Wireless Pers Commun (2017) 96:2039–2075

DOI 10.1007/s11277-017-4286-0

ACO and PSO Algorithms for Developing a New

Communication Model for VANET Applications
in Smart Cities

Yassine Hernafi1 • Mohamed Ben Ahmed1 • Mohammed Bouhorma1

Published online: 13 May 2017

 Springer Science+Business Media New York 2017

Abstract Traffic congestion is a major mobility problem, which generates enormous

economic resources and causes serious problems for the most dense of cities. Intelligent
transportation systems based on vehicular ad-doc network (VANET) communications will
improve many services, expressively, related to transport, security, reliability, management
and including the assistance in the reduction of traffic congestion. In this context, we
propose an intelligent system based, firstly, on a new clustering technique to control and
maintain the stability of routes during inter-vehicular communications, and secondly, a
bio-inspired systematically conducting mobility measurement of agent, intervening in
cases of traffic congestion, in order to find new routes to those offered by the global
positioning system. Obviously, by providing an appropriate route selection process, such
routing concept could be helpful in exchanging control messages to inform the nearest
medics willing to give first aid and routing of another alert to the nearest ambulance. Thus,
in order to deal with these sophisticated optimization techniques, we evaluate in this article
means of simulation experiments. According to that, we have anticipated particular inci-
dents suitable to estimate the pertinence of the proposed system. The analysis and
implementation in VANET will be based on three simulators, especially, SUMO, MOVE
and NS2. The results prove the effectiveness of the approach by reducing fuel consumption
and CO2 emissions along with the rest of pollutant emissions in the case of an incident.

Keywords Smart city  VANET  Clustering  Ant colony optimization

(ACO)  Particle swarm optimization (PSO)  Smart clustering-optimized
link state routing (SC-OLSR)

& Yassine Hernafi

[email protected]
1
LIST Laboratory, Department of Computer Science, Faculty of Science and Technology of
Tangier, Tangier, Morocco

123
2040 Y. Hernafi et al.

1 Introduction

In the context of economic and technological changes caused by globalization in addition

to the constant increase in population. This great increase will drive significant con-
sumption of more energy and need for more space to live. Thus, to limit the load, a new
concept of cities showed up. A smart city use the new information and communication
technologies to facilitate the access to urban services including minimizing the effect of
human activities on the environment [1]. The Smart city as the term would suggest, is a
city that works with sustainable and intelligent ways while incorporating a range of
facilities and services using intelligent devices to ensure efficiency, increase comfort and
quality of life of citizens with a minimum of resources.
In the transport field, the congestion of urban traffic is a serious problem that makes
several cities suffer. The phenomenon is even more alarming with the urban development
of the city, although the improvement marked in living standards of its inhabitants. The
situation is even more aggravated by the old complex architecture inherited, materialized
by narrow streets, in addition to the wrong exploitation of parking sapces and even more
problems.
To define conveniently the approach of smart mobility and intelligent transport, we will
focus on the application of VANET by introducing vehicular networking technologies. The
vehicular networks are characterized, among other things, by high dynamics due to the
rapid mobility of vehicles. Besides, vehicular network for intelligent transportation systems
that refers to the application of new information technologies and transport communica-
tion, can help to improve road safety and traffic efficiency through the exchange of
information between vehicles [2–5]. This type is characterized by a high mobility, in fact,
the complexity of this network resides in the expected difference between the speeds of
vehicles, which changes permanently due to the intense rapid network topology. That
being said, we must adapt the mechanisms to be able to function effectively in this
particular context. The different types of communication, i.e. vehicle to vehicle (V2V)
[2, 3, 6], is established between two or more vehicles. This approach is more or less
adapted to the short range vehicular networks. The second type of communication is
vehicles to infrastructure (V2I), communication is established between the equipment of
the existing network infrastructure such as wireless access points and vehicles available in
the range of equipment. Furthermore, another opportunity can lead to a communication,
namely, the hybrid architecture who can combine both of the previous possibilities, and
can allow the communication either in vehicle-to-infrastructure mode or in vehicle-to-
vehicle mode.
We propose a new model to contribute in the urban smart mobility by developing two
ideas: The first idea is to conceive a stable structure based on clustering in VANET. This
process is carried out by each vehicle that routes its information to all other neighboring
nodes. These vehicles select a cluster head to coordinate the communication among them
and effectively taking into account all the necessary information related to the node
mobility, by introducing an inspiration from social behavior of bird, in order to organize
and designate the cluster head. On the other hand, we propose an approach of urban
navigation using ant colony optimization algorithm. The ant agents are used to find the best
route from source to destination in terms of various parameters, as well as by re-routing the
itinerary delivered by the Global Position System (GPS) navigation in case the congestion
detection is broadcasted by other vehicles.

123
ACO and PSO Algorithms for Developing a New Communication... 2041

Our goal is to conceive a model that will exploit the communication to facilitate
exchanging messages, and discover an ideal route of a vehicle in case of an undesirable
incident and an unplanned event. The approach applies a routing management of the road
incident. Obviously, by providing an appropriate process in selecting itinerary, such
routing could be helpful in exchanging control messages to inform the nearest medics,
which are identified by a tagged label that marks the practitioner of medicine and informs
the ambulances who are close and ready to provide the first aid care and rapid medical
service on time.
However, the concept allows the detection of incidents in general, besides avoiding the
occurrence of similar accidents as well as alerting and involving medics and optimizing its
traffic.
In short, the main points of the contributions of the work are listed below:
• Ensure the fluidity of controlling communications and maintaining stability by
suggesting a clustering algorithm made for a perfect localization and high mobility.
• An inspiration from particle swarm optimization to elect the Cluster Head. With that
being said, modeling the vehicle’s interactions by two sub graphs based on graph
theory to ensure the communication and quality.
• In order to deal with these sophisticated optimizations, we have introduced the
suggested changes in the intern configuration of OLSR routing protocol.
• Avoiding route in case of an incident selected by the cluster head using ant colony
optimization.
• We use a realistic VANET mobility traffic taken from the metropolitan area of Tangier
(Morocco), in order to find and validate the best optimal configurations.
This article is organized as follows. In Sect. 2, we present a large state of the art, in order to
consolidate our literature followed by an appropriate motivation of the approach described
in Sects. 3 and 4. These Sections dissect the proposed mathematical models and algo-
rithms, while Sects. 5 and 6 describe the various simulations including presenting an
interpretation of results. Finally, we conclude the paper.

2 State of the Art

Various works have been conceived in previous literature, derived from the use of clus-
tering and routing algorithms in mobile along with vehicular networks. In this section, we
define the main clustering algorithms proposed for VANET, and the routing algorithms
based on Ant Colony Optimization since our proposed approach makes those two key
concepts harmonize with each other.

2.1 Clustering Algorithms for Vehicular Network

Zhang et al. [7] suggested a multi-hop clustering that represents a relative mobility
between vehicles. The cluster-head is elected according to the smallest aggregate mobility
value in order to construct stable vehicle clusters. To model the relative mobility, the
authors used the packet transmission delay to represent the logical distance between two
vehicles.
In [8], the authors proposed a clustering protocol that is able to maintain the stability of
the vehicular network, by introducing a quality of services based on clustering algorithm
that considers requirements and mobility constraints.

123
2042 Y. Hernafi et al.

The authors [9] suggested a clustering scheme using data clustering technique called affinity
propagation for VANET. The clustering scheme will form clusters with both minimum dis-
tance and minimum relative velocity between each cluster head and its members.
Several clustering algorithms have been presented in previous works, derived from the use
of an adaptive concept of clustering, however, the majority of them designed essentially for
classical ad hoc network and have different constraints that make them insufficient for
vehicular network. Our system focuses on assuring fluidity information for warning message
systems and re-route the itinerary in case of an incident, based on an inspiration from social
behavior of bird, referred to as particle swarm optimization (PSO) [10, 11], to elect the cluster
head by proposing a fitness function for finding the best probability. With an illumination of
the proposed approach, Hernafi et al. [12] suggested a mixture between the two techniques, the
goal is to present a method for compressing data by using aggregation based on the division
into sectors and the implementation of the clustering technique by mixing the entirety in a
dedicated architecture that is designed to optimize the interconnection.
Concerning the cluster head selection, different works have been presented, we can
define, among others, the lowest ID [13], where the selection of the CH is based of lowed
ID. We can also specify the highest degree [14], that selects the CH based on the node
neighboring connectivity. The node with the maximum number of neighbors is elected as
the CH. The problem of these previous algorithms is that in a vehicular network, this
concept of selection is not always the ideal proposition because the movement of the
vehicles is not considered, and not stable due to the nature of the nodes’ movement. The
utility function [15] it performs better than the preceding two approaches. This algorithm is
based on a multiple metric weighing algorithms to determine the most stable CH.

2.2 Graph Theory in Vehicular Network

Several works have been achieved coping with the use of graph theory to model vehicle’s
interactions [16, 17], however, the majority of them focus on the obvious concept of formation
and control. Our system focuses on assuring information fluidity for message warning systems,
and that includes assuring the quality of cluster head selection and transmission.
When we return to the literature, one of the important thematic being proposed, is an
elaboration of a system of communication protocol to propagate warning messages in order
to prevent collisions. Avvin et al. [18] suggested a model to represent the interactions
between multiple vehicles, it aims to provide contextual information for collision warning
applications. Yang et al. [19] proposed a communication protocol for emergency warning
dissemination, which uses a state transition mechanism and limits the forwarding distance
to reduce redundant messages. The aim is to develop a congestion control strategy for an
efficient multi-hop broadcast protocol.
The work of Biswas et al. [20] suggested an intelligent broadcast protocol that uses
implicit acknowledgment that randomly delays reducing the unnecessary rebroadcast. Our
work belongs to a different dimension of research compared to the existing works, the
ultimate goal is to develop a model that will exploit the message dissemination to facilitate
exchanging messages and update the route of a vehicle in case of an incident.

2.3 ACO-Based Routing Algorithms

The proposed routing protocol in [21] uses an adaptive quality of service routing protocol,
which aims to find the best routes from source to target intersection in terms of delay,

123
ACO and PSO Algorithms for Developing a New Communication... 2043

bandwidth and delivery ratio. The presented protocol selects, adaptively, the best routing
paths based on the periodically estimated relaying quality of road segments.
Baras et al. [22] proposed a probabilistic emergent routing algorithm (PERA) based on
the Swarm Intelligence paradigm that exploits the inherent broadcast capability. An active
method, that periodically broadcasts ants’ agent in order to find the best local solution. Our
work belongs to a different ideology of research compared to the existing works, the
ultimate goal is to develop a model that will exploit the message dissemination to facilitate
exchanging messages, and update the route of a vehicle in case of an incident.

3 The Application of VANET for Alerting the Nearest Medics

and Ambulance

Offering an emergency service is an indispensable public service for every city as it

protects citizens’ life, the need for forwarding safety messages to reach the maximum of
destinations and provide them with the information about the incident is so crucial to avoid
risking human lives.
The preeminence of accident management resides in ensuring the dissemination, in case
of any sudden problem, of alert messages and the ability to give apart information to the
first aid in particular to notify ambulance or to look for a specialized doctor in the interest
zone, in order to guarantee timely delivery of medical service for early intervention in
saving lives. A priori, to facilitate these challenges, the ambulances are equipped with
identification system to facilitate tracking. It’s evident that these health-care applications
would benefit greatly from routing information.
By combining V2V and V2I communications, we would be able to perform the model
that disseminates warning messages through the vehicular network. Obviously, the
important information is included in warning message, and nevertheless, the vehicular
networks can allow faster notification on the road propagated using multi-hop communi-
cation with the ambition of reaching all vehicles. Indeed, the model will exploit various
types of communication technology, namely, the dedicated short range communications
(DSRC), the wireless access in vehicular environments (WAVE), and so on. Admitting that
road users management would be with different wireless communications (certainly the
vehicular network, cellular and satellite communications). Additionally, the individual
vehicles are equipped with computing and communication devices, moreover, they can
transmit and receive the information on the vehicular networks. By using these types of
communications, vehicles could be involved in an accident and send alerts with other
important information about the accident, accompanying it with the proposed inspiration
for probability calculation, a bio-inspired agent-based system, intervening in case of
congestion of vehicles to find new routes to those offered by the GPS navigation system.
This combination has triggered our capability to conceive an improved responsiveness of
emergency services. Evidently, such information could be especially helpful in the case of
emergency response operations. For example, an approaching emergency vehicle could
send periodically warning messages to the neighboring vehicles, to inform them about its
current position and speed. In addition to that, any node can detect alert information, by
observing the location of the node after issuing an alert.
Related to the principle exposed earlier, we opted for an exploration of clustering
method that permits a wiser use in the case for emergency operations. Obviously, this
technique could be especially helpful because it is an efficient method of managing various

123
2044 Y. Hernafi et al.

vehicles in VANET. Moreover, clustering can be employed to increase vehicle coordi-

nation, and likewise, decrease the interference with each other. In addition, it can help in
load balancing efficient aggregation of data, and produce a reasonable performance in
mobility metrics.
The idea, already mentioned before, is to ensure communication and information
exchange between vehicles, this being established by routing protocols. That said, the role
of this mechanism is to ensure the communication of information between nodes in the
most efficient manner. Nevertheless, the strategy of routing must consider all the char-
acteristics of vehicular networks, specifically the frequently changing topology, the high
mobility, and other challenges that seem beneficial to offer a relevant strategy that provides
continuous network connectivity. The routing poses crucial networking research chal-
lenges for the VANET, which are characterized by a dynamic environment that makes a
regular change in the network topology, and causes instability in traffic road. The cluster
formation must play an integral role in ensuring mobility to achieve stability, this moti-
vates function by a clustering mechanism with the need for stable traffic. To widely use
optimized link state routing protocol (OLSR) the table driver proactive routing protocol,
we calibrate the use of Hello message and control TC messages in order to change the
information disseminated. Based on the fact that the OLSR has a global topology view, and
would know about a route to the destination.
The work proposes an improved intelligent transportation system, offering the oppor-
tunity to exploit the benefits of smart mobility. Two relevant solutions are considerable, the
first one is to treat an architecture based on clustering algorithm to conceive an appropriate
structure to enable fluidity in the exchange of warning message dissemination, while
winning in terms of metrics such as delay along with speed, in favor of warning the
destination in case there is an unforeseen. The second solution is an inspiration derived
from the ant colony to bring a probability calculation of agents tending to discover another
alternative itinerary based on the shortest route.
We represent a model to form stable clusters and maintain the steadiness during the
communications, especially to bring out the interaction between road vehicles, consisting
of positioning systems and wireless communication to exchange messages using periodic
broadcast to reduce traffic accident. We assume that all the vehicles are equipped with the
Global Positioning System (GPS) system, and sensors to obtain a vehicles position, speed
and acceleration. Also, based on parameters of vehicle’s consumption, and mobility
metrics. The Algorithms 1 and 2 demonstrate the desired inspiration, and Tables 1 and 2
dissect the notation and metrics used.
To elect the cluster head, we were inspired by the social behavior of birds, using particle
swarm optimization [10, 11]. When we observe the perception of the concept, the birds
randomly search for food by following the nearest bird to the objective. In PSO, sets of
solutions are nominated particles, that are initialized arbitrarily, each particle evaluates a
satisfied fitness. The proposed concept applied this context of ideas, the development
works will be as follows, each vehicle based on its current position, velocity and distance
between the current position and the best solution found by its neighbors, a fitness function
will be used to evaluate the quality of the solution.
Due to the significantly high mobility in vehicular networks, ensuring network con-
nection introduces a pertinent communication for exchanging and updating the architec-
ture. Thus, efficient clustering formation should take into consideration major VANET
characteristics. It’s clear that the cluster formation should consider a mixture of classical
qualities of service and a definition of the mobility metrics, among others we find the speed

123
ACO and PSO Algorithms for Developing a New Communication... 2045

Table 1 Notations
Symbol Significance

V Set of vehicles in the roads

FQ Fitness quality of vehicles on path
a Source cluster head
V2 ðiÞ Two-hop away vehicles from vehicles i
CHSet(i) Set of cluster head elected
MPRSet(i) Set of MPRs selected, the cluster gateway
x The ideal position
s, d Basic source and destination vehicles
Hti The transmission of the packets i
Hri The receiving of the packets i
Nr The total number of packets received
fCO2 Total of CO2 emissions
DF Total of fuel consumption
sij Vcib Value of the pheromone on path
Probij Vcib Probability of pheromone for path
quantity(i) Relayed quality collected by probability of the road
v1 ; v2 Positive constants
a1 ; a2 Random numbers
xid The actual position
pgd The global best position by neighborhood

Table 2 Fitness quality metrics

Notations and metric functions

Let i be a vehicle in the traffic road

FQðiÞ = Fitness quality metrics of vehicle i
F ðxi Þ = The best global fitness
N(i): Neighboring vehicles
Velocity(i) = The overall average velocity
Distance(i) = The average absolute distance
Quantity(i) = The quality collected by probability
Delay(i) = The average end to end delay
Vhop(i) =Number of vehicles in the path
Fuel(i) = The estimated fuel consumption
 
ðiÞ
FQðiÞ ¼ N ðiÞ  Distance
VelocityðiÞ
 
F ðxi Þ ¼ FQðiÞ þ v2  a2  ðx  xid Þ þ v2  a2  pgd  xid
qualityij Vcib d2  Dij þ d2  Nij þ d3  Fij
Dij ð1  aÞ  Dij þ a  delayij
 i i
Ht
Delayij ¼ HrNr
Nij ð1  cÞ  Nij þ c  Vhopij
Vhopij =Number of vehicles in the path
Fij ð1  bÞ  Fij þ b  Fuelij
DFuelij ¼ fCO2 DF

123
2046 Y. Hernafi et al.

including the residual distance. We believe that all the definition of these parameters may
not ensure the stability because of these complex probabilities.
We model a cluster head selection algorithm that allows to elect a set of optimal CHs as
well as dividing the network into clusters. The cluster structure is determined for the first
time by the geographic position of the vehicle, then in second place the most stable vehicle
that can act as a CH is a vehicle that has the minimum average distance among the other
vehicles in the cluster, the ideal closest speed to the average speed and the maximum
number of neighboring vehicles.

Algorithm 1: CO2 emissions and fuel consumption calculation

1: Initialization
2: D: Distance of the trajectory traveled by the vehicles
P
3: g for CO2 : unit consumption per kilometer
4: Fe : Emission Factors
5: procedureFuelConsumption
6: for each vehicles (i) in V do
P
7: fCO2 ðiÞ : g for CO2  D  Fe
8: fTotalCO2 ðiÞ : fCO2 ðiÞ=h km vehicle i
9: DFuelij : fTotalCO2 ðiÞ  DF
10: end for
11: end procedure

Algorithm 2: Speed and velocity calculation

1: Initialization
2: D: Distance of the trajectory traveled by the vehicles
3: t: Time of travel
4: Dd : Displacement between two points
5: AvgVelocity ¼ Dtd
6: AvgSpeed ¼ Dt
7: procedureSVRatio
8: for each vehicles (i) in V do
9: VelocityRatioðiÞ ¼ AvgVelocity
AvgSpeed
10: end for
11: end procedure
**Since speed is based on distance and velocity is based on
displacement, these two quantities are effectively the same.

3.1 A Modified Smart Clustering-OSLR

One of the important features to ensure smooth and relevant communication is the famous
principle of routing. Routing protocols can be classified according to several criteria that
depend on different families and classifications.

123
ACO and PSO Algorithms for Developing a New Communication... 2047

We decided to integrate our approach by modifying the strategy of routing protocol to

have a complementary study, we opted to make a comparison between two protocols,
namely the classical optimized link state routing (OLSR) protocol and the modified one,
the smart clustering optimized link state routing (SC-OLSR) protocol, while identifying
performance is needed to derive the benefits of modified protocols.
OLSR is among the proactive protocols [23], with essential strategy, roads are available
when needed. The advantage of this type of protocol is the absence of the prior discovery route.
That being said, the destination is stored in the routing table, protocols keep updating this
table in each node, and with every change of network topology, these routing tables are
updated. It is likely an optimized link state, this means that it is based on the technique of the
shortest path for choosing a road. The topology in this type of protocol experiences regular
changes caused by flooding, OLSR uses the technique of multipoint relays (MRP), the idea is
to reduce flooding of the broadcast, this mechanism consists in ignoring a set of links and direct
neighbors, which are redundant, for the calculation of the shortest path routes.
This protocol has a number of characteristics that make it suitable and exposes its use for
VANETs networks, among others, it reaches an adequate performance in relation to the delay
used to express the transmission of data packets in large networks, adapted to topology
changes, simple and can be applied into different types of systems. It is particularly suitable for
the network with applications that require short transmission delays, just like as the majority of
applications of vehicular networks, and especially exchanging warning message.
Besides the fact that it comes with several promising advantages, proactive protocols
are inapplicable, for the reason that too many signaling messages are required so that the
vision of the topology is updated, indeed the requirement to maintain the routing table for
all possible routes. Such a disadvantage may be insignificant for low-density scenarios.
However, for a congested network, this will trigger an overload control message that could
consume more bandwidth and cause network congestion.
Though, we can bounce back with an intelligent and lucid use. That being said, to
optimize the performance of routing decision, we realized some changes affecting the
internal configuration strategies of the protocol. Thus, calculating an optimal configuration
for the parameters of this protocol is crucial before deploying it, because it could decisively
improve the quality of service, with strong involvement of the expansion of the network’s
data rates and reduces the network load. Afterwards, we will be based on the basic idea of
electing the specialized nodes MPR in OLSR, the function of this particular entity is to
reduce the overhead of flooding messages by minimizing the duplicate transmission within
the same zone. Consider the MPR as a relay gateway, beneficial for ensuring the inter-
action between cluster head and classical members.
We assure basically certain modifications to the OLSR routing protocol, we made an
update of cluster head election with an inspiration from Particle Swarm Optimization and the
MPR selection. A priori, the OLSR contains a simple heuristic, in addition, we altered the path
determination algorithm using Ant Colony Optimization to avoid route in case of an incident.
Due to its proactive nature, OLSR works with a periodic exchange of messages, already
defined, the Hello and TC messages. The idea was to calibrate the quality of service (QoS)
routing extension for OLSR by modifying the selection of the entity and the TC messages to
spread the QoS information throughout the road. As a matter of fact in OLSR, only vehicles
that have symmetric links between them can be neighbors, particularly for exchanging
neighbor information. Considering that the generation of cluster in our architecture is based in
one-way circulation vehicles. Evidently, the cluster selection structure is determined in the
first place by the geographic position of the vehicle, then in the second place the most
stable vehicle that can act as a CH is a vehicle that has the lowest average distance among the

123
2048 Y. Hernafi et al.

other vehicle in the cluster, the ideal closest speed to the average speed and the maximum
number of neighboring vehicles. The candidate vehicle broadcasts periodically a Hello
message containing two constraints of qualities, the presumed ideal position (to accurately
estimate the current geographical location, the Global Positioning System can easily be
deployed on vehicles) and mobility quality metrics. Then each vehicle votes for it neighbor
having the maximum probability and the ideal location. Therefore, the Hello messages are
modified to support the exchange of QoS information, each vehicle sends this type of QoS
information to all its neighbors, the one-hop neighbor in order to apply such heuristic vehicles
to know the two-hop neighbors, the neighbor of their neighbors. To better understand this
routing protocol optimization, the TC messages are similar to those of OLSR but they carry
QoS information associated with links. The available mobility metrics of links are embedded
in TC messages, along with probability calculation of routes.

3.2 Cluster Head Election Algorithm

The improving idea of our work is presented as follows: The candidate vehicle broadcasts
periodically a Hello message containing two constraints of qualities, the presumed ideal
position and mobility metrics. Then, each vehicle vote for the neighbor, that has the maximum
probability value and the ideal location. Let taking for granted that the vehicular networks
represent vehicles as nodes dispersed over a circle of a certain radius R, as depicted in Fig. 1.
Moreover, the Fig. 12 as well as the Sect. 6 define an illustration of the cluster head election
algorithm. All the vehicles move together around the circle in order that they continue to be at
the same distance from each other, whatever was the variation of the angular speed with time.
The cluster head must imperatively be located in the ideal position of sectors. Let O be the
center of the circle and let r be the location of the candidate cluster head. Location of mobile
vehicles can be conveniently described in the polar coordinate system with pole O and polar
axis Oz. The road is divided into M rings, where each ring in the same width mR . All vehicles
along the road are expected to know their geography and can determine the limits of sector.
In particular, by adopting a polar coordinate system which is defined as a two-dimensional
coordinate system, where each point of the plane is fully determined, the coordinate of
every point can be designated by ðr; hÞ, where r is the distance between the point and the
polar (the position of the candidate vehicle) and h is the included angle with the polar axis.
It is interesting to note that the point M is perfectly located if we know the distance OM =
r, that correspond to the radial coordinate as exposed in Fig. 1. The coordinates of the
boundary of an area in each cycle can be represented by (1) and (2), the ideal position of
candidate cluster head in each sector of the ring must be located in the centroid of the
inferior arc of the sector (geometric center), the coordinates are (3).
 
R
ð i  1Þ  ; a ð1Þ
m
 
R
ði  1Þ  ; b ð2Þ
m
where a and b are the angles of the boundary of the sector.
 
R aþb
x ¼ h ði  1Þ  ; T ð3Þ
m 2

123
ACO and PSO Algorithms for Developing a New Communication... 2049

Fig. 1 Cluster head election in the ideal position

where h the position of cluster head source and T the multi-hop transmission delay to send
messages. However, if there is no vehicle located in the ideal position the nearest node of
this position will be the aggregation vehicles in sector. Then, all the others can transmit
their message dissemination to the destination vehicles by one of the routing protocols. The
warning message includes additional information containing the position, the total number
of vehicles on the road, the radius of the network R, the number of cycles m, and the
desirable number of vehicles. The connectivity valuable function is to increase the cov-
erage of CH and MPR’s, which is defined in our context as a cluster gateway, also the
velocity, and and the distance parameters that aim to maintain the the desirable stability of
the traffic. In the same algorithm, the elected cluster head interact with MPR nodes to
ensure a smooth communication. Note that in the internal configuration of routing protocol
we need some changes, some types of modifications are done on the original Hello election
and messages to make adequate job. Along with adding a pertinent flag to signal that a
vehicle has been designated as a CH and another flag in the configuration to denote that a
neighbor has been elected as a CH. The conception of the mechanism is exposed in
Algorithm 3.

Algorithm 3: Election cluster head

1: procedureCH
 
2: for each vehicles vi ; vj in V do
3: get all vehicles coordinates by the network topology
4: randomly initialize the position and velocity of the particles

123
2050 Y. Hernafi et al.

Algorithm 3: Election cluster head

5: broadcast Hello message containing Fitness Quality

6: for all V(i) in the neighbor list do
7: if aV2 ðiÞ [ fig is a cluster head vehicle then
8: if aVs;d nfs; dgthen
9: FQðaÞ : maxfFQð jÞjjV2 ðiÞ [ figg
10: evaluate the global fitness:F ðxi Þ
11: Vote for a as a pertinent election
12: CHSetðiÞ w MPRSetðiÞ
13: MPRSetðiÞ ¼ fag
14: end if
15: end if
16: for each elected candidate has aV do
17: broadcast a feedback
18: end for
19: end procedure

3.3 Road Assumption in Graph Theory

The representation of graph theory to model the VANET communication, designates each
equipped vehicle in correspondence to a node in the graph, the route is modeled as a
directed graph GðV; EÞ that exposes the interactions between vehicles, where V is a set of
nodes representing the vehicles, and E is the directed edges defining reaction between
vehicles, links in the graph are determined by the model based on vehicle-to-vehicle
communications. A cluster head a is coupled with some information package like the
position coordinates, width, and speed. The list of vehicles in a specific area is represented
as a set A ¼ fa1 ; a2 ; . . .; an g the set of vertices V is equivalent to the set of vehicles A.
Given a set of vehicles A, we can construct an interaction graph G by firstly assigning A to
V and generating the set of edges E by following the procedure. Figure 3 represents the
interaction graph for the traffic situation shown in Fig. 2.
For the transfer of a warning message and to compensate the packet loss using multiple
paths from source to destination, we conceive two sub graphs of G, namely, Gðs; aÞ and
Gða; dÞ, where:
GGðs; aÞ  Gða; d Þ ð4Þ

We define a cluster head aVnfs; dg whose goal is to guarantee message transfer from s to
d as soon as possible, the cluster head vehicle creates messages that conform to the
communication protocol. To be more successful, (a) needs to be in V ðs; d Þ, including being
part of all possible paths from s to d in Gðs; d Þ

aVs;d nfs; dg ð5Þ

We use Gðs; aÞ ¼ ðV ðs; aÞ; Eðs; aÞÞ to represent the preliminary transfers of message, that
means all forwarding paths starting at s to a, the cluster head which acts as a relay. V ðs; aÞ
contains all nodes in V that are connected by an edge Eðs; aÞ.

123
ACO and PSO Algorithms for Developing a New Communication... 2051

Fig. 2 Illustration of multi-hop routing for warning messages

Fig. 3 An interactive graph of the traffic situation shown in Fig. 2

Gða; dÞ ¼ ðV ða; d Þ; Eða; dÞÞ represents successful transfers from s to d. Thus V ða; d Þ
and Eða; dÞ contain all nodes and edges that are part of a path from s to d.
To conceive an intelligent system and to ensure the efficiency of alert messages transmission,
we can calculate the number of cluster heads on the road between source and destination.
S ¼ fjfvVnfs; d givSgjg ð6Þ

The presence of the number of Cluster Heads factor procreates the success for transmission
of the warning message on the paths, it aims to provide some intuition of routing protocol’s
decision.
S
PðaVnfs; dgÞ ¼ ð7Þ
jV fs; dgj

4 Bio-inspired Amelioration of GPS Navigation

The GPS [24] is a space-based navigation system consisting of orbiting satellites that assures a
precision and accuracy details of position, speed and time. The main solution for vehicular
network localization is to equip each vehicle node with a receiver, a clever solution since the
GPS act as a product able to receive information constantly being sent by the satellites. The
GPS receiver uses the time of arrival technique (ToA) to estimate its distance and compute the
position. We assume that all the vehicles are equipped with the GPS system, for launching
location to find as quickly as possible the destination and vehicle’s position. Besides, the
current vehicular positioning systems are not providing sufficient positioning accuracy to
support the required reliability of vehicular safety applications. After all, we meet massively
traffic on the roads causing daily congestion problems. Moreover, GPS can consider that
information to suggest recommendations of itineraries to avoid the increase in traffic. The
blotting and expansion in traffic on the planned route are thus detected, reported and displayed,
that said, traffic problems could therefore be circumvented quickly.
The growing impact of congestion causes so many complications, the saturation of
roads will have very serious consequences for the economy in general. Further, related to

123
2052 Y. Hernafi et al.

the environmental impact that accompanies rising traffic jam, the fuel consumption which
generates a proportional increase in CO2 emissions and other air pollutants, including
certainly the real risk in terms of health, by causing the anxiety and stress.
The idea was to address some observations for inspiration to offer a suitable product, the
aim of this work is to design and evaluate a bio-inspired process that take into account
information available in the vehicular network, such as the position and the vehicle’s
speed. Clearly, that information could be especially beneficial to drive the routing deci-
sions, so that they can be adapted to VANET applications.

Algorithm 4: Selection strategy based on ANT colony

1: Initialization
2: Initialization all vehicles and get all coordinates
3: Set an initial value of pheromone on every path
4: CHSetðaÞ :¼ CHSetðdÞ :¼ £
Forward Agent
5: procedureForwardPhase
6: for each CHsource a do
7: set flag of ANT Hello message to 0
8: broadcast Ant hello message
9: for each intermediate vehicle i do
10: insert information into Ant Hello message
11: end for
12: end for
13: end procedure
Backward Agent
14: procedureBackwardPhase
15: for each destination d do
16: set flag on Ant hello message to 1
17: for each path i do
18: calculate Dij
19: calculate average fuel consumption and emissions CO2
20: compute FQðiÞ : quantityðxÞjxi and quantityðxÞ :¼ minfquantityðujuiÞg
21: computepheromoneðiÞ ¼ sij Vcib  QosðiÞ
22: computeProbðiÞ :¼ pheromoneðiÞ=pheromoneð jÞ
23: end for
24: CHSetðd Þ :¼ f xjxjgprobð jÞ : maxf probðujuPÞg
24: send back the Ant hello messages 2 hop away
25: end for
26: end procedure
Final Phase
27: procedureFinal
28: for each source k do
29: CHSetðaÞ :¼ fxjxjgprobð jÞ : maxf probðujuPÞg
30: end for
31: end procedure

123
ACO and PSO Algorithms for Developing a New Communication... 2053

When we usually observe a colony of ants in searching for food in their everyday lives,
we realize that it solves such problems as finding the shortest paths [25, 26]. Ants solve
complex difficulties by relatively simple mechanisms to model. This makes it quite simple
to simulate their behavior by algorithms applying in the case of vehicular networks.
Walking from the nest to the food source, and inversely (which initially mainly done
randomly), ants deposit on the route in passing a fragrant substance called pheromones.
This substance therefore allows creating a chemical trail, which represents a mark for
tractability. In fact, other ants can detect pheromones through sensors on their antennae.
Pheromones have a role of path marker, when the ants choose their routes, they tend to
select the track that carries the highest concentration of pheromones. This allows them to
find their way back to their nests during the return. On the other hand, literally, the other
ant’s can use the odors to find the sources, and when choosing their way, they tend to
choose, in probability, paths anchored by strong pheromone.
Assuming that the ants are going to follow a pheromone trail, at one point, a sudden
appearance of an obstacle causes an unexpected shutdown of the road. The latter one,
which, are in front of this obstacle, must choose a new route. Since no pheromone is
deposited along the obstacle, there are so many ants that tend to look for another way out.
However, since the chosen path is shorter than the other, the ants that use it, will find a
faster starting pheromone trail. Pheromones of these ants will be superimposed along the
road. So, when they will join the original path, the path will be selected twice-impregnated
pheromone than the previous road, where the initial ants have not yet been able to join the
original track (the path is longer).
Ants who come to the obstacle at this moment prefer to follow the new road. The
number of ants passing through the latter will increase, which will further increase the
concentration of pheromones. Additionally, the evaporation of pheromones will be
stronger compared to the old road. It will therefore quickly abandon by cause of less
impregnation, ants spend instantaneously to the shortest one. Our inspiration applies this
idea to the intelligent transportation system.
In our case, it is assumed that the vehicles are equipped with a GPS module for the
initial location and VANET connection system (802.11p). Before launching the strategy of
routing, the user tends to launch location via GPS to find as quickly as possible the
destination, the Fig. 4 presents the concept of our routing decisions, likewise, the Fig. 5
depicts the global probability of the ideal itinerary of the alert system.
The route discovery is realized by broadcasting the message to each vehicle, the goal is
to find the highest pheromone neighbor along with an update of the message. The prob-
ability made of Ant colony, offer a package of performance like a smooth connectivity
related to a permanent reduction of average end-to-end delay, network latency and
increasing the probability of finding the roads as quickly as possible. We evaluate in a first
place the impact of road incidents by anticipating random crashes at a particular case, to
draw a relevant and comprehensive analysis of the evaluation results obtained, then we
apply the intelligent system and routing optimization, to allow vehicles to bypass the
blocked itinerary, including the possibility to alert the nearest medics in the area of interest,
in case of a problem that occurred in their routes.
Deducing from general principle, when a vehicle wants to establish a communication
with another on the vehicular network, normally, the road is between these two users. After
all, if during the communication, we found ourselves surrounded of a risk of an incident or
bottling, according to the agents created, artificial ants tend to apply the shortest path
mechanism. On the basis of the principal strategy, the new route can be deduced by the
probability calculation (the calculation of the intensity of the trace pheromones and

123
2054 Y. Hernafi et al.

Fig. 4 Concept of the proposed routing decisions

Fig. 5 Probability of the ideal itinerary

123
ACO and PSO Algorithms for Developing a New Communication... 2055

practicable time). Figure 6 describes the proposed re-routing strategy. To adapt these
concepts, the main objective of this part is to anticipate accident scenarios to design a
mechanism to ensure updating in real time upon detection of an abnormality. On the other
hand, distinctly, the mechanism of our routing could be especially beneficial in the case of
emergency response operations. The approach is to send periodically the warning messages
to the neighboring vehicle, in addition to find as well as to inform the nearest doctor user in
case of serious problems. In contrast, the routing is based on multi-hop communications
with the aim of reaching all nodes for which the warning may be relevant, in order that
represents an alerting system and a traffic optimization for a preliminary rescue. Never-
theless, this routing concept reduces the amount of times lacks in alerting and increases the
chance of saving lives of the accident victims.
Based on the principle presented by Dorigo [26], the probability of ants to choose a
particular path is given by Eq. (8) and the pheromone evaporation is calculated by formula
(9):
sij
Pij ¼ P ð8Þ
k¼1 nsik

sij ¼ð1  pÞ  sij ð9Þ

Our purpose reflects the principles and qualities to find the ideal paths in terms of delay, in
addition, to reduce the dynamic effect of the network topology on the routing protocol, in
order to avoid a route with an unplanned disruption. Based on the characteristics of
VANET, the desirable path will be the one with lower hops number of vehicles between
the source and destination, on the other hand, the communication will be privileged among
vehicles with a smaller delay between them.

Fig. 6 Flow diagrams illustrating the suggested re-routing strategy

123
2056 Y. Hernafi et al.

Afterwards, we opted to dissect other metrics, the idea was to estimate the lowest
average CO2 emissions and fuel consumption based on vehicle variables that can fre-
quently be obtained from a microscopic traffic simulator or another alternative source like
GPS.
A fitness functions are evaluated for each route, the best route is selected based on
smooth scores, and the evaluation is the sum of weighted cost function, so the route with
the lowest result is the adequate solution. We have the intuition that the highest impact on
the environment of CO2 emissions and fuel consumption in a particular itinerary tends to
predict that we have an unexpected problem in this path. Relatively for the reason that in
case of an incident, they tend to consume more resources for a longer time, the route with
the highest and desirable probability will be the one for the route discovery. Ancillary, this
strategy implements certain qualities to explore, furthermore, updates routing paths in the
best conditions in order to bounce problems in traffic.
Once determined, to interconnect and form a connected network, the cluster heads are
charged to select the optimal MPR as a cluster gateway, in order to manage the classical
cluster member. Consider a case where a vehicle wants to discover an alternative ideal path
during an incident to avoid the actual itinerary. It establishes the interactions between these
two entities. Then, the cluster head source generates more packets of the ant hello message
with a flag equal 0 indicating that these messages will be forwarded to the target cluster
heads destination exploring the network and finds the routing paths.
Each vehicle calculates its quality of service metrics value and inserts it in the message
until reaching the cluster head. Thereafter, this cluster head calculates the quantities of
pheromone for each path. Subsequently, it selects the path having the highest concentration
of pheromone.
Once reached, this cluster head sends back the ant hello messages with a flag 1 indi-
cating that these messages will be returned to the source. The source cluster head receives
the message and selects the optimal path and localization within its cluster and MPRs.
Technically speaking, at each vehicle along the path, the identifier of the vehicle Vi is
added to the header. The response will be gradually released to the next vehicle. If routing
information is available, Vi choose the next vehicle Vj with a probability PVij cib calculated as
follows:
 V d  1 a  1 c
sij cib  Delayij  hopij ðFuelÞx
Probij Vcib ¼ P h ia h ic ð10Þ
ke Nei
Vcib
½sik Vcib d Delay
1
ij
1
hopij ½Fuel x

where d, a, x and c are a control parameters. At time t, the ants are a compromise between
visibility and intensity of the trace of these pheromones on all routes to choose the next
destination j, sij Vcib represents the value of the pheromone of a path on Vj when the request
launches since Vi to Vcib . Nei Vcib is the set of neighbor Vi vehicle. a is the number of hops
influence factor and c is the delay influence factor.
After finding the probability of preferably paths, the cluster head source has several
paths to the destination. The route, which has a higher probability path, is selected for data
transmission.
When the request arrives at Vcib , a response is generated (using an unicast transmission
that depends on routing protocol strategy) to the source S following the corresponding
reverse paths. The requests aim to collect quality road segments, then remove and update
the pheromone. When response reaches Vi (since Vj ), it updates the value of pheromone
sij Vcib in Vi as follows:

123
ACO and PSO Algorithms for Developing a New Communication... 2057

sij Vcib ð1  pÞ  sij Vcib þ p  qualityij Vcib ð11Þ

where p is the regulatory factor of the influence of new measured values of the current
pheromone. qualityij Vcib is relayed quality collected by probability of the road from Vi to
Vcib over Vj .
The idea is to implement a quality mechanism for predicting a particular incident to find
out a path that satisfies certain requirements. The probability in addition to the pheromone
value must have a delay (refers to the time between sending a dissemination message and
its reception) and a hop count as small as possible to calculate an ideal probability to select
the destination route. So, it is desirable to find paths with minimum resources to avoid road
incident.

5 Simulation Parameters

In this section we evaluate our proposed solution by means of simulation experiments.

According to that, in the second part we have anticipated a specific initial event such an
accident to estimate the pertinence of our work when a problem occurred in road traffic.
We evaluate SC-OLSR comparing it with the optimized link state routing (OLSR) and
geographic source routing (GSR). With the eventual objective to dissect the effectiveness of
the proposed routing mechanism conducive to improve significantly the performances com-
pared to both other routing protocols. GSR [27, 28], the Geographic Source Routing is a
position routing protocol, which becomes a practicable routing solution for VANET, espe-
cially, routing in a city environment. The concept in this protocol is that it takes into con-
sideration the road layouts in forwarding packets. In GSR, each vehicle is equipped with a GPS
receiver to manage the selection. We briefly describe the concept, when the source vehicle
tends to elaborate a communication for ensuring the interconnection with the vehicle desti-
nation, the shortest path is based on the road layout position. It should be noticed that the data
packets are routed using greedy geographical forwarding along the path.
The analysis and implementation in VANET will be based on three simulators, to define
performance metrics and simulation parameters. We have used Network Simulator 2 (NS2)
[29] the network simulator of general purpose. In order to define realistic VANETs sim-
ulations, we have used simulation of urban mobility (SUMO) [30] the open source sim-
ulator, refers to urban mobility. It allows the user to build a custom route topology, and
also it helps to import different card formats, generally speaking, a free world map. To
ensure the interactions between those tools, we worked on the mobility model generator for
vehicular networks (MOVE) [31] to generate a model of mobility within the VANET
simulations and to ensure the relay of simulators. This tool is developed on the upper layer
of the open-source micro traffic simulation software.
Beneficial to evaluate the quality and fitness of our approach, we have defined important
quality of service metrics, namely, the packet delivery ratio (PDR), throughput, average
End-to-end Delay, and to calculate CO2 emissions and fuel consumption, we use a sta-
tistical model namely the EMIT model integrated in SUMO. The VANET instance defined
in this work, vehicles are inserted in roads of the highway at a constant rate, and all the
main parameters of the simulation are summarized in Table 3. For mobility model, we ran
experiments in networks with different node densities, and the vehicles are traveling in a
3 km2 from a real traffic network of the city of Tangier (Morocco), which clearly include
various intersections of urban roads, among others, the traffic lights.

123
2058 Y. Hernafi et al.

Table 3 Simulation parameters

Parameter Value

Simulators MOVE/ NS-2/SUMO

Routing protocols SC-OLSR/OLSR/GSR
Packet size 512 bytes
Mac layer 802.11p
Radio propagation Two ray ground model
Minimum speed 60 km/h
Maximum speed 120 km/h
Simulation area 3000  1000m2
CBR rate 1–5 packet(s)/s
Simulation time 1000 s
Traffic type UDP (CBR)
Map type Open Street Map (OSM)
Confidence interval 95%
Number of vehicles 100–1000
1000–2000–3000–6000
Movement type Random
Transmission time From 10 to 950 s (1000 s)

Nevertheless, we applied a realistic scenario for traffic simulations taken from the Open
Street Map as described in Fig. 7, by considering a road network section of city of Tangier
in Morocco and taking into account real parameters of traffic. As a next step, the Figs. 8, 9
and 10 define respectively the procedure to generate a real traffic simulation, since the
capabilities to import road networks from Open Street Map until the configuration and
visualization of the simulations.
On roads, an unexpected incident is often used more broadly to refer to any problem that
causes complications in traffic conforming to collision, roadwork, and weather conditions
potentially leading to a slow speed and congestion, along with other common issues. To
anticipate such an event in SUMO, we can arrange some option, among others, by
adjusting the possibility for the traffic stop over a period. In addition, defining other
alternatives, through handling the traffic lights so that remains red for a longer period than
the regular duration, furthermore, impeding traffic by driving too slowly, however, by
arranging speed state. We opted to adopt the first option for stopping vehicles, the most
profitable solution. Certainly, to make this challenge we need all the essential information,
namely, points and track identifiers, necessary to simulate a traffic jam.
As expected, among the importance remedies made for accident prevention is antici-
pation. To illustrate the simulator functionality, a simplified algorithm is also described,
including the desirable changes. The pseudo-code triggers a stop of a vehicle on a specific
track for a period, as a complement, the Fig. 11 shows a complete cessation of all cars.
Deducing from general principle, in simulation of continuous road traffic in common,
SUMO in particular, when a vehicle waits longer that 300 s, the car will be automatically
teleported by default. This means that when a vehicle recognizes a stop in our case to
simulate a break down, it will be stopped for a period, which is less than what is defined in
advance, then they moved automatically. To overcome this problem, the time was
increased imperatively, nonetheless, the main objective is to anticipate accident scenarios
in order to achieve the rating requirements in furtherance of a comprehensive study.

123
ACO and PSO Algorithms for Developing a New Communication... 2059

Fig. 7 Generalization of a real card of Tangier using Open Street Map

Fig. 8 Import the map of Tangier using Open Street Map by SUMO

Including the effectiveness evaluation of the proposed models, by designing a mechanism

able to provide an update in real time upon detection of an abnormality. Our ultimate
purpose is to deploy a model that will update the route of a vehicle accident.
Table 4 defines all the statistics based on the results of the simulation of Tangier city,
these outputs bring many of the ideas about the simulation states. As illustrated in the table,
we can interpret that the average mobility metrics of vehicles are significantly suggestible,
the averages are heavily depending on the number of vehicles, and the measures tend to
rise when the traffic increases.

123
2060 Y. Hernafi et al.

Fig. 9 Visualization of the real world of Tangier by SUMO

Fig. 10 Vehicle movements visualization of traffic roads

123
ACO and PSO Algorithms for Developing a New Communication... 2061

Pseudo-code: Anticipate accidents

hrouteid ¼ ‘‘100 edges ¼ ‘‘d04l23u14r12l45d2500 =i

hvehicledepart ¼ ‘‘100 id ¼ ‘‘f 000 route ¼ ‘‘100 =i

stoplane ¼ ‘‘fo000 endPos ¼ ‘‘5000 duration ¼ ‘‘60000 =
h=vehiclei
hrouteid ¼ ‘‘100 edges ¼ ‘‘d04l23u14r12l45d2500 =i
hvehicledepart ¼ ‘‘100 id ¼ ‘‘f 000 route ¼ ‘‘100 =i

stoplane ¼ ‘‘fo000 endPos ¼ ‘‘5000 duration ¼ ‘‘120000 =
h=vehiclei
hrouteid ¼ ‘‘100 edges ¼ ‘‘d04l23u14r12l45d2500 =i
hvehicledepart ¼ ‘‘100 id ¼ ‘‘f 000 route ¼ ‘‘100 =i

stoplane ¼ ‘‘fo000 endPos ¼ ‘‘5000 duration ¼ ‘‘240000 =
h=vehiclei
hrouteid ¼ ‘‘100 edges ¼ ‘‘d04l23u14r12l45d2500 =i
hvehicledepart ¼ ‘‘100 id ¼ ‘‘f 000 route ¼ ‘‘100 =i

stoplane ¼ ‘‘fo000 endPos ¼ ‘‘5000 duration ¼ ‘‘360000 =
h=vehiclei

6 Interpretation of Results

As an illustration to cluster head election algorithm, the Fig. 12 presents a network with a
ten vehicles and the Table 5 depicts the fitness quality value of each node using the model
proposed referring by the Table 2. After exchanging the Hello messages, a node votes for
its neighbor with the maximal metric value. In consideration of the proposed algorithm,

Fig. 11 Anticipate accidents in traffic

123
2062 Y. Hernafi et al.

Table 4 Traffic statistics via SUMO

Total number of vehicles 1000 2000 3000 6000

Total departure delay (s) 39 72 90 144

Total waiting time (s) 326 501 1631 3563
Average vehicular waiting time (s) 1 1 3 4
Total travel time (s) 21,405 33,176 39,814 62,333
Average vehicular travel time (s) 109 114 79 84
Total travel length (m) 227,065 348,459 472,387 718,621
Average vehicular travel length (m) 948 968 1164 1201
Average vehicular travel speed (m/s) 12 12 12 12

Fig. 12 An illustration example of the clusters’ relationship

node 5 and 6 are elected as cluster heads, furthermore, the cluster heads select the MPR
nodes for an eventual connection of all the heads. Indeed, the node with a better Fitness
metric value is picked out. Moreover, there are two choices to connect head 5 with head
node 6 either through (node 10–9) and (node 8–9). The entity 5 can selects the node, which
has the maximal Fitness quality and the presumed position. Then, by symmetry, head node
6 would selects node 9 as an MPR node.
In the Fig. 13, the simulation results show the performance of the different clustering
algorithms, effectively by addressing the Lower ID, Highest Degree, Utility Function and
our election approach. We examine the stability of the cluster, which depends on the
distance as well as velocity factors related to connectivity. The score is obtained by
counting the number of current vehicles based on duration travel time. If the average of
vehicles is still in the cluster, then it’s considered stable.
The results reveal that our algorithm performed better than the others, this clarification
can be justified by the fact that our model takes into account the distance factor

123
ACO and PSO Algorithms for Developing a New Communication... 2063

Table 5 The fitness quality models

Vehicle n1 n2 n3 n4 n5 n6 n7 n8 n9 n10

Fitness quality 350.5 101.6 123.4 196.8 590.2 486.3 159.1 260.4 390.1 101.5

Fig. 13 Percentage of cluster head stability for lowest ID, highest degree, utility function, and our selection,
depending on densities

proportional to the adopted mobility metric. For the lowed ID, the results obtained have
any impact on the selection of the cluster head, the movement and constraint of vehicular
network is not considered. Except for the highest degree and the utility function algorithm,
it is shown that they performed differently when the number of vehicles changed. The
density of vehicles and speed affect the number of nodes that are connected to the cluster
head, and both of these algorithms use the degree of connectivity, in order to choose the
right cluster head.
Figure 14 shows the analysis of the observed vehicles’ consumption depending on the
number of vehicles. The graph confirms that a fuel saving consumption is directly pro-
portional and has an affinity with the vehicle density. The proposed approach can decrease
the total fuel consumption rate and performs with great results for all scenarios. For
instance, when we vary the vehicle density, the fuel tends to increase and the reduction
triggered by the model is more than 30%.
It’s exactly the same thing happened to CO2 emissions, these emissions created by
burning fuel are directly related to fuel consumption rate. That said, when the percentage
of fuel increased, the CO2 emissions realizes a concrete high-rise. Generally speaking, the
traffic congestion does not affect the CO2 emissions consumption and fuel consumption of
our system, intuitively, for the reason that the alternative route used by vehicles in the
approach leads to a smooth traffic condition.

123
2064 Y. Hernafi et al.

Fig. 14 Fuel consumption rate and CO2 total emissions for classical model and our proposed intelligent
model, as a function of vehicle

In the scenario with accident condition, the fuel consumption has increased for both
solutions after the accident takes place. The total simulation period was 1000s, the accident
is triggered after 300 s and excluded at 700 s of simulation. When an accident appears
compared with the other solution, the results exposed that the lowest fuel consumption rate
is related to our solution in an accident scenario. Likewise, the Fig. 15 reveals that our
model can reduce the fuel consumption by offering at least 30% of benefits for all assessed
scenarios and consequently outperforms in total consumption of CO2 emissions.
In Fig. 16 we evaluated the average travel time for the overall increase in both densities
and times. It’s evident that the average travel time increases, when the number of vehicles
conduces of an exponential growth. However, the proposed model was performing great
results versus the classical model. The average time travel was almost the same for both
models in initial densities, and at higher densities, the proposed model has the best results.
Facing on incident conditions, it’s obvious that the duration of the accident tends to
increase the average travel over time. From the performances shown in the Fig. 17, we can
observe that our approach still works greatly, that being said, it has been shown that the
concept may bounce and avoid accidents. The mechanism started to reroute the vehicles
through another alternative. When the simulation cleared the duration of the accident, we
notice a rapid decline and all the graphs displayed their initial values.
Related already to the relative smartness, we demonstrate another aspects to evaluate
the model projected, as well as to enhance the quality of the air around cities by reducing
the emissions. The vehicle accident contributes significantly to fuel consumption and also
the CO2 emissions as mentioned previously. This lead consequently to large increases of
gas emissions. Among the described emissions, are cited, the particulate matter (PM),

123
ACO and PSO Algorithms for Developing a New Communication... 2065

Fig. 15 Fuel consumption rate for classical model and our proposed intelligent model, depending on
simulation time in case of an accident (from 300 to 700 s)

Fig. 16 Average travel time for the classical model and our model, as a function of vehicle density

123
2066 Y. Hernafi et al.

Fig. 17 Average travel time for the classical model and our model, depending on simulation time in case of
an accident (from 300 to 700 s)

nitrogen oxides NOx , carbon monoxide (CO), hydrocarbons (CH), and including the noise
emissions.
Moreover, the classical model produces negative effects, we can dissect a lead to large
increases in all emissions pollutants. The results of pollutant emissions can be depicted
respectively in the Figs. 18, 19, 20, 21 and 22. Comparing the results of the graphs, it can
be noticed that the quantity of gases emitted by the first scenario related to the proposed
model, is smaller than the classical model in the second scenario. It’s totally evidence and a
confirmation, as mentioned earlier in previous interpretations.
A quick glance in the Fig. 23 shows that in terms of time, by using the transmission
protocol user datagram protocol (UDP), the SC-OLSR protocol provides a reduced time of
average end-to-end delay compared, respectively, to OLSR and GSR routing protocols, we
find that the suggested protocol branded satisfying benefits face of both protocols. The SC-
OLSR protocol can automatically make routing decisions based on the updated pheromone
and fluidity of clustering mechanism, these clusters employ a higher data rate. It’s clear
that the metric of delay increases with increasing number of vehicles. For OSLR, expect
the problem of routes fail and the innumerable signaling messages, it’s despite a pre-
sentable performance in beginning compared to GSR who performed poorly. The average
delay of GSR increases constantly by cause of forwarding the data on road segments, the
problem in this method is that consequently, the road segment may become blocked and
congested, related to this dilemma, the overall communication performance could suffer.
Briefly, we can deduce that GSR, and also OLSR present highest delay compared to the
approach routing proposed.
The packet delivery ratio is affected by the increase of the traffic load, as we can catch
from the graphs depicted in the Fig. 24, as the vehicle density increases all routing

123
ACO and PSO Algorithms for Developing a New Communication... 2067

Fig. 18 The carbon monoxide (CO) emissions for classical model and our proposed intelligent model,
depending on vehicle density

Fig. 19 The hydrocarbons (HC) emissions for classical model and our proposed intelligent model,
depending on vehicle density

123
2068 Y. Hernafi et al.

Fig. 20 The noise emissions for classical model and our proposed intelligent model, depending on vehicle
density

Fig. 21 The nitrogen oxides (NOx ) emissions for classical model and our proposed intelligent model,
depending on vehicle density

123
ACO and PSO Algorithms for Developing a New Communication... 2069

Fig. 22 The particulate matter (PM) emissions for classical model and our proposed intelligent model,
depending on vehicle density

protocols clearly perceive a decrease. Using constant bit rate (CBR) traffic, the highest rate
that is anchored by the SC-OLSR protocol. It displays a higher PDR than OLSR and GSR,
with as much as a 40% increase compared with the classical OLSR and as much 25%
increase compared with GSR. In most cases, we also see some decline in the average
delivery ratio when data traffic increases. In contrast, as can be observed that in some
traffic load, the values have a tendency to decline. Undoubtedly, the SC-OLSR always
takes the lead with a mixture of the rise and stability, therefore, it performs better in
medium and dense networks than in simple traffic.
The gap in delivery ratio between these routing protocols is due to a problem inherent to
position-based routing and proactive one in general. We observe that the delivery ratio of
GSR shows a normal performance, the delivery increases as the network become denser,
this being by reason of that the protocol integrates the mechanism for road layout selection,
conceivably, a part of packets data cannot attain their destination due to an inadequacy of
connectivity performance in the grid layout. For OLSR, the highest PDR is even less, we
recognize that the performance is more affected by contention in the networks. The
expansion in the number of nodes translates into an increase of the link state updates. In
general, this increases a high volume of control overhead, hence many transmissions on
errors occurs as well as traffic congestion resulting in a lower delivery ratio. In our
measurements, SC-OLSR clearly outperforms the other protocols, which demonstrates the
relevance of the routing concept.
Concerning the throughput, when we observed the results identified as shown in the
Fig. 25. The results basically confirm conclusions of the PDR, its clear that the average
throughput of the SC-OLSR protocol is higher by comparing with the others protocols in

123
2070 Y. Hernafi et al.

Fig. 23 Average end-to-end

delay for SC-OLSR, OLSR, and
GSR depending on different node
densities. a 100 vehicles. b 300
vehicles. c 500 vehicles

123
ACO and PSO Algorithms for Developing a New Communication... 2071

Fig. 24 Average packet delivery

ratio for SC-OLSR, OLSR, and
GSR depending on different node
densities. a 100 vehicles. b 300
vehicles. c 500 vehicles

123
2072 Y. Hernafi et al.

Fig. 25 Throughput for SC-OLSR, OLSR, and GSR depending on different node densities

transmission traffic of CBR. Even with these little appreciable results these protocols tend
to decline, we see a relatively flow down of the number of vehicles.

7 Conclusion

The metamorphoses of Smart cities engender an appropriate requirement for safety

applications and pertinent communication in vehicular networks. That being said, in this
article, we opted for an improved intelligent transportation system, offering the opportunity
to exploit the benefits of smart mobility, that offers a wiser use in the case for emergency
operations. Indeed, we proposed an approach of a communication model to be used in
vehicular communications, leading to maintain stability, including avoiding the most
severe effect on traffic flow. Thereby, the improving idea of our work is to offer to vehicles
the convenience to escape the congested roads, and increase performance qualities by
reducing the fuel consumption, and likewise the CO2 emissions along with other specific
air pollutant emissions.
The suggested approach is composed of three components: an architecture based on
clustering algorithm, obviously, beneficial to ensure the fluidity of controlling communi-
cations and maintain the stability by highlighting a clustering algorithm made for a perfect
localization and great mobility.
Thereafter, the inclusion of a model in conjunction with interactions based on graph
theory to enable fluidity in the warning messages, by modeling the vehicle’s interactions by
two sub graphs based on graph theory to establish the communication and quality. In
addition, an inspiration was made from particle swarm optimization to elect the cluster
head. Efficiently, the cluster selection structure is determined in the first place by the
geographic position of the vehicle, then in the second place the most stable vehicle that can
act as a cluster head is a vehicle that has the lowest average distance among the other

123
ACO and PSO Algorithms for Developing a New Communication... 2073

vehicle in the cluster, the ideal closest speed to the average speed and the maximum
number of neighboring vehicles.
Finally, a lucidity from the Ant Colony to bring probability calculation of agents
tending to discover another alternative itinerary depending on the shortest route. The
interest of routing is to find the ideal path, with involving medics and ambulances who are
close and ready to provide the first aid care and rapid medical service on time. In order to
deal with these sophisticated optimizations, we have introduced the suggested changes in
the intern configuration of the OLSR routing protocol, namely, the smart cluster OLSR.
The idea was to calibrate the QoS routing extension for OLSR by modifying the selection
of the entity and the TC messages to spread the QoS information throughout the road.
The suggested model marks an acceptable benefits in terms of metrics performance and
optimization, by referring to accomplishment analysis along with simulation results. In the
future perspectives, we are planning to enhance the functionality from other more
sophisticated routing protocol and advanced design for implementing complex systems,
likewise, this work can be extended by aspects of artificial intelligence techniques with the
quality of service objectives. Furthermore, we intend to analyze the performances of our
deployed system in more realistic scenarios by using traces from real environments.

References
1. Nam, T., & Pardo, T. A. (2011). Conceptualizing smart city with dimensions of technology, people, and
institutions. In Proceedings of the 12th annual international digital government research conference:
Digital government innovation in challenging times (pp. 282–291). ACM. http://dl.acm.org/citation.
cfm?id=2037602.
2. Santa, J., Gmez-Skarmeta, A. F., & Snchez-Artigas, M. (2008). Architecture and evaluation of a unified
V2V and V2I communication system based on cellular networks. Computer Communications, 31(12),
2850–2861.
3. Bernardos, C. J., & Guerrero, C. (2013). Position-based routing in vehicular networks: A survey.
Journal of Network and Computer Applications, 36(2), 685–697.
4. Hernafi, Y., Ben Ahmed, M., & Bouhorma, M. An approaches’ based on intelligent transportation
systems to dissect driver behavior and smart mobility in smart city. In 2016 4th IEEE international
colloquium on information science and technology (CiSt) (pp. 886–895). IEEE.
5. Chaqfeh, M., Lakas, A., & Jawhar, I. (2014). A survey on data dissemination in vehicular ad hoc
networks. Vehicular Communications, 1(4), 214–225.
6. Hoque, M. A., Hong, X., & Dixon, B. (2014). Efficient multi-hop connectivity analysis in urban
vehicular networks. Vehicular Communications, 1(2), 78–90.
7. Zhang, Z., Boukerche, A., & Pazzi, R. (2011). A novel multi-hop clustering scheme for vehicular ad-
hoc networks. In Proceedings of the 9th ACM international symposium on mobility management and
wireless access (pp. 19–26). ACM.
8. Wahab, O. A., Otrok, H., & Mourad, A. (2013). VANET QoS-OLSR: QoS-based clustering protocol for
vehicular ad hoc networks. Computer Communications, 36(13), 1422–1435.
9. Shea, C., Hassanabadi, B., & Valaee, S. (2009). Mobility-based clustering in VANETs using affinity
propagation. In Global telecommunications conference, 2009. GLOBECOM 2009. IEEE (pp. 1–6).
IEEE.
10. Kennedy, J. (1999). Small worlds and mega-minds: Effects of neighborhood topology on particle swarm
performance. In Proceedings of the 1999 congress on evolutionary computation, 1999. CEC 99 (Vol. 3).
IEEE.
11. Trelea, I. C. (2003). The particle swarm optimization algorithm: Convergence analysis and parameter
selection. Information Processing Letters, 85(6), 317–325.
12. Hernafi, Y., Bouhorma, M., Ben Ahmed, M., Boudhir, A., & Alluhaidan, Y. A. (2015). Clustering
architecture for smart mobility control: Application to pilgrims in the Great Mosque in el Hajj. Journal
of Theoretical and Applied Information Technology, 78(1), 84.

123
2074 Y. Hernafi et al.

13. Stojmenovic, I., Seddigh, M., & Zunic, J. (2002). Dominating sets and neighbor elimination-based
broadcasting algorithms in wireless networks. IEEE Transactions on Parallel and Distributed Systems,
13(1), 14–25.
14. Ramalingam, A., Subramani, S., & Perumalsamy, K. (2002). Associativity based cluster formation and
cluster management in ad hoc networks. In 9th international conference on high performance com-
puting (HiPC02).
15. Fan, P., Haran, J. G., Dillenburg, J., & Nelson, P. C. (2005). Cluster-based framework in vehicular ad-
hoc networks. In Ad-hoc, mobile, and wireless networks (pp. 32–42). Springer.
16. Olfati-Saber, R., & Murray, R. M. (2002). Graph rigidity and distributed formation stabilization of
multi-vehicle systems. In Proceedings of the 41st IEEE conference on decision and control, 2002 (Vol.
3, pp. 2965–2971). IEEE.
17. Olfati-Saber, R., Dunbar, W. B., & Murray, R. M. (2003). Cooperative control of multi-vehicle systems
using cost graphs and optimization. In Proceedings of the American Control Conference. Citeseer.
18. Sebastian, A., Tang, M., Feng, Y., & Looi, M. (2009). Multi-vehicles interaction graph model for
cooperative collision warning system. In Intelligent vehicles symposium, 2009 IEEE (pp. 929–934).
IEEE.
19. Yang, X., Liu, J., Vaidya, N. H., & Zhao, F. (2004). A vehicle-to-vehicle communication protocol for
cooperative collision warning. In The first annual international conference on mobile and ubiquitous
systems: Networking and services, 2004. MOBIQUITOUS 2004 (pp. 114–123). IEEE.
20. Biswas, S., Tatchikou, R., & Dion, F. (2006). Vehicle-to-vehicle wireless communication protocols for
enhancing highway traffic safety. IEEE Communications Magazine, 44(1), 74–82.
21. Li, G., & Boukhatem, L. (2013). Adaptive vehicular routing protocol based on ant colony optimization.
In Proceeding of the tenth ACM international workshop on vehicular inter-networking, systems, and
applications (pp. 95–98). ACM.
22. Baras, J. S., & Mehta, H. (2003). A probabilistic emergent routing algorithm for mobile ad hoc
networks. In WiOpt’03: Modeling and optimization in mobile, ad hoc and wireless networks.
23. Plesse, T., Adjih, C., Minet, P., Laouiti, A., Plakoo, A., Badel, M., et al. (2005). OLSR performance
measurement in a military mobile ad hoc network. Ad Hoc Networks, 3(5), 575–588.
24. Hofmann-Wellenhof, B., Lichtenegger, H., & Collins, J. (2012). Global positioning system: Theory and
practice. Berlin: Springer.
25. Tangour, F., Saad, I., & Borne, P. (2009). Optimisation par colonie de fourmis. REE. Revue de l’lec-
tricité et de l’électronique, (4), 39–43. http://cat.inist.fr/?aModele=afficheN&cpsidt=21331936.
26. Dorigo, M., Birattari, M., & Sttzle, T. (2006). Ant colony optimization. IEEE Computational Intelli-
gence Magazine, 1(4), 28–39.
27. Lochert, C., Hartenstein, H., Tian, J., Fussler, H., Hermann, D., & Mauve, M. (2003). A routing strategy
for vehicular ad hoc networks in city environments. In Intelligent vehicles symposium, 2003. Pro-
ceedings. IEEE (pp. 156–161). IEEE.
28. Lochert, C., Mauve, M., Füßler, H., & Hartenstein, H. (2005). Geographic routing in city scenarios.
ACM SIGMOBILE Mobile Computing and Communications Review, 9(1), 69–72.
29. The Network Simulator—ns-2.
30. DLR—Institute of Transportation Systems—SUMO Simulation of Urban Mobility.
31. Karnadi, F. K., Mo, Z. H., & Lan, K. (2007). Rapid generation of realistic mobility models for VANET.
In Wireless communications and networking conference, 2007. WCNC 2007. IEEE (pp. 2506–2511).
IEEE.

123
ACO and PSO Algorithms for Developing a New Communication... 2075

Yassine Hernafi was born in Rabat, Morocco. He received his engi-

neering degree in computer science, in 2013 from Abdelmalek Essaadi
University in Tangier, Morocco. He is currently a Ph.D. Candidate in
Computer Science, Systems, and Telecommunications Laboratory. His
main research interests are in smart cities, intelligent transportation
systems and vehicular networks applications.

Mohamed Ben Ahmed Associate Professor of computer sciences at

Abdelmalek Essaadi University-Morocco, he received the Ph.D.
degree in computer sciences and Telecommunications, in 2010 from
Abdelmalek Essaadi University. His researches are about data mining,
routing in wireless sensor networks and smart cities. He is currently a
supervisor of several thesis and investigator in several international
research projects about smart cities. He is author of more than twenty
papers published in international journals and conferences.

Mohammed Bouhorma Prof. Dr. received the ‘‘Master degree in

Electronics’’ degree in 1990 from Abdelmalek Essaadi University in
Tetuan, Morocco, the ‘‘DEA’’ degree in Electronics and Telecommu-
nications and the ‘‘Doctorat d’Etat’’ degree in Telecommunications
with honors, respectively, in 1991 and 1995 from the ‘‘ENSEEIHT-
INPT’’ of Toulouse. He is a Professor of computer sciences and net-
works at Abdelmalek Essaadi University since 1999. He was chief of
computer sciences department and he is also responsible of the Master
titled Computer Sciences Systems and Computers. Prof. Bouhorma has
supervised several Ph.D. and Masters Theses and has been the prin-
cipal investigator and the project manager for several international
research projects dealing with different research topics concerned with
his research interests mentioned above.

123