Vehicular Communications 47 (2024) 100746

Contents lists available at ScienceDirect

Vehicular Communications
journal homepage: www.elsevier.com/locate/vehcom

Distributed and collaborative system to improve traffic conditions using

fuzzy logic and V2X communications
José Antonio Sánchez , David Melendi *, Roberto García , Xabiel G Pañeda , Víctor Corcoba ,
Dan García
Department for Informatics, University of Oviedo, EPV, Campus de Xixón, s/n, Xixón, Asturies 33203, Spain

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

Keywords: Nowadays, the increase in the number of vehicles on the roads has brought about several problems such as an
Collaborative System increase in traffic congestion and, consequently, in polluting emissions. These problems are especially severe in
Fuzzy logic urban environments. It is crucial to perform a sustainable urban mobility plan to improve the traffic and
Smart mobility
therefore, reduce the negative impacts caused by traffic jams. To this end, this paper presents a smart mobility
Vehicular communications
Vehicular networks
plan that employs a collaborative driving strategy. Each vehicle tries to infer traffic conditions using its own
status and the information shared by other peers. Using a fuzzy logic approach, vehicles perform decisions in
accordance with the traffic levels inferred in real time. The designed mobility plan has been tested through a
simulation environment and considering two types of urban areas in a typical European city (a peripheral area
and a more congested city centre). If we compare the performance of traffic with and without the system
designed, with our approach average speeds increase by up to 11.20 % and CO2 emissions are reduced by up to
12.27 %. Thus, our results show that the mobility plan has helped to enhance the ability of cars to be able to solve
problems caused by traffic congestion and traffic jams.

1. Introduction create concentrations of vehicles in reduced areas. Therefore, these ve­

hicles generate a huge quantity of pollutant gases in the affected area.
In recent years, the number of people living in metropolitan areas has Although, various approaches have been proposed to reduce traffic
increased significantly [1]. As a result, the number of vehicles in cities jams, such as the proposals of Lakas & Chaqfeh [7], Brennand et al. [8],
has grown constantly and both the fabrication and sales of vehicles Rocha-Filho et al. [9] or González-Aliste et al. [10], they are difficult to
continue to increase [2,3]. E-commerce has been another factor eliminate and, thus, they continue to be a great source of pollutants.
increasing the traffic in urban environments in recent years, because of Even in an electric vehicle scenario, traffic jams reduce the distance
the huge growth in these services [4] and, thus, in the number of ship­ which may be driven with the batteries due to the usage of other
ments performed, especially in urban environments. Due to the fact that electric-powered in-vehicle services [11] and are a source of stress for
vehicles are one of the most harmful agents for the environment [5], drivers [12]. Even in an autonomous vehicle scenario, traffic jams would
more cars lead to more pollution. Even though several countries have continue to be a source of stress for the passengers of the vehicle. Re­
planned the retirement of combustion cars [6], the alternative is still searchers must then continue to develop new methods in order to
unclear, and this retirement is yet without worldwide consensus. The improve traffic conditions and alleviate traffic jams and, consequently,
fact is that pollution levels are constantly rising, and more mechanisms reduce current air pollution, improve the performance of electric vehi­
are needed to reduce emissions to improve air quality. Due to this fact, cles and reduce the stress of the citizens.
governments have been deploying networks for monitoring the envi­ With the goal of alleviating the aforementioned problems, in this
ronment with the aim of controlling contamination levels. The situation paper we present a smart mobility solution based on a collaborative
has reached a point in which critical measures are been carried out, such driving strategy. We have implemented this system based on Vehicular
as applying traffic restrictions. Ad-hoc Network (VANET) communications using V2V (Vehicle to
Traffic jams are one of the major causes of air pollution because they Vehicle) and V2I/I2V (Vehicle to Infrastructure/Infrastructure to

* Corresponding author.
E-mail address: [email protected] (D. Melendi).

https://doi.org/10.1016/j.vehcom.2024.100746
Received 26 July 2023; Received in revised form 21 December 2023; Accepted 26 February 2024
Available online 27 February 2024
2214-2096/© 2024 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-
nc/4.0/).
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Vehicle) interactions. Moreover, a fuzzy logic inference system is used to a strategy like collaborative driving can be the best way to achieve an
detect and adjust the behaviour of the vehicles, according to several optimal use of the urban infrastructure, with the cooperation among
parameters such as speed, acceleration, number of vehicles in the same vehicles by using communications to navigate through urban traffic
area, etc. Through simulations, we have implemented two different [15].
scenarios to check the impact of our design in two different types of area Collaborative driving has gained importance in ITS (Intelligent
in a common city: one in the city centre with a slightly haphazard Transport Systems) because of the high number of applications that can
development, and another in the surroundings of the city in a recently make use of it, especially those that have to do with safety and effi­
constructed urban area. For this purpose, and with the aim of achieving ciency. For instance, Kitwiroon et al. [16] examined the impact of
the most realistic results possible, we have used the roadmap of Gijón, a various traffic measures on both traffic density and pollution emissions
city in the north of Spain with many projects aligned with the smart-city in London. The authors used a system called OSCAR. They apply several
concept. We have employed different vehicle creation rates in both techniques such as reducing the number of heavy-duty vehicles (HDVs)
scenarios to study different congestion levels in both types of areas. or the effects of speed, as we do. The authors achieve improvements of
Our results show the improvement, in terms of improved mobility up to 11 %, reducing HDVs a 20 %. Speed is a relevant factor in vehicular
and based on the reduction of greenhouse gas emissions, caused by the emissions. The impact of speed limits on the environment depends on
collaborative system proposed. We infer traffic conditions, detect traffic the impact of these limits on the behaviour of drivers and may be
jams, and warn other vehicles of problematic situations, so that they can different for different pollutants. Panis et al. [17] evaluate the impact of
find alternative routes to their destination. This leads to an increase in speed reductions on emissions, by using different modelling approaches
the speed of vehicles and an important reduction in the generation of (both microscopic and macroscopic). They found that speed limits in
pollution. Our design may be applied in the context of route planning or urban areas have a non-significant impact on CO2 and NOx emissions.
implemented in a real-time driving assistant. Furthermore, it could even But, in the case of PM (Particulate Matter) emissions, microscopic results
be applied in an autonomous vehicle scenario to automatically improve show a significant decrease, with a moderate increase in macroscopic
the routes followed. results. Other works, such as Mahmod et al. [18], study the reduction of
The main contributions of this research follow: emissions in urban intersections by using ITS countermeasures. They
consider restrictions in speed and heavy-duty vehicles. The analysis
• In this paper we present a novel system for improving traffic con­ show that, in a specific area, the CO2 emissions generated by vehicles
ditions in urban areas based on vehicular communications and fuzzy may be reduced up to 23 %, but it is necessary to reduce the traffic a 20
logic. %. We decrease car emissions by up to 12.27 %, but no traffic needs to be
• We have conceived a fuzzy logic system to infer traffic conditions reduced. Previous work has also designed techniques to allow collabo­
from speed and stationary times by reporting one of four values: Very ration in a VANET by means of clustering techniques taking into account
Low Traffic, Low Traffic, Medium Traffic and High Traffic. the particular characteristics of these networks (e.g., mobility) For
• We have designed a system to autonomously alter the behaviour of example, Mukhtaruzzaman & Atiquzzaman [19] analyse
the vehicles depending on the values reported by the fuzzy logic intelligence-based VANET clustering techniques and compare hybrid
inference system. architectures combining machine learning and fuzzy logic algorithms, or
• We present a communications model to allow vehicles to exchange Aissa et al. [20] who propose and analyse the performance of a clus­
information to take decisions collaboratively when traffic conditions tering technique based on fuzzy logic.
are considered to be suboptimal.
• Two simulation environments with different urban designs have 2.1. VANET applications
been implemented in order to perform experiments and verify the
performance of the proposed system in diverse geographical settings. Many VANET applications have been implemented since the incep­
• The results show that the designed system can improve the flow of tion of this type of networks. Taking into account the classification
traffic at the same time it reduces the generation of pollution up to performed by Toor et al. [21], we can observe 2 types of VANET ap­
12.27 % when combustion vehicles are used and, as a side benefit, plications: user and safety applications. The former may provide users
the stress of the drivers. The improved flow of traffic may also imply with information, warnings and entertainment. Mainly, there are two
an optimised usage of batteries in an electric vehicle scenario. types of user applications: Internet connectivity and peer-to-peer ap­
plications for sharing music, movies and even, playing games. On the
The rest of the paper is organized as follows: Section II shows related other hand, safety applications are more important than user applica­
work about collaborative systems and Fuzzy Logic in vehicular envi­ tions in the context of ITS. They are responsible for reducing road ac­
ronments. The design of the collaborative system is presented in Section cidents or traffic congestion problems, approach followed by us. Within
III, with details of the vehicular communications and the fuzzy inference this field, studies such as Wang & Thompson [22] show that 60 % of the
system. Section IV presents the simulation environment with the soft­ accidents could be avoided if the driver had received a collision warning
ware and parameters employed in this study. Section V shows the results half a second before. Taking into account safety applications for
obtained from the simulations in the aforementioned scenarios. Section collaborative and cooperative environments, we find other works such
VI provides a discussion about the results obtained and about the limi­ as Knorr et al. [23]. The authors use V2V communications to avoid
tations of the study. Finally, Section VII includes the conclusions and traffic jams by sending periodic beacon messages warning about
some guidelines for future work. congestion. Their results prove that VANET communications may be
suitable to improve traffic efficiency, but their solution has been
2. Related work designed for highway roads only. Also, Hafner et al. [24] design, with
safety in mind, a cooperative V2V communication solution to avoid
Previous work has demonstrated that a coordinated approach when collisions at intersections. They perform experiments with two instru­
trying to improve the efficiency in urban mobility benefits all the mented vehicles engaged in an intersection collision avoidance scenario
members involved. This has been shown by many studies such as Youn on a test track. Moreover, Ribeiro et al. [25] propose a low-cost
et al. [13], which demonstrates that in a group of drivers seeking the collaborative and opportunistic system that monitors traffic using
shortest path to a destination a Nash equilibrium is reached, so no single available IEEE 802.11 networks. The system provides information
driver can do any better by changing his or her strategy unilaterally, or regarding the location of vehicles, and, with this information, the au­
by Tientrakool et al. [14], stating the benefits of making use of sensors thors monitor traffic conditions collaboratively, although the solution
and V2V communications in comparison with using sensors alone. Thus, relies on a central system in which the information is stored and treated.

2
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Dannheim et al. [26], give examples of intelligent, networked and lead to the reduction of pollution generated by these vehicles and the
collaborative driving assistance systems that employ V2X communica­ level of stress of the drivers. Also, we do not rely on any central system,
tions with the aim of alerting the driver in time about external events in contrast with previous work such as Ranjita and Acharya [41].
and conditions so an increase in both safety and comfort can be ach­ Instead, vehicles interact with each other to perform decisions collab­
ieved. Drawil & Basir [27] present a collaborative technique used to oratively, reducing the infrastructure needed by the solution. Moreover,
improve the location accuracy of vehicles in a VANET. Two more ex­ we use V2X communications thanks to the use of IEEE 802.11p solu­
amples of collaborative transportation applications are presented by tions, due to the fact that this standard continues to be evolved and that
Piorkowski [28]. In this case, both applications pursue different aims to it generally outperforms other vehicular communication technologies
those seen previously. One of them has the goal of coordinating drivers such as those based on LTE-V2X [42]. This is in contrast with other
to free parking spaces and the other, of matching drivers who offer works which rely on Wi-Fi communications such as Stolfi and Alba [43]
empty seats in their cars with pedestrians who want a ride to reach their or the recent Ranjita and Acharya [41]. In fact, although we follow a
destinations. similar approach to that of Stolfi and Alba [43], their solution uses
Due to the fact that collaborative driving relies on the information several parameters included in our work, but they do not consider
shared among vehicles, it is necessary to take some measures in order to pollution. Moreover, they take into account accelerations and the speed
make cooperative driving applications function in a safer way. Thus, of vehicles, but as static parameters for each type of vehicle. In contrast,
protocols such as the one presented by Gu et al. [29] could be used with our work is focused on dynamic accelerations and speeds. We adjust
the aim of guaranteeing fail-safe operations in driving environments. them depending on the status of traffic.
Despite the problems arising from the dependency of these systems
on communications in order to work properly, the use of V2X commu­ 3.1. Differences with previous work
nications has many more significant advantages. The architecture of
most of the ITS used so far often separates sensing from computing, as In summary, our approach is novel as it is based on a totally
well as decision making from actuation. In fact, many of them, such as autonomous and distributed approach for urban scenarios, based on the
the traffic congestion detection and dissemination system proposed in collaboration between vehicles through the use of vehicular
Jayapal & Roy [30], use external data servers to process the information communications.
that they collect, which implies that a long period of time must be spent Table 1 shows aspects in previous work clearly differing from the
until the vehicle acts in accordance with the previously collected in­ approach followed by our design. If compared with previous research,
formation. Thus, if the information can be interchanged between vehi­ our proposal has been designed to improve traffic conditions and, thus,
cles, and the vehicles are able to process it themselves, these systems to reduce the effects of traffic density on the environment and the
could be used to solve traffic problems in real time. drivers. Most previous work focus on other aspects such as traffic pre­
diction, safety, passenger comfort or data communications. Also, we do
3. Fuzzy logic in VANET applications not apply restrictions on traffic such as denying certain types of vehicles,
blocking certain roads, or reducing speed limits as other proposals do.
Regarding the usage of fuzzy logic [31], there has been extensive Furthermore, our proposal is collaborative and decentralised and does
previous research carried out in the VANET field in which this technique not rely on any central system like other works do. There is previous
has been used. For example, that carried out by Dimitrou et al. [32], who work focusing on motorway traffic, whereas our proposal has been
use a fuzzy rule-based system with the aim of modelling and predicting designed to improve urban traffic, due to the fact that traffic density is
the traffic flow. Another example of these studies is Zrar-Ghafoor et al. particularly problematic in cities. Finally, there is previous work also
[33], which proposes an intelligent Adaptive Beaconing Rate (ABR) focused on providing a theoretical solution, in contrast to our proposal
approach based on fuzzy logic that takes traffic characteristics into which is eminently practical. Section 6 includes a comparison of our
consideration to control the frequency of beaconing. Also, research results with the achievements in other proposals.
carried out to detect network traffic congestion such as Sonmez et al.
[34] or Naja & Matta [35], where fuzzy logic is also used to provide 4. Design of the smart collaborative system
information for a vehicular admission control. Another example is Miao
et al. [36], employing fuzzy logic to manage resources such as text, Basically, our fuzzy logic inference system infers traffic conditions
audio and video in a VANET. Luo et al. [37] present a fuzzy logic from speed and stationary times by reporting one of four values: Very
collaborative solution to specifically allow file transfers in a VANET, Low Traffic, Low Traffic, Medium Traffic and High Traffic. These values
mainly designed for infotainment applications in highway vehicular allow vehicles to take decisions and adjust their behaviour:
networks. Balasubramani and Aravindhar [38] focus their research on
routing and present an optimized next hop node selection process for • When the fuzzy logic inference system considers that the traffic is
VANETs using fuzzy logic and compare their solution with Greedy Low or Very Low, it increases the speed of vehicles by 10 % or 30 %
Perimeter Stateless Routing (GPSR). Mukhtaruzzaman & Atiquzzaman respectively to take advantage of traffic conditions. It is also
[19] discuss intelligence based VANET clustering strategies and analyse important to take into account that the resulting speed after the in­
hybrid architectures combining fuzzy logic with other techniques. crease will never exceed the urban maximum speed in Spain, which
Similarly, Aissa et al. [20] propose and assess a fuzzy logic based clus­ is currently 13.889 m/s (50 km/h).
tering technique designed to select the best cluster head (CH) in a po­
tential safety scenario. Abbasi et al. [39] also focus on data transmission
issues and present a vehicle-weighted clustering model based on fuzzy
logic (FWDP), designed to improve data propagation in VANETs. Arena Table 1
et al. [40] present a fuzzy logic solution designed to control, dynami­ Main differences between our proposal and previous research.
cally, traffic lights in road intersections in a city, adapting their behav­ Proposed system Feature in previous work References
iour to traffic conditions.
Designed to improve Not focused on improving [15,19,20,22,24,
Although fuzzy-logic has been extensively used in VANET scenarios traffic conditions traffic conditions 26–29,32–39]
with different goals, in contrast with previous work our solution in­ No traffic restrictions Traffic restrictions applied [13,16–18]
cludes a fuzzy logic-based decision system to allow vehicles not only to Distributed system Centralised system [25,30,41,43]
infer traffic conditions, but also to perform real time decisions in order to Urban areas Motorway roads only [14,23,35]
Practical approach Theoretical approach [13–15,40]
use alternative routes to avoid congested areas of a city. This will also

3
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

• When the traffic level is categorized as Medium or High, V2X com­ simply increase their speed.
munications are employed. This permits a collaborative approach The increase and decrease speed reference values used by the fuzzy
when trying to avoid traffic jam formations. There are two different logic inference system have been chosen after some preliminary tests.
types of messages: “Alert” and “Traffic”. “Alert” messages are sent
using V2V communications to vehicles in range when a car detects 4.1. Collaborative driving and traffic congestion reduction
that the level of traffic is Medium. The cars will therefore be aware
that a 5 % speed decrease is needed in order to enhance the distance The designed system is based on the idea that speed and stationary
between vehicles and so, to avoid, if possible, traffic congestion. Only times are indicators of traffic conditions. Each vehicle is aware of its own
the vehicles that are travelling in the same street and direction speed and the time spent without moving and, thus, may infer the
reduce their speed, which is the reason why fields with the road conditions of traffic autonomously (without the need of a central sys­
identification and the direction are included. Furthermore, “Traffic” tem). Collaboration means that said information may be then shared
messages are sent when a car detects that the level of traffic is High. with other vehicles in an effort to alleviate the situation. In our
“Traffic” messages are broadcasted to other cars using V2V com­ approach, control information is shared when a vehicle considers that
munications and to RSUs using V2I communications. RSUs, using I2V traffic is high, so the system reacts to an already existing problem, or
communications, forward “Traffic” messages to other cars to ensure when the traffic is medium in order to try to improve a potential prob­
information propagation. This kind of message affects the behaviour lem. Reacting to a problem means that some vehicles may be stuck in a
of all the vehicles that receive them and that have a certain street in traffic jam, but other vehicles receiving the information generated by
their paths. As with what happened with “Alert” messages, they the former may avoid the traffic jam by selecting a different route. A fast
include the road identification but, in this case, when a car receives propagation of this information using combined V2V, V2I and I2V
this type of message, it employs it with the aim of changing its route communications will help to finish the congestion situation. Avoiding
in order to avoid the street in which dense traffic was detected. the problem means that when there is evidence of a deterioration in
traffic conditions, that situation may be improved by decreasing the
The design of the smart collaborative system is shown in Fig. 1. There speed of the vehicles and enhancing the distance between them. This
are cars stopped in a traffic jam inferring that traffic is High. They are information may be propagated backwards to other vehicles travelling
broadcasting “Traffic” messages which are also forwarded by a RSU in the same direction, so that the concentration of vehicles in a single
available in the area, warning the rest of the vehicles about a congested area is reduced and, thus, the traffic jam is ultimately avoided. Finally,
street. The cars that receive the messages that were going to go through when a vehicle infers that traffic is low or very low, an increase of speed
that road, calculate a new route to avoid said road. There are other will reduce trip times and, thus, the general time vehicles occupy city
vehicles inferring that the traffic is Medium. These cars broadcast “Alert” roads. This will improve overall traffic conditions reducing, at the same
messages (not forwarded by the RSU) warning other vehicles travelling time, pollution and the stress of the drivers.
in the same street and direction to decrease their speed. The rest of the The system is also able to react to sudden changes in traffic caused by
vehicles in the example infer Low or Very Low traffic conditions, so they accidents or other external factors. The vehicles driving in the affected

Fig. 1. Design of the smart collaborative system.

4
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

road would suddenly infer that traffic is high because of their reduced evaluations, the Aggregator then combines the results to obtain output
speed or because they have to stop for a long period. This will imme­ values for all the output parameters. Once the membership value for
diately trigger the generation of “Traffic” messages which will be each of the linguistic labels that compose the output parameters is
broadcasted by the affected cars and by the RSUs in the area. Vehicles known, the Defuzzifier block is able to obtain a resulting value out of it.
receiving these messages, which were going to traverse the affected Once the structure of a fuzzy inference system is known, the first step
road, proceed to calculate a new route to avoid the problem. Once the to design one is to determine the number of input parameters and output
traffic conditions are recovered, the affected vehicles would infer an variables that will be considered, as well as the shape of their mem­
improvement in the situation and, thus, stop sending “Traffic” messages bership functions. Since the purpose of cars employing this system is to
and gradually start to increase their speeds back to normality. detect traffic conditions, the parameters chosen are their speed and the
time that they are stationary. These parameters have been chosen
5. Usage of fuzzy logic because both of them are clearly affected by the level of traffic. The
higher the traffic level is, the lower the speed of the vehicles and, when a
Transportation problems, including traffic congestion, follow com­ traffic jam occurs the higher the time that vehicles are stationary. Also,
plex behaviours as they change over time depending on various factors. when designing the membership functions, it is necessary to take into
These behaviours are difficult to define accurately (with precise nu­ account that the more complex they are, the more accurate results can
merical data) and, thus, it is not easy to make decisions to solve said be obtained so, for that reason, four linguistic labels were employed for
problems since they are based on uncertain information. The most each input variable: low speed, medium speed, high speed, very high speed,
suitable artificial intelligence technology to deal with uncertainty and low time, medium time, high time and very high time. Furthermore, as the
pervasive fuzzy information is precisely fuzzy logic. This technology purpose of this system is to determine the level of traffic around the
may be used in order to make decisions in contexts in which there is no vehicle, this parameter is employed as output variable, also with four
accurate numerical data, as already stated by Zadeh [31], and an different labels: very low traffic, low traffic¸ medium traffic and high traffic.
approximation of the real world is adequate. Accordingly, fuzzy logic The shape of the membership functions designed is shown in Fig. 3,
can be seen as a useful tool when it comes to letting vehicles make de­ Fig. 4 and Fig. 5.
cisions regarding, for instance, traffic in an efficient way. Detecting The second step is to define the set of IF-THEN rules which will be
whether there is a traffic jam or not, will not depend on a combination of used to obtain the membership value of each linguistic label for the
binary decisions. It will be determined as the result of an inference output values. These rules connect the different values of input variables
system based on more interpretable and user-friendly concepts such as with values of the output variable based on expert knowledge. Table 2
the volume of traffic (e.g., very low, low, medium, or very high). presents the set of rules that make up the fuzzy logic base.
Furthermore, the system may consider parameters which are numerical As mentioned above, each of the rules will produce an output asso­
in nature, including the speed of the cars or the time they spend sta­ ciated to one of the linguistic labels of the output parameter, as shown in
tionary, without the need of classifying data into discrete categories. Table 2. According to that table, ri would be the specific value of each
This contrasts with other artificial technologies which relay on binary or rule, i being the rule number. The value of each ri is calculated by using
probabilistic representations of data. Moreover, most of these technol­ MINIMUM or MAXIMUM operations with the corresponding values of
ogies are based on doing predictions learnt from large datasets, but a the membership functions of input variables shown in Figs. 3 and 4. The
rule-based approach may be more suitable, self-explanatory, and more value of each ri is used to calculate Eqs. (1) to (4) and obtain the values
computationally-efficient without the need of extensive training pro­ corresponding to each output linguistic label. For example, the rules
cesses or massive datasets. It also important to take into account that needed to determine if the traffic is very low are r9, r10, r13 and r14, as
there has been extensive previous research carried out in the VANET shown in Table 2. Root square sum operations have been performed, as
field in which fuzzy logic has been used [19,32–40]. this type of operation is able to give a good, weighted influence on all the
implied rules. The resulting values are used by the Defuzzifier block to
6. Configuration of the fuzzy inference system calculate the final output value.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
A fuzzy logic inference system, based on Mamdani [46] and shown in very low traffic = r29 + r210 + r213 + r214 (1)
Fig. 2, is composed by three differentiated blocks: a Fuzzifier, an Infer­
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
ence Engine and a Defuzzifier.
low traffic = r25 + r215 (2)
The first block is in charge of converting the crisp input into fuzzy, so
it can be used by the Inference Engine. The second is comprised by the √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Rules block, in which its input is compared with the membership medium traffic = r21 + r22 + r26 + r211 + r212 + r216 (3)
functions to obtain the membership value of each linguistic label that
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
makes up the IF-THEN linguistic rules, and the Aggregation block. In this
high traffic = r23 + r24 + r27 + r28 (4)
way, once that the membership values have been obtained, they are
evaluated according to the Fuzzy Logic rules, so an output value is ob­ The Defuzzifier is able to obtain from the calculated values a partic­
tained with each one of them. It is worth mentioning that the resulting ular number with which the traffic conditions can be assessed. The
values are calculated by using MINIMUM or MAXIMUM operations in
accordance with the way the membership values or antecedents are
linked with either AND or OR operations. After completing all the

Fig. 2. Fuzzy inference system with inference engine included. Fig. 3. Speed membership functions.

5
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

6.1. Description of V2X communications

“Alert” and “Traffic” messages are both WSM (Wave Short Message).
These messages are sent using the WSMP (Wave Short Message Protocol)
through the CCH (Control Channel). This is the channel provided by the
802.11p standard [44] which has been designed to distribute critical
data packets.
Both messages are identified by the vehicles by using the PSID
(Provider Service Identifier) field according to the standard. Thus, the
value given to this field will be 1 when delivering “Alert” messages and
0, in the case of “Traffic” messages.
Fig. 4. Time membership functions. WSMs are sent in broadcast mode but it is important to take into
account that the characteristics of urban environments make their
dissemination difficult. Although this is not so critical in the case of
“Alert” messages, it is very important when dealing with “Traffic”
messages, since they should be received by the highest number of ve­
hicles possible in order to prevent traffic jams. Due to this, “Traffic”
messages are also spread by RSUs (Roadside Units), so each time that
any of these units receive one of these messages, it forwards the message
with the purpose that a higher number of vehicles can modify their
routes.
It must be considered that, due to the rapid variability of the traffic,
the validity of the messages is short. They should not be reporting about
traffic situations that no longer exist, e.g., traffic jams. For this reason,
Fig. 5. Traffic level membership functions. the messages are not always sent again each time a vehicle receives them
since the longer their dissemination takes, the lower their validity lasts.
Besides, it must also be kept in mind that a high amount of traffic in­
Table 2 formation in the network could have negative effects on the traffic since,
Set of rules for fuzzy logic base. for instance, if a vehicle receives many “Traffic” messages at the same
Rule Number IF THEN time it could lead it to adopt a much longer route unnecessarily. This
Speed Time Traffic change of route could mean that it will take longer to reach the desired
1 Low Low Medium destination and so could favour the increase of the traffic density in the
2 Low Medium Medium network.
3 Low High High
It is important to mention that, while the vehicles are always allowed
4 Low Very High High
5 Medium Low Low
to modify their speed, this does not happen with the delivery of the
6 Medium Medium Medium messages. It may be the case that a given car detects repeatedly that the
7 Medium High High traffic is low or medium, sending many times the same message.
8 Medium Very High High Moreover, it is important to bear in mind that the system has been
9 High Low Very Low
designed to be used in urban areas, where the number of vehicles is
10 High Medium Very Low
11 High High Medium usually high, and, for this reason, this problem may lead to causing a
12 High Very High Medium broadcast storm in the VANET [45]. Therefore, with the aim of avoiding
13 Very High Low Very Low these problems, a minimum inter-message time lapse was employed.
14 Very High Medium Very Low
Firstly, a vehicle may send its first message after a time lapse of 5 s since
15 Very High High Low
16 Very High Very High Medium
the start of the simulation. We consider this period as the minimum
necessary to consider a vehicle as part of the network. Moreover, after a
certain vehicle sends a message, it must wait for 300 s before it is
operation used to obtain this result is the centroid method. The crisp allowed to send another message. These values have been chosen thanks
output value is calculated according to Eq. (5), where centre refers to the to previous simulation experiments. They are high enough to avoid an
central point of the membership function of each linguistic label unnecessary generation of control messages and low enough to achieve
considered (ranging from 0 to 1) and strength, to the firing strength good results when traffic density is high.
associated with it. When Eq. (5) is applied to our study, we obtain Eq. It is worth mentioning that as RSUs are used to disseminate messages
(6). Eq. (6) shows that we have used different centre values for each within a specific area. They should be located in such a way that they
linguistic label, wheighin more the labels implying higher levels of can cover a wide area avoiding, at the same time, the overlapping of
traffic. their coverage areas. Besides, as will be explained in depth when
∑N describing the scenarios used in the study, when choosing their location,
(centeri ⋅strengthi )
output = i=1∑N (5) it is also necessary to consider that the best option is to deploy them in
i=1 strengthi spots near the points where traffic jams tend to take place.

0.1 very low traffic + 0.2 low traffic + 0.5 medium traffic + 0.9 high traffic
output = (6)
very low traffic + low traffic + medium traffic + high traffic

6
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Table 3 mobility models described in Harri et al. [47], the vehicles use a “car-­
Simulation and network parameters. following” mobility model. Specifically, we employ the default model
Simulation Parameters implemented in SUMO, which is based on traditional models such as
Gipps [48] and the Intelligent Driver Model [49].
Type of Maps Real (City of Gijón)
Vehicular traffic (packet size) 256 B
Vehicle creation rates 1v/s, 1v/1.25 s, 1v/1.5s
Number of vehicles Scenario #1 1894/1596/1276
Number of vehicles Scenario #2 2601/2096/1750
Simulation area Scenario #1 700 m × 400 m
Simulation area Scenario #2 500 m × 600 m
Number of RSUs 3

Network Parameters

Network type WAVE/802.11p

Frequency 5.890 GHz
Channel CCH (Control)
Channel Bandwidth 10 MHz
Transmission power − 89 dBm
Data Rate 18 Mbps
Antenna type Omnidirectional antenna

7. Simulation environment

With the purpose of testing the collaborative system, two different

scenarios were implemented. These implementations were performed
using Veins, a hybrid VANET simulator which makes use of the network
simulator OMNet++ and SUMO as a traffic simulator. We chose this
simulator not only because of its simulation characteristics as a whole,
but also because of the possibilities that SUMO offered to create net­
works based on real scenarios. The main parameters used in the simu­
lations of the scenarios chosen are shown in Table 3. These parameters
are not only related with the simulation environment, but also with the
vehicular network.
Furthermore, the origins and destinations of the routes followed by
the vehicles are generated randomly. During their generation we aim to
achieve that trips start and end at the fringe of the network. Also, a
shortest path computation mechanism establishes the path followed by
Fig. 7. Scenario 2 – Peripheral Area of the city of Gijón.
the vehicles. When needed, according to the classification of vehicular

Fig. 6. Scenario 1 – Centre of the city of Gijón.

7
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

7.1. Simulation scenarios the module in charge of emulating the application layer of the vehicles
will run two of the main parts of the system. Firstly, the functions
When choosing the real-based employed scenarios, their location responsible for the delivery and reception of the WSMs. Secondly, the
was taken into account, since they had to belong to an urban area, as method from which the function that contains the fuzzy inference sys­
well as the topology of their streets. It was necessary for them to be as tem is going to be invoked and, depending on its output, will make a
different as possible with the aim of checking the correct functioning of decision. The latter method is executed in every time step of the simu­
the mobility plan. Taking these considerations into account, the two lation, so that the vehicle can react to the changing traffic conditions.
chosen scenarios belong to the urban area of the city of Gijón, in Astu­ Regarding the RSU application layer, the module used for this pur­
rias, Spain. The first scenario (Scenario #1), shown in Fig. 6, is located in pose does not include some of the methods used in the same layer of the
the centre of the city while the second scenario (Scenario #2), which can vehicles since RSUs are not going to make decisions and their actions are
be seen in Fig. 7, belongs to a peripheral area. limited. They only need to have the functions necessary to process the
The different locations of these scenarios explain their unique char­ WSMs in such a way that they can detect their type and, depending on
acteristics. The first scenario is located in an old part of a typical Eu­ this type, forward or not the messages to the vehicles within their
ropean city. It is a complex area in which the roads are narrow and there coverage areas.
is an asymmetric disposition of the buildings. Besides, it also includes a In order to carry out the experiments, the routes of the vehicles need
roundabout that eases the change of direction of the vehicles. This sit­ to be generated. Therefore, to this end, the script randomTrips.py,
uation is quite different in the case of the second scenario, in which there included among the tools offered by SUMO, was used. This script per­
are wide avenues and most of the roads have multiple lanes, thanks to a mits to create source and destination edges uniformly at random for the
modern and better urban planning. The urban design in this second vehicles. A fringe factor of 20 allows us to increase the probability that
scenario is similar to the design found in cities around the world. We trips start/end at the fringe of the network. Also, in order to ensure that,
chose this second scenario to perform experiments with a geographical for every vehicle, the point of arrival can be reached from the point of
setting which may be found in any world city, allowing us to extrapolate departure, invalid (origin, destination) tuples are automatically dis­
the results to other places. carded. The output of this script is fed into the DUAROUTER module of
Apart from the differences that the roads of both scenarios present, SUMO, to calculate the routes between each origin and destination pair
the same happens with the buildings as they are not only more asym­ using a shortest path algorithm. When necessary the default “car-
metrically distributed in the centre of the city but are also taller. This following” mobility model implemented in SUMO is used [50]. When a
fact is especially relevant when choosing the location of the RSUs in the car needs to avoid a traffic jam, routes are changed using the TraCI
scenario since it is important that they are located in such a way that module of Veins, so that the street in which the traffic jam is located, is
they are able to ease the V2I communications. Due to this, when avoided.
distributing the RSUs in the scenario, in addition to their coverage area, In spite of the possibility of specifying some of the characteristics of
it is also important to take into account the height at which they are the vehicles, we decided to maintain the default values offered by the
located so the surrounding buildings do not disturb the communications. simulator. It must be considered that most of the vehicles that make up
Three RSUs have been deployed in both scenarios to cover their urban traffic are cars and usually with similar physical characteristics.
whole area. Regarding their height, this parameter does not represent a Besides, although the presence of motorcycles is also common, accord­
problem in the second scenario where, as mentioned above, the sepa­ ing to traffic legislation their behaviour regarding traffic should be the
ration between the blocks of buildings is considerable and they are not same as that of cars, so their consideration would not have any reper­
very high. However, in the first scenario we have carefully chosen their cussion in this study.
height in order to guarantee that no building in the nearby area impedes As mentioned above, since Veins is a hybrid simulator, it is possible
the communications. In order to place the RSUs, it was also important to to visualize it either by using the network simulator OMNet++ or by
consider their coverage area so, apart from choosing a location suitable making use of the graphic interface of SUMO, (SUMO-GUI). Thanks to
to cover the whole scenario, their coverage areas did not overlap, this, it is possible to use OMNet++ to visualize the exchange of messages
impeding the interchange of messages between each other. performed by the vehicles in detail, using SUMO-GUI at the same time to
It is important to keep in mind that despite the possibility of calcu­ view the movement of the cars in the scenario. OMNet++ offers the
lating the transmission range based on the transmission power of the possibility of introducing visual effects in the visualization tool during
antennas and the sensitivity of the radios, parameters that can be set in the simulation (Tkenv). In order to check whether the vehicles were
OMNet++, there is no guarantee that the vehicles located within this inferring traffic conditions correctly, we used colour codes to identify
range are going to receive the messages properly since transmission also the different traffic levels in the scenario. Thus, when vehicles detected
depends on path loss and fading effects. that the traffic level was low, it was indicated by using a green circle,
when it was medium, yellow and in the case it was low or very low, red
8. Implementation and violet respectively. Moreover, to indicate that the vehicles had
received the messages correctly, the colour of these circles was also
Once both locations had been chosen, maps were exported from altered on their reception. They turned pink in the case of receiving a
OpenStreetMap and transformed into networks that could be handled by “Traffic” WSM and orange, in the case of receiving an “Alert” WSM (see
SUMO. This was carried out using NETCONVERT, one of the applica­ Fig. 8). Also, message interchanges were presented in the global view of
tions included in the SUMO software package. SUMO-GUI, as shown in Fig. 9.
It is worth mentioning that before the importation of the maps, they
were edited with JOSM, a software that allowed us to delete wrong 9. Performance evaluation
lanes, unconnected lines, as well as other elements that were included in
the maps that had no relevance for creating the SUMO network. In order to check the effectiveness of the system designed, two sit­
We also used the OMNet++ network simulator. This simulator is uations were taken into account for each scenario: one without using the
extensible, so it is not only possible to add the necessary modules that collaborative system (i.e., the default behaviour) and another using the
allow the use of Veins, but it is also possible to create new modules in collaborative system. By comparing both situations it is possible to check
order to add complexity to the created scenarios. the improvements achieved with the collaborative system. Moreover, a
In the case of this study, new modules were created to implement the total amount of 12 simulations were carried out for each scenario. Six
smart collaborative system. These modules have allowed us to adapt the simulations were performed without using the collaborative system and,
application layers of the vehicles and the RSUs to our needs. In this way, therefore, without interchanging WSMs. Another six simulations were

8
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Fig. 8. Simulation view in Tkenv.

Fig. 9. Global view in SUMO-GUI.

performed using the collaborative system. When the collaborative sys­ randomTrips.py script to adjust the time interval used to create the cars
tem was used, all the vehicles were equipped with the designed func­ in the simulator, as explained in Section IV. Three vehicle creation rates
tionality and, thus, were allowed to make decisions depending on the were considered in the experiments. The first of them, one vehicle per
level of traffic detected and to collaborate with each other in order to second (1 v/s), allowed us to emulate a high density of traffic in the city.
avoid traffic congestion. The second, one vehicle each second and a quarter (1 v/1.25 s), allowed
Apart from considering the aforementioned situations for each sce­ us to create a considerable amount of traffic. Finally, the third was to
nario, we have also performed experiments with different traffic den­ create one vehicle every second and a half (1 v/1.5 s), to simulate a low
sities. Thus, we have considered different vehicle creation rates, having density of traffic.
more or less cars in the simulations and, accordingly, better or worse In both situations and scenarios, we allow all the vehicles to finish
traffic conditions. For this purpose, we have used the aforementioned their routes. Thus, the departure of the last car is scheduled to happen

9
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

3000 s after the beginning of the experiment, leaving enough time for it Table 5
to reach its destination. Number of vehicles that achieve maximum speed in Scenario #1 for different
Once the simulation has been launched, statistics are collected from creation rates (v/s).
both the vehicles and the RSUs. In order to compare both situations Rate of Vehicle Creation Default behaviour Collaborative system
(with or without the collaborative system) in both scenarios, we analyse x x
the differences in minimum, maximum and average speeds of the ve­ 1 v/s 153 (8.1 %) 1843 (97.3 %)
hicles, CO2 emissions and acceleration of the vehicles and the distance of 1 v/1.25 s 133 (8.3 %) 1571 (98.4 %)
their routes. It is worth mentioning that the amount of emissions of CO2 1 v/1.5 s 93 (7.3 %) 1254 (98.3 %)
of the cars is calculated thanks to a model developed by Capiello et al.
[51] which employs the speed and acceleration of the vehicles.
With the aim of determining whether the obtained data gathered for Table 6
both situations show significant differences or not, we began by studying Average and standard deviation values for minimum, maximum and average
their normality and homoscedasticity. In order to check the normality of speeds (m/s) in Scenario #1 for different vehicle creation rates (v/s).
data we performed Shaphiro–Wilk tests and to evaluate their homo­ Rate of Vehicle Creation Default behaviour Collaborative system
scedasticity we used Levene’s tests. Once the results of those tests were x σ x σ
obtained, since all the data sets turned out not to be normal and ho­ Minimum speed

moscedastic, Kruskal–Wallis tests using a 0.01 significance level were 1 v/s 0.056 0.103 0.075 0.113
employed in order to check if there is a statistically significant difference 1 v/1.25 s 0.077 0.114 0.086 0.117
1 v/1.5 s 0.083 0.115 0.098 0.121
between the usage or not of the collaborative system. If the p value of the
Maximum speed
test is lower than 0.01, there are statistically significant differences, 1 v/s 13.788 0.750 13.811 0.698
while if the p value is equal to or greater than 0.01, we fail to reject the 1 v/1.25 s 13.816 0.635 13.835 0.578
null hypothesis. The latter indicates that we do not have sufficient evi­ 1 v/1.5 s 13.816 0.657 13.824 0.646
dence to affirm that there is a statistically significant difference between Average speed
1 v/s 6.446 2.502 6.956 2.748
the situations being compared.
1 v/1.25 s 7.243 2.485 7.639 2.578
1 v/1.5 s 7.516 2.542 7.849 2.658

9.1. Scenario #1 – centre of Gijón city

This scenario, as explained before, was based on an area of the centre Table 7
Average and standard deviation values for acceleration (m/s2) in Scenario #1
of Gijón. The average distance covered by vehicles is showed in Table 4.
for different vehicle creation rates (v/s).
Another relevant result, is the number of vehicles that achieve the
maximum speed in the scenario, depending on whether they use the Rate of Vehicle Creation Default behaviour Collaborative system
x x
collaborative system or not. As we can see in Table 5, the percentage of σ σ

vehicles that achieve the maximum speed is much higher when the 1 v/s 0.215 0.299 0.243 0.314
collaborative system is used. Moreover, Table 6 shows the mean of the 1 v/1.25 s 0.236 0.288 0.254 0.298
1 v/1.5 s 0.253 0.300 0.275 0.313
minimum and maximum values of speed for all the cars that take part in
the simulation, collected when varying the traffic in the area.
It can be noticed that, although the length of the routes increase
trying to avoid congested streets, cars are able to move faster when using Table 8
Average and standard deviation values for CO2 emissions (g/s) in Scenario #1
the collaborative system designed. This can also be seen if we compare
for different vehicle creation rates (v/s).
the average speed of the vehicles, also shown in Table 6. Kruskal–Wallis
tests show that these differences are statistically significant for all the Rate of Vehicle Creation Default behaviour Collaborative system
x x
vehicle creation rates with p values < 0.01, proving the positive effects σ σ

caused by the collaborative system. Moreover, there are slight changes 1 v/s 241.695 119.871 225.418 125.261
in the acceleration of the vehicles. Table 7 shows average accelerations 1 v/1.25 s 219.192 103.542 205.378 101.741
1 v/1.5 s 207.367 97.177 196.828 99.211
for the three vehicle creation rates considered in the experiments. As in
the case of speeds, Kruskal–Wallis tests show statistically significant
differences for all the vehicle creation rates with p values < 0.01. These characteristics of the scenario. It is important to remember that it is
results are obtained thanks to the improvements in the flow of traffic located in the centre of a city, which implies that cars travel most of the
achieved by the system. time on one-way streets with a high number of intersections. Thus, they
Thanks to these parameters, it is possible to work out the total tend to suffer severe traffic problems when traffic jams appear and so
amount of CO2 emissions generated by the vehicles, according to Cap­ increase the generation of pollutants.
iello et al. [51]. This parameter, whose values for the simulations per­ Regarding the messages interchanged in the network, the number
formed are shown in Table 8, gives an idea of the pollution caused by the matches up to the expectations. In the scenario with the highest traffic,
traffic. As can be seen in this table, the emissions are significantly vehicles send, on average, 1.01 messages while they receive 10.78.
reduced when the collaborative system is used (up to 6.73 %). This However, when the lowest vehicle creation rates are used, the average of
difference in the emissions can also be explained by the particular messages sent by the vehicles is 1 and they receive an average of 5.72
messages, almost 50 % less. The reduction of the messages interchanged
Table 4 in accordance with the vehicle creation rate is also reflected in the
Average distance covered (meters) by vehicles in Scenario #1 for different messages sent by RSUs. Thus, the average number of messages sent with
vehicle creation rates (v/s). the highest rate is 203.33, 154.67 with the medium rate and 136.67
Rate of Vehicle Creation Default behaviour Collaborative system messages with the lowest.
x x

1 v/s 594.628 619.363 9.2. Scenario #2 – peripheral area of gijón city

1 v/1.25 s 580.953 598.302
1 v/1.5 s 561.158 578.230
This second scenario covers a peripheral area of Gijón. As mentioned

10
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Table 9 Table 13
Number of vehicles that achieve maximum speed in Scenario #2 for different CO2 emissions (g/s) for Scenario #2 for different vehicle creation rates (v/s).
creation rates (v/s). Rate of Vehicle Creation Default behaviour Collaborative system
Rate of Vehicle Creation Default behaviour Collaborative system x σ x σ
x x
1 v/s 273.564 160.131 240.013 122.299
1 v/s 229 (8.8%) 2557 (98.3%) 1 v/1.25 s 226.285 110.881 224.644 110.826
1 v/1.25 s 187 (8.9%) 2054 (98.0%) 1 v/1.5 s 212.931 90.456 218.229 107.053
1 v/1.5 s 151 (8.6%) 1717 (98.1%)

statistically significant with p values < 0.01.

As seen in the first scenario, the acceleration values gathered when
Table 10
the collaborative system is used are slightly higher, as shown in
Average distance covered (meters) by vehicles in Scenario #2 for different
vehicle creation rates (v/s).
Table 12.
Unlike what happened to the simulation data obtained in the first
Rate of Vehicle Creation Default behaviour Collaborative system
scenario, on this occasion there are not so many differences in pollution
x x
values if we compare the default behaviour with the collaborative sys­
1 v/s 690.699 700.171 tem. As shown in Table 13, CO2 emissions are lower if we use the
1 v/1.25 s 647.759 689.643
1 v/1.5 s 648.511 688.377
collaborative system when the highest and the medium creation rates
are used. This is also shown in Fig. 10, which represents the CO2 emis­
sions generated by the cars in scenario #2, when the 1v/s vehicle cre­
ation rate is used. Nevertheless, when the lowest creation rate is
Table 11
employed, there is a slight increase in CO2 emissions when the collab­
Average and standard deviation values for minimum, maximum and average
speeds (m/s) in Scenario #2 for different vehicle creation rates (v/s).
orative system is used. This can be explained by taking into account the
topology of the scenario. Since the scenario covers a peripheral area of
Rate of Vehicle Creation Default behaviour Collaborative system
the city, there is a higher separation between buildings, so cars have to
x σ x σ
Minimum speed travel longer distances in order to avoid congested streets and reach
their destination.
1 v/s 0.083 0.291 0.088 0.117
1 v/1.25 s 0.106 0.122 0.106 0.121
The use of the Kruskal–Wallis test on these values shows that when
1 v/1.5 s 0.114 0.123 0.116 0.123 the highest creation rate is employed, the emission values registered are
Maximum speed significantly different with p values < 0.01, so the collaborative system
1 v/s 13.808 0.682 13.829 0.603 clearly improves the situation. Nevertheless, when the lowest creation
1 v/1.25 s 13.801 0.705 13.814 0.695
rate is used, the p value obtained is 0.8486 (>0.05) which reflects that
1 v/1.5 s 13.809 0.714 13.817 0.708
Average speed the values do not present significant differences.
1 v/s 7.006 2.919 7.791 2.612 Regarding the communications, the collected data are consistent
1 v/1.25 s 7.985 2.618 8.311 2.478 with those obtained in the first scenario. The lower the traffic, the lower
1 v/1.5 s 8.370 2.325 8.601 2.348 the number of messages exchanged. Vehicles send, on average, 1.010
when one vehicle per second is created, 1.001 with the chosen medium
before, several vehicle creation rates are used in order to test the rate and 1, when the lowest rate is used. The same happens with the
collaborative system in this scenario. The same vehicle creation rates average number of messages received. When the traffic density is very
used in the previous scenario. The reason for choosing the same rates high, this number is 43.96, when it is medium, 31.01 messages and
again, despite the differences of topology in both scenarios, is that it when it is low, 24.43. Therefore, 55.58 % fewer messages are generated
makes it their comparison easier, while also checking the effects of the if we compare the highest and the lowest traffic densities.
different topologies on traffic. The number of vehicles achieving the Since this scenario is more prone to road congestion, it is especially
maximum speed is shown in Table 9. The results are similar to those important for the RSUs not only to be correctly located in order to cover
obtained in scenario #1. Also, the average distance covered by the ve­ the whole area but also to be close to the most problematic points. In this
hicles is shown in Table 10. As shown in the table, routes are longer way, the highest number of problems that might arise is reflected in an
when the collaborative system is used. This is caused by the system increase in the number of messages sent by the RSUs. Thus, when the
altering the initial routes of the vehicles when congestion is detected. highest vehicle creation rates are used, these devices send, on average,
Table 10 shows a maximum difference of 42 m with a 1v/1.25 s vehicle 396.33 messages while, in the case of the lowest rates, 348.67 messages
creation rate. are sent. These values are 94% and 155% higher than the same situa­
The gathered speed values, shown in Table 11, present the same tions in the first scenario.
pattern observed in the first scenario. The ability of the vehicles to react
to the traffic conditions when the collaborative system is used achieves 10. Discussion and limitations
an improvement in their speed. This not only affects the minimum and
maximum values but, as in the first scenario, the average speed of the Thanks to the simulation experiments carried out, we have been able
vehicles. Again, Kruskal–Wallis tests show that the differences are to demonstrate the effectiveness of the proposed system. By comparing
the behaviour of traffic with and without the collaborative system, using
several traffic densities (i.e., vehicle creation rates) on two different
Table 12 urban scenarios, we have checked the usefulness of the system designed.
Average and standard deviation values for acceleration (m/s2) in Scenario #2 When the collaborative system is used, most of the vehicles travelling
for different vehicle creation rates (v/s). in both scenarios are able to achieve the maximum legal speed, inde­
pendently of the volume of traffic and urban design. The worst condi­
Rate of Vehicle Creation Default behaviour Collaborative system
x σ x σ tions are met when one vehicle per second is created in the centre of the
city. With these conditions, an average of 97.3 % of the vehicles achieve
1 v/s 0.187 0.242 0.209 0.256
1 v/1.25 s 0.228 0.262 0.234 0.280 said speed. Nevertheless, when the system is not used, only 8.1 % of the
1 v/1.5 s 0.224 0.246 0.227 0.248 vehicles achieve this speed. Fig. 11 shows the average speeds of the

11
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Fig. 10. . CO2 emissions in Scenario #2 for a 1/v vehicle creation rate.

Fig. 11. Average speeds in Scenarios #1 and #2.

vehicles in both scenarios. When the system is used, average speed im­ Regarding the generation of pollutants, Fig. 12 shows the average
provements range from 2.75 % when one vehicle is created every 1.5 s in CO2 emissions generated in both scenarios. Pollutants are significantly
the city outskirts to 11.20 % when one vehicle per second is created in reduced when the collaborative system is used, achieving an improve­
the same urban environment. Speed improvements range from 4.43 % to ment to 12,26 % when 1 vehicle per second is created in the second
7.91 % in the city centre. The worse traffic conditions are, the higher the scenario. The greatest improvements are obtained when the volume of
increase in average speeds. This result is logical since the traffic flow traffic is high because of an increase in the number of vehicles or
improves, making it possible for vehicles to maintain higher speed because of the design of the urban topology (i.e., there are traffic jams)
values for most of their journey. The increase of speed also has an impact The centre of the city is full of one-way streets and intersections, so
on the acceleration of the vehicles since the changes of speed are vehicles tend to suffer severe problems when traffic jams appear,
translated in a slight increase in this parameter as well. increasing the pollutants generated. In this case, the system allows us to
As a result of the decisions performed by the collaborative system, reduce CO2 between 5 % and 6,73 %. Nevertheless, when there are no
average routes are up to 6.46 % longer. When congestion is detected, the traffic jams in a peripheral area of the city, the increase in the speed of
system changes the initial routes of the vehicles, forcing them to travel cars and in the distances travelled because of the decisions made by the
longer distances than those initially planned. Nevertheless, although the collaborative system get to produce a slight increase in the pollutants
length of the routes increase trying to avoid congested streets, cars are generated (not statistically significant). When traffic density is low, the
able to move faster when using the collaborative system. results between using and not using the collaborative system are very

12
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Fig. 12. Average CO2 emissions in Scenarios #1 and #2.

close. The reason can be attributed to the low number of traffic jams work has demonstrated that this policy is not very effective [17]. Other
when the density of cars is low. Furthermore, there are fewer changes in cities have created low emission zones -LEZ- to avoid certain vehicles to
the routes of the vehicles when the system is employed under low traffic enter specific areas of the city [53,54]. In some cases, those restrictions
conditions. Under these conditions, the collaborative system does not are not applied to cars (only to other heavier vehicles) or even if they are
take effect. applied, drivers may enter those areas paying a fee [53] In other occa­
The conclusions are in line with our expectations since the lower the sions, certain vehicles may enter without restrictions depending on their
traffic density in the network, the closer the results are between using engine, the activity of the owner, the route which will be followed or
and not using the collaborative system. Such behaviour is explained by other aspects [54] These policies clearly improve the situation but they
the decrease in the number of cars that are travelling in the scenarios, do not ultimately finish with traffic congestion and the generation of
since this brings with it a reduction in the number of traffic jams. This pollution, thus, our solution may help those policies to be more effective.
reduction means that vehicles make a lower use of the collaborative In view of the obtained results, it can be concluded that the smart
system designed to reduce the traffic congestion because they will detect collaborative mobility system designed meets expectations since it al­
low and very low traffic. Therefore, most of their decisions will be lows to decrease the number of traffic jams by improving the traffic flow
translated into changes of speed instead of changes of route. The and, at the same time, decreasing the generation of pollution. The sys­
collected experimental values can be used to prove the positive re­ tem is decentralized and autonomous, therefore, there are no single
percussions that these speed changes have on the traffic flow. Never­ points of failure nor the need of Internet connectivity. The usage of fuzzy
theless, since in the original situation there are hardly any traffic jams logic has allowed us to design a low-resource consuming solution, in
and most of the vehicles follow the same route, the improvements are contrast with other artificial intelligent techniques and without the need
less noticeable than when the traffic density is higher. It is necessary to of extensive training processes or massive datasets.
emphasize that we have demonstrated that the system is able to improve Our approach is novel as it is based on a totally autonomous and
traffic conditions reducing, at the same time, the stress of drivers and distributed approach for urban scenarios, based on the collaboration
pollution. Nevertheless, the exact grade of improvement depends on the between vehicles through the use of vehicular communications. A
complexity of the urban environment and the density of vehicles in the comparative study with other similar solutions follows. Tientrakool
scenario. A sudden and massive deployment of vehicles in a complex et al. [14] focus their study on collision avoidance in vehicles driving on
urban scenario cannot be handled simply because of the existence of a motorway, whereas our solution has been designed for urban traffic.
physical limitations. Kitwiroon et al. [16] achieve improvements of up to 11 % in vehicular
The system may help governments to reduce congestion and pollu­ emissions. Even though these values are similar to the improvements in
tion in urban areas. In recent years, public administrations have been pollutant emissions obtained with our solution, they have to reduce
carrying out different interventions to reduce pollution. Some of these heavy-duty vehicles a 20 %. Thus, they are restricting the most-polluting
policies are general and mainly imply a movement toward zero or low- vehicles in the road. Similarly, Mahmod et al. [18] are able to reduce
emission vehicles [52]. Other policies are primarily local and have been CO2 emissions up to 23 %, but by reducing the traffic a 20 %. In our
designed to improve traffic and pollution specially in the centre of the study, we decrease car emissions by up to 12.27 % but no traffic needs to
cities. General measures have been usually planned to be deployed on a be restricted or reduced. Knorr et al. [23] aim their study at avoiding
long term. For instance, the plan of the European Union to achieve traffic jams and, although they are able to improve traffic efficiency,
climate neutrality includes restrictions on new passenger cars and new their solution has been designed for highway roads only. The solution
light commercial vehicles from 2030 onwards. Nevertheless, vehicles proposed by Ribeiro et al. [25] is focused on communication aspects and
sold before that date will continue to be used for years. This same plan relies on a central system as opposed with our study, in which we present
also includes a 90 % reduction in transport emissions, but the deadline is a holistic solution. Furthermore, they perform experiments with IEEE
2050 [6]. Our solution can bring immediate benefits while those 802.11 b/g and report high association delays, whereas in our study we
long-term policies finish to be deployed. It is not an alternative to those employ IEEE 802.11p communications. Jayapal & Roy [30], present a
policies, but a complement. Regarding local policies, major cities around traffic congestion detection and dissemination system but relays on a
the world have been applying certain restrictions to traffic in recent central system, does not use vehicular communications and it does not
years. Some cities have reduced speed limits in urban areas, but previous provide any improvement metric. Dimitrou et al. [32] describe an

13
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

adaptive hybrid fuzzy rule-based system to predict the evolution of exchange control information. All these public vehicles may relay con­
traffic in an urban environment, but they do not take action to improve trol messages when they are stationary, or they may even implement a
traffic conditions. Naja & Matta [35] also present a method to avoid “store-carry-forward” communication model. These actions may be
traffic congestion, but their solution is tailored to motorways and not to included in the interventions targeted by public administrations at
cities. Ranjita and Acharya [41] proposed a method to detect congestion reducing pollution and improving urban mobility.
and to avoid it by proposing an alternative route. Although these ideas It is also necessary to take into account that there may be legal re­
are also part of our design, their study is based on a central system and strictions in the deployment of the solution since the communications of
has been specifically designed for emergency vehicles only. Stolfi and the system are based on the IEEE 802.11p standard and this technology
Alba [43] propose a rerouting mechanism for vehicles to optimize the uses a licensed frequency band. In fact, although this technology has
distribution of traffic. Their solution is infrastructure dependant as been specifically designed for vehicular communications, there has not
opposed to our solution which is autonomous. Also, their approach is been a significant deployment of this type of solutions. Although it is
different because they try to balance traffic by rerouting all the vehicles, optimal for the type of communications we perform, the fact that it is not
whereas in our case we only reroute vehicles when needed, adjusting available in current vehicles hinders the adoption of the proposed so­
speeds depending on the status of traffic. Finally, they rely on lution. On the hand, the fact that solutions of this type are proposed and
non-vehicular communication technologies which may be problematic used, may promote the usage of the technology.
in certain situations, and they do not report pollution improvements.
Even though the results are promising, this study has certain limi­ 11. Conclusions and future work
tations. Firstly, all the results are based on simulation experiments.
There are certain limitations inherent to the simulation environments In this paper we have presented a novel system for improving traffic
which may not fully replicate real-world complexities. For instance, the conditions in urban areas reducing, at the same time, the generation of
routes followed by the vehicles are generated randomly. Starting and pollution when combustion vehicles are used and, as a side benefit, the
ending points of every trip are generated in the fringe of the scenarios stress of the drivers. Furthermore, an improved flow of traffic may also
and a shortest-path algorithm is used to calculate the path. This does not imply an optimised usage of batteries in an electric vehicle scenario. Our
reflect how real traffic behaves, as certain avenues and streets usually solution can bring immediate benefits to society while the policies
concentrate most of the traffic because of the urban design. Moreover, designed by public administrations finish to be deployed.
trips do not necessarily start and finish in any preestablished border (e. The system is based on vehicular communications and fuzzy logic.
g., the boundaries of a suburb). Also, cars behave using the default “car- Given the obtained results, we can affirm that fuzzy logic is a better
following” mobility model implemented in SUMO when needed. While approach to determine and correct traffic jams than deterministic or
this approach is useful to perform simulations, it does not realistically random methods, providing a new system for traffic control. Moreover,
imitate the behaviour of real drivers. Finally, another limitation of the the designed system has proved the effectiveness of collaborative solu­
simulations carried out is that vehicle creation rates are fixed. This has tions in avoiding traffic jams, confirming the outcomes of previous work.
also been convenient because it has allowed us to examine the behaviour The deployment of the designed collaborative system using VANETs
of the system by gradually increasing the load in the scenarios with with both vehicles and infrastructure, and Fuzzy Logic may reduce
changing speeds, but it does not accurately reflect how real vehicles pollution in a city up to 12.27 % reducing, at the same time, the stress of
enter and exit a certain urban area. In order to fully check the efficacy of the drivers as stated by Gulian et al. [12] and optimizing the autonomy
the collaborative system designed it would be necessary to carry out a of electric vehicles according to Yan et al. [11].
pilot test in a real-world setting with real drivers. The simulations have been performed using realistic scenarios, and
When we carried out the experiments, we assumed that when the they have also shown that the designed system is effective not only in the
collaborative system is used, all the vehicles are equipped with the centre of a city, but also in peripheral areas with different traffic den­
designed functionality. In a mixed scenario with standard vehicles sities. Moreover, the fact that the urban designs chosen resemble both
travelling together with vehicles using the system, only the latter would the typical structure of an old European city and the design of a modern
benefit from the decisions made. Vehicles equipped with the system will city which may be found in any continent, has allowed us to demonstrate
continue taking decisions based on their speed, the time they stay sta­ that the system may be effective in any geographical setting. Never­
tionary, and the messages sent by similar cars, but the rest of vehicles theless, the experiments have only been carried out in a simulation
will continue with the default behaviour. Thus, the results in terms of environment. Therefore, in order to fully check the efficacy of the
pollutants reduced would not be so beneficial. But it is necessary to collaborative system designed, as future work it would be interesting to
consider that this is a decentralized collaborative system and needs the carry out a pilot test in a real-world setting and apply the solution in a
vehicles to exchange information. If the density of vehicles using the real scenario with real drivers. We aim to perform such pilot test by
system is very low, the solution is not effective because most of those deploying the system in professional bus fleets, thanks to our collabo­
vehicles would not be able to act proactively, as they are not receiving ration with the ADN Mobile Solutions company, similarly to what was
messages generated by other vehicles already entering a traffic jam. The done in the UrVAMM project [55]. In fact, the first actual outcomes of
lack of vehicles using the system may be alleviated by deploying a great the system may be obtained in said environment by helping to deal with
number of RSUs which, on the other hand, would help to deal with range congestion situations in heavy-loaded bus routes. To develop the model
limitations and adverse environmental conditions. The availability of in a real environment we would need devices to work as either On-Board
these systems is another assumption of the study. This may be feasible if Unit (OBU) or RSU stations. We have carried out a preliminary revision
current public infrastructure is upgraded for said purpose (e.g., street­ of commercial vehicle side devices (OBU) and found several options
lights and traffic lights) Nevertheless, considering the recent controversy used in the current market such as CohdaWireless MK6 [56] or Unex
about the deployment of 5 G infrastructure this may be definitely OBU-352 [57]. Possible solutions for RSU devices are the same Cohda­
polemical and social acceptance is not guaranteed. A not so noticeable Wireless MK6 [56] and Unex RSU-352 [58]. An implementation of our
option would be to deploy the system in public transport vehicles. fuzzy inference system could be deployed in these devices, in order to
Although they need to follow preestablished routes independently of the take decisions and generate the messages we need.
volume of traffic, they can behave like RSUs and, thus, operate as ex­ It is also worth researching into the acceptance of this type of systems
change points. Deploying the system in other types of public vehicles by drivers. New intelligent transportation systems may be used to reduce
such as shared scooters or electric bicycles would extend the model by pollution, but it is not clear whether users like them or not. Subjective
adding V2P (Vehicle-to-Pedestrian) communications. These vehicles evaluations may be a good option to check whether they accept them or
may operate as well as infrastructure elements, helping vehicles to not. This is something we would like to research with our pilot tests.

14
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

Even though they will be carried out with professional drivers in an [14] P. Tientrakool, Y.C. Ho, N.F. Maxemchuk, Highway capacity benefits from using
vehicle-to-vehicle communication and sensors for collision avoidance, in:
urban transport environment, their feedback may be useful to obtain
Proceedings of the IEEE Vehicular Technology Conference (VTC Fall), San
preliminary conclusions on the acceptance of the system. Francisco, USA, 2011, pp. 1–5.
Furthermore, in light of the current growth in the number of alter­ [15] S. Hallé, J. Laumonier, B. Chaib-Draa, A decentralized approach to collaborative
native means of transport in urban areas, such as bicycles, it could also driving coordination, in: Proceedings of the 7th International IEEE Conference on
Intelligent Transportation Systems, Washington, USA, 2004, pp. 453–458.
be interesting to work on the design of collaborative systems with the [16] N. Kitwiroon, Y. Yu, H. Mao, R.S. Sokhi, Traffic Management impact in London,
goal of facilitating the integration of these other means into the urban using OSCAR, in: Air Quality Assessment For Europe: From Local to Continental
traffic, improving several aspects such as the safety of their users. Scale (Air4EU), Deliverable D7.1, Part, 8, 2007.
[17] L.I. Panis, C. Beckx, S. Broekx, I. De Vlieger, L. Schrooten, B. Degraeuwe,
L. Pelkmans, PM, NOx and CO2 emission reductions from speed management
CRediT authorship contribution statement policies in Europe, Transp. Policy (Oxf) 18 (2011) 32–37, https://doi.org/
10.1016/j.tranpol.2010.05.005.
[18] M. Mahmod, B. Van Arem, R. Pueboobpaphan, R. De Lange, Reducing local traffic
José Antonio Sánchez: Conceptualization, Investigation, Software, emissions at urban intersection using ITS countermeasures, IET Intell. Transport
Writing – original draft. David Melendi: Conceptualization, Investiga­ Syst. 7 (2013) 78–86, https://doi.org/10.1049/iet-its.2011.0222.
[19] M. Mukhtaruzzaman, M. Atiquzzaman, Clustering in vehicular ad hoc network:
tion, Supervision, Writing – original draft, Writing – review & editing.
algorithms and challenges, Comput. Electr. Eng. 88 (2020) 106851, https://doi.
Roberto García: Formal analysis, Validation. Xabiel G Pañeda: org/10.1016/j.compeleceng.2020.106851.
Funding acquisition, Methodology, Project administration, Resources. [20] M. Aissa, B. Bouhdid, A. Ben Mnaouer, A. Belghith, S. AlAhmadi, SOFCluster:
safety-oriented, fuzzy logic-based clustering scheme for vehicular ad hoc networks,
Víctor Corcoba: Software, Visualization. Dan García: Software,
Trans. Emerg. Telecommun. Technol. 33 (2022) e3951, https://doi.org/10.1002/
Visualization. ett.3951.
[21] Y. Toor, P. Muhlethaler, A. Laouiti, A. De La Fortelle, Vehicle ad hoc networks:
applications and related technical issues, IEEE Commun. Surv. Tutor. 10 (3) (2008)
74–88, https://doi.org/10.1109/COMST.2008.4625806.
Declaration of competing interest [22] C.D. Wang, J.P. Thompson, Apparatus and method for motion detection and
tracking of objects in a region for collision avoidance utilizing a real-time adaptive
The authors declare that they have no known competing financial probabilistic neural network. Google Patents. https://www.google.com/patents/
US5613039, 1997 (Accessed 26 July 2023).
interests or personal relationships that could have appeared to influence
[23] F. Knorr, D. Baselt, M. Schreckenberg, M. Mauve, Reducing Traffic Jams via
the work reported in this paper. VANETs, IEEE Trans. Veh. Technol. 61 (8) (2012) 3490–3498, https://doi.org/
10.1109/TVT.2012.2209690.
Data availability [24] M.R. Hafner, D. Cunningham, L. Caminiti, D. Del Vecchio, Cooperative collision
avoidance at intersections: algorithms and experiments, IEEE Trans. Intell.
Transport. Syst. 14 (3) (2013) 1162–1175, https://doi.org/10.1109/
No data was used for the research described in the article. TITS.2013.2252901.
[25] J.G. Ribeiro, M.E.M. Campista, L.H.M. Costa, COTraMS: a collaborative and
opportunistic traffic monitoring system, IEEE Trans. Intell. Transport. Syst. 15 (3)
(2014) 949–958, https://doi.org/10.1109/TITS.2013.2291996.
References [26] C. Dannheim, M. Maeder, C. Icking, Collaborative traffic technologies and their
effect on driving safety, in: Proceedings of the 16th Int. IEEE Conference on
Intelligent Transportation Systems (ITSC 2013), The Hague, Netherlands, 2013,
[1] United Nations, World Population Prospects 2022, United Nations Department of
pp. 1203–1209.
Economics and Social Affairs, Population Division, New York, 2022.
[27] N. Drawil, O. Basir, Vehicular collaborative technique for location estimate
[2] International Organization of Motor Vehicle Manufacturers, Global Sales statistics,
correction, in: Proceedings of the 2008 IEEE 68th Vehicular Technology
2019–2022. https://www.oica.net/category/sales-statistics/, 2022 (Accessed 10
Conference (VTC), Barcelona, Spain, 2008, pp. 1–5.
July 2023).
[28] M. Piorkowski, Collaborative transportation systems, in: Proceedings of the 2010
[3] International Organization of Motor Vehicle Manufacturers, 2022 production
IEEE Wireless Communication and Networking Conference (WCNC), Sydney,
statistics. https://www.oica.net/category/production-statistics/2022-statistics/,
Australia, 2010, pp. 1–6.
2022 (Accessed 10 July 2023).
[29] Y. Gu, S. Lin, N.F. Maxemchuk, A fail safe broadcast protocol for collaborative
[4] Eurostat, E-commerce statistics for individuals. https://ec.europa.eu/eurostat/stati
intelligent vehicles, in: Proceedings of IEEE 16th International Symposium on a
stics-explained/index.php?title=E-commerce_statistics_for_individuals, 2022
World of Wireless, Mobile and Multimedia Networks (WoWMoM), Boston, USA,
(Accessed 14 July 2023).
2015, pp. 1–6.
[5] R.C. Flagan, J.H. Seinfeld, Fundamentals of Air Pollution Engineering, Dover
[30] C. Jayapal, S.S. Roy, Road traffic congestion management using VANET, in:
Publications Inc., Mineola, New York, 2012.
Proceedings of the 2016 International Conference on Advances in Human Machine
[6] European Union, Regulation (EU) 2023/851 of the European Parliament and of the
Interaction (HMI), Kodigehalli, India, 2016, pp. 1–7.
Council of 19 April 2023 amending Regulation (EU) 2019/631 as regards
[31] L.A. Zadeh, The role of fuzzy logic in the management of uncertainty in expert
strengthening the CO2 emission performance standards for new passenger cars and
systems, Fuzzy Sets Syst. 11 (1–3) (1983) 199–227, https://doi.org/10.1016/
new light commercial vehicles in line with the Union’s increased climate ambition,
S0165-0114(83)80081-5.
Off. J. Eur. Union 110 (2023) 5–20. http://data.europa.eu/eli/reg/2023/851/oj.
[32] L. Dimitriou, T. Tsekeris, A. Stathopoulos, Adaptive hybrid fuzzy rule-based system
[7] A. Lakas, M. Chaqfeh, A novel method for reducing road traffic congestion using
approach for modeling and predicting urban traffic flow, Transport. Res. Part C:
vehicular communication, in: Proceedings of the 6th International Wireless
Emerg. Technol. 16 (5) (2008) 554–573, https://doi.org/10.1016/j.
Communications and Mobile Computing Conference, Caen, France, 2010,
trc.2007.11.003.
pp. 16–20.
[33] K. Zrar-Ghafoor, K. Abu Bakar, M. van Eenennaam, R.H. Khokhar, A.J. Gonzalez,
[8] C.A.R.L. Brennand, A. Boukerche, R. Meneguette, L.A. Villas, A novel urban traffic
A fuzzy logic approach to beaconing for vehicular ad hoc networks, Telecommun.
management mechanism based on FOG, in: Proceedings of the 2017 IEEE
Syst. 52 (2013) 139–149, https://doi.org/10.1007/s11235-011-9466-8.
Symposium on Computers and Communications (ISCC), Heraklion, Greece, 2017,
[34] C. Sonmez, O.D. Incel, S. Isik, M.Y. Donmez, C. Ersoy, Fuzzy-based congestion
pp. 377–382.
control for wireless multimedia sensor networks, EURASIP J. Wirel. Commun.
[9] G.P. Rocha-Filho, R.I. Meneguette, J.R. Torres-Neto, A. Valejo, L. Weigang,
Netw. 63 (2014), https://doi.org/10.1186/1687-1499-2014-63, 2014.
J. Ueyama, G. Pessin, L.A. Villas, Enhancing intelligence in traffic management
[35] R. Naja, R. Matta, Fuzzy logic ticket rate predictor for congestion control in
systems to aid in vehicle traffic congestion problems in smart cities, Ad Hoc Netw.
vehicular networks, Wirel. Person. Commun. 79 (2014) 1837–1858, https://doi.
107 (2020) 102265, https://doi.org/10.1016/j.adhoc.2020.102265.
org/10.1007/s11277-014-1961-2.
[10] P. González-Aliste, I. Derpich, M. López, Reducing urban traffic congestion via
[36] Z. Miao, C. Li, L. Zhu, X. Han, M. Wang, X. Cai, Z. Liu, L. Xiong, On resource
charging price, Sustainability 15 (3) (2023) 2086, https://doi.org/10.3390/
management in vehicular ad hoc networks: a fuzzy optimization scheme, in:
su15032086.
Proceedings of the 2016 IEEE 83rd Vehicular Technology Conference (VTC
[11] J. Yan, J. Zhang, Y. Liu, G. Lv, S. Han, I.E. Gonzalez-Alfonzo, EV charging load
Spring), Nanjing, China, 2016, pp. 1–5.
simulation and forecasting considering traffic jam and weather to support the
[37] Q. Luo, X. Cai, T. Luan, Q. Ye, Fuzzy logic-based integrity-oriented file transfer for
integration of renewables and EVs, Renew. Energy 159 (2020) 623–641, https://
highway vehicular communications, EURASIP J. Wirel. Commun. Netw. 3 (2018),
doi.org/10.1016/j.renene.2020.03.17.
https://doi.org/10.1186/s13638-017-1009-x.
[12] E. Gulian, A.I. Glendon, G. Matthews, D.R. Davies, L.M. Debney, The stress of
[38] S. Balasubramani, D.J. Aravindhar, Optimizing data transmission in VANET using
driving: a diary study, Work Stress 4 (1990) 7–16, https://doi.org/10.1080/
fuzzy logic system with link residual time, in: Proceedings of the 3rd International
02678379008256960.
conference on IoT in Social, Mobile, Analytics and Cloud (I-SMAC), Palladam,
[13] H. Youn, M.T. Gastner, H. Jeong, Price of anarchy in transportation networks:
India, 2019, pp. 761–766.
efficiency and optimality control, Phys. Rev. Lett. 101 (12) (2008) 128701, https://
doi.org/10.1103/PhysRevLett.101.128701.

15
J.A. Sánchez et al. Vehicular Communications 47 (2024) 100746

[39] F. Abbasi, M. Zarei, A.M. Rahmani, FWDP: a fuzzy logic-based vehicle weighting [48] P.G. Gipps, A behavioural car-following model for computer simulation, Transport.
model for data prioritization in vehicular ad hoc networks, Veh. Commun. 33 Res. Part B: Methodol. 15 (2) (1981) 105–111, https://doi.org/10.1016/0191-
(2022) 100413, https://doi.org/10.1016/j.vehcom.2021.100413. 2615(81)90037-0.
[40] F. Arena, G. Pau, A. Ralescu, A. Severino, I. You, An innovative framework for [49] M. Liebner, M. Baumann, F. Klanner, C. Stiller, Driver intent inference at urban
dynamic traffic lights management based on the combined use of fuzzy logic and intersections using the intelligent driver model, in: Proceedings of the 2012 IEEE
several network architectures, J. Adv. Transport. 2022 (2022) 1383349, https:// Intelligent Vehicles Symposium (IV), Alcala de Henares, Spain, 2012,
doi.org/10.1155/2022/1383349. pp. 1162–1167.
[41] R. Ranjita, S. Acharya, A Fuzzy Logic-Based Congestion Detection technique for [50] M. Behrisch, L. Bieker, J. Erdmann, M. Knocke, D. Krajzewicz, P. Wagner,
vehicular ad hoc networks, in rout, R.R., Ghosh, S.K., Jana, P.K., Tripathy, A.K., Evolution of SUMO’s simulation model. Transportation Research Board Circular,
Sahoo, J.P., Li, KC. (Eds.), Advances in Distributed Computing and Machine Transportation Research Board, 2014, pp. 1–21.
Learning, Lecture Notes in Networks and Systems, vol 427, Springer, Singapore, [51] A. Cappiello, I. Chabini, E.K. Nam, A. Lue, M.A. Zeid, A statistical model of vehicle
2022. emissions and fuel consumption, in: Proceedings of the IEEE 5th International
[42] R. Molina-Masegosa, J. Gozalvez, M. Sepulcre, Comparison of IEEE 802.11p and Conference on Intelligent Transportation Systems, Singapore, 2002, pp. 801–809.
LTE-V2X: an evaluation with periodic and aperiodic messages of constant and [52] European Commission, Zero emission vehicles: first ‘fit for 55′ deal will end the sale
variable size, IEEE Access 8 (2020) 121526–121548, https://doi.org/10.1109/ of new CO2 emitting cars in Europe by 2035. https://ec.europa.eu/commissio
ACCESS.2020.3007115. n/presscorner/detail/en/ip_22_6462, 2022 (Accessed 06 October 2023).
[43] D.H. Stolfi, E. Alba, Red Swarm: smart mobility in cities with EAS, in: Proceedings [53] Visitlondon.com, Driving in London. https://www.visitlondon.com/traveller-info
of the 15th Annual Conference on Genetic and Evolutionary Computation, rmation/travel-to-london/driving-in-london, 2023 (Accessed 06 October 2023).
Amsterdam, Netherlands, 2013, pp. 1373–1380. [54] esMADRID.com, Driving in Madrid. https://www.esmadrid.com/en/driving-madri
[44] D. Jiang, L. Delgrossi, IEEE 802.11 p: towards an international standard for d, 2023 (Accessed 06 October 2023).
wireless access in vehicular environments, in: Proceedings of the 2008 IEEE 68th [55] A. Rionda, I. Marín, D. Martínez, F. Aparicio, et al., UrVAMM — A full service for
Vehicular Technology Conference (VTC), Barcelona, Spain, 2008, pp. 2036–2040. environmental-urban and driving monitoring of professional fleets, in: Proceedings
[45] Y.C. Tseng, S.Y. Ni, Y.S. Chen, J.P. Sheu, The broadcast storm problem in a mobile of the 2013 International Conference on New Concepts in Smart Cities: Fostering
ad hoc network, Wirel. Netw. 8 (2002) 153–167, https://doi.org/10.1023/A: Public and Private Alliances (SmartMILE), Xixón, Spain, 2013, pp. 1–6.
1013763825347. [56] Cohda Wireless, Cohda Wireless mk6 Obu and Rsu, https://www.cohdawireless.
[46] E.H. Mamdani, Application of fuzzy logic to approximate reasoning using linguistic com/solutions/mk6/, 2023 (Accessed 06 October 2023).
synthesis, IEEE Trans. Comput. C-26 (12) (1977) 1182–1191, https://doi.org/ [57] Unex Technology Corporation, V2X On-Board Unit https://unex.com.tw/pdf/OB
10.1109/TC.1977.1674779. U-352.pdf, 2023 (Accessed 06 October 2023).
[47] J. Harri, F. Filali, C. Bonnet, Mobility models for vehicular ad hoc networks: a [58] Unex Technology Corporation, V2X Roadside Unit, https://unex.com.tw/pdf/RS
survey and taxonomy, IEEE Commun. Surv. Tutor. 11 (4) (2009) 19–41, https:// U-352.pdf, 2023 (Accessed 06 October 2023).
doi.org/10.1109/SURV.2009.090403.

16