICCST), Orlando, FL, USA, 24–27 October 2016; pp. 1–8.
25. Babou, C.S.M.; Fall, D.; Kashihara, S.; Taenaka, Y.; Bhuyan, M.H.; Niang, I.; Kadobayashi, Y. Hierarchical Load Balancing and
Clustering Technique for Home Edge Computing. IEEE Access 2020, 8, 127593–127607. [CrossRef]
26. Lv, Z.; Yu, Z.; Xie, S.; Alamri, A. Deep Learning-Based Smart Predictive Evaluation for Interactive Multimedia-Enabled Smart
Healthcare. ACM Trans. Multimed. Comput. Commun. Appl. 2022, 18. [CrossRef]
27. Sangaiah, A.K.; Javadpour, A.; Ja’fari, F.; Pinto, P.; Ahmadi, H.; Zhang, W. CL-MLSP: The Design of a Detection Mechanism for
Sinkhole Attacks in Smart Cities. Microprocess. Microsyst. 2022, 90, 104504. [CrossRef]
28. Dong, C.; Xu, X.; Liu, A.; Liang, X. Load Balancing Routing Algorithm Based on Extended Link States in LEO Constellation
Network. China Commun. 2022, 19, 247–260. [CrossRef]
29. Jeyaraj, R.; Balasubramaniam, A.; Ajay Kumara, M.A.; Guizani, N.; Paul, A. Resource Management in Cloud and Cloud-Influenced
Technologies for Internet of Things Applications. ACM Comput. Surv. 2023, 55, 1–37. [CrossRef]
30. Tarahomi, M.; Izadi, M.; Ghobaei-Arani, M. An Efficient Power-Aware VM Allocation Mechanism in Cloud Data Centers: A Micro
Genetic-Based Approach. Cluster Comput. 2021, 24, 919–934. [CrossRef]
31. Saba, T.; Rehman, A.; Haseeb, K.; Alam, T.; Jeon, G. Cloud-Edge Load Balancing Distributed Protocol for IoE Services Using
Swarm Intelligence. Cluster Comput. 2023. [CrossRef] [PubMed]
32. Ghobaei-Arani, M.; Shahidinejad, A. A Cost-Efficient IoT Service Placement Approach Using Whale Optimization Algorithm in
Fog Computing Environment. Expert Syst. Appl. 2022, 200, 117012. [CrossRef]
33. Farahbakhsh, F.; Shahidinejad, A.; Ghobaei-Arani, M. Multiuser Context-aware Computation Offloading in Mobile Edge
Computing Based on Bayesian Learning Automata. Trans. Emerg. Telecommun. Technol. 2021, 32, e4127. [CrossRef]
34. Quy, V.K.; Van Hau, N.; Van Anh, D.; Ngoc, L.A. Smart Healthcare IoT Applications Based on Fog Computing: Architecture,
Applications and Challenges. Complex Intell. Syst. 2022, 8, 3805–3815. [CrossRef]
35. Khanh, Q.V.; Nguyen, V.-H.; Minh, Q.N.; Van, A.D.; Le Anh, N.; Chehri, A. An Efficient Edge Computing Management Mechanism
for Sustainable Smart Cities. Sustain. Comput. Inform. Syst. 2023, 38, 100867. [CrossRef]
36. Emmanuel, A.A.; Awokola, J.A.; Alam, S.; Bharany, S.; Agboola, P.; Shuaib, M.; Ahmed, R. A Hybrid Framework of Blockchain
and IoT Technology in the Pharmaceutical Industry: A Comprehensive Study. Mob. Inf. Syst. 2023, 2023, 3265310. [CrossRef]
37. Liu, Y.; Sun, Q.; Sharma, A.; Sharma, A.; Dhiman, G. Line Monitoring and Identification Based on Roadmap towards Edge
Computing. Wirel. Pers. Commun. 2021, 127, 441–464. [CrossRef]
38. Hakak, S.; Gadekallu, T.R.; Maddikunta, P.K.R.; Ramu, S.P.; M, P.; De Alwis, C.; Liyanage, M. Autonomous Vehicles in 5G and
beyond: A Survey. Veh. Commun. 2023, 39, 100551. [CrossRef]
39. Amr, M.E.; Al-Awamry, A.A.; Elmenyawi, M.A.; Eldien, A.S.T. Design and Implementation of a Low-Cost IoT Node for Data
Processing, Case Study: Smart Agriculture. J. Commun. 2022, 17, 99–109. [CrossRef]
40. Zaman, U.; Imran; Mehmood, F.; Iqbal, N.; Kim, J.; Ibrahim, M. Towards Secure and Intelligent Internet of Health Things:
A Survey of Enabling Technologies and Applications. Electronics 2022, 11, 1893. [CrossRef]
41. Khriji, S.; Benbelgacem, Y.; Chéour, R.; El Houssaini, D.; Kanoun, O. Design and Implementation of a Cloud-Based Event-Driven
Architecture for Real-Time Data Processing in Wireless Sensor Networks. J. Supercomput. 2022, 78, 3374–3401. [CrossRef]
42. Al Sohan, M.F.A.; Nahar, A. A Low-Power Wireless Sensor Network for a Smart Irrigation System Powered by Solar Energy.
In Proceedings of the 2nd International Conference on Computing Advancements, Dhaka, Bangladesh, 10–12 March 2022;
Association for Computing Machinery: New York, NY, USA; pp. 537–543.
43. Mousavi, S.M.; Khademzadeh, A.; Rahmani, A.M. The Role of Low-Power Wide-Area Network Technologies in Internet of
Things: A Systematic and Comprehensive Review. Int. J. Commun. Syst. 2022, 35, e5036. [CrossRef]
44. Bouguera, T.; Diouris, J.F.; Chaillout, J.J.; Jaouadi, R.; Andrieux, G. Energy Consumption Model for Sensor Nodes Based on LoRa
and LoRaWAN. Sensors 2018, 18, 2104. [CrossRef]
45. Song, Y.; Xin, R.; Chen, P.; Zhang, R.; Chen, J.; Zhao, Z. Identifying Performance Anomalies in Fluctuating Cloud Environments:
A Robust Correlative-GNN-Based Explainable Approach. Futur. Gener. Comput. Syst. 2023, 145, 77–86. [CrossRef]
46. Amro, D. Dynamic Energy-Efficient Routing Protocol for Wireless Sensor Networks. Palest. J. Tech. Appl. Sci. 2019, 6, 23–36.
47. Bomgni, A.B.; Ali, H.M.; Shuaib, M.; Mtopi Chebu, Y. Multihop Uplink Communication Approach Based on Layer Clustering in
LoRa Networks for Emerging IoT Applications. Mob. Inf. Syst. 2023, 2023, 5828671. [CrossRef]
48. Reshma, G.; Al-Atroshi, C.; Nassa, V.K.; Geetha, B.T.; Sunitha, G.; Galety, M.G.; Neelakandan, S. Deep Learning-Based Skin
Lesion Diagnosis Model Using Dermoscopic Images. Intell. Autom. Soft Comput. 2022, 31, 621–634. [CrossRef]
49. Kirola, M.; Memoria, M.; Shuaib, M.; Joshi, K.; Alam, S.; Alshanketi, F. A Referenced Framework on New Challenges and Cutting-
Edge Research Trends for Big-Data Processing Using Machine Learning Approaches. In Proceedings of the 2023 International
Conference on Smart Computing and Application (ICSCA), Hail, Saudi Arabia, 5–6 February 2023; pp. 1–5.
50. Alam, S.; Mohammad, O.K.J.; Alfurhood, B.S.; Saxena, K.k.; Anand, M.; Mahaveerakannan, R.; Savitha, V. Effective Sound
Detection System in Commercial Car Vehicles Using Msp430 Launchpad Development. Multimed. Tools Appl. 2023. [CrossRef]
Sensors 2023, 23, 5349 25 of 25

51. Shu, B.; Chen, H.; Sun, M. Dynamic Load Balancing and Channel Strategy for Apache Flume Collecting Real-Time Data Stream.
In Proceedings of the 2017 IEEE International Symposium on Parallel and Distributed Processing with Applications and 2017 IEEE
International Conference on Ubiquitous Computing and Communications (ISPA/IUCC), Guangzhou, China, 12–15 December 2017;
pp. 542–549.
52. Neelakandan, S.; Anand, J.G. Trust Based Optimal Routing in MANET’s. In Proceedings of the 2011 International Conference on
Emerging Trends in Electrical and Computer Technology, Nagercoil, India, 23–24 March 2011; pp. 1150–1156.
53. Asha, P.; Natrayan, L.; Geetha, B.T.; Beulah, J.R.; Sumathy, R.; Varalakshmi, G.; Neelakandan, S. IoT Enabled Environmental
Toxicology for Air Pollution Monitoring Using AI Techniques. Environ. Res. 2022, 205, 112574. [CrossRef] [PubMed]

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