International Journal of Interactive Mobile Technologies (iJIM) – eISSN: 1865-7923 – Vol 17 No 04 (2023)

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Clustering Technique for Mobile Edge Computing To
Detect Clumps in Transportation-Related Problems

https://doi.org/10.3991/ijim.v17i04.37801

J. Albert Mayan1(), S.V. Manikanthan2, Azham Hussain3,
S. Nithyaselvakumari4, A. Vinnarasi5

1 Department of Computer Science and Engineering, Sathyabama Institute of Science and
Technology, Chennai, India

2 Melange Academic Research Associates, Puducherry, India
3 School of Computing, Universiti Utara Malaysia, Sintok, Malaysia

4 Department of Medical Instrumentation, Saveetha school of Engineering, Thandalam, India
5 Department of ECE, SRMIST, Kattankulathur, Chennai, India

albert.cse@sathyabama.ac.in

Abstract—The daily functioning of civilization depends heavily on
transportation. In most cities, a sizable section of the working class consistently
attends both employment and education. There are many different things you
can do when commuting, such as unwinding, eating out, and other things. The
most popular means of transportation in North Cyprus, particularly in
developing cities, is island transportation, which includes the usage of both
private cars and commercial vehicles. The advent of edge computing, which
offers the opportunity to connect potent processing servers next to the mobile
device, is a significant step toward improving user experience and reducing
resource use. Mobile Edge Computing is the next trustworthy approach for how
mobile devices consume communications and computing. Offloading
computation is a key component developing mobile edge computing, which
enables devices to get around clustering techniques' limitations and get around
computing, storage, and energy constraints. However, computation offloading
is not always the best strategy to use; making choosing unloading is a critical
step that requires consideration of numerous factors. For instance, shifting the
high-resource node to an edge server and granting similar capabilities to the
low-resource nodes would delegate heavy duties to the external unit inside the
network. The evaluations' results were noteworthy and substantial. Problems
involving the vehicles of institutions and organizations can be resolved using
the suggested solution to the school bus routing issue. We also test the impact
of network latency on the delivery of a particular result using an Edge
Computing simulator.

Keywords—mobile edge computing (MEC), transportation problems, save
time, clustering technique, energy efficiency, resource optimization

iJIM ‒ Vol. 17, No. 04, 2023 47

https://doi.org/10.3991/ijim.v17i04.37801

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

1 Introduction

Any city's growth is largely dependent on the state of its transportation
infrastructure and the services it provides to its populace. With access to real-time
information, an intelligent transportation system offers citizens quick and simple
options for safe transport. These facilities influence how quickly people move through
their daily lives and benefit society in many other ways, such as by reducing
pollution, boosting the economy, hastening development, improving health, and much
more. Numerous researchers have suggested various methods and schemes to hold the
large data produced by smart metropolises to provide more clever, smart, and
sustainable outcomes. However, these solutions don't just concentrate on the big data
aspect of real-time smart freight forwarding. By 2050, 75% of the world's population,
or more than 7 billion people, were expected to live in cities and nearby suburbs. As a
result, the amount of the city's traffic will considerably grow. Elevated big data will
be produced by the significant increase in city traffic intensity that will result from the
overabundance of vehicular communication and road sensors in the CPS ecosystem.
A proportionate spike in on-road traffic accidents may also result from this massive
increase in traffic intensity. As a result, residents will have issues with traffic
congestion and delayed travel times.

Due to the rapid advancement of technologies, consumers may desire to obtain any
on-road city traffic information at any time and from any location. In order to prevent
traffic congestion, the authorities may also need to distribute city traffic by skilfully
rerouting it to fewer congested transportations in actual time. Additionally, this
dispersion lessens air contamination, improving community health. In a nutshell,
traffic authorities are required to manage the traffic system wisely with the least
amount of human intervention and resources. This clever transport infrastructure
makes it easier for residents to access real-time transportation facts and has a big
impact on how people live [1].

Fig. 1. Mobile Edge Computing

Edge computing often referred to as mobile edge computing has recently emerged
to overcome the processing limitations of mobile devices in response to an increase in

Local Processing

Mobile Cloud
Application

Cloud Data
Center

Cloud Data Center
Request Processing

Task Execution
Computation

48 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

the number of computationally intensive apps. Edge computing relieves network
capacity constraints by distributing the workload to dispersed computing clusters.
Connections to servers are nevertheless slow, although cloud technology allows users
to access a shared pool of computers and move expensive computations to the cloud.
One of the key reasons why edge computing appears to have the ability to support
portable virtual reality is because of this. Edge servers keep a local reference image
database and use it to identify objects in image pixels. If the edge fails, the request
goes to the cloud [2].

The main contribution of this paper is, namely, as follows:

• We bring forth a problem for minimising the amount of time a task takes to
complete by jointly optimising its computational and networking resources, such as
its edge computing capabilities.

• In order to handle the original optimisation problem's non-linearity and non-
convexity, we separated it into smaller issues and iteratively resolved each one.
Additionally, we also provide a closed-form method at MEC for task splitting and
computing allocation of resources.

• To show the effectiveness of the suggested approach, statistical solutions are
contrasted with those from an extensive search and other benchmark schemes. The
numerical findings demonstrate that the proposed method outperforms all existing
schemes and performs epsilon identically to the extensive search when task
computation energy and time consumption are taken into consideration as
outstanding development [3].

The essay's remaining sections are organised as follows. Section 2 presents the
research on the pertinent prior work. Section 3 describes the features of the proposed
system, including the proposed system architecture, implementation model,
characteristics of the graph-based technique, and data analysis. The implementation
environment is described and the system's effectiveness is rated in Section 4. Section
5 provides the resolution.

2 Related works

Liu, L., Zhao et.al [4] additionally, some academics talked about the use of transfer
culture in MEC. Reference used distributed knowledge to simultaneously decide
which MEC networks should be offloaded. In order to address the trade-off between
latency and energy consumption, Reference established a hierarchical computational
intelligence task distribution framework. In this framework, the lower layers of the
pertained CNN model are embedded in the autonomous aerial vehicles, while the
higher layers are handled by the MEC server. In addition to demonstrating the
potential of deep learning and edge computing, this study also emphasises the
importance of generalisation and resource problems in useful uses. These
aforementioned methods frequently require much iteration to arrive at a local
optimum, making them inappropriate for real-world applications involving compute

iJIM ‒ Vol. 17, No. 04, 2023 49

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

offloading. Additionally, when the MEC network scale increases, their computational
cost tends to increase dramatically.

Sarfraz, M., Alshahrani et.al [5] this method of computing is referred to as MEC.
Unfortunately, due to the compute offloading link's current shortcomings, the MEC
model has not yet realised all of its promise. For example, when unloading
computation rather than completing the task locally, devices at the cell's perimeter
have a notoriously poor offloading success rate and/or may face increased delay. As a
result, these devices are forced to rely on their computing capacity, which is
frequently insufficient to execute resource-demanding applications. The efficiency of
the MEC systems' communications must be improved because of this.

abd Al_kadum Hassan, H., Hasan et.al [6] Vehicle-to-Vehicle system removes the
need for a main station to manage the network architecture and provides a connection
between vehicles using an ad hoc wireless network. Because of their high speed,
vehicular ad hoc networks (VANETs) are distinguished by the self-organization of the
vehicles and quick changes in the network topology. As link connections failures
frequently occur in VANETs, ensuring the security of information in VANETs is
more challenging than in typical MANETs. Creating a cluster with a hierarchical
system within the network is an effective and affordable way to reduce the mobility
effect and improve VANET network access.

Ahn, J., Lee, J., Park et.al [7] For the MEC environment, we suggest a power-
efficient clustering method (PECS) it keeps the MEC servers' processing of incoming
delays within acceptable bounds. By analysing the impact of the size of clusters on
the CPU workloads, the optimization problem for choosing the clustering to lower the
energy consumption of MEC servers is established and addressed. A thorough
examination of the suggested model shows that it effectively determines the ideal
number of clusters under the predetermined circumstances that satisfy the convexity
of the energy consumption model.

Zhang, X., Shen et.al [8] Applications like virtual reality, vr technology, and
language processing are among the many new ones that have emerged with the
growth of the mobile web and the popularity of smart terminals. These applications
typically have asset resources and demand a lot of computing and storage resources to
operate, which has an impact on the quality of the service. Although the performance
of smart terminal processors is constantly improving, they are unable to quickly
process high-performance programs, which have a significant negative influence on
the user's service encounter.

Rathore, M. M., Attique Shah et.al [9] transportation guarantees to provide the
correct evidence at the proper period, location, and device to assist the public in
making any transportation-related decisions. The vehicular network and the system of
allow are two examples of these two components. To collect data and determine
traffic information, a small computer and road sensors are installed at each
intersection of the road, such as the volume of cars, average vehicle speed, and traffic
conditions intensity. In contrast, the vehicular network infrastructure is utilised to
obtain details about each individual vehicle, including its location and speed. Relay
nodes, coordinator, gates, integrators, and classifiers are all utilised by the suggested
system to connect the two subsystems.

50 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Potts, C. M. et.al [11] the second challenge, data prioritising, requires a researcher
or user to effectively comb through a big dataset in search of crucial data. We assume
that the dataset is too big to look through manually, necessitating the use of a
computer engine. It has to do with how we order and prioritize analysis and search
terms. The second objective is about deciding what data to offer a user, whereas the
first task is about generating knowledge from data. This interactive component is
crucial since experts may become quickly dissatisfied, possibly give up on the system,
or make a vital error with potentially serious repercussions if important information is
hidden or overshared.

3 Methods and materials

Many other approaches can be used to explain cluster stability; our method
computes stability from arguments of distance travelled, velocity variability, and
possibility. We carefully consider value stability when choosing a cluster head. In
order to communicate with other vehicles, we assume that all vehicles have IEEE
803.12p radio transceivers and GPS systems, which individually allow for the
acquisition of position-related data. Each car can use the clustering process in the
suggested algorithm. A car notifies its neighbours when it needs help locating a
cluster head. If there is no response, the process of creating groups is initiated. One
cluster contains all the vehicles that are travelling in the same direction. Each car
sends a message to the other members of the group, which includes the relocation,
Identification, and velocity. After that, it calculations by arranging the inputs in a
neighbourhood array:

• The separation between the actual vehicle Q and its Neighbour P is:

𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑄𝑄,𝑌𝑌 = ���𝑀𝑀𝑞𝑞 − 𝑀𝑀𝑝𝑝� + �𝑍𝑍𝑞𝑞 − 𝑁𝑁𝑝𝑝�� (1)

Where the location of the car is (m, n)

• The difference in speed between vehicle Q and its neighbour U.

∆𝑈𝑈𝑞𝑞,𝑝𝑝 = 𝑈𝑈𝑞𝑞 − 𝑈𝑈𝑝𝑝 (2)

• The likelihood that the car is the cluster head:

𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝐷𝐷𝑝𝑝𝑝𝑝𝐷𝐷𝐷𝐷𝑝𝑝 = (𝐹𝐹 + 3 ∗ 𝑑𝑑𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝑝𝑝 𝑝𝑝𝑜𝑜 𝐷𝐷ℎ𝐷𝐷 𝐷𝐷𝑝𝑝𝑑𝑑𝐷𝐷 + 𝑈𝑈)/𝑄𝑄𝑚𝑚𝑚𝑚𝑚𝑚 (3)

Where F is the node's power consumption when transmitting or getting a package
(here, energy is the electricity a node uses throughout the data communication), and U
is the vehicle's speed. Likelihood will be in the range of 1 and 3. The transport
density, speed, and range of F have all been normalised (1–3). The greatest value of
the calculated probability will be applied to all collected information. The likelihood
that a car will act as the cluster head will therefore be calculated each time by
dividing the result by the number 3 (𝑄𝑄𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏); the resulting ranging value of likelihood

iJIM ‒ Vol. 17, No. 04, 2023 51

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

will have a scale of 1-3. Remember that the value approaches 2 to suggest a poor
choice, while the value approaches 1 to indicate the ideal candidate because it would
be a problem of reduction; the best is the minimal.

Clusters are built after the cluster heads are chosen, and then communication
between the cluster heads begins. It would be more likely for one cluster head to
replace another if it used the least amount of energy throughout transmission. Energy
in this context refers to the electrical power used by a node during transfer.

𝐹𝐹 = 𝐿𝐿 ∗ 𝐹𝐹𝑐𝑐𝑐𝑐𝑚𝑚 (4)

Where 𝐹𝐹𝑐𝑐𝑐𝑐𝑚𝑚 describes the wireless send's energy consumption.

• The transmitting chain's RQ,P stability factor calculated for each neighbouring car
P is:

𝑅𝑅𝑞𝑞,𝑝𝑝 =∝ 1, ∝ 2 ∗ 𝑑𝑑𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷+∝ 3∇𝑈𝑈𝑈𝑈, 𝑝𝑝 ∗ 𝑄𝑄+∝ 4 (5)

Where the numbers are ∝ 1, ∝ 2, ∝ 3, 𝐷𝐷𝐷𝐷𝑑𝑑 ∝ 4 referred to as the real constituents
of the model.

|∝ 1. . ∝ 4| = (𝑚𝑚𝑗𝑗 ∗ 𝑚𝑚𝑙𝑙) ∗ (𝑅𝑅 ∗ 𝑌𝑌) (6)

Where the effects of the matrix are R, output, and Y are representing a cluster's
stability. Cluster stability is a significant objective that clustering techniques work to
achieve and is a valid metric for assessing the effectiveness of a clustering method.
There are several ways to explain cluster durability, and our method computes it
based on arguments about distance, velocity variation, and chance. Cluster head is
chosen carefully, keeping value consistency in mind. The experiment's findings were
utilized to modify each of the three variables (vehicle speed differential, distance U t,
and likelihood) within the following ranges:

Table 1. Variable findings

• Stability R component:

All cars in a cluster exchange stability measurement with their nearby counterparts.
The vehicle with the greater value of steadiness is the cluster head.

𝑅𝑅 = ∑𝑅𝑅𝑝𝑝,𝑞𝑞 (7)

Procedure:
The suggested methodology's first algorithm follows the following steps: The

storage space, transmission range, and energy requirements will be used to compile a
list of backup mobile edge nodes that could be used. A backup mobile edge-node

Small Grade: +2 Tall Degree: -2
Expanses Close distant
Speediness Short great
Probability Short great

52 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

notifies the cluster through email before moving head alerting it to start a fresh
selection process for a new backup node if it left its clusters (took another
intersection). Another backup node with a high score will be chosen by the cluster
leader from the list of candidates. After a minute, if the handoff is still not working,
the backup mobile edge-node welcomes the user. The area has been improved, and
the BSI list is handled and kept current.

Algorithm 1:
Input:
Vehicle group Yd1 = "Y11, Y12, Y13,..., Y1n"
Output:
The groups of vehicles to be chosen from set Yd1 as

cluster midpoints are Hd1 = "g11, g12, g13,..., g1l"
1. The vehicles travelling in a fixed direction (e1)

while inside the range of a certain
2. GAM is a member of the set Yd1
3. The cluster midpoints are chosen at random from set

Yd1 by 'M' cars, and
4. set ld1 sort order
5. Recap
6. Do for each k from 2 to M
7. The cluster's midway is one of the cars ck from set

Ld1
8. The separation, distance(di, yj), between each yj

vehicle in set yd1 and di is
9. Calculated
10. In the event when |dj| = 40 && max (distance(di,

yj)) = d do
11. The vehicles dj with the closest separation from xi

are connected to
12. Set xi has dj and arranged in it.
13. Complete
14. Complete
15. Till
16. The created clusters' new cluster midpoints are

determined using to find b)
17. better answer, saved as xi (unique)
18. To create new membership of xi(unique), which

substitutes for dj, repeat steps 3-7
19. dj adopting the new median, xi
20. Repeat steps 4 through 9 until two novel cluster

averages, dj (unique), are discovered.

Due to the complicated nature of a road network and the frequent entry and exit of
vehicles, the connection link is unstable due to the difficulties of mobile nodes that
exist with vehicular traffic, and there is a substantial danger that the link will fail [4].

iJIM ‒ Vol. 17, No. 04, 2023 53

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Due to the dynamic nature of our suggested DEBCK, which automatically forms the
backup cluster node and assumes control whenever a topology breakdown occurs as a
result of vehicle redirections, it was developed to address this issue.

3.1 Mobile edge computing

MEC is a workable solution since it provides on-demand access to high-volume
compute. As a result, users at the MEC split the server's processing power. The
computational resources allotted to ℎ𝑞𝑞

𝑓𝑓 by MEC are represented by E n. Similarly,
max denotes the maximum allowable computation on the MEC server. Two other
factors affect how long it took MEC to complete the work in total: (1) The duration of
data transfer from the UEs to the MEC server, as described in (2); (2) MEC http
computation. As a result, the following is how long it took to complete the assignment
on the MEC server [5]:

ℎ𝑞𝑞
𝑓𝑓 = ∝𝑛𝑛𝑑𝑑𝑛𝑛

𝑙𝑙𝑚𝑚
𝑓𝑓 +

∝𝑛𝑛𝑟𝑟𝑛𝑛
𝑏𝑏𝑛𝑛

(8)

The continual power from the grid also leads us to believe that the MEC's battery
life is limitless. As a result, the power consumption of the Edge servers while it is
handling queries is disregarded in this case. Similar to this, the authors of state that we
overlook the findings' transmission rate from the Edge servers to UE because of their
short size. The fundamental concept behind edge devices is to move the cloud
platform from the mobile core network's inside to the mobile communication
network's edge in order to achieve flexibility resource consumption. In order to enable
high computational and delay programs on edge devices that have limited resources,
portable edge computing brings portable computing [12], network management, and
storage systems to the edges of the network.

3.2 Formulation of a problem

In this study, we sought to reduce the task's computational time by jointly
maximising communications and computing capabilities between the user, MEC, and
task division variable. Additionally, for the sake of simplicity, we define {q = q1, q2,
qn} {k, = k 1, k 2, k n, and = 1, 2, n}. The computing time minimization problem can
therefore be represented formally as follows:

𝑄𝑄1: min 𝑚𝑚𝐷𝐷𝑚𝑚∝,𝑞𝑞 = �
2−∝𝑚𝑚
ℎ𝑚𝑚
𝑓𝑓 ,

∝𝑛𝑛𝑑𝑑𝑛𝑛
𝑙𝑙𝑚𝑚
𝑓𝑓 +

∝𝑛𝑛𝑟𝑟𝑛𝑛
𝑏𝑏𝑛𝑛
� (9)

𝐷𝐷1: 𝑞𝑞𝑚𝑚∝𝑚𝑚𝑟𝑟𝑚𝑚
𝑏𝑏𝑛𝑛

+ ∁𝑛𝑛(2 −∝𝑛𝑛)𝐷𝐷𝑛𝑛 ≤ 𝑜𝑜𝑚𝑚
𝑚𝑚𝑚𝑚𝑚𝑚,∇𝑛𝑛 (10)

𝐷𝐷2: ∑ ℎ𝑞𝑞
𝑓𝑓 ≤𝑀𝑀𝑚𝑚=2 𝑜𝑜𝑚𝑚

𝑚𝑚𝑚𝑚𝑚𝑚, (11)

𝐷𝐷3: 𝑈𝑈𝑚𝑚 ≤ 𝑄𝑄𝑚𝑚𝑚𝑚𝑚𝑚, ∝𝑛𝑛∈ (1,2), ∀𝑚𝑚, (12)

54 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

𝐷𝐷4: |∄𝑙𝑙| = 2, ∇𝑙𝑙 (13)

While constraint (8d) specifies that the computational resources allocated to UEs at
the MEC server should be less than the highest computational resources, constraint
(7b) ensures that the overall amount of energy used to calculate the task will be less
than the highest amount of energy the battery can The optimization problem that is
covered in sections (7a) through (7e) is also mixed-integer, non-convex, and non-
linear in nature. This is due to the logarithmic function present in the rate equation.
We divided the main problem into a number of smaller difficulties in order to handle
this challenge. Additionally, the sub-optimization problems are then resolved
repeatedly in order to identify the best possible solution. Hold E max n. The
transmission power limitations of the UEs are represented by restriction (5h), in a
similar manner. Then, pitch shift control of IRS components is provided by (6h).

3.3 The system model

In the Mobile nodes, clustering is the technique of assembling nearby moving cars
on a street into stable groups to speed up information sharing between vehicles. A
roadway model with a lot of automobiles moving in one direction is taken into
consideration. The network is made up of several vehicles that are located on the
street and are regarded to be members of various groups, with one of the members of
each group acting as the head of the group. Presuming that vehicles can connect with
one another to exchange information, such as safeguard texts, the network is made up
of vehicles that assume this assumption (CH). A number of characteristics, including
position, velocity, and steadiness, define the CH. Regardless of the speed and density
of the cars, all groups evolve over time because a new group leader is always selected.
The message was sent over the network via the cluster head from one cluster head
vehicle to another cluster head vehicle in the neighbouring cluster until it reached the
destination vehicle. Clustering model illustration is shown in Figure 2 [6].

Fig. 2. Clustering model illustrations

C1 C2 C3

Direction

Cluster 1 Cluster 2 Cluster 3

iJIM ‒ Vol. 17, No. 04, 2023 55

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

3.4 Cluster head selection and cluster composition

The network starts out as a collection of automobiles on a route, split up into
several vehicle groups within its communication range. The cluster head (CH), which
serves as the group's leader, is symbolized by the number of vehicles that make up
each cluster (CM). The first CH is selected at random, and the second cluster head is
decided upon as the vehicle that is placed furthest away from the target vehicle inside
the first head's range. All vehicles that are in the head's field of vision are regarded as
belonging to the group. The steps in cluster composition are shown in Figure 3.

Fig. 3. Steps in cluster composition

3.5 Method for selecting cluster heads

The proposed LEC-steps SEP's are shown in below Method. The CHs broadcast
themselves to the network after being chosen by the system in Figure 4. The non-
cluster head nodes decide to join the closest cluster after receiving the message and
joining the cluster. The following formula describes how to cluster data [10]. CH
number is the optimal cluster set as determined by the calculations, CH member is a
cluster member node, and CH count is the total number of clusters in the current
round.

Algorithm: Cluster head selection and cluster
composition:
1. All vehicles must be on the roadway
2. idly set each node's track
3. determine each vehicle's speed
4. choosing the first CH
5. Identify the sending vehicle
6. The first cluster should contain all vehicles within

the FCH's range.
7. as long as (nodes!=unfilled)
8. find the following CH
9. All nodes + node clustering
10. Clusters are updated by changing the CH during the

designated time.
11. Finish while
12. finish

Designated
The CH

Control The
Group

Member

Designated
The Next

56 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Fig. 4. Design of the suggested system's deployment for intelligent transportation

4 Implementation and experimental results

A MATLAB simulator is used to examine the energy usage according to the
number of clusters in order to evaluate the effectiveness of the proposed Pectorals.
The effectiveness is assessed using a random system made up of 150 servers that are
distributed randomly around a normalized circular area with mg flows [7]. User job
characteristics, channel capacity disparities across base stations, and accessible
computing resources from mobile edge computing servers are carefully taken into
account while keeping in mind user fairness for mobile systems that allow dense
connectivity [8].

Transportation Data

Doorway

WEB

Assemblage Part

Organization

Apprising

Chart Storing

Uploading

Chart Dispensation

Outcome Combination

Elucidation

Experts Residents

Demand Demand

Reply Grade

iJIM ‒ Vol. 17, No. 04, 2023 57

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Table 2. Determining the Multi-Stage Asset Allocation Server's Simulation Parameters

Limitations Assessment
capacity for base stations 550
D2D link speed 100
The processing power of an MEC server 40/SEC
The quantity of computing resources needed to finish the task [0.8-0.9]cycles
The capability of Server J mg
factor in heaviness 1.5
The ability to compute(mg) [0.11-0.13]cycles

The suggested method first analyses the traffic dataset that was obtained from

Aarhus, Denmark's second-largest city by people. An investigation of how traffic
density affects vehicle speed is shown in Figure 5. When the transportation intensity
is higher, that is, when there are more between two locations on the road, the average
vehicle speed is seen to be meaningfully lower between any two road sections with a
resolution of 500 m. Similar to this, the contrary is also seen, namely that the average
speed increases as fewer cars are on the road. The graph's unfilled circle line
illustrates how the average speed drops dramatically during different times of the day
when there are a substantial number of cars (550–650). However, at times when there
aren't many cars (1 000–4 000), yet the observed average speed of vehicles is
considerable, as indicated by the full circle line on the graph. These situations do not
always occur, and some abnormalities, like a relative low seed despite a low
automobile intensity, can also be seen. These circumstances are primarily attributed to
causes like construction on the roads, incidents, or weather conditions like rainfall,
foggy, etc [9].

Fig. 5. When mg = 1000, power consumption of MEC is dependent on the cluster centres

1000 2000 3000 4000

E
N

E
R

G
Y

C
O

N
SU

M
P

E
T

IO
N

AMOUNT OF CLUSTERS

mg=1000

mg=2000

mg=3000

58 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

This is due to the fact that handling focused requests by a small number of servers
while leaving a large number of others idle uses more energy than handling fairly
spread queries by all servers using PECS. If m is small, a small number of mg servers
handle all requests for m servers, and many member servers evenly split requests
between them. Since, in our system model, queries to mg servers and to member
server are constant. Thus, as shown in Figure 5, the load demand increases when k is
small and reduces as mg increases as a result of the distribution of load among MEC
servers. Beyond a certain threshold, however, the concentration of demands on a
limited number of members causes the energy usage to tend to increase. For instance,
the MEC's computational capabilities may not be available, or the offload may not be
carried out if it is not profitable [13].

When mg = 550 in Figure 6, the dynamic positioning of MVEs causes an increase
in power usage. The start of the graph is eliminated because l ought to be a part of
mg. Because more requests result in increased CPU demand, PECS similarly grows as
the amount of flows does.

Fig. 6. Comparison of energy use when mg = 550

In Figure 7 and 8, researchers contrast the PECS program's average latency and
power usage with that of two instances of PECS providing 5G service with Treq of
550 and 1000 mg, correspondingly. Pecs(l) rises in tandem with the number of flows.
When the k value that minimises the average delay deviates from K, we identify a
replacement mg value for l that is somewhat close to the original k value. The task
execution overhead of the system continues to decline as the MEC server's processing
capacity rises [8]. This is the cause of the saw-tooth pattern on the PECS graph. The
best performance for PECS occurs when Treq is 600ms. A 13.34% reduction in power
usage is attained with PECS when mg is 4000.

1000 2000 3000 4000

E
N

E
R

G
Y

C
O

N
SU

M
P

T
IO

N
S

TOTAL NUMBER OF FLOWS

PECS=550mg

PECS=650mg

PECS=750mg

iJIM ‒ Vol. 17, No. 04, 2023 59

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

Fig. 7. When mg = 2000, the power usage of PECS is determined by the number of clusters

The overall latency of inputs processed by PECS servers, mg, is sustained by
PECS at the acceptable level, T(request), independent of the mg value, as shown in
Figure 8. Because PECS searches for the average delay is smaller than L (request)
when there are few flows in the MEC environment, the optimal mg to minimise
energy usage in these circumstances has a longer average delay than the PECS
technique.

Fig. 8. When mg = 650, compare the average delay

0%
10%
20%
30%
40%
50%
60%
70%
80%
90%

100%

1000 2000 3000 4000

E
N

E
R

G
Y

C
O

N
SU

M
P

T
IO

N
S

AMOUNT OF CLUSTERS

mg(nf)=3000

mg(nf)=2000

mg(nf)=100

Additional
Increase Due To
Pecs
Replacement

0%
10%
20%
30%
40%
50%
60%
70%
80%
90%

100%

1.5 2.5 3.5 4.5

T
IM

E
(S

E
C

)

TOTAL NUMBER OF FLOWS

CNSC

PECS,T(req)=650mg

PECS,T(req)=550mg

60 http://www.i-jim.org

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

5 Conclusion

The country's economy and commuters' social lives are both strongly impacted by
transportation systems. This study concentrated on the tasks of traffic analysis and
prediction, and it provides a current collection of accessible datasets and tools as a
resource for people looking for open-source materials. In order to more accurately
estimate and predict traffic states, it is also advised to collect and use external data,
such as weather forecasts, calendar details, air reduced pollution, and sound. The
simulation findings demonstrate that, when the suggested algorithm is less impacted
by fluctuations in the number of cluster heads, it can more efficiently extend the
stability period and transfer more data.

6 References

[1] Rathore, M. M., Attique Shah, S., Awad, A., Shukla, D., Vimal, S., & Paul, A. (2021). A
cyber-physical system and graph-based approach for transportation management in smart
cities. Sustainability, 13(14), 7606. https://doi.org/10.3390/su13147606

[2] Uma, D., Udhayakumar, S., Tamilselvan, L., & Silviya, J. (2020). Client aware scalable
cloudlet to augment edge computing with mobile cloud migration service.

[3] Sarfraz, M., Alshahrani, H. M., Tarmissi, K., Alshahrani, H., Elfaki, M. A., Hamza, M. A.,
& Khurshaid, T. (2022). Intelligent Reflecting Surfaces Enhanced Mobile Edge
Computing: Minimizing the Maximum Computational Time. Sensors, 22(22), 8719.
https://doi.org/10.3390/s22228719

[4] Liu, L., Zhao, Y., Qi, F., Zhou, F., Xie, W., He, H., & Zheng, H. (2022). Federated Deep
Reinforcement Learning for Joint AeBSs Deployment and Computation Offloading in
Aerial Edge Computing Network. Electronics, 11(21), 3641. https://doi.org/10.3390/
electronics11213641

[5] Sarfraz, M., Alshahrani, H. M., Tarmissi, K., Alshahrani, H., Elfaki, M. A., Hamza, M. A.,
& Khurshaid, T. (2022). Intelligent Reflecting Surfaces Enhanced Mobile Edge
Computing: Minimizing the Maximum Computational Time. Sensors, 22(22), 8719.
https://doi.org/10.3390/s22228719

[6] abd Al_kadum Hassan, H., Hasan, Z. Y., & Al Taie, R. H. (2022). A simulation approach
to improve the VANETs communication. iJIM, 16(12), 137. https://doi.org/10.3991/ijim.
v16i12.31423

[7] Ahn, J., Lee, J., Park, S., & Park, H. S. (2019). Power efficient clustering scheme for 5G
mobile edge computing environment. Mobile Networks and Applications, 24(2), 643-652.
https://doi.org/10.1007/s11036-018-1164-2

[8] Zhang, X., Shen, H., & Lv, Z. (2021). Deployment optimization of multi-stage investment
portfolio service and hybrid intelligent algorithm under edge computing. PLoS One, 16(6),
e0252244. https://doi.org/10.1371/journal.pone.0252244

[9] Rathore, M. M., Attique Shah, S., Awad, A., Shukla, D., Vimal, S., & Paul, A. (2021). A
cyber-physical system and graph-based approach for transportation management in smart
cities. Sustainability, 13(14), 7606. https://doi.org/10.3390/su13147606

[10] Du, R., Liu, Y., Liu, L., & Du, W. (2020). A lightweight heterogeneous network clustering
algorithm based on edge computing for 5G. Wireless Networks, 26(3), 1631-1641.
https://doi.org/10.1007/s11276-019-02144-x

iJIM ‒ Vol. 17, No. 04, 2023 61

https://doi.org/10.3390/su13147606
https://doi.org/10.3390/s22228719
https://doi.org/10.3390/electronics11213641
https://doi.org/10.3390/electronics11213641
https://doi.org/10.3390/s22228719
https://doi.org/10.3991/ijim.v16i12.31423
https://doi.org/10.3991/ijim.v16i12.31423
https://doi.org/10.1007/s11036-018-1164-2
https://doi.org/10.1371/journal.pone.0252244
https://doi.org/10.3390/su13147606
https://doi.org/10.1007/s11276-019-02144-x

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

[11] Potts, C. M. (2022). Interactive Data Analysis for Experts: Tools, Techniques, and
Assessments.

[12] Wang, T., Lu, Y., Cao, Z., Shu, L., Zheng, X., Liu, A., & Xie, M. (2019). When sensor-
cloud meets mobile edge computing. Sensors, 19(23), 5324. https://doi.org/10.3390/
s19235324

[13] Mach, P., & Becvar, Z. (2017). Mobile edge computing: A survey on architecture and
computation offloading. IEEE communications surveys & tutorials, 19(3), 1628-1656.
https://doi.org/10.1109/COMST.2017.2682318

[14] Julian Menezes R, Albert Mayan J, Breezely George M. (2015). Development of a
Functionality Testing Tool for Windows Phones. Indian Journal of Science and
Technology, 8(22), 1-7. https://doi.org/10.17485/ijst/2015/v8i22/79101

7 Authors

Dr. J. Albert Mayan is an Associate Professor in the School of Computing,
Sathyabama Institute of Science and Technology, India. With 21 Years of experience
in teaching and R&D, Dr. Albert had published a number of papers in Indexed
Journals. Dr Albert had presented various research-based papers at several national
and international conferences. Dr Albert had obtained his Ph.D in the Field of
Software Testing and his areas of interest include Machine Learning, Image
processing and Data Mining. Dr Albert is an active member of ACM Student Chapter
and OWASP Student Chapter at Sathyabama Campus (email: albert.cse@
sathyabama.ac.in).

Dr. S. V. Manikanthan is a Director of Melange Academic Research Associates,
Puducherry, India. His area of Research Interest is Wireless Sensor Networks. He has
21 years of experience in Anna University Affiliated Colleges and in Industrial
Research Projects (email: prof.manikanthan@gmail.com).

Azham Hussain is the Associate Professor of Software Engineering at UUM
School of Computing. He is the founder and head of Human-Centered Computing
Research Group which is affiliated with the Software Technology Research Platform
Center at School of Computing, Universiti Utara Malaysia. Assoc. Prof. Azham
Hussain is a member of the US-based Institute of Electrical and Electronic Engineers
(IEEE), and actively involved in both IEEE Communications and IEEE Computer
societies. Azham is published in the areas of software evaluation and testing, user
behaviours, group collaboration, and ubiquitous and mobile technology design. He
has authored and co-authored more than 100-refereed technical publications, served
as reviewer and referee for refereed journals and conferences on computing as well as
the examiner for more than twenty doctoral and postgraduate scholars in his research
areas (email: azham.h@uum.edu.my).

Dr. S. Nithyaselvakumari presently working at Saveetha school of Engineering as
Assistant professor SG in the Department of Medical Instrumentation and
experienced around 11 years in teaching and 5 year in Industry. She has completed
Bachelor of Engineering in Electronics and communication Engineering at Mahendra
Engineering college Salem under madras university and done Master of Engineering
in Sathyabama university in the department of Applied Electronics. In 2021, she has

62 http://www.i-jim.org

https://doi.org/10.3390/s19235324
https://doi.org/10.3390/s19235324
https://doi.org/10.1109/COMST.2017.2682318
https://doi.org/10.17485/ijst/2015/v8i22/79101

Paper—Clustering Technique for Mobile Edge Computing to Detect Clumps in Transportation-Related…

completed Ph.D under Anna University. Her area of research is in Bio Medical,
wireless communication, Embedded system and Internet of Things (email:
nithyaselvakumari.s@gmail.com).

A. Vinnarasi is an Assistant professor at the Department of ECE, SRMIST,
Chennai. Doing PhD.in SRMIST with total working experience of more than 17 yrs.
M. Tech in Embedded Systems and Technology, in SRM 2005. Completed BE in
ECE 2002. Pursuing PhD in STMIST. Worked in Saveetha Engineering College as
Lecturer for 1yr 6 months. Worked in Sri Muthukumaran Institute of Technology as
Assistant professor for 4yrs 5 month. Currently working in SRMIST as Assistant
professor till date with 12yrs experience (email: vinnaraa@srmist.edu.in).

Article submitted 2022-11-05. Resubmitted 2022-12-16. Final acceptance 2022-12-28. Final version
published as submitted by the authors.

iJIM ‒ Vol. 17, No. 04, 2023 63