Computer Networks 229 (2023) 109785

Contents lists available at ScienceDirect

Computer Networks
journal homepage: www.elsevier.com/locate/comnet

A reliable position-based routing scheme for controlling excessive data

dissemination in vehicular ad-hoc networks
Zainab H. Ali a, *, Noha A. Sakr b, Nora El-Rashidy c, Hesham A. Ali d
a
Department of Embedded Network Systems and Technology, Faculty of Artificial Intelligence, Kafrelsheikh University, El-Geish st, Kafrelsheikh 33516, Egypt
b
Department of Computer Engineering and Control Systems, Faculty of Engineering, Mansoura University, Eldakahlia, Mansoura 33511, Egypt
c
Department of Machine Learning and Information Retrieval, Faculty of Artificial Intelligence, Kafrelsheikh University, El-Geish st, Kafrelsheikh 33516, Egypt
d
Department of Computer Engineering and Control Systems, Faculty of Engineering, Mansoura University, Egypt- Faculty Artificial Intelligence, Delta University for
Science and Technology, Gamasa 35712, Egypt

A R T I C L E I N F O A B S T R A C T

Keywords: In the past two decades, the automotive industry has undergone tremendous changes, as this field has become
Computing platform one of the fastest developing and growing fields, especially with the progress of global digitization and massive
Position-based routing research related to networks. So, it comes as no surprise that self-driving vehicles are ubiquitous mainly relying
Energy conservation
on Vehicular ad-hoc networks (VANETs). This transformation ensures improved road navigation and traffic
SDN
congestion avoidance by relying on Rapid Data Deployment (DD). On the other hand, although DD achieves high
Data Dissemination
VANETs connection reliability, it may affect network bandwidth and performance. In addition, excessive DD causes
frequent link outages resulting in reduced data delivery, massive packet loss, and premature end of network life.
This paper presents an integrated architecture proposal for a vehicle dynamic assistance architecture, based on
reliable methods to ensure steering accuracy while minimizing the energy expended for controlling the DD rate
via VANETs. The proposed architecture integrates between Software-Defined Networks (SDN) and fog computing
based on dealing with the mobility factors that exploit vehicle communication behaviors. Such integration will
aid in improving network performance in terms of packet delivery and DD. The study also discusses how to take
into consideration the Euclidean distance, geographical routing information, residual power ratio, and latency
time to maximize network stability and avoid possible link disruption. The simulation results prove that there is a
62% to 70% enhancement of the whole power consumption and network throughput, depending on the
implementation of the proposed position-based routing approach. Interestingly, the proposed routing protocol is
a dual-phase routing protocol with a 90% of SDN data packet delivery ratio and an 82% of SDN data loss
reduction. So, when the SDN fails to deliver packets, the proposed position-based routing handles them as a
parallel mechanism of SDN.

1. Introduction many automotive assistance services to meet the demands of passengers

and the constraints of Quality of Service (QoS), it may not be able to
Smart Transportation (ST) is a promising technology, in developing provide a comprehensive view of the nodes’ behavior due to the
countries with shallow road capacity, poor economic resources, imper­ fast-frequently fragmentation of its topology. Additionally, vehicles’
fect motorization, and insufficient infrastructure. ST allows vehicles to lack of transmission dependability has a negative impact on the rate of
increase dynamic real-time interaction through the propagation of road network lifetime and overall performance [4,5].
information among motorcars to assist drivers and enhance the quality Excessive data transmission across VANET to deliver effective
of traffic journey [1]. Vehicular Ad-Hoc Network (VANET) is a foun­ communication causes notable deterioration in lifetime and network
dation of ST, it allows the spread of information about road safety performance [6]. The nodes in the communication process are
among vehicles via emergency message transmission to aid drivers in commonly establishing a uniform-path connection through the sender
taking early preventive measures during hazardous troubles to avert and the receiver which causes the transmission journey to be difficult in
accidents and enhance road protection [2,3]. Although VANET offers VANETS because of the high dynamic moving of vehicles in VANET

* Corresponding author.
E-mail address: zainabhassan@ai.kfs.edu.eg (Z.H. Ali).

https://doi.org/10.1016/j.comnet.2023.109785
Received 29 September 2022; Received in revised form 21 March 2023; Accepted 17 April 2023
Available online 19 April 2023
1389-1286/© 2023 Elsevier B.V. All rights reserved.
Z.H. Ali et al. Computer Networks 229 (2023) 109785

therefore the determination of the optimal next-hop is not guaranteed improve the performance of the position-based routing protocol in
for all situations [7,8]. A topology-based routing has generally inno­ reducing rate of DD and obtain energy conservation. Moreover, the
vated to supply the best route for data gathering taking into consider­ proposed routing protocol is a dual-phase routing protocol with a 90% of
ation power conservation [2,9,10]. Recently, research on routing can be SDN data packet delivery ratio and an 82% of SDN data loss reduction.
classified into two categories. First, topology-based routing that works So, when the SDN fails to deliver packets, the proposed position-based
based on the information collected about link status and the packet routing handles them as a parallel mechanism of SDN. The paper con­
forwarding method such as Source-Tree Adaptive Routing (STAR) pro­ tributions can be organized as follows:
tocol, Optimized Link-State Routing (OLSR), and Ad-hoc vehicle
On-demand Distance Vector routing protocol (AODV). Second, i. Demonstrating a dynamic vehicular-assisted architecture based
position-based routing or geographical routing such as Fisheye State on SDN and fog computing to improve message reliability in
Routing (FSR) and the Temporally Ordered Routing Algorithm (TORA) VANET.
which are concerned with geographical metrics like vehicle position and ii. Presenting seamless collaboration between the SDN and fog
neighbor direction with no demand to carry on a static route or locating computing to provide a comprehensive view helps to reduce DD
the best path between sender and receiver. It is considered to be suitable caused by exchanging data among vehicles to obtain the optimal
schema for obtaining an effective path based on the physical location next-hop.
information in realistic networks [11]. iii. Introducing a new routing technique using mobility metrics such
In VANET, the optimal route selection has typically acquired by the as the Euclidean distance, geographical routing information,
node broadcasting its location information with other neighbors in the delay time, and the ratio of residual power to enhance the
network, which leads to easy routing management [12,13]. However, position-based routing performance.
several challenges have been faced with this type of routing technique iv. Proposing a mathematical energy conservation approach to
such as the optimum detection for the reliability of the route between decrease the power consumption between vehicular items, thus
sender and receiver which forces massive amounts of data to be increasing the network lifetime.
broadcasted. This raises not only communication congestion probability v. Presenting a mathematical model to maintain the quality of link
but also low data packet delivery and high-power consumption [14]. via the Signal-to-Interference-Plus-Noise (SIN) and loss rate.
Another important issue is continuous breaks in link connections due to Therefore, the network stability and data delivery are increased.
the loss of neighbors causes poor communication quality and high delay vi. The simulation results depict that there is a 62% to 70%
time [15,16]. enhancement of the overall power consumption and VANET
For reliable communication with sufficient network bandwidth, throughput.
Software-Defined-Network (SDN) has been currently employed in real­
istic networks to offer accurate information about the traffic status and The rest of this paper is organized as follows: Section 2 reviews the
neighboring nodes across the networks that relying mainly on the free up-to-date literature reviews. Section 3 introduces a dynamic vehicular-
exchange of data among the VANET items. According to [6,17,18], the assisted architecture using SDN and fog computing. Section 4 demon­
data transmission process consumes more than two-thirds of the total strates the new routing protocol called Preparing Position-based Routing
energy of sensor batteries; this affects not only network lifetime but also and Energy Conservation Methodology (PRECM) via the mathematical
packet delivery and delay time. One solution is alleviating the influence model. Section 5 measures the complexity of PRECM in terms of time
of high data transmission to select the optimal next-hop. Hence, deliv­ complexity, communication overhead, and space complexity. Section 6
ering complete network overviews has become necessary. An SDN is. It estimates the stability process of the network lifetime. Section 7 dis­
has successfully forged cooperation between various data management cusses the performance evaluation results. Section 8 wraps up the paper
units to furnish a local network overview [19,20]. with significant contribution points.
By reducing processing power and data transfer volume, an SDN
enhances the performance of the VANET as a whole [21] with permit­ 2. Literature review
ting the network operations to decouple data exchanges generated by
the vehicular units. Such partition affects network performance and The rate of DD adjustment techniques has been investigated widely
holds the process of the data transmission inline. By resource provi­ by researchers in the academic area being effective ways to deploy and
sioning, expediting internal decision-making, and streamlining VANET develop urban transportation systems. Collaboration between SDN and
governance procedures, an SDN supports the QoS requirements. The fog computing has been proved to be instrumental by different recent
SDN-controller significantly supports OpenFlow by maintaining the flow studies. Power consumption, network bandwidth, delay time, and
tables that contain a list of flow entries throughout the VANET. In order communication reliability among other factors side of the vehicular
to optimize the flow of traffic and provide a reliable method of tracing systems have been enhanced due to that combination.
the present status of the VANET traffic, it also introduces a compre­
hensive view of status and topology [22]. 2.1. Fog computing in VANETs
Fog computing expands local cloud services at the edge of the end-
user. It promotes realistic networks to extract sophisticated data To ensure reliable transmission, fog technology has been introduced
locally without the demand to transmit this extraordinary data volume to bring new opportunities to VANET. For instance, Jorge Pereira et al.
to the cloud server. This property not only reduces network bandwidth [25] suggested a solid evidence system-based fog technology for inte­
usage and power consumption but also boosts network lifetime and grating computing services and their applications in the field of intelli­
stability [23]. Moreover, fog computing accelerates location awareness, gent mobility along with contemporary vehicle design. By preserving
and rapid response to unpredictable traffic events [24]. local data analytics at the user edge and redesigning both Roadside Units
In this study, a dynamic vehicular-assisted architecture allowing the (RSUs) and On-Board Units (OBUs) to represent the nodes of the fog, this
cooperation of SDN and fog technology to improve VANET performance architecture prevents delay time. According to the trial findings, using
is introduced. The novelty of this architecture lies in its ability to provide fog technology to the field of intelligent mobility will enable faster
a unique design with a comprehensive view of network traffic. This acquisition of accurate information. In the architecture, improvements
design helps to adjust the rate of DD caused by excessive data exchanges to location awareness, message delivery dependability, and response
in the position-based routing technique. Moreover, the proposed archi­ time were all taken into account. Yet, energy consumption and battery
tecture uses the mobility metrics such as the Euclidean distance, lifetime have not been taken into account in the proposed architecture.
geographical routing information, residual power ratio, and latency to In their research, [26] worked on improving total energy usage and

2
Z.H. Ali et al. Computer Networks 229 (2023) 109785

network connectivity by presenting a geo-distributed computing archi­ and low delay time registered outstanding results in this protocol. [32]
tecture using fog computing. Likewise, network bandwidth maintains by suggested other routing protocol based on merging fog computing with
allowing the system to control the amount of data uploaded to the cloud SDN for controlling DD in V2V manner. Packet loss, packet delivery ratio
server. However, the scalability of the network needs to be taken into and delay time were the assessment features for evaluating the effec­
account. tiveness of the proposed algorithm. Moreover, Truong et al. [33] who
offered a fog and SDN technologies-based system which enhances the
2.2. SDN in VANETs performance of the vehicular systems regarding scalability, resource
utilization, delay time, reliability and location-awareness. In the pro­
On another hand, sharing information that preserves the overall posed architecture, response time is enhanced because of using SDN
network conditions is highly recommended to reinforce the collabora­ which separates the data control and the forwarding functions which in
tion among vehicular units and to handle the data transmission rates as turn simplifies the VANET management. However, rate of power con­
well. That brought the researchers to install a configured control device sumption still without admitted control.
such as an SDN located at the edge of the customers’ network. Certainly, VANET Efficiency whenever implementing SDN with fog computing
such an idea introduced local data processing and worked effectively was tested by Pushpa et al. [34] through providing a tool that evaluates
against high usage of the network bandwidth. For example, various the network bandwidth, connectivity and time latency. The proposed
techniques for enhancing the performance of the drone are presented in tool controlled the traffic flow and enhanced the response time duo to
[27]. The paper introduced a service aggregation technique that sim­ virtualization feature sustained by the utilized Network Function Vir­
plifies the amount of the transferred data through the bandwidth of the tualization (NFV). Integrated fog computing with SDN system that
drone. Moreover, they introduced an approach for a normalized handles network transmission flow, able to exploit the resources and
throughput of a network channel to control the interference ratio which offers Internet of Things (IoTs) services was proposed in [35]. Despite of
implies the medium access control (MAC) layer overhead. ignoring the scalability issue in the proposed system, the combined
Moreover, Kumar et al. [28] built a smart framework called features of the fog computing and SDN acted for the strength of the
Software-Defined Drone Network (SDDN) for configuring UAV units that networks intelligence.
monitors road traffic by offering strategies that avoid collisions. In their In [36], designed new service-based architecture that merges the
framework, they tried to shrink network bandwidth usage by managing models of SDN and fog computing in VANETs with presenting a dedi­
large amounts of DD. Furthermore, they extended the flight time of the cated data scheduling algorithm. Simplifying scheduling of data in
drone, and effectively reduced the communication overhead along with higher dynamic environments by merging both fog computing and SDN
extending the sensors’ battery lifetime. And the proposed road-aware was the main contribution of that architecture. SDN structure allows
routing strategy using SDN by Abbas et al. [29]. For multi-hop rescheduling for the requests generated by the units of data manage­
communication, they divided the road networks into a set of segments ment; hence the resources usage is getting improved. Fog computing was
including RSUs. The rate of DD in VANET was handled by SDN who employed to govern the fast growing in the DD through monitoring the
acted as a control unit for that purpose. In their strategy, the main network bandwidth usage. The proposed algorithm still suffers from
contribution was reducing the delay time. Nevertheless, there is a lim­ complexity issues. Table 1 summarizes all mentioned papers in this
itation with SDN when experiencing a global network overview section and reorder it according to its published year.
including all the road conditions. And there is the Routing framework
based on SDN which was introduced by [30] for controlling the rate of 3. Proposed dynamic vehicular-assisted architecture using SDN
DD over the vehicular networks. For calculating the optimal path in the and fog computing (DVA-SDNF)
network, SDN is flexible to support the switching concept; therefore,
using SDN in the routing minimizes the communication overhead. This section introduces a dynamic vehicular-assisted architecture
allowing collaboration of SDN and fog computing called (DVA-SDNF) to
2.3. Fog computing along with SDN in VANETs ameliorate VANET performance. The novelty of this architecture lies in
its ability to provide a unique design that supports a comprehensive
Further studies for combining the fog computing with SDN concepts view of the networks and overcomes the challenges raised by fast-
in alliance systems have been proposed in the literature to raise the frequently fragmentation in VANET. Moreover, the DVA-SDNF pre­
performance of VANET. For example, there is Darabkh et al. [31] sents a control strategy to adjust the excessive DD and achieves minimal
introduced Innovative Cluster-Based Dual-Phase Routing Protocol Using communication overhead over the network.
Fog Computing and SDN (ICDRP-F SDVN) to confound the weaknesses Data transmission and traffic control are two levels at which the
of traditional VANETs routing protocols. The benefit of fog computing DVA-SDNF architecture operates. The data transmission level has
and SDN combined in the proposed protocol delivers a robust archi­ controlled by FOG technology. The FOG improves the operational effi­
tecture to fulfill all the new requirements and overcome challenges ciency of the real-time networking systems by adjusting the volume of
raised by the high speed of vehicles. The advantage of the ICDRP-F data traded over the networks. The traffic control level is established by
SDVN is to provide an efficient management overhead reduction SDN. The SDN ensures uniform resource utilization of all routing paths
mechanism, thereby decreasing the messages’ exchanges imposed. among the FOG nodes (fn). The proposed DVA-SDNF architecture has a
Kadhim [22] investigated the transmission power reduction in VANET hierarchical structure with scale-free topology to deliver a degree of
by presenting an approach for a multicast routing which asserted distribution networking solution, see Fig. 1. A single procedure has
deadline and bandwidth constraints based on both fog and SDN tech­ broken down into several smallish activities. Each task has been
nologies. That fulfills the QoS constraints by applying the scheduling assigned to a layer in the proposed architecture that works stalwartly to
and classification algorithms of multicast requests priority based. Their process the task only. The suggested design involves four primary layers
contribution focused on shrinking the time complexity and the power will be listed as follows: the Internet of Vehicles (IoV), SDN, FOG, and
consumption; yet they neglected the scalability feature. cloud transport layers.
Another routing protocol that is fog and SDN based was proposed in The IoV layer builds on the dynamic network topology in VANET. It
[17] to minimize the bandwidth usage of the excessive network and consists of electronic vehicles equipped with cache memory, OBUs,
maximize the transmission reliability. This technique shall impact the Global Positioning System (GPS), Geographic Information System (GIS),
quality of vehicular systems and ensure different local services such as and RSUs scattered on the roads to reach high coverage (connectivity)
route planning, traffic alert dissemination, elastic vehicular cloud ser­ among vehicles and between vehicles and traffic management units. To
vices, traffic monitoring services, and content transfer. High stability provide operational resilience in delivering a network edge that serves

3
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Table 1
Summary of closely related works that use both technologies a FOG and SDN in improving the VANET performance.
Refs. Objective Solution QoS-Factor Drawbacks Year
#

[31] Innovative Cluster-Based Dual-Phase Routing Protocol Using Fog Computing and Fog-SDN • Network • Energy consumption needs to 2022
SDN (ICDRP-F SDVN) to confound the weaknesses of traditional VANETs routing throughput. be considered.
protocols was proposed. • Packet delivery.
• Packet loss.
• Scalability.
[27] The VANET performance was improved via introducing a management framework Fog-SDN • Bandwidth usage. • Scalability is limited. 2022
with two techniques: (i) a new aggregation approach for reducing the amount of DD • Energy
and (ii) a normalized strategy with high dominance of the interferences on the consumption.
wireless links. • Latency time.
[28] A vehicular architecture was proposed under name Software-Defined Drone SDN • Energy • Location-awareness is limited. 2021
Network (SDDN) with a new strategy of collision avoidance for efficient deploying consumption.
UAV units in order to monitor road traffic. • Bandwidth usage.
• Communication
overhead.
[29] A geographically computing framework was built using fog technology. Fog • Bandwidth usage. • Scalability needs to be 2020
• Energy considered.
consumption.
[26] Increasing the ratio of the data packet delivery using SDN. Based on IoV, a new road- SDN • Response time. • Network coverage is lacking. 2020
aware routing strategy was innovated to maximize the amount of packet delivery • Network throughput • Concerning geographically
with avoiding the excessive rate of DD. network overview is lacking.
[17] Improving the communication capabilities in VANET by introducing position Fog-SDN • Network • System is very complex. 2020
routing strategy based on the combination between SDN and fog computing throughput.
technology. • Communication
reliability.
[35] A vehicular system was built on IoT services to utilize network infrastructure and Fog-SDN • Resource utilization. • Scalability needs to be 2020
govern data transmission across the VANET. • Response time. considered.
[36] Reducing the amount of big data transferring across VANET by providing a new Fog-SDN • Resource utilization. • Bandwidth consumption is 2020
smart transport system based on the combination of SDN and fog computing. • Data scheduling. high.
[25] Fog computing services and their implementations in ST are supported by a new Fog • Location-awareness. • Energy consumption is high. 2019
vehicular framework that was given to reduce the amount of DD in VANET. • Latency time.
• Communication
reliability.
[22] A multicast VANET routing protocol that includes deadline and bandwidth Fog-SDN • Power consumption. • Figure out how to use the 2019
constraints to reduce transmission power was proposed. • Time complexity. resources more effectively.
• Network bandwidth.
[37] A new vehicular framework to measure the effect of the combination between SDN Fog-SDN • Network • Scalability is limited. 2019
and fog technology on the network throughput and DD was provided. Throughput.
• latency time.
[33] A new robust transport framework introduced for supporting the collaboration of Fog-SDN • Latency time. • Power transmission is relatively 2018
fog and SDN was proposed. • Scalability. high.
• Communication
reliability.
• Resource utilization.
• Location-awareness.
[34] A new vehicular framework to measure the effect of the combination of fog Fog-SDN • Latency time. • Scalability needs to be 2017
computing and SDN on reducing the rate of DD was introduced. • Network considered.
Throughput.
[30] A new routing protocol using SDN for adjusting the rate of DD across VANET was SDN • Routing overhead. • The performance of routing 2018
proposed. • Latency time. protocol needs to enhance.
• Network bandwidth.
[32] A new robust routing protocol using the collaboration between fog and SDN for Fog-SDN • Packet delivery. • Power transmission needs to be 2016
governing the rate of DD in the V2V communication was proposed. • Latency time. taken into account.
• Network
throughput.

as a centralized monitoring system of the forwarding message, SDN maximize link quality. To this end, SDN reproduces a periodical message
coexists with RSU in the DVA-SDNF architecture. Consider that the RSU for all vehicles in its R to facilitate vehicle registration operation. Then,
is installed at road intersections and reconfigured to perform as SDN it asks the registered vehicles about their locations, movements, and
(data switch) to realize affordable costs. Every SDN is in charge of of­ velocity. This information participates in constructing both flow and
fering its services on a predetermined scale set by the FOG layer. To routing tables. A flow table involves information about the status for­
make information on the state of the roads more easily propagated warding and dropping data. A routing table includes a neighborhood
throughout the VANET, each SDN is marked with a distinctive ID. table with knowledge about network topology, vehicle state, and well-
In SDN controller layer, the allowed frequency range of communi­ connected routes calculated via the routing protocol. Additionally,
cation (R) between SDN and vehicles is defined according to the SDN provides an abstract level of security for information transfer be­
responsible fn. So, achieving reliable communication between every tween the FOG and IoV layers.
SDN and a set of vehicles V = { va , vb , vc , …, vn } must be done in The FOG layer is the second level of data controller in the proposed
va|comm SDN < R,∀ va ∈ V. The IoV layer divides into a set of cells, every DVA-SDNF architecture. It offers a control and management tool to su­
SDN propagates its ID in a specific cell to allow vehicles to connect pervise the rate of DD between the lower layers and the cloud server.
SDN’s services. Building an up-to-date local overview of VANET is a goal With SDN assistance, the fn constructs a database that has information
to ensure a continuous improvement in the next-hop selection and thus about the whole VANET. The FOG layer introduces message oriented-

4
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 1. Conceptual design of the DVA-SDNF architecture.

communication through a FOG-to-SDN connection instead of streaming- the fn as seen in Table 2. This flow table enhances the navigation process
oriented communication, and thus the communication overhead is by allowing nowcasting rather than forecasting techniques. Surely, the
reduced across VANET. The fn receives the whole data from SDN only, whole packet delivery and transmission reliability are improved across
which is filtered by SDN, and thus there is no need to direct connection VANET.
with vehicles. The data stored in fn and it categorizes into two cate­ The cloud transport layer is the upper layer of the DVA-SDNF ar­
gories: (i) data sent to get local processing and (ii) data to construct a chitecture. It provides a set of computing and unlimited data storage for
global network overview; this last category of data is stored in a flow incoming traffic from FOG. It also builds historical data about the overall
table. VANET conditions that helps to improve the network performance in
The IoV is segmented by fn into a set of certain cells called FOG Zone decision making. Surly, constructing the FOG layer in proximity to end-
(FZ). To offer an uninterrupted connection with SDN, each FZ has been users decrease the risk of data travel for a long distance to process at the
labeled by the responsible fn. With a multicast technique, the fn prop­ cloud server as well as it helps the cloud server makes a better-informed
agates their ID into each FZ allowing SDN devices to register themselves decision.
in this zone. Individual SDN sends up-to-date information to the
responsible fn about the network status, topology, and road conditions.
This collected information will be saved into a flow table generated by

5
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Table 2
The parameters of mobility metrics in the flow table.
f nid Location Timestamp Next-hop
SDNid V id Time Date Destid Velocity TTL
x y z h m s mm dd yy

– – – – – – – – – – – – – – – –

4. Preparing position-based routing and energy conservation VANET. The primary units that are deployed in this network will be
methodology (PRECM) listed as follows:

In this section, the mathematical model of the network topology i. BSs/RSUs: BSs and RSUs offer network connectivity and dis­
establishment and proposed PRECM is introduced. The abbreviations in covery across the IoV. The BS acquires information regarding the
this section are listed in the following Table 3. vehicle’s status that get from RSUs. All information is transmitted
to SDN using IEEE 802.11p to create a comprehensive view of the
whole network.
4.1. Network model
ii. SDN/RSUs: RSUs are recognized on the road to control traffic
discovery. Others RSUs that are placed at the intersections will
We consider fns are scattered randomly in the (3000n ×3000m) di­
reconfigure to perform as an SDN. SDNs are in charge of extra
mensions, and it governs by a FOG Head (FH). Consider all fns to have
functions such as gathering information about traffic conditions,
the same R, and thus the scale of IoV will be segmented of identical
establishing dynamic routing paths, and transmitting routing
dimensions called FZ. The SDNs are installed at road junctions. Only
information to FHs.
SDN has permission to connect to the FH, and it can transmit informa­
iii. fn/SDN: fn transmits a multi-cast message to the nearest SDN to
tion about the road conditions and vehicles. On the other hand, the IoV
notify it about the fn’s energy level and the distance between fn
is a set of vehicles va , vb , … vn . For example, the vehicles va and vb
and SDN. The most appropriate node will announce itself as an
transmit a request to SDN to connect to FH. The FH propagates a message
FH. The position-based routing technique determines the best
carrying its ID called (FH|id ) in the FZ. After receiving this message, the
paths between FHs and SDN.
va and vb store it in their cache, and they will exchange their information
with the SDN. The SDN uses this information to construct a temporal
To manage the amount of data transfer, SDN will broadcast the most
database. This concept improves not only the operational efficiency of
recent information to fn in the form of a multi-cast message. Depending
the navigation process but also communication overhead across the
on the energy level and location degree on SDN, all fn in the FOG layer
have an identical chance of becoming FHs. The SDN creates the flow
Table 3 table to store the coming information about fn; it will enable the system
A list of symbols. to layer fn in the future. Certainly, this operation assets better-informed
Symbol Description decisions in selecting the next-hop. We consider the fn has GPS with
V A number of vehicles. Digital Maps for the distance calculations, thus improving the opera­
r The measured value for the declaration of an initial FH. tional efficiency in the vehicle navigation [38]. The Euclidean distance
rs The required time to realize the stability of the network. can be employed in the local computation of FHs for the node without
do A threshold distance of an SDN and FH node into a certain FZ.
GPS. The following list contains the fundamental definitions that were
P|breakdown The measured value for a critical level of power, which denotes
the rate of energy depletion of fn. employed in this study:
Efn The energy of a fn.
Definition 1. r is the time estimation for completing all necessary
rN The maximum lifetime of each fn.
Dfn The rate of energy drains of each fn. calculations to announce an FH and accomplish a single cluster forma­
EPE The prior assigned energy of a fn. tion of FZ. We consider the initial round of data transmission will take 10
ECE The status of energy level of a fn. r.
TCT The current time.
TPT The prior time.
Definition 2. rs is the time taken to accomplish the network stability
Linkc The capacity of Path/Link.
Psize The maximum size of data packet by bit. and announce the final FH by SDN.
Erem (fn, r − 1) The remaining energy of fn at r.
Emin The minimum bound of energy through the round (r − 1th). Definition 3. FH is a selected node to be the head of a cluster char­
Emax The maximum bound of energy through the round (r − 1th). acterized by a high energy level and a minimal value of location degree
B|delay The limitation of delay values.
with SDN to maximize the stability of the network as long as possible.
(n − 1)lp The transmission latency of a ith.
r
Definition 4. do = εmpfs is a threshold distance into a single FZ between
ε
lp The allowed packet length to transfer across the network.
T|rate The amount of the data transmission. an SDN and a specified FH node.
gFR The areas out of the R of FH.
Propagation|speed The rate of propagation speed.
Definition 5. P|breakdown is the volume of energy depletion of fn, which
dmin The minimum radius assigned to define the minimum bound of FZ.
dmax The maximum radius assigned to define the maximum bound of may affect the lifetime of the network and result in failure. The
FZ. P|breakdown is calculated by:
dFH − dSDN The difference distance between any FH and a specific SDN.
FZareai The FZ is determined for each FH. P|breakdown = Max (Efn|rN
dmax The maximum and minimum bounds of FZ.
1 − Here, Efn denotes the energy level of fn. rN denotes the maximum
dmin
Nnoise The noises in a certain link l. lifetime of each fn.
E(trans|rece) The rate of energy consumption on both transmitter and receiver
packets.
Definition 6. Dfn is the energy drain level for fn, and it can be
εmp The propagation of multipath.
εfs The propagation of free space. measured depending upon the amount of staying energy and the

6
Z.H. Ali et al. Computer Networks 229 (2023) 109785

draining speed of the fn in the FOG layer. It is mathematically expressed with three operations are the calculations of distance, the measurements
as: of residual power, and the measurements of location degree. These
measurements are required to construct a separate flow table on both
EPE − ECE
Dfn = FOG and SDN. The flow table on the FOG layer is essential to pre-install
TCT − TPT
routing paths that can simplify communication operations among fn
Here, EPE denotes the previously assigned energy of a fn, and ECE nodes. However, the flow table on the SDN controller layer is influential
denotes the current power level for a fn. The TCT is the current time, and in constructing the optimal route. After completing these calculations,
TPT is the earlier time. The FH typically produces a Time-Division Mul­ the system has entered a new phase divided into four sub-phases: delay
tiple Access (TDMA) schedule and sets a time slot for transmitting every time, FHs declarations, FZ initiation, and link quality.
packet. The network returns to the start-up phase after the transmission
process of the packet is carried out to announce a new FH. A-Distance

Definition 7. linkc is the maximum link throughput that ensures the All fns are dispersed at random throughout the FOG layer, and they
connection between announced FH and SDN is alive; it represents by: have a fully Digital Mapping System (DMS). Recently, the DMS has
Psize developed to provide automated real-time navigation services [38]. The
linkc = fns use DMS to calculate the distance via the Euclidean distance between
∇ (Tr − Ts )
themself and SDN. Moreover, the DMS permits all fns to exchange their
Here, Psize denotes the maximum size of data packet. The term ∇ (Tr position information, road ID (rid ), and curve-metric distance with the
− Ts ) is the time interval by sec., it refers to the distinction between the SDN. The Euclidean distance between FOG to SDN can be computed as:
time taken to successfully received a packet Tr and the total time for n,m
( ) ∑ ⃒ ⃒
packet sent. Dist γfn , γ SDN = ⃒vfni − vSDNi ⃒ (1)
i,j

4.2. Routing procedures in VANET

The terms of Dist (γfn , γ SDN ) indicate the curve-metric distance
through a fn to SDN, it calculates as the sum of the absolute values of the
In this section, the proposed PRECM uses a method to measure the
coordinate differences through the source and the current position of
constitutive routing approach and efficient FH selection in a mathe­ ⃒
destination ⃒(vfni − vSDNi ).
matical way. As shown in Fig. 2, the process in the PRECM has three
main operations: (i) neighbor discovery, (ii) SDN operations, and (iii)
data transmission. B-Direction

4.2.1. Neighbor discovery The optimal route between fn and SDN will be computed later by
The neighbor discovery is the preparation operation performs in a SDN based on accurate information about the correct address of fn. Since
decentralized manner. It includes a single phase called the start-up the SDNs have a stationary place, it thus becomes crucial to predict the
phase with three operations: distance, FOG directions, and residual current location for each fn nodes for determining the possible number
power calculations. In the beginning, the SDNs are distributed and of hops to reach the destination. This operation controls the power
reinstalled along the route at junctions. The communication between consumption in data transmission, therefore enhancing network life­
SDN and vehicles via a bidirectional connection. While the communi­ time. As shown in Fig. 3, consider fn nodes are traveling in a frequent
cation between SDN and the FOG layer via a multi-cast message to behavior in a distance of 2d. The distance between the first deployed
overcome broadcast issues. fn nodes share information about their status node will be equal to 2d
π . All possibilities for the movements of fn can be

with SDN. This information is a seed to construct the flow table at SDN, given by:
which generally holds the network flow entries, the real distance be­
tween a single fn to an SDN, and the amount of residual energy.

Start-Up phase. With the first round (r), the start-up phase will occur

Fig. 2. The process of routing construction.

7
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 3. The fn’s movements.

⎧
⎪
⎪ − 1 fn is immobile; round duration at time t. The μi carries the allowed energy bounds
⎪
⎪
⎪
⎪ π confined within the lowest bound is Emin , and the highest bound is Emax
⎪
⎪
⎪
⎪
⎨ 2d
for va = 0 δ vb > 0; during the round (r − 1th). Based on the bounds of P|breakdown , the energy
fn|address = 3π (2) levels will be adjusted. According to Eqs. (4) and (5), the bounds of
⎪
⎪
⎪
⎪ 2d
for va = 0 δ vb < 0; energy are defined for each fn. Thus, the fn that has an energy level less
⎪
⎪
⎪
⎪ than the Emin will be automatically blocked in the first round r. Only fn
⎪ 2d − 1
⎪
⎩ for va = 0 δ vb = ∞; that have an energy level more than or equal to Emax will participate in
2d
the next round (rs ). After completing this stage, each SDN will arrange
all eligible nodes to select the most proper one.
C-Residual Power
4.2.2. Dual phase using SDN and PRECM
With the network discovery, another essential parameter computed The essential role for creating a flow table in SDN is to the traffic
by fn is to Residual Power Ratio (RPR%), which is a necessity to sustain status and the topology managers collaboratively. Pre-installing a
the network alive and guarantee high data packet delivery. All fn nodes routing table with accurate topology prediction is an aim of the flow
deployed in the FOG layer have a requirement to interact with others to table update policy. Based on the information collected from the vehi­
transmit up-to-date information about the network. This operation of cles, the SDN controller can predict the vehicle’s potential trajectory and
data transmission should be done under real-time restrictions, to keep specify its topology adjustment when a vehicle is connected to the
their batteries up and the network lifetime inline. Therefore, calculating controller. This operation helps the fn to provide its services for IoV in an
RPR% for each fn is significant to avoid the network lifetime ending end-to-end manner, thus a considerable reduction in the communication
prematurely. RPR% is given by: overhead by controlling number of Route Request (RREQs) messages .
Even if the vehicle disconnected from SDN in a short period, the pre-
Erem (fn, r − 1)
RPR% = ≥ P|breakdown (3) installed paths in the flow table still are used to predict its position [39].
Etotal (fn, r − 1)
The controller in SDN is a software program created by OpenFlow to
Here, Erem (fn, r − 1) denotes the remaining energy of fn, r denotes the manage and monitor the table flow contents. The controller is the only
time taken to establish the first cluster and data gathering process device responsible for notifying fn about the connection corruption
among all fns. The term Erem is given by: Etotal − (I × Volt × T)Joule, the I instead of dominating the whole VANET with details about all the path
is a current (in Ampere), Volt is the total voltage, and T is the time taken changes. In this case, the process of routing convergence is less affected.
to transmit and receive the data across the network. The Etotal (fn, n − 1) The controller acquires the redirected path and forces the redirected
denotes the whole level of energy of fn. The value of the right side must path information to the concerned switches [40]. If the packet transmits
be greater than or equal to the breakdown value in the left side. The from the source to the destination, the flow table looks for the header
energy level of fn is given by the Eqs. (4) and (5) as follows: packet at the source and destination as well as matches the path that was
taken by the data packet, and another saved in the flow table. If the error
Etotal (fn, r − 1) = μi Δt (4)
occurs in this process, several actions will be taken such as the respon­
μi = rand(Emin (r − 1), Emax (r − 1)) (5) sible SDN-controller will notify fn and other controllers across the
network, and the OpenFlow will encapsulate the data packet and
Here, μi denotes the energy level of fn at the round (r − 1th), Δt is the re-transmit it or drop the data packet directly.

8
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Another aspect of the OpenFlow network is that the OpenFlow n denotes the maximum number of deployed fn, the lp denotes the length
switch can govern the traffic flow once it fits the flow table and does not of the data packet allowed to transfer across the VANET bandwidth, and
∑n− gap− 1
require further communication from the OpenFlow controller. This do+gR
T|rate is the amount of data transmission. The term Propagation
i=1
denotes
makes switches forward data more efficiently. However, the issues come |speed

up once the flow table’s laws no longer match the state of the network. the propagation delay time, do denotes a threshold of distance, the
Herein, the controller in SDN will filter the collected data from the fn gR denotes the number of areas out of R, and the Propagation|speed denotes
and recast it to obtain the optimal selection of an FH. An SDN can filtrate the total of propagation speed.
data according to a set of metrics. In this case, three measurable factors
are considered such as distance, latency time, and level of power. To B. FH Selection
reach a smooth data flow and thus deliver a better routing solution, the
weights for selected paths should be calculated. In this section, we This section is going to pick the much more proper fn to be an FH.
discover the procedure for selecting a suitable fn and a way for maxi­ OpenFlow will save the data gathered from the FOG layer into a flow
mizing link quality between the selected node and SDN. table. The nodes of fn are by default positioned and reconfigured in
accordance with their proximity to the closest SDN. As shown in Fig. 4, if
A. Delay Time the distance between the fn and SDN is less than or equal to the threshold
do , the fn will be in the nearest layer to SDN that is the 1st layer.
One important QoS indicator for assessing network performance and Otherwise, the fn will rearrange according to their do .
validity is End-to-End (E2E) delay. It is defined as the estimated time Once an SDN acquires the required information, an SDN initiates
taken for a packet to transmit from a source node (mean fn) to a desti­ instantly to the transformation of this information to be more conse­
nation node (mean SDN). Since the nature of VANET is the highly dy­ quential for announcing FH and producing the optimal routing path. To
namic topology and fast fragmentation. So, it becomes difficult to this end, the weights (W) will be calculated by the OpenFlow in the SDN.
estimate this time accurately. Here, several parameters must be taken The degree of location, RPR%, and E2E delay are significant parameters
into account such as the bit rate of data, distance, propagation time, and for weight calculations. The lower weights denote the short path be­
transmission time. The E2E can be expressed as: tween the SDN and fn. Therefore, the optimal route can achieve via a set
n ( ) of measurable values: the shortest distance between the fn and SDN,
∑
E2 E|totalDelay = OneHop|delay + B|delay (6) high RPR%, and low latency. Consequently, an SDN denotes the opti­
i=1 mization of path selection by a general equation, which can describe as
The i ∈ {1, 2, 3, 4, …, n} denotes the number of fn, the oneHop term follows:
will be measured as follows: OneHop|delay = t|processingDelay + t|propagationDelay ∑n,m [
max(RPR%)
]
+ t|transmissionDelay + tchannelDelay + treceptionDelay + t|queuingDelay [38]. B|delay de­ W(fn,SDN) = ωi ( ) (8)
min dist(fn, SDN) + T|delay
notes the allowed bounded delay values. Form the Eq. (6), the E2E delay
i,j=1

time estimation of fn can be formulated as: Based on Eqs. (1), (3), and (7), W between any pairs (fn, SDN) can be
∑n− gap− 1 calculated as follows:
(n − 1)lp do+gR ∑n− 1 ∑
n− 1
i=1
fn|delay = + + t|queuingDelay + t|processingDelay
T|rate Propagation|speed i=1 i=1

(7)
The fn|delay is a measured value determined by a set of variables: the
(n− 1)lp
term T|rate
denotes the delay of data transmission of a fn for each ith, the

Fig. 4. The nodes of fn layered by the controller in SDN.

9
Z.H. Ali et al. Computer Networks 229 (2023) 109785

RPR% RPR% We consider the area study has six intersections equipped with six
W(fni , SDNi ) = ω1 ( ) + ω2 ( ) +…
dist γ fn1 ,γ SDN1 + T|delayfn1 dist γ fn2 ,γ SDN2 + T|delayfn2 SDN devices. Each SDN controller must designate the most suitable fn as
an FH. The proper FH must be the closest to an SDN to decrease the
RPR%
+ωn ( ) latency; it also has the maximum energy balance to keep the network
dist γfnn ,γ SDNn + T|delayfnn
alive as long as possible. Additionally, the FH offers several functions
(9) that would prevent expanding access to the cloud server. This operation
Here, i ∈ {1, 2, 3, 4, …, n}, the n denotes the number of fn and j ∈ {1, enables the vehicular community to employ the computing services
2, 3, 4, …, m}, n denotes the number of SDN devices. W the weight of the from fn and saves bandwidth usage.
link between two nodes (fn, SDN) with different constant-coefficient The deduction of FH depends on the W calculation of edges. To
values (ω1 , ω2 , …, ωn ). The total should not be greater than (1), and accomplish this task, the highest rate of PRP%, the shortest distance
they can represent as follows: between each fn to SDN, and the lowest time of delay are the important
⎡ ⎤
ω1 metrics required to discover the optimal weight value. Suppose that the
⎢ ω2 ⎥ connection has been established between a set of fn and SDN1 , as seen in
⎣⋯⎦≈1
Coefficient factor (constant) (ω) = ⎢ ⎥
Fig. 5. The values of the QoS metrics are gained by the local computing
ωn unit in fn. The constant-coefficient values (ω1 , ω2 , …, ωn ) are assigned
Otherwise, the total value of W is determined by three factors: the based on both distance and delay time. While the SDN’s decision about
estimation of delay time taken from sending one packet from any fn until fn is exclusion/inclusion given based on the balance of PRP. Table 4
receiving this packet by an SDN, the Euclidean distance from an SDN to displays a sample of these values. Hence, the SDN starts to calculate Wi
each deployed fn, and the real value of energy. Fig. 5 shows the illus­ as follows:
tration of how to calculate W before selecting FH, and Fig. 6 shows the ( )
90%
flowchart of the FH election process. The pseudocode for selecting an FH W(fn22 , SDN1 ) = 0.3 = 13.2
200.44 + 3.8
is illustrated in the Algorithm 1.
( )
89%
- Illustrative Example W(fn23 , SDN1 ) = 0.6 = 21.3
245.11 + 5.0

Fig. 5. The illustration of how to calculate W and select FH.

10
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 6. Flowchart of the process of FH election.

(
70%
) C. Measurements
W(fn44 , SDN1 ) = 0.1 = 3.4
198.43 + 4.0
To achieve high connectivity sufficient to maximize packets deliv­
(
98%
) ered and avoid the interruption of communication across the network,
W(fn51 , SDN1 ) = 0.02
99.12 + 0.5
= 1.9 the coordinates of the FZ should be markable. As mentioned in Table 4,
SDN1 announces fn51 is an FH. The fn51 starts to initialize the bounds of
Based on the previous calculations, the fn with a number (51) has the its FZ according to its R. The SDN1 begins to be layered other fn nodes
lowest weight, and thus SDN controller can declare it as an FH. The within three levels relied on their signal strength. The fn51 is responsible
second step is to define the bounded of the FZ by fn51 . for tracking the limitation of bounds (internal and external) of these

11
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Algorithm 1 and (ii) maximum radius (dmax ) that has allocated to the third layer for
W calculation and an FH election. specifying the external edge of FZ. We consider the term dmin is defined
Result: W, an FH election as the difference distance through FH and SDN, and it measures as:
For T=anytime & T <= ΔTr − Ts {
For X = 1 to X= (fnTotalNumber) {
dmin = dfn − dSDN (10)
A flow table created by the collected information from fn;
In our case, the formula dfn − dSDN indicates the difference between
While Location (fn) ∈ Location (SDN) {
fn transmit information to SDN; fn51 and SDN1 that indicates the smallest range that covered by fn51 .
If Dist (γfn , γSDN ) ≤ do { Otherwise, the dmax should not exceed the allowed R for fn51 . The dmax
fn is located in 1st layer; indicates the difference through fn51 and the farthest node deployed in
// Calculate PRP%; the 3rd layer, and it will be calculated as follows:
If P|breakdown = Max(Efn|rN ) = 0 {

RPR% =
Erem (fn, r − 1)
≥ P|breakdown ;
dmax = dfn51 − dfarthest (11)
Etotal (fn, r − 1)
// Calculate fn|delay ; From the Eqs. (10) and (12), the total area of FZ can be modeled as:
∑n− gap− 1
(n − 1)lp i=1 do+gR ∑n− 1 ∑n− 1 [ ]
fn|delay = + + t + t ; dmax
T|rate Propagation|speed i=1 |queuingDelay i=1 |processingDelay FZareai = π 1 − (12)
// Calculate W; dmin
]
∑n,m [
The FZareai denotes the FZ for each FH, the π denotes a constant help
max(RPR%)
W(fn,SDN) = i,j=1 ωi ;
min(dist(fn, SDN) + T|delay )
}} to find the circumference of a circle by measuring the radius, and the
[ ]
{ term of 1 − ddmax uses to define the internal and external bounds of FZ.
min
fn does’t have any chance to be FH;
} The pseudocode for identifying the bounds of FZ is illustrated in Algo­
} rithm 2.
}
}
D. Link Quality

The link quality is an essential factor in achieving transmission

Table 4 reliability and network stability. Building a resilience channel between
A sample of mobility factors for calculating W and declaring fn as the first FH. the sender and the receiver increased the total message delivery ratio,
Sender Receiver Euclidean- E2E PRP ω Decision reduced link failure rate, and maximized network throughput. The
distance (μs) (%) concept of Channel State Information (CSI) is a measurable value for
fn21 SDN1 154.13 2.3 40 0.03 Exclusion adjusting the consumption of power transmission, attainable data
fn22 SDN1 200.44 3.8 90 0.3 Inclusion packet rate, and faultless link quality via the Signal-to-Interference-Plus-
fn23 SDN1 245.11 5 89 0.6 Inclusion Noise Ratio (SINR) and loss rate [38,41].
fn31 SDN1 199.15 4.1 10 0.02 Exclusion
fn44 SDN1 198.43 4.0 70 0.1 Inclusion
The link (l) is a direct line (or path) decided between transmitter and
fn51 SDN1 44.12 0.5 98 0.02 Inclusion receiver to permit the network systems to establish a successful trans­
mission operation. In general, if two wireless nodes are using the same
radio connection and pass through the same interference zone, their
levels to guarantee the degree of connectivity. As depicted in Fig. 7, communication activities on one link may interfere with those on the
there are two important calculations that are: (i) minimum radius (dmin ) other. The communication process in VANET accommodates varying
that has allocated to the first layer for specifying the internal edge of FZ,

Fig. 7. A layered system within the FZ.

12
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Algorithm 2 5. Complexity of PRECM

Determining the bounds of FZ.
Result: Determining the bounds of FZ This section aims to measure the complexity of PRECM in different
For T=anytime & T ≤ Δ Tr − Ts { terms include the time complexity, communication overhead, and space
// Each fn calculates its distance to SDN based on its signal strength; complexity.
For Y = 1 to Y= ∀fn {
∑n,m ⃒⃒ ⃒
Dist (γfn , γSDN ) = i,j=1 dfni − dSDNj ;
⃒
}
5.1. Time complexity
While dmin ≥ Y ≤ dmax do
//Determine distance via FH; The time complexity of PRECM consists of three factor of complexity
If Y ≤ do do (i) computational complexity (PathΘ ) to obtain all possible paths, (ii)
fn is located in 1st layer;
//FH determines the difference through itself and SDN; computation time complexity for selecting a reliable path (R Θ
l ), and (iii)
FH ← dmin ; running time for the PRECM.
else The PathΘ of PRECM for the undirect graph (G) with a set of vehicles
fn is located in 3rd layer;
V = { va , vb , vc , …, vn } and edges (e) could be determined by:
//FH determines the difference through itself and farthest node;
FH ← dmax ; PathΘ = O (V + e) (18)
//Set the bounds of FZ;
SDN ← FZareai ; Assume that R is a constant equal O (1). The cost of the computa­
Θ
}
l

} tion complexity of R Θl for N path can be calculated as:

} ( )
Cost R Θl = O (N ) (19)

Assume that the complexity of the computation time that use to

levels of interference, which results in sounds and distortion that have
determine the path score of all paths is a constant equal O (1). The time
adversely impacted the connection capacity. Consequently, network
complexity of PRECM for all available paths can be given as:
bandwidth usage balloons to uncontrollable heights. By identifying the
( ( )
route with the least amount of interference and bandwidth, position- T|complexity = (PathΘ + N ) × Cost R Θl + Cost|total ) (20)
based routing is a technique that could be used to achieve quick trans­
mission times [42]. T|complexity = O (O (V + e) + O (N × l)) (21)
To find the optimal route that has high quality of link and low SINR,
we consider the fn sends the number of packets (P) to SDNj via a certain Based on Eqs. (20) and (21), the worst case of the complexity of
link l. P is successfully received at SDNj based on the following Eq.: PRECM can be calculated as follows:
( )
( )
[
P|size
] T|complexity = O V 2 + (N × l) (22)
lm fni , SDNj = lthrput − ≥ψ (13)
Nnoise + BW + Dfn
5.2. Communication overhead
lthrput denotes the total throughput of link, and P|size denotes the packet
size requires for transmitting from any fn to SDN. Certainly, the packet
In the communication overhead, the energy consumption estimation
size should not exceed the total linkc . The Nnoise denotes the noises appear
(O E) via network lifetime is the proportion of the energy consumed in
in a certain link l. BW denotes the estimation of total bandwidth of l,
transmitting control packets to that used for data packet transmission.
which is given by:
Assume that the energy cost of network setup during one round r is
BW = IRl × linkc (14) Cost E|setup (r), and the energy cost of one complete data-gathering pro­
cess during one round r is Cost E|data− gathering (r), wherein each fn trans­
SINR(l,BW)
IR indicates the ratio of interference on l. IRl can be given by: SNR(l,BW) . So, mits at least one packet to SDN for complete communication. For low
P(l,BW) P(l,BW) communication overhead, the energy consumed in network construc­
SINR(l,BW) = and the SNR(l,BW) = Here, SNR(l,BW) indicates
1− N(noise ,l) N(noise ,l) tion must be significantly less than the energy consumed in the data-
the amount of signal-to-noise of l. Dfn indicates the energy drain for each gathering process Cost E|setup (r)≪ Cost E|data− gathering (r). Herein, it is
fn. The last term in the Eq. (13) uses ψ as a constant value. Path|loss in­ possible to compute the ratio of energy overhead O E% by:
dicates the path loss that measures the power gains and expected
∑
rs
attenuation in the signal in telecommunication systems. Here, Path|loss O E% =
Cost E|setup (r)
(23)
can measure by the differences between the packet loss in forward and r=1
Cost E|data− gathering (r)
reverse transmission, and it calculates by:
1 5.3. Space complexity
Path|loss = (15)
P|forwred − P|reverse
Space of complexity (SΘ ) is determined by identifying all neighbor­
Here, P|forwred and P|reverse are defined as the expected number of the hood in the network via V and all available paths N established among
packet loss in the forward and reverse situation in a li . While P|forwred and V. So, the SΘ can be calculated by:
P|reverse rely mostly on the possibility of successful delivering a data
packet P|success and undelivered packet P|unsuccess from SDN to fn. The SΘ = O (V + N ) (24)
P|unsuccess and P|success are given by:
( )( ) 6. Estimation of network stability and lifetime
P|unsuccess = 1 − 1 − P|forwred 1 − P|reverse (16)
( ) The major objectives of the different routing strategies used in
P|success = Ps− 1
1 − P|unsuccess (17) VANET are to maximize network stability and longevity. As a result,
network stability and the preservation of link quality among vehicles are
Where Ps− 1 indicates the total number of attempts to retransmit data
tightly related. Whereas the lifetime of the network refers to maintaining
packet across the l.
the power of nodes up to survive a long period [6,18]. According to 4, all
nodes of fn are linked to SDN at least once at the start of the connection,

13
Z.H. Ali et al. Computer Networks 229 (2023) 109785

and the packet transmission process continues until the end of the sta­ transportability speed of the vehicle [5], which was estimated between
bility period. The total power consumption in the negotiation operation 15 Mph to 85 Mph as an average. Eclipse SUMO uses the .net file that
between FOG and SDN can defined as follows: was generated by the Netconvert tool [43], Fig. 8 shows the study area
corresponds to the Giza, Egypt, which is 8 km2. The Netedit tool uses to
ΔE(fn,SDN) =
⎧ ( )( ) adjust the maximum vehicle speed on the road. Urban scenario pa­
⎪
⎪ lm f ni , SDNj ΔE(trans|receive) + εmp d4 (fn, SDN) + rameters using Eclipse SUMO is illustrated in Table 5. A representation
⎪
⎪
⎪
⎪ ( ) of the correlations between every pair in the topology of VANET besides
⎨ rs p|size ΔE(trans|receive) + εmp d4 (fn, SDN) + ΔEdagg , d(fni ,SDNj ) ≥ do (25)
modeling the entire topology has been simulated by using ns2.35 [44].
( )( )
⎪
⎪
⎪ lm f ni , SDNj ΔE(trans|receive) + εfs d2 (fn, SDN) + As an assumption, simulation time has been set to 1000 s, and for
⎪
⎪
⎪
⎩ ( ) designing the network topology and generating VANET range from 10 to
rs p|size ΔE(trans|receive) + εfs d2 (fn, SDN) + ΔEdagg , Otherwise
100 nodes, a file named Tool Command Language (TCL) has been in a 6
× 6 Manhattan grid road network. The PRECM has been implemented
E(fn,SDN) =
on NS2.32. The entire configurations of the network have been
⎧ ( )( )
⎪
⎪ lm f ni , SDNj E(trans|receive) + εmp d4 (fn, SDN) + demonstrated in Tables 6 and 7. The behavior of fn is configured by the
⎪
⎪
⎪
⎪ ( ) reference guide called Fog Hierarchical Deployment Model from
⎨ rs p|size E(trans|receive) + εmp d4 (fn, SDN) + Edagg , d(fni ,SDNj ) ≥ do (26)
OpenFog Reference Architecture.
( )( )
⎪
⎪
⎪ lm f ni , SDNj E(trans|receive) + εfs d2 (fn, SDN) For assessing the proposed PRECM, a scale of traffic rates (Pkt/Sec)
⎪
⎪
⎪
⎩ ( ) and vehicle speeds (km/h) has been prepared. For obtaining Constant
+rs p|size E(trans|receive) + εfs d2 (fn, SDN) + Edagg , Otherwise
Bit Rate (CBR) as a source of traffic between fn and the VANET, 10
The Δ denotes the various degrees of fn. The E(trans|receive) denotes the communication connections were installed randomly. Every 2 s, a traffic
energy consumption in the case of sending and receiving data. εmp de­ agent Node-UDP and 512 bytes packets are generated via CBR. The
notes the propagation of multipath, and εfs denotes the propagation of connections of the network have been tested by passing the HELLO test
free space. d4 denotes the multipath fading (power loss), and d2 denotes message between fn and RSUs, and the interval time of the message was
the free space (power loss). P|size denotes the estimation of data packet selected to be 1.5 s. Because of the probabilities of loss and possible
collisions, details from a HELLO-delivered message have been retained
size. rs denotes the time taken to network stability. do denotes the
for (2.5 × hellointerval) for each neighbor. Yet, it is in the identified R
allowed threshold distance from fn to SDN. Edagg denotes the power
for every fn node. In Section 4.2.2, a configuration for the values of the
consumption in the case of data aggregation. The network topology does
constant-coefficient has been shown as: ω1 = ω2 = ω3 ⋯ = ωn ≈ 1.
not substantially change after network stability has been obtained. Due
to the enormous amount of data that must be transferred through the
internet, energy losses continue to be a bottleneck. As a result, the
7.2. Assessment metrics and discussion
proposed PRECM’s usefulness has been demonstrated.
From the perspective of the network lifetime, the stability of the
Due to the obtained results, the proposed PRECM based on SDN, and
route denotes the link’s capacity to be endured for a long time without
Fog technologies has outperformed other competitive research in terms
failure. Herein, the link’s capacity does not affect the period of the
of the effectiveness of selecting the appropriate path and excessive data
network stability but also the effectiveness of the communication
dissemination management over VANET. Enhancing the VANET per­
operation and the lifetime network topology. In this case, we calculate
formance via exploiting the network bandwidth and the rate of power
the period of network lifetime for all nodes of fn that was already
consumption has been recognized due to the adoption of the proposed
declared as FH. linkc|max indicates the link capacity in case the connec­
PRECM. Many assessment metrics have been used to evaluate the per­
tion through (FH, SDN) is established. Therefore, the lifetime of the
formance of PRECM, including normalized routing overhead, power
network is defined based on Eqs. (25) and (26) as follows:
consumption, network throughput, percentage of the delivered and loss
⎧
⎪ of packets, and E2E delay time. During 12S experiments, the perfor­
⎪
⎪ (
⎪
⎪
⎪
)(
l f n , SDNj 2E(trans|receive) + εmp d4 (fn, SDN) + FZareai 4
) mance of proposed PRECM has been tested against EMHR with BwEst
⎪
⎪ m i
⎪
⎪ ( ) [38], EERD [27], SFSR [17], and enhanced AODV [45]. The network
⎪ +rs p|size E(trans|receive) + εmp d4 (fn, SDN) + Edagg ,
⎪
⎪
⎪
⎪ ( ) setting for these modern algorithms vs. the PRECM is discussed in the
⎪
⎪
⎪
⎪ FZ, d(fni ,SDNj ) ≥ do Table 8. The upcoming sections discuss with analysis of the out­
⎪
⎪
⎪
⎪
⎪
( )( 2 4
) performing results according to applying PRECM versus other compet­
⎨ lm f ni , SDNj 2E(trans|receive) + εmp d (fn, SDN) + FZareai
( ) itive proposals.
E|init = +rs p|size E(trans|receive) + εmp d2 (fn, SDN) + Edagg , (27)
⎪
⎪
⎪
⎪
⎪
⎪
⎪
d (fni ,SDNj ) < d o ≤ FZ 7.2.1. Impact PRECM in terms of routing overhead
⎪
⎪
⎪
⎪
⎪ l
(
f n , SDN
)(
2E + ε d2 (fn, SDN) + FZareai
2
) At the beginning, capability of PRECM to deliver the data packet
⎪ m i j (trans|receive) fs
⎪
⎪
⎪ ( ) efficiently had to be measured through monitoring the ratio of the
⎪ +rs p|size E(trans|receive) + εfs d2 (fn, SDN) + Edagg ,
⎪
⎪
⎪
⎪ network control packets. Normalized Routing Overhead (NRO) has been
⎪ Otherwise
⎪
⎪
⎩ exploited to study the impact of PRECM from this perspective. NRO can
be obtained by applying the following formula Eq. (27):
NRO = P|fail + T|meg + Tri|meg (27)
7. Performance evaluation
Where P|fail represents the number of failed data packets that
7.1. Simulation setting couldn’t reach the destination, T|meg is the periodic messages, and Tri|meg
represents the trigger messages for arriving at the packet to the
A discussion of simulation results along with our assessment tools is destination.
going to be introduced in this section. All setup has been conducted on For evaluating the strength of the proposed routing technique in the
the operating system Ubuntu 20.04.5 LTS that installed on the laptop VANET network in terms of routing overhead avoidance problems, the
with picked Intel’s 10th Gen Core i7–1065G7 on the 15-inch Book 3, efficiency of PRECM was tested on data traffic rate. As shown in Fig. 9
32GB RAM, Nvidia GeForce GTX 1660 Ti Max-Q. Simulation of Urban (a), the effect of adopting PRECM in VANET has been spotted on the rate
Mobility (Eclipse SUMO) has been employed for monitoring the of the data traffic. A comparative study has been conducted among

14
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 8. Giza, Egypt study area (8 km2).

Table 5 Table 7
Urban scenario parameters using Eclipse SUMO. IEEE802.11p configuration [42].
Parameter Value Parameter Value

Vehicle Number 100 vehicles Center Frequency 5.875 GHz

Simulation Time 1000 s. Receiver sensitivity - 95.2 bBm
Area 8 km2 Network Bandwidth 10 GHz
Avg. Speed 15 Mph to 85 Mph Antenna Gain 2 bBi
Junction Number 65 Channel 175
Setup TxPower 23 bBm/18 bBm

Table 6 enhanced AODV proposed in [45] were delivered in Fig. 9(b). As can be
Topology setting. seen in Fig. 9(b), noticeable improvement has been reached by PRECM
Parameter Value versus the enhanced AODV protocol proposed in [45]. For avoiding the
Simulation tool NS2.35 issues of the routing overhead, PRECM had come in the first place out­
Vehicle range 1–100 vehicles performing the enhanced AODV by a difference ratio estimated by 6%.
Simulation time 1000 s. PRECM’s selection methodology along with its link capacity and sus­
Transmission rage 150m
tainable stable path have ensured the advanced behavior of PRECM.
Mobility generator SUMO
Propagation Model Nakagami
PRECM has recorded a slight difference over EMHR with BwEst [38]. In
Mobility model Manhattan this comparison, EERD [27] stuck to the third rank, while enhanced
Interface type WirelessPhy AODV was the fourth and last ranked. Affirmatively, the performance of
Channel type WirelessChannel the routing protocol is highly affected by the speed of the vehicle. So, the
Velocity 5, 20, 60 km/h
NRO ratio is directly proportional to the average vehicle speed.
Number of seed 1
MAC protocol IEEE 802.11p
Max. packet size 512-Byte 7.2.2. Impact precm in terms of power consumption
Traffic agent Node-UDP Path optimization was not the only judging metric for testing the
Energy model Battery quality of PRECM. On the other hand, further analysis has been deliv­
Initial power Random [0,1000] Joule
CBR 1 packet/s
ered to evaluate the performance of the proposed routing algorithm
Traffic type CBR considering power conservation regarding the data transmission factor.
PRECM has been employed for decreasing the overall consumed energy
in BS via a wireless radio network. Additionally, normalization to the
different network systems before and after applying PRECMs. As shown number of roadside sensors is applied in VANET. Fig. 10(b) shows that
in 9(a), applying PRECM introduced a level of excellence in terms of maximum energy depletion has been reached by AODV, while the
NRO reduction. Further comparisons between PRECM and other earlier minimum value has been achieved by PRECM with an enhancement
routing algorithms such as EMHR with BwEst [38], EERD [27], and estimated by 7% for PRECM regarding the overall consumed power.

15
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Table 8
Network setting for all algorithms used in our experiments.
Algorithm Network Scale Packet size Mobility Model Propagation Model MAC Specification Simulation Tool

EMHR with BwEst [38] 15–100 512 Manhattan Nakagami IEEE 802.11p NS3
EERD [27] 10–100 512 Manhattan Shado IEEE 802.11p NS2
SFSR [17] 50–150 512 – – IEEE 802.11p NS2
Enhanced AODV [45] 10–100 512 Manhattan Nakagami IEEE 802.11p NS2
Our Proposed PRECM 10–100 512 Manhattan Nakagami IEEE 802.11p NS2

Fig. 9. The performance of PRECM in term of NRO is tested in Figure (a). The efficiency of PRECM against EMHR with BwEst, EERD, and AODV vs. the VANET scale
is tested in Figure (b).

Fig. 10(a) and (b) show the improvement that took place in the packet loss in VANET with close scores recorded by EMHR with BwEst.
consumed power rate achieved by PRECM because of setting SDN and Values recorded by PRECM, EMHR, and BwEst were significant
fog server nearest to the requests of the vehicles. Localized data pro­ compared with those scored by EERD [27] and AODV [45]. Eventually,
cessing and the minimum rate of data transmission along with an effi­ PRECM outperformed other state-of-the-art algorithms such as EMHR
cient computing model can be guaranteed by adopting the fog with BwEst [38], EERD [27], and enhanced AODV [45] to introduce a
computing concept with SDN. The PRECM has provided the minimum data computing system with minimum power consumption.

16
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 10. The performance of PRECM in term of power consumption is tested in Figure (a). The efficiency of PRECM against EMHR with BwEst, EERD, and AODV vs.
the VANET scale is tested in Figure (b).

7.2.3. Impact precm in terms of network throughput was represented on the y-axis as can be seen. Local data processing
An evaluation of the effectiveness of the proposed routing algorithm operations of PRECM raise the likelihood of keeping the bandwidth
considering link quality and network bandwidth is going to be demon­ controllable all the time. Thus, PRECM’s quality in terms of the network
strated by monitoring the network throughput. As shown in Fig. 11(a), throughput has outperformed that achieved by other traditional algo­
vehicles’ speeds varying from 15 to 85 Mph are represented on the x-axis rithms. It was observed that SFSR was relatively the closest algorithm to
by adjusting the network bandwidth at 10 MHz and representing the PRECM regarding the network throughput. EMHR with BwEst has been
network throughput on the y-axis. The obtained results revealed PRECM ranked in third place due to the exercised bandwidth estimation module
has outperformed other state-of-the-art vehicular systems, hence this that goes to define the packet size in advance, which leads to successful
maintains the network bandwidth controllable. High link quality which throughput packet normalization. Finally, the last rank has been occu­
guarantees the lowest interference rate provided by PRECM is the main pied by the EERD. Owing to its capability to provide packet size control
reason behind this outstanding performance of PRECM. A comparative and high-quality network links, PRECM outperformed other algorithms
analysis between PRECM versus EMHR with BwEst [38], SFSR [17], and in terms of network throughput despite not introducing techniques for
EERD [27] in terms of the average throughput of the network can be packet normalization.
found in Fig. 11(b). The scale of the VANET network (ranging from 10 to
100) was represented on the x-axis, while average network throughput

17
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 11. The performance of PRECM in terms of Throughput is tested in Figure(a). The efficiency of PRECM against EMHR with BwEst, SFSR, and EERD vs. the
VANET scale is tested in Figure (b).

7.2.4. Impact precm in terms of on delivery and loss of packets represented by y-axis in Fig. 12(b), however packet loss ratio (Pkt/sec) is
It is crucial to demonstrate packet delivery success with the lowest represented by y-axis in Fig. 12(d). The proposed algorithm affects the
loss ratio while evaluating the dependability of PRECM data trans­ vehicular network positively and outperforms earlier algorithms such as
mission. Fig. 12 approximates the performance of PRECM with other EMHR with BwEst [38], SFSR [17], and enhanced AODV [45] regarding
algorithms in terms of packet delivery and packet loss ratio. The x-axis in data delivery and packet loss. The excellence of PRECM over the
Figs. 12(a) and (c) depicts the range of vehicle speeds from 15 to 85 mentioned algorithms has been achieved owing to three reasons. Firstly,
Mph, while the y-axis shows the average data packet delivery (Pkt/sec) service provider identification is brought by the supported selection
in Fig. 12(a), however, it shows the packet loss ratio (Pkt/sec) in Fig. 12 methodology which causes enhanced oriented vehicular services. Sec­
(c). Fig. 12(a) shows a comparison between the network system before ondly, frequent monitoring for the entire set of the registered nodes in
and after applying PRECM. As can be noticed, the link quality and the the local zone by SDN as a routing method adopted by PRECM enhances
method of selection adopted by PRECM lead to the outstanding perfor­ the decision-making process by the routing algorithm in PRECM.
mance of PRECM over the conventional vehicular system in terms of Thirdly, link quality with low interference by PRECM enhances the
successful data packet delivery that works on avoiding data packet loss. usage of the network bandwidth. It can be deduced that through the
Fig. 12(b) and (d) show a comparison between our proposed algo­ conducted experiments, PRECM achieved the best data delivery without
rithm and other state-of-the-art algorithms for a crystal view of the effect loss due to its network bandwidth maintainability versus other tradi­
of PRECM. The range of vehicles number is represented by x-axis in both tional algorithms. As well, it is worth mentioning that PRECM has a
Fig. 12(b) and (d), while packet delivery ratio (Pkt/sec) number is stable performance against vehicle speed variations.

18
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 12. The performance of PRECM in terms of on delivery packet is tested in Figure(a). The efficiency of PRECM against EMHR with BwEst, SFSR, and EERD vs. the
VANET scale is tested in Figure (b).

7.2.5. Impact PRECM in terms of delay time BwEst [38], SFSR [17], and AODV [45]. Computing E2E delay time is
The final assessment of the effectiveness of PRECM can be performed counting on the delay time average for all the data packets that have
by evaluating network latency as a crucial performance metric. The crossed the journey from the source to the destination successfully.
average time consumed for transmitting the data packet between two Hence, packet loss ratio and packet delivery are the judging factors for
nodes over the network can be considered as the E2E delay time. As well obtaining this metric which illustrates the outperforming performance
as the consumed time for sending a packet of data from the source node of PRECM regarding the delay time rather than other compared algo­
and reaching the destination successfully with all the possible delays rithms. As shown in Fig. 13(b), PRECM was achieved a significant
because of processing, reservations, and acquisition of the available improvement in terms of the E2E delay time that estimated by 7%.
routes in the network or even the buffering [46]. As shown in Fig. 13(a), EMHR with BwEst [38] comes in the second order with an enhancement
the speed of the vehicles varying from 15 to 85 Mph has been repre­ estimated by 6% on SFSR [17], and AODV [45]. As can be noticed,
sented on the x-axis. The average E2E delay time or latency for trans­ PRECM has outperformed EMHR with BwEst, SFSR, and enhanced
mitting the packets of data has been represented on Y-axis. Fig. 13(a) AODV in terms of latency reduction. Furthermore, PRECM introduced
displays a clarification of the performance of VANET before and after the stable performance with network scalability changes.
adopting PRECM. PRECM was superior to other conventional method­
ologies in terms of transmitting the data packets over the network due to 8. Conclusion
its ability to adjust the packet size and the communication channel
improvements resulting from providing high link quality. Despite of ST brought to urban transportation systems, challenges
VANET scale varying from 10 to 100 nodes has been represented on including power consumption, network bandwidth, and response time
the x-axis, while the average E2E delay is represented on the y-axis in have been appeared due to adopting modern information and commu­
Fig. 13(b). For validating the performance of PRECM algorithm, a nication technologies to upgrade their services. This paper presents a
comparative study was conducted between PRECM and EMHR with real-time integrated dynamic framework for vehicle interaction that

19
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Fig. 13. The performance of PRECM in terms of E2E delay time is tested in Figure (a). The efficiency of PRECM against EMHR with BwEst, SFSR, and AODV vs. the
VANET scale is tested in Figure (b).

supports the alliance between fog computing and SDN technologies with link quality in terms of network throughput, data packet delivery, and
an effective path detection facility. Moreover, it introduces an energy- power consumption along with controlling the normalized routing
optimized routing algorithm with QoS awareness for taking control of overhead. As a result, PRECM affects packet loss rate and data packet
periodic fragmentation, shorter network lifetime, and network band­ delivery and enhances the entire consumption of power exceeding other
width usage across vehicular networks. An entire improvement in the traditional algorithms by a percentage estimated by 62% to 70%.
real time performance with controlling the data dissemination rate in Additionally, the proposed routing protocol is a dual-phase protocol
VANET is achieved successfully owing to the proposed approaches. The with 90% of SDN data packet delivery ratio and 82% of SDN data loss
experimental results show that the proposed DVA-SDNF framework has reduction. Therefore, when the SDN fails to deliver packets, the pro­
an outstanding performance that gets over the mentioned challenges by posed position-based routing handles them as a parallel mechanism of
adopting the concept of integration between SDN and fog technology SDN.
which efficiently adjusts the data transmission rate. The performance of Future work will explore the modification of our proposed frame­
the DVA-SDNF framework has been verified by conducting a set of work to consider increasing network performance by applying machine
experiments. learning and deep learning on a real dataset. Moreover, increasing the
Another contribution in this study has been introduced by imple­ network scale with different values of packet size (i.e., 127, 512, 265,
menting a new routing algorithm called PRECM that employs an effi­ 1024) are also required to make the proposed framework more effective.
cient mathematical model ensuring an effective performance for VANET

20
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Data availability [26] Z.H. Ali, M.M. Badawy, H.A. Ali, A novel geographically distributed architecture
based on fog technology for improving vehicular ad hoc network (vanet)
performance, Peer-to-Peer Netwo. Appl. 13 (5) (2020) 1539.
Data available on request due to ethical restrictions. The QoS metrics [27] Z.H. Ali, H.A. Ali, Eeoma: end-to-end oriented management architecture for 6 g-
that are used in this paper are available on the following public link: enabled drone communications, Peer-to-Peer Netwo. Appl. (2022) 1–23.
https://github.com/Z-HAli/Awk-Files. [28] A. Kumar, R. Krishnamurthi, A. Nayyar, A.K. Luhach, M.S. Khan, A. Singh, A novel
software-defined drone network (sddn)-based collision avoidance strategies for on-
road traffic monitoring and management, Vehicular Commun. 28 (2021), 100313.
[29] M.T. Abbas, A. Muhammad, W.C. Song, Sd-iov: sdn enabled routing for internet of
Declaration of Competing Interest vehicles in road-aware approach, J. Ambient. Intell. Humaniz. Comput. 11 (3)
(2020), 1265.
[30] M. Zhu, J. Cao, D. Pang, Z. He, M. Xu, Sdn-based routing for efficient message
The authors declare that they have no known competing financial propagation in vanet pp. 788–797 (2015).
interests or personal relationships that could have appeared to influence [31] K.A. Darabkh, B.Z. Alkhader, K. Ala’F, F. Jubair, M. Abdel-Majeed, Icdrp-f-sdvn: an
the work reported in this paper. innovative cluster-based dual-phase routing protocol using fog computing and
softwaredefined vehicular network, Vehicular Commun. 34 (2022), 100453.
[32] N. Noorani, S.A.H. Seno, in: Proceedings of the 8th International Conference on
References Computer and Knowledge Engineering (ICCKE), IEEE, 2018, pp. 339–344.
[33] N.B. Truong, G.M. Lee, Y. Ghamri-Doudane, Software defined networking-based
vehicular adhoc network with fog computing pp. 1202–1207 (2015).
[1] Z.H. Ali, H.A. Ali, Towards sustainable smart iot applications architectural
[34] J. Pushpa, P. Raj, Fog Computing, Springer, 2018, pp. 107–130.
elements and design: opportunities, challenges, and open directions,
[35] S. Tomovic, K. Yoshigoe, I. Maljevic, I. Radusinovic, Software-defined fog network
J Supercomput (2020) 1–58.
architecture for iot, Wirel. Personal Commun. 92 (1) (2017) 181.
[2] M.M. Islam, M.T.R. Khan, M.M. Saad, D. Kim, Software-defined vehicular network
[36] K. Liu, K. Xiao, P. Dai, V. Lee, S. Guo, J. Cao, Fog computing empowered data
(sdvn): a survey on architecture and routing, J. Syst. Archit. (2020), 101961.
dissemination in software defined heterogeneous vanets, IEEE Trans. Mob.
[3] Y. Chen, A. Ardila-Gomez, G. Frame, Achieving energy savings by intelligent
Comput. (2020).
transportation systems investments in the context of smart cities, Transp. Res. Part
[37] J. Khoury, H. Sami, H. Safa, W. El-Hajj, in: Proceedings of the 15th International
D Transp. Environ. 54 (2017) 381, https://doi.org/10.1016/j.trd.2017.06.008.
Wireless Communications & Mobile Computing Conference (IWCMC), IEEE, 2019,
[4] P. Shah, T. Kasbe, A review on specification evaluation of broadcasting routing
pp. 1198–1203.
protocols in vanet, Comput. Sci. Rev. 41 (2021), 100418.
[38] L.F. Mohaisen, L.L. Joiner, Interference aware bandwidth estimation for load
[5] G. Abbas, S. Ullah, M. Waqas, Z.H. Abbas, M. Bilal, A position-based reliable
balancing in emhr-energy based with mobility concerns hybrid routing protocol for
emergency message routing scheme for road safety in vanets, Comput. Netw. 213
vanet-wsn communication, Ad Hoc Netw. 66 (2017) 1.
(2022), 109097.
[39] N. Gupta, M.S. Maashi, S. Tanwar, S. Badotra, M. Aljebreen, S. Bharany,
[6] R.E. Mohemed, A.I. Saleh, M. Abdelrazzak, A.S. Samra, Energy-efficient routing
A comparative study of software defined networking controllers using mininet,
protocols for solving energy hole problem in wireless sensor networks, Comput.
Electronics 11 (17) (2022) 2715 (Basel).
Netw. 114 (2017) 51.
[40] M. Awais, M. Asif, M.B. Ahmad, T. Mahmood, S. Munir, in: Proceedings of the
[7] R.K. Yadav, R. Mahapatra, Energy aware optimized clustering for hierarchical
Mohammad Ali Jinnah University International Conference on Computing
routing in wireless sensor network, Comput. Sci. Rev. 41 (2021), 100417.
(MAJICC), IEEE, 2021, pp. 1–6.
[8] A. Ghaffari, Hybrid opportunistic and position-based routing protocol in vehicular
[41] T.L. Marzetta, B.M. Hochwald, Fast transfer of channel state information in
ad hoc networks, J. Ambient. Intell. Humaniz. Comput. 11 (4) (2020) 1593.
wireless systems, IEEE Trans. Signal Process. 54 (4) (2006) 1268.
[9] B.S. Kim, S. Ullah, K.H. Kim, B. Roh, J.H. Ham, K.I. Kim, An enhanced geographical
[42] Z.H. Ali, H.A. Ali, M.M. Badawy, A new proposed the internet of things (iot)
routing protocol based on multi-criteria decision making method in mobile ad-hoc
virtualization framework based on sensor-as-a-service concept, Wirel. Personal
networks, Ad Hoc Netw. 103 (2020), 102157.
Commun. 97 (2017) 1419–1443, https://doi.org/10.1007/s11277-017-4580-x.
[10] G. Paliwal, S. Taterh, A topology based routing protocols comparative analysis for
[43] P.B. Bautista, L.U. Aguiar, M.A. Igartua, How does the traffic behavior change by
manets, Int. J. Adv. Eng. Research and Science 3 (3) (2016), 258851.
using sumo traffic generation tools, Comput. Commun. 181 (2022) 1.
[11] J. Bhatia, J. Dave, M. Bhavsar, S. Tanwar, N. Kumar, Sdn-enabled adaptive
[44] T. Issariyakul, E. Hossain, Introduction to Network Simulator NS2, Springer, 2009,
broadcast timer for data dissemination in vehicular ad hoc networks, IEEE Trans.
pp. 1–18.
Veh. Technol. 70 (8) (2021) 8134.
[45] D. Sutariya, S. Pradhan, in: Proceedings of the IEEE-International Conference on
[12] M. Benguenane, A. Korichi, N. Azzaoui, in: Proceedings of the International
Advances in Engineering, Science And Management (ICAESM-2012, IEEE, 2012,
Conference on Industry 4.0 and Artificial Intelligence (ICIAI 2021), Atlantis Press,
pp. 575–581.
2022, pp. 158–163.
[46] M.M. Badawy, Z.H. Ali, H.A. Ali, Qos provisioning framework for service-oriented
[13] Z.H. Ali, H.A. Ali, , Eeoma: end-to-end oriented management architecture for 6 g-
internet of things (iot), Cluster Computing pp. 1–17 (2019).
enabled drone communications, Peer-to-Peer Netw. Appl. 15 (2) (2022) 1232.
[14] K. Satheshkumar, S. Mangai, Ee-fmdrp: energy efficient-fast message distribution
routing protocol for vehicular ad-hoc networks, J. Ambient. Intell. Humaniz.
Comput. 12 (3) (2021) 3877. Zainab H Ali is a lecturer at Embedded Network Systems
[15] C. Chen, L. Liu, T. Qiu, D.O. Wu, Z. Ren, Delay-aware grid-based geographic Technology Department, Faculty of Artificial Intelligence,
routing in urban vanets: a backbone approach, IEEE/ACM Trans. Netw. 27 (6) Kafrelsheikh University. In 2021, she participates in teaching
(2019) 2324. at the Faculty of Graduate Studies for Statistical Research,
[16] M. Mohammadnezhad, A. Ghaffari, Hybrid routing scheme using imperialist Cairo University. Zainab did her Ph.D. and master’s degree in
competitive algorithm and rbf neural networks for vanets, Wirel. Netw. 25 (5) telecommunication networks particularly, the vehicle trans­
(2019) 2831. portation and Internet of Things (IoT) field in 2016 and 2020
[17] N. Noorani, S.A.H. Seno, Sdn-and fog computing-based switchable routing using respectively.
path stability estimation for vehicular ad hoc networks, Peer-to-Peer Netw. Appl. GoogleScholar account: https://scholar.google.com.eg
(2020) 1–17. /citations?user=Uyr4QvkAAAAJ&hl=en
[18] A.S. Toor, A. Jain, Energy aware cluster based multi-hop energy efficient routing
protocol using multiple mobile nodes (meacbm) in wireless sensor networks, AEU-
Int. J. Electron. Commun. 102 (2019) 41.
[19] T. Mekki, I. Jabri, A. Rachedi, L. Chaari, Software-defined networking in vehicular
networks: a survey, Transactions on Emerging Telecommunications Technologies
p. e4265 (2021).
[20] Y.C. Liu, C. Chen, S. Chakraborty, in: Proceedings of the IEEE International Noha Sakr received PhD. and MSC degrees in Computer Sci­
Conference on Communications (ICC), IEEE, 2015, pp. 6559–6564. ence and Control Systems from MU (Mansoura University,
[21] N.B. Truong, G.M. Lee, Y. Ghamri-Doudane, in: Proceedings of the IFIP/IEEE Egypt) in 2016 and 2010 respectively. Since 2016 she works as
International Symposium on Integrated Network Management (IM), IEEE, 2015, a Lecturer in the Computer Science and Control Systems
pp. 1202–1207. Department in the Faculty of Engineering, Mansoura Univer­
[22] A.J. Kadhim, S.A.H. Seno, Energy-efficient multicast routing protocol based on sdn sity. Her research interests include algorithms, data structures,
and fog computing for vehicular networks, Ad Hoc Netw. 84 (2019) 68. big data applications, image processing, text mining, and in­
[23] J.C. Nobre, A.M. de Souza, D. Ros´ario, C. Both, L.A. Villas, E. Cerqueira, T. Braun, formation retrieval.
M. Gerla, Vehicular software-defined networking and fog computing: integration GoogleScholar account: https://scholar.google.com
and design principles, Ad Hoc Netw. 82 (2019) 172. /citations?hl=en&user=T3KoXTYAAAAJ
[24] K.A. Darabkh, B.Z. Alkhader, K. Ala’F, F. Jubair, M. Abdel-Majeed, Icdrp-f-sdvn: an
innovative cluster-based dual-phase routing protocol using fog computing and
softwaredefined vehicular network, Vehicular Commun. (2022), 100453.
[25] J. Pereira, L. Ricardo, M. Lu´ıs, C. Senna, S. Sargento, Assessing the reliability of fog
computing for smart mobility applications in vanets, Future Gener. Comput. Syst.
94 (2019) 317.

21
Z.H. Ali et al. Computer Networks 229 (2023) 109785

Nora Mahmoud El-Rashidy is a lecturer at Machine Learning Hesham Arafat Ali is a Prof in Computer Engineering and sys­
and Information Retrieval Department, Faculty of Artificial tems and an associated Prof. in Information systems and asso­
Intelligence, Kafrelsheikh University. She received PhD. and ciated Prof. in computer Engineering. Faculty Artificial
MSC degrees from the Faculty of Computer Science and Infor­ Intelligence, Delta University for Science and Technology,
mation, Mansoura University, Egypt in 2020 and 2016 Gamasa 35,712, Egypt. He received a BSc in electrical engi­
respectively. Machine Learning and Information Retrieval neering (electronics), and MSc and PhD in computer engi­
department, Faculty of Artificial Intelligence, Kafrelsheiksh neering and automatic control from the Faculty of Engineering,
University, Kafrelsheiksh, Egypt. Mansoura University, in 1986,1991 and 1997, respectively. He
GoogleScholar account: https://scholar.google.com was assistant professor at the University of Mansoura, Faculty
/citations?hl=en&user=Qyui5McAAAAJ of Computer Science In 1997 up 1999. From January 2000 up
to September 2001, he was joined as Visiting Professor to the
Department of Computer Science, University of Connecticut,
Storrs. From 2002 up to 2004 he was a vice dean for student
affair the Faculty of Computer Science and Information, University of Mansoura. He was
awarded with the Highly Commended Award from Emerald Literati Club 2002 for his
research on network security. Since 2003 he has been an associate professor at the Com­
puter Engineering Department, Faculty of Engineering, University of Mansoura. He is a
founder member of the IEEE SMC Society Technical Committee on Enterprise Information
Systems (EIS). He has many book chapters published by international press and about 160
published papers in international (conf. and journal). He (supervised and co- supervised)
more than 50 students for (M Sc and PhD degree). He has served as a reviewer for many
high-quality journals, including Journal of Engineering Mansoura University. Interna­
tional Arab Journal of Information technology- The International Journal of Information
Technology and Web Engineering- The Journal of Computers (JCP)- International Journal
of Information Processing and Management. His-interests are in the areas of network se­
curity, mobile agent, Network management, Search engine, pattern recognition, distrib­
uted databases, and performance analysis.
GoogleScholar account: https://scholar.google.com/citations?user=_nOkE-I
AAAAJ&hl=en

22