Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Journal of Ambient Intelligence and Humanized Computing

https://doi.org/10.1007/s12652-019-01253-x

ORIGINAL RESEARCH

Congestion avoidance through fog computing in internet of vehicles


Shumayla Yaqoob1 · Ata Ullah1 · Muhammad Akbar1 · Muhammad Imran2   · Muhammad Shoaib2

Received: 2 June 2018 / Accepted: 13 February 2019


© Springer-Verlag GmbH Germany, part of Springer Nature 2019

Abstract
Recently, internet of vehicles (IoVs) is captivating a lot of interest due to a wide range of applications in various domains.
These applications rely on up-to-date information of vehicles for provisioning various services. However, frequent message
transmissions by a sheer number of vehicles may not only engulf a centralized server but may also cause a severe congestion
which is not suitable for ongoing services specially in emergency situations. The aim of this study is to reduce congestion and
messaging overhead. This paper presents a fog-assisted congestion avoidance scheme for IoV named energy efficient message
dissemination ­(E2MD). Unlike most of the existing schemes, ­E2MD capitalizes the merits of fog computing to minimize
communication cost and manage services. Each vehicle is required to update their status to a fog server frequently, either
directly or through intermediate nodes. In case of an emergency, the fog server will timely inform upcoming traffic to slow
down, dispatch rescue teams to provide necessary services, and coordinate patrolling missions to clear the road. Moreover,
failure detection and recovery mechanisms are also presented to ensure availability of the fog server. The performance of
the proposed scheme is validated through NS 2.35 simulations. Simulation results confirm the performance reign of E ­ 2MD
compared to contemporary schemes in terms of latency and communication cost.

Keywords  Internet of vehicles · Fog computing · Congestion avoidance · Messaging overhead

1 Introduction communication network that allows mobile communication


between vehicles and other networks. Traditional vehicular
Internet of vehicles (IoVs) is a significant component of ad-hoc network (VANET) is growing into IoVs for the sake
intelligent transportation systems (ITS) for the sake of of intelligent and efficient transportation system in future.
road safety and other transportation services by saving data In fact IoVs help to manage huge data transmission, com-
at main online repositories for decision making (Yi et al. putation, and storage for users and owners (Xu et al. 2018).
2015). Recent research in sensing and communication tech- IoVs is an emerging realm with a number of applications in
nologies give way to IoVs in wide range. IoV is an active various domains such as smart traffic management, mobile
crowd sensing, accident reporting, weather alerts, monitor-
ing, parking alerts and audio/video streaming in vehicles.
* Ata Ullah In IoVs, vehicle highlights junctions, fuel stations, park-
aullah@numl.edu.pk
ing plazas and marts on the trajectory between source and
Shumayla Yaqoob destination locations (Fangchun et al. 2014) as shown in
shumaylayaqoob@gmail.com
different scenarios in Fig. 1. In real world IoV are used
Muhammad Akbar to give E-health applications in the terms of mobile hos-
makbar@numl.edu.pk
pitals (Lin et al. 2017). Vehicular networks confirm safety
Muhammad Imran on the road, organize traffic by using signals, and timely
cimran@ksu.edu.sa
intimate accidents by swapping useful information (Ahmad
Muhammad Shoaib 2017). To meet this ever-growing transportation needs, IoV
muhshoaib@ksu.edu.sa
is an attractive idea that needs to be explored for the sake
1
Department of Computer Science, National University of smooth traffic and road safety by resolving various con-
of Modern Languages, Islamabad 44000, Pakistan straints like congestion.
2
College of Computer and Information Sciences, King Saud
University, Riyadh, Saudi Arabia

13
Vol.:(0123456789)
S. Yaqoob et al.

In IoV, congestion during information dissemination is messages (Kumar and Dave 2013; Zrar et al. 2013). Direct
a major issue which is handled by existing schemes using communication among server and vehicles is not possible
congestion detection (Roma and Gayatri 2016; Wan et al. due to limited communication range of vehicles. Therefore,
2016) i.e., detect gradual blockage of message and timely vehicles may rely on multi-hop communication which may
decide about the alternate paths. Most of the existing con- result in message storm, lead to congestion. Timely delivery
gestion avoidance schemes to deal with message storm are is a prime factor in communication of safety messages in
proposed in context of VANET. With the increasing demand order to avoid catastrophic consequences and congestion that
of vehicles for wireless applications, IoVs need to work on results into packet loss, poor communication, low latency,
high mobility and limited lifetime (Paul et al. 2015). Moreo- energy overheads and delay in propagating emergency
ver, IoV-based schemes mainly rely on a centralized server messages. Message congestion in IoVs can lead to weak
to provide services, which might not be feasible for a large road safety. It occurs when a vehicle broadcast the packet
number of vehicles demanding for diverse applications. repeatedly. Repeated packets are inefficient for scalability
When messages are transmitted repeatedly to central server and causes packet collisions (Panichpapiboon and Pattara-
from various vehicles in IoV at the same time, it creates atikom 2012; Chaqfeh et al. 2014). Packet collisions lead to
message storm and hence causes congestion. It is a chal- loss of important data that results in poor communication
lenging issue in IoVs environment that may lead to service of EMs’ delivery.
degradation besides ruinous consequences. Basically, all IoV is a type of wireless network where vehicles behave as
vehicles are needed to update central server continuously mobile nodes for data collection and information distribution.
for safety and infotainment services (Chang et al. 2016), These mobile nodes have enough storage, and communication
sending emergency messages (EMs) about accident, road capabilities (Khabazian et al. 2011; Dandala et al. 2017). High
blockage, foggy weather or storm ahead (Syed et al. 2017; traffic density effects the performance of information diffusion
Schwarz and Rup 2016). In this case, the emergency vehicle models in one-hop and multi-hop scenarios. For efficient mes-
does not stop broadcasting messages until a rescue is not saging, it should avoid long delays during message distribu-
reached. It becomes worst when the nearby passing vehi- tion (Sushama and Choudhary 2017; Julio et al. 2016). IoVs
cles also keeps on flooding the emergency information until ensure driving safety, organize traffic and timely report acci-
a long distance that causes communication congestion and dents by exchanging useful information and achieve smooth
energy dissipation. This type of communication between traffic by handling congestion as well (Li et al. 2011) (Bihari
a central server and vehicles occurs through beaconing et al. 2010). In IoVs, role of edge computing is quite relevant

Fig. 1  Communication Sce-
narios in IoVs

Internet of Vehicles
Vehicle to Grid

Vehicle to Device

V2D
V2G+G2V
Telemacs Charging Staons

Parking Alerts

V2V

Communicaon

Communicaon
V2I

13
Congestion avoidance through fog computing in internet of vehicles

to ensure computation closer to the vehicles (Ahmed 2017; 2 System model and problem statement
Hassan 2018). It can be categorized as mobile computing and
fog computing (Yaqoob 2016; Park and Yoo 2017) where later IoVs includes number of moving vehicles as connected
is adopted to reduce the communication delays that become members in a network. These mobile vehicles may connect
worst in high mobility. It also demands location awareness and directly or via multi-hop communication. Vehicles exchange
direction of moving vehicles to improve information exchange all type of messages either normal message (NMs) or EMs
in less delays. It makes computation tasks easy by computing for routine messaging and emergency alerts about accidents
workload at the network edges (Merino and Vaquero 2014) respectively. These messages includes position of vehicle,
using fog servers positioned locally. velocity, heading information, and other emergency or safety
This paper presents a Fog-assisted congestion avoid- based information (Kang et al. 2017). We have assumed
ance scheme called energy efficient message dissemination that vehicles maintain a list of one hop neighbors based on
­(E2MD). This approach helps to prevent congestion and dis- their communication range. There are two types of major
seminate all type of messages effectively for reliable com- public safety messages including periodic safety (beacon)
munication. It ensures delivery of emergency messages and messages and event driven messages (Xiaomin et al. 2017).
timely intimation of accidents for safety. ­E2MD relies on These types of messages are meant to be forwarded to other
fog computing which helps to distribute data and process nodes. Event driven messages occur in the case of emer-
it locally. Vehicles are used to update fog server about their gency like accident. This type of messages is considered as
current status on regular basis. In proposed scenario, abnor- reliable in order to deliver in multi-hop neighbors. There is
mal vehicle (Av) sends messages to fog server and one hop a scenario where too many accidents are happening at the
neighbors directly. Hence, fog server is responsible to inform same time. Emergency alerts are sent to server and finding
upcoming traffic to slow down speed. Similarly, fog server is nearest server users so that it causes congestion as shown in
supposed to approach rescue team for medication and coor- Fig. 2. Continuously transmitted messages causes storm that
dinate with patrolling missions to remove accidental vehicles becomes severe in case of high vehicles density due to more
from road. If a fog server is not in direct communication packet delay and collisions (Ucar et al. 2016).
range of Av then, it communicates with nearby vehicles in Congestion avoidance schemes are essential in IoV. Main
one hop and selects a reliable vehicle that has shortest path influence of the problem is observed that messages occurs
towards the fog server and transmits messages to it. The reli- repeatedly on each transmission within one-hop and multi-
able vehicle informs fog server in order to approach nearby hop groups. When messages are transmitted blindly to cen-
vehicles. In case of emergency messages, rescue team and tral server from various vehicles in IoV at the same time, it
upcoming traffic could be the nearby resources. In cases, creates message storm and hence congestion. The vehicles
where server is not in the range of Av, then messages reach entering into a region that is suffering from congestion can
to server via V2V communication in IoVs. For the smooth be the cause of dropping established sessions for audio,
operability of fog server, failure detection and the recovery video streaming, monitoring or emergency reporting. This
algorithm is presented to timely identify the malfunctioning. message congestion leads to packet drop, communication
Simulation is performed in NS 2.35 for the validation of overhead and increase in delays due to unnecessary messag-
­E2MD by comparing it with existing schemes. It improves ing. These constraints are still tolerable in case of NMs while
packet delivery ratio and delay whereas reduce bandwidth in the case of EMs, packet drop and delay in message deliv-
consumption due to reduced messaging cost. ery during accident reporting is more critical. In existing
The rest of paper is organized as follows; Sect. 2 describes schemes, central units and neighbor vehicles causes delay
system model and problem statement. Literature review and in message delivery and repetition of messages that results
background knowledge is discussed in Sect. 3. Section 4 pre- in communication overhead.
sents the proposed E ­ 2MD scheme along with an algorithm
for message dissemination that manages the congestion as
per different vehicle types. It also includes the algorithm for 3 Related work
fog server failure and recovery detection. The results and
analysis for the proposed and existing schemes are presented In existing information dissemination and message conges-
in Sect. 5 along with simulation scenario and parameters. tion avoidance schemes, most of these schemes face high
Section 6 concludes our work and future directions are also message congestion and repetition rate. It arises due to
indicated. repetition and leads to information loss. Delay in message
delivery is also a main feature to be focused for communi-
cation. Most schemes followed static or limited dynamic
approaches that are not suitable in all cases. Congestion

13
S. Yaqoob et al.

Fig. 2  Message congestion in
IoVs

avoidance schemes along with their limits are discussed loss DL which effects message delivery as given in Eq. 1.
in this section. All schemes are categorized in a taxonomy In Park and Yoo (2017), it avoids congestion by inform-
under specific categories under static and dynamic schemes ing nearby vehicles (NbV) and stationary vehicles (StV) as
as depicted in Fig. 3. These schemes focused to reduce the shown in Eq. 2. Therefore, these schemes are limited where
rate of message congestion for reliable communication but high vehicle velocity effects message delivery. The Av only
still there are several limitations. transmit EMs to road side stationary nodes and neighbor
vehicles in predictable pattern respectively for congestion
3.1 Bandwidth control‑based schemes avoidance (CA):

The scheme (Munir et al. 2010) includes two major con-


CA = Vv ∝ DL, (1)
cerns including message storms and the disconnection due CA = acct → NbV and StV. (2)
to high bandwidth consumption. The scheme reduces num-
ber of repeated messages at local level. In this approach,
velocity of vehicles Vv is directly proportional to distribution

Messaging in VANET

Event Driven Beacon Messages


messages

Dynamic Beaconing Static Beaconing

Bandwidth Transmission Segmentation Encryption Integration


power based Position based
control based based based based

Fig. 3  Taxonomy of message congestion avoidance schemes

13
Congestion avoidance through fog computing in internet of vehicles

3.2 Transmission power based schemes direction with highest number of messages; (ii) rule 2 is used
to keep messages as active as possible. In case of second
In Munir et al. (2010), a scheme is presented to resolve mes- rule, if there is same number of messages then the message
sage congestion. Author of this scheme stated that message arrived earlier can be considered first:
storm is a result of high bandwidth power during transmis-
M−1 N−1
sion of repeated messages on unique channel such as 802.11 ∑ ∑
Score = diA − diB . (5)
network. This scheme reduces limited congestion by hav- i=0 i=0
ing control on transmission power but still need to provide
dynamic algorithms to control transmission power. In Djahel The CANCORE scheme restricts the distance between
and Ghamri-Doudane (2012), authors have presented a con- the sender and the receiver of a selected message to handle
gestion control scheme that includes priority assignment and the case of long delay between the arrival time for packets
congestion detection. The scheme adjusts the transmission of P1 and P2 when the beacons are received from vehicles that
beacon messages. It comprises of following three phases; (i) are out of direct transmission range. In Sindhu and Mit-
priority assignment of message for transmission; (ii) detect tal (2016), a distributed denial of service (DDOS) attack
the congestion caused by several repeated messages simul- detection and prevention scheme is presented to give better
taneously; (iii) adjust the beacon messages storm as given path for the safe V2V communication. The authors claim to
in Eq. (3). The scheme is quite hard to implement in real reduce the communication delay and communication loss.
deployments: Firstly, the work model created the controlled clusters and
{ } apply the several limits based analysis to identify the attack.
CA = min tp ∝ min{bt }. (3)
3.5 Integration based schemes
3.3 Segmentation based schemes
A probability based architecture (Kumar and Dave 2013)
Segmentation is a process of dividing large area into specific is presented to resolve EMs broadcast storm for the sake
segments where a small or specific area is considered as a of less congestion. It controls repeated EMs to achieve less
segment. A new segmentation based mechanism (Chuang broadcast rate. Limited bandwidth rate may help to improve
and Chang Chen 2013) is proposed that broadcasts messages packet delivery ratio during data broadcasting among vehi-
to resolve the EMs broadcast storm issue. It achieves less cles. Proposed framework is based on dictionary of local
delay in distribution and reliability but only for specific area. knowledge in order to save the all possible decisions for
In Xia et al. (2014), authors proposed a scheme for message data aggregation. This storage of decision making may
distribution among vehicles. It is based on segments where increase communication overhead. In Chang et al. (2015,
it divides time slots along transmission in order to improve 2016), authors presented a scheme to resolve congestion by
delay as shown in Eq. (4). But the scheme is also not suitable introducing clustering and aggregation. This scheme relies
for sparse environment: on two phases i.e., adaptive forwarding message (AFM) and
AFM including cooperative active safe driving (AFM-CSD)
CA = Div(Time) ∝ less(Delay). (4) where message sent to group head then that is responsible
to forward message by using V2V communication in a hop
3.4 Encryption based schemes
as given in Eq. (6). It only improves limited congestion for
CSD-Vehicles that are considered as smart vehicles:
A Context-Aware Network COded REpetition (CANCORE)
scheme is focused to improve reliability and effectiveness of 1hop

applications. It is based on following two points; (i) trans- CA = GH → V ∝ min(bandwidth, delay). (6)
mit the EMs vehicles extensively with the help of network i=1
coding; (ii) control repetition in order to avoid unnecessary
reporting growth which may lead to congestion caused by 3.6 Position based schemes
repetition of coded EMs. Beacon messages are updated regu-
larly as shown in Eq. (5) where each vehicle calculates the In Yang et al. (2004), a vehicle-to-vehicle locality based
score and also prepares encrypted packets P1 and P2. These broadcast communication protocol is presented where each
encrypted packets consist of two native messages using fol- vehicle produces emergency alerts at a fix rate. Considered
lowing two rules; (i) rule 1 helps to reduce overhead by giv- a scenario that all vehicles are not equipped with wireless
ing permission to large number of vehicles for the selection transceivers so urgent message transferring in lightly con-
of only one point from P1 and P2 at the receiving time. In nected ad hoc network having high mobile vehicles. When
first rule, the scheme chooses one simple packet from each an abnormal event occurs, mostly there are many vehicles

13
S. Yaqoob et al.

affected by the scenario. By summing up, MAC protocols clusters (Hou et al. 2016). Vehicles as infrastructure by
take channels among various vehicles; multi-hop forwarding using Fog causes to reduce the extra cost of main infra-
technique expands the reachable location for warning alerts structure. It consists of multi-hop and moving features
and collected management protocols identify the group of of vehicles to hold the information from one location to
vehicles in order to share it. Here the author focused on another. Sharma et al. proposed communication architec-
congestion control based on collision warning application ture based on fog nodes. This scheme helps to provide
and proposed a model to deal with emergency warning mes- strong connectivity for smooth communication. Mobile
sages (EWM) (Yang et al. 2004; Ullah et al. 2019). It also nodes are introduced to balance work load (Munir et al.
discussed how to avoid the natural emergency situations. 2018). These mobile nodes works on fast-moving plat-
This paper focuses on congestion control among vehicles form where fog servers are placed on edges (Hassan 2018;
using fog computing. Congestion is a major constituent of Sharma et al. 2017; Gupta and Yamaguchi 2018). Mobile
intelligent transportation system (ITS) for safety (Yusof and nodes are used to resolve and manage data on edges of the
Abu-Bakar 2011). In Gutiérrez-Reina et al. (2018) author network (Gupta and Yamaguchi 2018). To meet the up-
proposed a scheme to remove distance for data diffusion by growing requirements of transportation, need any scalable
reducing retransmission. This scheme used Euclidean dis- solution (Gianfranco et al. 2017). Fog computing provides
tance formula to increase reachability for smooth communi- this scalable solution for large data. A reliable model for
cation. In most of the schemes in literature, the major limita- flooding (RMFF) is based on restrictive flooding which
tion are as follows: (i) congestion control only for specific is proposed to overcome the message storm (Sattar et al.
area rather than real scenario, means reduced message; (ii) 2018). However, the scheme is still not efficient enough to
storm for small dense area rather than large dense area; (iii) control repeated message transmission. Through literature
Increased communication overhead, due to several interme- review, we have figured out that static schemes and fixed
diate nodes for communication and repeated messages; (iv) small segment slots are not efficient approaches to manage
authors assumed smart vehicles and has given only solution vehicles. These schemes are based on specific and limited
for smart vehicles integrated with V2I sensors and internet areas. These static and small road segments disturb the
other than the normal vehicles in real scenario. actual situation on roads. Actual situation refers to real
scenario with dense environment of vehicles. Whereas,
3.7 FoG‑oriented VANET schemes large segments lead to high communication load, results in
accuracy loss. Thus, decisions should be dynamic based on
A general and flexible design permits it to be used for sev- contents in order to collect data from vehicles. Dynamic
eral applications including management of traffic lights, schemes are accurate and reliable to collect data instead
detection of road conditions for accident avoidance. Fog of predefined roads. Generally vehicular networks consist
computing is used to support the mobile devices used by of two nodes named vehicles and road side units (RSUs)
humans for communication for example, augmented real- (Ahmad 2017). Vehicular communication is a component
ity and connected vehicles (Chiang and Zhang 2016). Fog of the intelligent transport system. Its aim is to attain
computing gives virtual platform that computes data on safety via intelligent transportation system by integrating
its edges. Fog computing has the concept of local server different information. Communications in an active trans-
that brought cloud to the edges (Ahmed 2017; Merino port system take place by using the combination of com-
and Vaquero 2014). It is used to reduce latency and act munications like vehicle-to-vehicle (V2V) and vehicle to-
as backbone for bandwidth saving in order to achieve bet- infrastructure (V2I). For V2V communication, Vehicles
ter quality of services. Fog design presents an efficient exchange information directly where the main problem is
resource allocation architecture and algorithm (ERA) on the distance between the two vehicles. If it is lesser than
cloud in order to test the performance. Thus, Fog is able to the communication range in order to build up a successful
handle resources in an optimized way in order to improve connection. Network connectivity is a basic issue for ena-
overall round trip time, data transfer cost and bandwidth bling information transmission for V2V communication.
utilization. In Kukreja and Sharma (2016), the authors Thus, we go to the fog assisted IoVs which takes moving
have shown that fog has proficient algorithm including vehicles as communication nodes in order to establish bet-
resources that is based on virtualization technique. It has ter network connectivity.
also proved that fog computing approach is able to allocate Message are disseminated through group heads or V2V
resources in optimistic manner and better than previous communication in a specific area that becomes more chal-
algorithms in terms of response time, data transfer cost lenging in real road scenario or dense environment. High
and bandwidth utilization. Vehicular Fog Assisted VANET message congestion rate leads to delay in message deliv-
is a structure that deals vehicles as permanent node or ery and increases communication cost due to repetition. In
infrastructure with a central powerful node to manage the

13
Congestion avoidance through fog computing in internet of vehicles

literature, we have identified a number of issues in different Table 1  List of notations for ­E2MD
schemes as follows. Notation Description

(i) Each vehicle node transmits messages to each other by Av Abnormal vehicle
using V2V and RSU before passing to server (Chang EmT Emergency message transmission
et al. 2015, 2016; Munir et al. 2010) EMs Emergency messages
(ii) More complex tree structure and cluster formation are N Set of packets
required in hierarchal architecture (Kumar and Dave N Nodes
2013; Ucar et al. 2016). NMs Normal messages
(iii) Limited congestion is considered despite of real sce- SN Sink node
narios with two sided traffic (Chang et al. 2015, 2016; MT Total messaging cost
Chuang and Chang Chen 2013. Vmn , Vmx Minimum and maximum velocity
(iv) Limited known pattern is considered for information Vc Current velocity
exchange rather than unknown patterns for traffic (Park Psize Packet size
and Yoo 2017). RPackets Received packets
(v) Additional overhead for encryption and decryption Rltv , Rhtv Range of LTV and HTV traffic
operations (Lin et al. 2017). k Constant
Cc Congestion control
This work is significant extension of our previous work GH Group head
presented in Shumayla et al. (2018). A number of enhance- tp Transmission power
ments and improvements in this article are introduced. For Sv Smart vehicle
example, we have categorized and classified a comprehen- Bv Basic vehicle
sive literature review conducted in this study. Moreover, Iv Intelligent vehicle
additional performance metrics such as packet delivery ratio, acct Accident transmit message
packet loss ratio, message overhead, delay and throughput bt Beacon transmission
are computed in this work. Ff Far-distance failure
Nf Near-distance failure

4 Energy efficient message dissemination


­(E2MD) scheme AV = Abnormal Vehicle
SN = Sink Node
GH = Group Head
Message congestion has significant influence on information N = Simple Node
SN
dissemination in IoVs. This paper proposed Fog assisted N
N N
model for IoVs, to avoid congestion by exchanging messages AV GH GH N
GH
including NMs and EMs between vehicles and fog server N
N N
on regular basis. Basically it helps to avoid repeated mes- N
sages when messages occur frequently. In fact E2 MD helps N AV GH
to deliver messages timely which may help to save precious N
AV
lives in the case of EMs. A list of notations for E ­ 2MD is N

provided in Table 1.
Proposed scenario includes one way roads, having three
­ 2MD
Fig. 4  Proposed model for E
lines including fast medium and slow respectively. Fog
server responds to normal and emergency events in order
to timely avoid potential losses. In case of emergency,
2019). Basically Fog server acts as local server in different
the urgent alert sent by Av is needed to be delivered to the
manners. In this scheme, the network contains three dis-
nearby vehicles and reporting server in order to arrange
tinct types of nodes. First one is SN (sink node) that act as
quick response. Motion of vehicles provides good message
fog server. SN has high energy, memory, speed and com-
carriers to transmit information by making new connec-
putation power than other types of nodes. Second one GH
tions continuously. Due to these communication hubs, the
(group head), having more energy, computation power and
fog is shaped instead of sending meaningful data to cloud
range than ordinary nodes. Third one is simple node (N)
servers. Fog server share communication resources locally,
having less energy, computation power and range than GH.
that includes both local decision making and geo distribu-
This network is based on priority and groups as shown in
tion characteristics in order to bring less delay (Ullah et al.

13
S. Yaqoob et al.

Fig. 4. Either group is controlled by only GH or by both SN time delay and less cost for geo related vehicles. Proposed
and GH. Simple nodes N have the ability to communicate solution is categorized in two sub sections to discuss mes-
with GH and neighbor nodes. GH acts as intermediate node saging by utilizing vehicle type support. A novel algorithm
between SN and N. It is based on three distinct scenarios like is also proposed to provide a stepwise solution for managing
in first case, communication takes place through intermedi- messaging as follows.
ate node GH. In Second case, normal nodes are involved
while GH communicate with each other in third case. In 4.1 Priority based vehicle selection for messaging
last case, GHs are responsible for communication, Av sends
message to GH and the GH is responsible to find fog server. Vehicles act as moving nodes as mobility is essential dur-
Therefore, GH transfers message to other GH until it finds ing transportation. An Av is reported when any mobile
the fog server. Here SN is serving as fog server. In addi- node acts unusually on the road. It includes motion of node
tion, this model is based on three cases that involve GH for with high velocity, sudden change in angle or position of
information exchange. In first case, Av informs SN through vehicle due to slippery roads, accident and break failure.
GH where SN stops Av to forwarding messages and informs A vehicle can be an Av when any mechanical breakdown
upcoming traffic through GH. In second case, communica- or sudden road hazards happen in the vehicles. If Iv or Sv
tion takes place through V2V. In third case, GHs communi- vehicle is considered as Av then it can transmit message to
cate with each other to forward message till SN. Proposed fog server directly for medication and road clearance. If BV
scheme is dynamic and considers real scenario with multi type vehicle is facing abnormal situation then it opts the
hop communication for the sake of reliable communication most reliable moving node from one hop neighbors, rather
and achieves reduced bandwidth and delay by avoiding mes- than blind transmission of repeated messages. Priority
sage repetition. based vehicle selection follow this equation Iv > Sv > Bv.
Our protocol considers three different types of vehicles Iv is at first as it has highest priority because it informs fog
according to their ability. First one is intelligent vehicle (Iv) server directly by using V2I sensors and internet. Sv has
that has ability to detect the hurdles and velocity by utiliz- higher priority than Bv. In fact, proposed solution saves
ing sensors in vehicle to find an accident and inform to fog the bandwidth rate by reducing transmission of repeated
server directly with the help of internet and V2I sensors. Fog messages from the Av. We have presented algorithm 1
server has ability to store important data like information of to disseminate the message by dynamically utilizing the
driving tracks, fuel alerts, heat, oil and water levels for future available capabilities of a vehicle. A stepwise description
use. For emergency case, fog server calls rescue and petro- of algorithm is presented as follows; When accident is
leum missions for medication road clearance respectively. occurred, then autoVehFlag and inetFlag is activated. It
Fog server response quickly as these servers are deployed means V2I sensors and internet is available in order to
locally on edges, ultimately help to avoid more accidents. transmit messages, then select vehicle type as Iv. If only
Second type is smart vehicle (Sv) that approaches fog server internet is available and there is no v2I sensors, then select
through wireless sensors which exchange required data to vehicle as Sv. Finally, if RSUFlag is activated then select
vehicle embedded application or driver’s smart phone. Bv, it shows absence of internet. Selection of vehicles will
Third one is basic vehicle (Bv) which shares the sensed be done on the basis of priority as illustrated in steps 1–7.
data via V2I communication where RSUs are involved to If the vehicle is Iv or Sv and internet is available then send
approach base station and then communicate with fog server message to fog server in order to approach rescue team
because of unavailability of internet. Moreover, in case of for medication and remove accidental vehicles from road,
the absence of near RSU, both type of vehicles can perform it helps to avoid more accidents. If the vehicle is Bv then
V2V communication. Basically purpose of this architecture severe condition flag are activated and NeighborCount
is reliable communication, if road side units are not available start working in order to count nearby vehicles having Iv
in that area then the message can be transfer through other and Sv. Then send message to fog server through Iv or
vehicles by using V2V concepts. Suppose there is no RSU in Sv and deactivate SevConditionFlag. In steps 19–28, it
the range of normal vehicle then it gives message to nearby checks severe condition flag if it is activated then Sent-
vehicles in one hop only at once its message is reached to MsgFlag will be turned on and start finding the vehicle
fog server and it gets back response back from fog server. to whom RSU is nearest in one-hop. In this way message
When the message is reached to fog server it contacts to that will reach to RSU through V2V communication. When
Av with in no time in order to stop messaging. Fog server messaged reached to its destination then deactivate the
replies in the same moment to that smart vehicle, in order SentMsgFlag. After that, SentMsgFlag is activated and
to inform Av. Fog server considered as local servers so that then message is transmitted to nearby vehicles unless it
they can compute issue and contact with in no time. Our Fog reaches at the server.
assisted architecture gives better communications with less

13
Congestion avoidance through fog computing in internet of vehicles

the fog server because any neighboring vehicle can help to


Algorithm 1: Algorithm for Message Dissemination (AMD)
connect or temporarily sustain the data to exchange with fog
// Accident occurred server. A nearby vehicle which is Iv or Sv or Bv can play a
1. If autoVehFlag is 1 AND inetFlag is 1 then vital role to exchange messages towards the destination node
2. Set VTYPE = IV
until fog server is accessible again. In case of Ff  , vehicle has
3. Else if inetFlag is 1
4. Set VTYPE = SV to change the track and find some other next-hop reliable
5. Else if RSUFlag is 1 then vehicle in its range to identify a new fog server in upcoming
6. Set VTYPE = BV region. In this scenario, the tasks that are non-delay toler-
7. End if ant may be discarded by the appropriate receivers. In this
8. If VTYPE equals IV OR SV AND inetFlag is 1 then
9. Send_Message (Pkt) to fog server scenario, all the requests that will expire shortly should be
10. Else prioritized first.
11. Set SevConditionFlag to ON
12. For i =1 to NeighborCount
13. If NeighborList[i]equals IV OR SV Algorithm 2: Fog Server Failure Detection and Recovery
14. Send_Message (Pkt) to fog server Set Ftype as Number // Failure Type
15. Set SevConditionFlag to OFF Set as Bool //Near-distance Failure
16. Break loop Set as Bool //Far-distance Failure
17. End If
18. End For 1. Procedure failureDetection Begin
19. If SevConditionFlag is ON then 2. Declaration NdF = Node distance from fog server;
20. Set SentMsgFlag to ON
Dm= Maximum distance;
21. For k =1 to NeighborCount 3. Initialization Initial Flag IF=0, Report Flag RF=0,
22. If NeighborList[k]contains RSU at one hop then Time Flag TF=0; Edge Node EN=1
23. Send_Message (Pkt) to RSU 4. Input accident report with GPS to EN. //Av report to fog
24. Set SentMsgFlag to OFF server.
25. Break loop 5. While EN is 0 do //Edge node is not in working
26. End If position
27. End For 6. If (NdF <= Dm) then
28. End if 7. Ftype=
29. If SentMsgFlag is ON then 9. Else if (NdF > Dm)
30. Send_Message (Pkt) to Neighbor ->Next_Vehicle 10. Ftype=
31. End if 11. End if
32. End if 12. End while
13. End Procedure

14. Procedure Recovery Begin


15. If (Ftype= ) then
16. Wait for connection recovery with fog server
4.2 Fog server failure detection and recovery 17. Else if (Ftype= ) then
18. Select next-hop reliable vehicle towards fog server
19. End if
Proposed scheme is an opportunity for reliable communica-
20. End Procedure
tion by avoiding message congestion. ­E2MD based on fog
server which gives concept of local server and helps data
computation locally, in its parallel there is a chance of server
failure. Failure of server may lead to distraction in communi-
cation. Fog server failure detection and recovery is a point to
5 Results and analysis
consider as explored in algorithm 2. A stepwise description
for algorithm 2 is explored as follows; initially, we declare
We have performed extensive simulation in NS2.35 for
and initialize variables. In case of accident reporting, Av
verification of the proposed scheme. We have used tool
informs to fog server in its range. When Av did not get
command language (TCL) for vehicles’ deployment, road
response back it means local server is not working properly.
segmentation, and clustering along with message ini-
Now it is time to detect type of failure. If distance of vehicle
tiation. In the similar vein, practical extraction and report
from fog server is less than the range of maximum distance
language (PERL) is used for result extraction from trace
then it is temporary and near-distance failure represented as
files. Moreover, the send and receive functions are imple-
Nf  . While if vehicle has greater distance from fog server it
mented using C language where proposed algorithms are
will be considered as far-distance failure Ff as illustrated in
also implemented. Each simulation scenario is executed 10
steps 4–13. After that, it needs to recover the fog server, if
times to get the trace files and then results are extracted by
failure is Nf then need to wait for connection recovery with

13
S. Yaqoob et al.

Table 2  Simulation parameters AV = Abnormal Vehicle


Message
Parameters Values
Messaging AV Drop
Repetition

Deployment region 100 × 2000 m


Tx radius for GH 400 m 3
18 10 6
Tx power at node 0.819 µJ
Receiving POWER 0.049 µJ 4
11
Transmission range 30 m 12 9 7 5
Queue type Queue/DropTail/PriQue
Antenna type Omni antenna
Routing vehicles Iv, Sv and Bv
Agent trace On
Router trace On
Vehicles initiated messages 20–50
Server

Fig. 5  Un-necessary repetition of messages in IoVs


taking average using PERL scripts. Two types of vehicles
including LTV and HTV are deployed where former has less
energy consumption and transmission range as compared to smooth communication is possible when there is less delay
later. Transmission range of LTV and HTV is set to 30 m as in packet delivery and less chance of packet drop.
per specification of radar sensors by NXP (Editors 2016).
Deployment region for nodes is set to 100 × 2000 m where 5.2 Communication overhead during messaging
100 m represents width of road and nearby area whereas
2000 m represents the road distance. A list of simulation Communication overhead means transmission overhead dur-
parameters is presented in Table 2. A real scenario is taken ing messaging. Transmission overhead refers to total no. of
where all types of vehicles are moving like Iv, Sv and Bv. messages in network. Total number of packets in a network
as per speed limits. Communication period for EMs among is also known as total energy consumption. This parameter
vehicles is set to 3 min during which vehicles are supposed (communication overhead) also compares proposed ­E2MD
to send and receive bulk of messages. We have simulated for with existing approaches in order to highlight the achieve-
04 communication periods with congestion scenario. During ment of this paper. MT is sum of all messaging cost as shown
simulation, nodes are configured in three types including in Eq. (9) where M𝛼 represents messaging cost due to direct
vehicles, GH and fog server. We have simulated proposed transmission by Av to passing by vehicles. M𝛽 represents
E2 MD scheme and compared its performance with AFCS messaging cost due to messaging from passing by vehicles
(Chang et al. 2015) and RMFF (Sattar et al. 2018). that forward the EMs to their neighbors. M𝛾 represents the
cost of indirect messaging to vehicles in opposite lane.and
5.1 Performance metrics M𝜔 represents their forwarding cost:

The number of connections between vehicles can be calcu- M T = M 𝛼 + M 𝛽 + M𝛾 + M 𝜔 . (9)


lated as |N|×|N−1|
2
 = 10(10−1)
2
 = 90
2
 = 45 when there are 10 vehi- Figure 6 elucidates the communication overhead in terms
cles near Av as shown in Fig. 5. In this case, total number of total messages. The results clearly demonstrates that
of vehicles in one-hop are represented as N and established E2 MD incurs far less communication overhead compared to
links are represented as L for Graph (N, L). All estab- AFCS and RMFF. This is mainly because there is no chance
lished connection links causes packet drop and congestion. of repeated data due to reliable vehicle selection for report-
Repetition
∑n of messages also helps to arise congestion as ing fog server and upcoming traffic.
(DM+RM)
L × i=1 100 where L represents established connection The messaging overhead increases with the increased
links, DM represents dropped messages and RM is repeated number of neighbors and more message sharing by mul-
messages. tiple sources via internet. It is worth noting that E2 MDsig-
Below mentioned metrics are considered to assess the nificantly reduces the communication overhead in congested
performance of proposed E ­ 2MD. These metrics are con- network as well. On the other hand, RMFF incurs highest
sidered to highlight the strength of proposed scheme. Less messaging overhead for the diffusion of messages due to
communication overhead, delay and packet loss ratio makes restrictive flooding. In case of 5 vehicles in neighborhood
any network reliable, while high packet delivery ratio and then in existing scheme AFCS, Av informs head node in one
throughput allows smooth communication in IoVs. In fact, message then head node informs other 4 vehicles in group so

13
Congestion avoidance through fog computing in internet of vehicles

Fig. 7  Packet delivery ratio for emergency messages


Fig. 6  Communication overhead during messaging

number of messages like 20, 30, 40 and 50. Number of mes-


M𝛼 = 5 . These 4 vehicles inform to 2-hop away neighboring sages depends upon no. of vehicles for example dense area has
vehicles means M𝛽 = 5 × 4 = 20 messages are exchanged. many vehicles so that no. of messages is large. When there is
After that, messages are also exchanged to vehicles in the sparse environment with less messages, then automatically it
opposite lane so M𝛾 = 5 when there are 5 vehicles in neigh- causes less congestion. Less congestion leads to good packet
borhood and then transmitted to further neighbors means delivery. Basically there is less chance of packet loss in lim-
M𝜔 = 5 × 5 = 25 messages. Finally, minimum MT = 45 are ited congestion. Dense environment has high congestion than
exchanged, it can be worst when more number of vehicles sparse environment with limited messages. In AFCS messages
are present where various vehicles transmitted 20, 30, 40 are forwarded through nearby vehicles means each vehicle
and 50 messages in different scenarios. Similarly, number transmits same messages in its range which causes conges-
of messages are transmitted in case of RMFF but vehicles tion. RMFF schemes creates flood of repeated messages which
drop repeated message but transmission cost is consumed. results in high congestion. As we have seen high congestion
Our proposed scheme sends message to reliable vehicle and leads to less PDR. However, proposed scheme handle repeated
neighbors in one hop means M𝛼 = 5 messages are transmit- messages by introducing three distinct type of vehicles and
ted to vehicles in direct range. Messages are not forwarded using fog server. The results clearly demonstrates that E2 MD
and hence also not reach till opposite lane. Meanwhile server incurs far high PDR compared to AFCS and RMFF. For the
inform other upcoming traffic and hence dominates as com- scenario where 30 vehicles initiated the messages, the PDR is
pared to counterparts. 44% and 53% for RMFF and AFCS respectively whereas our
proposed ­E2MD dominates by successfully delivering 96%.
5.3 Packet delivery ratio Our ­E2MD scheme achieves 118%, 81% better PDR in contrast
with RMFF and AFCS respectively:
Packet delivery ratio (PDR) is a measure of received pack- ∑
(Rpackets )
ets at destination against sent packets from source. Proposed PDR = × 100, (10)
scheme E2 MD gives good packet delivery ratio than existing Sent Packets
schemes due to fog server. Fog server helps to reduce conges- ∑
tion which causes due to repetition of messages. E2 MD selects (Rpackets )
reliable vehicle to pass messages to fog server. Fog server is PDR(RMFF) = ∑ � � × 100. (11)
Dropped + RPackets
responsible to inform rescue team and upcoming traffic. AFCS
and RMFF inform server and upcoming traffic through neigh-
bor vehicles that causes congestion due to repetition of mes-
sages. To calculate PDR, proposed scheme and AFCS divides 5.4 Packet loss ratio
received packets by sent packets while for RMFF received
packets divided by sum of received and dropped packets due Packet loss ratio (PLR) is a measure of dropped messages
to flooding as shown in Eqs. 10 and 11 respectively. Figure 7 during packets transmission from source to destination.
elucidates the PDR in terms of message transmission among Proposed scheme E2 MD provides less packet loss ratio as
number of neighbor vehicles successfully. This research compared to existing schemes AFCS and RMFF. In existing
considered real scenario where vehicles transmits different

13
S. Yaqoob et al.

schemes repeated messages causes congestion where mes- loaders can travel. Range of LTV vehicles must be smaller
sage congestion leads to packet loss. Packet loss is very dan- than HTV like Rltv ≤ Rhtv . Acceleration of current {vehicles
( )2
gerous in the case of emergency messages, if EMs cannot should not exceed than the limits as Acc ≤ k Vmx
deliver in time it may results in uncertain deaths. E2 MD ( )2 }
− VC  . Figure 9 explores the average delay during mes-
provides less PLR due to intelligent fog based scenario, hav-
ing three types of vehicles. This proposed scenario helps saging transmission from Av to the server. Average delay for
to reduce the rate of congestion due to less no. of repeated proposed scheme and AFCS is calculated as shown in Eq. 12
messages. Figure 8 elucidates the PLR in terms of message whereas average delay for RMFF is given in Eq. 13:
loss during packet transmission. To calculate the PLR need ∑
(End − to − End delay)
to subtract received messages from sent packets. In fact, Delayavg = ∑� � , (12)
dropped messages helps to calculate the average of packet RPackets
loss. This research considered real scenario having different
number of messages like 20, 30, 40 and 50 sent from source n
to destination. Results vary according to transmitted num- Delayavg =
2r
. (13)
ber of messages as small number of packets gives less PLR
than large number of transmitted packets. The results clearly
demonstrates that E2 MD incurs far less PLR compared to We explore different formula for delay calculation of
AFCS and RMFF. For the scenario where 30 vehicles initi- RMFF scheme where sent packets n are divided by 2 multi-
ated the messages, the packet loss ratio is 998 and 105 for ples of received packets 2r. In different cases when there are
RMFF and AFCS respectively whereas our proposed E ­ 2MD distinct neighbors across the Av then average delay vary in
dominates by having less PLR OF 4 percent. Our E ­ 2MD seconds and milliseconds. For the case when 30 vehicles ini-
scheme achieves 99%, 96% less PLR over RMFF and AFCS tiated messages, the delay is 0.018 and 0.007 for RMFF and
respectively. AFCS respectively whereas our proposed ­E2MD dominates
by having less PLR of 4 0.005. Our ­E2MD scheme achieves
5.5 Message transmission delay 72% and 28% less delay over RMFF and AFCS respectively.
Our scheme dominates by consuming less delay compared
Average delay is the average time from source to destination to AFCS and RMFF.
for n number of messages. With respect to real scenario that
have following limitations Vmn ≤ Vc ≤ Vmx where 5.6 Throughput
RL = 150mph ≥ VC ≥ 100mph   , ML = 100mph ≥ VC ≥
70mph and LL = 90mph ≥ VC ≥ 10mph . According to these Throughput is a measure of received packets at destina-
equations speed of current vehicles must be in limit and tion per unit time. It helps to analyze the performance of
according to rules, neither maximum nor minimum. In right network. Figure 10 explores the throughput in terms of
lane LTV traffic having small cars with fast speed can travel. message delivery in given time. This research considered
In medium lane vehicles can travel with normal speed while real scenario having different number of messages like 20,
in left lane HTV traffic with big vehicles having trucks and 30, 40 and 50 messages sent from source to destination.

Fig. 8  Packet loss ratio during message dissemination Fig. 9  Average delay for message dissemination

13
Congestion avoidance through fog computing in internet of vehicles

Fig. 11  Packet loss during fog server failure


Fig. 10  Throughput for the Emergency Message Initiation from Vehi-
cles

6 Conclusion

Average throughput is calculated by multiplying the IoVs is a network of vehicles used for reliable communica-
answers obtained after dividing received packets and tion in transportation system. Basically IoVs based on thou-
packet size with average delay and 1000 respectively as sands of vehicles for specific purposes. Message congestion
shown in Eq. 14. The results clearly demonstrates that is a major challenge and essential hurdle during communi-
E2 MD incurs good throughput compared to AFCS: cation in IoVs. Basically message congestion occurs when
same messages transmit repeatedly. This repeated messages
RPackets Psize leads to packet drop. Due to packet drop, information and
ThroughputAvg = × . (14)
Delayavg 1000 EMs are unable to deliver timely, resulted into more acci-
dents. To address this problem, this paper Proposed a E2 MD
scheme for efficient messages delivery and to reduce the
5.7 Fog server failure evaluation congestion. Proposed work is based on fog computing as it
gives the sense of local servers for calculation and informa-
During fog server failure, the vehicles may be either at tion distribution. This scheme is not only useful for smart
Nf or Ff conditions as per time quanta and number of vehicles but also for normal vehicles where internet is not
hops covered in the direction of fog server. Time quanta available. Therefore, it helps to prevent more accidents in
is represented as ∆t and total time to reach packets at fog real scenario our proposed solution also works in scenario
server is T. In case of time quanta ∆t = T − 1, the mes- where server is not in the range of vehicle. In this work, we
sages are near to fog server and vehicles can wait for the have presented several messaging schemes by focusing on
fog server’s availability in a limited threshold time. In congestion avoidance and control based schemes as well. A
this scenario, priority is given to those messages which taxonomy of related schemes is maintained that subdivided
are near to expire in terms of expected delay to reach at into static and dynamic schemes. Proposed Fog assisted IoV
fog server. Figure 11 elucidates that most of the messages architecture is well suitable for efficient messaging. This
were delivered to fog server as per slightly less PDR and paper explores the role of smart vehicles that also contain
meagerly more delay but within average delay constraints. vehicle synchronized cellphones and internet facility in addi-
Results illustrates that for 40 messages initiated by vehi- tion to RSUs. It also explores the level of various metrics
cles when fog server is failed at ∆t = T − 1, then 24, 22 and including delay, bandwidth and transmission range that are
4 messages are lost in case of AFCS, RMFF and ­E2MD mandatory to gain optimal performance. We have simulated
respectively. More packets are lost during fog server fail- proposed E2 MD scheme and compared its performance with
ure in AFCS and RMFF, because the PDR is already not AFCS and RMFF. Simulation results show dynamic and reli-
good due to dynamic mobility of vehicles failure to meet able message congestion avoidance schemes as well as level
delay threshold. of various metrics like PDR, PLR, delay, bandwidth, trans-
mission overhead and packet loss during fog server failure.
E2 MD improved the message delivery cost by 108% than
AFCS and RMFF while decrease messages overhead cost
by 73% and 98% than AFCS and RMFF respectively. In

13
S. Yaqoob et al.

future we plan to investigate the impact on variations of Julio A, Sanguesa M, Fogue P, Garrido F, Martinez J, Cano CT (2016)
queue size interface during congestion due to emergency A survey and comparative study of broadcast warning message
dissemination schemes for VANETs. Mob Inf Syst 2016:1–19
message dissemination. Kang H, Yoo H, Kim D, Chung Y-s (2017) CANCORE: Context-Aware
Network COded REpetition for VANETs. IEEE Access 5:1–9
Acknowledgements  This work is partially supported by the Deanship Khabazian M, Aissa S, Mehmet-Ali M (2011) Performance modeling
of Scientific Research, King Saud University through research group of message dissemination in vehicular ad hoc networks with prior-
project number RG-1439-036. ity. IEEE J Sel Areas Commun 29(1):61–71
Kukreja P, Sharma D (2016) A detail review on cloud, fog and dew
computing. Int J Sci Eng Technol Res (IJSETR) 5(5):1–9
Kumar R, Dave M (2013) A framework for handling local broadcast
References storm using probabilistic data aggregation in VANET. Wirel Pers
Commun 72(1):315–341
Ahmad IN (2017) Characterizing the role of vehicular cloud computing Li M, Zeng K, Lou W (2011) Opportunistic broadcast of event-driven
in road traffic management. Int J Distrib Sens Netw 13(5):1–14 warning messages in vehicular ad hoc networks with lossy links.
Ahmed EA (2017) Bringing computation closer toward the user Comput Netw 55:2443–2464
network: is edge computing the solution? IEEE Commun Mag Lin D, Tang Y, Labeau F, Yao Y, Imran M, Athanasios V, Vasilakos
55:138–144 (2017) Internet of vehicles for e-health applications: a potential
Bihari B, Dubey N, Chauhan P, Kumar (2010) A survey on data dis- game for optimal network capacity. IEEE Syst J 11(3):1888–1896
semination techniques used in VANETs. Int J Comput Appl Merino R, Vaquero LM (2014) Finding your way in the fog: towards a
10(7):5–10 comprehensive defininition of fog computing. SIGCOMM Com-
Chang B-J, Liang Y-H, Huang Y-D (2015) Adaptive message forward- put Commun Rev 44(5):27–32
ing for avoiding broadcast storm and guaranteeing delay in active Munir B, Asif M, Wagan A, Hasbullah H (2010) Efficient congestion
safe driving VANET. Wirel Netw 21(3):739–756 control in VANET for safety messaging. In: Information technol-
Chang B-J, Liang Y-H, Huang Y-D (2016) Efficient emergency for- ogy symposium (ITSim), vol 2, pp 654–659
warding to prevent message broadcasting storm in mobile society Munir A, Laskar MTR, Sakhawat M (2018) A localized fault toler-
via vehicle-to-X communications for 5G LTE-V. In: International ant load balancing algorithm for RFID systems. J Ambient Intell
computer symposium (ICS), Chiayi, Taiwan, pp 479–484 Humaniz Comput 2018(1):1–13
Chaqfeh M, Lakas A, Jawhar I (2014) A survey on data dissemination Panichpapiboon S, Pattara-atikom W (2012) A review of information
in vehicular ad hoc networks. Veh Commun 1(1):214–225 dissemination protocols for vehicular ad hoc networks. IEEE
Chiang M, Zhang T (2016) Fog and IoT: an overview of research Commun Surv Tutor 14(3):784–798
opportunities. IEEE Internet Things J 3(6):1–11 Park S, Yoo Y (2017) Network intelligence based on network state
Chuang M-C, Chang Chen M (2013) DEEP: density-aware emer- information for connected vehicles utilizing fog computing.
gency message extension protocol for VANETs. IEEE Access Mobile Information Systems, Cairo, pp 1–9
5(10):3504–3512 Paul A, Daniel A, Ahmad A, Rho S (2015) Cooperative cognitive intel-
Dandala TT, Krishnamurthy V, Alwan R (2017) Internet of vehicles ligence for internet of vehicles. IEEE Syst J 11(3):1–10
(IoV) for traffic management. In: International conference on com- Roma P, Gayatri J (2016) Traffic congestion detection and management
puter, communication and signal processing (ICCCSP), Chennai, using VANET. Int J Adv Res Innov Ideas Educ 2(3):1274–1280
India, pp 1–15 Sattar S, Qureshi KH, Saleem M, Mumtaz S, Rodriguez J (2018) Reli-
Djahel S, Ghamri-Doudane Y (2012) A robust congestion control ability and energy-efficiency analysis of safety message broadcast
scheme for fast and reliable dissemination of safety messages in VANETs. Comput Commun 119:118–126
in VANETs. In: IEEE wireless communications and networking Schwarz S, Rup M (2016) Society in motion: challenges for LTE and
conference: mobile and wireless networks, France, pp 2264–2269 beyond mobile communications. IEEE Commun Mag 54(5):76–83
Editors E (2016). Radar sensing for driverless vehicles. (Digi-Key). Sharma V, Lim JD, Kim JN, You I (2017) SACA: self-aware com-
https​://www.digik​ey.com/en/artic​les/techz​one/2016/nov/radar​ munication architecture for IoT using mobile fog servers. Mob
-sensi​ng-for-drive​rless​-vehic​les. Accessed 24 Nov 2018 Inf Syst 2017:1–17
Fangchun Y, Shangguang W, Jinglin L, Zhihan L, Qibo S (2014) An Shumayla Y, Ullah A, Akbar M, Imran M, Guizani (2018) Fog-assisted
overview of internet of vehicles. China Commun 11(10):1–15 congestion avoidance scheme for internet of vehicles. In: Interna-
Gianfranco EM, Veniero M, Trombetta A, Sacco M, Clemente S (2018) tional wireless communication and mobile computing (IWCMC).
Semantic based events signaling for AAL systems. J Ambient IEEE, Cyprus, pp 618–622
Intell Humaniz Comput 9:1311–1325 Sindhu G, Mittal P (2016) A novel model based on group controlled
Gupta BB, Yamaguchi DP (2018) Deep learning models for human observation for DDOS attack detection and prevention in VANET.
centered computing in fog and mobile edge networks. J Ambi- Indian J Sci Technol 9(36):1–6
ent Intell Humaniz Comput. https ​ : //doi.org/10.1007/s1265​ Sushama WP, Choudhary S (2017) Data dissemination in software
2-018-0919-8 defined vehicular ad hoc network: a review. Int Res J Eng Technol
Gutiérrez-Reina D, Sharma V, You I, Toral S (2018) Dissimilarity (IRJET) 4(1):1–4
metric based on local neighboring information and genetic pro- Syed A-H, Muddesar I, Atif S, Imran R, Hassan R (2017) An efficient
gramming for data dissemination in vehicular ad hoc networks channel access scheme for vehicular ad hoc networks. Hindawi
(VANETs). Sensors 18(7):1–18 Mob Inf Syst 2017:1–11
Hassan N, Gillani S, Ahmed E, Yaqoob I, Imran M (2018) The role Ucar S, Ergen S, Coleri, Ozkasap O (2016) Multihop-cluster-
of edge computing in internet of things. IEEE Commun Mag based IEEE 802.11p and LTE hybrid architecture for VANET
56(11):110–115 safety message dissemination. IEEE Trans Veh Technol
Hou X, Li Y, Wu D, Jin D, Chen S (JUNE 2016) Vehicular fog comput- 65(4):2621–2636
ing: a viewpoint of vehicles as the infrastructures. IEEE Trans Veh Ullah A, Yaqoob S, Imran M, Ning H (2019) Emergency message dis-
Technol 65(6):3860–3873 semination schemes based on congestion avoidance in VANET
and vehicular FoG computing. IEEE Access 7:1570–1585

13
Congestion avoidance through fog computing in internet of vehicles

Wan J, Liu J, Shao Z, Vasilakos A, Imran M, Zhou K (2016) Mobile Yaqoob IA (2016) Mobile ad hoc cloud: a survey. Wirel Commun Mob
crowd sensing for traffic prediction in internet of vehicles. IEEE Comput 16(16):2572–2589
Sens 16(1):1–15 Yi S, Li C, Li Q (2015) A survey of fog computing: concepts, applica-
Xia S, Cheng X, Yang L, Zhang R, Jiao B (2014) Data dissemination tions and issues. In: ACM proceedings of the 2015 workshop on
in VANETs: a scheduling approach. IEEE Trans Intell Transp mobile big data, Hangzhou, pp 37–42
Syst 15(5):2213–2223 Yusof DM, Abu-Bakar K (2011) A review of congestion control algo-
Xiaomin M, Gabe K, Kishor ST (2017) Application-level scheme to rithm for event-driven safety messages in vehicular networks. Int
enhance VANET event-driven multi-hop safety-related services. J Comput Sci Issues 8(5):49–53
In: Workshop on computing, networking and communications Zrar GK, Lloret J, Abu-Bakar K, Sadiq AS, Ben A, Mussa S (2013)
(CNC), Durham, pp 860–864 Beaconing approaches in vehicular ad hoc networks: a survey.
Xu W, Zhou H, Cheng N, Lyu F, Shi W, Chen J, Shen X (2018) Internet Wirel Pers Commun 73(3):885–912
of vehicles in big data era. J Autom Sin 5(1):19–35
Yang X, Liu L, Jie, Vaidya NH, Zhao F (2004) A vehicle-to-vehicle Publisher’s Note Springer Nature remains neutral with regard to
communication protocol for cooperative collision warning. In: jurisdictional claims in published maps and institutional affiliations.
International conference on mobile and ubiquitous systems: net-
working and services, Boston, pp 114–123

13

You might also like