IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO.

3, 1 FEBRUARY 2024 3757

An Improved Congestion-Controlled Routing

Protocol for IoT Applications in
Extreme Environments
Muhammad Adil , Member, IEEE, Muhammad Usman , Member, IEEE,
Mian Ahmad Jan , Senior Member, IEEE, Hussein Abulkasim , Senior Member, IEEE,
Ahmed Farouk, Senior Member, IEEE, and Zhanpeng Jin , Senior Member, IEEE

Abstract—The Internet of Things (IoT) has shown its presence improvement in these performance metrics, in the presence of
in applications that require monitoring extreme environments, the existing schemes. Despite that, we checked the computation
such as wildfires, military operations, and coastal areas, among complexity of the proposed approach with existing protocols,
others. In these applications, the IoT nodes are deployed in haz- which demonstrated noteworthy outcomes just like the other
ardous terrains where humanistic access is hard or not possible. metrics.
Hence, to ensure reliable data transmission in these applica-
tions, novel routing protocols need to be designed due to the Index Terms—Deep-Q-learning (DQL), DSDV, dynamic routing
multihop nature of communication possessed by the deployed protocols, Internet of Things (IoT), Quality of Service (QoS).
nodes. Currently, most of the routing protocols utilized by IoT
nodes follow traditional approaches, which creates congestion and I. I NTRODUCTION
contention in the network. As a result, the network performance
N THE recent past, Internet of Things (IoT) technology has
is degraded in terms of various communication metrics. To
address this problem and improve the communication statistics
in extreme environments, we propose a deep-Q-learning-enable-
I emerged as an effective paradigm for extreme environment
monitoring and operations, where humanistic access is hard
destination-sequenced distance-vector (DQL-DSDV) framework. or not possible [1], [2]. Given that, in the future, this tech-
DQL-DSDV focuses on selecting the next hop during communi-
cation. Initially, the DSDV protocol updates routing information nology could be a game changer in many domains that will
for connected nodes. This information is subsequently utilized by range from general to specific, such as smart cities, intelligent
the deep-Q-learning (DQL) algorithm to compute the next hop transportation, smart homes, healthcare, wildfire monitoring,
count. This computation is based on reward functions, known and secrete military operations, etc. Form the future perspec-
as Q-values, which are conceptualized as the distance between tive of this technology, it is clearly visible that how effective
connected nodes by taking into account the traffic flow. These
distinguishing operational features of DQL and DSDV ensure and productive it could be in several applications. But when it
that DQL-DSDV minimizes the packet lost ratio, congestion, end- comes to the network architecture, interconnectivity, and com-
to-end delay, and communication cost with improved Quality munication of this technology, then various factors need to be
of Service (QoS). During simulations, we observed significant considered and evaluated for its better operation. Because this
technology uses dynamic and wireless communication to col-
Manuscript received 27 July 2023; accepted 17 August 2023. Date of pub-
lication 31 August 2023; date of current version 24 January 2024. This work lect, process, and transmit data in the network. Following these
was supported in part by the Guangdong Provincial Key Laboratory of Human problems, some research efforts have been made in the recent
Digital Twin under Grant 2022B1212010004, and in part by the Endowed past by considering different aspects of the networks, such as
Professorship from Shenzhen Holdfound Foundation. (Corresponding author:
Zhanpeng Jin.) authentication, data privacy, interoperability, device-to-device
Muhammad Adil is with the Department of Computer Science and (D2D) communication, Quality of Services (QoS), and routing
Engineering, University at Buffalo, The State University of New York, protocols, but somehow they are unable to establish network
Buffalo, NY 14260 USA (e-mail: [email protected]).
Muhammad Usman is with the School of Science, Engineering and friendly environment [3], [4], [5].
Information Technology, Federation University, Ballarat, VIC 3350, Australia Chen et al. [6] provided a comprehensive survey of the
(e-mail: [email protected]). existing literature on routing protocols and QoS. They iden-
Mian Ahmad Jan is with the Department of Computer Science, College
of Computing and Informatics, University of Sharjah, Sharjah, UAE (e-mail: tified the limitations of current studies, and highlighted open
[email protected]). research questions awaiting the attention of relevant academic
Hussein Abulkasim is with the College of Engineering and Technology, and industry stakeholders. In addition, they proposed potential
University of Science and Technology of Fujairah, Fujairah, UAE, and also
with the Faculty of Science, New Valley University, El-Kharga 72511, Egypt research solutions that could address some of the outlined chal-
(e-mail: [email protected]). lenges. Kwon et al. [7] extended this discussion and proposed
Ahmed Farouk is with the Department of Computer Science, Faculty an agent-based deep learning approach to improve the com-
of Computers and Artificial Intelligence, South Valley University,
Hurghada 83523, Egypt (e-mail: [email protected]). munication infrastructure of IoT devices in the multipath
Zhanpeng Jin is with the School of Future Technology, South China communication environment. Furthermore, the authors men-
University of Technology, Guangzhou 511442, China, and also with the tioned that their model is useful to enrich the throughput of
Department of Computer Science and Engineering, University at Buffalo,
Buffalo, NY 14260 USA (e-mail: [email protected]). the network. But, from theoretical analysis of this work, we
Digital Object Identifier 10.1109/JIOT.2023.3310927 can observe that this model will create congestion problems,
2327-4662 
c 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
3758 IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO. 3, 1 FEBRUARY 2024

because it only focuses on the throughput rather than the The remainder of this article is organized as follows. In
packet lost ratio followed by latency. For redressal of rout- Section I, we familiarized readers with the concepts of rout-
ing protocol problems in IoT applications, Cong et al. [8] ing protocols and their importance in Extreme Environment
proposed a multioptimality routing criteria protocol (RLR-M). IoT applications. Section II discusses the related work that
In this article, the authors proclaimed that they have considered has been proposed to address the routing and QoS issues in
centralized and distributed environments, while the implemen- these applications. In addition, we highlighted the merits and
tation and analysis phases. Despite the protocol’s advantages, demerits of the present literature to pave the way for our con-
such as supporting device mobility, the authors did not suffi- tribution. Section III introduces the proposed DQL-enabled
ciently focus on communication aspects, which, in our opinion, DSDV prototype by detailing its functionality. Section IV
could hinder its adoption. evaluates the implementation and results of this work, while
Given that, we targeted the routing protocols domain, Section V summarizes and concludes this article.
because it has a direct impact on the interconnectivity and
communication of network-participating nodes. Even though
there exist several solutions in the literature, but most of them II. R ELATED W ORK
focus on the shortest path selection from source to destina- Routing in IoT applications employed in an extreme envi-
tion. With this, they are getting better results for this, but at ronment has emerged as a key research topic due to its
the same time, they somehow ignore important communica- potential to enhance the communication process in line with
tion and computation aspects of IoT networks, which as a consumer market demand. The existing literature presents sev-
result degrades the performance of these networks. To solve eral approaches intended to address the routing concerns in
this problem with the help of routing protocol, in this article, these applications with the objective to improve the QoS
we proposed a deep-Q-learning-enable-destination-sequenced metrics. However, these approaches often fail to balance all
distance-vector (DQL-DSDV) routing protocol. In this model, communication metrics effectively and achieve optimal results.
we ensured dynamic routing with the help of the DSDV and Despite the relentless efforts from the research community and
deep-Q-learning (DQL) algorithms to transmit the data from industry stakeholders to create reliable and communication-
source to destination in a congestion-free environment. friendly environments for these applications, there is still
The major contributions of this work are summarized below. significant work required to meet the future demands of this
1) In the first phase, we defined our network structure, and newborn technology.
connect the participating nodes with the help of DSDV Saha et al. [9] proposed an intelligent routing scheme
routing protocol. Given that, we ensured and checked designed specifically for software-defined network (SDN)
that every node updates their routing table with respect enabled IoT applications to meet their QoS requirements. In
to other nodes. this model, the authors used the features of SDN controllers
2) In the second phase, we employed the DQL algorithm to ensure the traffic management of IoT devices at the edge
to select the optimal hop-count node, using reward func- side of the heterogeneous network. Zhao et al. [10] extended
tions that facilitate priority-based traffic in a congestion- this discussion and developed an application-aware QoS rout-
controlled environment. However, during the training ing algorithm (AQRA) to improve the communication metrics
phase, the agent stores various information that includes of IoT networks. Hasan et al. [11] proposed a dynamic hop
observation state, action state, reward function, and selection routing scheme for IoT networks with the objec-
next state in the replay buffer. Which is then sampled tive to ensure the QoS standards in these networks. In this
for mini-batch information (experience) to ensure the model, the authors primarily focused on the delay sensitivity
optimal reward. of communicating devices. However, they overlooked cru-
3) Furthermore, it is important to highlight that our pro- cial factors, such as PLR, communication, and computation
totype addresses issues, such as network traffic, con- cost—each of which directly impacts the QoS metrics in these
gestion, contention, packet loss ratio (PLR), and delay. networks. Li et al. [12] proposed a routing scheme known as a
While these issues are often overlooked not only in “multiobjective chaotic elite adaptive ant colony optimization
traditional but also in some modified versions of the algorithm” to resolve the security and QoS issues in IoT
DSDV, ad-hoc on-demand distance vector (AODV), and applications. From the operation of this scheme, it is clearly
LEACH protocols. visible that the complex authentication process gives rise to
4) Even though the DQL-DSDV algorithm is model-free, several communication issues, including increased communi-
but it possesses the capability to manage communi- cation cost, latency, and potentially, reduce the throughput.
cation among interconnected nodes within congestion- Despite that, we noticed that the authors only focused on the
controlled environment. Additionally, it utilizes simul- communication after the authentication process, which also
taneous reward functions to further enhance communi- makes this model in a fuzzy state. This is because to the
cation metrics and fulfill the routing requirements of fact that the authentication process has directly connection
Extreme Environment IoT applications. with the network’s communication and QoS metrics, which
5) During experimental analysis, we found that the DQL- consequently influences the performance of the network.
DSDV algorithm outperformed rival protocols across Su et al. [13] proposed a hybrid routing protocol for
various metrics and use case scenarios by demonstrating IoT applications, known as a coding-aware routing protocol
remarkable results. (HCAR). In this scheme, the authors presented an interflow

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
ADIL et al.: IMPROVED CONGESTION-CONTROLLED ROUTING PROTOCOL FOR IoT APPLICATIONS 3759

network coding framework that cleverly combines reactive the most suitable node as the cluster head (CH). Although
routing information with opportunistic routing information the authors claimed the superiority of this framework, it
to effectively managing the network traffic. Similarly, poses a challenge for nodes that have been selected multiple
Nazaralipoorsoomali et al. [14] proposed a multisink rout- times as CHs to participate in the network on par just like
ing protocol for constraint-oriented networks, such as wireless other nodes due to limited resources. Nwadiugwu et al. [23]
sensor networks (WSNs) and IoT networks. This algorithm proposed a two-tier layered network paradigm to enhance
utilizes sink nodes to oversee the routing information of the the packet exchange rate in a constraint-oriented network to
employed devices, thereby enhancing communication met- minimize energy consumption. This work employed CHs to
rics. Liu et al. [15] proposed a localization-free-based routing evaluate and eliminate redundant packets within each clus-
scheme for underwater WSNs (UWSNs). In this protocol, ter, thereby improving traffic statistics in an employed IoT
the authors used the distance-vector information to guide the network. Furthermore, the routing protocols literature has been
devices involved in the context of hop count communication. discussed comprehensively in [24], [25], and [26], where the
Furthermore, the authors claimed that this approach mini- authors raised several open research questions that still need
mized the hop counts for a device while sending data to the response of the concerned stakeholders.
the destination, ensuring reliable data transmission. However,
applying such a routing protocol in a real environment appears III. P ROPOSED M ODEL
problematic as underwater transmissions are susceptible to
In this section, we generalize the collective steps of the
several interferences. Given that, if the hop count is min-
“DQL-DSDV” routing protocol. Before going into the detailed
imized, then the distance between communicating devices
discussion of our proposed model, herein, we would like to
increases, which potentially disrupts the communication pro-
present a flowchart diagram for visual representation. In Fig. 1,
cess due to external interferences. Consequently, the practi-
we have highlighted the steps how our proposed model will
cability of this algorithm for real deployment still remains
work. In the initial phase, we connect all nodes or devices via
unclear.
the DSDV routing protocol. Here, each node updates its rout-
Lu et al. [16] proposed a noncollision theory-based deter-
ing table periodically in coordination with participating nodes
ministic scheduling (NDS) algorithm for IoT applications. The
of the network. To explore, we denote participated nodes as Si ,
primary objective of this algorithm was to ensure ultralow
where the ith term represents the total number of nodes par-
latency in these networks. Chen et al. [17] proposed a novel
taking in the network. Additionally, we use F1 , F2 , F3 , . . . , Fn
two-hop cluster-routing protocol (D2CRP) for IoT applica-
to symbolize route advertisement (RA) information of these
tions to improve the network lifetime. In this model, the
nodes. Continuing this, we initialized the DQL agent with a
authors enabled each node to act as a cluster node when
neural network and a replay buffer. Next, we enabled the DQL
they have sufficient residual energy. Even though the authors
agent to select the next hop count for a source message trans-
claimed reasonable results, but their model is not applica-
mitting toward the destination node. By following the routing
ble in the real-world. Yao et al. [18] proposed a flow-guided
table, the agent estimates the Q-values of different nodes for
opportunistic routing (FGOR) protocol for IoT applications.
the next state based on the reward function to take an action.
Within this framework, the authors utilized the relative posi-
Given that, the agent updates the routing table information
tion (RP) of IoT devices to determine the criteria for candidate
with the specified next hop count node. This process repeats
relaying devices. These devices could serve as an intermedi-
randomly to adapt the changes related to the network’s partici-
ary, allowing other devices to transmit data via them from
pating devices routing information. During the training phase,
the source to the destination. Zier et al. [19] proposed a
the agent stored information, such as the observed state, action
new bio-inspired routing protocol for flying IoT applications.
state, reward function, and next state in the replay buffer. This
In this algorithm, the author used simple additive weight
buffer was then sampled for mini-batch information (experi-
(SAW) algorithm in conjunction with a multicriteria deci-
ence). The information from these samples was used to update
sion making (MCDM) process to ensure edge device traffic
the neural network’s weights according to the DQL algo-
categorization within a communication-friendly environment.
rithm. This process continues until the agent comprehended
Kumar et al. [20] proposed a path selection-based algorithm
an optimal routing policy.
known as the “Path Game Theoretic (PSGame) approach”
to tackle the routing challenges in IoT applications. During
problem formulation, the authors considered the communi- A. DSDV Operational Steps
cation cost in the context of multicast tree transmission, In this section, we discuss how the DSDV routing protocol
articulating it as a potential game to effectively evaluate the has been configured in the proposed model. To elaborate, we
proposed algorithm performance in a simulation environment. considered the same network topology, where Si number of
Chen et al. [21] proposed a federated learning approach nodes were connected. Given that, we assumed two nodes, Sk
known as “FedTor” to address the computation problems that and Sj , separated by a distance d, which is symbolized as d(k,j).
arise with the authentication of IoT devices and improve 1) In the start, we constructed a network of n nodes, and
the performance of these applications. Al-Sadoon et al. [22] connected them via DSDV routing protocol.
proposed a dual-tier clustering routing protocol for mobile 2) Thereafter, we initialized the communication process
WSNs to enhance communication efficiency. This prototype between source node Sk and destination node Sj to
divides the network area into virtual zones while selecting observe the routing table information of Si node.

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
3760 IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO. 3, 1 FEBRUARY 2024

Fig. 1. Flowchart diagram of DQL-DSDV routing algorithm.

3) Next, each node broadcasts a message packet to Algorithm 1 DSDV-Operational Steps

advertise its routing table information as a RA. This Require: Routing table of interconnected nodes.
routing table is generalized with Ri for all Si ∈ Sn Ensure: Periodic routing table updates
nodes. Upon receiving this, Sj node updates its prestored 1: Initialize routing
routing table with respect to Si ’s message packet. 2: Set-up broadcast packet time
4) In the next phase, the route discovery process is initiated, 3: Sk node ⇐ broadcast a RREQ to update it’s R
if the Si node want to communicate or send a packet to 4: Sk initiate communication process with Sj
node Sj node. And there is no entry for Sj in the Sk node 5: Sk RREQ contain ⇐ d(k) and SeqNum(Sk )
routing table or Sk node has a route with a higher metric 6: In response of Sk RREQ
than a new route, then it reinitiates a route discovery 7: Sj sends ⇐ RREP, contain d(j) and SeqNum(Sj )
process by broadcasting a route request (RREQ) packet 8: For (Sk match d(j) and SeqNum(Sj ) in memory
in the network. 9: If (SeqNum(Sj ) ∈
/ (Sk ) R)
5) Likewise, when a node receives the Si node RREQ mes- 10: And d(j) > other adjacent nodes distance
sage to facilitate him for destination Sj node. In this case, 11: Then, Sk choose an adjacent node for next hop
the Si node sends a route reply (RREP) packet with the 12: Else
sequence number, such as (SeqNum(k,j) >= s) to Sk 13: SeqNum(Sj ) ∈ R of Sk and d(j) <= d(adjacent
node. node)
6) Next, in the route maintenance phase: every node checks 14: Sk update it’s R, and the process repeats
and updates its routing table periodically by broadcast- 15: End If
ing a RA packet in the network. 16: End For
7) Finally, the process ensures a loop-free RA information 17: return Current routing information of Sk with connected
in the network. Given that, each node stores the next hop nodes
count address in its routing table. If Si node receives a
packet from intermediary nodes for destination Sj with
a lower metric than the one in the packet, it discards DSDV protocol. And how this combination can improve the
the message and sends a warning to the source node to communication statistics of IoT networks deployed in extreme
prevent potential routing loops. environments, where traditional or newly adopted algorithms
For simplification and easy understanding, we have provided may not work properly. We know that DQL is an advanced ver-
the corresponding pseudocode 1. sion of the Q-Learning algorithm, which employs deep neural
networks (DNNs) to resemble the Q-functions in a model-
B. Deep-Q-Learning Algorithm Operational Steps free environment. Given that, the DQL algorithm estimates
In this section, we summarize the operational steps of the the optimal action-value function (OAVF), which is gener-
DQL algorithm to show how it works in coordination with the ally known as Q-function [Q(st, a)] to calculate the expected

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
ADIL et al.: IMPROVED CONGESTION-CONTROLLED ROUTING PROTOCOL FOR IoT APPLICATIONS 3761

cumulative rewards (r) for an action “a” with state “st” by Algorithm 2 DQL-Framework-Operational Steps
following an optimal policy. To explore, we have summarized Require: Updated Q values.
these steps as below. 1: Initialize Q(st, a; r) with random weight θ
1) Initially, we defined a Deep Q-Network with random 2: update network ⇒ Q (st, a; r) for the next state
weights, θ , to approximate the Q-function as Q(st, a; r). 3: Initialize replay memory buffer (D)
Consequently, we generalized the target network as 4: for each episode of DQL do
Q (st, a; r) with weight θ , and replay memory buffer, 5: Choose the (st) of a node (Sk ) ∈ Si
which is symbolized as (D). 6: while (not terminal st) do
2) Next, the following steps have been adopted for each 7: Take an action a with  greedy policy
episode with the initial state “st.” An action “a” with 8: a = argmaxa Q(st, a; θ ) with probability 1 − 
time-step “t” in a particular episode follows the under- 9: Execute a ⇒ with st and best r
mentioned steps. 10: stores updated information in ⇒ D
3) First, we choosed an action “a” with respect to an - 11: Update mini-batch (B)
greedy policy following the Q values. This policy allows 12: if (B returns ⇒ Sk node R information) then
us to balance between exploration and exploitation (of 13: True
the routing table) based on the computed Q-values, 14: else
which represent the expected future rewards of each 15: False
action. Given that, we select a = arg maxa Q(st, a; r, θ ) 16: end if
with the probability 1 − . Next, we executed an action 17: Update θ with minimized loss function
“a” in the environment (in our case DSDV routing table), 18: Update target network as ⇒ Q (st, a; r)
and observe the reward “r” with (next hop count cost 19: Next update ⇒ st as st’
metrics and traffic statistics) to take an action to the 20: end while
new state (st ). 21: end for
4) Consequently, we stored the tuples (st, a, r, st’) as done 22: return Updated states (in our case routing table with
in the replay buffer (D), where st and st’ denote the respect to new state)
current state and next state, respectively, while the action
and reward is symbolized, such as a and r.
5) Continuing this, we generalized the target Q-values as
yi , for each tuple in the mini-batch (B). Given that, if destination node, while for the Q-values, we considered
yi values ∈ Sk node routing table (R), then, the store the output of each potential next-hop node.
information as True. Else, yi ∈ / Sk node routing table 3) Next, we set up the target network as a Q’(s, a; r) with
(R), then, discard the information as False. defined weight, such as θ  , and replay memory buffer D.
6) Finally, we update the new state as st of the target 4) Periodically, each node broadcast its routing tables
network [Q (st , a; r)]. information in the network, and the immediately con-
The pseudocode for the DQL-algorithm is provided below nected nodes update their prestore registers just like the
(see Algorithm 2). DSDV protocol. Following this, the DQN also updates
its state based on the new routing information, which
is simplified as a tuple of source, intermediary, and
destination node) in our framework.
C. DQL-Enabled-DSDV Collaborative Operation 5) At every iteration within an episode, a node determines
In this section, we discussed how the DSDV routing pro- the next-hop count using an  greedy strategy policy in
tocol and deep Q-learning algorithm were integrated into a the context of current Q-function at a given state.
cooperative platform. The primary goal of this framework is to 6) Next, the DQL-enabled-DSDV executes the chosen next-
enhance next-hop selections based on the accumulated rewards hop selection processes, and observes the reward based
to improve communication metrics. Given that, the suggested on factors, such as the next hop count distance and
prototype enabled the participating nodes to adapt with respect traffic.
to the changes in network topology and maintain efficient 7) Consequently, the concerned node stores the experience
routes from the source to the destination, incorporating fac- information, such as state, action, reward, next state, and
tors, such as communication delay, energy consumption, and terminal state in the replay memory buffer.
traffic statistics. The combined steps of the DSDV and DQL 8) Considering that, the DQL-enabled-DSDV prototype
algorithms are summarized as follows. samples the minibatch experiences from D, and trains
1) We Initialized the DSDV routing tables of each node the DQN using the loss function by following the dif-
with concerned parameters, such as (destination node, ference between current Q-values and target Q-values,
next-hop count, number of hop counts, and sequence just like the traditional deep Q-learning algorithm.
number). 9) Furthermore, this prototype ensures the periodic update
2) Thereafter, we initialized the DQN algorithm at each of the target network with weights θ  = τ * θ + (1 − τ ).
node routing table with random weights θ . For the input Herein, θ represent the weights of the primary network,
of the DQN, we used the current state of the source and while θ  demonstrate the weights of the target network,

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
3762 IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO. 3, 1 FEBRUARY 2024

Fig. 2. Visual representation of DQL-enabled-DSDV protocol operation steps.

and τ is a small constant that controls the rate at which to all vicinity connected nodes. Salah et al. [28] presented an
the target network tracks the primary network. advanced mathematical model of the DSDV protocol to enrich
10) The process repeats for 4-to-9 steps until the routing the communication statistic of IoT networks. In this proto-
tables and DQN models converge, or meet predefined type, the author used the probability density function to find
criteria. the best approximation values for next count selection in real
For superficial evaluation and visual representation, we add scenarios. Although, the authors claimed significant contribu-
Fig. 2 in this article. tions, but the proposed model is computationally expensive
and complex. Because it updates the routing information iter-
IV. P ERFORMANCE E VALUATION atively based on the density function, which consume extra
In this section, we talked about the results of DQL- energy resources with some computation complexities. Which
enabled-DSDV in the context of comparative metrics, when as a result degrade the performance of the network.
a node Sk need to send data/message to node Sj follow- Chen et al. [17] proposed a novel distributed D2CRP for
ing multihop communication. For comparative evaluation, we WSN and IoT networks to improve the network lifetime and
consider the advanced versions of AODV [27], DSDV [28], communication metrics. In this framework, the authors used
distributed 2-hop cluster-routing protocol (D2CRP) [17], and a two-hop distance for CH selection. During the evaluation
multihop-LEACH (M-LEACH) protocol [29]. It is evident that phase, the authors claimed that the proposed protocol ensures
traditional versions of these protocols have been enhanced the minimum energy consumption of the participating node.
to optimize communication metrics for resource-constrained But it has not been checked, when a node plays the role of
networks. However, most of them focus on one commu- CH, then how much energy he consumes. Will this affect the
nication aspect, and ignore the other aspects, which as a upcoming operation of that particular node? Despite that, does
whole degrade the network performance in terms of QoS this node accommodate all vicinity nodes, if not, then how we
metrics. can address this issue? Moreover, the authors mentioned that
Chavan and Venkataram [27] modified the classical AODV the CH nodes communicate in a chain till the BS. During this
protocol in the context of a multicast communication process, it is not clear how much distance was set for intra-
paradigm. In this article, the authors reported promising CH communication. In addition, is this protocol capable of
results, but we are wondering about the operation of this pro- ensuring reliable and transparent CH selection in real-world
totype, because they used multicast communication, which deployment? These concerns introduce ambiguities that casts
means a specific group of sensors. This becomes problem- some doubt on the reliability of the work. Mohapatra et al. [29]
atic when considering multipath and multihop communica- modified the existing LEACH protocol to multipath LEACH
tion, while ensuring reliable and cost-effective communication protocols. In this prototype, the authors targeted the resid-
among interconnected nodes. Because of the use of multicast ual energy of IoT nodes for data transmission from source to
communication, instead of broadcast, the RAs are not sent destination.

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
ADIL et al.: IMPROVED CONGESTION-CONTROLLED ROUTING PROTOCOL FOR IoT APPLICATIONS 3763

A. Simulation Set Up Likewise, we check the results statistics of [17] and [29]
In this section, we talked about how the proposed prototype for the considered scenarios. Given that, both of these
has been checked and verified. We first set up an IoT network references follow CH nodes for the transmission of data
to produce a routing table. Using this table, we then applied the from source to destination. However, during the CH selec-
Q-learning algorithm to choose the hop count that delivers the tion, these techniques consume unnecessary energy, which
maximum reward function. Consequently, we enabled commu- as a result increased transmission delay, and end-to-end
nication between two nodes, source (Sk ) and destination (Sj ), delay. Collectively, these factors adversely affect the over-
utilizing a 7-hop communication infrastructure. Considering all communication metrics’ in operational networks. To
that, node Sk must traverse at least seven hops to transmit explain this, let us assume a network participating node
data to the destination node Sj within the network. With this, advertises his candidacy of CH in the network. During
we defined energy consumption parameters for each message this process, it first broadcasts a “route candidacy request”
transmission and reception by following the energy consump- (RCREQ) which nearby nodes receives. Subsequently, these
tion model discussed in the upcoming sections. This enables nodes assess the distance and other pertinent criteria to
us to calculate the total energy consumption throughout the either accept or reject the RCREQ of this node. Based
communication process. Similarly, by setting the transmis- on defined parameters, they make a decision, but this pro-
sion time for each packet, we can compute the end-to-end cess involves some calculations, which as a result degrade
delay for transmitted messages. To validate the efficiency of the network performance. Despite that, another node in the
our prototype against competing protocols, we ensured that network concurrently sends its own RCREQ for CH can-
each node selects the next hop based on the reward function. didacy. This action further consumes resources in terms of
This function primarily incorporates distance parameters and computation and communication, which as a result affects the
traffic statistics, which helps to ensure delay-sensitive in the network’s lifetime, transmission delay, and end-to-end (E-2-E)
network. delay.
1) Energy Consumption: In this segment, we discussed that
how the proposed prototype has been evaluated for the energy
B. End-to-End Delay Results Statistics consumption of IoT nodes. Given that, we started with a typi-
In this section, we talked about the transmission delay, cal IoT node, which require three fundamental components for
propagation delay (Pdelay ), and round trip time (RTT) of mes- successful operation: 1) a sensing module to gather data; 2) a
sage sessions. Given that, we underline the advantages of our processor to manage both the sensor and the communication
prototype over competing protocols traditionally used in IoT module; and 3) a communication module to handle data trans-
applications to address the network traffic issues. Moreover, mission and reception within the network. Following this, we
we observed the simulation statistics of communicating nodes examined a scenario involving a transmitted message, sym-
to see the aforementioned metrics in the presence of existing bolized as Txm . This message comprises a packet (pkt) size
protocols, such as [27] and [28]. From [27] and [28], we can of 512 bits, inclusive of a packet overhead length, which is
see that, if a node Sk wants to send data to Sj , it strictly fol- symbolized as Nmpkt . The time required for the transmission
lows its routing table information, which is updated statically of this packet depends on the channel capacity and can be
and dynamically. However, these nodes neglect to consider formulated by the Shannon model [30] as follows:
optimal paths with maximum rewards immediately after their Nmpkt
routing table updates. Such an oversight can compromise com- Txpkt = . (1)
BW log2 (1 + )
munication attributes, which as a result lead to the network
performance degradation. In (1), BW and γ represent the bandwidth of the channel
For explanation, let us consider a scenario in which node Sk and the signal to interference and noise ratio (SINR), respec-
aims to transmit data to Sj . In doing so, Sk relies on its routing tively, of the connected IoT nodes. Furthermore,  has been
table information to determine the next hop for communication generalized as follows:
with node Sj . Notably, this selection process requires message RxPow
exchanges between Sk and the relaying node, as highlighted = . (2)
RIPow + Npow
in [27] and [28]. This message exchange introduces addi-
tional end-to-end delays and energy consumption, which as In (2), RIPow symbolizes the random interference power of an
a result negatively impacts the overall network performance. IoT node, while Npow and RxPow represents the measured noise
From the acquired results of our prototype, we observed that power and the received power, respectively. The RxPow can be
Sk utilizes a varying number of hops to transmit to Sj when calculated as follows:
employing AODV, DSDV, and DQL-enabled-DSDV, respec- RxPow = Lloss × PTx . (3)
tively. Notably, we fixed the each node’s transmission power
dissipation TTXp at 0.03 mW, and the transmission delay τdelay In (3), PTx represents the transmission power of Sk ∈ Sn−1 ,
to 0.005 msec. Utilizing these parameters, we derived the and Lloss symbolizes the loss factor between communicating
average end-to-end (E2E) delay statistics, as illustrated in nodes. This is further generalized using a log-distance path-
Fig. 3(b). In addition, we evaluated the energy consumption loss model, such as Lloss = Lconst × hrd−α [29]. In this model,
(shown in Fig. 3(a) and the (RTT - averaged for all sessions, Lconst serves as a constant that accounts for antenna gain and
presented in Fig. 3(c). frequency. The variable “h” is assumed to be a random variable

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
3764 IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO. 3, 1 FEBRUARY 2024

Fig. 3. Statistical analysis of (a) energy consumption, (b) end-to-end delay, and (c) session RTT.

accounting for channel fading, while “α” denotes the path- follows:
loss exponent, and “rd ” is the contact distance between Sk
and Sj . Considering this, the total energy consumed by Sk , Si × RREQhop = α × hop × Pkt × Sj × RREPhop . (6)
while transmitting a fixed-length message can be calculated And
as (MsgTE )
  α × hop × Pkt × Sj × RREPhop = α × hop × Pkt. (7)
MsgTE = Cf (Pt ) + Po Tm . (4)
In (6) and (7) pkt represents the packet size, which is
In (4), Cf represents the conversion power of the ampli- assumed 512 bytes in our case
fier and radio, while Po demonstrates the electronic power 
7
 
consumption of the communication module while encoding Total-Cost = CostRREQ (hop) + CostRREP (hop) . (8)
the message [30]. The aforementioned relationship has been hop=1
substituted to generate a concise formula representing the
After the cost functions substitution in the derived equation,
energy consumption model for extreme area deployed IoT
we will get
nodes
Cf L(PI + PN )γ + Po 
7
 
MsgTE = . (5) Total-Cost = (α × hop × Pkt) + (β × hop × Pkt) . (9)
Nm · pkt · BW · log2 (1 + ) hop=1

From (5), we can observe that the targeted SINR  is a free To simply (9), we factor out the common term p
variable, which can be controlled, while the other parame-
ters should be managed through hardware and the channel 
7
 
Total-Cost = P × (α + β) × hop . (10)
set-up in operational networks. Furthermore, we employed
hop=1
this paradigm to evaluate our proposed prototype in compar-
ison to other schemes by showcasing the results obtained in Next, we have summed up the weighted hop counts all hops
Table I. To summarize, we have noted in simulation that [27] from 1 to 7, which is generalized as follows:
consumes 1 × 6 time extra energy during the initiation of Total-Cost = Pkt × (α + β) × (1 + 2 + 3 + 4 + 5 + 6 + 7)
the communication process (only in the CH candidacy phase,
despite facilitating the communication process), while [28] Total-Cost = Pkt × (α + β) × 28. (11)
eradicates 1 × 5 energy than of our model while adver- We used this equation to calculate the communication cost of
tising multipath information in the network. Similarly, both proposed prototype and comparative protocols. Table II sum-
of these schemes have 1 × 5 times and 1 × 4 times maries the results statistics of our framework followed by the
extra E-2-E delay, while 1 × 4 and 1 × 3 transmission comparative frameworks.
delay.
D. Packet Lost Ratio Result Statistics
C. Communication Cost Result Statistics In this section, we discuss the PLR that we observed during
In this section, we talked about, that after bounding the our simulation. Our focus was primarily on the communi-
proposed prototype and comparative protocols to follow at cation between a source node Sk and a destination node Sj
least 7 hop counts, while transmitting data from source Si communication aspects. To carry out the simulation, we initi-
to the destination Sj . Given that, we set the packet size as ated random traffic among the interconnected nodes with the
512 bytes for each RREQ and RREP, which are generalized objective to assess the volume of packets that were success-
as α bytes/hop and β bytes/hop, respectively. Following the fully sent from a source node Sk to the destination node Sj .
7 hop count communication paradigm of a Si , we defined the Additionally, we sought to understand the number of packets
cost function for the route request and route reply packets as that were lost during the communication process. By doing

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
ADIL et al.: IMPROVED CONGESTION-CONTROLLED ROUTING PROTOCOL FOR IoT APPLICATIONS 3765

TABLE I
E NERGY C ONSUMPTION , E-2-E, AND RRT R ESULT S TATISTICS

TABLE II
C OMMUNICATION C OST R ESULT S TATISTICS

TABLE III
T IME C OMPLEXITY A NALYSIS

this, we were able to provide an analysis of the efficiency of

data transmission within the network. The simulation results
obtained within the set time frame for communication ini-
tiation and termination processes provided us a quantifiable
measure of network performance and reliability. Specifically,
we assessed the communication from Sk to Sj over the time
intervals of 10, 20, 30, 40, and 50 min, respectively, and
measured the PLR under random conditions. In addition, dur-
ing this process, we monitored the number of packets that
have reached successfully to the Sj and those that were lost.
This was achieved by considering the acknowledgment (ACK)
packets of Sj at Sk . The results obtained during the sim-
ulation process are visualized and can be seen in Fig. 4.
This visual representation of PLR offers a clear and imme-
diate understanding of the performance of the system under
observation.
Fig. 4. Comparative statistical analysis of PLR.
E. Time Complexity Analysis
For analyzing time complexity results, we considered the
is (nk+c) and (n2 ) for the aforestated two use cases, where
same scenario as of the communication costs of the hop-
c additionally represents the number of clusters, when Sk node
count communication with session initiation and establish-
sends data to Sj node. Given that, we have added Table III in
ment. Moreover, we posited the best and worst use-case time
the manuscript to provide a summary of the extracted results
complexities ( T) as [n + (k × s)] and (n2 + (k × s) + c),
in a comparative format.
respectively. In these formulas, “n” represents the number of
nodes in the network, while “k,” “s,” and “c” denotes the num-
ber of hop counts, sessions, and CHs, respectively, when data V. C ONCLUSION
is transmitted from node Sk to node Sj . In the case of [27], [28], In this article, we proposed a hybrid routing protocol known
and [17], the worst and best use cases are the same, but the as the DQL-enabled-DSDV protocol. In this prototype, we
traffic generation and path selection are different, which affects combined the DSDV and DQL algorithms to improve the route
the T of them. Likewise, Chavan and Venkataram [27], T selection process followed by communication statistics of IoT

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
3766 IEEE INTERNET OF THINGS JOURNAL, VOL. 11, NO. 3, 1 FEBRUARY 2024

networks employed in extreme environments. Initially, we gen- [12] C. Li, Y. Liu, J. Xiao, and J. Zhou, “MCEAACO-QSRP:
erated the routing table of connected nodes through DSDV A novel QoS-secure routing protocol for Industrial Internet of
Things,” IEEE Internet Things J., vol. 9, no. 19, pp. 18760–18777,
protocols. While in the second phase, we used this information Oct. 2022.
to calculate the next hop count based on reward functions [13] Y. Su, Y. Xu, Z. Pang, Y. Kang, and R. Fan, “HCAR: A
(Q-values) utilizing the Q-learning algorithm. Unlike the tra- hybrid-coding-aware routing protocol for underwater acoustic sensor
networks,” IEEE Internet Things J., vol. 10, no. 12, pp. 10790–10801,
ditional DSDV protocol, in our prototype, each node selects Jun. 2023.
the next hop count based on the calculated reward, which [14] M. Nazaralipoorsoomali, P. Asghari, and S. H. H. S. Javadi,
ensures the best candidate node selection with the congestion- “Performance improvement of routing protocol for low-power and
lossy networks protocol in an Internet of Things-based smart
controlled path in the network to facilitate delay-sensitive com- retail system,” Int. J. Commun. Syst., vol. 35, no. 10, 2022,
munication. During our simulations, we observed noteworthy Art. no. e5166.
contributions of the proposed prototype. Notably, we compared [15] Z. Liu, X. Jin, Y. Yang, K. Ma, and X. Guan, “Energy-
efficient guiding-network-based routing for underwater wireless sensor
the performance metrics with competing routing protocols and networks,” IEEE Internet Things J., vol. 9, no. 21, pp. 21702–21711,
found remarkable improvements in communication cost, end- Nov. 2022.
to-end delay, round-trip time, and energy consumption within [16] Y. Lu et al., “An intelligent deterministic scheduling method for
ultralow latency communication in edge enabled Industrial Internet of
the operational network. Based on these results, we are confi- Things,” IEEE Trans. Ind. Informat., vol. 19, no. 2, pp. 1756–1767,
dent that the proposed prototype would greatly benefit Extreme Feb. 2023.
Environment IoT applications, especially in scenarios where [17] C. Chen, L. C. Wang, and C. M. Yu, “D2CRP: A novel distributed 2-
hop cluster routing protocol for wireless sensor networks,” IEEE Internet
human access is impossible but delay-sensitive communica- Things J., vol. 9, no. 20, pp. 19575–19588, Oct. 2022.
tion is essential. Furthermore, we believe that the effective [18] X. W. Yao, Y. W. Chen, Y. Wu, K. Zhao, and J. M. Jornet,
performance of this prototype can significantly enhance the “FGOR: Flow-guided opportunistic routing for intrabody nanonet-
works,” IEEE Internet Things J., vol. 9, no. 21, pp. 21765–21776,
productivity of any IoT applications installed in Extreme Nov. 2022.
environments. [19] A. Zier, A. Abouaissa, and P. Lorenz, “FIRP: Firefly inspired rout-
ing protocol for future Internet of Things,” in Proc. IEEE Int. Conf.
Commun., May 2022, pp. 2948–2953.
R EFERENCES [20] S. Kumar, A. Goswami, R. Gupta, S. P. Singh, and A. Lay-Ekuakille,
“A game-theoretic approach for cost-effective multicast routing in
[1] X. Li, X. Hu, R. Zhang, and L. Yang, “Routing protocol design for the Internet of Things,” IEEE Internet Things J., vol. 9, no. 18,
underwater optical wireless sensor networks: A multiagent reinforce- pp. 18041–18053, Sep. 2022.
ment learning approach,” IEEE Internet Things J., vol. 7, no. 10, [21] Y. Chen, Y. Su, M. Zhang, H. Chai, Y. Wei, and S. Yu, “FedTor:
pp. 9805–9818, Oct. 2020. An Anonymous Framework of Federated Learning in Internet of
[2] M. Adil, M. A. Jan, Y. Liu, H. Abulkasim, A. Farouk, and H. Song, “A Things,” IEEE Internet Things J., vol. 9, no. 19, pp. 18620–18631,
systematic survey: Security threats to UAV-aided IoT applications, tax- Oct. 2022.
onomy, current challenges and requirements with future research direc- [22] Al-M. E. Sadoon, A. Jedidi, and H. Al-Raweshidy, “Dual-tier cluster-
tions,” IEEE Trans. Intell. Transp. Syst., vol. 24, no. 2, pp. 1437–1455, based routing in mobile wireless sensor network for IoT applica-
Feb. 2023. tion,” IEEE Access, vol. 11, pp. 4079–4094, 2023.
[3] M. S. Frikha, S. M. Gammar, A. Lahmadi, and L. Andrey, [23] W. P. Nwadiugwu, W. Ejaz, M. Kaneko, and A. Anpalagan, “Neural-
“Reinforcement and deep reinforcement learning for wireless Internet network-assisted packet accelerators for Internet of Things network
of Things: A survey,” Comput. Commun., vol. 178, pp. 98–113, systems,” IEEE Internet Things J., vol. 10, no. 17, pp. 15238–15251,
Oct. 2021. Sep. 2023.
[4] M. Adil, H. Alshahrani, A. Rajab, A. Shaikh, H. Song, and A. Farouk, [24] M. Hosseinzadeh, A. Hemmati, and A. M. Rahmani, “Clustering for
“QoS review: Smart sensing in wake of COVID-19, current trends and smart cities in the Internet of Things: A review,” Clust. Comput., vol. 25,
specifications with future research directions,” IEEE Sensors J., vol. 23, no. 6, pp. 4097–4127, 2022.
no. 2, pp. 865–876, Jan. 2023. [25] A. Malik, B. Bhushan, A. Kumar, and R. Chaganti, “Opportunistic
[5] M. Adil, M. K. Khan, M. Jamjoom, and A. Farouk, “MHADBOR: Internet of Things (OIoT): Elucidating the active opportunities of oppor-
AI-enabled administrative-distance-based opportunistic load balancing tunistic networks on the way to IoT,” in New Trends and Applications
scheme for an agriculture Internet of Things network,” IEEE Micro, in Internet of Things (IoT) and Big Data Analytics. Cham, Switzerland:
vol. 42, no. 1, pp. 41–50, Jan./Feb. 2022. Springer Int., 2022, pp. 209–224.
[6] W. Chen, X. Qiu, T. Cai, H. N. Dai, Z. Zheng, and Y. Zhang, “Deep [26] A. Zrelli, “Hardware, software platforms, operating systems and routing
reinforcement learning for Internet of Things: A comprehensive sur- protocols for Internet of Things applications,” Wireless Pers. Commun.,
vey,” IEEE Commun. Surveys Tuts., vol. 23, no. 3, pp. 1659–1692, vol. 122, no. 4, pp. 3889–3912, 2022.
3rd Quart., 2021. [27] C. P. Chavan and P. Venkataram, “Design and implementation of event-
[7] M. Kwon, J. Lee, and H. Park, “Intelligent IoT connectivity: Deep based multicast AODV routing protocol for ubiquitous network,” Array,
reinforcement learning approach,” IEEE Sensors J., vol. 20, no. 5, vol. 14, Jul. 2022, Art. no. 100129.
pp. 2782–2791, Mar. 2020. [28] S. Salah, R. Zaghal, and M. Abdeljawad, “A mathematical-based
[8] P. Cong et al., “A deep reinforcement learning-based multi-optimality model for estimating the path duration of the DSDV routing pro-
routing scheme for dynamic IoT networks,” Comput. Netw., vol. 192, tocol in MANETs,” J. Sens. Actuator Netw., vol. 11, no. 2,
Jun. 2021, Art. no. 108057. p. 23, 2022.
[9] N. Saha, S. Bera, and S. Misra, “Sway: Traffic-aware QoS routing in [29] S. Mohapatra, P. K. Behera, P. K. Sahoo, S. K. Bisoy, K. L. Hui,
software-defined IoT,” IEEE Trans. Emerg. Topics Comput., vol. 9, no. 1, and M. Sain, “Mobility induced multi-hop LEACH protocol in hetero-
pp. 390–401, Jan.–Mar. 2021. geneous mobile network,” IEEE Access, vol. 10, pp. 132895–132907,
[10] D. Zhao, Q. Zou, and M. B. Zadeh, “A QoS-aware IoT ser- 2022.
vice placement mechanism in fog computing based on open- [30] D. Chandler. “The transmission model of communication.” 1994.
source development model,” J. Grid Comput., vol. 20, no. 2, Accessed: Jul. 24, 2013. [Online]. Available: http://www.aber.ac.uk/
p. 12, 2022. media/Documents/short/trans.html
[11] N. Hasan, A. K. Ray, and A. Mishra, “Analysis of hybrid MAC protocols [31] A. Al-Hourani, S. Kandeepan, and E. Hossain, “Relay-assisted
in vehicular ad hoc networks (VANET) for QoS sensitive IoT appli- node-to-node communication: A stochastic analysis of energy sav-
cations,” in Proc. Comput. Netw. Inventive Commun. Technol., 2021, ing,” IEEE Trans. Mobile Comput., vol. 15, no. 12, pp. 3129–3141,
pp. 753–766. Dec. 2016.

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.
ADIL et al.: IMPROVED CONGESTION-CONTROLLED ROUTING PROTOCOL FOR IoT APPLICATIONS 3767

Muhammad Adil (Member IEEE) received the Hussein Abulkasim (Senior Member, IEEE)
B.S. and M.S. degrees in computer science from received the B.Sc., M.Sc., and Ph.D. degrees in com-
the Virtual University of Lahore, Lahore, Pakistan, puter science from South Valley University, Qena,
in 2017 and 2020, respectively. He is currently Egypt, in 2004, 2012, and 2016, respectively.
pursuing the Ph.D. degree with the Department From 2006 to 2011, he was a Web Developer
of Computer Science and Engineering, University with the Center for Information and Communication
at Buffalo, The State University of New York, Technology, South Valley University. From 2012 to
Buffalo, NY, USA. 2014, he was with the College of Computer Sciences
He has CCNA and CCNP Certifications. and Information Systems, Jazan University, Jizan,
He has many publications in prestigious jour- Saudi Arabia. From 2019 to 2021, he worked as
nals, such as IEEE I NTERNET OF T HINGS, a Postdoctoral Researcher with the Cybersecurity
IEEE T RANSACTIONS OF I NTELLIGENT T RANSPORTATION, IEEE Research Laboratory, Toronto Metropolitan University, Toronto, ON, Canada.
T RANSACTIONS ON I NDUSTRIAL I NFORMATICS, IEEE T RANSACTIONS He is currently working as an Assistant Professor of Computer Science
ON N ETWORK S CIENCE AND E NGINEERING , IEEE T RANSACTIONS with the Department of Mathematics and Computer Science, New Valley
ON I NTELLIGENT V EHICLES , IEEE S ENSORS J OURNAL , IEEE Network University, Kharga, Egypt. His current research interests include quantum
Magazine, IEEE Micro Magazine, IEEE Consumer Electronics Magazine, cryptography, cybersecurity, the IoT security, blockchain security quantum
ACM Transactions on Sensor Networks, Computer Networks (Elsevier), and computation and communication, blockchain technology, and the IoT security.
Sustainable Cities and Societies. His research interest includes networking,
cybersecurity, cyber-physical systems, unnamed aerial vehicles, Internet of
Things, and wireless sensor networks.
Mr. Adil received the Chair’s Fellowship from the Department of
Computer Science and Engineering, University at Buffalo, The State Ahmed Farouk (Senior Member, IEEE) is currently
University of New York, in 2022. He is reviewing for prestigious jour- an Assistant Professor with the Department of
nals, such as IEEE I NTERNET OF T HINGS J OURNAL, IEEE S ENSORS, Computer Science, Faculty of Computers and
IEEE S YSTEMS J OURNAL, IEEE T RANSACTIONS ON N ETWORK Artificial Intelligence, South Valley University,
S CIENCE AND E NGINEERING, IEEE T RANSACTIONS ON A RTIFICIAL Hurghada, Egypt. Before that he was a Postdoctoral
I NTELLIGENCE, IEEE T RANSACTIONS ON I NDUSTRIAL I NFORMATICS, Research Fellow with Wilfrid Laurier University,
IEEE T RANSACTIONS ON C OGNITIVE C OMMUNICATIONS AND Waterloo, ON, Canada, and Ryerson University,
N ETWORKING, IEEE T RANSACTIONS ON I NTELLIGENT T RANSPORTATION Toronto, ON, Canada. He is one of the Top
S YSTEMS, IEEE T RANSACTIONS ON G REEN C OMMUNICATIONS AND 20 Technical Co-Founders of the Quantum
N ETWORKING, and IEEE T RANSACTIONS ON S YSTEMS , M AN , AND Machine Learning Program by Creative Destruction
C YBERNETICS. In addition, he is a member of the IEEE Computer Society, Laboratory, University of Toronto. He has authored
IEEE Industrial Electronics, IEEE Cybersecurity, IEEE Young Professionals, or coauthored 62 articles in reputed and high-impact journals, such as
and London Journal Press Club—U.K., as an Honorary Member. Nature Scientific Reports and Physical Review. The exceptional quality of
his research is recognized nationally and internationally.
Dr. Farouk is also selected as Top 25 of Innovate to 150 Canada to show-
case the best of Toronto’s next generation of change-makers, innovators, and
entrepreneurs. He is exceptionally well-known for his seminal contributions
to theories of quantum information, communication, and cryptography. He
was selected by the scientific review panel of the Council for the Lindau
Nobel Laureate Meetings to participate in the 70th Lindau Nobel Laureate
Meeting. His volunteering work is apparent since he was appointed as the
Muhammad Usman (Member IEEE) received the Chair of the IEEE Computer Chapter for the Waterloo-Kitchener area and
Ph.D. degree from the University of Technology editorial board for many reputed journals, such as Nature Scientific Reports,
Sydney, Ultimo, NSW, Australia. IET Quantum Communication, and IEEE ACCESS. He is also selected for
He joined Federation University, Ballarat, VIC, IEEE and IET Young Professional Ambassador and as a Moderator for the
Australia, in 2019 as a Lecturer of Information new IEEE TechRxiv. He is currently an Associate Editor of IEEE Canadian
Technology. Prior to this, he has worked as a Review.
Postdoctoral Researcher with Swinburne University
of Technology, Melbourne, VIC, Australia. He has
also worked with the University of Technology
Sydney as a Research and Teaching Assistant. His
primary research interests are in the areas of cyber-
security (security and privacy), machine learning, QoS, QoE, video processing
and streaming, and Internet of Multimedia Things.
Zhanpeng Jin (Senior Member, IEEE) received
the Ph.D. degree in electrical engineering from the
University of Pittsburgh, Pittsburgh, PA, USA.
He is currently a Xinshi Endowed Professor
with the School of Future Technology, South China
University of Technology, Guangzhou, China, and
also an Associate Professor with the Department of
Computer Science and Engineering, University at
Mian Ahmad Jan (Senior Member, IEEE) received Buffalo, Buffalo, NY, USA. He was an Associate
the Ph.D. degree from the University of Technology Professor of Electrical and Computer Engineering
Sydney, Ultimo, NSW, Australia, and the master’s and Biomedical Engineering with Binghamton
degree in mobile computing from the University of University, Binghamton, NY, USA, as well as a Postdoctoral Research
Bradford, Bradford, U.K. Associate with the University of Illinois at Urbana–Champaign, Champaign,
He is currently an Assistant Professor with the IL, USA. He has published over 100 papers in international journals and
Department of Computer Science, Abdul Wali Khan conferences. His research interests include ubiquitous computing, human–
University Mardan, Mardan, Pakistan. He has pub- computer interaction, and AI-powered smart health and smart home.
lished his research in the world leading journals and Dr. Jin received the IEEE Region 1 Technological Innovation Award. His
conferences and has been a guest editor of numer- pioneer work in brain biometrics was selected and named as one of the “Future
ous special issues in various prestigious journals. His Technology: 22 Ideas About to Change Our World” by BBC Science Focus.
research interests include cyber security, energy efficient and secured com- He serves as an Associate Editor for the following journals: ACM Computing
munication for wireless sensor networks, and the Internet of Things. Surveys, Proceedings of the ACM on Interactive, Mobile, Wearable and
Dr. Jan has been among the World’s Top 2 2021 and 2022, respectively. Ubiquitous Technologies, and Computers in Biology and Medicine (Elsevier).
He has been the general chair of various highranked conferences. He is a Senior Member of ACM.

Authorized licensed use limited to: Kumaraguru College of Technology. Downloaded on April 05,2024 at 07:26:58 UTC from IEEE Xplore. Restrictions apply.