A Monte-Carlo Analysis between a Microscopic Model and a Mesoscopic Model for Vehicular Ad-Hoc Networks

International Journal of Wireless & Mobile Networks (IJWMN), Vol.16, No.6, December 2024
DOI:10.5121/ijwmn.2024.16604 47
A MONTE-CARLO ANALYSIS BETWEEN A
MICROSCOPIC MODEL AND A MESOSCOPIC MODEL
FOR VEHICULAR AD-HOC NETWORKS
Aslinda Hassan 1
, Wahidah Md. Shah 1
and Mohammed Saad Talib 2
1 Fakulti Teknolog Maklumat dan Komunikasi, Universiti Teknikal Malaysia Melaka
Hang Tuah Jaya, 76100 Durian Tunggal, Melaka, Malaysia
2
College of Administration and Economics, University of Babylon,
Babel, Iraq
ABSTRACT
Vehicular Ad-hoc Networks (VANETs) are crucial for advancing intelligent transportation systems,
enhancing road safety, and enabling efficient vehicle-to-vehicle and vehicle-to-infrastructure
communications. However, accurately simulating vehicular environments' dynamic and complex nature
remains a significant challenge. This study addresses this gap by benchmarking the performance of a
mesoscopic model, which incorporates a lane-changing technique, against a microscopic model using
Monte Carlo simulations. The microscopic model focuses on individual vehicle movements, considering
driver behaviour and interactions, while the mesoscopic model captures traffic flow at the road segment or
neighbourhood level. The updated mesoscopic model incorporates a lane change technique to better reflect
realistic vehicle movements. The updated mesoscopic model generated approximately 350 to 400 vehicles
in the simulations, with a narrow distribution and a peak frequency of about 120 vehicles. In contrast, the
original microscopic model produced around 800 vehicles and had a wider distribution but exhibited a
similar peak frequency. The revised model demonstrated a slight negative skewness of -0.1019, while the
original model showed a slight positive skewness of 0.0618. Both models displayed negative kurtosis
values, indicating lighter tails than a normal distribution. Notably, the original model had a more negative
kurtosis of -0.2931, compared to -0.1742 for the revised model. These findings suggest that the
microscopic model is more adept at capturing the variability of traffic flow, making it a more accurate
reflection of real-world scenarios where vehicle interactions significantly impact vehicle dynamics during
data transmissions.
KEYWORDS
VANET; mobility model; microscopic; mesoscopic; Monte-Carlo
1. INTRODUCTION
Vehicular Ad-Hoc Networks (VANETs) have become increasingly important in advancing
intelligent transportation systems, improving road safety, and establishing effective means for
vehicle-to-vehicle and vehicle-to-infrastructure communications [1, 2]. Given the dynamic nature
of vehicular environments with high mobility and diverse traffic conditions, simulations are
essential to develop efficient network protocols for VANETs [2]. These simulations rely on
mobility models to accurately represent real-world vehicle movements within the network. The
movement patterns of vehicles have a significant impact on VANET simulations, influencing
network protocol performance metrics such as throughput, delay, and packet delivery ratios [3].
These models aim to capture the dynamic behaviour of vehicles on roads, encompassing aspects
like acceleration, deceleration, lane changes, and route variations. Therefore, comprehensive

48
mobility models must consider numerous factors, from basic vehicle motions to complex
interactions with road systems and traffic conditions.
The selection of an appropriate mobility model is crucial for accurately representing real-world
simulation scenarios [2]. Researchers have developed various mobility models, each designed for
specific simulation requirements and levels of realism. These models can be broadly classified
into three categories: microscopic, mesoscopic, and macroscopic, based on their level of
complexity and the scale at which they operate [2, 4-6]. Microscopic models focus on the
intricate details of individual vehicle movements, considering factors such as driver behaviour,
vehicle acceleration, and interactions with other vehicles [2, 4-6]. In contrast, mesoscopic
models offer a moderate level of detail, suitable for simulating traffic flow and vehicle
movements at the level of road segments or neighbourhoods [7]. On the other hand, macroscopic
models abstract individual vehicle movements to concentrate on the overall traffic flow patterns
across larger areas, such as cities or regions [6, 8].
Each type of mobility model offers unique advantages and is suitable for different kinds of
VANET simulations. By carefully selecting and applying these models, researchers and engineers
can derive meaningful insights into the performance and behaviour of VANETs under a wide
range of conditions. Ultimately, the insights gained from simulations using accurate mobility
models are instrumental in designing and optimizing VANET protocols and applications for real-
world deployment, paving the way for smarter, safer, and more efficient transportation systems.
The mobility model presented in this paper has been adapted from a prior model developed by [9]
by incorporating a lane change technique to better capture the realistic movements of vehicles in
VANET environments. The formulation of this adapted model has been thoroughly described in
references [10, 11]. The current study aims to benchmark the performance of this revised model,
which is a mesoscopic mobility model against the original mobility model from [9] that is based
on a microscopic mobility model.
2. LITERATURE SURVEY ON VANETS MOBILITY MODELS
In VANETs, the dynamic vehicular movements trigger variations in the network architecture,
directly impacting key performance metrics such as throughput, transmission latency, and packet
loss rate [2, 3, 6]. Accurately reproducing realistic traffic flow in the simulation environment is
crucial for advancing research on VANET topology and routing protocols. Consequently, the
vehicular mobility model has become the primary focus in VANET simulation research,
concentrating on identifying the movement patterns of vehicle nodes to enhance the realism of
the simulations. This ensures that the conclusions drawn from the research are applicable to real-
world implementations. Therefore, researchers employ various methods to accurately simulate
vehicle mobility [3, 6].
One of the early influential works is by Wisitpongphan et al. [12]., which is still highly cited. In
this research, the authors developed an analytical framework based on an extended car-following
model, which falls under the category of the microscopic mobility model. The authors utilized
empirical data collected from the dual-loop detector on the eastbound I-80, a 5-lane highway
immediately east of the San Francisco-Oakland Bay Bridge between Emeryville, CA, and
Berkeley, CA. From the realistic mobility trace, the authors were able to approximate the
probability distribution of inter-arrival times and inter-vehicle spacing as exponential distribution,
and vehicle arrival as a Poisson distribution. The study conducted by Wisitpongphan et al. [12]
formed the basis for our revised mobility model, which incorporates Poisson distribution to
represent vehicle arrival rate and exponential distribution to account for vehicle inter-arrival
times.

49
Table 1 presents a summary of research conducted between 2019 and 2024 on empirical mobility
trace and mobility models. These works aim to accurately represent the dynamic nature of vehicle
movements in both highway and urban areas.
Table 1: Literature review on common approaches for generating vehicles' mobility for year 2019-2024
Vehicle
Mobility
Approach
Referenc
e
Dataset/Model/
Technique
Objectives
Mobility
Trace
[13]
Beijing taxi traces,
Singapore and Jakarta
e-hailing vehicle traces
This study examines data dissemination in a
vehicular network using empirical mobility traces
and incorporates the method of index coding to
minimize the number and size of transmissions.
[14]
Public transportation
mobility using General
Transit Feed
Specification (GTFS)
data for Dublin, Rome,
Seattle and Washington
The study reveals key characteristics of bus-based
networks and examines their network topology and
spatiotemporal impacts. The data is further used to
assess the design of routing protocols in bus-based
networks.
[15]
Two sets of GPS
traces. 1) over 4000
taxis and buses in
Shanghai, 2) over 7000
taxis in Shenzhen
The mobility trace is incorporated into the proposed
Dynamically Evolving Networking model, which
simulates the evolution of VANETs. This model
enables a more accurate representation of how
VANETs function in real-world scenarios. The
proposed model allows the network to evolve into a
scale-free network as it grows, self-organizes, and
even decreases in size. Additionally, it can ensure
good network connectivity and survivability.
[16]
Datasets of GPS traces
from 320 taxis in Rome
and 551 taxis in San
Francisco
The study employs three distinct approaches -
instantaneous, aggregated, and time-varying analysis
- to model vehicular ad-hoc networks using graph
theory. The authors analyze two large-scale traces,
which enables them to ground the theoretical models
in real-world data, thereby enhancing the validity and
applicability of the findings.
Microscop
ic
Mobility
Mode
[17]
Microscopic
Mechanism based on
Intersection Records
(MMIR)
The proposed mechanism aims to enhance the
accuracy of connectivity probability or estimated
delivery delay for street selection in VANETs using
a microscopic approach that leverages individual
vehicle data.
[18]
A microscopic traffic
modelling for road
networks
In this research, the authors use microscopic
modelling to focus on vehicle behaviour and
interactions rather than on aggregate traffic flow
analysis. A new formalism has been introduced to
describe the dynamics of vehicle traffic, particularly
in unidirectional traffic scenarios. This model
effectively captures vehicle interactions, including
acceleration, deceleration, and the maintenance of
safe distances. It reveals the behaviour of individual
vehicles and is adaptable to various road networks,
making it a valuable tool for analyzing traffic
systems and urban environments. This approach
benefits both scholars and practitioners who are
studying complex traffic situations due to its detailed
examination and broad applicability.

50
[19]
the Adaptive Driver
Model (ADM) and
Adaptive Lane
Changing Model
(ALC)
This study develops microscopic simulations that can
accurately represent realistic traffic configurations
and support large-scale applications in cyber-
physical systems
[20]
A Variable Speed
Limit (VSL) control
algorithm with the
Model Predictive
Control (MPC)
framework
This study examines Variable Speed Limit control
strategies that leverage microscopic traffic flow data
to enhance traffic flow predictions.
Mesoscopi
c Mobility
Model
[21]
A mesoscopic model of
vehicular mobility on a
multilane highway with
steady-state traffic flow
conditions
This study presents the mathematical modelling of 1)
the expected number of hops in a communication
link, 2) the distribution of successful multi-hop
forwarding, 3) the expected time delay, and 4) the
expected connectivity distance. The accuracy of the
proposed model is validated through simulations
conducted using an event-based network simulator
and a road traffic simulator
[22]
Probability-based
theoretical approaches,
including the Poisson
process, uniform
distribution, and
exponential
distribution, are applied
to deduce the
connectivity
probabilities of
cognitive vehicular
networks (CVNs)
The authors develop a robust analytical framework
using probability theory to analyze the connectivity
of cognitive vehicular networks under different
scenarios, including single-hop and multi-hop
clustering. By examining these scenarios
independently, the authors gain insights into how
connectivity varies with distinct communication
structures, namely the inter-cluster integration
process and the intra-cluster communication process.
[23]
The study uses a
stochastic geometry
model to characterize
the vehicular network's
spatial features. It
integrates the Poisson
Line Process to model
the distribution of
roadside infrastructure
and the 1D Poisson
Point Process to
represent vehicle
locations within the
network.
The paper proposes a comprehensive analytical
framework for investigating multi-hop relaying in
vehicular networks. The framework analyzes how
various network parameters, such as the number of
hops, impact the coverage provided by road-side
units (RSUs) to vehicles. This analysis provides
insights into the optimal RSU density required for
effective network-wide communication.
Additionally, the study assesses the delivery delays
introduced by multi-hop relaying.
[24]
The paper uses the
M/M/1 queuing model
to assess the VANET
system's performance,
assuming a Poisson
arrival process and
exponential service
times to understand
vehicle flow and
service dynamics on
highways.
The paper presents a mathematical model that
captures the dynamic nature of traffic and service in
VANETs. It includes the calculation of the average
number of vehicles in the queue and the distribution
of waiting times. The authors analyze the queuing
system's performance by measuring traffic flow in
highway lanes, with a specific focus on the impact of
vehicle collisions. The simulation helps understand
how vehicles interact within the network and their
overall impact on communication efficiency.

51
Macrosco
pic
Mobility
Model
[25]
A new macroscopic
model for Variable
Speed Limits (VSLs).
This study investigates the impact of variable speed
limits on the fundamental diagram of traffic flow.
The research reveals that the fundamental diagram
induced by VSLs does not conform to the traditional
triangular assumption, challenging the conventional
understanding of traffic flow under speed
restrictions. The authors aim to demonstrate how
changes in speed limits can alter the characteristics
of this diagram, particularly in terms of capacity and
density. Their findings indicate that reducing the
speed limit from 120 to 90 km/h leads to a decrease
in freeway capacity and an increase in critical density
for the studied segment of the A12 freeway in The
Netherlands. Furthermore, the proposed model is
rigorously compared with two well-established
macroscopic models for VSLs
[26]
The authors have
modified the original
Greenberg model to
improve its accuracy
and applicability in
real-world scenarios.
The authors employed a modified Greenberg model
to simulate traffic variables across different
scenarios, incorporating intermediate inputs/outputs
as well as a viscosity term in the motion equation.
The primary goal of this work was to investigate the
impact of viscosity on traffic variables by conducting
simulations using the modified model and leveraging
measured traffic data as the initial and boundary
conditions.
[27]
The authors adopt the
macroscopic first-order
Lighthill-Whitham-
Richards model to
describe the overall
dynamics of the traffic
flow, coupled with two
ordinary differential
equations that
characterize the
trajectories of the front
and back endpoints of
the platoon.
The model effectively captures the interaction
between a platoon of vehicles and the surrounding
traffic flow, representing a novel approach in this
research area. The paper highlights the dynamics of
how slow-moving vehicles, depicted as a platoon,
can create point-moving bottlenecks in traffic.
Additionally, the authors have developed a finite
volume scheme to compute approximate solutions to
the coupled partial differential equation and ordinary
differential equation system. This computational
method is evaluated within the paper, demonstrating
its effectiveness in simulating the dynamics of
vehicle platooning.
[28]
This work improves the
established Lighthill-
Whitham-Richards
model by incorporating
vehicle spacing in both
time and distance
domains.
This paper introduces a novel model that focuses on
the analysis of inhomogeneous traffic flow during
transitional periods. The proposed model is designed
to more effectively characterize the evolution of
traffic dynamics and provide deeper insights into the
behavior of traffic under varying conditions. The
authors employ the Godunov numerical technique to
evaluate both the established LWR model and their
newly developed model, ensuring numerical stability
and reliability through adherence to the Courant-
Friedrichs-Lewy condition. The performance of the
proposed model is rigorously assessed and compared
to the LWR model using Greenshields and
Underwood target velocity distributions

52
2.1. Real-World Mobility Trace
Based on a survey done by Clayson et al. [29], a mobility trace is a dataset that records vehicle
positions over time using Global Navigation Satellite System (GPS) and cellular networks. Real-
world traces provide accurate vehicle movement data, aiding in creating realistic simulations and
improving vehicular network solutions. As a result, these traces offer advantages that range from
creating more realistic simulation scenarios to identifying information that improves solutions for
vehicular networks.
According to Clayson et al. [29], most vehicular mobility traces originate from academic research
or are provided by traffic control organizations. In their survey, the authors identified three types
of mobility traces that contain vehicle movement data from real-world scenarios: bus mobility
traces, taxi mobility traces, and private car mobility traces. The authors present several vehicular
motion traces and conducts a qualitative comparison among them. All of the traces are publicly
accessible and can be categorized as either real-world or synthetic. Real-world traces comprise
positioning data captured by location devices such as GPS receivers. However, due to privacy
and security concerns, the majority of these traces are of the movements of anonymous taxis or
buses.
The literature review presented in Table 1 clearly shows that researchers are likely using mobility
trace data from public transportation due to significant concerns about security and privacy.
2.2. Mobility Models
As mentioned in Section 1, VANET mobility models are typically classified into three main
categories: 1) microscopic, 2) mesoscopic, and 3) macroscopic.
2.2.1. Microscopic Mobility Models
The microscopic mobility model focuses on individual vehicle motion, typically at the level of a
single road segment. Several models have been developed under this category, including the Car
Following Model, Intelligent Driver Model, Krauss Model, Wiedemann Model, and Cellular
Automata Model [6].
2.2.2. Mesoscopic Mobility Models
Mesoscopic traffic models represent an intermediate level of abstraction, capturing the overall
properties of traffic flows through probability distributions while still accounting for individual
vehicle interactions [6, 21]. For instance, these models may employ a uniform distribution to
characterize the velocity distribution at a given time and location or an exponential distribution
for the vehicular arrival rate.
The mesoscopic mobility model is commonly used in VANET research for simulating vehicular
communication networks. It captures steady-state traffic flow conditions and predicts
connectivity metrics, essential for designing efficient VANET applications [6, 21]. While
mesoscopic models are practical, there's a growing interest in more detailed microscopic models
that capture finer-grained mobility patterns, providing insights into individual vehicle behaviours'
impact on network performance for specific applications like autonomous driving and pedestrian
safety.

53
2.2.3. Macroscopic Mobility Models
This model takes into account more than just vehicles. It considers the flow of a large number of
vehicles from a global perspective, taking into account road topology, features and conditions,
traffic density and distribution, traffic signals, and traffic flow [6]. This approach allows for the
calculation of road capacity and traffic distribution in the road network.
3. MOBILITY MODELS ANALYSIS
This section provides an overview of how the behavioural characteristics in the mobility model
from [9] and [10, 11] influence the overall traffic flow and patterns. Comprehensive
documentation of both the original and revised mobility models can be found in the references
cited. Table 2 listed the summary on the analysis of both mobility models.
Table 2: Analysis on mobility models by [9] and [10, 11]
Assumption Mobility Model by [9] Mobility Model by [10, 11]
Vehicle Arrival Rate
Traffic flow based on microscopic
approach
Poisson process
Inter-Arrival Rate
between vehicle
batches
Not stated Exponential distribution
Vehicles Speed
Distribution
Vehicles’ speed in both model use an equation that is influenced by
acceleration factor. The factor is determined by a set of equations
involving several random variables and an aggressiveness (AGG)
parameter
Aggressiveness of
Vehicle Mobility
Behaviour (AGG)
The AGG parameter affects the
predictability of vehicle movements,
thereby influencing the performance of
the routing protocol proposed by the
authors. The parameter AGG is
represented by the single value of 0.2.
The impact of the parameter in a
traffic scenario is illustrated
using three different AGG
values: 0.2, 0.5, and 0.8.
The revised mobility model utilizes a batch arrival process to simulate vehicle arrivals, with
vehicles being generated in groups according to a Poisson process, as stated in Table 2. The
parameters of the Poisson distribution, including the mean and standard deviation, are employed
to manage the expected batch size and its variability. From the definition given in Section 2, we
can confidently categorize our revised model as a mesoscopic mobility model.
Conversely, the mobility model in [9] does not explicitly mention batch generation for vehicle
arrival. Instead, it uses traffic flow that is based on a microscopic approach to focus on
maintaining a continuous flow of vehicles on the highway.
As shown in Table 2, the revised model in [10, 11] employs an exponential distribution to
represent the inter-arrival rate between vehicle batches. This probabilistic distribution effectively
captures the randomness and variability in the time intervals between successive vehicle arrivals,
mirroring the stochastic nature of traffic flow. The rate parameter of the exponential distribution
determines the expected time interval between batches, allowing for flexibility in simulating
diverse traffic conditions.
On the other hand, the authors in [9] does not explicitly define an inter-arrival rate distribution.
Instead, the authors use a discrete-time model where vehicles continuously recalculate their
acceleration at regular intervals. This approach focuses on the uninterrupted movement of

54
vehicles rather than discrete arrival events, aligning with the model's emphasis on maintaining
prescribed speed limits and lane dynamics.
Both models employ a similar approach to modelling speed distribution, drawing from the
methodology outlined in [9]. For each time interval, the models calculate each vehicle's speed by
incorporating an acceleration factor to determine whether vehicles should accelerate or
decelerate. Additionally, the models leverage random variables and aggressiveness (AGG)
factors to simulate realistic speed variations among vehicles, capturing the dynamic nature of
highway traffic.
As both mobility models employ the same approach to determine vehicle speeds, they also utilize
the same definition for the aggressiveness parameter [10, 11]. The AGG parameter, as defined by
Kesting et al. [30], is used to control and simulate the aggressive driving behavior of vehicles in
the proposed model. Higher values of AGG indicate more aggressive driving, which significantly
affects the overall mobility pattern of vehicles on the highway. Therefore, AGG parameter has a
direct impact on the performance of the proposed routing protocol in [9], influencing its ability to
accurately predict route lifetimes and proactively create new routes before existing ones fail. The
authors for the revised mobility model in [10, 11] investigated the impacts of varying the AGG
parameter, which represents driver aggressiveness, at specific values of 0.2 (low), 0.5 (medium),
and 0.8 (high), on the performance of the proposed clustering algorithm.
3.1. The Lane Changing Technique
In our revised mobility model, we incorporate a lane-changing technique to enhance the
performance of the clustering algorithm proposed in [9]. This technique involves the introduction
of two key probabilities:
𝑝1: The probability of a vehicle maintaining its current lane when the relative distance
between the vehicle and its preceding vehicle is lower than a specific threshold,
known as the “safety distance.”
𝑝2: The probability of a vehicle preserving its lane when the relative distance between
the vehicle and its preceding vehicle exceeds a certain value, denoted as the 𝑑𝑡ℎ
parameter
The lane-changing technique divides the complementary probabilities 𝑝1 and 𝑝2 into two equal
probabilities. For vehicles not in the border lanes, 𝑝1: represents the probability of a lane change
to the right, while 𝑝2: represents the probability of a lane change to the left. However, for
vehicles in the border lanes, there is only one lane change option, either to the right or to the left.
The mathematical expressions for the two key probabilities, 𝑝1 and 𝑝2 are presented in Equation
1. Equation 1 formalizes the lane change decision-making process, where the vehicle's lateral
position 𝑦 is updated based on the probabilities of executing a lane change to the right or left.
These probabilities are determined by the distance 𝑑 between the vehicle and its leading vehicle,
as well as the values of the probabilities 𝑝1 and 𝑝2.
𝑦𝑖,𝑡+1 =
{
𝑦𝑖,𝑡+1 𝑖𝑓 (𝜀 ≥ 𝑝1 𝜀 < 𝑝1 +
1 − 𝑝1
2
 𝑑 < 𝑑𝑡ℎ) 𝑜𝑟 (𝜀 ≥ 𝑝1 𝜀 < 𝑝2 +
1 − 𝑝2
2
 𝑑 ≥ 𝑑𝑡ℎ)
𝑦𝑖,𝑡−1 𝑖𝑓 (𝜀 < 1  𝜀 ≥ 𝑝1 +
1 − 𝑝1
2
 𝑑 < 𝑑𝑡ℎ) 𝑜𝑟 (𝜀 < 1  𝜀 ≥ 𝑝2 +
1 − 𝑝2
2
 𝑑 ≥ 𝑑𝑡ℎ)
𝑦𝑖,𝑡 𝑖𝑓(𝜀 < 𝑝1  𝑑 < 𝑑𝑡ℎ)𝑜𝑟(𝜀 < 𝑝2  𝑑 ≥ 𝑑𝑡ℎ)
(1)
With

55
𝑑 = |𝑥𝑖+1,𝑡 − 𝑥𝑖,𝑡|
Where
𝑦𝑖,𝑡+1: the updated position of vehicle 𝑖 at time 𝑡 + 1
𝜀: a random number generated between 0 and 1
𝑑: denotes the distance separating the vehicle of interest and its preceding vehicle
𝑑𝑡ℎ: denotes a distance threshold between the vehicle 𝑖 and the lead vehicle that triggers a lane change
for probability 𝑝1.
4. MOBILITY MODELS BENCHMARK
A Monte Carlo simulation, comprising 1000 iterations, was conducted to assess the performance
of the two mobility models. The simulated scenario involved a 6-lane unidirectional highway
spanning 10 kilometres. Communication configuration and delay factors were excluded from this
simulation. The purpose of the Monte Carlo simulation was to evaluate and analyse the
differences between the two mobility models based on the discussion presented in the preceding
section.
4.1. Results on Vehicles' Generation
The first step in the Monte Carlo simulation for both mobility models is to analyze the
characteristics and the distribution for the vehicle generation.
Figure 1. Comparison between two models on number of vehicles generated
Figure 1 presents the histograms depicting the frequency distributions of the number of vehicles
generated by the two modelling approaches. The histogram on the left represents the distribution
of the number of vehicles generated by the revised model, while the histogram on the right
represents the distribution of the number of vehicles generated by the original model. A
histogram is a graphical representation of the frequency distribution of a quantitative variable.
The x-axis depicts the variable values, with each bar representing a discrete value or a class of
continuous values arranged in ascending order. The height of the bars on the y-axis corresponds
to the frequency distribution of the respective variable values.

56
The revised model, as indicated in Table 2 and referenced in [10, 11], employs a Poisson process
for generating the number of vehicles and an Exponential distribution for inter-arrival time. In
contrast, the original model described in [9] adopts a microscopic approach to determine the
number of vehicles. As depicted in Figure 1, the distribution for the revised model is centred
around 350-400 vehicles with a relatively narrow spread and a higher peak frequency of
approximately 120 vehicles. On the other hand, the distribution of the original model is centred
around 800 vehicles with a wider spread but a similar peak frequency, highlighting greater
variability in vehicle generation. This aligns with the model's emphasis on individual vehicle
behaviours and dynamic routing. In summary, the revised model assumes a random arrival of
vehicles based on an average rate, making it more suitable for modelling traffic flow at a macro
level or for less congested scenarios. Conversely, the original model accounts for individual
vehicle behaviours and interactions, allowing for more densely packed vehicles and potentially
providing a more realistic representation of congested highway scenarios.
Figure 2. Number of vehicles generated statistical comparison between the two models
Furthermore, Figure 2 presents the findings on skewness and kurtosis. Skewness is a measure
used to assess the symmetry, or lack thereof, in a distribution or dataset [31]. An asymmetric
distribution appears the same on both sides of its central point. Both models exhibit relatively
small skewness values, indicating that their distributions are nearly symmetrical. The revised
model has a slightly negative skewness of -0.1019, suggesting a subtle leftward asymmetry,
whereas the original model demonstrates a slightly positive skewness of 0.0618, indicating a mild
rightward asymmetry.
Kurtosis is a statistical measure that indicates whether the data exhibits a heavy-tailed or light-
tailed distribution relative to a normal distribution [31]. A kurtosis value of 0 corresponds to a
distribution with a peak and tails similar to a normal distribution. Both models exhibit negative
kurtosis values, indicating that their distributions have lighter tails and are more platykurtic
(flatter) compared to a normal distribution. The original model has a slightly more negative
kurtosis value of -0.2931 compared to the revised model's -0.1742, suggesting an even flatter
distribution with fewer extreme values in the tails than the revised model. This implies a slightly
flatter peak and lighter tails for the original model.
4.2. Impact on the AGG Parameters on the Traffic Flow
This section examines the impact of the aggressiveness behaviour parameter on traffic flow using
the Monte Carlo simulation. Figure 3 illustrates the relationship between the aggressiveness
(AGG) parameter and different vehicle traffic flows in the network. The AGG parameter is used
to control and simulate the aggressive driving behaviour of vehicles in the network. As described

57
in Section 3, this parameter influences the acceleration and deceleration dynamics of vehicles,
where higher AGG values correspond to more aggressive driving. This parameter has a
significant impact on the overall mobility patterns exhibited by vehicles on the highway.
Figure 3. Comparison of Two Models Examining the Impact of Aggressiveness Parameters on Different
Traffic Flows
In terms of the distribution shape, the revised model exhibits a more symmetric and narrower
distribution, which is consistent with the Poisson distribution used for vehicle arrivals. In
contrast, the original model displays a wider range of vehicle counts, reflecting a more complex,
microscopic approach to vehicle movement. As we progress through the rows in Figure 3, both
models demonstrate an increase in vehicle count as the AGG parameter rises. However, this
effect is more pronounced in the original model, indicating that the microscopic approach is more
sensitive to changes in driver aggressiveness.
As the traffic flow increases as we move right across the columns of Figure 3, both models
demonstrate a rise in vehicle generation. The original model exhibits a wider spread of vehicle
counts at higher flow rates, indicating that the microscopic approach is able to capture a greater
degree of variability in traffic patterns. Conversely, the revised model's narrower distribution
suggests that it may be more adept at predicting average traffic conditions, but could
underestimate extreme scenarios. The wider distribution seen in the original model implies that it
encompasses a wider range of traffic scenarios, potentially making it more suitable for the study
of edge cases or unusual traffic patterns.

58
The findings illustrated in Figure 3 suggest that the original model displays higher peak densities,
particularly at elevated AGG values and flow rates. This implies that the original model's
microscopic approach may be more effective in capturing traffic clustering or congestion effects.
Additionally, both models exhibit sensitivity to changes in AGG and flow rate, but the original
model appears to be more responsive. This indicates that the original model's microscopic
approach may be better equipped to adapt to diverse traffic conditions and driver behaviours.
In summary, both models exhibit the expected behaviour as flow rates and aggregation
parameters increase. The original model, which utilizes a microscopic approach, consistently
yields higher vehicle counts, possibly due to its more detailed representation of vehicle
interactions. In contrast, the revised model, which employs a Poisson process, demonstrates
slightly more right-skewed distributions compared to the original model.
While both models produce distributions that closely resemble a normal distribution, they still
significantly differ from a perfectly normal distribution. These results emphasize the distinctions
between the two modelling approaches and their implications for traffic flow simulation. The
original model appears to capture more variability in traffic flow, potentially making it more
representative of real-world scenarios where vehicle interactions play a significant role. On the
other hand, the revised model, although simpler, still captures the general trends of increasing
traffic density and the effects of the AGG parameter.
4.3. Results on Lane Changing Technique
This section presents the findings from the extended version of the Monte Carlo simulation
conducted in the previous section. The extended simulation incorporated a lane-changing
technique for both the original and revised models. For the revised model, the lane-changing
technique was based on Equation 1. However, the authors in [9]. did not specifically mention the
model or technique for lane changing in the original model. Therefore, it is assumed that the
authors of the original model employed a microscopic approach to model lane-changing
behaviour. The scenario for this extended simulation remains consistent with the previous
section, featuring a 10 km highway with six unidirectional lanes.
Table 3: List of parameters used in the extended Monte-Carlo simulation
Parameter Value
𝑣𝑚𝑖𝑛 5.56 m/s (20 km/h)
𝑣𝑚𝑎𝑥 33.33 m/s (120 km/h)
𝑑𝑡ℎ 52 meter
𝑎𝑚𝑎𝑥 2.0 m/s2
𝑑𝑚𝑎𝑥 -2.0 m/s2
Table 3 presents the parameter values employed in the extended Monte-Carlo simulation. The
parameters 𝑣𝑚𝑖𝑛 and 𝑣𝑚𝑎𝑥 correspond to the speed ranges observed for vehicles on a highway
setting. Similarly, 𝑣𝑚𝑖𝑛 and 𝑣𝑚𝑎𝑥 reflect the acceleration and deceleration limits typically
exhibited by highway vehicles. The parameter 𝑑𝑡ℎ represents the distance threshold used to
trigger lane-changing decisions, and its value is based on the commonly used 3-second rule for
safe following distances on highways [32].

59
Figure 4. CDF of Distances Between Vehicles
Figure 4 presents the cumulative distribution functions of the inter-vehicle distances for both
models. The x-axis in Figure 4 indicates that the distances between vehicles range from very
small (close to 0 meters) to quite large (over 5000 meters), reflecting the variability in traffic
density along the highway. The relatively smooth curve suggests a continuous distribution of
distances with no sharp jumps at any particular distance, indicating that vehicles are spread out
along the highway rather than clustering at specific intervals.
A notable feature is the relatively steep increase in the CDF for short distances (0-100 meters),
suggesting that a significant portion of vehicles are quite close to each other, potentially
representing areas of higher traffic density or potential congestion. Furthermore, the curve
flattens out for larger distances (beyond 500 meters), indicating fewer instances of very large
gaps between vehicles. From this CDF, it is evident that only a small fraction of vehicle pairs
(less than 5%) is within the threshold distance (67 meters) of each other
.
Figure 5. Average Vehicles Speed over Simulation Time
In Figure 5, the average speeds of both models throughout the simulation period are illustrated.
The revised model starts with a slightly lower average speed compared to the original model.
However, both models rapidly stabilize and maintain consistent average speeds throughout the
simulation. The revised model demonstrates minor fluctuations in speed, suggesting a more
controlled traffic flow with a gradual decline in average speed over time, but within a consistent
range. In contrast, the original model starts with a higher average speed than the revised model
but then experiences a sharp drop in average speed early in the simulation. The output for the

60
original model in Figure 5 shows more pronounced variations in speed throughout the simulation,
with a decreasing trend in average speed over time and greater variability.
5. CONCLUSION
This study compares a microscopic model and a mesoscopic model for vehicular ad hoc networks
(VANETs), highlighting their respective strengths and applications. The microscopic model
offers a detailed representation of individual vehicle behaviours, making it suitable for scenarios
that require high accuracy in vehicle interactions and driver behaviour. In contrast, the
mesoscopic model captures broader traffic flow patterns, making it better suited for simulations
at the level of road segments or neighbourhoods. Including a lane-changing technique in the
revised mesoscopic model enhances its realism and applicability.
While the microscopic model excels in reflecting the variability of traffic flow and is more
responsive to changes in driver aggressiveness, it also presents higher computational complexity,
which may limit its use in certain situations. On the other hand, although less detailed, the
mesoscopic model provides a practical approach for macro-level traffic flow modelling; however,
it may need to fully address extreme traffic scenarios or the sensitivity to driver behaviour.
The Monte Carlo simulations show that the microscopic model provides a more accurate
representation of vehicle movements, which is crucial for developing effective VANET
protocols. Future work could further investigate these models' performance in more complex
scenarios, such as urban environments with traffic signals, intersections, and pedestrian
interactions. Insights gained from simulations using accurate mobility models are essential for
designing and optimizing VANET protocols and applications for real-world deployment, paving
the way for smarter, safer, and more efficient transportation systems.
ACKNOWLEDGEMENTS
The authors would like to thank Fakulti Teknologi Maklumat dan Komunikasi (FTMK),
Universiti Teknikal Malaysia Melaka (UTeM) for the financial support
REFERENCES
[1] H. Peng, L. Le, X. Shen, and G. Y. Li, "Vehicular Communications: A Network Layer Perspective,"
IEEE Transactions on Vehicular Technology, vol. 68, no. 2, pp. 1064-1078, 2019, doi:
10.1109/tvt.2018.2833427.
[2] S. Babu and A. Raj Kumar P, "A comprehensive survey on simulators, emulators, and testbeds for
VANETs," International Journal of Communication Systems, vol. 35, no. 8, 2022, doi:
10.1002/dac.5123.
[3] S. I. Boucetta, Y. Guichi, and Z. C. Johanyák, "Review of Mobility Scenarios Generators for
Vehicular Ad-Hoc Networks Simulators," Journal of Physics: Conference Series, vol. 1935, no. 1,
2021, doi: 10.1088/1742-6596/1935/1/012006.
[4] F. J. Ros, J. A. Martinez, and P. M. Ruiz, "A survey on modeling and simulation of vehicular
networks: Communications, mobility, and tools," Computer Communications, vol. 43, pp. 1-15,
2014, doi: 10.1016/j.comcom.2014.01.010.
[5] J. Treurniet, "A Taxonomy and Survey of Microscopic Mobility Models from the Mobile
Networking Domain," ACM Computing Surveys, vol. 47, no. 1, pp. 1-32, 2014, doi:
10.1145/2616973.
[6] H. Hartenstein and K. P. Laberteaux, Eds. VANET: Vehicular Applications and Inter‐Networking
Technologies. John Wiley & Sons, Ltd, 2010.

61
[7] E. Papadimitriou, J. M. Auberlet, G. Yannis, and S. Lassarre, "Simulation of Pedestrians and
Motorised Traffic," International Journal of Interdisciplinary Telecommunications and Networking,
vol. 6, no. 1, pp. 57-73, 2014, doi: 10.4018/ijitn.2014010105.
[8] A . De Marchi and M. Gerdts, "Traffic Flow on Single-Lane Road Networks: Multiscale Modelling
and Simulation," IFAC-PapersOnLine, vol. 51, no. 2, pp. 162-167, 2018, doi:
10.1016/j.ifacol.2018.03.028.
[9] V. Namboodiri and G. Lixin, "Prediction-Based Routing for Vehicular Ad Hoc Networks," Ieee T
Veh Technol, vol. 56, no. 4, pp. 2332-2345, 2007, doi: 10.1109/tvt.2007.897656.
[10] M. S. Talib et al., "A Center-Based Stable Evolving Clustering Algorithm With Grid Partitioning
and Extended Mobility Features for VANETs," IEEE Access, vol. 8, pp. 169908-169921, 2020, doi:
10.1109/access.2020.3020510.
[11] M. S. Talib, A. Hassan, Z. A. Abas, A. H. yassir, M. Alndiwee, and M. F. Ali, "Analysis of Mobility
Model Effect on Stable and Coverage Nodes List in VANETs Clustering," International Journal of
Advanced Science and Technology, vol. 29, no. 8, 2020.
[12] N. Wisitpongphan, F. Bai, P. Mudalige, V. Sadekar, and O. Tonguz, "Routing in Sparse Vehicular
Ad Hoc Wireless Networks," IEEE Journal on Selected Areas in Communications, vol. 25, no. 8,
pp. 1538-1556, 2007, doi: 10.1109/jsac.2007.071005.
[13] E. R. Magsino, G. P. Arada, and C. M. L. Ramos, "Investigating Data Dissemination in Urban
Cities by Employing Empirical Mobility Traces | IEEE Conference Publication | IEEE Xplore,"
2020 IEEE 12th International Conference on Humanoid, Nanotechnology, Information Technology,
Communication and Control, Environment, and Management (HNICEM), 2020, doi:
10.1109/HNICEM51456.2020.9400028.
[14] C. Celes, A. Boukerche, and A. A. F. Loureiro, "Generating and Analyzing Mobility Traces for
Bus-Based Vehicular Networks," IEEE Transactions on Vehicular Technology, vol. 72, no. 12, pp.
16409-16425, 2023, doi: 10.1109/tvt.2023.3290154.
[15] X. Qu, E. Liu, R. Wang, and H. Ma, "Complex Network Analysis of VANET Topology With
Realistic Vehicular Traces," IEEE Transactions on Vehicular Technology, vol. 69, no. 4, pp. 4426-
4438, 2020, doi: 10.1109/tvt.2020.2976937.
[16] B. Ravi and J. Thangaraj, "Stochastic traffic flow modeling for multi-hop cooperative data
dissemination in VANETs," Physical Communication, vol. 46, 2021, doi:
10.1016/j.phycom.2021.101290.
[17] Z. Cai, M. Liang, and Q. Sun, "MMIR: a microscopic mechanism for street selection based on
intersection records in urban VANET routing," EURASIP Journal on Wireless Communications and
Networking, vol. 2019, no. 1, 2019, doi: 10.1186/s13638-019-1475-4.
[18] R. M. Colombo, H. Holden, and F. Marcellini, "On the Microscopic Modeling of Vehicular Traffic
on General Networks," SIAM Journal on Applied Mathematics, vol. 80, no. 3, pp. 1377-1391, 2020,
doi: 10.1137/19m1270896.
[19] A. C. Nae and I. Dumitrache, "Adaptive Models for Microscopic Traffic Modelling," in 2021 23rd
International Conference on Control Systems and Computer Science (CSCS), Bucharest, Romania,
2021, doi: 10.1109/CSCS52396.2021.00051. [Online]. Available:
https://ieeexplore.ieee.org/document/9481060
[20] J. Fang, H. Ye, and S. M. Easa, "Modified Traffic Flow Model with Connected Vehicle
Microscopic Data for Proactive Variable Speed Limit Control," Journal of Advanced
Transportation, vol. 2019, no. 1, pp. 1-11, 2019, doi: 10.1155/2019/8151582.
[21] M. Zarei, "Traffic-Centric Mesoscopic Analysis of Connectivity in VANETs," The Computer
Journal, vol. 63, no. 2, pp. 203-219, 2020, doi: 10.1093/comjnl/bxz094.
[22] X. Li, R. Zhou, T. Zhou, L. Liu, and K. Yu, "Connectivity Probability Analysis for Green
Cooperative Cognitive Vehicular Networks," IEEE Transactions on Green Communications and
Networking, vol. 6, no. 3, pp. 1553-1563, 2022, doi: 10.1109/tgcn.2022.3158953.
[23] H. A. Ammar, A.-K. Ajami, and H. Artail, "A Poisson Line Process-Based Framework for
Determining the Needed RSU Density and Relaying Hops in Vehicular Networks," IEEE
Transactions on Wireless Communications, vol. 19, no. 10, pp. 6643-6659, 2020, doi:
10.1109/twc.2020.3004387.
[24] B. Ravi, J. Thangaraj, and S. Petale, "Data Traffic Forwarding for Inter-vehicular Communication in
VANETs Using Stochastic Method," Wireless Personal Communications, vol. 106, no. 3, pp. 1591-
1607, 2019, doi: 10.1007/s11277-019-06231-2.

62
[25] J. R. D. Frejo, I. Papamichail, M. Papageorgiou, and B. De Schutter, "Macroscopic modeling of
variable speed limits on freeways," Transportation Research Part C: Emerging Technologies, vol.
100, pp. 15-33, 2019, doi: 10.1016/j.trc.2019.01.001.
[26] E. J. Lopez-Sanchez, N. Yanet Sanchez-Torres, and P. E. O. Martinez, "The Viscosity Effect on
Velocity of a Macroscopic Vehicular Traffic Model," Communications - Scientific letters of the
University of Zilina, vol. 23, no. 2, pp. A80-A93, 2021, doi: 10.26552/com.C.2021.2.A80-A93.
[27] G. Piacentini, P. Goatin, and A. Ferrara, "A Macroscopic Model for Platooning in Highway
Traffic," SIAM Journal on Applied Mathematics, vol. 80, no. 1, pp. 639-656, 2020, doi:
10.1137/19m1292424.
[28] Z. H. Khan, T. A. Gulliver, and W. Imran, "A Macroscopic Traffic Model Based on the Safe
Velocity at Transitions," Civil Engineering Journal, vol. 7, no. 6, pp. 1060-1069, 2021, doi:
10.28991/cej-2021-03091710.
[29] C. Celes, A. Boukerche, and A. A. F. Loureiro, "Mobility Trace Analysis for Intelligent Vehicular
Networks," ACM Computing Surveys, vol. 54, no. 3, pp. 1-38, 2021, doi: 10.1145/3446679.
[30] A. Kesting, M. Treiber, D. Helbing, A. Kesting, M. Treiber, and D. Helbing, "Agents for Traffic
Simulation," in Multi-Agent Systems: Simulation and Applications, A. M. Uhrmacher and D.
Weyns Eds., 1st ed.: CRC Press, 2009.
[31] R. Bono, J. Arnau, R. Alarcón, and M. J. Blanca, "Bias, Precision, and Accuracy of Skewness and
Kurtosis Estimators for Frequently Used Continuous Distributions," Symmetry, vol. 12, no. 1, 2019,
doi: 10.3390/sym12010019.
[32] J. F. McDonald, "Following Rules and Traffic Flow: What If Everyone Used the Three-Second
Rule?," SSRN Electronic Journal, 2023, doi: 10.2139/ssrn.4540900.

A Monte-Carlo Analysis between a Microscopic Model and a Mesoscopic Model for Vehicular Ad-Hoc Networks

More Related Content

Similar to A Monte-Carlo Analysis between a Microscopic Model and a Mesoscopic Model for Vehicular Ad-Hoc Networks (20)

Recently uploaded (20)

A Monte-Carlo Analysis between a Microscopic Model and a Mesoscopic Model for Vehicular Ad-Hoc Networks