SDN-Based VANET Routing A Comprehensive Survey On Architectures Protocols Analysis and Future Challenges

This article has been accepted for publication in IEEE Access.
This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2024.3355313
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.Doi Number
SDN-Based VANET Routing: A Comprehensive

Survey on Architectures, Protocols, Analysis, and
Future Challenges
Nehad Hameed Hussein 1, Siaw Paw Koh 2,*, Chong Tak Yaw 2,*, Sieh Kiong Tiong 2, Senior
Member, IEEE, F. Benedict 3, Talal Yusaf 4,5, K. Kadirgama 6,7,8 and Tan Chung Hong 2
1
College of Graduate Studies (COGS), Universiti Tenaga Nasional (The Energy University), Jalan Ikram-Uniten, Kajang 43000, Selangor, Malaysia
2
Institute of Sustainable Energy, Universiti Tenaga Nasional (The Energy University), Jalan Ikram-Uniten, Kajang 43000, Selangor, Malaysia
3
Enhance Track Sdn. Bhd., No. 9, Jalan Meranti Jaya 12, Meranti Jaya Industrial Park, Puchong 47120, Malaysia
4
School of Engineering and Technology, Central Queensland University, Brisbane, QLD 4009, Australia
5
College of Engineering, Almaaqal University, Basra 61003, Iraq
6
Advance Nano Coolant‑Lubricant (ANCL), College of Engineering, Universiti Malaysia Pahang, Pekan 26600, Malaysia
7
Faculty of Mechanical and Automotive Engineering Technology, Universiti Malaysia Pahang, Gambang 26300, Malaysia
8
Centre for Research in Advanced Fluid and Processes, Universiti Malaysia Pahang, Pekan 26600, Malaysia
Corresponding author: Siaw Paw Koh ([email protected]); Chong Tak Yaw ([email protected]).
This work was supported by Tenaga Nasional Berhad (TNB) and Universiti Tenaga Nasional (UNITEN) through BOLD Refresh Publication Fund (J510050002-IC-6
BOLDREFRESH2025-Centre of Excellence)
ABSTRACT As the automotive and telecommunication industries advance, more vehicles are becoming connected,
leading to the realization of intelligent transportation systems (ITS). Vehicular ad-hoc network (VANET) supports
various ITS services, including safety, convenience, and infotainment services for drivers and passengers. Generally,
such services are realized through data sharing among vehicles and nearby infrastructures or vehicles over multi-hop
data routing mechanisms. Vehicular data routing faces many challenges caused by vehicle dynamicity, intermittent
connectivity, and diverse application requirements. Consequently, the software-defined networking (SDN) paradigm
offers unique features such as programmability and flexibility to enhance vehicular network performance and
management and meet the quality of services (QoS) requirements of various VANET services. Recently, VANET
routing protocols have been improved using the multilevel knowledge and an up-to-date global view of traffic
conditions offered by SDN technology. The primary objective of this study is to furnish comprehensive information
regarding the current SDN-based VANET routing protocols, encompassing intricate details of their underlying
mechanisms, forwarding algorithms, and architectural considerations. Each protocol will be thoroughly examined
individually, elucidating its strengths, weaknesses, and proposed enhancements. Also, the software-defined vehicular
network (SDVN) architectures are presented according to their operation modes and controlling degree. Then, the
potential of SDN-based VANET is explored from the aspect of routing and the design requirements of routing
protocols in SDVNs. SDVN routing algorithms are uniquely classified according to various criteria. In addition, a
complete comparative analysis will be achieved to analyze the protocols regarding performance, optimization, and
simulation results. Finally, the challenges and upcoming research directions for developing such protocols are widely
stated here. By presenting such insights, this paper provides a comprehensive overview and inspires researchers to
enhance existing protocols and explore novel solutions, thereby paving the way for innovation in this field.
INDEX TERMS Vehicular network, Energy, Vehicular ad-hoc network (VANET), Software defined network
(SDN), Data routing, SDVN, IoV, V2V, V2I, V2X
I. INTRODUCTION that approximately 92% of road accidents are caused due to

Nowadays, cities are becoming more crowded, with over 50% errors in human perception (such as drivers’ negligence,
of the world population existing in urban cities, and this inadequate surveillance, and distraction) or human decision-
number is expected to rise. Despite the implementation of new making (such as excessive speed, delayed reactions, and
traffic regulations and safety measures by governments, the misjudgment of safety distance) [1]. Despite the advancements
rate of improvement in road safety remains low. Studies show in in-vehicle safety technologies such as anti-locking braking
VOLUME XX, 2017 1
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
systems (ABS), seat belts, airbags, and rear-view cameras, network by minimizing the number of broadcasted control
many people still lose their lives due to road traffic accidents messages to find the routing paths [10].
yearly [2]. Most of these traffic accidents can be alleviated by By providing more details on the data routing under the
realizing ITS, empowering many road traffic management and SDN-based VANET model, this review aims to contribute to
safety services [1], [2], [3]. the understanding and advancement of this evolving research
Vehicular Ad-hoc Networks (VANETs) are a crucial area by shedding light on the underlying mechanisms,
component of ITS, and they are expected to significantly algorithms, and architectural considerations of such protocols
improve road safety and enhance the driving experience. and discovering their robustness and limitations, and
VANETs are wireless networks formed spontaneously by suggesting the appropriate improvements decisions. This is
vehicles to facilitate data exchange among vehicles, roadside what will be presented in this review.
units (RSUs), and pedestrians through message forwarding
over multi-hop inter-vehicular communication. VANET can A. MOTIVATIONS
provide several safety-related applications that help prevent Nowadays, more motivation towards developing autonomous
accidents and reduce traffic congestion. Also, it can support vehicles (AVs) is reinforced by the possibility of saving lives and
many infotainment services, such as weather conditions, video reducing collisions and fatal accidents. Autonomous vehicles
streaming, and nearby parking zones [4]. use embedded sensors with intelligent systems to observe their
Based on the routing protocol, vehicles can form a mobile surroundings and control their movement autonomously.
communication network, allowing various services to be Unfortunately, wrong decisions may be made, mainly when
deployed without using fixed networking infrastructures. undetected objects aren’t identified properly, especially when
Unfortunately, the self-organization and high node mobility they depend entirely on the embedded sensors for decision-
make the network topology changes frequently, generating a making without sharing the sensing data with neighboring
challenging task regarding data delivery and service reliability vehicles, causing fallacious results and disastrous impacts. This
[4]– [6]. Therefore, one of the main challenges of the VANET critical dependence paves VANET developers and researchers
research community is the development of routing protocols to explore and devise novel methods to ensure high data
with high adaptability to various VANET scenarios, intending reliability with minimal transmission delay. Using the current
to enhance the reliability and packet delivery ratio and decrease VANET infrastructure for enabling connected autonomous
message delay [4]. In this context, conventional VANET vehicles (CAVs) cannot meet the ambitions of deploying reliable
routing protocols have utilized either network topology services for such vehicles. SDN, as an emerging technology, can
information or geographical vehicle distribution to decide the support CAVs with more programmability and manageability.
path from source to distention [2]. Accordingly, some routing However, the subsequent inquiries need to be addressed
protocols are proposed to address the unpredictable vehicle carefully to ensure reliable deployment of SDN in the existing
mobility and network fragmentation caused by uniform VANET infrastructure:
vehicles distribution [5]. Also, many routing protocols have 1. Can the VANET infrastructure solely improve the
been suggested to improve data routing decisions by gathering connectivity among autonomous vehicles?
information about both network topology and vehicles' 2. What situations ensure the SDN role in VANET is helpful?
geographical locations. Consequently, the traditional routing 3. What is SDN's role in improving vehicular data delivery?
protocols may not effectively handle the high mobility, 4. What are the merits and demerits of different SDN
intermittent connectivity, and resource constraints inherent in architectural designs in VANET infrastructure?
VANETs. As a result, there has been a growing interest in 5. How can SDN realize highly reliable data routing solutions
exploring the potential of emerging techniques and exploiting in a high-mobility environment?
their abilities to address these challenges and enhance routing Generally, conducting a thorough examination of current
performance in vehicular communication [4]. methods and proposals and identifying their limitations and
Recently, researchers have focused on utilizing the software- shortcomings can facilitate the development of novel systems
defined networks (SDN) paradigm to improve data routing and and solutions. So, this review will comprehensively study how
increase data delivery rates. Due to its programmability and SDN technology can enhance and optimize VANET routing.
ability to provide global network management, SDN The primary objective of this study is to furnish
technology can address the multiple and unpredictable comprehensive information regarding the current SDN-based
disconnections in VANET networks and deal with the future VANET routing protocols, encompassing details of their
mobility of VANET nodes. Thus, SDN can give better underlying mechanisms, forwarding algorithms, architectural
scalability, programmability, and routing decisions than considerations, and proposed improvements. In addition, full
traditional networking systems. Utilizing SDN global network comparative analysis will be achieved to analyze the protocols
topology, network congestion, and infrastructure-less issues regarding performance, optimization, and simulation results.
are reduced substantially [7] [8]. SDN can predict the vehicles' Finally, the challenges and upcoming research directions for
geographical location and future mobility and reduce the developing such protocols are widely stated here. By presenting
routing failure caused by continuous vehicle mobility [9]. such insights, this paper can inspire researchers to enhance
Additionally, SDN can reduce the overhead of the entire existing protocols and explore state-of-the-art solutions, thereby
paving the way for innovation in this field.
6 VOLUME XX, 2017
TABLE 1
LIST OF ABBREVIATIONS
TABLE 2
COMPARISON OF SDVN-BASED VANET DATA ROUTING SURVEYS. THE SYMBOL ✓ INDICATES THAT THE TOPIC IS DISCUSSED; THE BLANK CELL INDICATES A TOPIC HAS
NOT BEEN DISCUSSED, AND THE SYMBOL ∂ MEANS A TOPIC IS SLIGHTLY COVERED.
Main Topic
Year
Reference
Protocols Details
Protocols Taxonomy
Discussion of SDN Architectures
Discussion of role of SDN in data Routing
Data Routing Optimization Metrics
Protocols Application Area
Protocols Limitations
Protocols Robustness
Proposed Improvements
Comparison of Optimization Criteria
Analysis of performance evaluation
Simulation & Testbeds Tools
Future Challenges
New Trends and Upcoming Technologies
2022 [11] ✓ ✓ ✓ ✓ ∂ VANET/V2X routing using non-learning- and learning-based approaches

[12] ✓ ✓ ✓ ∂ ✓ ✓ Data dissemination techniques in VANET
2021 [10] ∂ ✓ ∂ ✓ ∂ ∂ ✓ ∂ ✓ SDVN architecture and routing techniques
[13] ✓ ✓ ✓ ✓ ✓ ∂ ✓ QoS routing protocols in VANET.
[14] ∂ ✓ ✓ ∂ ∂ ∂ ✓ ✓ Technologies to solve routing problem in IoV
[15] ✓ ∂ ∂ ✓ ✓ Cross layer routing methods in VANET
[16] ✓ ✓ ✓ ✓ ✓ ∂ Evaluation of VANET broadcasting routing
[17] ✓ ✓ ✓ ✓ ✓ ∂ ✓ Cross‑layer design and QOS routing in IoV
2020 [18] ∂ ✓ ✓ ∂ ✓ ✓ ✓ ✓ Security and data routing in SDN-VANET
[19] ∂ ✓ ✓ ✓ ✓ ✓ ✓ Position-based data routing in fog-based VANETs
[20] ✓ ✓ ✓ ✓ ✓ ∂ ✓ ✓ Location based routing protocols in VANET
2019 [21] ✓ ∂ ∂ ✓ ∂ ✓ QoS and scalability of SDN-IoV routing protocols
[22] ✓ ✓ ✓ ✓ ✓ ✓ ∂ ✓ ✓ Performance analysis of VANET routing protocols
[23] ✓ ✓ ∂ ∂ ∂ ✓ ✓ ∂ Challenges of data routing in IoV
2018 [24] ✓ ✓ ✓ ✓ ✓ ✓ ✓ VANET geographic routing protocols
[25] ✓ ✓ ✓ ✓ ∂ ∂ Comparative analysis of position based VANET routing
[26] ✓ ✓ ∂ ∂ ∂ ✓ ✓ Data routing techniques in VANETs
[27] ∂ ✓ ✓ ✓ ✓ ∂ ∂ ✓ Challenges, applications and use cases of SDN in VANET
2017 [28] ✓ ✓ ∂ ✓ ✓ ✓ ∂ ∂ ✓ Single- and cross-layer VANET routing techniques
[29] ✓ ∂ ∂ ∂ ✓ ✓ ∂ Position-based VANET routing protocols
2016 [30] ✓ ∂ ✓ ✓ ∂ ∂ ∂ ✓ Heuristic-based VANET routing methods
[31] ∂ ✓ ∂ ∂ ∂ ✓ Architectures of routing protocols of VANET
2015- [32] ✓ ✓ ✓ ∂ ∂ ✓ Position-based routing protocols in VANET
2012 [33] ✓ ✓ ∂ ∂ ∂ ✓ ✓ ✓ IoV data routing techniques and designs
[34] ✓ ✓ ✓ ✓ ∂ ✓ ✓ Application oriented DTN VANET protocols
[35] ✓ ✓ ✓ ✓ ✓ ∂ ∂ ✓ Design and architecture of VANET routing
2023 Our ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Comprehensive study about VANET data routing using SDN Technology.
Survey
6 VOLUME XX, 2017
FIGURE 1. Survey Structure.
B. RELATED WORKS IV. SDN-based VANET design and networking: this section
In the last decade, the number of surveys related to routing details the concepts and architectures of SDN-based
protocols in VANETs has increased incredibly. In this context, VANETs to help grasp the next sections.
the researchers primarily focused on deterministic routing V. SDN-based VANET routing: SDN-based VANET
areas in their works, such as topology-based routing, positional routing protocols will be reviewed, showing their
routing, and secure routing methods. None has underlying mechanisms, forwarding algorithms,
comprehensively studied SDN-based VANET data routing advantages, limitations, application area, and potential
with complete performance analysis. However, table 1 presents improvements.
the abbreviations list that was used throughout the survey. VI. Comparative analysis: it includes a full comparisons
Table 2 compares previous surveys and our survey based on among reviewed protocols. Open challenges and
multiple comparison criteria such as technical details, SDN upcoming research directions will be provided here.
architecture, routing optimization metrics, proposals for future VII. Section VII will present the main conclusions.
improvements, optimization criteria, and so on.
D. CONTRIBUTIONS
C. SURVEY STRUCTURE In the last two decades, many surveys have been published to
Fig. 1 illustrates how the survey is organized, where the study VANET data routing from different aspects, such as
survey is structured into seven main sections; each one covers routing mechanisms, forwarding techniques, or security
different aspects: methods. However, these studies still need to thoroughly explore
I. Section I, the introduction, explains why we conducted the role of SDN in VANET data routing, which inspired us to
this work and how it can offer fresh research conduct this comprehensive survey. First, an overview of how
opportunities to researchers. Additionally, the structure of SDN can help VANET routing protocols is provided. Then, each
the survey is illustrated in this section. routing protocol will be technically discussed, highlighting its
II. Section II covers the fundamentals of VANET networks. underlying mechanism, architectural design, strengths,
It provides a comprehensive overview of VANET limitations, application area, and potential improvements.
networking and data transmission techniques. Also, this Finally, a comparative analysis with upcoming challenges will
section discusses various characteristics and challenges be presented. So, this survey extensively deliberates the SDVN
associated with VANETs. data routing, aiming to provide new researchers with a deep
III. VANET routing background: here, the main concepts and understanding of the subject. However, this survey can
the high-level classification of VANET data routing will contribute to the literature as follows:
be detailed.
6 VOLUME XX, 2017
1. A brief review of VANET from different aspects, such as wireless links to facilitate high-speed, large-capacity
data communication, services, and deployment challenges. communication between VANET nodes. RSUs are deployed at
In addition, the key concepts of VANET data routing and the intersections or some points along a road for different roles, such
conventional classification of such protocols are presented. as data dissemination, decision-making, traffic data analysis,
A detailed discussion of metrics used to evaluate the security management, and localization services [37].
performance of routing algorithms is also presented. This To be part of VANET, vehicles must include OBUs (on-board
will be input to grasp the significance of SDN with VANET units), the network devices fixed on vehicles to help them in
and its implication on the design of reliable protocols. wireless connectivity and localization services [12]. It can offer
2. A full discussion about the SDVN architecture models and wireless communication over short distances using the IEEE
their role in VANET data routing. Also, the integration of 802.11p radio technology. Also, they can incorporate other
SDN with VANET architecture will be covered here. network devices that employ different radio technologies, such
3. An in-depth discussion of the SDVN data routing with a as IEEE 802.11a/b/g/n, for data transmission purposes. In
comprehensive analysis and novel taxonomy of SDN-based addition to wireless radio access, OBUs can help in channel
VANET routing algorithms is presented. The aim is to congestion control, localization, data security, and message
qualitatively compare the different routing methods by encoding [38]. Besides, vehicles use embedded sensors and
evaluating their key features, characteristics, performance, multimedia devices to sense the surroundings and distinguish
simulation results, and limitations. In addition, the most nearby objects to avoid crashes or unexpected breaks. Still, the
appropriate areas of application and potential improvement sensing data must be shared with nearby vehicles and VANET
will be identified separately for each protocol. infrastructure [39]. Recently, vehicles have embedded human-
4. The open issues and upcoming research challenges with machine interfaces to make a cognitive VANET paradigm [40].
probable limits and solutions will be outlined and discussed. The VANET concept was initially introduced to deliver safety
This discussion is very relevant not just for reviewed messages for drivers and passengers, providing information such
protocols but also for new challenges or optimization as accident details, road safety messages, congestion
requirements. This research will aid decision-making by information, and violation warnings to allow vehicles to make
offering valuable insights into the most appropriate SDVN- alternative decisions and save time in congested traffic. Also,
based routing schemes. VANETs are developed to provide entertainment and comfort
services, such as video streaming, weather forecasts, music
II. VANET BASICS downloads, online gaming, and commercial advertisements,
A. VANET DEFINITION enabling travelers to plan their journeys more efficiently [41].
VANETs are among the most studied areas in mobile ad hoc Utilizing real-time information, VANET will provide these
networks (MANETs). It is an efficient networking solution that services through direct or multi-hop communications [41].
allows vehicles to share data among themselves and with nearby
infrastructure and pedestrians [2][3]. VANETs possess diverse B. VANET ARCHITECTURE
attributes that differentiate them from MANETs, including but In VANETs, real-time messages can be transmitted over
not limited to high node mobility, restricted mobility patterns, wireless communication through vehicles directly or by using
unpredictable network topology, and frequent battery other mediums such as handhelds, BSs, RSUs, drones, etc.
recharging. In general, VANETs can be spontaneously Accordingly, VANET wireless communication can be achieved
established among mobile vehicles, either with or without the through various types of communication links (see Fig. 2) that
need for any pre-existing infrastructure. Infrastructure- can be classified according to the communicating systems into:
independent VANETs, or self-organizing VANETs, operate 1. Vehicle-to-Vehicle (V2V): V2V links arise wirelessly
without relying on existing infrastructure such as RSUs or fixed between vehicles directly. These links are used mainly in
access points. One of the key characteristics of such VANETs is infrastructure-independent VANETs to share vehicle safety-
decentralized communication. Messages can propagate through related messages. Also, it can be used for multi-hop
the network by hopping from vehicle to vehicle, allowing the communication in infrastructure-dependent VANETs.
collaborating dissemination of critical information like 2. Vehicle-to-infrastructure (V2I): V2I links will help vehicles
emergency alerts or traffic updates. However, infrastructure- share their data with VANET infrastructure. VANETs can
independent VANETs face many challenges. Maintaining utilize V2I to share messages and help vehicles access
continuous communication links can be difficult in sparse or Internet services.
highly dynamic scenarios, affecting communication reliability 3. Infrastructure-to-infrastructure (I2I): I2I links exchange real-
and data dissemination efficiency. time data and traffic patterns among VANET infrastructures
In contrast to infrastructure-independent VANETs, in particular areas, such as vehicle mobility patterns and
infrastructure-dependent VANETs are distinctively traffic density data.
characterized by the significant presence of back-end 4. Vehicle-to-Pedestrian (V2P): Through V2P, vehicles will
infrastructure, particularly RSUs [36]. RSUs act as intermediary share data with nearby Pedestrians. V2P communications are
nodes that facilitate the connection of VANETs to external essential in avoiding or minimizing accidents resulting from
networks, including the Internet. Moreover, RSUs enable pedestrians’ mental distractions or when people are in a
establishing a hybrid routing path that combines wired and position that is hidden from the driver's vision [42].
6 VOLUME XX, 2017
5. Vehicle-to-Barrier (V2B): V2B links help to access data directly among the vehicles and UAVs over aerial
from the barriers installed on the roadside to evade run-off- communication [45].
road accidents [43]. 8. Vehicle-to-Sensors (V2S): These links provide intra-vehicle
6. Vehicle-to-Cloud (V2C): V2C helps direct communication communication between the vehicles and built-in sensors. It
between RSUs and cloud systems. It provides many services, can be realized between OBUs and embedded or roadside
such as decision-making, big data processing, and traffic sensors [46].
density prediction [44].
7. Vehicle-to-UAV (V2U): In the future, UAVs will be a
crucial part of ITS systems. Over V2U, the data are shared
FIGURE 2. Basic VANET Communications
C. VANET COMMUNICATION (ETSI) designated 30 MHz in the 6-sub-GHz band for VANET
VANET have gained importance, leading to efforts by communications [47].
academic and government entities to establish standardized As shown in Fig. 3, the bandwidth for DSRC has been
vehicular communication. In 1999, the US federal divided into seven channels, each with a 10 MHz. Six of them,
communications commission (FCC) initiated standardization by known as service channels (SCH), are utilized for transmitting
allocating 75 MHz of dedicated short-range communication safety and non-safety-related packets. In contrast, the remaining
(DSRC) spectrum for V2V and V2I communication. Similarly, channel, known as the control channel (CCH), is used to
in 2008, the European Telecommunications Standards Institute broadcast control data and critical safety services.
FIGURE 3. DSRC spectrum band channels.

6 VOLUME XX, 2017
To empower adaptability and flexibility, IEEE agreed with the participants limits the performance of data processing and
DSRC standard and named it Wireless Access in Vehicular modeling. Besides, ever-increasing security and control
Environment (WAVE) in 2003, intending to enable direct access challenges have verified the possible effect of
vehicular communications up to a range of 1 km at regular road heterogeneous networks [38]. Besides, VANET applications
speeds [47]. Although DSRC/WAVE standards offer low are designed to serve various purposes, each with its own
transmission delay in a licensed bandwidth, the high node QoS requirements. For example, safety-related applications
mobility and the dynamic nature of VANET can cause like collision alerts require low latency, typically less than
significant overhead and delay. All seven DSRC channels are 100ms. On the other hand, non-safety applications can
used by vehicles regularly for their data exchange, which means tolerate more latency, up to 500ms. This heterogeneity of
that they must compete for access to these channels to send their data flows must be considered while designing VANETs to
packets. This competition can result in network congestion, ensure each application receives the appropriate QoS to work
increased packet latency, and reduced throughput, significantly effectively [53].
degrading the overall QoS. [48]. The growing number of 5. Unlimited power and computation resources: The VANET
VANET services and the need for continuous and scalable network faces no power or storage restrictions. The
communication have led to the exploration of various types of embedded OBUs can utilize continuous and unrestricted
wireless communications, such as cellular communication power sources from vehicle batteries, thereby eliminating
(4G/LTE), ultra-reliable and low-latency communications (5G, computation power-related issues. This enables the
mmWave, and THz channels), satellite communications, and execution of various power-consuming methods, such as
cognitive radio systems [49][50][51]. intelligent models and cryptography algorithms [51]. Also, it
helps in increasing the network coverage by utilizing
D. DISTINGUISHED CHARACTERISTICS OF VANET
multiple antennas in VANET communications [54].
The design requirements of VANETs are distinct from other
6. Unbounded network size: The VANET network can cover a
MANETs. To provide services effectively, the unique
single urban area, multiple urban areas, or even large cities,
characteristics of VANET must be considered including:
meaning its geographical scope can be unlimited. Ensuring
1. Mobility variation: The VANET network comprises
that the network can handle increasing demands of data
stationary entities such as RSUs and BSs, slow-moving
traffic and expand its coverage area while maintaining
vehicles, and high-speed vehicles. This variation in node
reliable communication is a crucial problem that requires to
mobility poses significant challenges to VANETs. For
be addressed. [24]. However, the unbounded VANET
example, in high-mobility scenarios, the chance of successful
network results in more issues, such as node management,
communication between VANET nodes is slight, where the
data security and privacy, and vehicle tracking.
transmission range will be more limited [19]. In scenarios
7. Spectrum Scarcity: Recent studies on DSRC-based VANET
where the mobility of vehicles is moderate, communication
have identified reliability and scalability challenges when
can be negatively impacted by several factors, such as the
operating in large-scale, dense areas [28] [50]. The challenge
Doppler effect, frequent link breakages, and increased
of DSRC bandwidth scarcity comes from the increased
latency. These challenges can decrease the communication
demand for VANET services. However, cooperation with
quality and overall network performance [35]. Furthermore,
other network infrastructures may compensate for the
several challenges arise when the environment becomes
bandwidth scarcity but bring additional challenges such as
denser, such as more collided data, highly utilized
channel access, routing schemes, and data security [48].
bandwidth, channel fading, and signal interference. [52].
8. Environmental Effect: Unlike MANETs, VANETs operate
2. Movement restriction: In VANET, nodes mobility is
outdoors, where the surrounding environment can
influenced by the public transportation distribution, the
significantly impact electromagnetic signals [55]. Various
nature of roads, traffic density, buildings, and other
obstructions, including buildings, vehicles, and trees, may
obstacles. However, this variation can result in some
interfere with VANET signals, leading to impairments such
challenges for reliable data delivery [3].
as multipath propagation, signal shadowing, and channel
3. Highly network disconnectivity: Traffic density may vary
fading [56]. Additionally, weather conditions, such as rain,
depending on the location and daytime, with higher densities
ice, and snow, can affect the conductivity of surfaces,
at intersections, roads near offices and markets, and during
resulting in altered reflection paths that may degrade
the day. Low vehicle densities can lead to network
VANET communication performance [57]. These conditions
fragmentation, hindering communication and packet
may also lead to flooded or snow-covered roads, negatively
delivery. Vehicles with high velocity can increase VANET
impacting node distribution. Also, the presence of sandy
dynamicity, causing network fragmentation into many
grains can cause high attenuation in microwave signals,
disconnected fragments. Considering these factors when
potentially reducing the effectiveness of data transfer [58].
designing VANETs is essential to ensure continuous and
9. Data privacy: All VANET protocols have supposed the
effective communication [2].
vehicles will be part of data communication over V2V multi-
4. Heterogeneity: VANET includes various entities regarding
hop communications [24]. Due to data security and control
networking access, applications, and properties. The
access, all participating nodes need to be identified and
heterogeneity in hardware capabilities of VANET
known to VANET [59]. In this context, individuals may be
6 VOLUME XX, 2017
reluctant to share information about their vehicles or III. VANET ROUTING BACKGROUND
intended destinations to avoid potential privacy breaches. As This section provides background about the VANET routing
a result, addressing data privacy concerns in VANETs concept, the main routing optimization parameters, and the
requires a balance between protecting personal information existing metrics for evaluating VANET routing. This will help
and respecting people's privacy preferences. to understand the main ideas of VANET routing protocols before
10. Data security: Ensuring security in VANET is critical as a understanding the methodologies targeted in this study.
network breach can result in hackers taking control of
vehicles, leading to traffic errors and fatal accidents [60]. To A. DEFINITION OF DATA ROUTING
safeguard data transmission, packets must be protected from A routing protocol is a set of guidelines that govern how
tampering to prevent eavesdropping and ensure the intended nodes connect by selecting the best communication path
sender sent them. Generally, due to fundamental VANET between nodes, facilitating data exchange with minimal latency
characteristics, such as high topology changes, ensuring data and maximal throughput. This is accomplished by following
security and non-repudiation pose significant challenges predefined rules and specified constraints in the protocol [22].
[61]. Key management is also challenging in vehicular To enhance vehicular safety, the routing protocol must prioritize
communication. This can lead to a long list of revocable keys low latency and minimal packet loss when forwarding data
and significant overhead in revocation, especially when the packets. The primary challenge researchers have faced is
number of nodes is extendable [59]. designing an effective, reliable, and secure routing mechanism
11. Data Inconsistency: Generally, VANETs have used data that can address the unique characteristics and limitations of
dissemination to inform nearby vehicles about traffic VANETs with minimal overhead. This involves improving
situations and road conditions. Most VANET protocols use traditional routing protocols and adapting them in VANETs [22]
global positioning system (GPS) for localization purposes. [24].
Unfortunately, GPS systems provide inaccurate position Many researchers studied the feasibility of using different
information where the buildings and covered tunnels can MANET routing protocols in vehicular communication, such as
result in GPS signal blockage, leading to many technical destination-sequenced-distance-vector (DSDV), dynamic-
challenges [62] [63]. As expected, the subsequent VANET source-routing (DSR), optimized link state routing (OLSR), and
applications require high-efficiency systems to analyze data ad-hoc-on-demand distance-vector (AODV). Due to VANETs
and provide future decisions instantaneously. characteristics, not all routing protocols designated for MANETs
12. Hard delay constraints. The importance of the delay are appropriate for VANETs. Yet, some researchers have
constraint in VANETs depends on the application type and adapted MANET routing protocols to suit the distinct attributes
user requirements. Safety-critical applications demand low- of VANETs and developed routing protocols specifically for the
latency communication, making delay a critical factor, while vehicular environment [64] [65]. The performance analysis
non-critical services can tolerate some delay. VANET often results show that these routing models cannot give efficient
need to balance ensuring rapid communication for safety results when applied to VANETs data routing [66] [67].
applications and efficient data delivery for other services.
Routing Strategy
FIGURE 4. Taxonomy of conventional VANET data routing.

6 VOLUME XX, 2017
B. CLASSIFICATION OF VANET DATA ROUTING reactive routing protocols can save bandwidth with low memory
Due to the various architectures, requirements, and applications requirements, they may have high latency and be unsuitable for
involved in VANETs, many routing algorithms have been security applications in VANET [35]. Due to the high response
developed. However, classifying VANET routing protocols can to link failures, these protocols are suitable for large-size and
be complex, and researchers have used various methods and frequently changing VANETs [28] [71].
parameters to categorize them. In Fig. 4, we broadly classify For optimal results, hybrid protocols combine proactive and
VANET routing protocols into three main classes based on the reactive methods to reduce the overhead in proactive models and
VANET communication architecture: decrease the delay in reactive models [33]. Hybrid protocols
divide the network into multiple zones to improve the reliability
I. Routing information-based routing protocols. of the path discovery and maintenance processes. Despite their
II. Routing mechanism-based routing protocols. advantages, hybrid protocols are not better suited for high
III. Transmission strategy-based routing protocols. mobility and frequent changes in VANET topology [66] [72].
Conversely, position or geographic routing methods utilize
1) CLASSIFICATION BASED ON ROUTING INFORMATION nodes' geographical location instead of the IP addresses in the
This class is based on how the routing information will be routing method. These protocols differ from topology-based
transmitted. These protocols can be further categorized into two routing because they do not maintain a routing table or share
sub-categories: topology-based, position-based, and hybrid information about link states with neighboring nodes. Each
routing protocols vehicle must know its location and the location of its neighbors
Topology-based routing protocols rely on traditional MANET through GPS assistance or other location-determining methods
routing protocols, where information about links is maintained [68]. This helps to transmit packets directly to the destination
in a routing table and used as a basis to move packets from the without performing the route discovery process or maintaining
source vehicle to the destination. In this method, each vehicle link state information and network topology status [19] [73].
should know the network layout and use information about Thus, they can be suitable and stable for high-mobility VANETs.
available vehicles and links in the VANET to make routing Yet, the effectiveness of such protocols relies intensely on the
decisions [35]. Though these routing methods can search the best accuracy and availability of location information, which can be
possible shortest routes, they have more limitations regarding affected by weather conditions and surrounding buildings [74].
scalability, overhead, and route discovery latency [66]. So, they Generally, geographic routing protocols can be categorized into
may not be a proper choice for the high dynamic VANETs where delay tolerant networks (DTN), non-delay tolerant networks
the regular network fragmentation and partitioning demand (Non-DTN), and hybrid routing methods.
frequent re-calculation of the topology information, which can DTN can address the technical issues arising from the lack of
lead to significant overhead [22]. Generally, these methods are continuous network connectivity, such as VANET networks
divided into three categories [68] [66]: proactive, reactive, and [75]. The DTN protocol employs the store, carry, and forward
hybrid routing protocols. strategy, where each node stores data packets for a certain period
Proactive routing protocols, or table-driven protocols, enable before forwarding them to nearby access points for further
nodes to have a routing table that stores route details of all other forwarding [76]. This way, all nodes collaborate in delivering
nodes in the network. It is essential to update the table regularly data packets, allowing the network to cope with high
to keep it up-to-date with changes in the network topology and disconnectivity. However, when the vehicles maintain the
send it periodically to all neighbors [17]. Although these packets if the connection with other vehicles is lost will result in
protocols have low latency, storing unexploited paths in the increased packet delay [34].
routing table at each node will cause high network overload and Non-DTN protocols are designed for high-density VANETs
more consumed frequency, ultimately reducing overall system with continuous connectivity, assuming a sufficient number of
performance. Moreover, as the network grows, maintaining nodes are always present to facilitate successful communication
routing tables becomes increasingly complex [69]. Accordingly, [22]. These protocols utilize the basic greedy scheme, which
these routing methods are not well-suited for VANETs as they allows a node to forward its message to the nearby neighbor
are inefficiently responding to link failures [28] [69]. toward the destination. However, this method may fail if there is
Reactive or on-demand routing protocols discover the route to no nearby neighbor to the destination other than the present node
a particular destination using information about other vehicles. [77]. With the critical goals of these approaches to explore the
It does not need to maintain information about the network shortest path toward the destination and reduce the required time
topology [11]. When a route is unavailable, the sender node for packet routing, the shortest path models may not constantly
begins the path discovery process by requesting information guarantee timely forwarding, particularly in sparse VANETs
from neighboring nodes in the network. Neighboring nodes with [22] [78].
relevant information send a route reply packet back to the sender To provide more adaptability, hybrid position-based routing
node, providing information about the path. Once the sender has protocols combine DTN and non-DTN routing algorithms to
received enough information to generate a path to the destination benefit from the ability of DTN routing protocols to maintain
node, data transmission can begin along the established path network connectivity and the minimum latency caused through
[70]. Here, the flooding method will initiate more routing data transmission using non-DTN protocols [22].
overhead, leading to network clogging [22]. Even though
6 VOLUME XX, 2017
2) CLASSIFICATION BASED ON THE ROUTING MECHANISM handles intra- and inter-cluster management and data routing
The classification of routing mechanisms may be based on tasks. While such protocols can improve scalability for large-
whether they employ beacon messages as part of their routing scale VANETs, they may cause more delay and overhead in the
protocols [28]. In a VANET, vehicles periodically broadcast presence of high mobility in VANET [84] [14].
beacons to announce their presence, share status information, Broadcast-based routing protocols are generally utilized to
and update their positions. Beacon-based routing protocols use disseminate messages on road conditions, weather, and disasters,
beacon messages to exchange information among neighboring as well as for advertising and announcements [33]. These routing
nodes and update their data before transmitting packets. These solutions employ a straightforward flooding scheme in which
routing models are sender-based, where the sender can choose each vehicle resends the packet to other vehicles. This ensures
the optimal vehicle to forward data toward the destination using the packets reach all destinations but causes a higher overhead
the information gathered by the beaconing system [79]. Unlike [85]. Furthermore, further messages are broadcasted as node
data packets, beacons are typically small and can easily pass density increases, causing more collisions, increased bandwidth
through weakly connected links. However, the beacon overhead utilization, and reduced overall network reliability [83].
will mainly rise with higher traffic densities, causing channel
congestion and beacon flooding challenges. C. PERFORMANCE EVALUATION METRICS FOR ROUTING
On the other side, beaconless-based protocols are receiver- PROTOCOLS
based routing protocols. Unlike the previous class, this In VANET, along with the common metrics designated to
mechanism does not rely on exchanging beacon messages; assess routing algorithms, such as end-to-end delay (E2E),
instead, the receiving nodes choose whether to contribute to data packet loss ratio (PLR), and packet delivery ratio (PDR), there
routing. This approach reduces overhead and packet collision as are many metrics that were considered in literature when
long as the packet loss rate by eliminating the need for redundant evaluating VANET routing methods [86]. However, such
beacons to flow over the network [80]. Nevertheless, the absence metrics including:
of the beacon messages means the nodes lack information about 1. End-to-End Delay (E2E): it is the time required for a packet
their neighbors and cannot directly know the next-hop relay in to forward from its source until it reaches its destination [13].
the routing path. This results in more delay as compared to 2. Packet Delivery Ratio (PDR): PDR is determined by
beacon-based protocols. Besides, beaconless protocols are more dividing the number of packets successfully delivered to the
susceptible to multipath formation, leading to redundant packets destination by the total number of packets sent by the source.
traveling in the network [79]. 3. Packet Loss Ratio (PLR): PLR is the ratio of transmitted
packets that fail to reach their intended destination due to
3) CLASSIFICATION BASED ON TRANSMISSION STRATEGY
faults in data transmission or network congestion.
Generally, the packets can be sent in four forms: unicast,
4. Throughput: It measures the data transmission rate over a
broadcast, multicast, and geocast. Unicast routing protocols
communication channel and is typically expressed as the
transfer the packet from the sender to a unique receiver over
average number of bits delivered per unit of time.
multi-hop transmission or carry-and-forward methods [81]. The
5. Jitter: It is a metric that quantifies the variance between the
carry-and-forward scheme involves the source vehicle carrying
maximum and minimum delays experienced by packets
the packet as far as possible until it comes within the
traveling over a network. It is caused by variations in the
transmission range of a vehicle nearer to the destination [23].
queuing delay of consecutive packets, which can lead to
Generally, most topology-based routing methods are unicast
differences in packets arrival time at the destination.
protocols. On the other hand, multicast routing schemes enable
6. Routing overhead: The routing overhead refers to the
data routing from a single sender to many receivers within a
additional data and control traffic generated by a routing
particular geographic area. However, multicast routing schemes
protocol that is not part of the actual data being transmitted
are divided into geocast and cluster routing methods [23].
but is necessary for the routing process.
Geocast-based routing protocols are position-based multicast
7. Bit Error Rate (BER): The number of bit errors divided by
routing methods that send messages from the source to all
the total number of sent bits during a specified period.
vehicles within a specific geographical zone of relevance (ZOR)
8. Network load (NL): It is the proportion of nodes receiving a
[70]. The node membership is updated when outside the
duplicate copy of a packet and total hello packets required
predefined geographical area, and here, it discards the packet
for packet forwarding.
[82]. However, by directing the packet flooding, the message
9. Normalized routing load (NRL): NRL expresses the
overhead and channel congestion generated by broadcasted data
proportion of routing packets to packets that deliver to their
can be minimized. Also, network partitioning and the existence
destination, with each hop counted distinctly [87].
of undetected vehicles may delay the appropriate message
10. Normalized overhead load (NOL): The fraction of routing
forwarding in geocasting mechanism [83].
packets to successfully reached packets, indicating the extra
Cluster-based routing protocols involve clustering techniques
bandwidth used for routing packets [88].
to group vehicular nodes with similar characteristics, such as
11. Routing request ratio (RReq): This is obtained by dividing
traveling in the same direction with similar speeds. A cluster
the number of routing requests transmitted by the source
head (CH) is designated to manage the other nodes within the
vehicle by the number of routing packets received by the
cluster that are referred to as cluster members (CM). The CH
receiver node [90].
6 VOLUME XX, 2017
12. Average routing reply ratio (ARRr): The proportion of increasing communication ranges with higher transmission
routing reply packets transferred from all nodes in the power to allow every node to reach its destination without
network that act as destinations of routing requests [89]. the support of the RSUs. In contrast, the urban area is vast,
13. Average routing discovery time (ARDt): The period between crowded, and has a diverse range of vehicles. Therefore, the
transmitting a route request to a particular destination and routing method must find a path that minimizes congestion
getting a route reply from that destination [89]. in such environments. In this context, an adaptive routing
14. Link failure (LF): This metric measures the average number scheme would be preferable to adjust its operation based on
of link failures during the routing process. A low LF value the traffic conditions in real-time.
indicates the protocol successfully evades link failures [90]. 7. Predictability: It is significant for the routing protocol to have
15. Route lifetime: The average amount of time a discovered the ability to predict the traffic density or next movement of
route remains valid, showing the efficiency of the routing the vehicle to take the necessary action and avoid packet loss
method in maintaining stable and reliable routes [90]. and increased latency.
8. Dynamic and high mobility: The vehicle's mobility is
D. OPTIMIZATION PARAMETERS USED IN VANET restricted according to the road distributions and traffic rules.
ROUTING PROTOCOL So, the consideration of node mobility has a high impact on
In the literature, many optimization parameters were the overall routing efficiency.
considered in designing and developing VANET routing 9. Scalability: Due to unpredictable network size, the routing
methods. Unfortunately, even with many suggested VANET protocol should be able to employ the increased number of
routing schemes, non-existing routing protocols satisfied all the routing requests and reduce resource consumption to
optimization parameters [13], [28], [91], [92]. However, guarantee optimal paths. Also, the routing algorithm should
VANET heterogeneity and high diversity in node mobility, link be able to address dynamic challenges, including broken
disconnectivity, applications, and QoS needs are the main links, network congestion, signal interference, and other
reasons behind such challenges. This subsection introduces the factors [28].
fundamental QoS parameters that enable the routing protocol to 10. Fault Tolerance: Due to rapidly changing topology, the
select the optimal relay node or the most efficient route toward vehicles are entering or leaving the network frequently. So,
the receiver [13]. the routing protocol must be able to handle such changes by
1. Delay: The delay is the most critical parameter that routing analyzing and predicting potential route failures in advance
protocols consider. Typically, it refers to four separate terms: to prevent communication disruptions.
access delay, transmission delay, propagation delay, and 11. Multimodality: It is highly encouraged for routing algorithms
processing delay [41]. It is best to consider all these delays to to consider different specifications in making data routing
reduce the total routing delay. decisions such as link status, vehicle kinematic information,
2. Communication distance: This parameter represents the traffic density, application category, and packet priority.
physical distance between the current node and the next 12. Security enhancement: Security in VANET routing is a
forwarding node [41]. critical aspect of ensuring the integrity, confidentiality, and
3. Neighbor node discovery: A vehicle may have several one- availability of communication in these networks. VANETs
hop neighbor vehicles within its communication range [85]. are vulnerable to various security threats due to their
So, selecting the node with the most stable and reliable links dynamic and open nature, making secure routing protocols
is essential for efficient data routing. Generally, and mechanisms essential to prevent malicious nodes from
neighborhood discovery is achieved through beacon entering the network and causing fatal results.
messages during route establishment. However, selecting the 13. Context awareness: It is highly recommended to develop
periodic interval for beaconing poses challenges, as small intelligent routing solutions to deal with unpredicted
intervals increase control overhead, while large intervals situations, like driver behavior, vehicle type, sudden route
provide stale information. To ensure timely and accurate change, building distribution, and intersection status.
network status updates, the routing protocol must select a 14. Resources utilization: The network resources are shared by
suitable interval to address this tradeoff. multiple nodes simultaneously. Proper resource utilization
4. Link reliability: In vehicular communication, the ensures that resources are used judiciously and the system
measurement of link reliability is crucial for optimizing operates reliably and efficiently. The routing protocol must
routing protocols to ensure data is transmitted reliably and be aware of resource utilization and load balancing,
efficiently [93]. efficiently using available network resources to ensure that
5. Hop count (HC): By minimizing the number of hops, routing data is transmitted without congestion and delays.
protocols can reduce the latency and increase the 15. Energy optimization: Recently, electric and autonomous
effectiveness of data transmission. vehicles have been nominated as influential nodes in
6. Vehicular traffic awareness: The VANET routing protocol VANETs. Such nodes are restricted by their embedded
should be adaptable and can provide high reliability in both battery energy. Energy-aware routing protocol aims to
urban and rural environments, even in sparse traffic. The maximize the network lifetime, so it is preferred to take the
rural area may sometimes have fewer vehicles without RSUs node energy when developing new routing protocols.
deployment. Here, fewer power constraints can be used by
6 VOLUME XX, 2017
FIGURE 5. Components of Software-Defined Vehicular Network.

TABLE 3
COMPARISON OF TRADITIONAL VANET AND SDVN
IV. SDN-BASED VANET NETWORKS A. SDVN BASIC DEFINITION

In this section, an overview of SDVN will be presented. Then, SDN is an emerging network paradigm that divides the
a general description of the SDVN architectures will be detailed, network into two logical planes: control planes and data planes.
showing their structural design, advantages, and disadvantages. This architecture is designed to provide flexibility,
6 VOLUME XX, 2017
programmability, and reliability, making it easier for developers development effort, and improved compatibility across
to deploy network applications. SDN network architecture is different SDN implementations.
logically centralized, where the control units are distributed but
work as a whole [94] [95]. In the context of VANET, SDN can B. SDVN ARCHITECTURE
be leveraged to address various challenges related to resource First, SDN is planned for wire-based computer networks [27].
management and data routing, allowing reliable deployment of Implementing SDN in VANETs without any adaptation poses
different VANET applications with better access decisions [10]. several challenges because of the inherent characteristics of
[7]. It can allocate resources to critical services and applications, VANET, such as a highly mobile network and a dynamic
ensuring they have sufficient bandwidth and low-latency topology [18] [98]. The unique aspect of SDVN is that the data
connections. Also, the centralized controller can reconfigure plane consists of vehicular devices, unlike static switches
network devices and routes in real time to respond to traffic typically in traditional SDN networks [99]. However, depending
patterns, network failures, or emerging events, making vehicular on the control level of the SDN controller, SDVN architecture is
networks more agile and responsive [96]. classified into three types: centralized, hierarchical, and hybrid.
As revealed in Fig. 5, the layered architecture of SDVN In centralized SDVN architecture, the control logic flow is
managed by a central controller. The nodes in the data plane run
comprises three layers: a management layer, a control layer, and
actions based on flow rules delivered from the SDN controller,
a data layer [10]. As compared with conventional network
which prepares and distributes the rules to the intending vehicles.
architecture, where all three layers are included in the same layer
The vehicles then run actions based on the rules they receive
at each network device [27]. Table 3 presents the comparison of [94]. The SDN controller gathers status messages from the data
SDVN and VANET networks. However, the layered plane in a centralized control architecture, enabling it to create a
architecture of SDVN has the following layers and interfaces. global network topology and make informed decisions.
1. Management plane: This plane comprises various end-user However, network performance will suffer if the controller is a
applications for managing the network, enhancing its single point of failure and the vehicles cannot access the
security, traffic flow, and overall performance [10]. These controller. Fig. 6(a) provides an overview of the architecture
applications may include security tools, routing mechanisms, [10]. Scalability is also a challenge with this architecture. So,
topology management, network monitoring utilities, with the increase in vehicles, managing the growing number of
balancing tools, and more. They use the northbound interface received requests is a big challenge.
(NBI) to provide the SDN controller with logical commands Hierarchical architecture is presented to address the
to control forwarding device behavior. challenges of centralized SDN architecture by helping to make
2. Control plane: It is a core component of SDN, including the decisions hierarchically. Here, the SDN controlling is clustered
SDN controller and other modules like firewall, system into multiple controllers on physically distributed servers [100].
status, failure control, and flow tables. It offers the necessary The hierarchical SDVN architecture is designed to reduce the
services and programming interfaces to function applications load on the controller by delegating flow decisions to the end
effectively, providing a comprehensive view of the whole nodes. Vehicles initially attempt to discover routes
network [97]. SDN controller is a virtual component that independently, and if unsuccessful, they can request assistance
gathers information from data layer, such as vehicle location, from the local controller (e.g., RSUs). If the local controller
velocity, and traffic load, and transmits it to the control plane cannot find a route, the request is sent to the SDN controller to
for more processing [18]. prepare and distribute flow rules. This approach is illustrated in
3. Data plane: This layer comprises devices that forward data Fig. 6(b), which shows a different process for interacting with
packets based on instructions received from the SDN the SDN controller compared to the centralized mode. With this
controller [97]. The communication between the controller approach, vehicles can still find routes independently or through
and data plane components is achieved through an open the local controller even if the SDN controller is inaccessible or
communication protocol called OpenFlow protocol that is the network size increases. It can increase system flexibility and
responsible for securing the communication channel and improve the overall system performance. Conversely, this SDN-
sending rules and data associated with topology variations controlling architecture requires more time for route discovery
and communication formats [7]. In SDVN, OpenFlow- [101].
enabled switches consist of vehicles, RSUs, BSs, and other Finally, the hybrid SDN control architecture can address the
participant nodes. problem of the long time required for route discovery in the
4. Southbound interface (SBI): SBI is the communication hierarchal architecture model [10]. In this architecture, the
interface between the controller and infrastructure devices at control level can be adjusted to the network's specific needs. The
the data plane [97]. SBI uses OpenFlow protocol to provide controller can deploy all the flow rules or give some control to
API for data communication between OpenFlow controller the end entities. For instance, the main SDN controller can
and OpenFlow-enabled switches. control the RSUs centrally while allowing the vehicles in the
5. Northbound interface (NBI): NBI is the interface between data plane to operate hierarchically [102]. As illustrated in Fig.
the SDN applications plane and the SDN controller. The NBI 6(c), the SDN controller helps the local SDN controller and the
interface provides an abstract view of the complete system. vehicles to deploy the control flow independently without access
The NBI is not standardized yet [10]. Standardizing the NBI to the main SDN controller [98].
will bring benefits such as increased flexibility, reduced
6 VOLUME XX, 2017
FIGURE 6. SDVN controlling architectures (a) Centralized architecture. (b) Hierarchal architecture. (c) Hybrid architecture.
FIGURE 7. The basic principles of data routing in SDN-based VANETs.

6 VOLUME XX, 2017
V. SDN-BASED VANET ROUTING and distributes rules to RSUs or vehicles. The intelligence of
A. BASIC PRINCIPLE OF ROUTING IN SDVNS RSUs or vehicles is then used to forward messages. For
hierarchical networks, the local controller, typically an RSU,
Vehicular networks typically use individual routing schemes to
will address routing queries within its coverage area. If a route
send/receive data with nearby infrastructure or other vehicles
cannot be found, the query broadcasts to other local controllers
when located out of their transmission coverage range. In
until one is found that can access the destination node. To
VANET, the traditional routing process involves exchanging
manage local requests, node clustering is commonly employed
node statuses, selecting routes, and maintaining/repairing routes.
[101] [8]. The controller will determine the members of each
However, the SDVN routing process is similar with more
cluster independently [106]. Once vehicles receive the member
complexity, including further steps such as associating nodes,
list, the most stable one is chosen as CH based on different
gathering network information, selecting a routing mode,
conditions, such as location and communication stability. The
calculating the flow table, and others [8]. Using multilevel
CH is responsible for data collection and managing local routing.
knowledge and an up-to-date global overview of traffic
Firstly, the CH gathers the status information of all cluster
conditions, data routing in SDVN will be more reliable [10]. In
members and then relays that information to the controller.
an SDVN, the controller manages routing and switching, while
Secondly, the CH may act as a local controller and oversee
the vehicles will be used as data forwarding nodes [103]. The
the intra-cluster routing. For instance, the V2 vehicle represents
network is partitioned into multiple zones, and each zone can
cluster heads with two vehicles as cluster members (V3 and V4).
have its local controller that shares and updates the local
In an alternative hybrid mode, the controller can furnish vehicle
topology with the central controller that will these data to build
flow rules for a limited period. This enables vehicles to
the global topology. However, before further discussing the
determine optimal paths for routing data messages in a
existing SDVN routing schemes and revealing their
distributed manner based on the information received from the
classifications and challenges, the standard SDVN routing
controller [107].
procedure will be presented first.
The SDN controller can create the network topology by
As revealed in Fig. 7, the vehicle associates with the most
considering the quality of the links. QoS metrics are utilized to
suitable RSU to receive its status beacon and routing query. The
determine the weight of a link, which is crucial in determining
vehicle could select the optimal RSU to connect using various
the shortest path for both static graphs and VANET
measurement means. For instance, a V6 vehicle can access RSU
communication. The shortest path techniques, such as Dijkstra
#1 and RSU #2, so it can select the appropriate RSU using either
[108], Eppstein’s K-shortest [109], and Bellman-Ford and Floyd
channel state information (CSI) or accessibility.
algorithm [110], are employed to calculate the route using a
To control the network flow, the controller in each region
static graph that is constructed from previous timeslot beacons.
must gather topology data to construct the network graph. This
To maintain high data delivery rates, the SDN controller can
data can come from periodic beacon messages or prediction
generate multiple paths per source-destination pair [108] [109].
models. Periodic beacon messages include information such as
This method is designed to mitigate the shortcomings of static
velocity, location, direction, and cluster status. The status beacon
graphs in dynamic VANET, where the validity of routes
can also contain relative mobility tables for neighboring vehicles
extracted from static graphs may need to be improved for
[8]. Based on the SDVN architecture, the beacon messages can
transmitting data packets in real-time scenarios. Though the
be transmitted to the local SDN controller or directly to the
multi-path method can ensure the best performance regarding
central controller. The controller can use this topology data to
reliability and data rates, the controller overhead and highly
maintain traffic flow, react to routing queries, and deploy routing
consumed power will be a big challenge in such routing
policy. The central or local controller will handle the routing
algorithms.
process through various shortest-path algorithms. Link
prediction models are an alternative to the periodic beacon mode.
B. SDN-BASED VANET ROUTING PROTOCOLS
They help the SDN controller to build the network graph based
Using SDN in VANET routing can solve the problems of present
on limited gathered data and mobility prediction-based links
infrastructure-less VANET routing methods [7]. Using SDN
[104]. The controller can predict the next data transmission route
global network topology, packet drops, and congestion can be
using historical trajectory, the surrounding vehicles, and road
significantly reduced compared to conventional VANET
traffic status, even if there are unstable links initially. As an
networks [8]. SDVN helps predict the vehicles' location,
example, in Figure 7, the controller can predict that V10 will
reducing routing failure caused by continuous vehicle mobility
move away from the coverage area of RSU2 due to a change in
[9]. Additionally, in SDVN, only some vehicles broadcast
its direction.
beacon messages to find the route from the source to a
SDVN allows two methods of route computation: centralized
destination where the routing information does not need to be
and hybrid. The former involves the controller finding the entire
exchanged among vehicles [10]. Thus, SDN-based VANETs can
path from the sender to the receiver upon receiving a route query
also reduce the overhead of the whole network.
from the sender node. The latter, however, uses a combination
However, this section thoroughly discusses most SDVN
of ad-hoc routing and controller-aided routing to determine the
routing algorithms and divides them into novelty based on
optimal path [105]. In SDVN hybrid mode, the controller creates
different considerations, as shown in Fig 8.
the routing policy, which outlines the general routing behaviors
6 VOLUME XX, 2017
FIGURE 8. SDVN routing protocols taxonomy.

6 VOLUME XX, 2017
1) INNOVATIVE CLUSTER-BASED DUAL-PHASE ROUTING with the fast-moving nature of AVs. The BreakCR algorithm
PROTOCOL USING FOG COMPUTING AND SDVN (ICDRP- checks whether any AV in a CR is moved in the opposite
F-SDVN) [111] direction of the CR or whether the range threshold has been
ICDRP-F-SDVN is a cluster-based routing algorithm that surpassed. If the AV changes direction or the distance between
combines fog computing and SDN to meet the requirements of it and the CR exceeds the threshold, the UpdateCR algorithm
smart vehicle communication. The protocol uses H-SDVN removes the AV from its current CR and adds it to a new CR that
architecture where one central SDN controller is used for matches its direction and is closer in distance. The CombineCR
network management and fog node orchestration, and many algorithm merges CRs that share the same direction and velocity.
SDN-RSUs controllers (RSUC) are linked to the fog servers. In addition, the protocol offers a handoff strategy to transfer the
The area is partitioned into identical clusters based on predefined feature set of a specific CR from one gNB to the next gNB.
dimensions. The vehicle with the lowest mobility and the
maximum residual distance is selected as the CH for that cluster. Advantages: The protocol maintains the clusters continuously
When the cluster member wants to send a packet, it first sends it through many algorithms, ensuring high reliability and
to the CH node. The CH first checks its routing table to search if meaningfully decreasing the cost of cluster formation.
the destination is in the same cluster. If so, the CH will use the Disadvantages: Cluster maintenance will cause high
AODV protocol to deliver it. Otherwise, CH will send a routing computation overhead on the central SDN controller. Inter-
request to the fog node to invoke the SDN for packet forwarding. clusters communication is not mentioned here.
If the SDN switch can access it, it will forward the packet to the Application Area: This protocol must be implemented in a
fog node of the next cluster to deliver it to CH of the destination secure, dense VANET network with 5G coverage.
node. Otherwise, the routing request will be moved to the central Future Improvements: Trajectory prediction can increase
controller, which will use the flooding technique to search the cluster lifetime and decrease protocol computation overhead.
destination in all SDN switches. If none is responded to, the
conventional AODV algorithm will be used as a fallback 3) COOPERATIVE DATA ROUTING AND SCHEDULING
mechanism. The protocol also provides a mechanism to reduce SCHEME (CDRS) [113]
the routing overhead by lowering the packet size and the number In [113], the authors proposed a cooperative data routing and
of exchanged hello messages. If a CM departs from the cluster, scheduling mechanism, leveraging SDVN to optimize the
the CH will be informed by possessing the lifetimes of all the delivery of infotainment packets with minimum latency and
members and can recognize when they are leaving the cluster. maximum throughput. As shown in Fig. 9, CDRS protocol uses
H-SDVN architecture, with two tiers of control plane: the top
Advantages: The protocol ensures a high PDR by providing a tier located on the Internet and the lowest one at the RSU side.
fallback mechanism. The technique of optimizing overhead and RSUs will collect network information by monitoring the
mitigating broadcasts can conserve bandwidth. broadcasted safety data and then send the collected information
Disadvantages: It is not a DTN routing protocol. Also, it to the local controllers. The incremental and maximum weighted
neglects the link failure effect on the routing path. independent set (MWIS) method has been used to accomplish a
Application Areas: It is most appropriate for urban zones with computationally feasible solution while keeping near-optimal
a suitable density of vehicles. Due to its various recovery results. The incremental method prioritizes packet transmission
options, it is also ideal for sparse zones. based on their dwell times, traffic type, arrival time, and message
Future Improvements: It may be more efficient if considering type. It finds K shortest paths for each packet. Then, the feasible
link prediction and store, carry, and forward (SCF) mechanism. function is utilized to check the feasibility of each path in terms
of link existence and potential conflicts. If a feasible path is
2) SDVN-ASSISTED MIGRATING CONSIGNMENT REGION
(MICR) (SDVN-MICR) [112] found, the algorithm identifies the minimum delay possible for
In [112], S. Prathiba et al. proposed the use of SDN and the packet over that path. The path with the lowest delay and
migrating consignment region (MiCR) model to deliver safety- satisfying all constraints is nominated as the optimal path for
messages to AVs on highways via 5G-V2X communication. The data routing, which is then inserted into the network graph.
MiCR approach is based on federated K-means clustering and
uses a three-tier architecture, including a centralized SDN Advantages: The consideration of packet dwell time can
controller, gNBs/edge server, and AVs layer. First, all AVs minimize the vehicle mobility effect. It improves PDR and
provide the edge server with information regarding their minimizes packet collisions by simultaneously choosing the path
mobility. Then, the SDN controller processes the collected data for routing and scheduling the channel.
using the federated K-means clustering algorithm to group AVs Disadvantages: Packet prioritizing without considering
into consignment regions (CRs) based on their velocity and application type may lead to high packet loss and increased
direction. The SDN controller selects a seed-AV (SAV) with latency for safety-related packets.
maximum remaining energy and transmission range near the Application Area: This protocol suits a secure, dense
group's center. The SAV can reach all the other members in the environment with multiple data access requests.
cluster (MAVs) through a single V2V link. To reduce network Future Improvements: Traffic prediction can help decide the
traffic, the MiCR protocol includes BreakCR, UpdateCR, and future link dwell time changes. A routing recovery method is
CombineCR algorithms to keep clusters up-to-date and cope required in case of a link failure with the central controller.
6 VOLUME XX, 2017
PT-GROUT routing protocol aims to enhance routing

Routing Security Mobility Traffic Engineering Applications
performance and computation efficiency by integrating the
hidden Markov model (HMM) and the temporal graph. The
Regional Controller Regional Controller
HMM technique is used to predict future routing information,
Internet
Control Plane
which is then used as input for the optimal routing algorithm.
Backbone
Top Tier
The future temporal graph is built for the VANET network using
historical routing statistics and localization data. When a packet
Broadband LTE Bottom Tier is routed, a routing request is forwarded to the control plane to
RSU
find the optimal routing path and sends it back to the requesting
Local Controller Local Controller
node. Then, the corresponding vehicles are updated and start in
packet forwarding based on their flow tables. If vehicle density
is too sparse, some BSs are chosen as relay nodes to participate
Data Plane
in the routing process. Acknowledgement (ACK) messages are
used to indicate data delivery. If a vehicle did not receive an
FIGURE 9. HD-SDVN system architecture is used in CDRS. ACK from the next-hop node after a predefined period, the
4) DELAY-EFFICIENT MULTICASTING BASED ON PARKED
vehicle encounters the failure and sends an error message to the
VEHICLES, FOG COMPUTING, AND SDVN (DMPFS) [114]
controller to recalculate the forwarding path. If the controller
DMPFS protocol utilizes parked vehicles as stationary fog cannot be reached, GPSR protocol is utilized for packet routing
computing nodes with SDN to minimize transmission delay and until they are received by a node that can access the controller.
improve link stability for multicast routing. Considering the type
of vehicle that initiated the request, the incoming requests are Advantages: Incorporating temporal information in the graph
prioritized into different categories: ambulance, police, ensures that the routing paths are more efficient and realistic. It
firefighting, and regular requests. The protocol divides the area provides computational efficiency with stable QoS.
into several segments of fixed size. Each set of RSUs/BSs in a Disadvantages: Using a centralized SDN will be the single point
specific geographic area is controlled by a local SDN controller of failure; it may be overloaded with many routing requests.
to make local routing decisions and network management. The Application Areas: It is best to deploy in an urban area with
local controllers then send their data to the central SDN well-distributed roads and a suitable vehicular density.
controller. The protocol employs a joining/leaving method to Future improvements: It can provide further efficiency if it
determine whether to add or remove a vehicle from an active considers the link lifetime in routing decisions. Vehicular
multicast session. The fog computing device periodically density prediction can also improve its efficiency.
compares the velocity received from the beacon packets for all 6) SOCIAL COMPUTING INSPIRED PREDICTIVE ROUTING
vehicles within its region to determine whether the vehicle is (SPIDER) [116]
parked. DMPFS selects the optimal multicast path by SPIDER is a centralized SDVN-based routing protocol with the
considering several constraints, such as data size, resource aim to provide low-latency, reliable data routing in a dynamic
congestion, link bandwidth, relay buffer, transmission rate, and VANET environment. As shown in Fig 10, the protocol consists
network flooding. When a source node wants to multicast data, of three components: context feature mining, one-shot
it sends a unicast request to the nearby RSU/BS, which forwards prediction, and social-based optimization routing. Context
it to the local SDN controller. If all destination vehicles are feature mining is designated to decide vehicle similarity using
within the controller coverage region, it builds the multicast tree multiple factors such as direction, acceleration, speed, and
and floods flow tables to all nodes in the tree. If some movement angle. The one-shot prediction algorithm is used to
destinations are outside the coverage area, the request is directed predict link lifetime based on the contextual features known
to the central controller to search the optimal multicast routing. from the nodes. The expected link lifetime is then used in the
Advantages: Requests classification and scheduling can achieve social-based optimization routing model called time-constrained
the QoS requirements. FC can assist with mobility and enhance influence maximization (Trim) model, which uses social
location awareness in multicast routing. computing techniques to identify relay vehicles with high data-
Disadvantages: Vehicular density and node mobility are not spreading capabilities. When a vehicle wants to send data to
considered in the routing decisions. The complexity and multiple another vehicle, it forwards a routing request to the nearby
constraints in the computing of optimal routing path may cause controller with the destination vehicle information. The
high data delivery latency. controller will find the best path and send it back to the
Application Area: The protocol needs a fully covered area with requesting node to initiate multi-hop data routing along the path
high node density with some parked FC vehicles. based on the flow table. The optimal path is calculated using
Future Improvements: Using ML methods for mobility several metrics, including delay, jitter, congestion, and PDR.
prediction can enhance protocol operation. Also, data validity
models can minimize the amount of data routed. Advantages: The consideration of driver behavior in the routing
5) PREDICTION-BASED TEMPORAL GRAPH ROUTING decision makes it more proper for real-life operation.
ALGORITHM (PT-GROUT) [115]
6 VOLUME XX, 2017
Disadvantages: The whole network is affected if the centralized Disadvantages: Even if it realizes a substantial performance
SDN controller fails for any reason. Unsecured data transmission improvement, the protocol complexity will increase with the rise
can produce unreliable outcomes. in the network size.
Application Area: This algorithm necessitates a secure and Application Area: It is best for deployment in secure and well-
well-distributed urban environment for complete functionality. developed urban regions with less complex road structures.
Future Improvements: The protocol scalability and efficiency Future Improvements: Traffic density prediction can help in
can be improved by introducing traffic density prediction along avoiding sudden disconnection. Also, area segmenting with a
the practical zone and implementing routing recovery and scoring strategy can enhance the performance.
flooding mitigation mechanisms.
8) SDN EMPOWERED LOCATION AWARE ROUTING
… Application (SELAR) [118]
APP APP APP
Plane SELAR is a location-aware multipath routing protocol using 5G
and fog computing, along with the SDN paradigm, to improve
data transmission and recover connection failures. To minimize
SDVN Notrhbound Interface energy consumption, the protocol disables redundant devices
during off-peak hours, where the path satisfying all constraints
Vehicle Model
Neural and minimizing the number of active networking devices will be
Info. Param. Control
Static Dynamic Network
Plane selected as the optimal one. Based on data size and path capacity,
Graph Graph
the protocol can transfer the data over a single path or divide the
Context Model
Feature Gener.
data into multiple fragments and transfer those fragments over
Mining Social-Based multiple routes. The procedure stays active until all data requests
Real-time Optimization
information
are achieved. Then, it updates the status of each device, the path,
Prediction Routing
collection Social Context Model
and the number of activated networking devices for each data
Feature Features Routes demand. The controller uses path quality and vehicle mobility
patterns to decide whether a link will be lost and whether that
SDVN Southbound Interface link can be recovered. It classifies the connection loss as severe
or temporary failure. In the case of temporary failure, the failed
Data
vehicle waits for the link to recover. In severe failure, the nodes
Plane terminate monitoring of the present forwarding table and execute
their routing rule before the disconnection. The forecast-based
FIGURE 10. SPIDER protocol architecture.
selective routing method (FSR) is introduced, based on multiple
7) GREEDY ROUTING WITH LINK STABILITY (GLS) [117] parameters, to predict the routing failures in advance.
GLS is a link stability-constraint, intelligent fuzzy-based routing
algorithm for urban VANETs, focusing on finding a path with Advantages: Energy-ware consideration can be a practical
maximum stability and minimum latency. GLS protocol is a approach for satisfying green VANET. The use of the FSR with
semi-centralized, intersection-and hierarchical-based approach. SDN can minimize delay and increase performance.
The central controller handles a routing table that maintains the Disadvantages: The switching off for multiple nodes may
priorities of packets routed from one area to another. It selects conflict with VANET's unpredictable network size nature,
the optimal path according to link stability, where the selected where there may be a sudden extensive increase in the data
relay node is a neighbor geographically closer to the destination demands, hence decreasing scalability.
and has a stable link with the present node. To initiate packet Application Area: This protocol requires complete coverage
transmission, the SDN controller performs two algorithms: area of urban areas to work properly.
selection (AS) and relay selection (RS) algorithms. In AS, the Future improvements: Security challenges should be
controller uses fuzzy logic to determine a consecutive sequence improved. Traffic density prediction can help proactively
of areas that will be utilized for data forwarding, considering the determine the required active nodes (e.g., RSUs).
area capacity, packet success rate, and the distance between 9) PENICILLIUM REPRODUCTION-BASED ONLINE
areas. Then, using the RS algorithm, the protocol will use the LEARNING ADAPTIVE ROUTING SCHEME (POLAR) [119]
real-time traffic pattern to determine the best path over the POLAR is an adaptive routing approach for hybrid SDVNs,
selected areas. Finally, the computed path is issued to the source employing online sequential learning and swarm intelligence.
vehicle for packet forwarding to the next selected hop until the The local SDN controllers are deployed to process global
packet reaches the destination. To adapt to changes in network information and dynamically choose the best routing strategy
topology, the protocol uses a reinforcement learning (RL) model based on real-time traffic conditions and road network layouts.
to dynamically update the routing table and adjust routing Depending on the traffic scenario, it can switch between multiple
policies based on past routing experiences. routing strategies such as AODV, OLSR, GPSR, DSR, and
DSDV. A Geohash technique is used to divide the large zone
Advantages: Focusing on intersections is a practical way to into multiple grids. A penicillium reproduction algorithm (PRA)
improve connectivity between road segments. Also, using RL enhances the learning efficiency of an online sequential extreme
can avoid redundant exploration when choosing the next step. learning machine. This model is then sent to local controllers for
6 VOLUME XX, 2017
regional management. To map traffic patterns to an ideal routing Future Improvements: An intelligent model is preferred to
method, a data processing module is introduced that extracts determine the optimal sub-protocol based on network traffic.
geographical features and traffic patterns, such as node density, 11) CONTEXT-AWARE COOPERATIVE DATA SHARING IN
road capacity, and maximum road speed, and labels these data EDGE COMPUTING ASSISTED 5G-VANET (CCDEC) [122]
sets for learning decision-making models in the SDN controller. In [122], Luo et. al proposed a context-aware protocol for
Based on this model, the controller determines the performance cooperative data sharing in the mobile edge computing (MEC)-
metrics of various routing schemes to select the best one. based SDVN network. The 5G network will provide Internet
connectivity and be used as an interface to collect contextual
Advantages: Using real-time vehicular data with traffic information from mobile vehicles. Besides, the DSRC technique
patterns can increase protocol adaptability. Multi-learning will enable cooperative data sharing among nearby vehicles. The
features will help in making more optimal routing decisions. architecture of the protocol is shown in Fig 11. CCDEC protocol
Disadvantages: Multi-hop V2V communication is not uses graph theory with a balanced, greedy algorithm to distribute
discussed here. content more evenly. The protocol operates in three phases. In
Application Area: This protocol must be implemented in a
the first phase, vehicles exchange beacon messages to help them
well-organized urban area with different traffic patterns.
sense contextual information, such as cached and required data
Future improvements: Neighbor quality, link lifetime, and
items, as well as neighboring nodes and channel capacity. In the
communication stability must be considered in low-level
routing decision selection. second phase, all vehicles communicate with the BS to share
contextual information via the cellular link. In the third phase,
10) SECURE SDN-BASED ROUTING PROTOCOL (SURFER) after deciding the set of vehicles to send and the set of data items
[120] that must be sent for each sender, the selected senders will use
K. Mershad presented a secure IoV routing scheme in [120], V2I or V2V communication for content transfer.
which uses SDN controlling with blockchain technique for
packet routing securely. The protocol is an improved version of Advantages: Using conflict graphs, decreasing the search area,
the author's previous protocol (ROAMER) [121], which utilized and managing the restrictions is possible. Contextual
RSUs for geographical routing and the store, carry, and forward information can improve the protocol's adaptability.
(SCF) mechanism. The protocol has two sub-routing Disadvantages: The protocol did not consider nodes' mobility
mechanisms: SURFER-1, which uses SDN to improve data and network dynamicity, where all the nodes are supposed to
transmission over the ROAMER routing technique, and be in the same neighborhood in a time interval.
SURFER-2, which deploys SDN across the entire IoV. To Application Area: This protocol must be deployed in a highly
secure both the routing actions and data transactions, a covered area by the cellular network and VANET.
blockchain model is integrated with the high-performance Future improvements: Expanding the scope of application
blockchain consensus (HPBC) method. For each geographical scenarios to include multiple channel access is beneficial. For
zone, RSUs are clustered, with each cluster being controlled by more reliability, it is best to integrate computation offloading.
an SDN controller. Some of the RSUs are selected to form a
Internet 5G Core
blockchain network to maintain the transactions using two
Network
blockchains: a routing blockchain and a message blockchain. Remote
Each RSU holds a table of nearby vehicles, while the controller Cloud Server Edge Backhaul
of each RSU cluster records and updates a table of RSUs that are Server
managed by it. When a node needs to send a message to another
Macro BS Macro BS
one, initially, it checks if the destination is within its routing
table. If not, the packet is transferred to the nearest RSU. If the
destination is within the RSU vicinity, the RSU will forward the
packet using geographic routing or SCF methods. If not, the RSU Small-cell
Cellular BS
RSU will create a multicast request for all nearby RSUs. If no Link
reply, the request is forwarded to the SDN controller to check if
V2I
it can access the destination and create flow rules accordingly. In
SURFER-2, an optimized objective function is introduced to V2V
search for the best routing path with maximum network
connectivity, less latency, and minimum network traffic.
Advantages: The performance analysis in both urban and FIGURE 11. The CCDEC protocol architecture.
highway presented the scheme's efficiency in terms of E2E
12) SDN AND FOG COMPUTING-BASED SWITCHABLE
delay and PDR, along with scalable security solutions. ROUTING (SFSR) [123]
Disadvantages: The higher network traffic overhead is SFRS protocol combines SDN and fog computing technology to
observed here. SURFER-1 efficiency may be unreliable with facilitate data routing over more stable links. Using the
no constraints in the next relay selection.
information received by periodic beacons, the SDN-enabled
Application Area: Due to the DTN mechanism, the protocol
RSU controllers (RSUCs) will make a local routing decision.
can be deployed in urban and highway situations.
Wireless switches are deployed at every junction to facilitate
6 VOLUME XX, 2017
packet transmission through VANET communication. Fog Application Area: it is suitable in a fully connected secure
nodes collect information periodically and calculate the weight environment with an average number of requests.
of each road edge based on constraints such as road length, Future improvements: Computation offloading can help
Euclidean distance between intersections, delay, node density, when the local EC is overloaded with increased requests.
and stability. This information is delivered to the local RSUC to 14) INFLUENCE MAXIMIZATION-BASED DYNAMIC
measure streets suitability and select proper paths under different FORWARDING NODE SELECTION SCHEME IN SDVN (IM-
situations. For data routing, when a new data flow reaches the DOS) [125]
switch, it sends a request to the controller, which uses its global In [125], Zhao et al. proposed the IM-DOS routing protocol,
view of the network status to calculate the optimal greedy which integrates social communication with data routing in
routing path. Then, the decision rules are sent to switches over SDVN networks, considering timeliness and dynamics among
DSRC or Internet communication. If the stability period of the nodes to ensure stable routing paths among adjacent nodes. The
path streets is larger than the time required for data transmission, protocol consists of three algorithms: a link lifetime prediction
packets are sent via VANET. Otherwise, packets are sent via the algorithm, a subgraph generation algorithm, and a forwarding
Internet when the vehicle density is low. In that time, the algorithm. The link duration prediction method uses dynamic
controller examines streets that previously lacked sufficient parameters such as distance, speed, acceleration, location, and
vehicle density. If data transmission is possible, the controller direction to predict the link's lifetime. Then, based on the
determines a new VANET path. Otherwise, the CSF strategy is predicted lifetime and single-hop delay, the subgraph generation
used for data transmission if data routing is impossible via algorithm removes some invalid links in the entire graph. The
VANET and the Internet. last algorithm calculates the optimal path using the time-
constrained influence maximization method inspired by social
Advantages: Roadside switches can increase reliability and computing, where the nodes with the maximum influence in
ensure high PDR even with low vehicular density. terms of link timeliness and transmission probability are selected
Disadvantages: The protocol is unsuitable for less- as relay nodes. Finally, the controller will forward the routing
infrastructure VANETs such as highways. Roadside switches path to the requesting node, and each node participating in the
increase security vulnerabilities and deployment costs. route to update its routing table and forward the packets.
Application Area: The suitable area should have well-
organized roads with pre-installed required infrastructure. Advantages: Link duration prediction can improve protocol
Future improvements: Future vehicle mobility and link reliability and decrease packet loss ratio.
lifetime analysis are good options for protocol improvement in Disadvantages: Although some redundant nodes are
scalability and adaptability. eliminated by subgraph generation, it did not consider the
13) EFFICIENT ROUTING ALGORITHM (ERA) [124] routing recovery during data transmission and future hops.
An ERA is a routing protocol designed for the MEC- SDN-based Application Area: The performance evaluation shows that the
IoV paradigm to predict the shortest, stable path by forecasting protocol can improve the performance of multi-hop data
transmission in highly dynamic SDVN environments.
the nodes' trajectory using an artificial neural network (ANN).
Future Improvements: Considering the vehicle density, the
The control plane includes a centralized SDN controller and
protocol efficiency can be improved significantly.
multiple edge servers that collect real-time messages from the
vehicles for future position prediction locally. Each edge server 15) HYBRID SOFTWARE-DEFINED NETWORKING
contains multiple modules. The input manager (IM) module is GEOGRAPHIC ROUTING APPROACH (HSDN-GRA) [126]
designated to retrieve and process vehicle information. The HSDN-GRA is a clustering-based geographical routing
synchronizer module will synchronize the data into a distributed protocol. The protocol utilizes a multi-criteria approach to select
local table, including the real-time node status pattern. Finally, the most reliable relays while ensuring connection availability,
the routing module (RM) will generate the routing path by including the contact interval between nodes, the available load
considering the information received from other modules. After of each node, and the log of encountered communication errors
getting all available routing paths, the RM module will order the embedded in each cluster head. It comprises five algorithms:
paths in ascending order using the number of hops. The path with contact duration and free load calculation algorithm, cluster head
a greater lifetime is chosen for data routing. Once the optimal election algorithm, log update algorithm, geographic routing
path is selected, the flow rules are forwarded to the algorithm, and data dissemination algorithm. The protocol
corresponding nodes and RSUs in the chosen route. If the incorporates a multi-agent method, where each vehicle has two
destination vehicle is out of the edge coverage area, the request independent agents: a controller agent and a data transfer agent.
is forwarded to the SDN controller, which will use the global The first agent runs the routing and network control plane, while
network topology to calculate the optimal routing path. the second agent follows the rules given by the first agent for
Advantages: Introducing edge controllers can minimize the data transfer. The node with the highest load is designated as a
SDN controller burden by reducing the number of requests sent cluster head that will maintain an error log that records
from nodes to the SDN controller. communication anomalies during the routing process. For data
Disadvantages: using centralized SDN architecture will cause routing, first, the controller agent of the sender vehicle will check
a single point of failure. The SDN is not fully utilized here, the error log to check the link failure with the destination node.
where most routing decisions are made in EC. If so, the controller sends the information with the respective
6 VOLUME XX, 2017
packet to the data agent. Then, a reply message is transmitted determined through link lifetime and location to the nearest
when the data agent successfully relays the packet. The absence RSU. The cluster head aggregates the packets of its cluster and
of a reply message will trigger the update of the communication drops identical or invalid packets. After that, a two-stage
anomaly log of each vehicle. heuristic method is used to determine the transmission mode
through cellular or DSRC communication technology.
Advantages: The duration of inter-vehicle contacts allows for
taking proactive measures to prevent early link failures. Also, Advantages: Ensuring QoS is achieved through the formation
the use of an error log can help avoid unstable links. of clusters while communication costs are minimized by
Disadvantages: Obtaining the quantitative values of multiple balancing the tradeoff between data latency and cost.
criteria in selecting the next relay will lead to more delay with Disadvantages: Free vehicles (non-clustering) cannot access
high overload on the cluster head. the controller. The protocol did not mention any routing
Application Area: it is suitable in dense environments with recovery and maintenance mechanism.
low to medium vehicle mobility. Application Area: It is suitable on straight roads with high
Future improvements: The protocol must be simulated using cellular and DSRC infrastructure coverage.
a powerful network simulator. Cluster size optimization and Future improvements: Link lifetime prediction can enhance
routing maintenance can increase protocol efficiency. the clustering process. Adopting a robust routing maintenance
model can improve protocol efficiency.
16) ANT COLONY ALGORITHM-BASED ROUTING
PROTOCOL IN SDVN (ACA-RP-SDVN) [127]
In [127], Kong et al. designed a routing algorithm for SDN-
based urban vehicular networks, which employs the ant colony
algorithm to determine the best routing path. The algorithm
deploys forward ants to discover the complete routing path from
the source to the destination. Besides, the backward ants are
forwarded to update the path pheromone. The area is divided into
multiple sections. Each section is managed by a local SDN
controller located at the intersection to provide routing and
management rules. Instead of searching for the routing path
between the source and destination vehicles, the protocol
explores the route between two vertex intersections, as revealed
in Fig 12. When a routing request is received, the local controller
checks if the destination IP address is within the control area. If FIGURE 12. The ACA-RP-SDVN protocol [127].
so, it generates a path based on the topology location information
18) SDN-ENABLED SPECTRAL CLUSTERING-BASED
and selects nearby intersections for data forwarding. When the
OPTIMIZED ROUTING (SESCR) [129]
destination is outside the control area, the request is relayed to
Nahar et al. [129] leveraged spectral clustering and deep learning
the main controller, which will use an ant colony algorithm to
to maintain cluster stability and path selection in SDVN
find an inter-area routing link based on the destination location.
networks and reduce the impact of arbitrary node distribution. A
Finally, the computed path is sent to the local controller to
Laplacian graph is used to categorize vehicles into clusters based
initiate the routing process.
on eigenvalues. Initially, each vehicle analyzes the mobility data
of neighboring vehicles and computes its cluster head eligibility
Advantages: The issue of reaching a local maximum in the ant
score (CES), considering the velocity difference, the Euclidean
optimization algorithm is addressed by incorporating the global
distance, the adjacency value, and the weight matrix values.
network topology of SDN.
Afterward, vehicles share their CES with their neighboring
Disadvantages: The route optimality may not be guaranteed
since the best relay is defined based on a heuristic method, and vehicles. Once all CESs have been received, the node with the
ant colony optimization can be computationally intensive. maximum CES is chosen as the CH. After forming clusters, SDN
Application Area: It requires uniform-distributed utilizes the deep deterministic policy gradient (DDPG) algorithm
transportation roads with high vehicle availability. to identify the best path to the destination vehicle based on the
Future improvements: For more reliability, load balancing quality of available routes. The learning process continues at
can enhance the protocol efficiency and minimize traffic load. each forwarding node until the packet delivers its target.
Advantages: Deep learning can take sensory information as
17) TRAFFIC DIFFERENTIATED CLUSTERING ROUTING input from the practical area to provide output with
(TDCR) [128] approximations, thereby minimizing latency and overhead.
In [128], Qi et al. introduced a TDCR scheme for data collecting Disadvantages: Using a single cluster head can cause a single
and routing in a hybrid SDVN architecture to minimize cellular point of failure. The nodes' capability and buffer size are
bandwidth cost and guarantee QoS over a centralized one-hop neglected when selected as cluster heads.
clustering method. The vehicles with the same mobility pattern Application Area: Due to the focus on the performance
are grouped into a particular cluster. The head of each cluster is analysis of protocol in urban traffic circumstances, the protocol
can be deployed in such an environment.
6 VOLUME XX, 2017
Future Improvements: The protocol can be optimized by 20) SDN-BASED MULTI-ACCESS EDGE COMPUTING
utilizing link reliability and mobility prediction techniques. FRAMEWORK FOR THE VEHICULAR NETWORKS (SDMEV)
[99]
19) THREE-LEVEL ROUTING HIERARCHY BASED ON SDVN In [99], Nkenyereye et al. presented an SDN-MEC-based data
AND MOBILE EDGE COMPUTING (V-TLRH) [130] routing scheme for VANET communications. The network
In [130], Ji et al. proposed a three-level routing hierarchy in architecture consists of four layers: the forwarding layer, the
an SDN-MEC-VANET architecture. The edge devices are control layer, the MEC layer, and the access layer. Within the
distributed in network segments, and the vehicles periodically MEC layer, RSUs are deployed inside eNB to form eNB-RSU.
send status information to the closest edge device. When an edge The eNB-RSUCs act as local controllers that manage the
device receives a beacon from a vehicle without an associated network and provide the best communication modes to nodes at
edge device, it updates its management table and sends a the network edge. The eNB-RSUCs also have a communication
message to that vehicle to update the domain in its beacon packet decision module that monitors link status and implements
and notifies other vehicles and edge devices that it has found the routing decisions. If the eNB-RSUC fails to create routing rules,
target edge device for access. Meanwhile, the SDN controller is the SDN controller generates new rules and updates the
informed to record each edge device's access status. The protocol forwarding devices' flow tables. The SDMEV framework
architecture is shown in Fig. 13. The protocol supports three employs two algorithms: choose neighboring vehicles that
levels for data transmission. Level I is used if the source node receive in-vehicle messages over V2V or V2I links and update
has a forwarding entry to the destination to transmit data directly the forwarding device flow tables. Vehicular nodes in each MEC
to the destination. Level II is utilized when a vehicle is associated group are communicated with eNB-RSU via an LTE network.
with an edge device but has no route to the target destination. So, Then, the clusters are formed based on periodic Hello messages.
it uploads a request to its edge device to search for a path to the The fuzzy logic model selects the head of each MEC based on
destination. If found, the edge device uses the Dijkstra algorithm vehicle location, velocity, and SNIR metric. The CH requests in-
to calculate the shortest path and distribute the routing path to vehicle wireless access to collect data and service-based IVI
the source vehicle and all participating nodes. Level III is used if messages from cluster members. If the message belongs to the
the edge device of the requester vehicle cannot access the warning category, it sends it to neighboring vehicles via V2V
destination. In this case, the routing request is forwarded to the communication. For vehicles that cannot comply with V2V, V2I
SDN controller to calculate the route over multiple edge nodes. communication over LTE is used to disseminate the warning
The controller chooses the edge-based route and informs the message. The CH forwards the packets to the eNB-RSU to
edge devices to participate in data transmission along the path. request flow entries if they are not predefined. If the eNB-RSUC
Then, the destination edge will use the Dijkstra algorithm to find fails to create routing rules, it transfers the request to the central
the shortest path toward the intending vehicle. V-TLRH uses controller to generate new rules for forwarding packets to
vehicle mobility information to calculate the working duration vehicles. If there is signal interference among neighboring
of potential links, where the link with longer durations is vehicles, the controller states the flow decisions and sends to a
preferred for data transmission. node located near the RSU to forward the data to the destination.
Advantages: The MEC can function autonomously when a
controller fails, enhancing system resilience.
Advantages: The protocol can achieve high PDR and low E2E
Disadvantages: The protocol supposes the ability of MEC to
and computation latency compared to the baselines.
make routing decisions without intelligibility. The Dijkstra Disadvantages: The protocol neglected the impact of vehicle
algorithm may result in a high delay in routing calculation. mobility and network sparsity on routing decisions. Due to
Application Area: The protocol can be suitable for secure multi-layer architecture, it suffers high complexity.
VANETs with high node-infrastructure connectivity. Application Area: It can be adopted in high-density VANET
Future Improvements: Providing the protocol with load-
where the node distribution is sufficient for clustering.
balancing technique can increase the efficiency of edge servers. Future Improvements: Considering link reliability and node
SDN Controller
mobility can improve routing decisions and increase protocol
effectiveness.
21) TRIBRID ROUTING PROTOCOL [131]
Local EC Local EC
In [131], Liyanage et al. developed a link connectivity-aware
SDN-based routing protocol to determine the shortest and most
stable paths between source and destination nodes. The routing
framework comprises centralized and distributed routing
mechanisms. The routing scheme can handle routing requests in
unicast, broadcast, or SCF methods. In network uncertainties, the
Local EC
broadcasting-based technique is integrated with unicast routing.
The broadcasting mechanism finds an agent node based on the
last known location of the destination node to replace it in
centralized routing. Once a node with a stable path is found,
FIGURE 13. The architecture of V-TLRH protocol.
packets are forwarded to it via unicast, after that, the agent node
6 VOLUME XX, 2017
will broadcast the packets to the estimated destination location. QRA is a multi-metric geographical routing protocol that can
For sparse network conditions, the model seeks to use the SCF determine reliable and connected routing paths using
technique to deliver packets. The routing scheme follows an connectivity probability and signal-to-noise-plus-interference
incremental algorithm with a bidirectional shortest path ratio (SINR) metrics in the SDN-IoV paradigm. When the local
algorithm to discover the shortest paths and check for stability controller receives a data routing request, it will check if a route
and latency. When the packet reaches the starting node of the to the destination already exists and is updatable. If so, it will
link, the path with the minimum residual lifetime is selected to send the path to the participating nodes. If not, QRA initiates the
transmit packets until the feasibility criteria are met, depending route discovery process by sending route discovery packets and
on link lifetime, effective velocity, and transmission time. recording each traveled segment's intersection identifier and
SINR value. Then, using a modified laying chicken algorithm,
Advantages: The protocol can deal with VANET issues, such the best path is selected over the closest intersection to the
as obsolete or network information unavailability by applying destination with the highest vehicular density and greediness
SCF and broadcasting techniques. factor. The RSU computes a greediness factor based on the
Disadvantages: Due to the broadcasting technique, network closeness of neighboring intersections to the destination
congestion will increase. There is no mechanism to show the intersection. The route establishment process is re-initiated if no
role of SDN in knowing the black hauls in the network. route meets the SINR metric.
Application Area: Due to its dealing with different conditions,
it can be adopted in different VANET situations. Advantages: It aims to identify the most reliable and
Future Improvements: Using vehicle mobility prediction connected routing paths that ensure high QoS, considering
models can help predict the future network state and reliability and connectivity as key factors.
proactively select the best routing mechanism. Disadvantages: QRA efficiency is exposed in high-mobility
22) SDN-ENABLED ROUTING FOR INTERNET OF networks. The overhead is high in a sparse traffic scenario.
VEHICLES IN ROAD-AWARE APPROACH (SD-IOV) [132] Application Area: The protocol is appropriate for urban
SD-IoV is a road-aware SDN-based routing scheme with the aim VANET services that need stable data transmission.
to send data over the shortest and most reliable path. It leverages Future Improvements: The protocol can be improved by
the cellular links to share control messages between the utilizing more constraints on the routing selection such as
controller and vehicles. Edge controllers are deployed to vehicle load and link duration.
periodically gather real-time vehicle information and analyze it 24) VANET DATA ROUTING BASED ON DEEP
to remove redundant data and forward only necessary REINFORCEMENT LEARNING (VDR-DRL) [134]
information to the controller. Whenever a vehicle needs to In [134], Yang et al. suggested a data routing and distribution
transmit data, it inspects its routing table for the destination method based on a deep reinforcement learning model. The
address. If the destination is found, the vehicle uses a greedy vehicular area is divided into multiple segments, with a cluster
method to forward the data to its destination. Otherwise, a head selected for each segment. The SDN controller uses the
request is moved to the edge controller and then to the SDN neural episodic control method to select edge cluster head nodes
controller to compute the routing path and deliver it. First, the and the Q-learning algorithm to choose a gateway cluster head
SDN controller searches all routes with the help of hop count, vehicle from multiple edge heads. Each agent selects the next-
direction, and relative speed; then, it selects one path as optimal hop relay as its action, and the hops number and signal quality
if it has a vehicle density between 25-80%. After that, the between the node and the RSU determines the reward. LTE-
computed routing path is delivered to the source vehicle to based V2I data transmission is used to distribute data to the
initiate the data transmission. The routing recovery is invoked gateway cluster heads, while DSRC is used for V2V data
when the link expires before the data is fully transferred. When transmission. In the beginning, vehicles send their status
the edge control receives a failure report, it checks the type of information with signal quality factors to RSUs to compute the
failure and forwards a new route if it is within its coverage area SINR of each node. These data are then sent to the SDN
or requests a new routing from the controller. controller to choose heads nodes based on vehicle distribution,
velocity, and channel conditions.
Advantages: The cellular network can offload the VANET
from massive data transmission while ensuring its accessibility Advantages: Using multiple nodes as cluster heads can
with minimum delay requirements. increase communication reliability, especially with the
Disadvantages: Using cellular networks will bring more dynamicity of the VANET network.
challenges, such as resource access and data security. Disadvantages: The protocol neglects the effect of vehicle
Application Area: it is suitable in well-distributed urban areas mobility when selecting the cluster heads.
with uniform roads. Application Area: It can be deployed in a secure environment
Future improvements: Trajectory prediction can help in with fully covered by an LTE network.
protocol optimization. Adopting secure SDVN architecture can Future Improvements: Using routing recovery schemes can
be a good choice for protocol improvement. help with protocol improvements. Also, it requires to be tested
23) Quality of Service Aware Routing Algorithm (QRA) [133] and compared with well-known VANET protocols.
6 VOLUME XX, 2017
25) CROSS-LAYER ROUTING HANDOFF MECHANISM Advantages: It allows for fine-tuning the optimality of V2V
WITH LOAD BALANCING IN SDVN (CLHLB) [135] routes between path length and congestion rate. Only updates
In [135], Gao et al. proposed a path connectivity-based cross- to the neighboring table are sent to the central controller,
layer unicast geographic routing handoff method using SDVN to resulting in bandwidth savings.
ensure efficient data delivery at different traffic densities. The Disadvantages: it requires high computation for routing
cooperation between the V2V cross-layer and V2I routing decisions. The lack of link reliability and vehicle mobility may
occurs based on the number of backbone paths whose challenge the application of the protocol.
connectivity probability exceeds a threshold. If the connectivity Application Area: It is suitable to apply in urban areas with
probability is below the threshold, V2I routing is used; good DSRC and cellular communication coverage.
otherwise, the system switches to V2V cross-layer routing mode. Future Improvements: The congestion insensitivity value can
When a node has data to transfer, it first checks if the destination be dynamically adjusted based on the QoS needs and current
is a neighboring node. If it is, the source vehicle directly vehicular situations.
transmits the packet to the destination. Otherwise, the routing 27) NETWORK SELECTION AND DATA DISSEMINATION IN
request is transferred to the central controller to calculate the HETEROGENEOUS-SDVN NETWORKS (NSDD-SDVN) [137]
optimal path. Using the path cost function, the controller selects In [137], Chahal et al. introduced an NSDD-SDVN protocol for
the most optimal routing path based on the total rate of all next- distributing vehicular data over multiple network interfaces. For
hop nodes and the total load of all relay nodes on a backbone network selection, a two-stage single-leader multiple-follower
path. The vehicles with the highest transmission rate are selected Stackelberg game theory is employed using the application
as the relay nodes on the candidate backbone path. After requirements and network parameters such as bandwidth, cost,
choosing the relay nodes, the central controller issues the routing delay, throughput, and signal range as selection constraints. As
flows to the source, destination, and relay nodes on the chosen illustrated in Fig. 14, the network selection involves a network
path. To avoid the ping-pong effect, only the transmission rate detection manager, a priority manager, a network filter, and a
of neighbor nodes closer to the destination is compared. network selection manager. For data dissemination, the protocol
selects the path with the maximum link duration value as the
Advantages: Using the degree of road density in routing optimal one. When a node needs to route a packet, it forwards a
decisions can improve protocol efficiency and increase its routing request to the RSU. If the RSU has an entry to the
chance of application in real-time VANETs. destination in its flow table, it transfers a reply packet to the
Disadvantages: The protocol did not include mechanisms for source node; otherwise, it transmits a request packet to the
computing road density. Also, the lack of a routing recovery central controller. If the source node is outside the coverage area
mechanism will increase the routing overhead and delay. of the local controller, it sends a hello packet to its neighbors,
Application Area: It is suitable to apply in a secure, fully
then it computes the link duration for each neighbor and selects
covered VANET environment.
the one with the maximum value for data transmission.
Future Improvements: Using mobility prediction and link
lifetime estimation can improve the protocol's effectiveness. Decision Network Network
Parameters Access Detection
26) MULTI-FLOW CONGESTION-AWARE ROUTING
(MFCAR) [136] Technologies Manager
MFCAR protocol utilizes a hierarchal SDVN with graph theory Network
to find relay nodes with low congestion and short paths in the Parameters
VANET network. Each vehicle can exchange network control Throughput Available Applications
messages with the nearly SDN edge controller through a 5G-NR Signal Range Networks Priority
interface. Each vehicle maintains a list of neighboring vehicles Max. BW Filtration Manager
Delay
and shares the list information with nearby SDN controllers Cost
using delta compression. For data routing, first, the source
Net. Selection: Active
vehicle requests the SDN edge controller for data routing. If the
Applications Utility Fun. & Applications
destination vehicle is within the coverage area of the SDN edge
Requirements Game theory List
controller, it determines the best route and adjusts the forwarding
tables of all vehicles along the path. If not, the central SDN Throughput
controller is requested to find the optimal route. SDN controller Min. Cost
BW Select Best Data
employs a uniform-cost search algorithm to compute the optimal
Delay Value Dissemination
path based on congestion insensitivity and QoS requirements. Network on Select Net.
Range
Starting from the source node, the node with the minimum
objective function value is chosen as the next relay until the
FIGURE 14. The flow work of NSDD-SDVN protocol.
destination is reached. Then, the SDN controller updates the
forwarding tables and network topology image to reflect the Advantages: Using application requirements with network
effect of the recently assigned flow on network congestion. parameters can lead to efficient data delivery and maintain high
When the flow ends, the SDN controller removes its impact from QoS for different VANET services.
the network connectivity graph.
6 VOLUME XX, 2017
Disadvantages: The calculation of the utility function requires for the destination node. If so, it will compute the routing path
high computational power, which may impact the network using the local network topology image. Otherwise, the request
efficiency. Also, the data dissemination process is done will be forwarded to the main controller to search for the optimal
hierarchically, which can cause delays in the system. routing path. After that, the reply packet is delivered to all
Application Area: Due to the lack of routing maintenance and vehicles on the routing path through local controllers, limited by
recovery, it can apply with high vehicular density. a specific period and a maximum number of hops. The protocol
Future improvements: The improvement can be achieved by has routing recovery and maintenance mechanisms, where the
providing an efficient handover technique to allow the data to path is checked for repair upon link damage, and the routing
be transmitted seamlessly over multiple interfaces. discovery process is restarted if repair is impossible.
28) CENTRALIZED ROUTING SCHEME WITH MOBILITY
PREDICTION (CRS-MP) [138] Advantages: A reliable routing path is obtained with multiple
CRS-MP is a centralized unicast geo-less routing scheme that considerations such as channel conditions, link stability,
calculates the best routing path based on the global SDVN vehicle velocity, and location.
topology image and mobility prediction. The local SDN Disadvantages: With high density, the main SDN controller is
controller uses a back-propagation neural network (BANN) for burdened by the numerous steps required for route calculation.
mobility prediction. Based on the estimated node mobility, the Application Area: it is suitable in an urban city with uniform
protocol will make routing decisions and assign the transmission roads and high availability of cellular infrastructure.
method over V2I or V2V data transmission. For data routing, the Future improvements: Using data offloading techniques with
source vehicle sends routing requests containing the IP addresses edge computing can decrease the computation overhead and
delay on SDN controllers.
of both the sender and receiver vehicle to the RSU/BS. The
RSU/BS will make routing decisions when both the 30) INFLUENCE MAXIMIZATION-BASED CLUSTER
communicating vehicles are in their transmission range. ROUTING ALGORITHM FOR SDVN (IMCR) [140]
Otherwise, the request will be forwarded to the local SDN In [140], Wang et al. presented a double-head clustering
controller to make routing decisions. The SDN controller will algorithm for influence maximization in a hierarchal SDVN
use the road segment identification of the communicating nodes network to reduce overhead and improve transmission
to instruct the particular RSU/BS to select the transmission efficiency. Adjacent vehicles are divided into logical clusters.
method for each node using a bipartite matching scheme. If the The vehicle with the most influential role on all other vehicles is
source and destination vehicles are moved in different RSU designated as the primary cluster head which is responsible for
communication ranges, I2I communication will be used. Finally, collecting members' mobility information and sending it to the
if both the communicating nodes are out of the transmission SDN controller. If the primary cluster head is left, the backup
range of any RSU, multi-hop transmissions over BSs are used, cluster head immediately takes over to eliminate the issue of a
with BSs acting as intermediate relay nodes. single point of failure. The influence maximization strategy is
invoked to select the new backup cluster head. Both cluster
Advantages: Different V2V and V2I data transmission heads update information with each other so that the backup head
channels and vehicle mobility prediction can enhance QoS and can act as the primary one if necessary. SDN controller will
decrease bandwidth utilization issues. exchange the flow table and relevant vehicle information with
Disadvantages: Cellular networks will increase resource the cluster heads only if the vehicles are in two clusters. When a
access and data security challenges. vehicle that does not belong to any cluster needs to communicate
Application Area: Using DSRC and LTE, the protocol can with a specific cluster, it communicates directly with the local
adapt to urban and highway environments. controller to find the path toward the destination. The controller
Future improvements: Using edge computing can be a good replies with the flow table to the requesting node and sends it to
choice for protocol optimization. Also, adopting link reliability the cluster head of destination node to initiate data packet
when selecting the next hop relay can be a good choice. exchange by following the flow table.
29) CROSS-LAYER SDVN ROUTING PROTOCOL (CLR-
SDVN) [139] Advantages: The selection of double cluster heads can better
In [139], You et al. introduced an SDN-based cross-layer routing solve problems such as the method of re-affiliation of cluster
strategy to find the optimal path in an urban environment by and packet loss rate, especially when the existing CH fails.
leveraging the global network topology, channel status, link Disadvantages: The periodic execution of the clustering
lifetime, and vehicle mobility pattern. The network comprises process results in high control overhead. SDN's role in data
three core systems: the main SDN controller, local controllers, routing and vehicle mobility is not considered sufficiently.
and forwarding nodes. The local controller (e.g., RSUs) stores Application Area: it is preferable to be applied in dense
information about the local network topology of its coverage environments with restricted vehicle mobility.
area by maintaining a database of vehicle state information. Future Improvements: Using more constraints on cluster
Utilizing cloud computing technology, the databases of all RSUs management can increase cluster stability.
are combined in a global network state vector. For data routing, 31) HETEROGENEOUS SDVN-BASED COOPERATIVE
the source vehicle first transmits a request message to the nearest TEMPORAL DATA DISSEMINATION (CTDD) [141]
RSU. The RSU will check its database to see if it has an entry
6 VOLUME XX, 2017
In [141], Die et al. introduced a CTDD protocol for data routing a route breakdown, the vehicle requests the controller for the
in a heterogeneous, decentralized SDVN architecture, routing re-initiation process.
considering the temporal details of data, the heterogeneity of
network interfaces, and the delay limitations on service requests. Advantages: The established routes are stable, ensuring
For each service, the protocol achieves scheduling decision that reliable packet delivery.
includes broadcast rules, bandwidth allocation strategies, and Disadvantages: It does not consider scalability and traffic
routing paths. The network interfaces modify their operation heterogeneity which may result in higher overhead and latency.
rules once they get the SDN control message. Accordingly, a Application Area: Given network-wide information, the
priority-based task assignment algorithm dynamically protocol can be deployed in highly-density environment.
distributes the transmission tasks of each request over multiple Future Improvements: Adopting reliable vehicular density
interfaces. When the vehicle enters the transmission range of a prediction model can improve routing stability.
network interface, it sends a service request and mobility pattern 33) LINK AVAILABLE TIME PREDICTION-BASED BACKUP
to nearby RSU/BS, which sends it to the SDN controller for CACHING AND ROUTING (LBR) [143]
further processing. Based on node mobility and the distribution In [143], X. Yan et al. presented the LBR routing algorithm,
of network interfaces, the controller decides the set of interfaces which aims to improve the performance of flow delivery and
for each vehicle and determines the set of data units to be routed data routing performance in high-speed V2I networks. The LBR
over each one. Schedulable requests that can efficiently decrease architecture comprises two planes: the SDN control plane, which
bandwidth utilization and be satisfied before their delay utilizes an SDN controller to manage and control the entire
requirements are selected. Then, a priority function will VANET network, and the data plane, which includes various
prioritize the requests based on allocation ratio, service limit, and radio access networks and basic infrastructure like RSUs, BSs,
data productivity. Finally, the routing decision and data flows are and vehicles (see Fig. 15). SDN controller includes many
delivered to the corresponding vehicles to initiate data delivery. modules which enable it to obtain the vehicle information and
predict the link lifetime for each vehicle based on its mobility
Advantages: The utilization of request scheduling along with information. The topology management module will gather the
data routing allows for the efficient allocation of data delivery status information of vehicular nodes and connected RSUs. The
tasks with less waiting time. link availability prediction module will utilize up-to-date status
Disadvantages: The protocol results in high computation information to calculate the duration of available links. The
overhead and latency due to multi-constraint data scheduling. information awareness module is integrated with the data
Application Area: Due to network-wide information from forwarding function in RSUs to collect a dataset about network
various network interfaces, it can be deployed in urban covered metrics, channel metrics, and vehicle mobility information.
by multiple network infrastructures. Finally, the data forwarding module updates flow rules and
Future Improvements: The protocol neglected the security routing decisions based on flow tables in RSUs.
issues associated with heterogonous VANETs. Also, multi-hop
V2V data transmission is not considered here.
32) LINK STABILITY BASED OPTIMIZED ROUTING
PROTOCOL (LSB-OR) [142]
LSB-OR is a distributed SDN-based routing protocol that selects
the optimal routing path by considering both the shortest and
most stable links. The source vehicle divides the data into
multiple units and sends them through multiple shortest paths
identified by the SDN controller under link stability constraint.
The protocol employs an incremental packet allocation scheme,
which explores nodes from both source and destination nodes
until a middle one is found. The protocol then extracts the
bottleneck link of the path to estimate the remaining link
lifetimes and find the link with the lowest rate before assessing
the stability of the path. If the bottleneck link fulfills the link FIGURE 15. Data transmission in LBR protocol [143].
capacity limitation, then all links in the path successfully satisfy
the link capacity limit. The protocol then checks the next higher Advantages: Considering link duration prediction improves
the rate at which flows are delivered.
path until all data units are assigned or the maximum acceptable
Disadvantages: LBR did not give any flexibility in the multi-
paths are analyzed. If multiple paths exist with the same number
hop V2V data routing, where it merely applied in high-speed
of hops, the protocol chooses the path with less mobility. When
V2I communication.
the requested vehicle gets the flow rules, it extracts the path Application Area: Considering high-speed mobility with link
information and inserts it in the buffered packet to send to the duration prediction can maintain protocol efficiency in a
next hop. All participating nodes test the validity of the received highway environment.
path and store the path in their flow tables if positive. If there is
6 VOLUME XX, 2017
Future Improvements: ML-based solutions for link duration connected sub-region and chooses a vehicle from each sub-
prediction can provide more efficiency—the need to develop it region to act as a region head to broadcast the message in its
to allow multi-hop V2V communication in urban areas. connected sub-region via a V2V link and send it to the controller
using LTE links. When the controller gets the message, it
34) SOFTWARE-DEFINED COGNITIVE ROUTING FOR THE retransmits it to all region heads to deliver it to all vehicles within
INTERNET OF VEHICLES (SDCoR) [144]
their sub-regions.
In [144], Cheng et al. introduced the SDCoR routing protocol,
which utilizes reinforcement learning technology in SDN-based
Advantages: It can decrease the latency and overhead by
IoV network to deploy different routing protocols in various minimizing the number of broadcasting vehicles.
traffic scenarios. The protocol is based on a software-defined Disadvantages: The protocol creates high overhead on
cognitive IoV network architecture. The SDN controller network resources. WAVE standard may cope with high
comprises three modules for sensing and learning purposes: the channel congestion, especially in high-density environments.
sensing module, the learning module, and the routing decision Application Area: The protocol can be deployed in different
module. By sensing the current environment, Q-learning- urban areas with moderate vehicular density.
enabled cognitive data forwarding is implemented in the SDN Future Improvements: Intelligent models can optimize the
controller to learn the best routing strategy that can achieve number of recipient nodes, resulting in optimization the delay
efficient routing performance in a current environment based on and routing overhead.
the nodes' velocity and vehicular density. The protocol defines a
36) HIERARCHICAL ROUTING SCHEME WITH LOAD
set of actions, including GPSR and AODV protocols. The Q- BALANCING (HRLB) [101]
learning reward function employs two routing performance In [101], Gao et al. suggested an SDN-aided hierarchical routing
metrics: PDR and E2E. method to determine the optimal path based on the traffic density
and the node transfer probability from one grid to another. First,
Advantages: Instead of following a specific routing strategy, the central SDN controller will use vehicle status patterns to
this protocol switches between multiple strategies based on build a global network connectivity graph. When a source node
traffic situations.
has data to be sent, it first checks for a matching routing entry in
Disadvantages: This scheme can only select between two
its table. The data is sent directly using that routing rule if a
routing protocols, and the rise in routing requests may lead to
match is present. Otherwise, the routing request is uploaded to
a high computation overhead on the SDN controller.
Application Area: It is preferable to be applied in a vehicular the central controller to find the optimal routing path using three
environment suited for GPSR and AODV protocols. algorithms: grid selection, path selection, and relay node
Future Improvements: The actions can be improved by selection (see Fig. 16). Initially, the area is segmented into
integrating more routing strategies. Defining reward functions smaller grids, and the well-connected grids are identified using
with more specifications, such as link lifetime and routing traffic density and historical transfer probability. Next, a path
overhead, can increase protocol efficiency. cost function is used to determine the two routes with the least
cost, considering path length, traffic density, nearby node
35) GEOCAST PROTOCOL FOR SOFTWARE-DEFINED distance, and network load. Finally, for load balancing, the relay
VEHICULAR NETWORKS (GEO-SDVN) [145]
nodes with low traffic loads are selected on both paths by
In [145], Sousa et al. suggested a geocast routing algorithm that
considering the remaining buffer and distance to the receiver
utilizes LTE and WAVE standards to optimize data routing in a
node. Once the routing path is received, the node starts data
specific geographic area. The SDN controller is logically
routing while the controller monitors the route load status. If the
centralized and does not depend on the existence of RSUs.
load is more than 70%, the controller instructs the node to use
Vehicles communicate with each other over WAVE standard
the second link. If not, the SCF method is used.
and share packets with the SDN controller using LTE. Each
vehicle has a flow table with two matching fields: vehicle ID and
a geocast ID that identifies the corresponding vehicle
geographical zone. To send a message, the vehicle checks its
routing table first, if it has an entry in its table, it uses the
corresponding flow for routing the message to its nearby nodes
using a one-hop broadcast. If not, a table-miss message is sent to
the SDN controller to compute the routing path. If the response
is positive, it performs the action provided in the reply message.
If not, it discards the packet. First, the controller computes the
minimum connected dominating set (MCDS) of the graph in the
region of interest. Vehicles in the MCDS will broadcast the
packet to all vehicles in the next hop list, while vehicles not in
the MCDS will discard the packet. The next hop nodes are
determined using a depth-first search starting from the sender
vehicle. In the low-density scenario, the controller detects each
6 VOLUME XX, 2017
FIGURE 16. Data transmission in HRLB protocol [101]. offloading continues until the lifetime of the path expires or
Advantages: It can reduce latency and overhead by avoiding when the path is disrupted due to vehicle mobility. Once V2V
new routing discovery. Network congestion will be reduced by routing is completed, the communicating vehicles must return to
using a backup route instead of computing a new one. cellular communication. If the vehicle detects a broken link
Disadvantages: The distance of the SDN controller from before the data delivery is complete, it sends a repair message to
vehicular nodes makes finding efficient routes for time- the SDNi-MEC server to fix the failure. To recover the path
sensitive applications difficult. failure, a lifetime-based path recovery mechanism has been
Application Area: It is suitable in urban areas with high presented to replace the drive-away vehicle with another
vehicle connectivity. neighboring vehicle that can directly communicate with the
Future Improvements: Studying the routing with high node intended vehicle. If not, a packet is dropped, and the vehicles
mobility may require some improvements, such as using must revert to the cellular data routing from the V2V links.
mobility prediction models to know the future grid of nodes. Advantages: The protocol can maintain high network
37) SDN-ENABLED SOCIAL-AWARE CLUSTERING (SESAC) connectivity and data transmission using dual networking
[146] infrastructures for data routing.
W. Qi et al. [146] introduced an SDN-aided social-aware Disadvantages: The multiple switching between cellular and
clustering-based routing protocol for 5G-VANETs. The social VANET channels can lead to high packet loss and latency.
pattern prediction model is utilized to enhance cluster stability. Application Area: It is suitable in highway environments
The nodes' mobility is modeled as a discrete time-homogeneous where the number of cellular users seems small.
semi-Markov model, with the social pattern being presented as a Future Improvements: Mobility prediction can help produce
set of road segments and corresponding sojourn times that the proactive handover. QoS-based offloading decisions can
node will follow the next time. At first, the road is divided into balance the networking load of both networks.
rectangular areas, and vehicles are grouped based on location. 39) COGNITIVE ROUTING PROTOCOL FOR SOFTWARE-
The vehicles record the road segments they travel through and DEFINED VEHICULAR NETWORKS (CR-SDVN) [148]
transmit it to the nearby BS. The SDN controller groups nodes Ghafoor et al. introduced the CR-SDVN routing protocol to find
with similar routes into clusters and selects cluster heads stable routing paths using cognitive radio technology and
according to the distance with other nodes, relative velocity, and hierarchal SDVN architecture to select the channel and relay
vehicle attributes. The data exchange between cluster members node efficiently. The protocol allows data transmission only
is performed via V2V links, and the cluster head sends the when the source and destination agree on the transmission range
aggregated data to the BS, which includes control messages and over a common idle channel. The method is divided into two
flow rules dissemination. phases: the registering phase and the route prediction phase. In
Advantages: Vehicle clustering can reduce network the registering phase, a subset of RSUs is selected as local
congestion and improve PDR. Using vehicle patterns and controllers, arranged in a tree structure according to transmission
social awareness can increase clusters' lifetime. delay, propagation delay, number of hops, and expected number
Disadvantages: Many issues are not addressed here, such as of reachable vehicles. In the route prediction phase, the vehicle
security issues, connection loss with the central controller, and uploads the routing request to the local controller, which tries to
the impact on vehicle mobility. reply with the optimal path to the source node. If no path is
Application Area: Particularly, the protocol can work more available, the request is sent to the central controller to find the
effectively in work environments with regular vehicle traffic, best routing path. If the source vehicle is out of range of any local
such as universities or government complexes. controller, it transmits beacon messages to locate the next-hop
Future Improvements: Mobility prediction models can help relay towards the nearest controller. The source vehicle
know future nodes' mobility and lead to stable clusters.
computes the link duration prediction for each neighboring
38) LIFETIME-BASED NETWORK STATE ROUTING (LT- vehicle to identify the optimal relay node with minimum value
NSR) [147] to establish a path to the local controller responsible for data
LT-NSR protocol employs MEC in SDVN to enable traffic routing. The number of users on the road is predicted using the
offloading from the cellular network to V2V communication in energy detector scheme in the spectrum sensing mechanism.
highway VANETs. The protocol deploys the control plane of
SDN within the MEC architecture, introducing the SDNi-MEC Advantages: The selection process of local controllers can
server. This server has a context database that maintains vehicles reduce the main controller burden. A cognitive radio system
context information, such as position, velocity, and neighboring can overcome the issues of bandwidth scarcity in SDVN.
vehicles. Initially, when the source node intends to transmit Disadvantages: Multi-hop communication is not discussed
packets, it forwards a data request to the SDNi-MEC server. here. All data transfer will be based on controllers that may
Then, it checks whether a V2V routing path exists between the result in high overload with increased routing requests.
communicating nodes that have already communicated over the Application Area: The protocol can give good results when
cellular links. If multiple V2V routing paths are available, the used in minimum building environments with low crowds.
V2V routing path with the highest lifetime links is selected as Future Improvements: It requires some routing technique as
the optimal path. Finally, the peered vehicles will be informed to a backup in case of a link failure with the central controller.
use V2V data routing instead of cellular communication. V2V
6 VOLUME XX, 2017
40) FLEXIBLE ROUTE AND PROACTIVE UPDATING Disadvantages: Cloud-enabled SDN can cause high latency in
ROUTING PROTOCOL (FR-PU) [149] routing computing. Network connectivity and delayed
In [149], Yang et al. introduced the FR-PU routing mechanism transmission monitoring cause high complexity and overload.
for data routing in SDVN framework efficiently by considering Application Area: It is suited to be deployed in an urban city
multiple link factors with a proactive update scheme to improve with multiple road sections and high vehicular density.
network performance and continually monitor the route path. It Future Improvements: Vehicular density estimation may
uses the Dijkstra algorithm to identify stable and short path links. provide efficient results rather than network monitoring.
The protocol includes two schemes: FR and PU. The FR scheme
42) SDN-ENABLED CONNECTIVITY-AWARE GEOGRA-
determines an effective route by considering the relative distance PHICAL ROUTING PROTOCOL (SCGRP) [151]
between communicating nodes, link stability rate, and successful The SCGRP protocol employs a cloud-based SDN controller to
reception ability. The PU scheme enables the SDN controller to choose the most efficient path for packet forwarding. The SDN
monitor and analyze the entire application duration to mitigate controller uses an updatable network topology to predict the
the impact of continuous vehicle mobility. Accordingly, the connectivity and link duration between communicating nodes.
protocol will constantly update the routes in response to vehicle When a source node has data to be sent, it first checks its routing
motion changes. The controller measures the predicted table to see if it has a routing path to the destination vehicle. If a
expiration of the computed route path and sends it to the source routing path exists, the data packet is forwarded accordingly. If
vehicle. This route is only valid for the current data transmission. not, it requests the SDN controller for a routing rule. SDN
The new routing request needs to be invoked to generate a new controller calculates a forwarding region around the source node,
route version for the next generated packet. Upon identifying and selects the next hop relay within this region, considering
abrupt changes in vehicle mobility, the controller recalculates multiple metrics that are traffic density, distance, velocity, and
the optimal paths for all unexpired routes using the same link duration. Considering the speed difference between the
algorithm. If the original path fails, the controller sends an source and potential next-hop nodes, the SDN controller will
updated report to the source node to update its cached route table. prioritize nodes with lesser speed differences to find a stable
routing path with maximum link duration. After that, the
Advantages: The PU mechanism will minimize the effect of calculated path is sent back to the source node to start the data
abrupt mobility changes in the vehicles by allowing the SDN delivery process. Finally, if there is no connection with the SDN
controller to take responsibility for the application duration. controller, the data packet will be flooded to all nearby vehicles.
Disadvantage: Computing proactive routing paths can bring
more delay and SDN overhead. Node mobility makes the
Advantages: Using vehicle distance, velocity difference, and
computing of proactive routing paths worthless.
link duration in routing decisions can decrease packet loss rate.
Application Area: It is best for the FR-PU protocol to be
Disadvantages: The protocol did not consider the junction
deployed in a secure urban environment.
while messages were broadcasting/beaconing.
Future Improvements: Mobility prediction models can
Application Area: The protocol can be applied in an urban city
enhance the efficiency of the proactive routing computation.
with moderate vehicular density.
41) OPTIMAL RESOURCE UTILIZATION ROUTING SCHEME Future Improvements: Data security must be adopted,
(ORUR) [150] especially when using cellular networks for data delivery.
The ORUR protocol utilizes a cloud-enabled SDN model to
43) SOFTWARE-DEFINED TRUST-BASED AD-HOC ON-
select optimal routing paths and mitigate congestion of V2V DEMAND DISTANCE VECTOR ROUTING (SD-TAODV) [152]
communications by balancing the load of communication paths In [152], Zhang et al. proposed the SD-TAODV routing
across the entire urban road network. The SDN controller keeps algorithm to enhance the security and throughput of the SDN-
track of existing data communication paths by monitoring real- based VANET. Along with data routing, the SDN controller is
time connectivity and transmission delays on road segments. also responsible for trust management in VANETs. The SDN
When a vehicle needs traffic data, it sends a request packet to the controller computes the routing path and updates it if a better
SDN controller to find the best path for data forwarding. If the path with a higher trust value is found. To build the network
data is available in the cloud database, it is sent to the vehicle topology, the SDN controller broadcasts topology request
through LTE downlink. Otherwise, the SDN controller uses the messages to record the trust value of nodes. SDN controller will
WAVE network to compute the optimal routing path based on a evaluate the node trust value of each vehicle through forwarding
list of road segments to be followed. The vehicle then adds the ratio and node trust computation methods. The packets are
optimal path to the header of each data packet and sends it to the divided into control packets (Trusted-RREQ, Trusted-RREP)
destination using V2V communication. Once the data is and data packets. If a node receives a T-RREQ packet from a
delivered, the vehicle sends a finish message to the SDN neighbor, it checks if it has already received the request and then
controller. If no optimal path is available over the LTE or WAVE checks its routing table for a new route to the destination. If a
network, the SDN controller obtains the data from the source via new route exists, it updates its routing table and sends a T-RREP
an LTE link and sends it to the destination via LTE. packet back to the source. Otherwise, it broadcasts the T-RREQ
packet to its neighbors. If a node gets multiple T-RREQ
Advantages: This method includes a load balancing and messages, it selects the path with the better trust value from its
congestion prevention routing system.
routing table. The destination vehicle sends a T-RREP message
6 VOLUME XX, 2017
back to the source via the relay nodes, and the forwarding paths Future Improvements: Mobility prediction consideration can
are established when the T-RREP message passes through the enhance AoA and decrease beamforming search latency.
relay nodes. After receiving the T-RREP message, the source
transmits the packets following the designated forwarding path.
Uplink UMTS BS
Advantages: It reduces routing overhead by updating the
routing path only when receiving a new path with the best trust.
Disadvantages: The protocol performance can be affected
CH
regarding latency as it still relies on the conventional AODV
algorithm with minor adjustments. Backup
CH
Operational Environment: It can be applied in dense
environments with moderate vehicle distribution.
CH
Future Improvements: An efficient method to detect black
hole nodes through modifying routing requests and reply
Backup
packets in AODV can be one of the required improvements. CH
Downlink
44) ADAPTIVE VEHICLE CLUSTERING & BEAMFORMED OF Switches
TRANSMISSION FOR AGGREGATED TRAFFIC (AVC-BTAT)
[153]
AVC-BTAT is an adaptive clustering routing protocol for 5G- 5G Users
LTE BS
VANET with the aim to improve network management and
handle the growing traffic by predicting the arriving road traffic. SDN Controller
Using SDN global information, the vehicles are clustered based Cellular Link Floating car data backup
5G-VANET Coverage Ph. 1: Cooper. decoding
on their mobility and real-time road conditions. The cluster head
IEEE 802.11p Traffic Ph. 2: Traffic Distribution
is selected according to three metrics including angle of arrival
(AoA), received signal strength (RSS), and inter-vehicular FIGURE 17. The system architecture of AVC-BTAT protocol [153].
distance. Also, the backup head is selected to ensure 45) SDN-BASED GEOGRAPHIC ROUTING PROTOCOL
communication continuity. As shown in Fig. 17, the SDN (SDGR) [154]
controller controls BSs/RSUs using high-capacity fiber optic SDGR leverages the node location, vehicle density, and digital
links. To obtain the cell load conditions, each BS maintains a maps to find the shortest routing paths using Dijkstra's algorithm.
local database (LDB) that stores information about the vehicles Considering the road dimensions, the central SDN controller
in its cell, including clustering information, vehicle locations, uses periodic beacon messages to calculate the traffic density for
traffic requirements, and transmission schemes. The information each road. SDGR protocol consists of two main algorithms:
from multiple LDBs is combined to form a global database forwarding path algorithm and packet forwarding algorithm. The
(GDB), which the SDN controller uses to design network former algorithm calculates the shortest path with higher
policies and update local application modules. The protocol forwarding progress and vehicle density, based on network state
proposes an adaptive transmission scheme with selective vector and digital map. In the packet forwarding algorithm, the
modulation and power control to improve the trunk link SDN controller creates a subgraph for the source and destination
capacity. So, if the traffic exceeds the trunk-link capacity, the nodes to select the best next-hop relays. There are two packet
cluster head removes some vehicles with high traffic forwarding modes: forthright and junction modes. The forthright
requirements to guarantee communication quality. Here, mode delivers data packets based on the next hop position,
cooperative communication is used by sharing antennas with the velocity, and direction. In contrast, the junction mode uses a
cluster head as virtual antenna arrays to enhance the quality of congestion detection mechanism to address the load balance
communication and reduce traffic distribution delay. Adaptive problem in intersections where the nodes with lower buffer
beamforming is also suggested to enhance coverage range by limits are used to ensure smooth traffic flow. Finally, the SDN
using a wider beam for complete cluster coverage and a narrow controller sends this path to the source vehicle to help it deliver
beam to reduce interference and improve the trunk link packets to the destination vehicle.
throughput rate when multiple clusters coexist.
Advantages: Junction forwarding positively impacts the traffic
Advantages: Using a dual cluster head and dynamic signal in road junctions. Packet latency will be reduced by
beamforming coverage improves trunk link communication using digital map information to elect the hop node.
quality and clusters' network robustness. Disadvantages: Overhead minimizing mechanism is not
Disadvantages: Protocol complexity can lead to high latency adopted here. Placing the central controller far away from the
in dense VANET. Using cooperative communication can lead end vehicles leads to an increase in route discovery time.
to increased latency and cluster head management overhead. Operational Environment: It is good that it is applied in a
Operational Environment: It is preferable to be deployed in well-defined transportation area so that it is possible to build
an urban city with high cellular communication coverage. sub-graphs by SDN controller.
6 VOLUME XX, 2017
Future Improvements: More analysis is required to show its Disadvantages: Due to local routing computation, the
efficiency with different transportation layouts and situations. computational complexity and overhead will be increased.
Application Area: The protocol is unsuitable for sparse
46) DISTRIBUTED SOFTWARE-DEFINED
INFRASTRUCTURE LESS VEHICULAR NETWORK-BASED
VANET, so it is good to be deployed in dense urban areas.
ROUTING (DSDiVN) [155] Future Improvements: Using RSU for local computation and
In [155], Alioua et al. introduced dSDiVN, a distributed multi- vehicle information management will reduce the workload of
hop SDN-based architecture for data routing in infrastructure- the local domain controller.
less VANETs. The approach assigns a dedicated mobile SDN 48) SDN-BASED ON-DEMAND ROUTING PROTOCOL
controller for each zone partition. Based on IEEE 802.11p, the (SVAO) [110]
SDN controllers will obtain a global view of the network state. In [110], Baihong et al. proposed the SVAO routing protocol to
The road is segmented into fixed-size fragments, representing utilize SDN to enhance the efficiency of data routing in the
virtual cluster. The vehicle with the most extended lifetime is VANET network. The system separates the SDN architecture
selected as the cluster head. Once elected, each cluster head into local and global control systems. Local SDN controllers are
activates its local mobile controller to manage a backup installed at every crossroad to collect the vehicles' information
candidate list to prepare for potential failure and establish a and perform local routing decisions. The global SDN controller
recovery controller. The knowledge base of each local controller utilizes a centralized level with an improved AODV method to
is compressed and backed up on the recovery controller to help calculate the optimal route among several road segments along
the recovery controller to continue services if a mobile controller which a message should be forwarded. Initially, the source
fails. Once activated, the recovery controller uses the knowledge vehicle sends a route request message to the local controller. If
base replication to respond to routing requests immediately. the destination is found within the local scope, the local
controller will send the route information to the source;
Advantages: Through zone partitioning and distributed otherwise, it will request the next controller level to inquire about
controllers, it can improve scalability, ensure a reasonable the destination vehicle information. If none in the current SDN
delay, and provide better support for delay-sensitive services. controlling level can access the destination vehicle, the request
Disadvantages: It did not show the backup controller election. will be repeated until the destination vehicle is located. If none
A high number of hops causes a longer flow setup time. of the local controllers can access the destination vehicle, the
Operational Environment: It can be applied in urban, less- global level will request to calculate the position of the route to
infrastructure VANET environments. find the final route among the road segments. After selecting the
Future Improvements: The protocol requires more attention forwarding road segments, the corresponding local controllers
toward the minimization of routing overhead and quality of will be informed to search the shortest transmission path and the
neighbors when selecting the next hop relay node. forwarding nodes that participate in the routing path by utilizing
47) HIERARCHICAL SOFTWARE-DEFINED VEHICULAR the Bellman-Ford algorithm. The candidate path will be selected
ROUTING (HSDVR) [156] according to the link stability, the relative velocity, and the
Correia et al. [156] proposed a cluster-based routing strategy to number of nodes.
address the issue of connectivity loss with the SDN controller
and provide virtual infrastructure. There are two types of SDN Advantages: Employing hierarchical SDN controlling can
controllers: the local controller and the primary controller. If alleviate the burden of the central controller.
there is no connection with the primary SDN controller, the local Disadvantages: In a high-speed scenario, the performance
SDN controller will take on the role of the primary controller and degrades rapidly. Also, the assumption that all roads are one-
handle the network in its domain until the primary controller way is impractical.
reconnects. For data routing, the source vehicle checks whether Application Area: The protocol requires well-distributed
it already has a path to the destination in the routing table. If so, urban, so it is only suitable for denser traffic situations.
it redirects the packets towards the destination. Otherwise, it Future Improvements: using mobility prediction can enhance
broadcasts the request path message to neighboring nodes. When the protocol efficiency by helping to determine the next
a vehicle receives a routing request message, it checks if it has a intersection before data transmission.
route to the destination. If it has, then it checks if it is a controller. 49) INTERSECTION DYNAMIC VANET ROUTING (IDVR)
If yes, then it sends the request path to the requesting vehicles. [157]
Otherwise, if it is not a controller but has a route, it sends a route IDVR is an intersection-based geographical routing protocol that
update request to the local controller to update it and then utilizes a centralized SDN architecture to enhance route stability
forwards the updated path to the requested vehicle. If there is no and minimize delay. The controller gathers real-time traffic
route and the receiving vehicle is not a controller, the local information and selects one vehicle as an intersection cluster
controller forwards the message to the primary controller to find head (ICH) for each intersection according to the maximum
the path toward the destination node. lifetime until it exits the cluster zone. When a packet reaches an
intersection, the protocol applies recursively between the current
Advantages: Using a filtering technique to determine the next and desired destination intersections. For each intersection, the
relay will reduce transmission overhead. The route can be protocol computes the threshold point distance, which represents
found if the connection with the central controller is lost. the final point of the handover process to be invoked. Besides,
6 VOLUME XX, 2017
the SDN controller periodically calculates the average graph (TD-G). The protocol determines whether there is a time-
throughput for each road segment. When packets arrive at the dependent path (TD-P) between two nodes and stores all
ICH, it uses Dijkstra’s algorithm to search the shortest route reachable nodes in a set. To select the best routing path with the
based on the current location, destination location, and the minimum total cost from the set of available paths, the protocol
maximum of the minimum average throughput among the set of employs the time-dependent shortest path (TD-SP) algorithm
candidate shortest routes. When the current ICH reaches a that identifies the earliest reachable path and then updates it if a
threshold point, a new ICH should be elected, and all data should shorter path exists.
propagate to the new ICH. When no vehicles are in the cluster
zone, the current ICH follows the SCF rule until it reaches Advantages: Using trajectory prediction and a time-dependent
another ICH closer to the destination intersection. graph, the routing overhead and network cost can be decreased.
Disadvantages: The protocol complexity is relatively high,
Advantages: The protocol can ensure high reliability by making it less suitable for denser networks.
utilizing intersections with a minimum number of segments. Application Area: Due to protocol complexity, it is preferred
Disadvantages: Routing re-calculation at each intersection to be applied in a moderate-density vehicular environment.
will cause high latency due to multiple routing computations. Future Improvements: Considering multiple constraints,
Application Area: It is suitable for urban environments. such as vehicle mobility, context features, and link lifetime can
Future Improvements: The performance can be improved by improve the performance of data multicasting.
considering multi-constraints such as communication ability
52) COOPERATIVE DATA SCHEDULING-ROUTING
and resource computing. PROTOCOL IN HYBRID SDVN (CDSRP) [160]
50) COST-EFFICIENT SENSORY DATA TRANSMISSION IN In [160], Liu et al. proposed a centralized cooperative data
HETEROGENEOUS SDVN NETWORKS (CESDT) [158] dissemination scheme by utilizing a hybrid of I2V and V2V
He et al. [158] proposed a CESDT protocol to minimize communications. The protocol employs a hierarchal SDVN
communication costs and bandwidth requirements by utilizing infrastructure where the local controllers are embedded in RSUs.
various wireless resources and scheduling the data over them. The local controller performs both data routing and scheduling
The protocol uses cloud-based SDN as a resource manager to decisions by selecting the communicating vehicles and
centrally manage all network resources. It is based on the vehicle instructing them on which channel to tune and which data to
trajectory prediction to estimate network availability and share. Data routing decisions consider both communication
required bandwidth. Based on the bandwidth requirements, the constraints and application requirements, providing a higher
SDN controller selects the optimal network interface from all priority to vehicles with shorter remaining dwell times in the
available candidates and determines the optimal routing path service region. The communication process is divided into three
over single or multi-hop transmission links. The single-hop phases: first, all vehicles use V2V mode to broadcast beacon
communication is modeled using a network availability matrix, messages to identify a list of neighbors; second, all vehicles
while a time-dependent graph models the multi-hop routing. For switch to I2V mode to inform the RSU of their current neighbors
multi-hop data routing, it employs a polynomial time and their cached and requested data items; third, for data
approximation scheme to select the optimal shortest path among delivery, each vehicle uses either I2V or V2V mode. Here, the
several paths that can fulfill the application requirement in a vehicles on the I2V channel can receive data items from the
time-dependent manner. To find the optimal single-hop routing RSU, while other vehicles can simultaneously communicate
paths, the protocol uses a greedy approximation algorithm, over the V2V channel.
where all available network interfaces are initially selected and
Advantages: Using dwell time-driven priority can ensure a
then gradually removed one by one in each iteration until the
high packet delivery ratio in a vehicular environment.
bandwidth requirement is slightly above the specified limit.
Disadvantages: The limitation of RSU to broadcast a single
data unit at every scheduling period will cause high latency and
Advantages: The communication cost can be minimized by decrease the overall QoS.
directing the data over various communication interfaces. Application Area: It is best to be deployed in urban VANET
Disadvantages: Large-scale networks may not achieve with a reasonable number of vehicles.
optimality due to using greedy methods in route selection. Future Improvements: MIMO technology and full-duplex
Application Area: Due to its interoperability, it can be useful communication can increase its efficiency and delivery rate.
in an environment covered by multiple network infrastructures.
Future Improvements: Using link lifetime and routing 53) CENTRALIZED ROUTING PROTOCOL (CRP) [108]
duration can improve protocol reliability and scalability. CRP routing protocol uses a centralized SDN model for data
routing efficiently while minimizing routing overhead. The
51) PREDICTIVE TIME-DEPENDENT MULTICAST ROUTING protocol presented the minimum optimistic time (MOT)
PROTOCOL (PRETTI) [159]
algorithm to handle changes in network density by allowing the
In [159], He et al. developed a multicast routing protocol to make
protocol to switch between multi-hop forwarding and SCF
efficient multicast scheduling decisions and reduce network models according to the current network density. Using global
delay and communication costs. Communication overhead is network information, the MOT algorithm can estimate the
reduced by leveraging vehicle trajectory prediction, which minimum time required for any vehicle to transmit a packet to
models the predicted topology change with a time-dependent
6 VOLUME XX, 2017
another one along the best route. When a vehicle receives or searched RSUs are computed and sent to the corresponding
generates a packet, its routing client application checks its RSUs. Finally, these RSUs will broadcast the message among
routing table for a path to the destination. If a path exists, the the vehicles within their broadcast area. To prevent duplicate
packet is sent along that path. If no path exists, a route query packets from being sent to two RSUs within the transmission
message is sent to the SDN controller, which uses the MOT range of the source vehicle, a sequence number is used to
algorithm to determine the shortest route from the current identify duplicate packets. The controller processes only one
vehicle to the destination vehicle. Based on the global network packet and adds a drop rule to one of the RSU flow tables. The
image and digital map, the SDN controller computes the Floodlight OpenFlow controller is used as the SDN controller,
optimal route and sends it back to the requesting node to start including multiple modules such as topology management, RSU
the data routing process. location management, and GeoBroadcast routing.
Advantages: The switch between multi-hop forwarding and
Advantages: Geographical-based message broadcasting can
CSF models can ensure high network performance.
reduce communication overhead.
Disadvantages: Using Wi-Max with Wi-Fi can bring more
Disadvantages: The protocol did not consider the need for
security and bandwidth utilization issues. V2V data transmission in less-infrastructure VANETs. Also,
Application Area: The protocol requires a digital map with message reliability and validation are not considered here.
special in-vehicle embedded interfaces.
Application Area: With I2I links, the protocol can provide
Future Improvements: H-SDVN architecture can help in fault
good results in a fully covered area with enough RSUs.
tolerance and reduce the overload on the central controller.
Future Improvements. Routing maintenance and recovery
54) EFFICIENT MULTIPLE-COPY ROUTING IN SDVN models are suitable solutions to improve protocol reliability.
(SPRAY-AND-PREY) [107]
In [107], Ming et al. proposed an SDN-based multiple-copy VI. COMPARATIVE ANALYSIS AND DISCUSSION
protocol to improve the delivery rate and minimize network This section provides a qualitative comparison of SDN-assisted
resource costs. A graph-based utility function is leveraged to VANET routing protocols based on their state-of-the-art ideas,
determine the best carriers based on global network topology, features, optimization principles, performance evaluation
the distance to the destination, and the number of hops. The techniques, and results analysis. The comparison is presented in
protocol uses the Dijkstra algorithm to find the minimum Tables 4, 5, 6, 7, and 8. Table 4 summarizes the protocols based
communication hops from any carrier to the destination. When on various parameters and factors. Table 5 outlines the key
two carriers have the same distance, the one with a larger angle features of the protocols, including their performance objectives,
with existing copies is chosen. In the spray phase, the source innovative ideas, and emerging techniques used. Table 6 lists the
node duplicates a fixed number of messages and transmits them performance metrics and evaluation techniques for assessing the
to one-hop neighbor carriers. The controller then uses the prey discussed protocols. Table 7 presents the optimization
algorithm to choose the next carriers. The prey phase continues parameters used in the existing routing protocols, while Table 8
until the destination receives the message, and then the controller compares the protocols based on their performance evaluation
clears all other copies. A cooperative elimination scheme and result analysis techniques. The tables provide researchers
removes duplicates with poor utility and reduces delivery delays. with a valuable resource for selecting an appropriate protocol or
designing a new one. The following subsection thoroughly
Advantages: By spraying multiple copies of packets, PDR will discusses the comparable factors and special features.
be high. Using distance/angle in the utility function can reduce
the latency and the number of sprayed copies. A. DISCUSSION ON GENERAL CHARACTERISTICS OF
Disadvantages: It did not consider nodes' mobility and low- PROTOCOLS
density situations. Table 4 compares the reviewed routing protocols based on their
Application Area: The protocol requires a highly dense operational characteristics related to various QoS parameters,
vehicular environment to ensure many carriers. performance measurements, and techniques. In addition to the
Future Improvements: The protocol requires an efficient tabulated data, a qualitative comparison is mentioned here.
algorithm to minimize the number of sprayed messages and Routing protocols, such as SPIDER, CLHLB, HRLB, and
communication overhead. SCGRP, utilize a centralized SDN controller to provide per-flow
55) SDN-BASED GEOGRAPHICAL BROADCASTING routes to vehicles based on the computation of the global
ROUTING PROTOCOL (GEOBROADCAST) [9] topology controller. While these protocols can maintain more
GeoBroadcast is designed to broadcast messages under SDN efficient routes with a high PDR, the central controller's burden
supervision within a specific geographical region. For data and route computation latency present significant challenges. On
routing, the vehicles will send their messages to the nearest RSU. the other hand, several protocols, such as DMPFS, OCDEC,
If a flow entry is found in the RSU flow table, the RSU will use SDMEV, and QRA, adopt hierarchical SDN controlling to
the cached flow rule for data routing. If not, a packet-in message compute routing paths locally. These schemes ensure low
is sent to the SDN controller to retrieve the geographical latency of local routing paths and efficient load balancing. Still,
information of the destination and find all RSUs within the their applicability in real-time is doubtful due to low data
destination geographical area. The shortest routing paths to the delivery and high latency of global routing paths. To address
6 VOLUME XX, 2017
these issues, hybrid SDN control is used in many protocols, such using SDN programmability to establish multi-hop V2V routing
as POLAR, TDCR, SeScR, and LSB-OR, to move the local protocols for vehicles to share their data. However, using SDVN
routing computation from the central controller to local routing protocols to find the optimal route for a particular
controllers and vehicles. This approach can minimize routing communication model (e.g., V2I or V2V) is a weakness.
latency and communication overhead. However, computing Adopting both models in data routing, as in PT-GROUT,
routing paths locally with local view controlling can decrease CCDEC, SFRS, and SDMEV protocols, can ensure high PDR
protocol reliability, especially when using cached routing paths. and improve scalability, but it may increase complexity and
In most protocols, such as ICDRP-F-SDVN, GLS, POLAR, controller overhead.
and V-TLRH, the controller finds just one path for data Many existing SDVN routing protocols, such as IM-DOS,
transmission. This single-path approach is simple and suitable TDCR, LSB-OR, and SDGR, use static shortest path algorithms,
for low-traffic or sparse networks. The dynamic nature of such as Dijkstra's algorithm, to determine the shortest routes.
VANET can cause links to be temporary, which may not provide Generally, in SDVNs, the links between communicating
enough time to transmit the necessary data. This can lead to vehicles are valid for a short time, rendering such algorithms for
increased latency and decreased throughput. To address this static networks inadequate. Consequently, solving the routing
issue, some schemes, such as SELAR, TRIBRID, and Pretti, problem becomes equivalent to finding the fastest route in a
establish multiple routes to ensure high PDR and less latency. dynamic temporal graph. The researchers have explored the
However, multiple paths can cause high network congestion, routing paths problem in temporal graphs, as in the PT-GROUT,
especially in dense VANET. CDRS, SPIDER, and SeScR, which primarily focus on refining
Existing SDVN routing schemes use either a beacon-based the online query effectiveness of network pre-computations.
method or a prediction-based method to maintain the dynamic Numerous approaches have been proposed to decrease the
topology of the network. Beacon-based protocols like SDVN- number of control messages exchanged with the SDN controller,
MiCR, SURFER, SFRS, SDMV, and CRS-MP use regular which are required to convey mobility and status information of
beacons to update the network topology. In contrast, prediction- vehicles or clusters. The increased number of exchanged packets
based protocols, such as CDRS, V-TLRH, CESDT, and Pretti, generates high uplink/downlink communication overhead, and
use historical data to predict link status and reduce the number concerns regarding data security and node privacy can further
of beacons sent to the controller. While beacon-based methods raise the number of control messages and packet size. Vehicle
can cause high uplink overhead, they maintain efficient routes trajectory prediction has been identified as the upper solution for
and high data delivery. Prediction-based methods, conversely, decreasing overhead in protocols such as SDVN-MiCR,
are more scalable with minimum network overhead in dense SPIDER, LSB-OR, and VDR-DRL. This approach significantly
networks. However, complex prediction algorithms can cause reduces the amount of beacon status while the controller
high latency and controller overhead. constructs the network graph based on gathered beacons and
SDVN routing protocols, such as POLAR, IM-DOS, HSDN- mobility predictions. Vehicle clustering has also been applied in
GRA, and MFCAR, provide unicast routing paths from source many protocols, such as TDCR, IMCR, HSDVR, and AVG-
to destination, which can offer high data delivery with minimum BTAT routing methods, to minimize the overhead by allowing
network overhead in multi-hop communication. However, the only the cluster head to communicate with the controller.
computation of routing paths is restricted to SDN controllers, However, the size of beacon packets is not adequately reduced
and vehicles cannot participate in routing computation. This as the cluster status messages are still being sent. Protocols that
limitation can result in a high packet loss rate in sparse or less- use vehicle clustering and mobility prediction can be classified
infrastructure VANETs. In contrast, some routing protocols, as having the lowest communication overhead. Conversely,
such as CDRS, SDMEV, and ORUR, allow vehicles to search protocols that use broadcasting and data dissemination, such as
for routing paths toward the destination using packet SDMEV, LBR, Spray-and-Prey, and GeoBroadcast protocols,
broadcasting. While this approach increases network overhead, are classified as having high communication overhead where
it can ensure a high packet delivery ratio, especially when many control packets are broadcasted to find next-hop nodes.
routing flows are not received, or the communication link with To cater to the requirements of SDVNs, routing schemes need
the SDN controller is unavailable or disconnected. Some to tackle not only the problem of broadcast storms in congested
protocols, such as SCGRP, SD-TAODV, dsDiVN, and HSDVR, networks but also the challenges of routing in intermittently
use adaptive modes for data transmission to improve routing connected networks. As a result, some works focused on routing
efficiency. However, such schemes require more tools for for both scenarios, such as POLAR, SURFER, and CLHLB,
communication management and decision-making. TRIBRID is instead of solely dense networks, as seen in DMPFS, ERA, and
an example of a routing scheme that uses the SCF mode to CRP. However, only a few studies are designed for sparse
improve PDR and minimize network overhead. scenarios, such as LBR, LT-NSR, and dSDiVN.
Many VANET applications realize this by broadcasting safety VANET allow vehicles to connect with RSUs to obtain real-
messages to nodes within the communication range using V2I time traffic data, warning reports, and entertainment data. Some
links. Several SDVN routing protocols have been suggested to proposed protocols, such as CDRS, DMPFS, CTDD, and
compute V2I routing paths and ensure high PDR, including NSDD-SDVN routing protocols, consider packet parameters,
DMPFS, STDD, LBR, and ORUR. Conversely, other protocols such as type, required bandwidth, and maximum latency, in
such as POLAR, ACA-RP-SDVN, HRLB, and QRA focus on routing decisions. The priority of some packets over others can
6 VOLUME XX, 2017
ensure a high PDR. However, the vast majority of SDVN addressed data transfer for both types of applications, with some
protocols do not consider message priorities. Protocols such as prioritizing critical applications over entertainment applications,
SFRS, SD-IoV, CRS-MP, and IMCR prioritize access to such as HSDN-GRA, TDCR, CTDD, and CRP. Others only
entertainment data at the expense of essential data, such as safety mentioned the ability to send packets of both types without
messages and weather warning reports. specifying the data transmission and queue management
SDVN routing protocols need to consider the required mechanism, as in the QRA, CLR-SDVN, and IMCR protocols.
computation power and implementation complexity. Some
protocols such as PT-GROUT, SDVN-MiCR, SPIDER, and IM- B. DISCUSSION ON PERFORMANCE AND INNOVATIVE
DOS protocols have addressed these issues through prediction- IDEAS
based routing schemes. However, running intelligent or time- Incorporating multi-constraints with context awareness into
constraint routing can also be complex due to the high routing decisions makes the protocol more practical and
computation requirements in a controller. On the other hand, appropriate for real-world applications. For instance, while
cluster-based routing schemes are generally less complex, as the protocols developed for highway scenarios may be able to avoid
SDN role is limited to forming clusters and performing inter- considering traffic distribution, those developed for urban
clustering routing. Examples of such protocols include TDCR, scenarios must consider several road segments and intersections
IMCR, ORUR, and dsDiVN. When comparing the complexity distribution. Besides, the protocols with more complex
of protocols, factors such as run-time assumptions, optimization mechanisms tend to utilize more control messages, which can
criteria, and coverage area should be considered. result in more controller burden. Therefore, performance metrics
The reviewed routing algorithms can be categorized are chosen and evaluated accordingly depending on the
according to their ability to handle increasing routing queries in objective. In the same context, even though emerging
the network. Centralized routing schemes such as SELAR, IM- technologies can enhance the reliability and scalability of routing
DOS, VDR-DRL, and Geo-SDVN protocols are not scalable due protocols, they require more technology access and management
to the heavy computation required by the central controller. This costs with more security challenges. In general, the protocols
makes them unsuitable for large-scale networks. In contrast, that adopt centralized SDN have the highest delay and the least
hybrid routing schemes such as CDRS, POLAR, SFRS, and scalability. Nevertheless, protocols that are specifically designed
TDCR are designed to be scalable with a hierarchical SDN for urban environments are expected to exhibit lower delays and
architecture. Vehicle clustering is one of the scalable solutions. improved connectivity in high-density scenarios. Some
Vehicular communications require reliable data transmission protocols prioritize the optimization of traffic load and
between fixed RSUs and mobile vehicles or between them. The computation overhead, while others attempt to achieve optimal
reviewed SDVN routing protocols have introduced various control by distributing the SDN controllers hierarchically.
schemes to ensure reliable data delivery. For routing protocols, However, the following discussion focuses on innovative ideas,
the communication reliability is evaluated based on the recovery emerging techniques, and performance objectives. The strengths
strategy, packet drop ratio, link duration, and forwarding and weaknesses of creative ideas, key features, and performance
mechanism. SDN accomplishes this by using a linear graph with criteria are also discussed technically here.
weighted vertices and branches as the probabilistic model. For ICDRP-F-SDVN [111] aims to maximize throughput and
instance, SCGRP, SPIDER, and CDRS routing protocols have reduce communication overhead using fog computing and
used link lifetime to ensure reliability on multi-hop data routing. vehicle clustering in SDVN data routing. The protocol was
However, given VANET mobility, such a method may be evaluated through simulation and compared with other methods
insufficient since vehicle mobility may cause link breakage. using five metrics: throughput, E2E delay, packet handling,
Conversely, other protocols use acknowledgment messages to control messages overhead, and cluster head switching.
guarantee the delivery of a message packet exactly once while However, the protocol does not consider high vehicle mobility
detecting errors or packet loss. Reliable protocols typically incur and network sparsity in its performance.
more overhead than unreliable protocols, resulting in slower and SDVN-MiCR [112] protocol aims to deliver vehicular data
less scalable operations. with low delay and high reliability in highway situations by
VANETs provide many applications with specific constraints reducing channel congestion caused by the high rate of data
and QoS requirements, such as delay, bandwidth, and dissemination. The protocol was tested through real-time
transmission rate. However, most of the reviewed protocols did simulation, demonstrating its superiority in communication
not consider the application type in their forwarding decisions, overhead, PDR, latency, and collision rate. Although the
potentially leading to suboptimal results where diverse protocol outperforms current protocols by 18.3% in terms of
application requirements are not adequately addressed, as is the PDR and latency, it did not consider the effect of mobility on
case with protocols like SPIDER, ERA, SeScR, CLHLB, routing decisions.
HSDVR, and Spray-and-Prey. Some protocols focused solely on The CDRS [113] aims to enhance packet delivery by
routing safety messages, such as ICDRP-F-SDVN, DMPFS, minimizing average service time and preventing conflicts during
SDMEV, and GeoBroadcast, or only on infotainment data, such data transmission. The protocol was evaluated in different
as the CDRS protocol. So, prioritizing some applications over scenarios with various metrics including packet drop ratio,
others may negatively impact transfer balance, resulting in poor deadline missed ratio, PDR, service delay, computational time,
quality of experience (QoE). Conversely, some routing protocols
6 VOLUME XX, 2017
and solution quality. Although the protocol targets infotainment based on real-time conditions. The protocol is compared with
applications, it was not tested with real-time infotainment classical routing strategies, including AODV, OLSR, GPSR,
applications. Besides, the computation time may become DSR, and DSDV, using PDR and E2E delay as comparison
challenging in dense environments where the number of vehicles parameters. Although the protocol maintains high QoS regarding
exceeds the predefined threshold of 25 vehicles per controller. PDR and E2E delay, the simulation has not considered extreme
DMPFS [114] employs stationary vehicles as fog nodes to traffic conditions like highway roads and sparse networks.
participate in multicasting data routing. However, the real-world SURFER [120] combines blockchain technology with SDN to
environment may not guarantee a consistent number of vehicles enhance the reliability and security of data routing in IoV.
to be as fog nodes, resulting in an inability to ensure the high Simulation results in urban and rural situations have
reliability and scalability of data routing. Moreover, packet demonstrated the protocol efficiency regarding E2E delay, PDR,
categorization and zone segmentation techniques minimize routing overhead, and traffic load. Following this analysis, when
overhead and time complexity, but these methods introduce the number of packets increases significantly, the packet drop
controller computation overhead issues. The protocol is rate will be high where the packets need to stay more times in
evaluated using PDR, E2E delay, and communication overhead the waiting queues at relay nodes. Besides, the blockchain
across varying mobility rates. Nevertheless, the protocol technology will produce additional packets with larger sizes due
performance does not evaluate with no or few parked vehicles to encryption processes.
on the roadside. CCDEC [122] is an SDN-based context-aware data
PT-GROUT [115] protocol employs temporal graphs for cooperative dissemination strategy. The protocol performance is
routing decisions with high efficiency in routing computation evaluated and compared to other protocols using multiple
and time expense for the routing process. The protocol is metrics, including protocol overhead, channel capacity gain,
assessed using various performance metrics including delay, and data delivery ratio. The communication quality and
computation cost, PDR, delivery delay, and jitter. The algorithm efficiency may be questionable due to the assumption of
consistently outperforms others in high-traffic density situations. communication ability only between neighboring vehicles with
However, under low-density scenarios, the four metrics LoS. SFSR [123] protocol aims to reduce packet loss rate
evaluated for Dijkstra's algorithm may resemble those of PT- through switching messages transmission over the VANET
GROUT. Additionally, as the number of vehicles increases, the infrastructure and Internet network via roadside fog switches
SDN controller workload will rise due to the more extensive placed at intersections. However, deploying fog systems to
computations required for temporal graph processing. SPIDER participate in V2V data routing will increase deployment costs
[116] is a predictive routing scheme, inspired by social and system complexity. The protocol is simulated and evaluated
computing to enable efficient and reliable data exchange in regarding PDR, packet loss ratio, E2E delay, and routing
dynamic vehicular networks. The protocol is evaluated using overhead. However, various considerations, such as vehicle
average delay, jitter, network congestion, average calculation mobility and traffic features, are not adequately evaluated. ERA
time, and PDR metrics. However, it is essential to consider [124] protocol leverages machine learning and edge computing
multiple factors in routing decisions such as high mobility and techniques to improve data routing in the SDN-IoV network.
controller overload. GLS [117] considers historical routing Based on the traffic conditions and vehicle mobility information,
experiences and real-time mobility patterns to identify routes the vehicle location can be accurately predicted. Although ERA
with high link stability and short delays. The protocol outperforms other protocols in terms of PDR, other QoS metrics,
performance is evaluated using three variations: GLS-no-update, such as E2E delay and throughput, are not considered. However,
GLS-update, and GLS with various metrics, including PDR, the protocol efficiency may be affected in the sparse network
average delay, path length, and communication overhead. The where insufficient mobility information will cause to inaccurate
GLS form performs better in providing efficient and stable prediction model. IM-DOS [125] uses social computing and a
transmission than other variations. However, the protocol link lifetime prediction model to select the optimal relay for
efficiency decreases with increased transmission distance or broadcasting safety messages in the SDN framework. The
reduced node count. SELAR [118] employs SDN to manage the simulations demonstrated the scheme's efficiency regarding E2E
load of fog nodes and reduce energy consumption in 5G- delay, PDR, and package utilization. Yet, the results become
VANET networks. The novelty stems from its adaptive control more inconsistent as the vehicle number increases and the
paradigm that optimizes traffic load and determines active coverage area enlarges. Besides, the scheme does not consider
networking devices. The effectiveness of the protocol is backup transmission during the transmission process and
evaluated using simulation-based experiments. However, the upcoming hops.
protocol performance has not been evaluated using primary QoS HSDN-GRA [126] addresses the connectivity issue in
metrics, such as PDR, delay, and throughput. Also, the protocol VANET through clustering techniques and multi-constraints
applicability in dense environments may experience high delay relay selection. The protocol is evaluated using the JADE multi-
and packet loss. POLAR [119] leverages edge computing, agent platform, measuring routing overhead, packet drop rate,
swarm intelligence, and machine learning models to select the and throughput. However, for more accurate results, the protocol
optimal routing strategy with better PDR and E2E delay across performance must be evaluated under real-time traffic situations
several VANET scenarios. The protocol's strength lies in its using network simulation software such as NS3 and OPNET.
adaptability, as it can switch between several routing protocols ACA-RP-SDVN [127] employs an ant colony strategy for data
6 VOLUME XX, 2017
routing in dynamic urban situations, alleviating the concern of QRA [133] utilizes the connectivity rate, road traffic
low success rates and poor routing performance. Although the information, and SINR metric to determine the optimal path for
ant colony scheme can improve the data routing by continuously message routing over reliable and stable path in urban vehicular
adapting routing tables with channel congestion, their environments. The performance evaluation is realized through
applicability in VANETs can be inefficient. Nonetheless, mathematical simulation and compared with other strategies in
experimental results demonstrate that the method outperforms terms of PDR, average delay, routing overhead, and controller
AODV and GPSR protocols regarding routing lifetime, routing overhead. However, the protocol does not consider the hops
change rate, and communication performance. However, the count, which may result in undesirable delays in vehicular data
protocol did not consider network dynamicity and node mobility. transmission. VDR-DRL [134] combines SDN and DRL
TDCR [128] aims to meet QoS requirements for delay-sensitive techniques to select edge and gateway head nodes. Numerical
services in terms of delivery rate and E2E delay through DSRC analysis shows that the scheme can enhance data throughput for
and C-V2X hybrid communication. The protocol is simulated different vehicle densities and velocity ranges. However, the
and evaluated regarding routing overhead, hops count, average dual cluster head selection may increase controller overhead and
delay, and bandwidth cost. The simulations demonstrated that packet latency. Furthermore, the protocol must be evaluated and
the protocol accomplishes low latency for delay-sensitive analyzed using various conventional QoS metrics. CLHLB [135]
applications and low bandwidth cost for high data-rate ones. combines V2I and V2V cross-layer routing to deliver packets
However, the assumption that there are always enough vehicles with high efficiency at different traffic densities using path
within the DSRC range cannot be achieved in most realistic connectivity probability. Simulation results demonstrated a
environments. SeScR [129] employs the spectral clustering significant gain in PDR and average delay compared with many
technique with a deep learning model to improve cluster stability routing strategies. However, the protocol may experience
and routing reliability. Extensive analysis has been performed to increased latency due to the gradual increase in multi-hops in the
ensure protocol efficiency against other clustering-based final routing path. MFCAR [136] leverages the management
schemes regarding cluster stability, lifetime, and association capabilities of SDN to decide the best routing paths using
time restrictions. Also, it is compared with SDN- and ML-based cellular and Wi-Fi standards. The numerical analysis
schemes using throughput, latency, and computation delay. demonstrates that the protocol outperforms Dijkstra's algorithm
However, it didn’t consider road conditions and vehicle regarding PDR, throughput, and average delay. However, the
parameters in cluster formation and routing decisions. V-TLRH simulations were carried out with fixed geographical positions
[130] uses a three-level routing hierarchy to improve routing of communicating vehicles, which may not fit the nature of
performance and data delivery in a dynamic VANET network. VANET networks. Additionally, the protocol needs to be
The results analysis shows the protocol can perform well under compared with the most studied routing algorithms, such as
different network scenarios, such as network change and node AODV and GPSR.
mobility. It has a higher PDR and reasonable delay than other NSDD-SDVN [137] utilizes multiple wireless interfaces to
approaches in different VANET scenarios. However, the enable efficient data transfer under different application
performance testing did not consider the complexity, load requirements. The SDN controller will select the network
distribution, and synchronization of multiple edge servers. interface using a two-stage Stackelberg game theory method.
SDMEV [99] protocol utilizes edge storage and computation The numerical analysis demonstrates the protocol's efficiency
abilities to minimize packet latency and loss rate for SDN- regarding average delay, PDR, throughput, and routing
enabled V2X communication. Performance analysis shows that overhead. However, the multi-metrics used in the utility function
the data routing over SDMEV can satisfy the delay requirements may result in high computational power and increased delay.
for V2X services. However, the results indicate that vehicles far CRS-MP [138] leverages the ANN technique to predict mobility
from the congestion area can receive messages in over 1 second, patterns and anticipate the nodes' arrival rate in the network. The
which may be insignificant in some situations. Moreover, the protocol is simulated and compared with other existing routing
controller will suffer a high burden in generating forwarding methods regarding transmission delay with varying vehicle
rules and knowing the actions that must be employed on the velocity. However, the protocol's performance is not analyzed
vehicles. TRIBRID [131] aims to find stable and shortest routes when the connection is loss with the controller or in the node
for packet routing with minimum latency. It utilizes the sparsity scenario. CLR-SDVN [139] can generate optimal
centralized and distributed routing models for unicast, broadcast, routing paths over the most stable links in urban inter-vehicle
and SCF data delivery. The protocol efficiency is analyzed and networks using cross-layer parameters such as forwarding
verified using PDR, latency, routing overhead, and hop number. probability, bandwidth availability, and link lifetime. The overall
However, the impact of node mobility is not considered in the evaluation of the protocol is demonstrated and compared with
performance analysis. SD-IoV [132] utilizes edge-enabled other protocols in terms of routing overhead, E2E delay, packet
SDVN technology to minimize routing overhead and improve drop ratio, and average throughput. However, the performance
scalability in highly dynamic VANETs. The results analysis analysis did not address high mobility and low vehicle density
showed a significant improvement in PDR, E2E delay, and situations. IMCR [140] aims to reduce overhead and improve
routing overhead. Also, the edge controllers can minimize link transmission efficiency by applying an SDN-powered influence
failure in the network. However, the E2E delay may increase maximization algorithm to select double-head clusters. The
with the high changes in vehicle mobility. protocol is compared under three structures: no cluster, single-
6 VOLUME XX, 2017
CH, and double-CH. However, the protocol complexity needs to and V2V communications. Even though such a method can
be analyzed with the ordinary situations of VANETs. improve PDR and reliability, the average delay will increase due
CTDD [141] routing strategy enables unified management of to multiple offloading between network architectures. The
heterogeneous network resources, capitalizing the number of numerical simulation uses multiple metrics, including offloading
routing quires completed within a delay constraint. The protocol fraction rate, throughput, link average lifetime, and delivered
considers the properties of temporal data, the wireless interface data volume. Even though the protocol can achieve high
heterogeneity, and the delay constraints. Although the throughput when the density of the nodes is average, its
simulation results demonstrated the superiority of the algorithm efficiency will decrease in low- or high-density situations. CR-
over other competitors, the impact of increased data requests in SDVN [148] uses SDN to select the most stable vehicle routing
high-density VANETs and the impact of service requests from path over cognitive vehicular networks. The algorithm is
other nearby network interfaces were not considered. LSB-OR simulated and compared with other counterparts using PDR,
[142] introduces an optimization-based routing algorithm with a E2E delay, and routing overhead for performance analysis. Even
source routing-based forwarding initiation strategy to efficiently with extensive simulation, the impact of the increased rate of
deliver and cache flow information in the appropriate nodes routing requests for the primary users was not realized. FR-PU
using route validity times. The protocol is evaluated using [149] studied the effect of sudden changes in vehicle motion on
various metrics such as time complexity, accuracy, PDR, data routing by using vehicle mobility information. It aims to
latency, and overhead. Although the protocol has a very low find the reliable, shortest paths before data transmission
processing delay, it may negatively impact packet reception due proactively. Numerical results indicated the protocol can be
to its failure to consider node density and controller overhead. more efficient and reliable than AODV and OLSR protocols
LBR [143] aims to avoid re-routing waste in high-speed based on E2E delay, PDR, and routing overhead. The protocol
SDVN networks by accelerating forwarding rules through a link performs satisfactorily even at high vehicle mobility and
lifetime prediction method. The simulation is set to a high-speed extended application duration. However, the protocol
train network with the cellular network as roadside performance will degrade when no nearby vehicles are available
infrastructure. The analysis demonstrated that the protocol when the primary relay changes motion or direction.
outperforms the GPSR-L protocol with successful data delivery. ORUR [150] leverages cloud-based SDN technology to
SDCoR [144] uses RL with SDN to adaptively select the best optimize bandwidth utilization and channel congestion through
routing strategy in IoV under different traffic situations. data transmission over multiple road segments. The protocol
Although the analysis has demonstrated that the protocol can utilizes edge servers to minimize latency and reduce bandwidth
achieve better PDR than AODV and GPSR strategies, the usage from multiple cloud access. The protocol evaluation
method's effectiveness must be verified with other QoS metrics, requires additional analysis in terms of delay and PDR. SCGRP
such as E2E delay and throughput. Also, the overhead of the [151] aims to reduce transmission delay and overhead over
SDN controller will increase intensely with the increased SDN-based vehicular communication. The protocol is evaluated
number of routing requests. using PDR, E2E delay, NRL, and hop counts ratio. However, the
Geo-SDVN [145] uses an SDVN architecture without protocol performance will decrease with high node mobility or
depending on the existence of RSUs for data transmission. sparse networks. SD-TAODV [152] is an AODV-based secure
Accordingly, it can reduce the cost of infrastructure deployment routing protocol. To minimize routing overhead, the controller
and improve node accessibility. The protocol was evaluated calculates the routing path and updates it only if a new path with
using mathematical model methods in terms of PDR, average a higher trust value is presented. The numerical analysis shows
delay, and transmission overhead. However, the performance the protocol can outperform the traditional AODV algorithm
was not analyzed in various VANET scenarios with varying regarding throughput and routing overhead. However, the E2E
node mobility. HRLB [101] uses a three-level hierarchical delay is higher than that of AODV due to using the traditional
routing approach with load balancing to maintain high PDR in AODV protocol with minor modifications. Also, the controller
real-time situations. However, this approach may increase overhead will increase in dense VANETs where the trust model
routing overhead and latency, especially in non-uniform requires more computations. AVC-BTAT [153] utilizes an
transportation areas. The evaluation of the algorithm is verified adaptive transmission method with selective modulation and
through simulation using PDR, throughput, average delay, and power control to improve the data delivery over the trunk link
average hop count metrics. Generally, the protocol needs to between the cluster head and BS and minimize packet delay
consider vehicle mobility and density in performance analysis to during transmission over V2I links. The analysis demonstrated
achieve the best results. the protocol efficiency in supporting high mobility VANET by
SESAC [146] combines a social-aware clustering technique improving cluster stability and QoS performance regarding
with a 5G network to improve cluster stability, prevent BER, SNR, and throughput. However, node mobility may pose
congestion, and reduce packet loss. The protocol is simulated more challenges in real-time applications.
and evaluated regarding cluster lifetime and clustering overhead. SDGR [154] utilizes node position, traffic density, and digital
However, the analysis does not include other critical QoS map to determine the optimal routing paths and avoid
metrics, such as PDR, delay, and throughput. LT-NSR [147] connectivity issues such as local maximum and sparse
finds the routes with the longest lifetime based on the current disconnections. The protocol was compared to AODV and
traffic status and vehicle context information using over cellular GPSR protocols regarding PDR and delay, using a moderate
6 VOLUME XX, 2017
number of vehicles (50 to 200). However, further analysis and routing path with minimum delay and low overhead in dynamic
evidence are needed to confirm its suitability for sparse network density. While this model increases protocol
VANETs. dSDiVN [155] deploys mobile SDN controllers to fill adaptability with various scenarios in VANETs, the calculation
the gap of the lack of SDN in infrastructure-less VANET areas. of the next hop may generate high delay and controller overhead.
Even though the protocol shows high reliability in mathematical The protocol is evaluated regarding PDR, E2E delay, and routing
simulation, further experimentation is needed to confirm the overhead. However, the impact of high mobility on routing
protocol's efficiency regarding latency, throughput, and routing computation was not considered in the performance analysis.
overhead. HSDVR [156] uses a clustering technique to address Spray-and-Prey [107] protocol utilizes centralized SDN to
the potential disconnections with the main SDN controller. The generate multiple copies of data in the network. The network
protocol is simulated and evaluated in an urban scenario using overhead and buffer expenses are minimized by removing other
several metrics, including PDR, E2E delay, throughput, and copies in the network when a single copy reaches the intended
routing overhead. However, further evaluations are needed to destination. While this protocol reduces packet latency and
verify the protocol’s reliability in various situations, such as network resource costs, the delay and bandwidth utilization will
partial RSU distribution and correlated mobility patterns. SVAO increase with the number of nodes. GeoBroadcast [9]
[110] leverages SDN to improve transmission efficiency by implements a module to automatically control the geographical
utilizing a two-level design to find the road segments and routing places of RSUs to be used as a basis for data forwarding. The
path along which a packet should be forwarded. The protocol is performance analysis showed its efficiency in reducing the
evaluated through simulations and compared to OLSR, DSR, controller overhead and bandwidth consumption with minimum
DSDV, and DB routing protocols. It considers the influence of delay. However, the protocol did not analyze the key challenges
node density and velocity on the data transmission rate and of VANET, such as high mobility and low density.
average delay. However, due to a slight change made to However, Fig. 18 presents the statistical analysis of standard
traditional AODV, the protocol generates many control packets metrics used for performance evaluation in the reviewed works.
in the event of a link failure as the network density increases. It is observed that most of the works use metrics such as E2E
IDVR [157] utilizes a centralized SDN to collect real-time delay, PDR, throughput, overhead, and network load to analyze
traffic knowledge and select the optimal routing path from a set the routing protocols.
of candidate shortest paths. The protocol is evaluated and Moreover, a statistical study on the optimization parameters
compared to other protocols, including IRTIV, VDLA, and used to improve the reviewed works in this survey has been
GPCR, using E2E delay and throughput metrics. However, the conducted. As depicted in Figure 19, delay minimization is the
computation of multiple paths may lead to high overhead on the most significant parameter for routing protocol optimization,
SDN controller, which requires further optimization. CESDT owing to the critical importance of delivering data packets with
[158] leverages service bandwidth requirements to schedule and the least possible delay for safety-related services. Routing
route transmission requests over multiple paths of various protocols such as SPIDER, CDRS, TDCR, CLR-SDVN, Pretti,
network interfaces in heterogeneous vehicular networks. To and CDSRP prioritize delay optimization.
evaluate its performance, a prototype urban traffic monitoring Furthermore, protocols such as PT-GROUT, POLAR, ERA,
application is developed that collects overall traffic conditions CRS-MP, and VDR-DRL consider communication distance a
from crowd-sourced taxi traces in Shenzhen, China. The key parameter due to its impact on link reliability and data
protocol is evaluated using PDR and compared to OLSR and delivery. So, the optimal selection of the next relay can minimize
GPSR. Other QoS metrics were not considered in evaluating the packet loss and increase reliability. However, the protocols
protocol's effectiveness. Besides, data offloading over multiple considering this optimization parameter use several parameters
interfaces may cause more delays and complexity. Pretti [159] and constraints, such as node mobility, communication distance,
applies SDN and trajectory prediction to minimize and buffer size. DMPFS, CCDEC, SeScR, SD-IoV, and HRLB
communication costs and ensure message delivery for multicast are examples of such protocols.
data routing. Extensive experiments are conducted using Regarding overhead control, the reviewed algorithms can be
different traffic scenarios. While the evaluation demonstrates the divided into two classes. Some protocols aim to optimize the
protocol's effectiveness in various scenarios in terms of latency overhead of the SDN controller through hierarchical control
and PDR, there is room for improvement regarding message architecture or load balancing, such as ICDRP-F-SDVN,
validity and node mobility considerations. TRIBRID, SVAO, SURFER, SDVN-GRA, CTDD, and ORUR
CDSRP [160] presented an SDN-based scheduling and protocols. Others optimize the overhead by utilizing mobility
routing algorithm for data dissemination over hybrid I2V and prediction to minimize the number of exchanged beacons, such
V2V communication links. The protocol is simulated, and its as IM-DOS, SPIDER, ERA, SD-IoV, and SESAC protocols.
performance is evaluated under various traffic situations using However, other optimization parameters have received less
various metrics such as scalability gain, broadcast productivity, attention from researchers due to the difficulty of modeling these
gains distribution, delivery rate, and latency. However, further parameters or the lack of sufficient data from the network
examination at the MAC layer is needed to validate the model in infrastructure. Therefore, it is necessary to intensify efforts to
real-time VANET situations. CRP [108] protocol represents one optimize protocols with more parameters to create more reliable
of the early works that deal with SDN in VANET routing. The protocols that can adapt to different network situations and
algorithm leverages global network information to determine the communication requirements.
6 VOLUME XX, 2017
TABLE 4.
CHARACTERISTICS OF SDVN ROUTING SCHEMES. THE SYMBOLS “SP”, “MP”, “U”, “MC”, “B”, “L”, “M”, “H”, “D”, S”, ✓, “-“AND  INDICATE SINGLE PATH,
MULTI-PATH, UNICAST, MULTICAST, BROADCAST, LOW, MEDIUM, AND HIGH, DENSE, SPARSE, INCLUDED, UN-INCLUDED, AND UN-MENTIONED RESPECTIVELY.
6 VOLUME XX, 2017
TABLE 5.
SUMMARY OF KEY FEATURES OF SDVN ROUTING PROTOCOLS.
6 VOLUME XX, 2017
TABLE 6.
COMPARISON OF PERFORMANCE METRICS AND TECHNIQUES USED IN SDVN ROUTING PROTOCOLS. THE SYMBOL ✓ INDICATES THAT THE METRIC IS INCLUDED;
THE BLANK SPACE INDICATES THE METRIC HAS NOT BEEN INCLUDED.
End-to-End Delay
Packet Delivery Ratio
Packet Loss Ratio
Throughput
Bit Error Rate
Jitter
Routing Overhead
Connectivity Probability
Network Load
Normalized Routing Load
Average Routing Reply Ratio
Average Routing Discovery Time
Routing Request Ratio
Link Failure
Percentage Idle Time
Route Lifetime
Availability of Recovery Process
Routing Maintenance
DTN Support
Scoring Technique
Flooding Mitigation Concern
QoS Support
AI Adoption
Protocol
ICDRP-F-SDVN ✓ ✓ ✓ ✓ ✓ ✓
SDVN-MiCR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CDRS ✓ ✓ ✓ ✓ ✓ ✓ ✓
DMPFS ✓ ✓ ✓ ✓ ✓ ✓
PT-GROUT ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SPIDER ✓ ✓ ✓ ✓ ✓ ✓ ✓
GLS ✓ ✓ ✓ ✓ ✓ ✓
SELAR ✓ ✓ ✓ ✓ ✓
POLAR ✓ ✓ ✓ ✓ ✓
SURFER ✓ ✓ ✓ ✓ ✓ ✓ ✓
CCDEC ✓ ✓ ✓
SFRS ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
ERA ✓ ✓ ✓ ✓ ✓ ✓ ✓
IM-DOS ✓ ✓ ✓ ✓ ✓ ✓
HSDN-GRA ✓ ✓ ✓ ✓ ✓ ✓
ACA-RP-SDVN ✓ ✓ ✓ ✓
TDCR ✓ ✓ ✓ ✓ ✓ ✓ ✓
SeScR ✓ ✓ ✓ ✓ ✓
V-TLRH ✓ ✓ ✓ ✓
SDMEV ✓ ✓ ✓ ✓ ✓
TRIBRID ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SD-IoV ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
QRA ✓ ✓ ✓ ✓ ✓ ✓ ✓
VDR-DRL ✓ ✓ ✓ ✓
CLHLB ✓ ✓ ✓ ✓
MFCAR ✓ ✓ ✓ ✓
NSDD-SDV ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CRS-MP ✓ ✓ ✓ ✓
CLR-SDVN ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
IMCR ✓ ✓ ✓
CTDD ✓ ✓ ✓ ✓
LSB-OR ✓ ✓ ✓ ✓ ✓ ✓
LBR ✓ ✓ ✓ ✓
SDCoR ✓ ✓ ✓
GEO-SDVN ✓ ✓ ✓
HRLB ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SESAC ✓ ✓ ✓ ✓
LT-NSR ✓ ✓ ✓ ✓ ✓ ✓ ✓
CR-SDVN ✓ ✓ ✓ ✓
FR-PU ✓ ✓ ✓ ✓ ✓ ✓
ORUR ✓ ✓ ✓ ✓ ✓
SCGRP ✓ ✓ ✓ ✓ ✓ ✓
SD-TAODV ✓ ✓ ✓
AVC-BTAT ✓ ✓ ✓ ✓ ✓ ✓
SDGR ✓ ✓ ✓ ✓
dSDiVN ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
HSDVR ✓ ✓ ✓ ✓ ✓ ✓ ✓
SVAO ✓ ✓
IDVR ✓ ✓ ✓ ✓ ✓ ✓
CESDT ✓ ✓ ✓ ✓ ✓ ✓ ✓
Pretti ✓ ✓ ✓ ✓
CDSRP ✓ ✓ ✓ ✓ ✓ ✓
CRP ✓ ✓ ✓ ✓
Spray-&-Prey ✓ ✓ ✓
GeoBroadcast ✓ ✓ ✓ ✓
6 VOLUME XX, 2017
TABLE 7.
OPTIMIZATION PARAMETERS USED IN THE EXISTING ROUTING PROTOCOLS FOR SDN-BASED VANETS. THE SYMBOL ✓ INDICATES THAT THE PARAMETER IS
INCLUDED; THE BLANK SPACE INDICATES THE PARAMETER HAS NOT BEEN INCLUDED.
Delay Optimization
Quality of Neighbors
Link Reliability
Hop Count Accountability
Vehicles Clustering
Zone Segmentation
RSU Needs Consideration
Road intersection consideration
Communication Distance Consider
Localization Inf. Requirements
Traffic Density Awareness
Mobility Predication Consideration
Security Consideration
Vehicles Mobility Concern
Increased Requests Efficiency
Fault Tolerance
Traffic Condition Adaptability
Multi-Constraints Decisions
Context Awareness
Overhead Control
Node Energy Awareness
Load Balancing
Protocol
DRP-F-SDVN ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SDVN-MiCR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CDRS ✓ ✓ ✓ ✓ ✓ ✓
DMPFS ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
PT-GROUT ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SPIDER ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
GLS ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SELAR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
POLAR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SURFER ✓ ✓ ✓ ✓ ✓
CCDEC ✓ ✓ ✓ ✓ ✓ ✓ ✓
SFRS ✓ ✓ ✓ ✓ ✓ ✓ ✓
ERA ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
IM-DOS ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
HSDN-GRA ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
ACA-RP- ✓ ✓ ✓ ✓ ✓
SDVN
TDCR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SeScR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
V-TLRH ✓ ✓ ✓ ✓ ✓
SDMEV ✓ ✓ ✓ ✓ ✓ ✓
TRIBRID ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SD-IoV ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
QRA ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
VDR-DRL ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CLHLB ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
MFCAR ✓ ✓ ✓ ✓
NSDD-SDVN ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CRS-MP ✓ ✓ ✓ ✓ ✓ ✓ ✓
CLR-SDVN ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
IMCR ✓ ✓ ✓ ✓
CTDD ✓ ✓ ✓ ✓ ✓ ✓
LSB-OR ✓ ✓ ✓ ✓ ✓ ✓ ✓
LBR ✓ ✓ ✓ ✓ ✓ ✓
SDCoR ✓ ✓ ✓ ✓ ✓
GEO-SDVN ✓ ✓ ✓ ✓ ✓
HRLB ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SESAC ✓ ✓ ✓ ✓ ✓ ✓ ✓
LT-NSR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
CR-SDVN ✓ ✓ ✓ ✓ ✓ ✓
FR-PU ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
ORUR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SCGRP ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
SD-TAODV ✓ ✓
AVC-BTAT ✓ ✓ ✓ ✓ ✓ ✓
SDGR ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
dSDiVN ✓ ✓ ✓ ✓ ✓ ✓
HSDVR ✓ ✓ ✓ ✓
SVAO ✓ ✓ ✓ ✓ ✓ ✓
IDVR ✓ ✓ ✓ ✓ ✓ ✓
CESDT ✓ ✓ ✓ ✓ ✓
Pretti ✓ ✓ ✓ ✓
CDSRP ✓ ✓ ✓ ✓ ✓ ✓ ✓
CRP ✓ ✓ ✓ ✓ ✓ ✓
Spray-&-Prey ✓ ✓ ✓ ✓ ✓
GeoBroadcast ✓ ✓ ✓ ✓ ✓ ✓
6 VOLUME XX, 2017
with SDVN routing is inherently challenging and severely

limited due to various factors, including cost, regulations, time
limitations, and the necessity of open space [163], [164]. As a
result, researchers rely heavily on simulators and emulators to
assess the effectiveness of their routing algorithms by utilizing
comparable or identical traffic scenarios and management levels.
Unfortunately, only a limited number of simulators offer support
for SDN in the VANET environment [164]. For example,
NetSim ver.11 has a module that supports the use of SDN in
implementing VANET protocols [163], [164]. Regarding the
Veins simulator, due to its compatibility and OpenFlow module
integration, numerous papers have employed it in the context of
SDVN [165]. Eclipse MOSAIC is a powerful tool combining
multiple simulators in a single package to model and evaluate
ITS solutions. However, we have not found substantial research
on using this simulator for SDVN routing [162]. Also, no
research is related to using the VENTOS simulator in SDVN
routing [166]. Furthermore, there is no mention of SDN support
FIGURE 18. The statistical analysis of standard QoS metrics in the documentation of VANETsim simulator that comes with
used for performance evaluation in the reviewed protocols. built-in modules. So, it poses a challenge in extending the
simulator to accommodate emerging technologies [167].
B. EVALUATION TOOLS AND TESTBEDS USED IN
Fig. 20 illustrates a statistical analysis of the most utilized
RESULTS ANALYSIS.
simulation tools and testbeds employed to assess SDVN data
Now, we focus on the simulation tools used to analyze the
routing. The graph highlights that most researchers rely on the
effectiveness of the techniques reviewed in section V. Table 8
NS3 simulator to validate their proposals. Besides, Python,
compares the reviewed protocols regarding performance
OMNET++, and Matlab were the secondary options for most
evaluation and analysis techniques employed. Since the VANET
researchers to test their protocols.
routing protocols are a highly interdisciplinary field, they require
In terms of testbeds, only a few studies utilized them to
flexible tools for performance evaluation. These tools should be
explore the application of SDVN data routing [164]. Testbeds
adaptable to various domains, including network design,
integrate various SDN features to enable researchers to test
communication strategies, optimization theory, and emerging
protocols under authentic conditions and obtain experiment
fields like ML, stochastic geometry, and game theory [161].
results. However, there are several limitations to conducting
such realistic experiments in SDVN models, which include:
1. Security challenges of deploying protocols.

2. Weather situations.
3. Communication channels issues.
4. The cost of used infrastructures.
5. Network infrastructure adaptability and compatibility.
6. The challenge of retrieving data from the network.
Therefore, tackling all these problems and obstacles is
essential when utilizing realistic experiments. Another approach
is to create alternative simulation tools that enable researchers to
effectively assess various VANET scenarios and SDN
programmability. Such tools must simulate mobility patterns and
networking protocols with more realistic parameters to reach
FIGURE 19. Optimization parameters used in reviewed SDVN good results.
routing protocols.
Several mobility simulators are available for evaluating and
Network simulators help researchers investigate a network's modeling traffic systems, including SUMO [168], VISSIM
operation under varying situations that might be challenging or [169], SimMobility [170], and PARAMICS [171]. SUMO, an
expensive to realize using real systems and entities, specifically open-source traffic simulation application, has gained popularity
in the context of VANETs. So, they are mainly focused on in recent decades. Given its high efficiency and compatibility,
testing new protocols or proposing modifications to current ones SUMO is widely used in traffic modeling in VANET routing
in a controlled and repeatable fashion [162]. Compared to the works. It can model various types of vehicles, public transport,
time and expense of establishing a complete real-time testbed, and even pedestrians in traffic systems. Additionally, it provides
network simulators are moderately fast and affordable. accurate vehicle traces in any map imported into the system,
Consequently, conducting extensive real-world experiments such as open street maps.
6 VOLUME XX, 2017
TABLE 8.
COMPARISON OF PERFORMANCE EVALUATION AND RESULT ANALYSIS TECHNIQUES.
The selection of the simulation area can significantly impact experiment. This may lead to questionable results, as adjusting
the evaluated results and performance analysis. For instance, to the environment requires working with various scenarios.
when a large area is chosen, the vehicles may move far apart, Moreover, specific protocols did not mention the simulation
which can hinder the presence of next-hop nodes near the data area in their analysis, including CDRS, DMPFS, CCDEC,
carrier. To tackle this issue, some studies, such as GLS, LBR, and HSDVR. On the other hand, selecting a small area
SURFER, CTDD, CDRS, and CLHLB have increased the can reduce the computational overhead caused by the
number of nodes or reduced the default speed in the simulation simulator and provide better outcomes for routing data in
VOLUME XX, 2017 9
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4
urban areas. However, the results may not be reliable for real- VIII. OPEN ISSUES AND FUTURE CHALLENGES
time data transmission where the area of protocol deployment The former sections have summarized most SDVN routing
is increased. These protocols include SELAR, MFCAR, FR- protocols and comprehensively analyzed their performance
PU, SDGR, and dSDiVN. Furthermore, most of the reviewed regarding delivery guarantees, routing efficiency, and other
algorithms consider urban scenarios for performance analysis. factors. Despite the efforts to develop SDVN routing
Nevertheless, urban areas' population, traffic density, road protocols, several gaps should be resolved. The incorporation
distribution, and communication requirements vary due to of SDN and VANETs is still in its early stages, and many
geographical features. Nonetheless, some protocols that have issues need to be refined. These issues must be thoroughly
considered the features of the geographical area and investigated to ensure flexible and robust usage of SDN in
parameters include SPIDER, CESDT, HRLB, and CLR- VANET routing with efficient allocation of physical
SDVN. Also, only three protocols have been evaluated in the resources. Therefore, there is still a performance gap between
highway scenario, namely SDVN-MiCR, LBR, and LT-NSR the current routing approaches and the ideal ones that can fully
protocols. Finally, the protocols should be compared with leverage the advantages of SDN. Table 9 presents upcoming
similar routing mechanisms for performance analysis and research directions, investigated problems, recommended
efficiency validation. Fig. 21 depicts a statistical analysis of solutions, and suggested references to supplement this study.
the most used protocols for comparison. However, no specific So, the open issues and future challenges are outlined here:
routing protocols can be deemed as a reference for analyzing 1) Lack of high mobility consideration: While SDN offers
the performance of all SDVN routing algorithms. GPSR, flexible and programmable network control, its
AODV, and OLSR protocols are the main benchmarks for the suitability for VANETs is still at a primary stage. The
performance analysis of many reviewed works. TDCR, AVC- rapid changes in topology make it challenging to gather
BTAT, and dSDiVN did not compare with other works. network information correctly. As a result, an efficient
solution is necessary to maintain high performance in
highly dynamic VANET and decrease latency in the
network. Nonetheless, many routing schemes have been
developed to address the issue of rapid network topology
updates [172] [173]. To ensure efficient routing and
traffic management, it is critical to maintain a reliable
and updatable global topology view at the SDN
controller. In recent years, the fog and edge computing
techniques have been introduced to support the global
controller. In addition, proactive mobility management
algorithms and hybrid SDN architecture can assist in
delegating partial load for mobility management,
resulting in new mobility management techniques.
2) Management of flow rule policies: In SDN, the data
forwarding rules comprise general rules. The
improvement of the flow rule policies is necessary to
FIGURE 20. Simulation tools used in reviewed works.
maintain seamless services. For instance, the real-time
data are uploaded to BSs/RSUs for processing using
general flow policies rather than specific policies related
to data forwarding. BSs/RSUs generate local data
routing policies using their local network topology
information. Once data is processed, BSs/RSUs forward
it to the central controller over the SBI interface.
3) Lack of research in dedicated trajectory prediction
methods: Trajectory prediction refers to the estimation
of the future state of a vehicle based on its current state.
Conventional trajectory prediction models typically rely
on different models of motion analysis, such as the
Markov chain, kinematic model, constant speed
mobility model, and others using extensive historical
data from the vehicle and neighboring nodes [8]. In
SDN, limiting the collected data is essential to avoid
overwhelming the controller. So, it is optimal to utilize
limited available vehicle mobility information to predict
the vehicle trajectory within a limited period to satisfy
FIGURE 21. Testbench protocols used in reviewed works.
the SDVNs routing requirements.
VOLUME XX, 2017 9
TABLE 9.
OPEN RESEARCH CHALLENGES OF SDN-BASED VANET DATA ROUTING
VOLUME XX, 2017 9
4) Further reduction of communication overhead: Beacons the monitoring of link status can help in re-routing data
and data packets transmitted to/from the controller result through previously recognized paths, which may
in high communication overhead. Predicting the vehicle minimize the delay. Generally, neither approach is a
trajectory in the controller is the most effective solution feasible solution for entirely resolving the failure. So,
to reduce this overhead. In this scenario, the number of future failures can be prevented by tracing back and
status messages is significantly minimized, and the determining the underlying cause. By incorporating
controller constructs the network graph using the historical data and current link status into machine
received beacons and link estimates. Numerous relevant learning methods, the links with the most appropriate
studies have utilized clustering techniques to minimize connectivity can be selected for data routing.
the overhead [174]. Future research should empower the 8) The need for multicast routing: Most current studies
controller to allocate routing queries to local controllers concentrate on delivering unicast routing. Nevertheless,
with many routing metrics or explicit routing strategies multicasting is an essential routing method for many
based on the network situation to reduce communication critical VANET applications, such as collision
overhead further. Developing intelligent schemes to avoidance and platooning driving. Coordinating
evaluate the trade-off between centralized and multicast requests within the global view controller is
distributed routing is worthwhile, such as adaptively crucial for effective one-to-many data delivery.
switching the SDVN routing method under different 9) Security consideration: Security is a critical concern in
network conditions [130]. VANET communications where it directly impacts the
5) Lack of research in forwarding algorithms in controllers: safety of in-car travelers, vehicles, pedestrians, and
Many current studies utilize the static shortest path others. SDVNs, with the presence of a controller, can be
strategy, such as Dijkstra's algorithm, to find the optimal less vulnerable to cyber-attacks than other types of
routing paths. Still, in VANETs, most links between wireless vehicular paradigms. Most researchers have
nodes have a limited validity period, rendering prioritized the confidentiality and integrity requirements
conventional shortest-path schemes unsuitable for static for the network and SDN controller access.
networks. This results in the routing problem being Consequently, any false information can result in low
converted to finding the best path in a dynamic timetable delivery rates and high latency. However, many cyber-
graph [115], [148]. In an SDVN, the controller may have attacks, such as GPS spoofing, Denial of Service (DoS)
to manage millions of routing queries concurrently, attacks, tunneling attacks, and gray hole attacks, can
making it crucial for route planning algorithms to be significantly impact the SDVN efficiency [176], [177].
efficient enough to fulfill networking requirements. Furthermore, the integration with other technologies
However, the current algorithms rely on pre-computed such as cloud/fog computing and 5G networks will lead
network data, which is unattainable in dynamic SDVNs. to more security risks [178]. In general, security for
Therefore, to achieve an effective timetable-dependent SDVN routing has not been extensively studied. So, it is
routing scheme, it is essential to adopt road route crucial to develop preventive measures and strategies to
planning algorithms or develop an alternate method as a safeguard the network against malicious access and
future research direction. ensure the security of both the SDN controller and
6) Lack of AI in routing management: There is a lack of communication data.
research on utilizing AI to improve the routing 10) Network scalability: The majority of SDVN routing
performance of SDVNs. Specifically, increasing data protocols are not scalable, as they are typically designed
traffic presents a highly promising method to address the for urban or highway scenarios [179]. It is preferable for
dynamic and large-scale nature of SDVNs. For instance, the routing protocol to be scalable to handle the
statistical learning, neural networks, or deep learning unexpected increasing in the number of nodes and
could be utilized to create the routing metric, routing support reliable services in various network situations.
policy, or even per-flow QoS parameters based on However, the scalability of SDVN networks can be
historical and real-time data. Despite the potential improved by adopting logical and decentralized
benefits of AI in SDVNs, new challenges may arise, controller architecture. For instance, a logical controller
such as computational complexity, learning and manned must be assigned in a scenario with varying vehicular
strategies, and pre-processing real-time data [175]. density to ensure the best data delivery in high-traffic
7) The need for a recovery strategy of failed routing: In scenarios. Besides, the integration of fog and edge
real-world scenarios, computed routes may become systems can help in the minimization of the controller
unusable if a link breaks due to node dynamicity. overload. For instance, standby offloading resources
Unfortunately, current works do not provide effective such as drones and parked or low-mobility vehicles can
recovery mechanisms from route failures. The most reduce the load on resident controllers. Also, the traffic
straightforward approach to repair a failed route is to prediction algorithms can be utilized to predict traffic
initiate local recovery requests or send the routing query density and load balancing to enhance scalability in
to the controller for a new route. Accordingly, the E2E scenarios with many nodes managed by a single
delay will be more with high data loss rate. Additionally, controller.
VOLUME XX, 2017 9
11) Controller bottleneck consideration: The increasing To optimize the number of controllers, it is essential to
volume and complexity of traffic patterns generated by determine the best locations for placing them. With
vehicles may overwhelm the SDN controller. In such a SDVN, the controller placement problem becomes more
situation, the SDN controller becomes a limiting factor complex as new candidates, such as BSs and RSUs, can
in terms of processing capacity, communication be considered potential controllers [182]. One potential
bandwidth, or overall performance. One of the best solution to address the challenges of controller
solutions is to explore distributed control strategies to placement is to develop reward-based cooperation
delegate decision-making processes to edge devices or solutions that utilize BSs or RSUs as SDN controllers.
distributed controllers to enhance reliability and reduce Machine learning algorithms such as reinforcement
the load on the central controller. This can help improve learning (RL) can be employed to determine the optimal
performance and ensure the system can handle a higher localization of SDN controllers, including the number
density of vehicles [179]. Besides, the complex routing and distribution of controllers [183]. However, none of
mechanisms and intelligent decision-making processes the SDVN routing proposals address this issue which is
can also contribute to controller bottlenecks, particularly still doubtful and needs more research effort.
when utilizing historical and real-time data [175]. So, the 15) Lack of infrastructural and economic considerations:
controller bottleneck should be considered in upcoming Although the cellular infrastructure is already in place
SDN-based data routing work. and can provide extensive coverage areas with high
12) Further reduction of communication latency: Achieving throughput and capacity, there are still financial barriers
low latency is crucial for effective vehicular to achieving uninterrupted SDN services via such
communication. Several studies have been proposed to networks [184]. The main costs are associated with the
leverage cloud/edge computing techniques to provide deployment of SDN controllers and the need for new
caching services and address the various routing needs applications and technologies, such as edge services and
[44] [99]. Due to various factors, such as information cybersecurity models. Accordingly, the development of
collection, channel conditions analysis, computational an effective economic model can enhance the integration
workload, and vehicles tracking techniques, the network of SDN and nearby infrastructures. Besides, studying
complexity will increase with more latency [180]. Some vehicle traffic in real-time is crucial for determining the
studies have suggested using multiple interfaces like hardware and software requirements and planning the
DSRC and C-V2X communications to minimize services that can be deployed.
latency. Frequent handovers caused by high-density 16) Lack of optimal localization mechanisms: All the
vehicle traffic can delay transmission and degrade reviewed protocols rely on the GPS to collect vehicle
performance. So, to improve latency, the optimization of coordinates and for localization purposes. However,
average delay from different angles, such as data vehicles have high speeds and different mobility
processing, routing establishment, flow rule patterns, requiring exact localization at small intervals.
propagation, and failure recovery is important. In many communication protocols, the GPS update time
13) Lack of real-world implementations: Despite the is 1 second, which is ineffective for accurate vehicle
numerous theoretical concepts and ideas proposed in the localization [185]. The imprecise vehicle localization
literature, they have yet to be validated through real- resulting from the 1-second GPS update time can lead to
world implementations. As an exception, in [164], more challenges and incorrect routing decisions.
Secinti et al. provided a testbed for SDVN using However, various models have been proposed to
Raspberry Pi for both RSU and vehicles that are Wi-Fi determine the precise position of vehicles. A cooperative
enabled. In [181], Sadio et al. proposed a more advanced localization system can help mitigate faults resulting
testbed for SDVN with multiple controllers, including from GPS imprecision [186]. Also, a video-based
the Twinand Wi-5 controllers. Four Zodiac FX switches positioning system and distance estimation are samples
are used to build the SDN backbone. To implement of optional solutions here [187]–[189]. Unfortunately,
SDVN in the real world, a comprehensive assessment of none of these models have been utilized and tested with
the vehicular network is required to identify which SDVN data routing.
control layer component can be added or removed from 17) Lack of weather affects consideration: Adverse weather
the infrastructure plane to leverage SDVN's benefits conditions, such as heavy rainfall, high winds, and
fully. In the future, greater emphasis should be placed on snowfall, can affect vehicle mobility and traffic density,
developing real-world implementations and simulation causing a high impact on signal propagation and data
tools to evaluate the performance of SDVN systems. transmission [190] [185]. Developing a mechanical
This will help determine which control layer component system that can effectively cope with such conditions
should be decoupled from the infrastructure plane. can improve data routing performance. Also,
14) The consideration of multiple controller placement: establishing short P2P communications among
Several SDVN routing protocols have proposed using controllers and vehicles, as well as among vehicles
numerous controllers to maintain the proper functioning themselves, can help minimize the impact of weather
of the network in the event of a single controller failure. changes on data routing.
VOLUME XX, 2017 9
18) Further reduction of computation load: The deployment VIII. CONCLUSIONS

of SDVN in real-time environments, particularly in SDN, as an emerging technology, can potentially resolve
dense traffic, can result in a significant computation load several problems in vehicular networks. Its implementation in
where a larger number of requests will be generated VANETs has increased considerable interest from the
which leads to potential performance issues [191], [192]. academic world and industry. First, we provide full details
Therefore, offloading computation workloads to the about VANET networks, including their architecture,
vehicles or edge servers can accelerate the computation communications, and challenges. Next, a general description
and save energy. However, the scheduling of offloading of traditional routing protocols used in VANET is provided.
workloads in SDVN routing is still challenging and More specifically, this research paper offers a thorough review
needs more investigation. Besides, the development of of the presenting SDVN architecture and routing schemes,
big data processing techniques can help in the mitigation along with related information on data routing under SDVN.
of such challenges. This survey has provided a comprehensive examination of
19) Integration of other techniques: The cooperation of SDN-based VANET routing protocols, shedding light on their
SDVN with other technologies, such as cloud/edge underlying mechanisms, forwarding algorithms, and
computing, and blockchain techniques, can enhance architectural considerations. The survey explored each
communication reliability and overall QoS in various protocol individually, highlighting their robustness,
VANET scenarios. However, a comprehensive limitations, and suggested improvements. SDVN-enabled
investigation is necessary to assess the feasibility and routing protocols have been divided into various categories
effectiveness of such techniques in SDVN models. Such based on their modes of operation, routing mechanisms,
integration can also introduce new security and privacy transmission modes, and others. By comparing and analyzing
vulnerabilities, which must be investigated. various optimization criteria, performance evaluation
20) Deficiency of simulation platforms: VANET and SDVN techniques, and simulation tools, this survey has contributed
networking models have different properties, including to a better understanding of the current state of SDN-based
signal coding, MAC, security requirements, and VANET routing protocols. Lastly, the research identifies the
congestion handling. Therefore, it is essential to challenges of SDVN and provides future directions for
establish a general simulation platform that includes researchers interested in SDVN architecture or data routing.
typical vehicular network scenarios to facilitate fair Overall, this survey not only provides a comprehensive
evaluations and comparisons of SDVN routing overview of SDN-based VANET routing protocols but also
solutions. Researchers often use different network serves as a platform for researchers to build upon and
simulation tools and mobility models to evaluate innovate. It lays the foundation for further exploration,
proposed solutions. However, many simulators do not enabling the development of more efficient, reliable, and
support SDVN modules, and most lack integration with scalable routing solutions in the dynamic and challenging
emerging technologies, critical attributes for SDVN vehicular ad hoc network environment.
routing design. Significant improvements are necessary
to meet SDVN routing requirements, including REFERENCES
[1] S. Sharma and B. Kaushik, “A survey on internet of vehicles:
developing new modules and simulation tools [163]. Applications, security issues & solutions,” Vehicular
Further, simulators should include mobility models that Communications, vol. 20, p. 100182, 2019, doi:
mimic real vehicle mobility to establish a realistic 10.1016/j.vehcom.2019.100182.
[2] M. Lee and T. Atkison, “VANET applications: Past, present, and
simulation situation. future,” Vehicular Communications, vol. 1, p. 100310, 2020,
21) Lack of standardization: Interworking gaps among doi: 10.1016/j.vehcom.2020.100310.
heterogeneous networks are challenging in SDVN [3] H. Zemrane, Y. Baddi, and A. Hasbp, “Mobile adhoc networks
networking. As the future of SDVN is not restricted to for intelligent transportation system: Comparative analysis of the
routing protocols,” Procedia Comput Sci, vol. 160, no. 2018, pp.
communication among vehicles, it will include new 758–765, 2019, doi: 10.1016/j.procs.2019.11.014.
technologies and entities with several features from [4] A. Dua, N. Kumar, and S. Bawa, “A systematic review on
different manufacturers. Defining the boundaries of how routing protocols for Vehicular Ad Hoc Networks,” Vehicular
Communications, vol. 1, no. 1, pp. 33–52, 2014, doi:
far one can integrate SDN into VANET is yet to be 10.1016/j.vehcom.2014.01.001.
established. As a result, mutual exclusiveness can occur, [5] L. Zhao, K. Yang, Z. Tan, X. Li, S. Sharma, and Z. Liu, “A Novel
leading to communication failures between vehicles. Cost Optimization Strategy for SDN-Enabled UAV-Assisted
Vehicular Computation Offloading,” IEEE Transactions on
Standardizing technologies can be a solution to Intelligent Transportation Systems, vol. 22, no. 6, pp. 3664–
overcome this issue [193]. Besides, A standardized and 3674, 2021, doi: 10.1109/TITS.2020.3024186.
vendor-agnostic SBI can facilitate interoperability [6] B. S. Yelure and S. P. Sonavane, “QoS Evaluation of VANET
Routing Protocol,” Proceedings of the 4th International
between different SDN controllers and network devices
Conference on Communication and Electronics Systems, ICCES
[172]. Developing ML-based assessment solutions can 2019, no. Icces, pp. 813–818, 2019, doi:
aid in proving the feasibility of SDN in VANET 10.1109/ICCES45898.2019.9002115.
environments. These solutions can provide insights and [7] M. Fogli, C. Giannelli, and C. Stefanelli, “Software-Defined
Networking in wireless ad hoc scenarios: Objectives and control
suggestions to address some of the expected challenges architectures,” Journal of Network and Computer Applications,
and issues in SDVN.
VOLUME XX, 2017 9
vol. 203, no. April, p. 103387, 2022, doi: [24] S. Boussoufa-Lahlah, F. Semchedine, and L. Bouallouche-
10.1016/j.jnca.2022.103387. Medjkoune, “Geographic routing protocols for Vehicular Ad hoc
[8] L. Zhao, J. Li, A. Al-Dubai, A. Y. Zomaya, G. Min, and A. NETworks (VANETs): A survey,” Vehicular Communications,
Hawbani, “Routing Schemes in Software-Defined Vehicular vol. 11, pp. 20–31, 2018, doi: 10.1016/j.vehcom.2018.01.006.
Networks: Design, Open Issues and Challenges,” IEEE [25] A. K. Saggu and K. Pandey, “Comparative Analysis of Position-
Intelligent Transportation Systems Magazine, vol. 13, no. 4, pp. Based Routing Protocols for VANETs,” Proceedings of the 8th
217–226, 2021, doi: 10.1109/MITS.2019.2953557. International Advance Computing Conference, IACC 2018, pp.
[9] Y. C. Liu, C. Chen, and S. Chakraborty, “A Software Defined 119–125, 2018, doi: 10.1109/IADCC.2018.8692111.
Network architecture for GeoBroadcast in VANETs,” IEEE [26] I. Wahid, A. A. Ikram, M. Ahmad, S. Ali, and A. Ali, “State of
International Conference on Communications, vol. 2015-Septe, the Art Routing Protocols in VANETs: A Review,” Procedia
pp. 6559–6564, 2015, doi: 10.1109/ICC.2015.7249370. Comput Sci, vol. 130, pp. 689–694, 2018, doi:
[10] M. M. Islam, M. T. R. Khan, M. M. Saad, and D. Kim, 10.1016/j.procs.2018.04.121.
“Software-defined vehicular network (SDVN): A survey on [27] M. Chahal, S. Harit, K. K. Mishra, A. Kumar, and Z. Zheng, “A
architecture and routing,” Journal of Systems Architecture, vol. Survey on software-defined networking in vehicular ad hoc
114, no. September 2020, p. 101961, 2021, doi: networks: Challenges, applications and use cases,” vol. 35, no.
10.1016/j.sysarc.2020.101961. July, pp. 830–840, 2017, doi: 10.1016/j.scs.2017.07.007.
[11] T. Chatterjee, R. Karmakar, G. Kaddoum, S. Chattopadhyay, and [28] A. Awang, K. Husain, N. Kamel, and S. Aissa, “Routing in
S. Chakraborty, “A Survey of VANET/V2X Routing From the Vehicular Ad-hoc Networks: A Survey on Single- and Cross-
Perspective of Non-Learning- and Learning-Based Layer Design Techniques, and Perspectives,” IEEE Access, vol.
Approaches,” IEEE Access, vol. 10, pp. 23022–23050, 2022, 5, pp. 9497–9517, 2017, doi: 10.1109/ACCESS.2017.2692240.
doi: 10.1109/access.2022.3152767. [29] N. Goel, G. Sharma, and I. Dhyani, “A study of position-based
[12] H. Shahwani, S. Attique Shah, M. Ashraf, M. Akram, J. (Paul) VANET routing protocols,” Proceeding - IEEE International
Jeong, and J. Shin, “A comprehensive survey on data Conference on Computing, Communication and Automation,
dissemination in Vehicular Ad Hoc Networks,” Vehicular ICCCA 2016, pp. 655–660, 2017, doi:
Communications, vol. 34, p. 100420, 2022, doi: 10.1109/CCAA.2016.7813803.
10.1016/j.vehcom.2021.100420. [30] R. Hajlaoui, H. Guyennet, and T. Moulahi, “A Survey on
[13] F. Belamri, S. Boulfekhar, and D. Aissani, “A survey on QoS Heuristic-Based Routing Methods in Vehicular Ad-Hoc
routing protocols in Vehicular Ad Hoc Network (VANET),” Network: Technical Challenges and Future Trends,” IEEE Sens
Telecommun Syst, vol. 78, no. 1, pp. 117–153, 2021, doi: J, vol. 16, no. 17, pp. 6782–6792, 2016, doi:
10.1007/s11235-021-00797-8. 10.1109/JSEN.2016.2583382.
[14] T. Kayarga and S. A. Kumar, “A Study on Various Technologies [31] M. Dixit, R. Kumar, and A. K. Sagar, “VANET: Architectures,
to Solve the Routing Problem in Internet of Vehicles (IoV),” research issues, routing protocols, and its applications,”
Wirel Pers Commun, vol. 119, no. 1, pp. 459–487, 2021, doi: Proceeding - IEEE International Conference on Computing,
10.1007/s11277-021-08220-w. Communication and Automation, ICCCA 2016, pp. 555–561,
[15] K. H. Rashmi and R. Patil, “Survey on Cross Layer Approach 2017, doi: 10.1109/CCAA.2016.7813782.
for Robust Communication in VANET,” Wirel Pers Commun, [32] S. Kumar and A. K. Verma, “Position Based Routing Protocols
vol. 119, no. 4, pp. 3413–3434, 2021, doi: 10.1007/s11277-021- in VANET: A Survey,” Wirel Pers Commun, vol. 83, no. 4, pp.
08414-2. 2747–2772, 2015, doi: 10.1007/s11277-015-2567-z.
[16] P. Shah and T. Kasbe, “A review on specification evaluation of [33] J. Cheng, J. Cheng, M. Zhou, F. Liu, S. Gao, and C. Liu,
broadcasting routing protocols in VANET,” Comput Sci Rev, “Routing in internet of vehicles: A review,” IEEE Transactions
vol. 41, p. 100418, 2021, doi: 10.1016/j.cosrev.2021.100418. on Intelligent Transportation Systems, vol. 16, no. 5, pp. 2339–
[17] M. A. Gawas and S. Govekar, State-of-Art and Open Issues of 2352, 2015, doi: 10.1109/TITS.2015.2423667.
Cross-Layer Design and QOS Routing in Internet of Vehicles, [34] S. M. Tornell, C. T. Calafate, J. C. Cano, and P. Manzoni, “DTN
vol. 116, no. 3. Springer US, 2021. doi: 10.1007/s11277-020- protocols for vehicular networks: An application oriented
07790-5. overview,” IEEE Communications Surveys and Tutorials, vol.
[18] O. S. Al-Heety, Z. Zakaria, M. Ismail, M. M. Shakir, S. Alani, 17, no. 2, pp. 868–887, 2015, doi:
and H. Alsariera, “A Comprehensive Survey: Benefits, Services, 10.1109/COMST.2014.2375340.
Recent Works, Challenges, Security, and Use Cases for SDN- [35] B. T. Sharef, R. A. Alsaqour, and M. Ismail, “Vehicular
VANET,” IEEE Access, vol. 8, pp. 91028–91047, 2020, doi: communication ad hoc routing protocols: A survey,” Journal of
10.1109/ACCESS.2020.2992580. Network and Computer Applications, vol. 40, no. 1, pp. 363–
[19] A. Ullah, X. Yao, S. Shaheen, and H. Ning, “Advances in 396, 2014, doi: 10.1016/j.jnca.2013.09.008.
Position Based Routing towards ITS Enabled FoG-Oriented [36] D. Kim, Y. Velasco, W. Wang, R. N. Uma, R. Hussain, and S.
VANET-A Survey,” IEEE Transactions on Intelligent Lee, “A new comprehensive RSU installation strategy for cost-
Transportation Systems, vol. 21, no. 2, pp. 828–840, 2020, doi: efficient VANET deployment,” IEEE Trans Veh Technol, vol.
10.1109/TITS.2019.2893067. 66, no. 5, pp. 4200–4211, 2017, doi:
[20] A. Srivastava, A. Prakash, and R. Tripathi, “Location based 10.1109/TVT.2016.2598253.
routing protocols in VANET: Issues and existing solutions,” [37] Z. Wang, J. Zheng, Y. Wu, and N. Mitton, “A centrality-based
Vehicular Communications, vol. 23, p. 100231, 2020, doi: RSU deployment approach for vehicular ad hoc networks,”
10.1016/j.vehcom.2020.100231. IEEE International Conference on Communications, 2017, doi:
[21] K. Smida, H. Tounsi, M. Frikha, and Y. Q. Song, “Software 10.1109/ICC.2017.7996986.
defined internet of vehicles: A survey from QoS and scalability [38] A. Zekri and W. Jia, “Heterogeneous vehicular communications:
perspectives,” 2019 15th International Wireless A comprehensive study,” Ad Hoc Networks, vol. 75–76, pp. 52–
Communications and Mobile Computing Conference, IWCMC 79, 2018, doi: 10.1016/j.adhoc.2018.03.010.
2019, pp. 1349–1354, 2019, doi: [39] X. Li, R. Zhou, Y. J. Angela Zhang, L. Jiao, and Z. Li, “Smart
10.1109/IWCMC.2019.8766647. vehicular communication via 5G mmWaves,” Computer
[22] S. A. Ahmad and M. Shcherbakov, “A survey on routing Networks, vol. 172, no. January, p. 107173, 2020, doi:
protocols in vehicular ad hoc networks,” 2018 9th International 10.1016/j.comnet.2020.107173.
Conference on Information, Intelligence, Systems and [40] E. Yurtsever, J. Lambert, A. Carballo, and K. Takeda, “A Survey
Applications, IISA 2018, doi: 10.1109/IISA.2018.8633700. of Autonomous Driving: Common Practices and Emerging
[23] O. Senouci, Z. Aliouat, and S. Harous, “A review of routing Technologies,” IEEE Access, vol. 8, pp. 58443–58469, 2020,
protocols in internet of vehicles and their challenges,” Sensor doi: 10.1109/ACCESS.2020.2983149.
Review, vol. 39, no. 1, pp. 58–70, 2019, doi: 10.1108/SR-08- [41] N. Ganeshkumar and S. Kumar, “Study on Routing in VANETs
2017-0168. - A QoS Perspective,” 2021 4th International Conference on
Electrical, Computer and Communication Technologies,
VOLUME XX, 2017 9
ICECCT 2021, 2021, doi: [58] L. Kong, J. He, Y. Ai, S. Chatzinotas, and B. Ottersten, “Channel
10.1109/ICECCT52121.2021.9616910. Modeling and Analysis of Reconfigurable Intelligent Surfaces
[42] T. T. de Almeida, J. G. R. Júnior, M. E. M. Campista, and L. H. Assisted Vehicular Networks,” 2021 IEEE International
M. K. Costa, “Wi-Fi direct performance evaluation for V2P Conference on Communications Workshops, ICC Workshops
communications,” Journal of Sensor and Actuator Networks, 2021 - Proceedings, pp. 4–9, 2021, doi:
vol. 9, no. 2, 2020, doi: 10.3390/JSAN9020028. 10.1109/ICCWorkshops50388.2021.9473681.
[43] S. Temel et al., “Vehicle-to-barrier communication during real- [59] V. O. Nyangaresi, A. J. Rodrigues, and N. K. Taha, Mutual
world vehicle crash tests,” Comput Commun, vol. 127, pp. 172– Authentication Protocol for Secure VANET Data Exchanges,
186, 2018, doi: 10.1016/j.comcom.2018.05.009. vol. 382, no. May. Springer International Publishing, 2021. doi:
[44] B. Sliwa, R. Falkenberg, T. Liebig, N. Piatkowski, and C. 10.1007/978-3-030-78459-1_5.
Wietfeld, “Boosting Vehicle-to-Cloud Communication by [60] A. Khan, M. Ishtiaq, S. Anwar, and M. A. Shah, “A survey on
Machine Learning-Enabled Context Prediction,” IEEE secure routing strategies in VANETs,” ICAC 2019 - 2019 25th
Transactions on Intelligent Transportation Systems, vol. 21, no. IEEE International Conference on Automation and Computing,
8, pp. 3497–3512, 2020, doi: 10.1109/TITS.2019.2930109. no. September, pp. 5–7, 2019, doi:
[45] O. S. Oubbati, N. Chaib, A. Lakas, P. Lorenz, and A. Rachedi, 10.23919/IConAC.2019.8895221.
“UAV-Assisted Supporting Services Connectivity in Urban [61] F. Tang, Y. Kawamoto, N. Kato, and J. Liu, “Future Intelligent
VANETs,” IEEE Trans Veh Technol, vol. 68, no. 4, pp. 3944– and Secure Vehicular Network Toward 6G: Machine-Learning
3951, 2019, doi: 10.1109/TVT.2019.2898477. Approaches,” Proceedings of the IEEE, vol. 108, no. 2, pp. 292–
[46] N. Cheng et al., “Air-Ground Integrated Mobile Edge Networks: 307, 2020, doi: 10.1109/JPROC.2019.2954595.
Architecture, Challenges, and Opportunities,” IEEE [62] F. B. Günay, E. Öztürk, T. Çavdar, Y. S. Hanay, and A. ur R.
Communications Magazine, vol. 56, no. August, pp. 26–32, Khan, Vehicular Ad Hoc Network (VANET) Localization
2018, doi: 10.1109/MCOM.2018.1701092. Techniques: A Survey, vol. 28, no. 4. Springer Netherlands,
[47] X. Wu et al., “Vehicular communications using DSRC: 2021. doi: 10.1007/s11831-020-09487-1.
Challenges, enhancements, and evolution,” IEEE Journal on [63] R. K. Jaiswal and C. D. Jaidhar, “A Performance Evaluation of
Selected Areas in Communications, vol. 31, no. 9, pp. 399–408, Location Prediction Position-Based Routing Using Real GPS
2013, doi: 10.1109/JSAC.2013.SUP.0513036. Traces for VANET,” Wirel Pers Commun, vol. 102, no. 1, pp.
[48] K. Z. Ghafoor, M. Guizani, L. Kong, H. S. Maghdid, and K. F. 275–292, 2018, doi: 10.1007/s11277-018-5839-6.
Jasim, “Enabling Efficient Coexistence of DSRC and C-V2X in [64] B. P. Maratha, T. R. Sheltami, and K. Salah, “Performance Study
Vehicular Networks,” IEEE Wirel Commun, vol. 27, no. 2, pp. of MANET Routing Protocols in VANET,” Arab J Sci Eng, vol.
134–140, 2020, doi: 10.1109/MWC.001.1900219. 42, no. 8, pp. 3115–3126, 2017, doi: 10.1007/s13369-016-2377-
[49] K. Zrar Ghafoor et al., “Millimeter-Wave Communication for y.
Internet of Vehicles: Status, Challenges, and Perspectives,” [65] K. C. Purohit, S. C. Dimri, and S. Jasola, “Performance
IEEE Internet Things J, vol. 7, no. 9, pp. 8525–8546, 2020, doi: evaluation of various MANET routing protocols for adaptability
10.1109/JIOT.2020.2992449. in VANET environment,” International Journal of System
[50] R. Li and P. Zhu, “Spectrum Allocation Strategies Based on QoS Assurance Engineering and Management, vol. 8, pp. 690–702,
in Cognitive Vehicle Networks,” IEEE Access, vol. 8, pp. 2017, doi: 10.1007/s13198-016-0507-2.
99922–99933, 2020, doi: 10.1109/ACCESS.2020.2997936. [66] P. K. Shrivastava and L. K. Vishwamitra, “Comparative analysis
[51] S. Goli-Bidgoli and N. Movahhedinia, “Towards Ensuring of proactive and reactive routing protocols in VANET
Reliability of Vehicular Ad Hoc Networks Using a Relay environment,” Measurement: Sensors, vol. 16, no. May, p.
Selection Techniques and D2D Communications in 5G 100051, 2021, doi: 10.1016/j.measen.2021.100051.
Networks,” Wirel Pers Commun, vol. 114, no. 3, pp. 2755–2767, [67] F. Li and Y. Wang, “Routing in vehicular ad hoc networks: A
2020, doi: 10.1007/s11277-020-07501-0. survey,” IEEE Vehicular Technology Magazine, vol. 2, no. 2,
[52] P. Sharma and S. Jain, “Review of VANET challenges and pp. 12–22, 2007, doi: 10.1109/MVT.2007.912927.
protocol for architecture design and intelligent traffic system,” [68] S. M. Hanshi, T. C. Wan, M. M. Kadhum, and A. A. Bin-Salem,
2nd International Conference on Data, Engineering and “Review of geographic forwarding strategies for inter-vehicular
Applications, IDEA 2020, pp. 13–16, 2020, doi: communications from mobility and environment perspectives,”
10.1109/IDEA49133.2020.9170685. Vehicular Communications, vol. 14, pp. 64–79, 2018, doi:
[53] K. Smida, H. Tounsi, M. Frikha, and Y. Q. Song, “Software 10.1016/j.vehcom.2018.09.005.
defined internet of vehicles: A survey from QoS and scalability [69] Sohail, Muhammad, Zohaib Latif, Shahzeb Javed, Sujit Biswas,
perspectives,” 2019 15th International Wireless Sahar Ajmal, Umer Iqbal, and Mohsin Raza. "Routing protocols
Communications and Mobile Computing Conference, IWCMC in Vehicular Adhoc Networks (VANETs): A comprehensive
2019, pp. 1349–1354, 2019, doi: survey." Internet of Things (2023): 100837.
10.1109/IWCMC.2019.8766647. [70] M. Altayeb and I. Mahgoub, “A Survey of Vehicular Ad hoc
[54] S. A. Rashid, L. Audah, M. M. Hamdi, M. S. Abood, and S. Networks Routing Techniques,” Int J Innov Appl Stud, vol. 3,
Alani, “Reliable and efficient data dissemination scheme in no. 3, pp. 829–846, 2013.
VANET: A review,” International Journal of Electrical and [71] Salman, O., Morcel, R., Al Zoubi, O., Elhajj, I., Kayssi, A. and
Computer Engineering, vol. 10, no. 6, pp. 6423–6434, 2020, doi: Chehab, A., “Analysis of topology-based routing protocols for
10.11591/IJECE.V10I6.PP6423-6434. VANETs in different environments” In 2016 IEEE International
[55] B. Iancu, I. Illyes, A. Peculea, and V. Dadarlat, “Pollution Probes Multidisciplinary Conference on Engineering Technology
Application: The impact of using PVDM messages in VANET (IMCET) (pp. 27-31), 2016.
infrastructures for environmental monitoring,” Proceedings - [72] Z. S. Houssaini, I. Zaimi, M. Oumsis, and S. E. A. Ouatik,
2019 IEEE 15th International Conference on Intelligent “Comparative study of routing protocols performance for
Computer Communication and Processing, ICCP 2019, pp. 443– vehicular Ad-hoc networks,” International Journal of Applied
449, 2019, doi: 10.1109/ICCP48234.2019.8959532. Engineering Research, vol. 12, no. 13, pp. 3867–3878, 2017.
[56] H. Brahmia and C. Tolba, “Nakagami Fading Impact on the [73] Y. Sarada Devi and M. Roopa, “Performance analysis of routing
Performances of VANET Routing Protocols in a Realistic protocols IN vehicular adhoc networks,” Mater Today Proc, no.
Urban,” International Journal of Advanced Networking and xxxx, 2021, doi: 10.1016/j.matpr.2021.02.556.
Applications, vol. 11, no. 04, pp. 4330–4335, 2020, doi: [74] M. A. Hossain, I. Elshafiey, and A. Al-Sanie, “High precision
10.35444/ijana.2020.11042. vehicle positioning: Towards cooperative driving based on
[57] F. Li, W. Chen, and Y. Shui, “Study on Connectivity Probability VANET,” Journal of Internet Technology, vol. 19, no. 1, pp.
of Vanets under Adverse Weather Conditions at 5.9 Ghz,” IEEE 289–295, 2018, doi: 10.3966/160792642018011901028.
Access, vol. 8, pp. 547–555, 2020, doi: [75] S. Dhankhar and S. Agrawal, “VANETs: A Survey on Routing
10.1109/ACCESS.2019.2962089. Protocols and Issues,” Int J Innov Res Sci Eng Technol, vol. 3,
no. 6, pp. 13427–13435, 2014.
VOLUME XX, 2017 9
[76] H. Kang, S. H. Ahmed, D. Kim, and Y. S. Chung, “Routing [94] N. Cardona, E. Coronado, S. Latre, R. Riggio, and J. M.
protocols for vehicular delay tolerant networks: A survey,” Int J Marquez-Barja, “Software-Defined Vehicular Networking:
Distrib Sens Netw, vol. 2015, 2015, doi: 10.1155/2015/325027. Opportunities and Challenges,” IEEE Access, pp. 219971–
[77] O. Alzamzami and I. Mahgoub, “Geographic routing 219995, 2020, doi: 10.1109/ACCESS.2020.3042717.
enhancement for urban VANETs using link dynamic behavior: [95] H. Trivedi, S. Tanwar, and P. Thakkar, Software Defined
A cross layer approach,” Vehicular Communications, vol. 31, p. Network-Based Vehicular Adhoc Networks for Intelligent
100354, 2021, doi: 10.1016/j.vehcom.2021.100354. Transportation System: Recent Advances and Future
[78] A. Sarlak and Y. Darmani, “An Approach to Improve the Quality Challenges, vol. 958. Springer Singapore, 2019. doi:
of Service in DTN and Non-DTN based VANET,” Journal of 10.1007/978-981-13-3804-5_24.
Information Systems and Telecommunication, vol. 8, no. 32, pp. [96] J. Joshi, K. Renuka, and P. Medikonda, “Secured and energy
263–271, 2020, doi: 10.29252/jist.8.32.240. efficient data transmission in SDN-VANETs,” 2018 22nd
[79] Y. Sasaki, W. C. Lee, T. Hara, and S. Nishio, “On alleviating International Computer Science and Engineering Conference,
beacon overhead in routing protocols for urban vanets,” ICSEC 2018, pp. 16–19, 2018, doi:
Proceedings - IEEE International Conference on Mobile Data 10.1109/ICSEC.2018.8712714.
Management, vol. 1, pp. 66–76, 2013, doi: [97] R. Sultana, J. Grover, and M. Tripathi, “Security of SDN-based
10.1109/MDM.2013.17. vehicular ad hoc networks: State-of-the-art and challenges,”
[80] M. Hu, Z. Zhong, M. Ni, and A. Baiocchi, “Design and analysis Vehicular Communications, vol. 27, p. 100284, 2021, doi:
of a beacon-less routing protocol for large volume content 10.1016/j.vehcom.2020.100284.
dissemination in vehicular ad hoc networks,” Sensors [98] R. Amin, I. Pali, and V. Sureshkumar, “Software-Defined
(Switzerland), vol. 16, no. 11, 2016, doi: 10.3390/s16111834. Network enabled Vehicle to Vehicle secured data transmission
[81] A. Srivastava, A. Prakash, and R. Tripathi, “Location based protocol in VANETs,” Journal of Information Security and
routing protocols in VANET: Issues and existing solutions,” Applications, vol. 58, no. February, 2021, doi:
Vehicular Communications, vol. 23, p. 100231, 2020, doi: 10.1016/j.jisa.2020.102729.
10.1016/j.vehcom.2020.100231. [99] L. Nkenyereye, L. Nkenyereye, S. M. R. Islam, C. A. Kerrache,
[82] Das, I. and Das, S., “Geocast routing protocols for ad-hoc M. Abdullah-Al-Wadud, and A. Alamri, “Software Defined
networks: comparative analysis and open issues”, In Cloud- Network-Based Multi-Access Edge Framework for Vehicular
Based Big Data Analytics in Vehicular Ad-Hoc Networks (pp. Networks,” IEEE Access, vol. 8, pp. 4220–4234, 2020, doi:
23-45). 2021, IGI Global. 10.1109/ACCESS.2019.2962903.
[83] S. Ahn, “Geographic information-based data delivery in [100] A. Kazmi, M. A. Khan, and M. U. Akram, “DeVANET:
vehicular networks: A survey,” ICT Express, vol. 3, no. 1, pp. Decentralized software-defined VANET architecture,”
22–26, 2017, doi: 10.1016/j.icte.2017.03.003. Proceedings - 2016 IEEE International Conference on Cloud
[84] W. Farooq, M. A. Khan, S. Rehman, N. A. Saqib, and M. Abbas, Engineering Workshops, IC2EW 2016, pp. 42–47, 2016, doi:
“AAGV: A Cluster Based Multicast Routing Protocol for 10.1109/IC2EW.2016.12.
Autonomous Aerial and Ground Vehicles Communication in [101] Y. Gao, Z. Zhang, D. Zhao, Y. Zhang, and T. Luo, “A
VANET,” Proceedings - 2017 International Conference on Hierarchical Routing Scheme With Load Balancing in Software
Frontiers of Information Technology, FIT 2017, vol. 2017- Defined Vehicular Ad Hoc Networks,” IEEE Access, vol. 6, pp.
Janua, pp. 315–320, 2017, doi: 10.1109/FIT.2017.00063. 73774–73785, 2018, doi: 10.1109/ACCESS.2018.2884708.
[85] S. Latif, S. Mahfooz, B. Jan, N. Ahmad, Y. Cao, and M. Asif, [102] S. Rehman and N. Kapoor, “A review on delay efficient
“A comparative study of scenario-driven multi-hop broadcast architecture for software defined vehicular networks (SDVN),”
protocols for VANETs,” Vehicular Communications, vol. 12, Proceedings of the 2019 6th International Conference on
pp. 88–109, 2018, doi: 10.1016/j.vehcom.2018.01.009. Computing for Sustainable Global Development, INDIACom
[86] A. Ahamed and H. Vakilzadian, “Issues and Challenges in 2019, pp. 1094–1100, 2019.
VANET Routing Protocols,” IEEE International Conference on [103] M. Karakus and A. Durresi, “Quality of Service (QoS) in
Electro Information Technology, vol. 2018-May, pp. 723–728, Software Defined Networking (SDN): A survey,” Journal of
2018, doi: 10.1109/EIT.2018.8500180. Network and Computer Applications, vol. 80, pp. 200–218,
[87] N. M. Al-Kharasani, Z. A. Zulkarnain, S. Subramaniam, and Z. 2017, doi: 10.1016/j.jnca.2016.12.019.
M. Hanapi, “An efficient framework model for optimizing [104] S. Din, A. Paul, and A. Rehman, “5G-enabled Hierarchical
routing performance in vanets,” Sensors (Switzerland), vol. 18, architecture for software-defined intelligent transportation
no. 2, 2018, doi: 10.3390/s18020597. system,” Computer Networks, vol. 150, pp. 81–89, 2019, doi:
[88] A. Lakas, M. E. A. Fekair, A. Korichi, and N. Lagraa, “A 10.1016/j.comnet.2018.11.035.
multiconstrained QoS-compliant routing scheme for highway- [105] M. Abolhasan, J. Lipman, W. Ni, and B. Hagelstein, “Software-
based vehicular networks,” Wirel Commun Mob Comput, vol. defined wireless networking: Centralized, distributed, or
2019, 2019, doi: 10.1155/2019/4521859. hybrid?,” IEEE Netw, vol. 29, no. 4, pp. 32–38, 2015, doi:
[89] X. Wang, C. Liu, Y. Wang, and C. Huang, “Application of ant 10.1109/MNET.2015.7166188.
colony optimized routing algorithm based on evolving graph [106] X. Huang, R. Yu, J. Kang, Y. He, and Y. Zhang, “Exploring
model in VANETs,” International Symposium on Wireless Mobile Edge Computing for 5G-Enabled Software Defined
Personal Multimedia Communications, WPMC, vol. 2015- Vehicular Networks,” IEEE Wirel Commun, vol. 2, no.
Janua, no. c, pp. 265–270, 2015, doi: December, pp. 55–63, 2017.
10.1109/WPMC.2014.7014828. [107] M. Zhu, Z. Cai, J. Cao, and M. Xu, “Efficient multiple-copy
[90] Mahmoud Hashem Eiza and Qiang Ni, “An Evolving Graph- routing in software-defined vehicular networks,” in IET
Based Reliable Routing Scheme for VANETs,” IEEE Trans Veh Conference Publications, 2015, vol. 2015, no. CP664. doi:
Technol, vol. 62, no. 4, pp. 1493–1504, 2013. 10.1049/cp.2015.0222.
[91] L. Urquiza-Aguiar, C. Tripp-Barba, and M. A. Igartua, “A [108] M. Zhu, J. Cao, D. Pang, Z. He, and M. Xu, “SDN-based routing
geographical heuristic routing protocol for VANETs,” Sensors for efficient message propagation in VANET,” Lecture Notes in
(Switzerland), vol. 16, no. 10, pp. 1–28, 2016, doi: Computer Science (including subseries Lecture Notes in
10.3390/s16101567. Artificial Intelligence and Lecture Notes in Bioinformatics), vol.
[92] P. Sathya Narayanan and C. S. Joice, “Vehicle-to-Vehicle (V2V) 9204, pp. 788–797, 2015, doi: 10.1007/978-3-319-21837-3_77.
Communication using Routing Protocols: A Review,” 6th IEEE [109] K. L. K. Sudheer, M. Ma, and P. H. J. Chong, “Link dynamics
International Conference on Smart Structures and Systems, based packet routing framework for software defined vehicular
ICSSS 2019, 2019, doi: 10.1109/ICSSS.2019.8882828. networks,” 2017 IEEE Global Communications Conference,
[93] D. S. Hotkar and S. R. Biradar, A review on existing qos routing GLOBECOM 2017 - Proceedings, vol. 2018-Janua, pp. 1–6,
protocols in vanet based on link efficiency and link stability, vol. 2017, doi: 10.1109/GLOCOM.2017.8254597.
31. Springer Singapore, 2019. doi: 10.1007/978-981-10-8911- [110] B. Dong, W. Wu, Z. Yang, and J. Li, “Software Defined
4_10. Networking Based On-Demand Routing Protocol in Vehicle Ad
VOLUME XX, 2017 9
Hoc Networks,” Proceedings - 12th International Conference on Journal of Communication Systems, vol. 33, no. 15, pp. 1–16,
Mobile Ad-Hoc and Sensor Networks, MSN 2016, pp. 207–213, 2020, doi: 10.1002/dac.4521.
2017, doi: 10.1109/MSN.2016.041. [127] D. Kong and G. Zhang, “Ant Colony Algorithm Based Routing
[111] K. A. Darabkh, B. Z. Alkhader, A. F. Khalifeh, F. Jubair, and M. Protocol in Software Defined Vehicular Networks,” ACM
Abdel-Majeed, “ICDRP-F-SDVN: An innovative cluster-based International Conference Proceeding Series, pp. 200–204, 2020,
dual-phase routing protocol using fog computing and software- doi: 10.1145/3390557.3394324.
defined vehicular network,” Vehicular Communications, vol. 34, [128] W. Qi, B. Landfeldt, Q. Song, L. Guo, and A. Jamalipour,
p. 100453, 2022, doi: 10.1016/j.vehcom.2021.100453. “Traffic Differentiated Clustering Routing in DSRC and C-V2X
[112] S. B. Prathiba, G. Raja, A. K. Bashir, A. A. Alzubi, and B. Gupta, Hybrid Vehicular Networks,” IEEE Trans Veh Technol, vol. 69,
“SDN-Assisted Safety Message Dissemination Framework for no. 7, pp. 7723–7734, 2020, doi: 10.1109/TVT.2020.2990174.
Vehicular Critical Energy Infrastructure,” IEEE Trans Industr [129] A. Nahar and D. Das, “SeScR: SDN-Enabled Spectral
Inform, vol. 18, no. 5, pp. 3510–3518, 2022, doi: Clustering-Based Optimized Routing Using Deep Learning in
10.1109/TII.2021.3113130. VANET Environment,” 2020 IEEE 19th International
[113] K. L. K. Sudheera, M. Ma, and P. H. J. Chong, “Real-time Symposium on Network Computing and Applications, NCA
cooperative data routing and scheduling in software defined 2020, 2020, doi: 10.1109/NCA51143.2020.9306690.
vehicular networks,” Comput Commun, vol. 181, no. August [130] X. Ji, W. Xu, C. Zhang, and B. Liu, “A Three-level Routing
2021, pp. 203–214, 2022, doi: 10.1016/j.comcom.2021.10.003. Hierarchy in improved SDN-MEC-VANET Architecture,” IEEE
[114] A. J. Kadhim, S. A. H. Seno, J. I. Naser, and J. Hajipour, Wireless Communications and Networking Conference, WCNC,
“DMPFS: Delay-efficient multicasting based on parked vol. 2020-May, 2020, doi:
vehicles, fog computing and SDN in vehicular networks,” 10.1109/WCNC45663.2020.9120486.
Vehicular Communications, vol. 36, p. 100488, 2022, doi: [131] K. S. Kalupahana Liyanage, M. Ma, and P. H. J. Chong,
10.1016/j.vehcom.2022.100488. “Connectivity aware tribrid routing framework for a generalized
[115] L. Zhao et al., “A Novel Prediction-Based Temporal Graph software defined vehicular network,” Computer Networks, vol.
Routing Algorithm for Software-Defined Vehicular Networks,” 152, pp. 167–177, 2019, doi: 10.1016/j.comnet.2019.01.040.
IEEE Transactions on Intelligent Transportation Systems, pp. 1– [132] M. T. Abbas, A. Muhammad, and W.-C. Song, “SD-IoV: SDN
16, 2021, doi: 10.1109/TITS.2021.3123276. enabled routing for internet of vehicles in road-aware approach,”
[116] L. Zhao, T. Zheng, M. Lin, A. Hawbani, J. Shang, and C. Fan, J Ambient Intell Humaniz Comput, May 2019, doi:
“SPIDER: A Social Computing Inspired Predictive Routing 10.1007/s12652-019-01319-w.
Scheme for Softwarized Vehicular Networks,” IEEE [133] K. Z. Ghafoor, L. Kong, D. B. Rawat, E. Hosseini, and A. S.
Transactions on Intelligent Transportation Systems, pp. 1–12, Sadiq, “Quality of service aware routing protocol in software-
2021, doi: 10.1109/TITS.2021.3122438. defined internet of vehicles,” IEEE Internet Things J, vol. 6, no.
[117] L. Zhao, Z. Bi, M. Lin, A. Hawbani, J. Shi, and Y. Guan, “An 2, pp. 2817–2828, 2019, doi: 10.1109/JIOT.2018.2875482.
intelligent fuzzy-based routing scheme for software-defined [134] Y. Yang, R. Zhao, and X. Wei, “Research on Data Distribution
vehicular networks,” Computer Networks, vol. 187, no. January, for VANET Based on Deep Reinforcement Learning,”
p. 107837, 2021, doi: 10.1016/j.comnet.2021.107837. Proceedings - 2019 International Conference on Artificial
[118] K. Renuka, D. S. Roy, and K. H. K. Reddy, “An SDN Intelligence and Advanced Manufacturing, AIAM 2019, pp.
empowered location aware routing for energy efficient next 484–487, 2019, doi: 10.1109/AIAM48774.2019.00102.
generation vehicular networks,” IET Intelligent Transport [135] Y. Gao, T. Luo, Y. Guo, and X. He, “A connectivity probability
Systems, no. May 2020, pp. 308–319, 2021, doi: based cross-layer routing handoff mechanism in software
10.1049/itr2.12026. defined VANETs,” IEEE Vehicular Technology Conference,
[119] L. Zhao et al., “Novel Online Sequential Learning-Based vol. 2019-April, 2019, doi: 10.1109/VTCSpring.2019.8746653.
Adaptive Routing for Edge Software-Defined Vehicular [136] A. Di Maio, M. R. Palattella, and T. Engel, “Multi-flow
Networks,” IEEE Trans Wirel Commun, vol. 20, no. 5, pp. congestion-aware routing in software-defined vehicular
2991–3004, 2021, doi: 10.1109/TWC.2020.3046275. networks,” IEEE Vehicular Technology Conference, vol. 2019-
[120] K. Mershad, “SURFER: A Secure SDN-based Routing Protocol Septe, pp. 1–6, 2019, doi: 10.1109/VTCFall.2019.8891465.
for Internet of Vehicles,” IEEE Internet Things J, vol. 8, no. 9, [137] M. Chahal and S. Harit, “Network selection and data
pp. 7407–7422, 2021, doi: 10.1109/JIOT.2020.3038465. dissemination in heterogeneous software-defined vehicular
[121] K. Mershad, H. Artail, and M. Gerla, “ROAMER: Roadside network,” Computer Networks, vol. 161, pp. 32–44, 2019, doi:
Units as message routers in VANETs,” Ad Hoc Networks, vol. 10.1016/j.comnet.2019.06.008.
10, no. 3, pp. 479–496, 2012, doi: 10.1016/j.adhoc.2011.09.001. [138] Y. Tang, N. Cheng, W. Wu, M. Wang, Y. Dai, and X. Shen,
[122] G. Luo et al., “Software-Defined Cooperative Data Sharing in “Delay-Minimization Routing for Heterogeneous VANETs with
Edge Computing Assisted 5G-VANET,” IEEE Trans Mob Machine Learning Based Mobility Prediction,” IEEE Trans Veh
Comput, vol. 20, no. 3, pp. 1212–1229, 2021, doi: Technol, vol. 68, no. 4, pp. 3967–3979, 2019, doi:
10.1109/TMC.2019.2953163. 10.1109/TVT.2019.2899627.
[123] N. Noorani and S. A. H. Seno, “SDN- and fog computing-based [139] Z. You, G. Cheng, Y. Wang, P. Chen, and S. Chen, “Cross-layer
switchable routing using path stability estimation for vehicular and SDN Based routing scheme for P2P communication in
ad hoc networks,” Peer Peer Netw Appl, vol. 13, no. 3, pp. 948– vehicular Ad-hoc networks,” Applied Sciences (Switzerland),
964, 2020, doi: 10.1007/s12083-019-00859-4. vol. 9, no. 22, 2019, doi: 10.3390/app9224734.
[124] M. A. Jibran, M. T. Abbas, A. Rafiq, and W. C. Song, “Position [140] C. Wang, X. Ma, W. Jiang, L. Zhao, N. Lin, and J. Shi, “IMCR:
prediction for routing in software defined internet of vehicles,” Influence maximization-based cluster routing algorithm for
Journal of Communications, vol. 15, no. 2, pp. 157–163, 2020, SDVN,” Proceedings - 21st IEEE International Conference on
doi: 10.12720/jcm.15.2.157-163. High Performance Computing and Communications, 17th IEEE
[125] L. Zhao, T. Zheng, M. Lin, A. Hawbani, and J. Shang, “A novel International Conference on Smart City and 5th IEEE
influence maximization-based dynamic forwarding node International Conference on Data Science and Systems,
selection scheme in SDVNs,” Proceedings - 2020 IEEE HPCC/SmartCity/DSS 2019, vol. 1, pp. 2580–2586, 2019, doi:
International Symposium on Parallel and Distributed Processing 10.1109/HPCC/SmartCity/DSS.2019.00361.
with Applications, 2020 IEEE International Conference on Big [141] P. Dai, K. Liu, X. Wu, Z. Yu, H. Xing, and V. C. S. Lee,
Data and Cloud Computing, 2020 IEEE International “Cooperative Temporal Data Dissemination in SDN-Based
Symposium on Social Computing and Networking and 2020 Heterogeneous Vehicular Networks,” IEEE Internet Things J,
IEE, pp. 674–682, 2020, doi: 10.1109/ISPA-BDCloud- vol. 6, no. 1, pp. 72–83, 2019, doi: 10.1109/JIOT.2018.2872432.
SocialCom-SustainCom51426.2020.00110. [142] K. L. K. Sudheera, M. Ma, and P. H. J. Chong, “Link Stability
[126] L. Alouache, N. Nguyen, M. Aliouat, and R. Chelouah, “HSDN- Based Optimized Routing Framework for Software Defined
GRA: A hybrid software-defined networking-based geographic Vehicular Networks,” IEEE Trans Veh Technol, vol. 68, no. 3,
routing protocol with multi-agent approach,” International pp. 2934–2945, 2019, doi: 10.1109/TVT.2019.2895274.
VOLUME XX, 2017 9
[143] X. Yan, P. Dong, X. Du, T. Zheng, J. Sun, and M. Guizani, Networks,” IEEE Sens J, vol. 16, no. 20, pp. 7342–7354, 2016,
“Improving flow delivery with link available time prediction in doi: 10.1109/JSEN.2016.2562699.
software-defined high-speed vehicular networks,” Computer [159] Z. He, D. Zhang, S. Zhu, J. Cao, and X. Liu, “SDN enabled high
Networks, vol. 145, pp. 165–174, 2018, doi: performance multicast in vehicular networks,” IEEE Vehicular
10.1016/j.comnet.2018.08.019. Technology Conference, vol. 0, pp. 6–10, 2016, doi:
[144] C. Wang, L. Zhang, Z. Li, and C. Jiang, “SDCoR: Software 10.1109/VTCFall.2016.7881215.
Defined Cognitive Routing for Internet of Vehicles,” IEEE [160] K. Liu, J. K. Y. Ng, V. C. S. Lee, S. H. Son, and I. Stojmenovic,
Internet Things J, vol. 5, no. 5, pp. 3513–3520, 2018, doi: “Cooperative Data Scheduling in Hybrid Vehicular Ad Hoc
10.1109/JIOT.2018.2812210. Networks: VANET as a Software Defined Network,”
[145] R. S. De Sousa, F. S. Da Costa, A. C. B. Soares, L. F. M. Vieira, IEEE/ACM Transactions on Networking, vol. 24, no. 3, pp.
and A. A. F. Loureiro, “Geo-SDVN: A Geocast Protocol for 1759–1773, 2016, doi: 10.1109/TNET.2015.2432804.
Software Defined Vehicular Networks,” IEEE International [161] Celes, Clayson, Fabricio A. Silva, Azzedine Boukerche, Rossana
Conference on Communications, vol. 2018-May, 2018, doi: Maria de Castro Andrade, and Antonio AF Loureiro. "Improving
10.1109/ICC.2018.8422755. vanet simulation with calibrated vehicular mobility traces."
[146] W. Qi, Q. Song, X. Wang, L. Guo, and Z. Ning, “SDN-Enabled IEEE Transactions on Mobile Computing 16, no. 12 (2017):
Social-Aware Clustering in 5G-VANET Systems,” IEEE 3376-3389.
Access, vol. 6, pp. 28213–28224, 2018, doi: [162] F. J. Martinez, C. K. Toh, J. C. Cano, C. T. Calafate, and P.
10.1109/ACCESS.2018.2837870. Manzoni, “A survey and comparative study of simulators for
[147] C. M. Huang, M. S. Chiang, D. T. Dao, W. L. Su, S. Xu, and H. vehicular ad hoc networks (VANETs),” Wirel Commun Mob
Zhou, “V2V Data Offloading for Cellular Network Based on the Comput, vol. 11, no. 7, pp. 813–828, Jul. 2011, doi:
Software Defined Network (SDN) Inside Mobile Edge 10.1002/wcm.859.
Computing (MEC) Architecture,” IEEE Access, vol. 6, pp. [163] J. S. Weber, M. Neves, and T. Ferreto, “VANET simulators: an
17741–17755, 2018, doi: 10.1109/ACCESS.2018.2820679. updated review,” Journal of the Brazilian Computer Society, vol.
[148] H. Ghafoor and I. Koo, “CR-SDVN: A Cognitive Routing 27, no. 1, Dec. 2021, doi: 10.1186/s13173-021-00113-x.
Protocol for Software-Defined Vehicular Networks,” IEEE Sens [164] G. Seçinti, B. Canberk, T. Q. Duong, and L. Shu, “Software
J, vol. 18, no. 4, pp. 1761–1772, 2018, doi: defined architecture for VANET: A testbed implementation with
10.1109/JSEN.2017.2788014. Wireless Access Management,” IEEE Communications
[149] S. Yang, J. Liu, and X. Yan, “Flexible Routing-Proactive Magazine, vol. 55, no. 7, pp. 135–141, Jul. 2017, doi:
Updating Mechanism for Software Defined Vehicle Networks,” 10.1109/MCOM.2017.1601186.
2018 10th International Conference on Wireless [165] McKeown, Nick, Tom Anderson, Hari Balakrishnan, Guru
Communications and Signal Processing, WCSP 2018, 2018, doi: Parulkar, Larry Peterson, Jennifer Rexford, Scott Shenker, and
10.1109/WCSP.2018.8555633. Jonathan Turner. "OpenFlow: enabling innovation in campus
[150] M. S. Rayeni and A. Hafid, “Routing in heterogeneous vehicular networks." ACM SIGCOMM computer communication review
networks using an adapted software defined networking 38, no. 2 (2008): 69-74.
approach,” 2018 5th International Conference on Software [166] M. Amoozadeh, B. Ching, C. N. Chuah, D. Ghosal, and H.
Defined Systems, SDS 2018, pp. 25–31, 2018, doi: Michael Zhang, “VENTOS: Vehicular network open simulator
10.1109/SDS.2018.8370418. with hardware-in-the-loop support,” in Procedia Computer
[151] D. K. Nooji Venkatramana, S. B. Srikantaiah, and J. Moodabidri, Science, 2019, vol. 151, pp. 61–68. doi:
“SCGRP: SDN-enabled connectivity-aware geographical 10.1016/j.procs.2019.04.012.
routing protocol of VANETs for urban environment,” IET [167] A. Tomandl, D. Herrmann, K. P. Fuchs, H. Federrath, and F.
Networks, vol. 6, no. 5, pp. 102–111, 2017, doi: 10.1049/iet- Scheuer, “VANETsim: An open source simulator for security
net.2016.0117. and privacy concepts in VANETs,” in Proceedings of the 2014
[152] A. B. Dajun Zhang, F. Richard Yu, Zhexiong Wei, “Trust-based International Conference on High Performance Computing and
Secure Routing in Software-defined Vehicular Ad Hoc Simulation, HPCS 2014, Sep. 2014, pp. 543–550. doi:
Networks,” arXiv preprint arXiv:1611.04012, pp. 189–190, 10.1109/HPCSim.2014.6903733.
2017, doi: 10.1145/1235. [168] Lim K.G., Lee C.H., Chin R.K.Y., Yeo K.B., and Teo K.T.K.,
[153] X. Duan, Y. Liu, and X. Wang, “SDN enabled 5G-VANET: “SUMO Enhancement for Vehicular Ad Hoc Network
Adaptive vehicle clustering and beamformed transmission for (VANET) Simulation,” in 2017 IEEE 2nd International
aggregated traffic,” IEEE Communications Magazine, vol. 55, Conference on Automatic Control and Intelligent Systems
no. 7, pp. 120–127, 2017, doi: 10.1109/MCOM.2017.1601160. (I2CACIS 2017): Sabah, Malaysia, 21 Oct., 2017, pp. 87–91.
[154] X. Ji, H. Yu, G. Fan, and W. Fu, “SDGR: An SDN-Based [169] M. Fellendorf and P. Vortisch, “Microscopic traffic flow
Geographic Routing Protocol for VANET,” Proceedings - 2016 simulator VISSIM,” in International Series in Operations
IEEE International Conference on Internet of Things; IEEE Research and Management Science, vol. 145, Springer New
Green Computing and Communications; IEEE Cyber, Physical, York LLC, 2010, pp. 63–93. doi: 10.1007/978-1-4419-6142-
and Social Computing; IEEE Smart Data, iThings-GreenCom- 6_2.
CPSCom-Smart Data 2016, pp. 276–281, 2017, doi: [170] C. L. Azevedo et al., “Simmobility short-term: An integrated
10.1109/iThings-GreenCom-CPSCom-SmartData.2016.70. microscopic mobility simulator,” Transp Res Rec, vol. 2622, no.
[155] A. Alioua, S. M. Senouci, and S. Moussaoui, “dSDiVN: A 1, pp. 13–23, 2017, doi: 10.3141/2622-02.
distributed software-defined networking architecture for [171] G. D. B. Cameron and G. I. D. Duncan, “PARAMICS-Parallel
infrastructure-less vehicular networks,” Communications in Microscopic Simulation of Road Traffic,” Kluwer Academic
Computer and Information Science, vol. 717, pp. 56–57, 2017, Publishers, 1996.
doi: 10.1007/978-3-319-60447-3_5. [172] J. Bhatia, Y. Modi, S. Tanwar, and M. Bhavsar, “Software
[156] S. Correia, A. Boukerche, and R. I. Meneguette, “An architecture defined vehicular networks: A comprehensive review,”
for hierarchical software-defined vehicular networks,” IEEE International Journal of Communication Systems, vol. 32, no.
Communications Magazine, vol. 55, no. 7, pp. 80–86, 2017, doi: 12, Aug. 2019, doi: 10.1002/dac.4005.
10.1109/MCOM.2017.1601105. [173] Z. He, J. Cao, and X. Liu, “SDVN: Enabling Rapid Network
[157] A. Abuashour and M. Kadoch, “An intersection dynamic Innovation for Heterogeneous Vehicular Communication,”
VANET routing protocol for a grid scenario,” Proceedings - IEEE network 30, no. 4, 2016.
2017 IEEE 5th International Conference on Future Internet of [174] H. Tao et al., “SDN-assisted technique for traffic control and
Things and Cloud, FiCloud 2017, vol. 2017-Janua, pp. 25–31, information execution in vehicular adhoc networks,” Computers
2017, doi: 10.1109/FiCloud.2017.19. and Electrical Engineering, vol. 102, p. 108108, Sep. 2022, doi:
[158] Z. He, D. Zhang, and J. Liang, “Cost-Efficient Sensory Data 10.1016/J.COMPELECENG.2022.108108.
Transmission in Heterogeneous Software-Defined Vehicular
VOLUME XX, 2017 9
[175] S. Sharma, “Towards Artificial Intelligence Assisted Software [193] S. Schaller and D. Hood, “Software defined networking
Defined Networking for Internet of Vehicles,” 2021, pp. 191– architecture standardization,” Comput Stand Interfaces, vol. 54,
222. doi: 10.1007/978-3-030-76493-7_6. pp. 197–202, 2017.
[176] M. Arif et al., “SDN-based VANETs, security attacks, [194] F. B. Günay, E. Öztürk, T. Çavdar, Y. S. Hanay, and A. ur R.
applications, and challenges,” Applied Sciences (Switzerland), Khan, “Vehicular Ad Hoc Network (VANET) Localization
vol. 10, no. 9. MDPI AG, May 01, 2020. doi: Techniques: A Survey,” Archives of Computational Methods in
10.3390/app10093217. Engineering, vol. 28, no. 4, pp. 3001–3033, Jun. 2021, doi:
[177] T. Zhang et al., “How to Mitigate DDoS Intelligently in SD-IoV: 10.1007/s11831-020-09487-1.
A Moving Target Defense Approach,” IEEE Trans Industr [195] A. Maria Christina Blessy and S. Brindha, “A Comprehensive
Inform, Jan. 2022, doi: 10.1109/TII.2022.3190556. Survey of Various Localization Methods in Vehicular Ad Hoc
[178] W. Ben Jaballah, M. Conti, and C. Lal, “Security and design Network,” International Journal of Computer Networks and
requirements for software-defined VANETs,” Computer Applications, vol. 8, no. 6, pp. 818–829, Nov. 2021, doi:
Networks, vol. 169, p. 107099, Mar. 2020, doi: 10.22247/ijcna/2021/210729.
10.1016/J.COMNET.2020.107099. [196] S. Khatri et al., “Machine learning models and techniques for
[179] A. Mahmood, W. E. Zhang, and Q. Z. Sheng, “Software-defined VANET based traffic management: Implementation issues and
heterogeneous vehicular networking: The architectural design challenges,” Peer Peer Netw Appl, vol. 14, no. 3, pp. 1778–1805,
and open challenges,” Future Internet, vol. 11, no. 3, 2019, doi: May 2021, doi: 10.1007/s12083-020-00993-4.
10.3390/fi11030070. [197] K. H. Rashmi and R. Patil, “Survey on Cross Layer Approach
[180] L. Nkenyereye, L. Nkenyereye, B. A. Tama, A. G. Reddy, and J. for Robust Communication in VANET,” Wireless Personal
Song, “Software-defined vehicular cloud networks: Communications, vol. 119, no. 4. Springer, pp. 3413–3434, Aug.
Architecture, applications and virtual machine migration,” 01, 2021. doi: 10.1007/s11277-021-08414-2.
Sensors (Switzerland), vol. 20, no. 4, Feb. 2020, doi: [198] T. Mekki, I. Jabri, A. Rachedi, and L. Chaari, “Software-defined
10.3390/s20041092. networking in vehicular networks: A survey,” Transactions on
[181] O. Sadio, I. Ngom, and C. Lishou, “Design and prototyping of a Emerging Telecommunications Technologies, vol. 33, no. 10,
software defined vehicular networking,” IEEE Trans Veh Oct. 2022, doi: 10.1002/ett.4265.
Technol, vol. 69, no. 1, pp. 842–850, 2019. [199] J. V. Leon, O. G. Bautista, A. Aydeger, S. Mercan, and K.
[182] L. Alouache, S. Yassa, and A. Ahfir, “A Multi-objective Akkaya, “A General and Practical Framework for Realization of
Optimization Approach for SDVN Controllers Placement SDN-based Vehicular Networks,” In 2021 IEEE International
Problem,” in 2022 13th International Conference on Network of Performance, Computing, and Communications Conference
the Future (NoF), 2022, pp. 1–9. (IPCCC) (pp. 1-7). IEEE., pp. 1–7, 2021.
[183] N. S. Radam, S. T. F. Al-Janabi, and K. S. Jasim, “Multi- [200] E. Qafzezi, K. Bylykbashi, P. Ampririt, M. Ikeda, K. Matsuo,
controllers placement optimization in sdn by the hybrid hsa-pso and L. Barolli, “An Intelligent Approach for Cloud-Fog-Edge
algorithm,” Computers, vol. 11, no. 7, p. 111, 2022. Computing SDN-VANETs Based on Fuzzy Logic: Effect of
[184] N. H. Hussein, C. T. Yaw, S. P. Koh, S. K. Tiong, and K. H. Different Parameters on Coordination and Management of
Chong, “A Comprehensive Survey on Vehicular Networking: Resources,” Sensors, vol. 22, no. 3, Feb. 2022, doi:
Communications, Applications, Challenges, and Upcoming 10.3390/s22030878.
Research Directions,” IEEE Access, vol. 10, 2022, doi: [201] H. Li, F. Liu, Z. Zhao, and M. Karimzadeh, “Effective Safety
10.1109/ACCESS.2022.3198656. Message Dissemination with Vehicle Trajectory Predictions in
[185] F. B. Günay, E. Öztürk, T. Çavdar, and Y. S. Hanay, “Vehicular V2X Networks,” Sensors, vol. 22, no. 7, Apr. 2022, doi:
ad hoc network (VANET) localization techniques: a survey,” 10.3390/s22072686.
Archives of Computational Methods in Engineering, vol. 28, no. [202] P. Rani, N. Hussain, R. A. H. Khan, Y. Sharma, and P. K.
4, pp. 3001–3033, 2021. Shukla, “Vehicular Intelligence System: Time-Based Vehicle
[186] F. Lobo, D. Grael, H. Oliveira, L. Villas, A. Almehmadi, and K. Next Location Prediction in Software-Defined Internet of
El-Khatib, “Cooperative localization improvement using Vehicles (SDN-IOV) for the Smart Cities,” in Intelligence of
distance information in vehicular ad hoc networks,” Sensors, Things: AI-IoT Based Critical-Applications and Innovations,
vol. 19, no. 23, p. 5231, 2019. Springer International Publishing, 2021, pp. 35–54. doi:
[187] S. Kumari and D. Agrawal, “A Review on Video Based Vehicle 10.1007/978-3-030-82800-4_2.
Detection and Tracking using Image Processing,” Int. J. Res. [203] J. Leon, A. Aydeger, S. Mercan, and K. Akkaya, “SDN-enabled
Publ. Rev, vol. 2582, p. 7421, 2022. vehicular networks: Theory and practice within platooning
[188] W. Ahmad, S. Ahmed, N. Sheeraz, A. Khan, A. Ishtiaq, and M. applications,” Vehicular Communications, vol. 39, p. 100545,
Saba, “Localization Error Computation for RSSI Based Feb. 2023, doi: 10.1016/J.VEHCOM.2022.100545.
Positioning System in VANETs,” in 2019 International [204] M. A. Mujahid, K. A. Bakar, T. S. J. Darwish, and F. T. Zuhra,
Conference on Advances in the Emerging Computing “Cluster-based location service schemes in VANETs: current
Technologies (AECT), 2020, pp. 1–6. state, challenges and future directions,” Telecommunication
[189] N. Saeed, W. Ahmad, and D. M. S. Bhatti, “Localization of Systems, vol. 76, no. 3. Springer, pp. 471–489, Mar. 01, 2021.
vehicular ad-hoc networks with RSS based distance estimation,” doi: 10.1007/s11235-020-00732-3.
in 2018 International Conference on Computing, Mathematics [205] J. Patil, Suresha, and N. Sidnal, “Comparative study of
and Engineering Technologies (iCoMET), 2018, pp. 1–6. intelligent computing technologies in VANET for delay
[190] M. N. Tahir, M. Katz, and U. Rashid, “Analysis of VANET sensitive applications,” Global Transitions Proceedings, vol. 2,
wireless networking technologies in realistic environments,” in no. 1, pp. 42–46, Jun. 2021, doi: 10.1016/j.gltp.2021.01.007.
2021 IEEE Radio and Wireless Symposium (RWS), 2021, pp. [206] M. Chiesa, A. Kamisinski, J. Rak, G. Retvari, and S. Schmid, “A
123–125. Survey of Fast-Recovery Mechanisms in Packet-Switched
[191] E. Qafzezi, K. Bylykbashi, P. Ampririt, M. Ikeda, K. Matsuo, Networks,” IEEE Communications Surveys and Tutorials, vol.
and L. Barolli, “An Intelligent Approach for Cloud-Fog-Edge 23, no. 2. Institute of Electrical and Electronics Engineers Inc.,
Computing SDN-VANETs Based on Fuzzy Logic: Effect of pp. 1253–1301, Apr. 01, 2021. doi:
Different Parameters on Coordination and Management of 10.1109/COMST.2021.3063980.
Resources,” Sensors, vol. 22, no. 3, p. 878, 2022. [207] L. Jorge, P. Melo, and T. Gomes, “Multiclass data plane
[192] C.-M. Huang and J.-J. Lin, “The k-hop V2V data offloading recovery using different recovery schemes in SDN: a simulation
using the predicted utility-centric path switching (PUPS) method analysis,” Procedia Comput Sci, vol. 204, pp. 305–314, 2022,
based on the SDN-controller inside the multi-access edge doi: 10.1016/j.procs.2022.08.037.
computing (MEC) architecture,” Vehicular Communications, [208] H. P. Joshi, M. L. Sichitiu, and M. Kihl, “Distributed Robust
vol. 36, p. 100496, 2022. Geocast Multicast Routing for Inter-Vehicle Communication,”
arXiv preprint arXiv:2204.00054., Mar. 2022, [Online].
Available: http://arxiv.org/abs/2204.00054
VOLUME XX, 2017 9
[209] S. Babu and A. R. K. Parthiban, “DTMR: An adaptive [227] F. Li, W. Chen, and Y. Shui, “Study on connectivity probability
Distributed Tree-based Multicast Routing protocol for vehicular of VANETs under adverse weather conditions at 5.9 ghz,” IEEE
networks,” Comput Stand Interfaces, vol. 79, p. 103551, Jan. Access, vol. 8, pp. 547–555, 2019.
2022, doi: 10.1016/J.CSI.2021.103551. [228] M. N. Tahir and U. Rashid, “Intelligent transport system (ITS)
[210] N. M. Yungaicela-Naula, C. Vargas-Rosales, J. A. Pérez-Díaz, assisted road weather & traffic services,” in 2020 IEEE
and M. Zareei, “Towards security automation in Software Vehicular Networking Conference (VNC), 2020, pp. 1–2.
Defined Networks,” Comput Commun, vol. 183, pp. 64–82, Feb. [229] N. Wan, Y. Luo, G. Zeng, and X. Zhou, “Minimization of
2022, doi: 10.1016/j.comcom.2021.11.014. VANET execution time based on joint task offloading and
[211] X. Zhang, H. Zhong, J. Cui, C. Gu, I. Bolodurina, and L. Liu, resource allocation,” Peer Peer Netw Appl, pp. 1–16, 2022.
“AC-SDVN: An Access Control Protocol for Video Multicast in [230] Y. He et al., “D2D-V2X-SDN: Taxonomy and architecture
Software Defined Vehicular Networks,” IEEE Trans Mob towards 5G mobile communication system,” Ieee Access, 2021.
Comput, 2022, doi: 10.1109/TMC.2022.3180809. [231] E. Qafzezi, K. Bylykbashi, P. Ampririt, M. Ikeda, K. Matsuo,
[212] Alrehan, Alia Mohammed, and Fahd Abdulsalam Alhaidari, and L. Barolli, “An Intelligent Approach for Cloud-Fog-Edge
“Machine Learning Techniques to Detect DDoS Attacks on Computing SDN-VANETs Based on Fuzzy Logic: Effect of
VANET System: A Survey,” in 2019 2nd International Different Parameters on Coordination and Management of
Conference on Computer Applications & Information Security Resources,” Sensors, vol. 22, no. 3, p. 878, 2022.
(ICCAIS), 2019, pp. 1–6. doi: 10.1109/CAIS.2019.8769454. [232] Mohamed A, Hamdan M, Khan S, Abdelaziz A, Babiker SF,
[213] P. Tam, S. Math, and S. Kim, “Intelligent massive traffic Imran M, Marsono MN. “Software-defined networks for
handling scheme in 5G bottleneck backhaul networks,” KSII resource allocation in cloud computing: A survey.” Computer
Transactions on Internet and Information Systems, vol. 15, no. Networks, 4;195:108151, 2021.
3, pp. 874–890, Mar. 2021, doi: 10.3837/tiis.2021. [233] Zhao, L., Li, X., Gu, B., Zhou, Z., Mumtaz, S., Frascolla, V.,
[214] A. M. El-Shamy, N. A. El-Fishawy, G. Attiya, and M. A. A. Gacanin, H., Ashraf, M.I., Rodriguez, J., Yang, M. and Al-
Mohamed, “Anomaly Detection and Bottleneck Identification of Rubaye, S., “Vehicular communications: standardization and
The Distributed Application in Cloud Data Center using open issues,” IEEE Communications Standards Magazine, vol.
Software–Defined Networking,” Egyptian Informatics Journal, 2, no. 4, pp. 74–80, 2018.
vol. 22, no. 4, pp. 417–432, Dec. 2021, doi:
10.1016/J.EIJ.2021.01.001. Nehad Hameed Hussein received his B.S. degree
[215] S. A. Syed et al., “QoS Aware and Fault Tolerance Based in Computers and Software Engineering from College
Software-Defined Vehicular Networks Using Cloud-Fog of Engineering at Al-Mustansiriya University of Iraq
Computing,” Sensors, vol. 22, no. 1, Jan. 2022, doi: in 2007. Received his M.S. degree in Information
10.3390/s22010401. Engineering from College of Information Engineering
[216] G. N. Nurkahfi et al., “Cross-Layer Design on Software Defined in Al-Nahrain University (Iraq) in 2011. Currently, he
Vehicular Network (SDVN) based on Radio Access Technology is a lecturer in the computer techniques engineering
(RAT) IEEE 802.11 bd to Improve End-to-End Communication department in Baghdad College of Economic Sciences
Performance,” in 2022 8th international conference on wireless University/Iraq, as well as doing his PhD in Electrical
and telematics (ICWT), 2022, pp. 1–6. and Communication Engineering from College of
[217] K. Praveen kumar, P. Sivanesan, and R. Thirumurugan, “Flow‐ Engineering in UNITEN University/Malaysia. He has participated in many
rule integration for quality-of-service enhancement in software‐ research workshops and Scientifics conferences sponsored by IEEE,
defined vehicular network,” International Journal of Springer, and IOP. His research interests include VANET, Ad-hoc routing
Communication Systems, vol. 35, no. 18, p. e5350, 2022. protocols, SDN, wireless communication optimization, and digital signal
[218] K. Smida, H. Tounsi, and M. Frikha, “Intelligent and resizable processing.
control plane for software defined vehicular network: a deep
reinforcement learning approach,” Telecommun Syst, vol. 79, Prof. Dr. Johnny Koh Siaw Paw is currently a
no. 1, pp. 163–180, 2022. Professor in the Institute of Sustainable Energy in
[219] N. S. Radam, S. Al-Janabi, and K. Shaker, “Optimisation Universiti Tenaga Nasional. He received Bachelor
Methods for the Controller Placement Problem in SDN: A degree (1st Class Honour) in Electrical & Electronic
Survey,” Webology, vol. 19, no. 1, pp. 3130–3149, 2022. Engineering (2000), M.Sc degree (2002), and Ph.D.
[220] N. Lin, Q. Zhao, L. Zhao, A. Hawbani, L. Liu, and G. Min, “A degree (2008) from Universiti Putra Malaysia. His
novel cost-effective controller placement scheme for software- areas of interest are in machine intelligence,
defined vehicular networks,” IEEE Internet Things J, vol. 8, no. automation technology and renewable energy.
18, pp. 14080–14093, 2021.
[221] K. S. K. Liyanage, M. Ma, and P. H. J. Chong, “Controller Ts. Dr. Yaw Chong Tak received his Bachelor’s
placement optimization in hierarchical distributed software degree with honours from Universiti Tenaga Nasional
defined vehicular networks,” Computer Networks, vol. 135, pp. (UNITEN), Malaysia in electrical and electronics
226–239, 2018. engineering in 2008. He received his Master’s degree
[222] I. Maity, R. Dhiman, and S. Misra, “Mobiplace: Mobility-aware with honours from UNITEN in electrical and
controller placement in software-defined vehicular networks,” electronics engineering in 2012. He earned his Ph.D.
IEEE Trans Veh Technol, vol. 70, no. 1, pp. 957–966, 2021. in 2019 from UNITEN in artificial neural network. His
[223] M. Adnan et al., “Towards the design of efficient and secure research interests include artificial neural networks
architecture for software-defined vehicular networks,” Sensors, and renewable energy. Currently, he is working as a
vol. 21, no. 11, p. 3902, 2021. post-doctoral researcher at Institute of Sustainable Energy in UNITEN.
[224] A. Boukerche and N. Aljeri, “Design guidelines for topology
management in software-defined vehicular networks,” IEEE Prof. Ir. Ts. Dr. Tiong Sieh Kiong is currently a
Netw, vol. 35, no. 2, pp. 120–126, 2021. professor in the College of Engineering. He is also the
[225] F. Luo, S. Wang, Y. Gong, X. Jing, and L. Zhang, “Geographical Director for Institute of Sustainable Energy (ISE),
information enhanced cooperative localization in vehicular ad- Universiti Tenaga Nasional. He received his
hoc networks,” IEEE Signal Process Lett, vol. 25, no. 4, pp. 556– B.Eng.(Hons), MSc and PhD, in Electrical, Electronic
560, 2018. and System Engineering from the Nasional University
[226] S. K. Parveen and R. Singh, “GPS-Free Localization in of Malaysia (UKM) in year 1997, 2000 and 2006
Vehicular Networks Using Directional Antennas,” Data respectively. His research interests are renewable
Engineering for Smart Systems: Proceedings of SSIC 2021, vol. energy, artificial intelligence, data analytics, microcontroller system and
238, p. 173, 2022. communication system. He is currently a Professional Engineer registered
with the Board of Engineers Malaysia (BEM). He is also a Member of the
Institute of Electrical & Electronic Engineers (IEEE)
VOLUME XX, 2017 9
Dr. Tan Chung Hong received his Bachelor's degree

Ts. Dr. Benedict Foo is currently the Managing with honours from the University of Nottingham
Director of Enhance Track Sdn. Bhd. He Malaysia (UNM) in Chemical Engineering in 2013.
is also the founder of Enhance Track Sdn. Bhd. He He earned his Ph.D. in 2020 from UNM as well in
has been managing the company for Chemical Engineering. His research interests include
17 years since its establishment in 2005. He has renewable energy and machine learning. Currently, he
received two bachelor’s degrees in mechanical is working as a post-doctoral researcher at the Institute
engineering, a Graduate Certificate in Mechanical of Sustainable Energy in Universiti Tenaga Nasional
Engineering from the University of Southern (UNITEN).
Queensland, Australia, and recently attained his
Doctor of Philosophy in Engineering from Universiti Malaysia Pahang. As
the managing director of Enhance Track Sdn. Bhd., he has vast experience
in many industries such as oil and gas, renewable energy, education, and
laboratory and testing equipment. In this role, he has also personally been
involved with all the research and development projects that the company
has undertaken and improving the outcome for clients in Malaysia,
Australia, and the Middle East. He is also active in associations and
professional bodies such as the Associated Chinese Chamber of Commerce
and Industries of Malaysia (ACCCIM), and the Selangor and Kuala Lumpur
Foundry and Engineering Association (SFEIA). He is also a Professional
Technologist registered with the Malaysia Board of Technologists (MBOT).
Professor Dr. Talal Yusaf currently hold the

position of Executive Dean - Higher Education and
Emerging Technologies, Aviation Australia,
Queensland Government, Australia, and Adjunct
Professor at Central Queensland University.
Professor Yusaf commenced his role as Pro Vice
Chancellor, International and Partnership at
Federation University Australia in September 2018,
and was the Executive Director (International and
Research Partnership) at University of Southern Queensland (2008-2018).
Professor Yusaf’s international university career has included teaching,
research and management positions in Southeast Asia, the Middle East, the
United Kingdom and in Queensland for the last 30 years. Professor Yusaf
holds PhDs in renewable energy and biotechnology, and has a strong
research background in Renewable Energy, Hydrogen Technologies,
environmental pollution, alternative fuels for IC engines. Professor Yusaf’s
academic qualifications include BSc (Hons) UoT Baghdad 1987, master’s
in engineering science 1994 (UKM), PhD1 USQ Queensland Australia and
PhD2 UKM Malaysia 1999. Professor Yusaf has supervised over 50 PhD
and Postgraduate students and has written and contributed to over 200
journals and 10 books. Prof. Yusaf’s professional memberships and
affiliations include Pro Vic Chancellor / Deputy Vice Chancellor Forum
(Australia), Australian University International Director Forum AUIDF,
Chair of QLD international Director Forum, Director- International Sponsor
Engagement and Research Partnerships (ISERP), Bio-Energy Research
Group Leader at the National Centre for Engineering in Agriculture
Queensland. He is Associate Editor-in-Chief for Sustainability Journal,
Applied Sciences Journal, Energies and Groundwater for Sustainable
Development.
Prof. Ts. Ir. Dr. Kumaran Kadirgama is a Professor

and research fellow of the Advanced Nano Coolant
Lubricant Laboratory (ANCoL), the Automotive
Engineering research group at the Faculty of Mechanical
& Automotive Engineering Technology, University
Malaysia Pahang (UMP). He is a Professional Engineer
registered under the Board of Engineers Malaysia
(BEM); and a Charted Engineer (UK) under the
Institution of Mechanical Engineers (IMechE). In
research, he is involved in the supervision of postgraduate students at the
Master’s and PhD levels and has mentored numerous students to graduation.
He has also published and presented various technical papers and journals
at an international and national level. He has an h-index of 31 with 4199
citations. He has received grants totaling RM 6.02 million from various
agencies and institutions. He has won gold medals in the International
Invention, Innovation & Technology Exhibition (ITEX), Seoul International
Invention Fair, Korea (SIFF), and British Invention Show (BIS).
VOLUME XX, 2017 9

SDN-Based VANET Routing A Comprehensive Survey On Architectures Protocols Analysis and Future Challenges

Uploaded by

Copyright:

Available Formats

SDN-Based VANET Routing A Comprehensive Survey On Architectures Protocols Analysis and Future Challenges

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

SDN-Based VANET Routing A Comprehensive Survey On Architectures Protocols Analysis and Future Challenges

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in IEEE Access.

SDN-Based VANET Routing: A Comprehensive

I. INTRODUCTION that approximately 92% of road accidents are caused due to

Discussion of SDN Architectures

Discussion of role of SDN in data Routing

Data Routing Optimization Metrics

Protocols Application Area

Comparison of Optimization Criteria

Analysis of performance evaluation

Simulation & Testbeds Tools

New Trends and Upcoming Technologies

2022 [11] ✓ ✓ ✓ ✓ ∂ VANET/V2X routing using non-learning- and learning-based approaches

6 VOLUME XX, 2017

FIGURE 1. Survey Structure.

FIGURE 2. Basic VANET Communications

FIGURE 3. DSRC spectrum band channels.

FIGURE 4. Taxonomy of conventional VANET data routing.

FIGURE 5. Components of Software-Defined Vehicular Network.

IV. SDN-BASED VANET NETWORKS A. SDVN BASIC DEFINITION

FIGURE 7. The basic principles of data routing in SDN-based VANETs.

FIGURE 8. SDVN routing protocols taxonomy.

PT-GROUT routing protocol aims to enhance routing

6 VOLUME XX, 2017

6 VOLUME XX, 2017

6 VOLUME XX, 2017

6 VOLUME XX, 2017

6 VOLUME XX, 2017

Packet Delivery Ratio

Packet Loss Ratio

Bit Error Rate

Normalized Routing Load

Average Routing Reply Ratio

Average Routing Discovery Time

Routing Request Ratio

Percentage Idle Time

Availability of Recovery Process

Flooding Mitigation Concern

6 VOLUME XX, 2017

Hop Count Accountability

RSU Needs Consideration

Road intersection consideration

Communication Distance Consider

Localization Inf. Requirements

Traffic Density Awareness

Mobility Predication Consideration

Vehicles Mobility Concern

Increased Requests Efficiency

Traffic Condition Adaptability

Node Energy Awareness

6 VOLUME XX, 2017

with SDVN routing is inherently challenging and severely

1. Security challenges of deploying protocols.

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

VOLUME XX, 2017 9

18) Further reduction of computation load: The deployment VIII. CONCLUSIONS

VOLUME XX, 2017 9