2019 Performance Analysis and Optimization of Next Generation Wireless Networks
2019 Performance Analysis and Optimization of Next Generation Wireless Networks
Emmanouil Skondras
Department of Informatics
School of Information and Communication Technologies
University of Piraeus
The publications made in the context of this PhD thesis have been partly supported by the
University of Piraeus Research Center (UPRC) and the Technological Educational Institute of
Western Macedonia.
Abstract
The Fifth Generation (5G) networks, including the 5G Vehicular Cloud Computing (5G-VCC)
systems, have evolved rapidly offering multiple services to users. The operating principles of
vehicular networks, Cloud Computing (CC), Fog Computing (FC), Mobile Edge Computing
(MEC) and Software Defined Networks (SDN) are applied to 5G infrastructures. In a 5G-VCC
system, the vehicles are equipped with On-Board Units (OBUs) which communicate with each
other as well as with Road Side Units (RSUs). Each RSU interacts with a Cloud infrastructure
which offers vehicular services with strict Quality of Service (QoS) requirements, including
Driver Assistance (DA), Passengers Entertainment and Information (PEnI) and Medical (MED)
services. Dense deployments of 5G access networks are also implemented, called Ultra Dense
Networks (UDNs), aiming to support high data rates produced by an increased number of
vehicular users. In this environment, heterogeneous technologies are used to transfer the
network services to vehicles. Optimal manipulation of the communication resources is required,
while at the same time vehicular users should always obtain connectivity to the most appropriate
network access technology, in order the constraints of the vehicular services to be satisfied. In
this thesis, existing schemes for resource allocation as well as for mobility management are
studied, while novel solutions are proposed for each topic.
Initially, the theoretical background of the 5G wireless networks and 5G-VCC systems
is described. Subsequently, the available delivery models for providing cloud services in 5G
infrastructures are mentioned, while the available 5G-VCC architectures and the communication
models are presented.
Regarding the topic of the manipulation of the available network resources in 5G-VCC
systems, the available Medium Access Control schemes (MAC schemes) are studied, while
at the same time resource scheduling algorithms implemented at the MAC layer of 5G-VCC
systems are described. Subsequently, two novel solutions are proposed to optimize the resource
allocation process in modern networks. The first one is called FLS Advanced (FLSA) and opti-
mizes the resource allocation for real-time services. Additionally, the second one is called FLS
Advanced - Cross Carrier (FLSA-CC) and is specialized for LTE-Advanced (LTE-A) access
networks where carrier aggregation is applied in order to provide widened bandwidth to the
x
users. The evaluation of the proposed algorithms is performed through extended experiments.
Simulation results show that the proposed algorithms outperform existing solutions.
The mobility management requirements arising at the 5G-VCC systems are also studied.
As a result of this study, initially three methods for calculating the significance of the criteria
affecting the mobility management are proposed, namely the Trapezoidal Fuzzy Analytical
Network Process (TF-ANP), the Trapezoidal Fuzzy Adaptive Analytical Network Process
(TF-AANP) and the Pentagonal Fuzzy Analytical Network Process (PF-ANP). Subsequently,
three network selection algorithms are proposed, namely the Trapezoidal Fuzzy TOPSIS
(TFT), the Trapezoidal Fuzzy TOPSIS with Adaptive Criteria Weights (TFT-ACW) and the
Pentagonal Fuzzy TOPSIS (PFT). Following this study and regarding the research results of the
implemented algorithms, a mobility management scheme is proposed. This scheme takes into
consideration the Signal to Noise plus Interference (SINR) parameters as well as the velocity of
the vehicular users to evaluate the necessity of performing a handover. Then, it applies the PFT
algorithm to select the most appropriate network for the user. Experimental results showed that
the proposed scheme outperforms the existing solutions described in the research literature.
Περίληψη
Η μελέτη της διατριβής εντάσσεται στην ερευνητική περιοχή των ασύρματων τηλεπικοινω-
νιακών συστημάτων και ειδικότερα των δικτύων Πέμπτης Γενιάς (Fifth Generation -5G).
Τα δίκτυα 5G, συμπεριλαμβανομένων των συστημάτων 5G Vehicular Cloud Computing (5G-
VCC) , έχουν εξελιχθεί με ταχείς ρυθμούς προσφέροντας πολλαπλές σύγχρονες υπηρεσίες
στους χρήστες. Σε μία υποδομή 5G, συνηθίζεται να εφαρμόζονται οι αρχές λειτουργίας
που διέπουν τα οχηματικά δίκτυα, το Cloud Computing (CC) , το Fog Computing (FC) , το
Mobile Edge Computing (MEC) και τα Software Defined Networks (SDN) . Επιπρόσθε-
τα, σε ένα σύστημα 5G-VCC, τα οχήματα είναι εξοπλισμένα με υπολογιστικές μονάδες επί
του οχήματος (On Board Units - OBU) οι οποίες επικοινωνούν μεταξύ τους καθώς και με
υπολογιστικές μονάδες εγκατεστημένες δίπλα στο οδικό δίκτυο (Road Side Units - RSU).
Κάθε μονάδα RSU αλληλεπιδρά με μία υποδομή Cloud, η οποία προσφέρει υπηρεσίες με
αυστηρές απαιτήσεις ως προς την Ποιότητας της Υπηρεσίας (Quality of Service - QoS).
Ενδεικτικά, υπηρεσίες υποστήριξης του οδηγού (Driver Assistance - DA) , υπηρεσίες ψυχα-
γωγίας και ενημέρωσης των επιβατών (Passengers Entertainment and Information - PEnI) ,
καθώς και ιατρικές υπηρεσίες (Medical - MED) συνηθίζεται να παρέχονται στα οχήματα
από την προαναφερθείσα υποδομή Cloud. Επίσης, σε μία αρχιτεκτονική 5G, συνηθίζεται η
ανάπτυξη πυκνών υποδομών πρόσβασης στο δίκτυο, οι οποίες αναφέρονται ως Ultra Dense
Networks (UDN) . ΄Ενα UDN αναπτύσσεται με στόχο την υποστήριξη των υψηλών ρυθ-
μών δεδομένων που παράγονται από τον αυξημένο αριθμό οχηματικών χρηστών. Στο εν
λόγω περιβάλλον ασύρματης πρόσβασης χρησιμοποιούνται ετερογενείς τεχνολογίες για τη
μεταφορά των δικτυακών υπηρεσιών από το Cloud στα οχήματα.
΄Οπως γίνεται σαφές, απαιτείται βέλτιστος χειρισμός των τηλεπικοινωνιακών πόρων,
ενώ ταυτόχρονα οι χρήστες των οχημάτων θα πρέπει πάντα να αποκτούν συνδεσιμότητα με
την καταλληλότερη τεχνολογία πρόσβασης στο δίκτυο, προκειμένου να ικανοποιούνται οι
περιορισμοί των υπηρεσιών που χρησιμοποιούν.
Στο πλαίσιο της παρούσας διδακτορικής διατριβής μελετώνται τα μοντέλα κατανομής
πόρων, καθώς και τα μοντέλα διαχείρισης της κινητικότητας των χρηστών, που έχουν προ-
ταθεί από την ερευνητική κοινότητα, ενώ νέες βελτιστοποιημένες λύσεις προτείνονται για
κάθε ένα από τα εν λόγω ζητήματα. Συγκεκριμένα, αρχικά αναλύεται το θεωρητικό υ-
xii
πόβαθρο στο οποίο βασίζεται η διατριβή. Περιγράφονται οι βασικές αρχές που διέπουν τα
συστήματα 5G, και πραγματοποιείται μνεία στα συστήματα 5G Vehicular Cloud Computing
(5G-VCC). Πραγματοποιείται επίσης μία επισκόπηση των διαθέσιμων μοντέλων παροχής
υπηρεσιών υπολογιστικού νέφους, καθώς και κατηγοριοποίησή τους. Ακολούθως, πραγ-
ματοποιείται εκτενής επισκόπηση των ασύρματων δικτύων 5G, με έμφαση στα συστήματα
5G-VCC που έχουν προταθεί στην ερευνητική βιβλιογραφία, όπου αναλύονται οι αρχιτε-
κτονικές τους, τα μοντέλα επικοινωνίας μεταξύ των συστατικών τους, καθώς και οι αρχές
που εφαρμόζονται για την διάθεση των υπηρεσιών που αυτά προσφέρουν.
Στη συνέχεια, εξετάζεται ο τομέας της διαχείρισης των δικτυακών πόρων των ασύρμα-
των δικτύων επόμενης γενιάς, συμπεριλαμβανομένων των δικτύων 5G και των συστημάτων
5G-VCC. Στο πλαίσιο της μελέτης αυτής, προτείνονται δύο νέοι αλγόριθμου χρονοπρο-
γραμματισμού για ασύρματα δίκτυα 5G. Συγκεκριμένα, αρχικά προτείνεται ο αλγόριθμος
χρονοπρογραμματισμού FLS Advanced (FLSA) , ο οποίος βελτιστοποιεί την εξυπηρέτηση
υπηρεσιών πραγματικού χρόνου. Στη συνέχεια προτείνεται ο αλγόριθμος χρονοπρογραμμα-
τισμού FLS Advanced - Cross Carrier (FLSA-CC) ο οποίος αποτελεί εξέλιξη του αλγορίθμου
FLSA του οποίου η εφαρμογή εξειδικεύεται για δίκτυα LTE-Advanced (LTE-A) όπου πραγ-
ματοποιείται συνένωση πολλαπλών φερουσών συχνοτήτων (carrier aggregation) με σκοπό
τη διεύρυνση του διαθέσιμου εύρους ζώνης.
Εξετάζεται επίσης ο τομέας της διαχείρισης της κινητικότητας των χρηστών ασύρματων
δικτύων επόμενης γενιάς. Αρχικά, πραγματοποιείται μελέτη των απαιτήσεων που υπάρ-
χουν στα ασύρματα δίκτυα 5G, με έμφαση στα συστήματα 5G-VCC, ως προς τη διαχείριση
της κινητικότητας των χρηστών. Ακολούθως, προτείνεται ένα σύνολο μεθόδων για τον
υπολογισμό της σημαντικότητας των κριτηρίων που μπορεί να ληφθούν υπόψη κατά τη δι-
άρκεια της διαχείρισης της κινητικότητας των χρηστών. Οι εν λόγω μέθοδοι αναφέρονται
ως Trapezoidal Fuzzy Analytic Network Process (TF-ANP), Trapezoidal Fuzzy Adaptive
Analytic Network Process (TF-AANP) και Pentagonal Fuzzy Analytic Network Process
(PF-ANP) . Επίσης, προτείνεται ένα σύνολο αλγορίθμων που αφορούν την επιλογή του
καταλληλότερου δικτύου πρόσβασης για τον εκάστοτε χρήστη, οι οποίοι αναφέρονται ως
Trapezoidal Fuzzy TOPSIS (TFT), Trapezoidal Fuzzy TOPSIS with Adaptive Criteria Weights
(TFT-ACW) και Pentagonal Fuzzy TOPSIS (PFT). Ως αποτέλεσμα της εν λόγω έρευνας,
προτείνεται ένα νέο μοντέλο, το οποίο εξειδικεύεται στη διαχείριση της κινητικότητας των
χρηστών που αλληλοεπιδρούν με συστήματα 5G-VCC. Το μοντέλο αυτό, λαμβάνει υπόψη
το Signal to Noise plus Interference (SINR) καθώς και την ταχύτητα κίνησης του χρήστη,
ώστε να αξιολογήσει την ανάγκη πραγματοποίησης διαπομπής, ενώ στη συνέχεια εφαρμόζει
τον αλγόριθμο PFT για την επιλογή του καταλληλότερου δικτύου για τον χρήστη. Το μο-
ντέλο αξιολογείται ενδελεχώς μέσω προσομοιώσεων, μέσω των οποίων αποδεικνύεται ότι
xiii
1 Introduction 1
1.1 Delivery Models for Cloud Services . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Software as a Service (SaaS) based Delivery Models . . . . . . . . . 4
1.1.2 Platform as a Service (PaaS) based Delivery Models . . . . . . . . . 5
1.1.3 Infrastructure as a Service (IaaS) based Delivery Models . . . . . . . 5
1.1.4 Comparison of the Delivery Models . . . . . . . . . . . . . . . . . . 5
1.2 5G Vehicular Cloud Computing Systems (5G-VCC) . . . . . . . . . . . . . . 7
1.2.1 5G-VCC Architectures . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.1.1 Vehicular Cloud (VC) . . . . . . . . . . . . . . . . . . . . 7
1.2.1.2 Vehicles using Cloud (VuC) . . . . . . . . . . . . . . . . . 8
1.2.1.3 Vehicles using Fog (VuF) . . . . . . . . . . . . . . . . . . 8
1.2.1.4 Software Defined Vehicular Architectures (SDN-V) . . . . 9
1.2.1.5 Hybrid Vehicular Architectures (HVA) . . . . . . . . . . . 10
1.2.2 Communication Models for 5G-VCC Systems . . . . . . . . . . . . 11
1.2.2.1 Vehicle to Vehicle (V2V) . . . . . . . . . . . . . . . . . . 11
1.2.2.2 Vehicle to Infrastructure (V2I) . . . . . . . . . . . . . . . 12
1.2.2.3 Vehicle to Pedestrian (V2P) . . . . . . . . . . . . . . . . . 12
1.2.2.4 Hybrid Vehicular Communication (HVC) . . . . . . . . . . 12
1.3 Thesis Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 Thesis Novelty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3 Mobility Management 39
3.1 Related Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.1 Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.1.1 The Interval-Valued Trapezoidal Fuzzy Numbers (IVTFN) 39
3.1.1.2 The Interval-Valued Pentagonal Fuzzy Numbers (IVPFN) . 40
3.1.1.3 Creating Fuzzy Numbers: The Equalized Universe Method
(EUM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1.1.4 The Mamdani Pentagonal Fuzzy Inference System (MPFIS) 41
3.1.2 Estimating the Mobility Management Criteria Importance . . . . . . 43
3.1.2.1 The Analytic Network Process (ANP) . . . . . . . . . . . 43
3.1.3 The Trapezoidal Fuzzy Analytic Network Process (TF-ANP) . . . . . 45
3.1.3.1 The Trapezoidal Fuzzy Adaptive Analytic Network Process
(TF-AANP) . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.3.2 The Pentagonal Fuzzy Analytic Network Process (PF-ANP) 49
3.1.4 Network Selection Methods . . . . . . . . . . . . . . . . . . . . . . 52
3.1.4.1 The Trapezoidal Interval-Valued Fuzzy TOPSIS (TFT) . . 52
3.1.4.2 The Trapezoidal Fuzzy TOPSIS with Adaptive Criteria Weights
(TFT-ACW) . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.4.3 The Pentagonal Fuzzy TOPSIS (PFT) . . . . . . . . . . . . 57
Contents xvii
5 Conclusion 133
5.1 Directions for Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
Bibliography 141
2.15 The FLSA-CC Scheduler Evaluation in terms of the Packet Loss Ratio of the
Real Time Flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.16 The FLSA-CC Scheduler Evaluation in terms of the Throughput of the Real
Time Flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.17 The FLSA-CC Scheduler Evaluation in terms of the Fairness Index of the Real
Time Flows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.18 The FLSA-CC Scheduler Evaluation in terms of the Throughput and the Fair-
ness Index of the Best Effort Flows. . . . . . . . . . . . . . . . . . . . . . . 38
C1 Number of vehicles that start from each LTE network and their corresponding
velocities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
C2 Number of vehicles that start from each WiMAX network and their correspond-
ing velocities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
C3 Number of vehicles that start from each WAVE network and their corresponding
velocities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
C4 Vehicles count per service and SLA. . . . . . . . . . . . . . . . . . . . . . . 174
Chapter 1
Introduction
In a Fifth Generation Vehicular Cloud Computing (5G-VCC) system the operating principles
of vehicular networks, Cloud Computing (CC) [1] and Software Defined Networks (SDN) [2],
considered as the key enabling technologies for the 5G networks, are combined. In a typical
VCC system, vehicles are equipped with On-Board Units (OBUs) with computational, storage
and communication resources. Vehicles communicate with each other as well as with Road
Side Units (RSUs). Also, each RSU interacts with a Cloud infrastructure which offers a variety
of modern services with strict Quality of Service (QoS) requirements. Vehicles such as cars,
motorcycles, buses and trains provide a wide variety of cloud services to their passengers.
Each vehicle could serve many passengers with different services and various requirements.
Such services may include Driver Assistance (DA), Passenger Entertainment and Information
(PEnI), and Medical (MED) services. The DA services include Navigation Assistance (NAV)
[3] and Parking Assistance (PRK) [4] services. The PEnI services include Conversational
Video (CV) [5], Voice over IP (VoIP) [6], Buffered Streaming (BS) [7] and Web Browsing
(WB) [8] services. Moreover, the MED services include Live Healthcare Video (LHVideo) [9],
Medical Images (MedImages) [10], Health Monitoring (HMonitoring) [11] and Clinical Data
Transmission (CData) [12] services.
To support the communication needs of the vehicles, dense deployments of 5G networks
are implemented, called Ultra Dense Networks (UDNs) [13]. UDNs aim to support the high
data rates produced by the increased number of vehicular users. Heterogeneous network
access technologies, such as the 3GPP Long Term Evolution Advanced (LTE-A) [14], the
IEEE 802.16 Worldwide Interoperability for Microwave Access (WiMAX) [15] or the IEEE
802.11p Wireless Access for Vehicular Environment (WAVE) [16] RSUs are used to provide
network services to vehicles. In addition, a large number of small cells, such as Femtocells,
is deployed inside the network coverage area in order to increase the overall capacity of the
network [17][18]. Network operators use the Anything as a Service (ANYaaS) [19] delivery
2 Introduction
Service Layer
Cloud & Fog services
Network Layer
Wireless communication equipment
including Road Side Units (RSUs)
and Base Stations (BSs)
Device Layer
On-Board Units (OBUs), Sensors,
Navigation devices, Smart phones,
Tablets etc.
model in order to deploy and orchestrate network management algorithms and services using
the Cloud infrastructure. Specifically, the Mobility Management as a Service (MMaaS) [20]
[21] subcategory of ANYaaS, could be applied to accomplish mobility management operations
by the Cloud infrastructure. The MMaaS model allows for each vehicular user the creation of a
Mobility Instance (MI) at the Cloud, offering mobility management services and manipulating
the user’s mobility. In this way, the user equipment resources will experience a decreased
workload and will consume less energy during the mobility management. The 5G architecture
could be improved by applying the operating principles of Fog or Mobile Edge Computing
(MEC) [22–26]. Specifically, WAVE RSUs, as well as LTE and WiMAX Base Stations
(BSs) are equipped with additional computational and storage resources and thus they are
called as micro-datacenter RSUs (md-RSUs) and micro-datacenter BSs (md-BSs), respectively.
Vehicular services are provided directly from the access network. Thus, the durability and the
response latency of the services are improved, compared with the traditional centralized cloud
server approach.
5G-VCC systems could be described considering the three-layer stack [27][28] presented
in Figure 1.1, where the Service, the Network and the Device layers are defined. The Device
layer contains the end-user equipment such as On-Board Units (OBUs), smart phones, tablets,
navigation devices and sensors. The Network layer contains the communication equipment,
such as Macrocell/Femtocell Base Stations (BSs) or Road-Side Units (RSUs), as well as in-
vehicle network devices. The Service layer contains Cloud or Fog infrastructures that offer
vehicular services.
The vehicular users should always obtain connectivity to the most appropriate network
access technology, according to the requirements of their services. In general, the mobility
management process consists of three subprocesses, namely the VHO initiation, the network
selection and the VHO execution. The VHO initiation decides when a handover must be
performed. Subsequently, the network selection deals with the selection of the most appropriate
network for the vehicle to handover. Finally, during the VHO execution two general techniques
1.1 Delivery Models for Cloud Services 3
are considered. The first one is called Soft Handover or "Connect before Break". It defines that
the OBU is firstly connected to the new network and then it is disconnected from its previous
network. The second technique is called Hard Handover or "Break before Connect". In this
case, the OBU is firstly disconnected from its previous network and then it is connected to the
next network.
Delivery Models
Information as a Computing as a
Service (INaaS) Service (COaaS)
Sensing as a Service
(SEaaS)
Warning as a
Service (WaaS)
• Data as a Service (DaaS): In a 5G network architecture, data sourcing and data manipu-
lation could be performed in the cloud infrastructure using the appropriate applications.
DaaS allows users to retrieve data from the cloud, in order to avoid the permanent storage
of big data sets on their devices [30].
• Information as a Service (INaaS): Users often need information related to events, points
of interest (POIs) or emergency situations. The INaaS delivery model provides that
information to the users. Indicatively, in [32] users share information about their location
with other users as well as with a cloud infrastructure. Then, if a user needs information
about a specific location, he/she interacts with the cloud and subscribes to the INaaS in
order to receive it.
• Warning as a Service (WaaS): The cloud infrastructure collects information about emer-
gency situations or disasters. Subsequently, warning messages are transmitted to users
depending on their locations [35].
1.1 Delivery Models for Cloud Services 5
• Discovery as a Service (DISCaaS): This delivery model allows users to discover cloud
recourses with specific characteristics [38].
• Network as a Service (NaaS): In NaaS the users with internet access can offer this facility
to the users that do not have any access [39].
• Hardware as a Service (HaaS): In HaaS the user devices or the cloud with plenty
computational resources can offer these to the users that need more resources [40]. The
HaaS model contains two subcategories, the Computing as a Service (COaaS) [36]
and the Storage as a Service (STaaS) [41]. COaaS provides computational resources,
such as Central Processing Unit (CPU) cores to users, in order to develop and run their
applications. STaaS provides storage space to users to deploy their applications.
cost effective delivery model, due to the fact that the user leases only the software that he
uses and not the resources that host it. In addition, the SaaS services are easy in use and they
can be rapidly provided on-demand, because they are already deployed on the cloud by their
provider. Also, the user does not have to worry about the services management, as this is the
provider’s responsibility. On the other hand, the user has no control neither over the application
implementation and parameterization, nor on the data processing functionalities. Also, the user
has limited control over the application deployment and upgrade processes, while integration
with other user’s software or systems is difficult, considering the fact that the integration must
be presupported by the service provider.
PaaS is less cost effective in comparison to SaaS, while at the same time it is more cost
effective than IaaS, as the user is leasing only the software platform (e.g. an Operating System
installation) and not the entire infrastructure that hosts the platform. Unlike SaaS, a user
can deploy his own applications to run on the aforementioned platform and, thus, he has full
control over the software that he runs. Therefore, the user has also full control over the rights
of the users accessing his software as well as over the data processing functionalities of his
applications. Furthermore, the integration between user applications and external systems
can be done at the application level, since the user has full control over the source code of
his applications. Another PaaS advantage is the minimal Virtual Machine (VM) management
that is required, since such processes are handled by the infrastructure provider. However, the
lack of control over the VM could create security risks in terms of application data privacy.
Also, the provided platform is usually a shared platform and other users could be running their
applications at the same time over it, creating even more privacy risks as well as overloading
the underlying infrastructure with additional workflows.
In IaaS, the user has full control over his VM, while he can deploy and run anything he
wants inside it. Furthermore, the user has full control of data processing functionalities inside
the entire VM and not only inside his applications as it happens in PaaS. As a consequence,
IaaS simplifies ever more the integration with other systems compared with PaaS. In addition,
it is considered as the most privacy aware cloud service due to the fact that the user can deploy,
run and control his own virtual infrastructure with full control to his VMs. However, IaaS is
more expensive than SaaS and PaaS, since the user is leasing physical resources, such as CPU
cores, RAM and storage space. Also, unlike SaaS or PaaS, in IaaS the user is responsible for
the entire VM manipulation processes, which may lead to additional tasks for him. Table 1
summarizes the main delivery models’ advantages and disadvantages.
1.2 5G Vehicular Cloud Computing Systems (5G-VCC) 7
Communication
Architectures
Models
According to the VC architecture [38, 41–46], the cloud is consisted of vehicles with computing
resources (Figure 1.4). Vehicles directly interact with each other, while data forwarding is
supported by proprietary protocols such as the 802.11s [47].
Vehicles’ resources [48] remain idle for a long period of time in vehicles parked for many
hours each day, or even for many days in some cases, as well as in vehicles that remain stuck
due to congested traffic roads. The resources of such vehicles could be used for constructing
a cloud infrastructure [49][50]. In such cases, additional resource manipulation factors are
arising, since the constructed cloud is based on an ever-changing architecture where vehicular
VMs (VVMs) [51] are continuously added or removed. Thus, in order to assure service
continuity, multiple instances of each service should be distributed simultaneously in several
VVMs [52]. Furthermore, the authorization of each vehicle’s owner is required [53], for the
vehicles’ resources to be provided to the VC architecture.
8 Introduction
Service Layer
Service Layer
Network Layer
Device Layer
Cloud
RSU/BS RSU/BS
Vehicular Environment
According to the VuC architecture [27, 28, 35, 54–60], the vehicles interact with a Cloud
infrastructure through Macrocell/Femtocell BSs or RSUs, as presented in Figure 1.5. Compared
to the VC architecture, in a VuC architecture the vehicles are considered as the end users.
The VuF architecture [61] presented in Figure 1.6, could be considered as an evolution of the
aforementioned VuC architecture, where the operating principles of Fog computing [22–26]
are applied. Specifically, the VuF architecture defines that the BSs or RSUs are equipped with
additional computational and storage resources, while they are also referred as micro-datacenter
BSs (md-BSs) and micro-datacenter RSUs (md-RSUs), respectively. In this case, the vehicles
1.2 5G Vehicular Cloud Computing Systems (5G-VCC) 9
Service Layer
Network Layer
Fog Device Layer
micro-datacenter micro-datacenter
RSU/BS RSU/BS
micro-datacenter
RSU/BS micro-datacenter
RSU/BS
Vehicular Environment
interact with computing and storage resources characterized by proximity, low latency and high
bandwidth.
Each VCC architecture could be enhanced with advanced manipulation functionalities using
the Software Defined Networks (SDN) technology [62]. More specifically, the SDN technology
simplifies the network management procedures, while at the same time it reduces the network
equipment cost. Its operating principles include:
• The network control plane and the data plane are decoupled.
• The interaction with network equipment from different vendors is achieved by proving
Open Application Programming Interfaces (Open APIs).
SDN control could be implemented either at the Cloud or the Fog infrastructure. The
SDN components could be virtualized [63], considering the operating principles of Network
Functions Virtualization (NFV) [64][65], in order to reduce the required SDN hardware as
well as to improve the overall system sustainability. In particular, concerning the Cloud-
based SDN control, in [66], [67] and [68] the SDN controller is implemented at the Cloud
infrastructure, by one or more Virtual Machines (VM). On the other hand, in [69] and [70]
a Fog infrastructure implementing the SDN control is described. Specifically, an md-RSU
contains a computing device and an OpenFlow (OF) Switch. The computing device include
10 Introduction
The previous main models are combined in hybrid 5G-VCC architectures. Indicatively, in [73]
a hybrid 5G-VCC architecture combining both the VuC and VuF models, is discussed. Services
are offered by both Cloud and Fog infrastructures, while the Cognitive Radio Networks (CRN)
[74] communication principles are adopted. More specifically, the traffic flows generated by
the Clouds are considered as primary flows, while the ones generated by the Fog are considered
as secondary flows. Also, vehicles are separated into two groups, the primary vehicles which
use the Cloud services and the secondary vehicles which use the Fog services. Primary
vehicles have higher priority on the spectrum usage, while secondary vehicles connecting to
the Fog use the remaining time and frequency holes. Additionally, in [75] and [31] an HVA
architecture (Figure 1.7) is introduced, combining the design characteristics of VC, VuC and
VuF architectures. Its main advantage is that the entire computation and storage resources
are virtualized. The complete virtualization provides enhanced system flexibility as well as
resources sustainability. Also, in [76] a HVA architecture which combines both VC and VuC
architectures is described. More specifically, a VC provides vehicle cooperation functionalities
such as Internet access via multihop traffic routing (mesh networking). Also, the VC collects
information about its vehicles, such as fuel consumption, GPS coordinates, CO2 emissions
and traffic data. The VC evaluates this information and extracts traffic related conclusions
about its area. Subsequently, the conclusions are sent to the central cloud. Thus, vehicles that
does not belong to the VC can obtain the traffic related information concerning the VC area.
Furthermore, the extended version of this architecture includes multiple VCs could. Each VC
evaluates traffic related information of its geographic area. The central cloud collects the traffic
related results from all VCs and extracts conclusions for the whole area.
1.2 5G Vehicular Cloud Computing Systems (5G-VCC) 11
Service Layer
Network Layer
Device Layer
Cloud
Fog
micro-datacenter
micro-datacenter RSU/BS
RSU/BS
micro-datacenter
micro-datacenter
RSU/BS
RSU/BS
Vehicular Cloud
Figure 1.7 A hybrid architecture which combines the design characteristics of both VC, VuC
and VuF architectures.
V2V communications [77–83] comprises a wireless network where vehicles exchange infor-
mation each other. Such information could include vehicle location, speed, direction, braking,
stability loss as well as any other information including traffic or multimedia. V2V could
applied in a mesh network, meaning that every vehicle could receive and forward messages
from/to other vehicles. The first V2V implementations warned the driver about traffic condi-
tions but not took control of the vehicle, while later implementations improve the braking or
steering capabilities of the vehicles.
12 Introduction
Cloud
RSU/BS RSU/BS
Pedestrian
Vehicular
Environment
Pedestrian
Pedestrian
V2I communication [84–90] is performed through the RSUs, while the information gathered
by the infrastructure could include traffic or road conditions and Points of Interest (PoI).
Indicatively, the vehicles velocities and accelerations as well as the inter-vehicle distances
could be suggested by the infrastructure considering the traffic conditions and optimizing the
traffic manipulation, the fuel consumption and the driver safety.
In a 5G-VCC architecture, multiple communication models could also be used together [98–
104] (Figure 1.8).
The second chapter, which deals with the management of network resources in 5G-VCC
systems, firstly describes the available Medium Access Control schemes (MAC schemes).
Subsequently, existing resource scheduling algorithms implemented at the MAC layer of 5G-
VCC systems are described, while two novel solutions are proposed to optimize the resource
allocation process.
In the third chapter, the requirements of the 5G-VCC systems for the management of
mobility of vehicular users are described. Three methods for calculating the significance of the
criteria affecting the mobility management are proposed. Furthermore, three network selection
algorithms are proposed. Subsequently, the proposed algorithms are evaluated.
The fourth chapter describes a proposed scheme for performing mobility management in 5G
wireless networks. The scheme takes into consideration the Signal to Noise plus Interference
(SINR) while at the same time it performs optimized supervision of the user’s velocity, in order
to evaluate the necessity of performing a handover. Subsequently, it selects the most appropriate
network for the user. Experimental results showed that the proposed model outperforms the
existing solutions described in the research literature.
Finally, in the fifth chapter a brief review of the thesis is made. The conclusions extracted
from the performed research are mentioned, while guidance for future work is given.
– Study, classification and evaluation of the most used medium access mechanisms
for 5G wireless networks and specifically for 5G-VCC systems.
– Study, classification and evaluation of the most used cloud delivery models that can
be applied to 5G network architectures.
– Three novel methods for calculating the weights of mobility management criteria
are proposed. Experimental results showed that the proposed methods optimize the
calculation of the significance of each mobility management criterion.
– Three novel network selection algorithms for 5G wireless network, including 5G-
VCC systems, are proposed. The experimental results showed that these algorithms
select the optimal network considering the requirements of both users and network
operators.
– An integrated mobility management scheme for 5G wireless networks, including
5G-VCC systems, is proposed. Extensive experimental results demonstrated that it
optimizes the mobility management process by minimizing the number of required
handovers, while at the same time it ensures that the user is connected to the optimal
access network considering both his requirements as well as network operators
constraints.
Chapter 2
ACFM HER-MAC
Hybrid CBRC-MAC OBV
CS-TDMA R-MAC
is reduced to improve channel utilization when the vehicle density becomes low, or increased
when the vehicle density becomes high to ensure that each vehicle can access the medium.
Time slots are divided in two sets, namely the Left and the Right set. A slot management
mechanism based on a binary tree model is used. The vehicles on the left sub-tree can compete
for the Left time slots, while the vehicles on the right sub-tree can compete for the Right time
slots. When a vehicle receives slot allocation information from its neighbors, it discovers which
slots are in use. Thus, the remaining slots are available to compete for.
The Cluster Based TDMA (CBT) [106] scheme provides a mechanism for intra-cluster
and inter-cluster communication to minimize the packet collisions that could occur when two
clusters are moving close to each other. In each cluster, the vehicles are synchronized in time
using their GPS devices, while one vehicle is elected as the Cluster Head (CH). A TDMA
related technique is used where each frame consists of N time slots. The CH maintains a
Slot Allocation Map (SAM) allocating time slots to vehicles. Moreover, the CH periodically
broadcasts its SAM and a beacon frame to its cluster’s vehicles. The cluster remains in the
intra-cluster communication state if the beacon frames from the CHs of the other clusters are
not received. However, if a beacon frame comes from an external CH, the two neighboring
clusters’ CHs exchange their SAMs in order to prevent inter-cluster interference. The CH that
successfully sends first its SAM is considered as the Winner, while the other is considered as
the Loser. The Loser must reschedule its own SAM.
Another TDMA-based MAC scheme is the Cross-layer Extended Sliding Frame Reservation
Aloha (CESFRA) [107]. In CESFRA safety information is disseminated up to the third hop
neighboring vehicles without any routing scheme. This scheme divides each frame into N
time slots. All the vehicles are considered to be synchronized in time using their Geographical
Positioning System (GPS) devices. When a vehicle has packets to transmit, it senses the
medium in order to find an idle time slot. Once an idle time slot is found, the vehicle starts
transmitting its packets. The time slot is also reserved by the vehicle in the subsequent frames
in order to transmit the remaining packets.
The Collision Free Reservation MAC (CFR-MAC) [108] is another mechanism that provides
TDMA based medium access. It considers the vehicles’ traffic flows as well as their velocities.
As in the ATSA, the time slots are divided into two sets, the Left and the Right set. The Left set
is assigned to vehicles that are moving to the one direction, while the Right set is assigned to
vehicles moving to the opposite. In general, when multiple vehicles are moving on the same
street with different velocities the interference levels on the wireless environment are constantly
changing leading on unpredictable changes in the medium quality. CFR-MAC addresses this
problem by dividing each slot’s set into three subsets. Each subset is associated to a specific
2.1 Medium Access Control (MAC) Schemes for 5G-VCC systems 17
velocity, namely the High, the Medium and the Low velocity. In this way the interference levels
inside each subset become less variable and the medium quality more resistant.
The Vehicular MAC (VeMAC) [109] supports broadcast services. Similar to the CBT and
CFR-MAC schemes, the two vehicles’ moving directions are considered, namely the Left and
the Right direction. A set of time slots is assigned to vehicles that move in the Left direction
and the Right direction, respectively. Using these time slots, the vehicles of each direction
communicate with each other, while the vehicles’ time synchronization is performed using
their GPS devices. Although VeMAC supports reliable transmission, it is not fully applicable
in vehicular networks with parallel transmissions. In [110] an enhanced version of VeMAC
scheme, the e-VeMAC, is proposed. The e-VeMAC scheme is based on the insight of the
one-hop neighboring vehicles in order to improve the performance of the VeMAC scheme
when parallel transmission occurs.
The Enhanced TDMA Cluster-based MAC (ETCM) [111], the Prediction-based TDMA
MAC (PTMAC) [112] and the Unified TDMA-Based Scheduling Protocol (UTSP) [113]
schemes also implement TDMA based multiple access. The ETCM scheme defines that the
vehicles are organized into clusters, while a vehicle of each cluster is defined as the CH.
Subsequently, the CH applies a TDMA based method to assign time slots to the cluster’s
vehicles.
The main operating principle of PTMAC is the packet collision prediction. PTMAC
consists of three parts, namely the collision prediction part, the collision detection part and the
collision elimination part. According to the collision prediction part, data traffic and vehicles
mobility information is used in order to predict potential future collisions. Furthermore, the
collision detection uses time slot information to detect collisions that occurred in cases where
two vehicles transmit data using the same time slot. Finally, the collision elimination part
reschedules the slots considering the information obtained from both the collision prediction
and the collision detection parts, in order to eliminate the packet collisions.
The UTSP scheme implements a centralized resource allocation mechanism for V2I commu-
nication. Initially, the RSU collects information about the channel state, the vehicles’ velocities
and the priorities of the vehicles’ services. Then, it uses a weighted function to compute a
score for each vehicle. Finally, the RSU assigns TDMA time slots to each vehicle according
to its score, where the amount of time slots assigned to each vehicle is proportional to the
corresponding vehicle’s score.
Similar to the aforementioned MAC schemes, other TDMA-based schemes are the Cooper-
ative ADHOC MAC (CAH-MAC) [114], the Improved Generalized Prime Sequence Based
MAC (IGPS-MAC) [115], the Self-organizing Time Division Multiple Access (STDMA) [116],
and the TDMA Cluster-based MAC (TC-MAC) [117].
18 Resource Allocation - Scheduling
predefined threshold, the RTS/CTS mechanism is applied in order to avoid the hidden terminal
and the exposed terminal problems, otherwise the position information is utilized in order to
decide whether to start the transmission.
warning messages transmission (e.g. in case of an accident) while the TDMA sub-segment is
used for non-safety data transmission.
It should also be noted that in some hybrid mechanisms, the TDMA or CSMA/CA are com-
bined with the Space Division Multiple Access (SDMA) that considers the geographic position
of each vehicle. For example, the CSMA-based Self-Organizing TDMA (CS-TDMA) [127]
combines TDMA, CSMA/CA and SDMA to support both safety and non-safety applications. A
CCH channel is used for the data transmissions of the safety applications, while a SCH channel
is used for the non-safety applications. The ratio between the CCH and SCH lengths is adjusted
taking into consideration the vehicle density in each location. If the vehicles density is high,
the CCH length is increased and the SCH length is decreased in order to guarantee a maximum
delay for safety applications. On the contrary, if the vehicles density is low, the CCH length is
decreased and the SCH length is increased in order to improve the throughput for non-safety
applications.
Finally, in the OFDMA-based MAC scheme for VANETs (OBV) [128] the CSMA/CA is
combined with the OFDMA mechanism. During a resource negotiation phase, vehicles allocate
resources using CSMA/CA. Thereafter, the data transmissions are performed using OFDMA,
while at the same time the vehicles are synchronized using their GPS receivers in order to
guarantee the orthogonality between the OFDMA subcarriers.
Another important factor is the control type applied in each scheme. Some schemes apply
distributed control, while other schemes apply centralized control. Although the distributed con-
trol distributes the resource manipulation workload, the centralized control could be considered
as more appropriate in some cases, since it simplifies the manipulation of the communication
resources. In 5G-VCC architectures where a centralized Software Defined Network (SDN)
[129] controller supervises the manipulation of the entire system, centralized MAC schemes
could be easily configured considering the complete view of the system’s resources. Distributed
schemes could also be configured in 5G-VCC systems considering the operating principles of
Fog computing [130].
Cognitive Radio Networking (CRN) [74] is another factor that is proposed in 5G-VCC
systems to organize the vehicles into two groups, namely the Primary and the Secondary
vehicles. Primary vehicles obtain immediate access to network resources as determined in
their Service Level Agreements (SLAs) [131]. However, sometimes Primary vehicles do not
need the entire network resources provided according to their SLAs. In such cases, Secondary
vehicles could use the free network resources, which are also called White Spaces. Furthermore,
whenever a Primary vehicle requires the resources defined in its SLA, it immediately reserves
them, while the Secondary vehicles should obtain access to other free network resources.
Considering the CRN operating principles, some MAC schemes (e.g. MQOG) take advantage
of the free white spaces into the available spectrum in order to serve more vehicles. Finally, the
communication type (V2V or V2I) that each scheme supports is considered. Table 4.2 presents
the characteristics of the discussed MAC schemes considering the aforementioned factors.
Positioning Centralized
Multiple Cluster CRN
MAC Scheme System /Distributed Communication
Access based Support
Required control
FDMA,
ACFM [123] No No Centralized No V2I
TDMA
ATSA [105] TDMA No No Distributed No V2V
CAH-MAC [114] TDMA No Yes Distributed No V2V
CA-MAC [118] CSMA/CA No No Distributed No V2V, V2I
FDMA,
CBRC-MAC [124] Yes No Centralized No V2I
TDMA
CBT [106] TDMA Yes Yes Centralized No V2V
CESFRA [107] TDMA No Yes Distributed No V2V
CFR-MAC [108] TDMA No Yes Distributed No V2V
CSMA/CA,
CS-TDMA [127] SDMA, No Yes Distributed No V2V
TDMA
ETCM [111] TDMA Yes Yes Centralized No V2V
CSMA/CA,
HER-MAC [125] No Yes Distributed No V2V
TDMA
IGPS-MAC [115] TDMA Yes Yes Centralized No V2V
MMAC-CL [120] CSMA/CA No Optional Distributed No V2V, V2I
MQOG [121] CSMA/CA No No Distributed Yes V2V, V2I
CSMA/CA,
OBV [128] No Yes Distributed No V2V, V2I
OFDMA
PTMAC [112] TDMA No Yes Distributed No V2V
QL-MAC [122] CSMA/CA No Yes Distributed No V2V
CSMA/CA,
R-MAC [126] Yes Yes Centralized No V2V, V2I
TDMA
STDMA [116] TDMA No Yes Distributed No V2V
TC-MAC [117] TDMA Yes Yes Centralized No V2V
UTSP [113] TDMA No Yes Centralized No V2I
VeMAC [109] TDMA No Yes Distributed No V2V
e-VeMAC [110] TDMA No Yes Distributed No V2V
allocates resources using the metric calculated by formula 2.1, where dki (t) represents the
available throughput in the kth RB of the ith Transmission Time Interval (TTI). The PF aims at
fair distribution of network resources to users. Its metric is estimated using formula 2.2, where
R̄i (t − 1) denotes the past average throughput. Finally, the TTA metric is calculated using
formula 2.3 where d i (t) represents the available throughput in the ith TTI. Simulation results in
[132], [133] showed that the MT achieves higher throughput, the TTA has better performance
in terms of the fairness index while the PF provides satisfactory throughput and fairness.
mMT i
i,k = dk (t) (2.1)
dki (t)
mPF
i,k = (2.2)
R̄i (t − 1)
dki (t)
mTi,kTA = (2.3)
d i (t)
The Blind Equal Throughput (BET) non-QoS aware scheduler described in [134] distributes
equally the throughput to the users using formula 2.4. The authors of [135] compare the MT,
2.2 Overview of Downlink Packet Schedulers 23
the PF and the BET schedulers using TCP traffic in a vehicular environment. Numerical results
showed that the MT and the PF schedulers achieve similar throughputs, while their performance
is better than the one achieved by the BET algorithm.
1
mBET
i,k = (2.4)
R̄i (t − 1)
Moreover, the authors of [136] evaluate the performance of video streaming using the MT,
the PF and the Round Robin (RR) which allocates resources sequentially to users. Simulation
results showed that the PF algorithm achieves a higher Mean Opinion Score (MOS) and
throughput for a large number of users, while the MT performs better in terms of the Freezing
Delay Ratio (FDR).
mM−LW
i,k
DF
= ai · DHOL,i · mPF
i,k (2.5)
24 Resource Allocation - Scheduling
EXP/PF ai · DHOL,i − X
mi,k = exp( √ ) · mPF
i,k (2.6)
1− X
logδi
αi = − (2.7)
τi
1 Nrt
X= · ∑ ai · DHOL,i (2.8)
Nrt i=1
The LOG RULE and the EXP RULE schedulers presented in [141] take into account
the head of line packet delay and the channel quality reported by UEs, to support delay
sensitive flows. The LOG RULE metric is estimated using formula 2.9, where Γik is the spectral
efficiency for the ith user on the kth subchannel. The EXP RULE metric is estimated using
formula 2.10, where bi and c are configurable parameters. Simulation results showed that
both LOG RULE and EXP RULE could guarantee delay constraints by configuring scheduler
parameters according to the users’ target delays.
mLOGRULE
i,k = bi · log(c + ai · DHOL,i ) · Γik (2.9)
ai · DHOL,i
mEXPRULE
i,k = bi · exp( q ) · Γik (2.10)
c + (1/Nrt ) ∑ j DHOL, j
The FLS scheduler described in [142] implements a two level QoS aware strategy. The
upper level uses formula 2.11 to estimate the ui (k) quota of data that the ith real time flow must
transmit at the kth frame to satisfy its QoS constraints. In this formula, qi (k) represents the
queue length in the kth frame, Mi the number of coefficients used and ci (n) the nth coefficient
value. Coefficients are used in order to guarantee the required delay constraints for real time
flows. The number Mi of coefficients is estimated using formula 2.12, where τi represents the
target delay. Additionally, the coefficient value ci (n) is determined by formula 2.13. The lower
level uses the PF metric to allocate network resources to real time flows for transmitting their
quota of data, whereas the remaining resources are allocated to best effort flows. Simulation
results showed that the proposed model improves the performance of real time flows compared
to the EXP RULE and LOG RULE schedulers.
Mi
ui (k) = qi (k) + ∑ [qi · (k − n + 1)
n=2
(2.11)
− qi · (k − n + 2) − ui · (k − n + 1)] · ci (n)
τi = (Mi + 1) · T f (2.12)
2.3 The Proposed FLS Advanced (FLSA) Scheduler 25
Real-time queues
ui(t)
YES
0 , when n = 0
ci (n) = 1 , when n = 1 (2.13)
ci (n − 1)/2 , when n ≥ 2
The work described in [133] classifies the LTE downlink schedulers into five categories
namely the channel-unaware, the channel-aware/QOS-unaware, the channel-aware/QoS-aware,
the semi-persistent for VoIP support and finally the energy aware scheduling algorithms.
Simulations that performed for the channel-aware/QoS-aware category showed that the FLS
scheduler outperforms the M-LWDF, the EXP/PF, the EXP RULE and the LOG RULE in terms
of packet losses and packet delays as the number of video flows increases.
Table 2.2 The Parameters considered in each Scheduler in comparison with the ones considered
by the FLSA.
Figure 2.3 The Topology Simulated for the Evaluation of the FLSA Scheduler.
In the simulations, an LTE network topology composed of 19 cells is considered. Each cell
contains 1 eNB with 43 dBm transmission power and 10 MHz downlink bandwidth providing
50 RBs available for allocation per TTI. The system sub-frequencies have been distributed to
cells using a reuse factor equal to 2 as presented in Figure 2.3, to avoid interference between
neighbor cells. A full bandwidth periodic Channel Quality Indication (CQI) reporting scheme
is applied and each user equipment (UE) reports its downlink Signal to Interference plus Noise
Ratio (SINR) to eNB, in every TTI. The eNB quantizes the reported SINR value and calculates
the CQI as described in [145], Then, it uses the CQI to guarantee a maximum block error
rate (BLER) less than 10% regardless of the scheduling strategy applied. A number of users,
between the range [10, 40], is moving inside the borders of each eNB according to the random
way-point mobility model. Each user receives two real time flows, including an H264 video
with bitrate equal to 440 kbps and a voice over IP (VoIP) using the G.729 codec. Furthermore,
one best effort flow is added as background traffic. Table 2.6 summarizes the simulation
parameters.
In general, QoS aware schedulers increase the packet loss ratio (PLR) as they try to
maintain the required target delay. This strategy is based on the assumption that real time
services such as VoIP and video have no advantages of receiving expired packets. Thus, since
the delay constraint is satisfied, the algorithms are evaluated in terms of PLR, so as to have a
28 Resource Allocation - Scheduling
Table 2.3 The Simulation Parameters for the Evaluation of the FLSA Scheduler.
Parameter Value
Simulation time 100 seconds
Downlink bandwidth 10 MHz
Modulation QPSK, QAM-16 and QAM-64
Number of cells 19
Cell radius 0.5 km
Number of users up to 40 users per cell
Users mobility Random way point
Real time traffics:
H264 video at 440 kbps,
Traffic models VoIP using G.729 codec
Figure 2.4 Evaluation of the FLSA Scheduler in terms of VoIP Packet Loss Ratio using Different
Target Delays.
comprehensive view about the performance improvements. Figures 2.4 and 2.5 illustrate the
impact of the target delay parameter in the PLR for VoIP and video flows, respectively, for the
case of having 20 users per cell. While the target delay increases from 50ms to 150ms, the PLR
decreases. Additionally, it may be observed that FLSA compared with FLS exhibits a lower
PLR by up to 7%.
In Figures 2.6 and 2.7, the PLR for the VOIP and video flows is presented as the number of
cell users varies from 10 to 40. The considered target delays are set to 100ms and 150ms for
VoIP and video flows respectively, as determined by the LTE QoS class specifications for these
service types [146]. As shown, FLSA results in a lower PLR than the rest of the algorithms.
FLSA shows a marginal decrease of its PLR compared to FLS for VoIP flows. Also, its PLR is
10% lower than that of FLS for video flows.
The analysis of the throughput provides an important insight on the performance of the
FLSA in comparison with the other schedulers. As presented in Figures 2.8 and 2.9 the FLSA
outperforms the rest of the schedulers, independently of the number of users for VoIP and video
2.3 The Proposed FLS Advanced (FLSA) Scheduler 29
Figure 2.5 Evaluation of the FLSA Scheduler in terms of Video Packet Loss Ratio using
Different Target Delays.
Figure 2.6 Evaluation of the FLSA Scheduler in terms of VoIP Packet Loss Ratio.
Figure 2.7 Evaluation of the FLSA Scheduler in terms of Video Packet Loss Ratio.
30 Resource Allocation - Scheduling
flows, respectively. This is expected due to the recalculation of formula 2.11 in each TTI by the
upper level of FLSA as well as the inclusion of the lower level in the scheduler. The FLSA
achieves higher throughputs than the rest of the algorithms providing rates of up to 330kbps for
VoIP and up to 4.7 Mbps for video services.
The proposed scheduler is also evaluated in terms of the Jain fairness index, which is
estimated using formula 2.14 where n is the number of the service flows and xi is the throughput
of the ith flow. n
(∑i=1 xi )2
Jain_Fairness = (2.14)
n · ∑ni=1 xi2
Flows with the same service constraints must receive similar QoS to avoid the situation of
having a set of satisfied users against a set of dissatisfied ones of the same service type. The
maximum value of fairness is 1 while the more a scheduler accomplishes a value close to 1, the
more the resource allocation is fair. As presented in Figure 2.10 the fairness for VoIP flows is
2.3 The Proposed FLS Advanced (FLSA) Scheduler 31
Figure 2.10 Evaluation of the FLSA Scheduler in terms of VoIP Fairness Index.
Figure 2.11 Evaluation of the FLSA Scheduler in terms of Video Fairness Index.
very close to 1. Additionally, Figure 2.11 demonstrates that the FLSA scheduler improves the
fairness for the video flows compared with the rest of the schedulers.
32 Resource Allocation - Scheduling
In each TTI, the middle level uses a cross carrier version of the MLWDF scheduler, called
MLWDF-CC, to allocate RBs to real time flows for transmitting their ui (x) quota of data which
have been calculated by the upper level. This scheduler extends the PF-CC metric [147] [148]
mPF−CC
i,k given in formula (2.16). The MLWDF-CC metric is evaluated by formula (2.17)
where the αi value is determined by formula (2.7). The use of the cross carrier QoS aware
MLWDF-CC scheduler achieves an improved resource distribution among the real time flows
in comparison with the FLS scheduler which at the second level uses the non-cross carrier
principle as well as the non-QoS aware PF algorithm.
The third level has been added to allocate the remaining RBs of each TTI to best effort
flows using formula (2.16) of the PF-CC algorithm.
dki (t)
mPF−CC
i,k = (2.16)
∑Nj=1 R̄i, j (t − 1)
2.4 The Proposed FLS Advanced - Cross Carrier (FLSA-CC) Scheduler 33
mMLW
i,k
DF−CC PF−CC
= ai · DHOL,i · mi,k (2.17)
The 3GPP urban channel model [152] [153] is used. Since the channel between DeNB or
RN and UE encounters non-line-of-sight (NLOS) transmission, its propagation loss is estimated
using formula (2.18) and (2.19), for DeNB-UE and RN-UE communication respectively, where
d represents the distance among the nodes and fc the carrier frequency. Also, since the channel
between a DeNB and an RN encounters line-of-sight (LOS) transmission, its propagation loss
is estimated using formula (2.20).
NLOS
PLDeNB→UE = 37.6 · log10 (d) + 58.94 + 21 · log10 ( fc ) (2.18)
NLOS
PLRN→UE = 36.7 · log10 (d) + 22.7 + 26 · log10 ( fc ) (2.19)
LOS
PLDeNB→RN = 22 · log10 (d) + 28 + 20 · log10 ( fc ) (2.20)
The Cloud contains a set of virtual machines (VMs) and implements the functionalities
of the LTE Evolved Packet Core (EPC). Additional VMs with user applications are created.
Specifically, one VM is created for each UE running three applications namely one VoIP, one
video and one best effort.
Flow forwarding as well as resource scheduling in each DeNB and RN are performed using
a centralized global controller placed into the Service Gateway (SGW) having a wide view of
the entire system. The simulated topology is presented in Figure 2.13.
Figure 2.13 The Topology Simulated for the Evaluation of the FLSA-CC Scheduler.
the reported SINR value and calculates the CQI as described in [145]. Then, it uses the CQI to
guarantee a maximum BLER less than 10% regardless of the scheduling strategy applied.
Table 2.5 The Parameters considered in each Scheduler in comparison with the ones considered
by the FLSA-CC.
A number of users move inside the borders of each RN according to the random way-point
mobility model. Each user receives two real time flows, an H264 video with bitrate equal to
440 kbps and a Voice over IP (VoIP) using the G.729 codec. Furthermore, one best effort flow
is added as background traffic. Table 2.6 summarizes the simulation parameters.
Due to the fact that the simulation environment includes an LTE topology with RNs, a two hop
target delay for real time flows τi = τi,DeNB→RN + τi,RN→UE is considered, where τi,DeNB→RN
36 Resource Allocation - Scheduling
Table 2.6 The Simulation Parameters for the Evaluation of the Proposed Algorithm.
Parameter Value
Simulation time 100 seconds
Downlink bandwidth 2*20 = 40 MHz
Modulation QPSK, QAM-16 and QAM-64
DeNBs number / radius 7 / 1 km
Relay nodes number / radius 4 per DeNB / 100 m
Number of users up to 100 users per relay node
Users mobility Random way ( point
H264 video at 440 kbps
Real time:
Traffic models VoIP using G.729 codec
Best effort: Web
Figure 2.14 The FLSA-CC Scheduler Evaluation in terms of the Packet Loss Ratio of the Real
Time Flows using Different Target Delays.
represents the target delay between a DeNB and an RN, and τi,RN→UE represents the target
delay between an RN and a UE. We consider τi,DeNB→RN = τi,RN→UE . In general, QoS aware
schedulers increase the packet loss ratio (PLR) to guarantee the required τi . This strategy is
based on the assumption that real time services such as VoIP and Video can not use expired
packets. Thus, since the delay constraint is satisfied, the algorithms are evaluated in terms of
PLR, so as to have a comprehensive view about the performance improvements. Figure 2.14
illustrates the impact of the target delay parameter τi in the PLR for VoIP and video flows, for
the case of having 100 users per RN. As the target delay increases from 50ms to 150ms, the
PLR decreases. Additionally it may be observed that FLSA-CC compared with FLSA exhibits
lower PLR independently of the target delay parameter.
In Figure 2.15, the PLR for VOIP and video flows is presented while the number of
cell RN users varies from 20 to 100. In this case, the considered target delays are set to
100ms and 150ms for VoIP and video flows respectively, as determined by the LTE QoS class
specifications for these service types. As shown, FLSA-CC results in a lower PLR than the rest
2.4 The Proposed FLS Advanced - Cross Carrier (FLSA-CC) Scheduler 37
Figure 2.15 The FLSA-CC Scheduler Evaluation in terms of the Packet Loss Ratio of the Real
Time Flows.
Figure 2.16 The FLSA-CC Scheduler Evaluation in terms of the Throughput of the Real Time
Flows.
of the algorithms. Specifically, FLSA shows a marginal decrease of its PLR for VoIP flows as
well as up to a 3% lower PLR for video flows compared to FLSA.
The analysis of the throughput offered to real time services provides an important insight
on the performance of the FLSA-CC in comparison with the other schedulers. As presented in
Figure 2.16 the FLSA-CC outperforms the rest of the schedulers, independently of the number
of users for VoIP and video flows. This is expected due to the cross carrier scheduling operating
principle applied as well as due to the recalculation of formula (2.11) in each TTI by the upper
level of the FLSA-CC. More specifically, the FLSA-CC achieves higher throughputs than the
rest of the algorithms providing rates of up to 800kbps for VoIP and up to 28Mbps for video
services.
The proposed scheduler is also evaluated in terms of Jain fairness index, which is estimated
using formula 2.14. Figure 2.17 demonstrates that the FLSA-CC scheduler improves the
fairness for both VoIP and video flows.
38 Resource Allocation - Scheduling
Figure 2.17 The FLSA-CC Scheduler Evaluation in terms of the Fairness Index of the Real
Time Flows.
In this subsection tne FLSA-CC, FLSA and FLS schedulers, which accomplish better per-
formance for real time services are evaluated for best effort flows in terms of the throughput
and the fairness index. As presented in Figure 2.18, the FLSA-CC outperforms the other two
schedulers and provides a throughput up to 1.5Mbps for best effort flows even when the number
of users increases, while the FLSA accomplishes only a 100kbps throughput. Additionally, the
FLSA-CC scheduler significantly improves the fairness index of the best effort flows.
Figure 2.18 The FLSA-CC Scheduler Evaluation in terms of the Throughput and the Fairness
Index of the Best Effort Flows.
Chapter 3
Mobility Management
The interval-valued pentagonal fuzzy number is a general IVFN case (Figure 3.2) defined
as: A = [AL , AU ] = [(x1L , x2L , x3L , x4L , x5L , vLA , vLA1 , vLA2 ), (xU U U U U U
1 , x2 , x3 , x4 , x5 , vA ,
vU U L U L L L L L
A1 , vA2 ))] where: A ⊂ A , 0 6 x1 6 x2 6 x3 6 x4 6 x5 6 1, 0 6 x1 6 x2 6 x3 6 x4 6 x5 6 1,
U U U U U
vLA 6 vU L L L U U U L L L U U U
A , vA1 6 vA2 6 vA , vA1 6 vA2 6 vA and vA , vA1 , vA2 , vA , vA1 , vA2 ∈ [0, 1]. The operational
rules of the interval-valued pentagonal fuzzy numbers are defined in [160].
1
uAU AU
U
uA1, uAU
2
uAL AL
uAL1, uAL2
Fuzzy Inference (FI) [163–165] involves the mapping of a given input to an output using fuzzy
logic. The Mamdani Pentagonal Fuzzy Inference System (MPFIS) considers two inputs (Input1
and Input2 ) to estimate the value of the Out put. Both Input1 and Input2 have normalized values
within the range [0, 1]. Also, the MFInput1 , MFInput2 , MFOut put membership functions (MF) are
defined, indicating the linguistic terms and the corresponding Interval Valued Pentagonal Fuzzy
Numbers (IVPFN) for the fuzzy representation of the Input1 , Input2 and Out put, respectively.
42 Mobility Management
Thus, for each crisp value, two membership degrees are determined in the corresponding
MF, one for the upper pentagon and one for the lower pentagon. The MPFIS implements the
following methods:
Step 1. Membership Functions Definition: The Membership Functions (MF) for Input1 , Input2
and Out put parameters are defined.
Step 2. Fuzzy rule (or knowledge) base: A set R of fuzzy rules is defined, where each rule r ∈ R
is a simple if-then statement with a condition and a conclusion. The rule’s condition
consists of MFInput1 and MFInput2 membership functions, while its conclusion indicates
an MFOut put membership function.
Step 3. Fuzzification: The Input1 and Input2 crisp values are converted to degrees of member-
ship indicated as Input1′ and Input2′ by a lookup in MFInput1 and MFInput2 membership
functions respectively.
Step 4. Combination of the fuzzified inputs (Fuzzy Operations): A Zr′ degree is calculated
considering the rule’s condition, that indicates the strength of the r rule. Furthermore,
in case that a rule’s condition has multiple parts, the fuzzy operators ‘AND’ and ‘OR’
may be used to combine more than one conditions. The ‘Algebraic product’ and the
‘Algebraic sum’ are applied for the ‘AND’ and the ‘OR’ operators respectively. In our
study, the ‘Algebraic product’ is calculated using formula 3.8, while the ‘Algebraic sum’
is calculated using formula 3.9.
′
Zu,i,r = Input1′ ˆ·Input2′ = Input1′ · Input2′ (3.8)
′
Zu,i,r = (Input1′ + Input2′ ) − (Input1′ · Input2′ ) (3.9)
Step 5. Implication method: The implication method estimates the consequence MFOut putrc of
the rule conclusion, considering both the rule conclusion MFOut putr and the rule strength
Zr′ . The MFOut putr height is trimmed with respect to the Zr′ degree, using formula 3.10,
which applies the Min method.
Step 6. Aggregation method: The aggregation method combines the R rules’ consequences to
calculate the Out put A fuzzy set, using formula 3.11, which applies the Max method.
Step 7. Defuzzification: During the defuzzification, the Out put A fuzzy set is transformed to the
crisp value Out put. Formula 3.31 that applies the Weighted Average method is used,
where µr is the height and hr is the centroid of each rule obtained from the Out put A . Also,
symbols U and L represent the upper and the lower pentagon of each rule respectively .
R
∑ (µrU · hUr + µrL · hLr )
r=1
Out put = R
(3.12)
∑ (hUr + hLr )
r=1
The Analytic Network Process (ANP) was introduced by Saaty [166] to deal with decision
problems with criteria and alternatives that depend on each other. ANP is a generalization
of the Analytic Hierarchy Process (AHP). A decision problem that is analyzed with the ANP
can be designed either as a control-hierarchy or as a non-hierarchical network. The nodes of
the network represent the components (or clusters) of the system while the arcs denote the
interactions between them. All the interactions and the feedbacks within the clusters are called
inner dependencies, while the interactions and the feedbacks between clusters are called outer
dependencies. ANP is composed of four major steps [167]:
Step 1. Model construction and problem structuring: During this step the problem is analyzed
and decomposed into a rational system, like a network.
Step 2. Pairwise comparison matrices and priority vectors: During this step, the pairwise
comparison matrix is derived using Saaty’s nine-point importance scale based (Table
3.1).
Importance Definition
1 Equal Importance
3 Moderate Importance
5 Strong Importance
7 Very Strong Importance
9 Extreme Importance
2, 4, 6, 8 Intermediate Values
Step 3. Supermatrix formation: During this step, the supermatrix of the ANP model is con-
structed to represent the inner and the outer dependencies of the network. This is a
44 Mobility Management
partitioned matrix, where each matrix segment represents a relationship between two
clusters in the network. To construct the supermatrix the local priority vectors obtained
in Step 2 are grouped and placed in the appropriate positions in a supermatrix based on
the flow of influence from one cluster to another, or from a cluster to itself, as in the
loop. Then, the supermatrix is transformed to a stochastic one, the weighted superma-
trix. Finally, the weighted supermatrix is raised to limiting powers until all the entries
converge to calculate the overall priorities, and thus the cumulative influence of each
element on every other element with which it interacts is obtained [168]. At this point,
all the columns of the new matrix, the limit supermatrix, are the same and their values
show the global priority of each element of the network.
For example, if we assume a network with n clusters, where each cluster Ck , k =
1, 2, . . . , n, and has mn elements, denoted as ek1 , ek2 , . . . , ekmk , then the standard form, W ,
for a supermatrix can be expressed as:
C1 ... Ck ... Cn
e11 . . . e1m1 ... ek1 . . . ekmk ... en1 . . . enmn
e11
..
C1 .
W11 ... W1k ... W1n
e1m1
.. .. .. .. .. ..
.
. . . . .
ek1
W= (3.13)
..
Ck .
Wk1 ... Wkk ... Wkn
ekmk
.. .. .. .. .. ..
.
. . . . .
en1
..
. ... ...
Cn Wn1 Wnk Wnn
enmn
Step 4. Selection of the best alternatives: If the supermatrix formed in Step 3 covers the whole
network, then the priority weights of the alternatives can be found in the column of
the alternatives in the normalized supermatrix. Otherwise, additional calculations are
required in order to obtain the overall priorities of the alternatives. The alternative with
the largest overall priority should be selected, as it is the best alternative as determined
by the calculations made using matrix operations.
3.1 Related Methodologies 45
Step 1. Model Construction and Problem Structuring: During this step the problem is analyzed
and decomposed into a rational system, consisting of a network of nodes.
Step 2. Pairwise comparison matrices and priority vectors: In this step, initially the fuzzy
pairwise comparison matrix à is derived for each TF-ANP cluster using the linguistic
terms presented in Table 3.2, which corresponds to the nine-point importance scale
introduced in [166]. The standard form of the à matrix is expressed as follows:
1 ... ã1 j ... ã1n
.. .. ..
. . .
1/ã1i ... 1 ... ãin
à =
. (3.14)
. .. ..
. . .
1/ã1n ... 1/ã jn ... 1
where n denotes the number of the cluster elements. Subsequently, the geometric mean
rÃi of each row i in à is estimated according to formula 3.34, where ⊗ denotes the
multiplication operator of two fuzzy numbers as defined in [169].
1
rÃi = (ãi1 ⊗ ãi2 ⊗ ... ⊗ ãin ) n (3.15)
where each ω̃i = (ω1U , ω2U , ω3U , ω4U , vU L , ω L , ω L , ω L , vL ) is calculated using for-
i ); (ω 1 2 3 4 i
mula 3.36. The ⊕ indicates the addition operator of two fuzzy numbers as defined in
[169].
Step 3. Construction of the Supermatrix: In this step, the fuzzy supermatrix W̃ of the TF-ANP
model is constructed. This matrix represents the inner and outer dependencies of the
TF-ANP network. It is a partitioned matrix, with each matrix segment representing the
relationship between two clusters of the network. To construct the supermatrix, the local
priority vectors Ω̃ are grouped and placed in the appropriate positions in the supermatrix
46 Mobility Management
based on the flow of influence from one cluster to another, or from a cluster to itself,
as in the loop. For example, if we assume a TF-ANP network of q clusters, Ck with
k = [1, 2, . . . , q] and that each cluster has nq elements, denoted as ek1 , ek2 , . . . , eknk , then
the supermatrix W̃ is expressed as:
C1 ... Ck ... Cq
e11 ...e1n1 ... ek1 ...eknk ... eq1 ...eqnq
e11
..
C1 .
W̃11 ... W̃1 j ... W̃1q
e1n1
.. .. .. .. .. ..
.
. . . . .
ek1
W̃ = (3.18)
..
Ck .
W̃k1 ... W̃k j ... W̃kq
eknk
.. .. .. .. .. ..
.
. . . . .
eq1
..
. ... ...
Cq W̃q1 W̃q j W̃qq
eqnq
Step 4. Construction of the Weighted Supermatrix: During this step, the supermatrix is trans-
formed to a stochastic one,the Weighted Supermatrix W̃ ′ using formula 3.19.
Step 5. Calculation of the Limited Supermatrix: In this step, initially the deffuzified Weighted
Supermatrix W is estimated using the Weighted Average method. According to this
method formula 3.20 is used, where the parameters v and d represent the height and the
centroid of each W̃k,′ j trapezoid respectively. Subsequently, W is raised to limiting powers
until all the entries converge. By this way, the overall priorities are calculated, and thus
the cumulative influence of each element on every other interacting element is obtained
[168]. At this point, all the columns of the produced Limit Supermatrix are the same and
their values show the estimated weight for each element of the TF-ANP network.
vU · dU + vL · d L
Wk, j = (3.20)
dU + d L
A decision problem that is analyzed with the TF-AANP can be represented as a network
of nodes. Each node represents a component (or cluster) of the system while arcs denote
3.1 Related Methodologies 47
interactions between them. Interactions and feedbacks within clusters are called inner depen-
dencies, while interactions and feedbacks between clusters are called outer dependencies. The
TF-AANP is composed of seven major steps:
Step 1. Estimation of the importance of each service: The fuzzy pairwise comparison matrix
P̃ is derived for the services using the linguistic terms presented in Table 3.2, which
correspond to the nine-point importance scale introduced in [166]. The standard form of
the P̃ matrix is expressed as follows:
1 ... p̃1 j ... p̃1S
.. .. ..
. . .
1/ p̃1s ... 1 ... p̃sS
P̃ =
. (3.21)
. .. ..
. . .
1/ p̃1S ... 1/ p̃ jS ... 1
where S denotes the number of the services. Subsequently, the geometric mean rP̃s of
each service (row) s in P̃ is estimated according to formula 3.22, where ⊗ denotes the
multiplication operator of two fuzzy numbers as defined in [169].
1
rP̃s = ( p̃s1 ⊗ p̃s2 ⊗ ... ⊗ p̃sS ) S (3.22)
Then, the priority vector Ω̃P̃s of services is constructed as follows:
Ω̃P̃s = [ ω̃ p̃1 ω̃ p̃2 ... ω̃ p̃S ] (3.23)
h i
where each ω̃ p̃s = (ω U
p̃1 , ω U , ω U , ω U , vU ); (ω L , ω L , ω L , ω L , vL ) is calculated us-
p̃2 p̃3 p̃4 ps p̃1 p̃2 p̃3 p̃4 ps
ing formula 3.24. The ⊕ indicates the addition operator of two fuzzy numbers as defined
in [169].
ω̃ p̃s = rP̃s /(rP̃1 ⊕ rP̃2 ⊕ ... ⊕ rP̃s ⊕ ... ⊕ rP̃S ) (3.24)
Step 2. Model Construction and Problem Structuring for each service: During this step, for each
service s ∈ S the problem is analyzed and decomposed into a rational system, consisted
of a network of nodes.
Step 3. Pairwise comparison matrices and priority vectors: In this step, for each service s ∈ S,
the fuzzy pairwise comparison matrix Ãs is derived for each TF-AANP cluster using the
linguistic terms presented in Table 3.2. The standard form of the à matrix is expressed as
follows:
1 ... ãs1 j ... ãs1n
.. .. ..
. . .
1/ãs1i ... 1 ... ãsin
Ãs =
. (3.25)
. .. ..
. . .
1/ã1sn ... 1/ãs jn ... 1
48 Mobility Management
where n denotes the number of the cluster elements. Subsequently, the geometric mean
rÃsi of each row i in Ãs is estimated according to formula 3.26.
1
rÃs i = (ãsi1 ⊗ ãsi2 ⊗ ... ⊗ ãsin ) n (3.26)
Then, the priority vector Ω̃si of the cluster elements is constructed as follows:
Ω̃si = [ ω̃s1 ω̃s2 ... ω̃sn ] (3.27)
U U U U U L , ω L , ω L , ω L , vL ) is calculated using
where each ω̃si = (ωs1 , ωs2 , ωs3 , ωs4 , vsi ); (ωs1 s2 s3 s4 si
formula 3.36. The ⊕ indicates the addition operator of two fuzzy numbers as defined in
[169].
Step 4. Construction of the Supermatrices: In this step, the fuzzy supermatrix W̃s of the TF-
AANP model is constructed for each service representing the inner and the outer depen-
dencies of the TF-AANP network. It is a partitioned matrix, with each matrix segment
representing the relationship between two clusters of the network. To construct the
supermatrix, the local priority vectors Ω̃s are grouped and placed in the appropriate
positions in the supermatrix based on the flow of influence from one cluster to another,
or from a cluster to itself, as in the loop. For example, if we assume that we have a
TF-AANP network of q clusters, Ck , with k = [1, 2, . . . , q] and that each cluster has nq
elements, denoted as ek1 , ek2 , . . . , eknk , then the supermatrix is expressed as:
C1 ... Ck ... Cq
e11 ...e1n1 ... ek1 ...eknk ... eq1 ...eqnq
e11
..
C1 .
W̃s11 ... W̃s1 j ... W̃s1q
e1n1
.. .. .. .. .. ..
.
. . . . .
ek1
W̃s = (3.29)
..
Ck .
W̃sk1 ... W̃sk j ... W̃skq
eknk
.. .. .. .. .. ..
.
. . . . .
eq1
..
. ... ...
Cq W̃sq1 W̃sq j W̃sqq
eqnq
Step 5. Construction of the Weighted Supermatrices: During this step, the supermatrix of each
service is transformed to a stochastic one,the Weighted Supermatrix W̃s′ using formula
3.1 Related Methodologies 49
3.38.
′
W̃s,k, j = W̃s,k, j /q (3.30)
Step 6. Calculation of the Limited Supermatrices: In this step, initially the deffuzified Weighted
Supermatrix Ws of each service is estimated by applying the Weighted Average method.
According to this method formula 3.31 is used, where the parameters vs and ds represent
the height and the centroid of each W̃s,k,′
j trapezoid respectively. Subsequently, each
Ws is raised to limiting powers until all the entries converge. By this way the overall
priorities are calculated, and thus the cumulative influence of each element on every other
interacting element is obtained [168]. At this point, all the columns of each produced
Limit Supermatrix Ls , are the same and their values show the importance of each element
e of the TF-AANP network for the corresponding service s.
vUs · dsU + vLs · dsL
Ws,k, j = (3.31)
dsU + dsL
Step 7. Estimation of the Criteria Weights: In this step, the weight we for each element e of the
TF-AANP network is calculated using formula 3.32 where the importance ω̃Ps˜ of each
service s is considered.
S
we = ∑ ω̃P̃,s ∗ Ls,e (3.32)
s=1
The Pentagonal Fuzzy Analytic Network Process (PF-ANP) method is the IVPFN version of
the typical ANP [166] method, used for the calculation of the criteria weights. PF-ANP allows
complex relationships within and among clusters of selection criteria using a network model of
dependencies. A decision problem that is analyzed with the PF-ANP can be represented as a
network of nodes. Each node represents a component (or cluster) of the system while the arcs
denote the interactions between them. The interactions and the feedbacks within the clusters are
called inner dependencies, while the interactions and the feedbacks between clusters are called
outer dependencies. Indicatively, we could consider one cluster with network characteristics
criteria such as throughput, delay, jitter and packet loss, as well as one cluster with operator
policy criteria such as service reliability, security and price. In this situation, the PF-ANP will
consider two clusters. The PF-ANP is composed of five major steps of selection criteria:
Step 1. Model Construction and Problem Structuring: During this step the problem is analyzed
and decomposed into a rational system, consisting of a network of nodes.
Step 2. Pairwise comparison matrices and priority vectors: Initially the fuzzy pairwise compari-
son matrix à is derived for each cluster using the linguistic terms presented in Table 3.2,
50 Mobility Management
which are calculated using the EUM method and correspond to the nine-point importance
scale introduced in [166]. The standard form of the à matrix is expressed as follows:
1 ... ã1 j ... ã1n
.. .. ..
. . .
1/ã1i ... 1 ... ãin
à =
. (3.33)
. .. ..
. . .
1/ã1n ... 1/ã jn ... 1
where n denotes the number of the cluster elements. Subsequently, the geometric mean
Table 3.2 The Linguistic Terms that are used for the Criteria Pairwise Comparisons.
rÃi of each row i in à is calculated according to formula 3.34, where ⊗ denotes the
multiplication operator of two fuzzy numbers as defined in [169].
1
rÃi = (ãi1 ⊗ ãi2 ⊗ ... ⊗ ãin ) n (3.34)
Then, the priority vector Ω̃i of each cluster element is constructed as follows:
Ω̃i = [ ω̃1 ω̃2 ... ω̃n ] (3.35)
Step 3. Construction of the Supermatrix: In this step, the fuzzy supermatrix W̃ of the PF-
ANP model is constructed representing the inner and outer dependencies of the PF-
ANP network. This is a partitioned matrix, with each matrix segment representing the
3.1 Related Methodologies 51
relationship between two clusters of the network. To construst the supermatrix, the local
priority vectors Ω̃ are grouped and placed in the appropriate positions in the supermatrix
based on the flow of influence from one cluster to another. For example if we assume a
network of q clusters, where each cluster Ck , k = [1, 2, . . . , q] has nk elements, denoted
as ek1 , ek2 , . . . , eknk , then the supermatrix is expressed as:
C1 ... Ck ... Cq
e11 ...e1n1 ... ek1 ...eknk ... eq1 ...eqnq
e11
..
C1 .
W̃11 ... W̃1 j ... W̃1q
e1n1
.. .. .. .. .. ..
.
. . . . .
ek1
W̃ = (3.37)
..
Ck .
W̃k1 ... W̃k j ... W̃kq
eknk
.. .. .. .. .. ..
.
. . . . .
eq1
..
. ... ...
Cq W̃q1 W̃q j W̃qq
eqnq
Step 4. Construction of the Weighted Supermatrix: During this step, the supermatrix is trans-
formed to a stochastic one,the Weighted Supermatrix W̃ ′ using formula 3.38.
Step 5. Calculation of the Limited Supermatrix: In this step, initially the deffuzified Weighted
Supermatrix W is estimated by applying the Weighted Average method using formula
3.39. The parameters v and d represent the height and the centroid of each W̃k,′ j pentagon
respectively. Subsequently, W is raised to limiting powers until all the entries converge.
In this way the overall priorities are calculated, and thus the cumulative influence of
each element on every other interacting element is obtained [168]. At this point, all the
columns of the produced Limited Supermatrix, are the same and their values show the
global priority of each element of the network.
vU · dU + vL · d L
Wk, j = (3.39)
dU + d L
52 Mobility Management
TOPSIS, introduced by Hwang and Yoon [170], is based on the concept that the best alternative
should have the shortest distance from the positive ideal solution and the longer distance
from the negative ideal solution. In the present work, network selection is performed using a
proposed fuzzy version of TOPSIS, namely the Trapezoidal interval-valued Fuzzy TOPSIS
(TFT). This method assumes that the linguistic values of the criteria attributes are represented
by interval-valued trapezoidal fuzzy numbers.
Suppose that A = {A1 , A2 , . . . , An } is the set of possible alternatives, C = {C1 ,C2 , . . . ,Cn } is
the set of criteria and w1 , w2, . . . , wm are the weights of each criterion. The steps of the method
are as follows:
Step 1. Construction of the decision matrix: Each element xi j of the n × m decision matrix
D is an interval-valued trapezoidal fuzzy number which expresses the performance of
alternative i for criterion j. Thus
C1 . . . Cm
A1 x11 . . . x1m
D= .. .. .. .. (3.40)
. . . .
An xn1 . . . xnm
h i
L L L L L U U U U U
where: xi j = (xi j1 , xi j2 , xi j3 , xi j4 , vi j ), (xi j1 , xi j2 , xi j3 , xi j4 , vi j )
In case there are Q decision makers, the decision matrix and the criteria weights include
the average of the performance values and of the weights of the decision makers, re-
spectively. Hence, assuming that for the k-th decision maker xi jk is the performance of
alternative i for criterion j, and w jk is the importance weight for criterion j, the average
of the performance values and weights are given by
" !
1 Q 1 Q L 1 Q L 1 Q L 1 Q L L
xi j = ∑ xi jk = ∑ xi jk1 , Q ∑ xi jk2 , Q ∑ xi jk3 , Q ∑ xi jk4 , vi jk ,
Q k=1 Q k=1 k=1 k=1 k=1
!#
1 Q U 1 Q U 1 Q U 1 Q U
∑ xi jk1 , Q ∑ xi jk2 , Q ∑ xi jk3 , Q ∑ xi jk4 , vUijk
Q k=1
(3.41)
k=1 k=1 k=1
and
1 Q
wj = ∑ w jk . (3.42)
Q k=1
Step 2. Normalization of the decision matrix: Consider that Ωb is the set of benefits attributes
and Ωc is the set of costs attributes. Then the elements of the normalized decision matrix
are computed as:
3.1 Related Methodologies 53
a) " ! !#
xiLj1 xiLj2 xiLj3 xiLj4 L xU U U U
i j1 xi j2 xi j3 xi j4 U
ri j = , , , ,v , , , , ,v (3.43)
bj bj bj bj ij bj bj bj bj ij
where b j = maxi xU
i j4 for each j ∈ Ω b and
b) " ! !#
cj cj cj cj L cj cj cj cj U
ri j = , , , ,v , , U , U , U , vi j (3.44)
xiLj4 xiLj3 xiLj2 xiLj1 i j xU
i j4 xi j3 xi j2 xi j1
Step 3. Construction of the weighted normalized decision matrix: The weighted normalized
decision matrix is constructed by multiplying each element of the normalized decision
matrix ri j with the respective weight w j according to the formula
Step 4. Determination of the positive and negative ideal solution: The positive ideal solution X +
is defined as follows:
h i
X+ = xi+L , x +L +L +L +L
, x , x , v
j1 i j2 i j3 i j4 i j , x +U +U +U +U +U
i j1 , xi j2 , xi j3 , x i j4 , vij =
" ! !#
^ ^ ^ ^ ^ ^ ^ ^
uLij1 , uLij2 , uLij3 , uLij4 , vLij , uUij1 , uUij2 , uUij3 , uUij4 , vUij (3.46)
i i i i i i i i
h i
X− = xi−L −L −L −L −L
j1 , xi j2 , xi j3 , xi j4 , vi j , xi−U −U −U −U −U
j1 , xi j2 , xi j3 , xi j4 , vi j
(3.47)
" ! !#
_ _ _ _ _ _ _ _
= uLij1 , uLij2 , uLij3 , uLij4 , vLij , uUij1 , uUij2 , uUij3 , uUij4 , vUij
i i i i i i i i
Step 5. Measurement of the distance of each alternative from the ideal solutions: The distances
+ +
di1 and di2 of each alternative from the positive ideal solution are evaluated as follows:
m
1 2 2 2 2 21
+ L +L L +L L +L L +L
di1 = ∑ ui j1 − xi j1 + u i j2 − xi j2 + u i j3 − x i j3 + ui j4 − x i j4 (3.48)
j=1 4
1 m 2 2 2 2 21
+ U +U U +U U +U U +U
di2 =∑ ui j1 − xi j1 + ui j2 − xi j2 + ui j3 − xi j3 + ui j4 − xi j4 (3.49)
j=1 4
54 Mobility Management
− −
Likewise, the distances di1 and di2 of each alternative from the negative ideal solution
are estimated as follows:
m
1 2 2 2 2 21
− L −L L −L L −L L −L
di1 =∑ ui j1 − xi j1 + ui j2 − xi j2 + ui j3 − xi j3 + ui j4 − xi j4 (3.50)
j=1 4
m
1 2 2 2 2 21
− U −U U −U U −U U −U
di2 =∑ ui j1 − xi j1 + ui j2 − xi j2 + ui j3 − xi j3 + ui j4 − xi j4 (3.51)
j=1 4
Consequently, similarly to [158] the distance of the alternatives from the positive and the
+ + − −
negative ideal solutions are expressed by intervals such as [di1 , di2 ] and [di1 , di2 ], instead
of single values. In this way less information is lost.
Step 6. Calculation of the relative closeness: The relative closeness of the distances from the
ideal solutions are computed as:
−
di1
RCi1 = + − (3.52)
di1 + di1
and −
di2
RCi2 = + − (3.53)
di2 + di2
The compound relative closeness RCi is obtained from the average of the above values:
RCi1 + RCi2
RCi = (3.54)
2
Step 7. Alternatives ranking: The alternatives are ranked according to their RCi values. The best
alternative is the one with the higher RCi value.
3.1.4.2 The Trapezoidal Fuzzy TOPSIS with Adaptive Criteria Weights (TFT-ACW)
The candidate networks are ranked using the TFT-ACW algorithm, which improves the TFT
[171] by using the adaptive weights that estimated from the TF-AANP method. Thus, the
importance of the opinion of each service (decision maker) is considered, since the opinions of
some decision makers could have a higher importance than the ones of the other decision makers.
In general, similar to TFT, the TFT-ACW method is based on the concept that the best alternative
should have the shortest distance from the positive ideal solution and the longer distance from
the negative ideal solution. Also, it assumes that the linguistic values of the criteria attributes
are represented by interval-valued trapezoidal fuzzy numbers. More specifically, suppose that
AL = {AL1 , AL2 , . . . , ALz } is the set of possible alternatives, CR = {CR1 ,CR2 , . . . ,CRn } is the
set of criteria and w1 , w2 , . . . , wn are the importance weights of the respective criteria obtained
from the application of the TF-AANP algorithm. The steps of the method are as follows:
3.1 Related Methodologies 55
Step 1. Construction of the decision matrix: Each element g̃ie of the z × n decision matrix D̃ is
an IVTFN number expressing the performance of alternative i for criterion e. Thus,
where g̃ie = (gLie1 , gLie2 , gLie3 , gLie4 , vLie ), (gUie1 , gUie2 , gUie3 , gUie4 , vUie ) .
In the case that there are S services (decision makers) the decision matrix include the
average of the performance values. Hence, assuming that for the sth decision maker g̃iex is
the performance of alternative i for criterion (element) e, the average of the performance
values is given by formula 3.56.
S
g̃ie = ∑ (g̃ies · ω̃ p̃s ) (3.56)
s=1
Step 2. Normalization of the decision matrix: Let Γb be the set of benefit attributes and Γc the
set of cost attributes. Then, the elements of the normalized decision matrix are calculated
using either formula 3.57 or 3.58, where be = maxi gU L
ie4 for each e ∈ Γb and ce = mini gie4
for each e ∈ Γc .
gU gU gU
U
gLie1 gLie2 gLie3 gLie4 L
g
g̃′ie = , , , , vie , ie1 , ie2 , ie3 , ie4 , vUie (3.57)
be be be be be be be be
ce ce ce ce L ce ce ce ce U
g̃′ie = , , , , v , , , , , v (3.58)
gLie4 gLie3 gLie2 gLie1 ie gUie4 gUie3 gUie2 gUie1 ie
Step 3. Construction of the weighted normalized decision matrix: The weighted normalized
decision matrix is constructed by multiplying each element of the normalized decision
matrix g̃′ie with the respective weight we according to the formula 3.59.
h
ũie = g′L ′L ′L ′L ′L
ie1 · we , gie2 · we , gie3 · we , gie4 · we , vie ,
i (3.59)
g′U
ie1 · we , g′U
ie2 · we , g′U
ie3 · we , g′U
ie4 · we , vU
ie
Step 4. Determination of the positive and negative ideal solution: The positive ideal solution
is defined in formula 3.60, where ≡ maxi in case e ∈ Γb and ≡ mini in case e ∈ Γc .
V V
i i
≡ mini
W
Correspondingly, the negative ideal solution is defined in formula 3.61, where i
56 Mobility Management
Step 5. Measurement of the distance of each alternative from the ideal solutions: The distances
of each alternative from the positive ideal solution are evaluated using formulas 3.62 and
3.63.
n
1 2 2 2 2 12
p+
i1 =∑ L +L L +L L +L L +L
uie1 − gie1 + uie2 − gie2 + uie3 − gie3 + uie4 − gie4 (3.62)
e=1 4
n
1 2 2 2 2 12
p+
i2 = ∑ uU
ie1 − g+U
ie1 + uU
ie2 − g+U
ie2 + uU
ie3 − g+U
ie3 + uU
ie4 − g+U
ie4
(3.63)
e=1 4
Likewise, the distances of each alternative from the negative ideal solution are estimated using formulas
3.64 and 3.65.
n
1 2 2 2 2 12
p−
i1 = ∑ uL
ie1 − g−L
ie1 + uL
ie2 − g−L
ie2 + u L
ie3 − g−L
ie3 + u L
ie4 − g−L
ie4
(3.64)
e=1 4
n
1 2 2 2 2 12
p−
i2 =∑ U −U U −U U −U U −U
uie1 − gie1 + uie2 − gie2 + uie3 − gie3 + uie4 − gie4 (3.65)
e=1 4
Consequently, the alternative distance from the positive and from the negative ideal
− −
solutions are expressed by intervals such as [p+ +
i1 , pi2 ] and [pi1 , pi2 ], instead of single
values, thus less information is lost.
Step 6. Calculation of the relative closeness: The relative closeness of the distances from the
ideal solutions are calculated using formula 3.66 and 3.67.
p−
i1
RCi1 = (3.66)
p+
i1 + p−
i1
p−
i2
RCi2 = − (3.67)
p+
i2 + pi2
3.1 Related Methodologies 57
RCi1 + RCi2
RCi = (3.68)
2
Step 7. Alternative networks ranking: The alternative networks are ranked according to their RCi
values. The best alternative is that with the higher RCi value.
The Pentagonal Fuzzy TOPSIS (PFT) algorithm is an improved version of the TOPSIS using
interval-valued pentagonal fuzzy numbers. PFT assumes that the linguistic values of the criteria
attributes are represented by IVPFN numbers.
Suppose that A = {A1 , A2 , ..., An } is the set of possible alternatives, C = {C1 ,C2 , . . . ,Cm } is
the set of criteria and w1 , w2, . . . , wm are the weights of each criterion. The steps of the method
are as follows:
Step 1. Construction of the decision matrix: Each element xi j of the n × m decision matrix
D is an interval-valued pentagonal fuzzy number which expresses the performance of
alternative i for criterion j. Thus
C1 . . . Cm
A1 x11 . . . x1m
D= .. .. .. .. (3.69)
. . . .
An xn1 . . . xnm
where: xi j = [(xiLj1 , xiLj2 , xiLj3 , xiLj4 , xiLj5 , vLij , vLij1 , vLij2 ), (xUij1 , xUij2 , xUij3 , xUij4 , xUij5 , vUij , vUij1 , vUij2 )]. In case there
are Q decision makers the decision matrix and the criteria weights include the average
of the performance values and of the weights respectively, of the decision makers.
Hence, assuming that for the k-th decision maker xi jk is the performance of alternative
i for criterion j, and w jk is the importance weight for criterion j, the average of the
performance values and weights are given by:
1 Q
xi j = ∑ xi jk (3.70)
Q k=1
and
1 Q
wj = ∑ w jk . (3.71)
Q k=1
Step 2. Normalization of the decision matrix: Consider that Ωb is the set of benefit attributes and
Ωc is the set of cost attributes. Then, the elements of the normalized decision matrix are
58 Mobility Management
computed as:
U
xiLj1 xiLj2 xiLj3 xiLj4 xiLj5 L L L xU U U U
i j1 xi j2 xi j3 xi j4 xi j5 U U
ri j = [( , , , , , vi j , vi j1 , vi j2 ), ( , , , , , v , v , vU )] (3.72)
bj bj bj bj bj b j b j b j b j b j i j i j1 i j2
where b j = maxi xU
i j4 for each j ∈ Ω b , or
cj cj cj cj cj L L L cj cj cj cj cj U U U
ri j = [( , , , , , v , v , v ), ( , U , U , U , U , vi j , vi j2 , vi j1 )] (3.73)
xiLj5 xiLj4 xiLj3 xiLj2 xiLj1 i j i j2 i j1 xUi j5 xi j4 xi j3 xi j2 xi j1
Step 3. Construction of the weighted normalized decision matrix: The weighted normalized
decision matrix is constructed by multiplying each element of the normalized decision
matrix ri j with the respective weight w j according to:
Step 4. Determination of the positive and negative ideal solutions: The positive ideal solution is
given by:
X + = [(xi+L +L +L +L +L +L +L +L +U +U +U +U +U +U +U +U
j1 , xi j2 , xi j3 , xi j4 , xi j5 , vi j , vi j1 , vi j2 ), (xi j1 , xi j2 , xi j3 , xi j4 , xi j5 , vi j , vi j1 , vi j2 )] =
^ ^ ^ ^ ^ ^ ^ ^ ^ ^
[( uLij1 , uLij2 , uLij3 , uLij4 , uLij5 , vLij , vLij1 , vLij2 ), ( uUij1 , uUij2 , uUij3 , uUij4 , uUij5 , vUij , vUij1 , vUij2 )]
i i i i i i i i i i
(3.75)
X − = [(xi−L −L −L −L −L −L −L −L −U −U −U −U −U −U −U −U
j1 , xi j2 , xi j3 , xi j4 , xi j5 , vi j , vi j1 , vi j2 ), (xi j1 , xi j2 , xi j3 , xi j4 , xi j5 , vi j , vi j1 , vi j2 )] =
_ _ _ _ _ _ _ _ _ _
[( uLij1 , uLij2 , uLij3 , uLij4 , uLij5 , vLij , vLij1 , vLij2 ), ( uUij1 , uUij2 , uUij3 , uUij4 , uUij5 , vUij , vUij1 , vUij2 )]
i i i i i i i i i i
(3.76)
Step 5. Measurement of the distance of each alternative from the ideal solutions: The distances
of each alternative from the positive ideal solution are evaluated as follows:
m
+ 1 1
di1 = ∑ { 5 [(uLij1 − xi+Lj1 )2 + (uLij2 − xi+Lj2 )2 + (uLij3 − xi+Lj3 )2 + (uLij4 − xi+Lj4 )2 + (uLij5 − xi+Lj5 )2 ]} 2 (3.77)
j=1
m
+ 1 1
di2 = ∑ { 5 [(uUij1 − xi+U 2 U +U 2 U +U 2 U +U 2 U +U 2 2
j1 ) + (ui j2 − xi j2 ) + (ui j3 − xi j3 ) + (ui j4 − xi j4 ) + (ui j5 − xi j5 ) ]} (3.78)
j=1
3.1 Related Methodologies 59
Likewise the distances of each alternative from the negative ideal solution are estimated
by:
m 1
−
di1 = ∑ {15[(uLij1 − xi−Lj1 )2 + (uLij2 − xi−Lj2 )2 + (uLij3 − xi−Lj3 )2 + (uLij4 − xi−Lj4 )2 + (uLij5 − xi−Lj5 )2 ]} 2 (3.79)
j=1
m
− 1 1
di2 = ∑ { 5 [(uUij1 − xi−U 2 U −U 2 U −U 2 U −U 2 U −U 2 2
j1 ) + (ui j2 − xi j2 ) + (ui j3 − xi j3 ) + (ui j4 − xi j4 ) + (ui j5 − xi j5 ) ]} (3.80)
j=1
Consequently, similarly to [158] the distance of the alternatives from the positive and
+ + − −
negative ideal solutions are expressed by intervals such as [di1 , di2 ] and [di1 , di2 ], instead
of single values. In this way less information is lost.
Step 6. Calculation of the relative closeness: The relative closeness RCi1 and RCi2 of the dis-
tances from the ideal solutions are computed as:
−
di1
RCi1 = + − (3.81)
di1 + di1
and −
di2
RCi2 = + − (3.82)
di2 + di2
The compound relative closeness RCi is obtained from the average of the above values
RCi1 + RCi2
RCi = (3.83)
2
Step 7. Alternatives ranking: The alternatives are ranked according to their RCi values. The best
alternative is that with the higher RCi value.
Criteria Groups
Network QoS Network Policy
Characteristics Characteristics
Criteria
Throughput Price
Throughput Price
Service
Delay Jitter
Reliability
Figure 3.4 The Relations of the Criteria considered by the ANP Network Model.
3.1 Related Methodologies 61
slightly worse decision criteria values than those offered by SLA1. SLA3 supports only the
buffered streaming and the web browsing services and satisfactory QoS characteristics and
policies, whereas the low price SLA4 supports only the web browsing service while providing
acceptable decision criteria values.
Table 3.4 The ANP Supermatrix for the SLA1 VoIP Service.
Table 3.5 The ANP Weighted Supermatrix for the SLA1 VoIP Service.
The criteria weights per service and SLA obtained by the limit supermatrices are presented
in Figures 3.5, 3.6, 3.7 and 3.8. As illustrated, the weights are proportional to the constraints
of each service as well as to the agreements of each SLA. In particular, the weight of the price
62 Mobility Management
Table 3.6 The ANP Limit Supermatrix for the SLA1 VoIP Service.
criterion is low for SLA1, in which the service reliability and the network QoS characteristics
are considered as the most important factors. In SLA2 the price criterion is more important than
in SLA1, thus the respective weight is greater than that of SLA1. Consequently, the weights
of the service reliability and QoS characteristics criteria in SLA2 are lower compared to the
relative weights of SLA1. In SLA3 the weights of the price and the service reliability criteria
are balanced as they are almost of equivalent importance. Finally, in SLA4 the price is the most
important criterion resulting in a high estimated weight value.
SLA1
0.8
Throughput
Delay
Jitter
0.7 Packet Loss
Price
Reliability
Security
0.6
0.5
Weight
0.4
0.3
0.2
0.1
0
VoIP ConversationalVideo BufferedStreaming RealTimeGaming Web
SLA2
0.45
Throughput
Delay
Jitter
0.4
Packet Loss
Price
Reliability
0.35 Security
0.3
0.25
Weight
0.2
0.15
0.1
0.05
0
ConversationalVideo BufferedStreaming Web
The ranking of the networks alternatives is performed using the TFT algorithm described
in the subsection 3.1.4.1. The weights of the network selection criteria are obtained from
Figures 3.5 - 3.8. The linguistic terms for the criteria attributes are represented by interval-
valued trapezoidal fuzzy numbers as shown in Table 3.7. The network policy specifications
are expressed directly using linguistic terms. Additionally, crisp values of network QoS
characteristics are converted into linguistic terms which correspond to specific ranges of values
per service type. Table 3.8 presents a relative example for the VoIP service, illustrating the
correspondence between ranges of network QoS characteristics values and linguistic terms.
The available - candidate networks in our simulations at the time of network selection per
service and SLA, as well as, their specifications expressed by linguistic terms are depicted in
Tables 3.9 and 3.10.
The case of having several services of different QoS constraints running at the user site is
being addressed by the TFT algorithm and network selection is performed in a way satisfying
multiple groups of criteria per user. We consider the case where nine users need to select a
network which satisfies the requirements of their services as presented in Table 3.11 and at the
same time comply with their respective SLA agreements. To achieve this goal the proposed
64 Mobility Management
SLA3
0.5
Throughput
Delay
Jitter
Packet Loss
Price
0.4 Reliability
Security
0.3
Weight
0.2
0.1
0
BufferedStreaming Web
TFT algorithm is applied for each user and the available networks are ranked as shown in
Figure 3.9. The positive and the negative ideal solutions are represented by unary and null
trapezoidal fuzzy numbers respectively to eliminate the ranking abnormality problem.
From the obtained results it is clear that the ranking of the network alternatives is in
accordance with the users expectations. For example, user 1 who requires increased QoS
provisioning selects LTE 1 network which guarantees the best QoS characteristics and service
reliability. As Figure 3.9 depicts, LTE 1 achieves higher ranking than the other networks,
due to the high values of the QoS characteristics and service reliability factors bearing higher
importance according to the relative ANP weights in SLA1. On the contrary, user 9 whose
prior selection criterion is the price of the service, selects the WiFi 1 network which satisfies
his requirements as expressed in his SLA agreement.
The performance of the TFT algorithm was evaluated against the original TOPSIS method,
as well as, against the method presented in [172], the Fuzzy AHP - ELECTRE (FAE) method.
The FAE method calculates the criteria weights using the Fuzzy AHP and performs the network
selection by applying the ELECTRE algorithm. We consider the scenario of the nine users of
Table 3.11. A critical weakness of TOPSIS and FAE is that they do not support users with
3.1 Related Methodologies 65
SLA4
0.45
Throughput
Delay
Jitter
0.4
Packet Loss
Price
Reliability
0.35 Security
0.3
0.25
Weight
0.2
0.15
0.1
0.05
0
Web
more than one service. In these cases, the TOPSIS and FAE methods consider only the most
demanding service of the user. Specifically, for users 2 and 3 they applied only considering the
VoIP service and for user 7 are applied only considering the BStreaming service, whereas for
the rest users the methods are applied respectively for each single user service defined in Table
3.11.
Table 3.12 presents the network classification performed by TFT, TOPSIS and FAE.
From the analysis of the results we conclude that when a user has only one service, the
methods usually provide similar results. However, when a user requires multiple services, TFT
accomplishes more reliable results than TOPSIS and FAE, since it considers the weights of
each service. For example, the results concerning user 1 using only the VoIP service, are similar
for TFT and TOPSIS with the exception of the evaluation sequence of the WiFi 1 and WiFi
2 networks. Also, FAE accomplishes quite similar network rates with TFT and TOPSIS for
this user. Nevertheless, TFT achieves more reliable results for user 4, compared with TOPSIS
and FAE. In this case, only the RTGaming service is used and the most important criteria are
service reliability, throughput and delay. TFT selects the WiMAX2 network which provides
AG for service reliability, VG for throughput and AG for delay criterion. On the other hand,
66 Mobility Management
Table 3.7 Linguistic Terms and the corresponding Interval-Valued Trapezoidal Fuzzy Numbers
used for the Evaluation of the TFT algorithm.
Table 3.8 Relation of the Network QoS Characteristics and the Linguistic Terms for VoIP.
Linguistic term Throughput range (Kbps) Delay range (ms) Jitter range (ms) Packet loss range
AP ≤ 164 ≥ 116 ≥ 65 ≥ 0.4
VP 165 - 174 111-115 55 - 64 ≥ 0.2 - 0.4
P 175 - 184 106-110 45 - 54 >10−1 - <0.2
MP 185 - 194 100 - 105 40 - 44 10−1
M 195 - 204 95 - 99 35 - 49 10−2
MG 205 - 214 86 - 94 30 - 34 10−3
G 215 - 224 66 - 85 25 - 29 10−4
VG 225 - 239 41 - 65 20 - 24 10−5
AG ≥ 240 ≤ 40 ≤ 20 ≤ 10−6
TOPSIS selects the LTE1 network, which has similar values with the WiMAX2 for service
reliability and delay criteria but worse performance for throughout by providing G instead of
VG. Moreover, FAE does not provide a clear choice for user 4 and results to equal rankings for
both WiMAX2 and LTE1 networks. Finally, the classification of the networks obtained by the
three methods is quite different for user 7 who requests both BStreaming and web browsing
services. In this case, the TFT accomplishes more reliable results by taking into account the
weights of both services.
3.1.5.1.3 Sensitivity Analysis : In this subsection, the sensitivity of TFT is evaluated when
the number of the available access networks changes frequently. We consider three different
network configuration scenarios for the users defined in Table 3.11. In the first scenario all the
networks defined in Tables 3.9 and 3.10 are available. In the second and third scenarios the
LTE 1 and the WiFi 2 networks respectively are not reachable. The graphs of Figures 3.10-3.18
include three column types of different patern indicating the ranking of network alternatives in
each case. In the first case, user 1 selects the LTE 1 network. In the second case, the remaining
networks improve their ranking order, thus user 1 selects the WiMAX 2 network. Furthermore,
in the third case only the last rated WiFi 3 network increases its rank, since the WiFi 2 network
preceded WiFi 3 in the other two cases. Similar behavior is observed in the ranking of the
3.1 Related Methodologies 67
0.1
0.05
0
User1 User2 User3 User4 User5 User6 User7 User8 User9
network alternatives for the other users. From the aforementioned analysis we conclude that the
ranking results of the proposed method are normally adjusted with respect to the heterogeneous
network environment changes, highlighting thus the method’s sensitivity.
68 Mobility Management
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.10 The TFT’s Network Ranking for User 1 in case of Network Environment Changes.
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.11 The TFT’s Network Ranking for User 2 in case of Network Environment Changes.
3.1 Related Methodologies 69
Packet Service
SLA Service Network Throughput Delay Jitter Price Security
Loss Reliability
LTE 1 MG AG VG VG VP AG VG
LTE 2 AG MG AG MG VP VG AG
WiMAX 1 M M MP AG P VG AG
VoIP WiMAX 2 G G G G P AG VG
WiFi 1 VG VG MG AG MP G G
WiFi 2 MG M MG VG MP MG MG
WiFi 3 M MP M AG MP G G
LTE 1 MP MG VG G AP AG VG
LTE 2 AG AG AG VG AP VG AG
WiMAX 1 MP M MG AG AP VG AG
CVideo WiMAX 2 MG MG G AG VP AG VG
WiFi 1 M MG M VG P G G
WiFi 2 VG VG VG AG P MG MG
WiFi 3 G G M VG P G G
LTE 1 M G VG VG AP AG VG
LTE 2 VG VG AG AG AP VG AG
WiMAX 1 M MG MG VG VP VG AG
SLA1 BStreaming WiMAX 2 MG G MG G P AG VG
WiFi 1 VG G M AG P G G
WiFi 2 AG AG G VG P MG MG
WiFi 3 G VG VG AG MP G G
LTE 1 G AG AG VG VP AG VG
LTE 2 G MG VG AG VP VG AG
WiMAX 1 MP MG G AG P VG AG
RTGaming WiMAX 2 VG AG AG VG VP AG VG
WiFi 1 AG VG VG VG VP G G
WiFi 2 M M MG AG MP MG MG
WiFi 3 P M M AG MP G G
LTE 1 AG AG AG AG VP AG VG
LTE 2 MG M G VG MP VG AG
WiMAX 1 G M G AG P VG AG
Web WiMAX 2 VG G VG AG P AG VG
WiFi 1 MG MP MG VG MP G G
WiFi 2 VG G M VG MP MG MG
WiFi 3 AG VG AG AG MP G G
LTE 1 MG G VG AG MP G G
LTE 2 MP M MG VG M G G
CVideo WiMAX 1 M MG G AG MP MG MG
WiMAX 2 MP M M AG M MG MG
WiFi 2 G VG VG AG MG G M
WiFi 3 MP G M VG MG P M
LTE 1 M G G VG MP G G
LTE 2 MG MG AG G MP G G
WiMAX 1 M MG MP AG MP MG MG
SLA2 BStreaming WiMAX 2 G G MG VG MP MG MG
WiFi 1 G VG MG AG MP MP MP
WiFi 2 AG AG VG VG MP M M
WiFi 3 MG VG VG AG M P M
LTE 2 M MP MG VG M G G
WiMAX 1 MG M G AG MG MG MG
WiMAX 2 VG G AG AG M MG MG
Web
WiFi 1 MG MP M VG MG MP MP
WiFi 2 MG M G VG MG M M
WiFi 3 VG VG AG AG MG P M
70 Mobility Management
Table 3.12 Networks’ Classification in respect of TFT, TOPSIS (T) and FAE Results.
User 1
User 2
User 3
User 4
User 5
User 6
User 7
User 8
User 9
FAE
FAE
FAE
FAE
FAE
FAE
FAE
FAE
FAE
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
❵❵ Method
❵❵❵
T
Networks ❵❵❵❵❵
LTE 1 1 1 1 2 1 1 1 1 1 2 1 1 4 2 2 4 2 5 4 2 5 2 1 4 4 6 4
LTE 2 3 3 3 3 3 3 3 3 3 4 4 2 2 3 6 1 1 4 1 3 1 3 2 5 3 3 4
WiMAX 1 5 5 5 5 5 5 5 5 5 5 5 4 3 4 3 5 4 7 2 1 4 1 3 1 6 7 3
WiMAX 2 2 2 4 1 2 4 2 2 4 1 2 1 5 5 4 2 3 2 - - - 5 5 3 5 5 4
WiFi 1 4 6 2 4 6 2 4 6 2 3 3 3 - - - 6 6 3 3 5 3 4 4 2 1 1 1
WiFi 2 6 4 6 7 4 6 6 4 6 6 7 5 1 1 1 3 5 1 5 4 2 6 6 6 7 4 4
WiFi 3 7 7 6 6 7 6 7 7 6 7 6 6 6 6 5 7 7 6 - - - - - - 2 2 2
3.1 Related Methodologies 71
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.12 The TFT’s Network Ranking for User 3 in case of Network Environment Changes.
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.13 The TFT’s Network Ranking for User 4 in case of Network Environment Changes.
72 Mobility Management
6
Scenario 1
Scenario 2
Scenario 3
4
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.14 The TFT’s Network Ranking for User 5 in case of Network Environment Changes.
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.15 The TFT’s Network Ranking for User 6 in case of Network Environment Changes.
3.1 Related Methodologies 73
5
Scenario 1
Scenario 2
Scenario 3
3
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.16 The TFT’s Network Ranking for User 7 in case of Network Environment Changes.
6
Scenario 1
Scenario 2
Scenario 3
4
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.17 The TFT’s Network Ranking for User 8 in case of Network Environment Changes.
74 Mobility Management
7
Scenario 1
Scenario 2
Scenario 3
6
5
Network Score
0
LTE1 LTE2 WiMAX1 WiMAX2 WiFi1 WiFi2 WiFi3
Available Networks
Figure 3.18 The TFT’s Network Ranking for User 9 in case of Network Environment Changes.
3.1 Related Methodologies 75
3.1.5.2 Network Selection using the TF-ANP and the TFT Algorithms for supporting
Medical Services
3.1.5.2.2 Performance Evaluation : During the network selection process the TF-ANP
method is applied initially, in order to estimate the decision weights per service type and patient
health status, considering the ANP network model proposed in [171]. The criteria used include
throughput, delay, jitter, packet loss, service reliability, security and price. Pairwise comparison
decision matrices are created to evaluate the priority vectors of criteria and to form the fuzzy
Supermatrix per service type and patient health status. Subsequently, the fuzzy Weighted
Supermatrix and, finally, the Limit Supermatrix are obtained. Indicatively, when the patient
health status is evaluated as Non-Urgent, for the Sensor Data service the initial, the weighted
and the limit supermatrices are presented in Tables 3.4, 3.5 and 3.6 respectively. For each
health status, the criteria weights per healthcare service obtained by the corresponding Limit
Supermatrix are presented in Figure 3.19. As illustrated, the weights are proportional to the
constraints of each service as well as to the health status of each patient. In particular, the
weight of the price criterion is low for Immediate health status, resulting in a weight value
which is very close to 0. Accordingly, when the health status is evaluated as Non-Urgent, the
medical risk for the patient is very low and the price criterion becomes more important.
The ranking of the networks alternatives is performed using the TFT algorithm using the
weights of the network selection criteria obtained using the TF-ANP method. The linguistic
terms for the criteria attributes are represented by IVTFNs as shown in Table 4.1. Furthermore,
the available-candidate networks in our simulations at the time of network selection per service
as well as their specifications expressed by linguistic terms are depicted in Table 3.14.
The case of having several medical services of different QoS constraints running at the
vehicle site is also addressed. The network selection is performed in a way that satisfies
multiple groups of criteria per vehicle. The vehicles need to select a network which satisfies
76 Mobility Management
Figure 3.19 The TF-ANP Weights per Service and Patient Health Status.
3.1 Related Methodologies 77
Table 3.13 Linguistic Terms and the Corresponding Interval-Valued Trapezoidal Fuzzy Numbers
used for the Criteria Attributes for the Evaluation of the TFT Algorithm in cases where Medical
Services are used.
0,1
0,08
0,06
0,04
0,02
0
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
LTE1 LTE2 LTE3 LTE4 WAVE1 WAVE2 WAVE3 WAVE4 WiMAX1 WiMAX2
the requirements of their healthcare services as presented in Table 4.8 considering the health
status of each patient. To achieve this goal the proposed TFT algorithm is applied for each
vehicle. The available networks are ranked as shown in Figure 3.20.
From the obtained results it is clear that the ranking of the network alternatives is in
accordance with the vehicles expectations. For example, vehicle 1 requiring increased QoS
provisioning selects the LTE 1 network which guarantees the best QoS characteristics and
service reliability. In particular, LTE 1 achieves higher ranking than the other networks, due
to the high values of the QoS characteristics and the service reliability factors bearing higher
importance according to the relative ANP weights for the Live Video service. On the contrary,
vehicle 4 whose prior selection criterion is the throughput of the service, selects the WAVE 1
network which satisfies his requirements considering the Clinical Data service constraints and
the Non-Urgent health status of its patient.
The performance of TFT, was evaluated against the ANST [174] and the FAS [175] methods.
A critical weakness of ANST is that it considers only the throughput criterion, while a weakness
of FAS is that it does not support vehicles with more than one service. Therefore, in the case of
FAS only the most demanding service of the vehicle is considered when more than one service
is required. Specifically, for vehicles 5, 7, 9 and 10 FAS is applied only for the Sensor Data
78 Mobility Management
Table 3.14 The Available Wide Coverage Networks for each Service.
service. Additionally, for vehicle 6 it is applied only for the Live Video service, whereas for
vehicle 8 it is applied only for the Medical Images service.
Table 3.20 presents the network classification performed by the proposed TFT, the ANST
and the FAS algorithms, respectively. From the analysis of the results we conclude that the
proposed method achieves better performance for the entire vehicles. Indicatively, the TFT
results concerning vehicle 2 are better than the ones obtained from both ANST and FAS, due
to the fact that TFT selects the LTE 1 network which offers AG packet loss, while the ANST
selects the LTE 3 and the FAS select the WAVE that offer G and VG packet loss respectively.
Additionally, TFT provides more reliable results for vehicle 10 by taking into account the
criteria importance for the Medical Images, the Sensor Data and the Clinical Data services as
3.1 Related Methodologies 79
Table 3.15 The Simulated Vehicles for the Evaluation of the TFT Algorithm.
Table 3.16 Networks’ Classification in respect of TFT, ANST and FAS Results.
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
ANST
ANST
ANST
ANST
ANST
ANST
ANST
ANST
ANST
ANST
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
FAS
FAS
FAS
FAS
FAS
FAS
FAS
FAS
FAS
FAS
❵❵❵
Networks
❵❵Method
❵❵
LTE 1 1 2 1 1 8 3 3 9 7 4 8 9 2 3 7 2 3 1 2 8 7 3 8 3 4 8 7 2 8 7
LTE 2 3 4 2 3 2 10 5 7 5 2 4 4 4 7 5 3 2 2 4 5 5 1 2 10 3 2 5 4 2 5
LTE 3 2 1 4 2 1 9 1 2 1 3 5 1 1 1 1 1 1 4 1 1 1 2 1 9 1 4 1 1 1 1
LTE 4 6 9 7 6 10 2 7 4 10 6 1 6 8 9 10 5 8 7 6 10 10 6 10 2 8 1 10 6 10 10
WAVE 1 4 6 5 4 9 7 2 8 3 1 3 3 3 6 3 4 5 5 3 9 3 4 7 7 2 6 3 3 7 3
WAVE 2 10 10 10 9 5 6 8 1 4 8 6 8 10 10 4 10 9 10 8 4 4 8 5 6 9 3 4 7 4 4
WAVE 3 9 8 9 10 6 4 10 3 9 10 10 7 9 8 9 9 7 9 10 6 9 10 9 4 10 10 9 10 9 9
WAVE 4 7 7 6 5 4 1 4 6 2 5 7 2 5 4 2 7 6 6 5 2 2 5 4 1 5 5 2 5 5 2
WiMAX 1 5 3 3 8 7 5 9 5 8 7 2 10 6 2 8 6 3 3 9 7 8 7 6 5 6 7 8 8 6 8
WiMAX 2 8 5 8 7 3 8 6 3 6 9 9 5 7 5 6 8 4 8 7 3 6 9 3 8 7 9 6 9 3 6
well as the entire set of criteria, while ANST considers only the throughput criterion and FAS
takes into account only the Sensor Data service.
3.1.5.3 Network Selection using the TF-AANP and the TFT-ACW Algorithms for sup-
porting Medical Services
Live Healthcare Video (LHVideo) [9], Medical Images (MedImages) [10], Health Monitoring
(HMonitoring) [11] and Clinical Data Transmission (CData) [12] services. Furthermore, a
Software Defined Network (SDN) controller provides a centralized control of the entire system.
Table 3.17 Linguistic Terms and the corresponding Interval-Valued Trapezoidal Fuzzy Numbers
used for the Criteria Attributes for the Evaluation of the TFT-ACW Algorithm.
LTE Femto 1
X X LTE Femto 2 AG G AG G VG G G
X LTE Femto 3 G MG G MG G AG P
X LTE Femto 4 AG AG MG AG G G AP
X X X WiMAX Macro G AG G VG AG VG VP
X X X WAVE 1 VG MG VG AG VG AG G
(NAV)
X WAVE 2 MG MG VG VG G G AP
X X X WAVE 3 MG MG G VG MG MG M
X WAVE 4 M M MG VG MG MG MP
X X X LTE Macro MG M G VG VG AG MP
X X X LTE Femto 1 AG AG AG AG AG VG G
Parking Assistance
X X LTE Femto 2 VG AG M VG AG G AG
X LTE Femto 3 G VG MG AG VG AG MG
X LTE Femto 4 MG G M G VG G G
X X X WiMAX Macro G M M AG MG M MP
(PRK)
X X X WAVE 1 M MP MG VG G G VG
X WAVE 2 MG M M AG M M M
X X X WAVE 3 AG G G AG MG MG P
X WAVE 4 G M AG AG MG G G
X X LTE Macro AG AG AG VG VG AG G
X X LTE Femto 1 MP MG VG AG AG VG AP
Conversation Video
X X LTE Femto 2 G G MG VG AG MG M
X LTE Femto 3 AG VG AG G VG G P
X LTE Femto 4 G VG VG AG MG AG MG
X X WiMAX Macro MP M MG G G G MP
X X WAVE 1 G G VG VG G G M
(CV)
X WAVE 2 MG MG AG VG G MG AG
X X WAVE 3 MG MG G AG MG VG MP
X WAVE 4 MP MP MG AG MG G P
X LTE Macro VG VG AG AG VG AG AP
X LTE Femto 1 M G VG VG AG VG G
X LTE Femto 2 G VG MG AG VG G MG
Voice over IP
X LTE Femto 3 MG G G VG G MG MP
X LTE Femto 4 VG AG VG AG VG VG P
(VoIP)
X WiMAX Macro M G MG MG G G M
X WAVE 1 M MG AG AG G G VG
X WAVE 2 MG M VG AG MG M M
X WAVE 3 VG AG VG AG MG VG G
X WAVE 4 G VG G AG MG G AG
X X LTE Macro G MG VG AG VG AG VG
X X LTE Femto 1 VG AG AG VG AG VG P
Buffered Streaming
X X LTE Femto 2 AG VG VG MG MG AG G
X LTE Femto 3 G MG VG G VG G M
X LTE Femto 4 AG AG AG VG G G P
X X WiMAX Macro G AG G VG AG VG VP
X X WAVE 1 VG MG VG AG VG AG G
(BS)
X WAVE 2 MG MG VG VG G G AP
X X WAVE 3 MG MG G VG MG MG M
X WAVE 4 M M MG VG MG MG MP
X X X LTE Macro MG M G VG VG AG MP
X X X LTE Femto 1 AG AG AG AG AG VG G
X X LTE Femto 2 G G M G G VG G
Web Browsing
X LTE Femto 3 AG AG MG AG VG MG M
X LTE Femto 4 VG G G VG MG G MG
X X X WiMAX Macro G VG MG G M VG M
(WB)
X X X WAVE 1 M MP MG VG G G VG
X WAVE 2 MG M M AG M M M
X X X WAVE 3 AG G G AG MG MG AP
X WAVE 4 G M AG AG MG G G
X X X LTE Macro MG MG G VG MG VG MP
X X X MP MG VG AG AG VG AP
Live Healthcare Video
LTE Femto 1
X X LTE Femto 2 G VG AG AG VG G G
X LTE Femto 3 VG AG G AG G MG M
X LTE Femto 4 MG G VG G AG AG G
X X X MP M MG G G G MP
(LHVideo)
WiMAX Macro
X X X WAVE 1 G G VG VG G G M
X WAVE 2 MG MG AG VG G MG AG
X X X WAVE 3 AG AG AG AG VG AG G
X WAVE 4 MP MP MG AG MG G P
X X X LTE Macro M MG AG AG G AG VG
X X X LTE Femto 1 M G VG VG AG VG G
X X LTE Femto 2 AG AG VG AG VG AG P
Medical Images
X VG MG G MG G MG MP
(MedImages)
LTE Femto 3
X LTE Femto 4 G G VG G AG G MP
X X X WiMAX Macro M G MG VG G G M
X X X WAVE 1 VG VG AG AG VG AG AP
X WAVE 2 MG M VG AG MG M M
X X X WAVE 3 VG AG VG AG MG VG G
X WAVE 4 G VG G AG MG G MP
X X X LTE Macro G MG VG AG VG AG VG
X X X LTE Femto 1 VG AG AG VG AG VG P
Health Monitoring
X X LTE Femto 2 AG G G G MG G VG
X VG AG MG AG MG VG G
(HMonitoring)
LTE Femto 3
X LTE Femto 4 G VG VG G G G VG
X X X WiMAX Macro G AG G VG AG VG VP
X X X WAVE 1 VG MG VG AG VG AG G
X WAVE 2 MG MG VG VG G G AP
X X X WAVE 3 MG MG G VG MG MG M
X WAVE 4 M M MG VG MG MG MP
X X X LTE Macro MG M G VG VG AG MP
Clinical Data Transmission
X X X LTE Femto 1 AG AG AG AG AG VG P
X X LTE Femto 2 AG VG MG VG VG G M
X LTE Femto 3 G AG M G G MG M
X LTE Femto 4 VG G G AG VG AG G
X X X WiMAX Macro G MG VG G VG MG AG
X X X WAVE 1 M MP MG VG G G AG
(CData)
X WAVE 2 MG M M AG M M M
X X X WAVE 3 AG G G AG MG MG P
X WAVE 4 G M AG AG MG G G
82 Mobility Management
Table 3.19 The Simulated Vehicles for the Evaluation of the TFT-ACW Algorithm.
Table 3.20 The Classification of the Networks according to TFT-ACW, TFT and FSAW Results.
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
TFT-ACW
FSAW
FSAW
FSAW
FSAW
FSAW
FSAW
FSAW
FSAW
FSAW
FSAW
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
TFT
❵❵❵
Networks
❵❵Method
❵❵
LTE Macro 4 3 3 2 2 2 1 1 1 2 2 1 2 2 1 2 1 1 1 2 2 1 1 2 2 3 4 3 2 3
LTE Femto 1 3 2 1 1 1 1 - - - - - - - - - - - - - - - - - - 3 1 1 2 1 1
LTE Femto 2 2 1 2 3 4 3 - - - - - - - - - - - - - - - - - - - - - - - -
LTE Femto 3 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
LTE Femto 4 - - - - - - - - - 1 4 5 - - - - - - - - - - - - - - - - - -
WiMAX Macro 7 7 8 6 7 7 3 3 2 4 5 3 4 4 4 4 4 4 3 3 3 3 3 3 4 2 3 4 5 5
WAVE 1 5 4 4 4 3 4 2 2 3 3 1 2 3 3 3 1 3 3 2 1 1 2 2 1 1 4 2 1 3 2
WAVE 2 6 6 7 8 8 8 5 5 5 6 7 7 - - - - - - - - - - - - - - - - - -
WAVE 3 1 5 5 5 5 6 4 4 4 7 6 6 1 1 2 3 2 2 4 4 4 4 4 4 5 5 5 5 4 4
WAVE 4 8 8 6 7 6 5 6 6 6 5 3 4 - - - - - - - - - - - - - - - - - -
3.1 Related Methodologies 83
Cloud
...
VM
... Services VM
VM Services
Services
...
SDN
controller
Vehicle 1
Vehicle 2 Vehicle 3
Vehicle 10
LTE LTE
Femto1 Femto2
Vehicle 9 WAVE RSU1
WAVE RSU2
WiMAX
Macro
LTE Macro
Vehicle 4
Vehicle 8
LTE
Femto4
WAVE RSU4
LTE
Femto3 WAVE RSU3
Vehicle 7
Vehicle 6
Vehicle 5
Figure 3.21 The Simulated Topology for the Evaluation of the TF-ANP and the TFT Algorithms.
Three Service Level Agreements (SLAs) are defined. Each SLA determines the available
networks for each service type. SLA1 supports all the available networks while SLA3 supports
only a small subset of the available networks.
Table 4.1 presents the lingustic terms and the corresponding interval-valued trapezoidal
fuzzy numbers used for the criteria attributes of the available networks, while Table 3.18
presents the corresponding specifications per service and SLA of each network, in terms of
throughput, delay, jitter, packet loss ratio, price, security and service reliability.
We consider the case where 10 vehicles with patients are moving inside the network
environment and need to be connected to a network which satisfies the requirements of their
services and at the same time complies with their patient health status, as well as with their
respective SLA agreements. The health status of each patient is evaluated using the Manchester
Triage System (MTS) [173] healthcare classification system, which defines 5 health statuses,
called Non-Urgent, Standard, Urgent, Very-Urgent and Immediate. The Non-Urgent status has
the lower risk about patient’s life, while the Immediate status has the higher one.
3.1.5.3.2 Performance Evaluation : During the network selection process initially the
relative importance ω̃ p̃s of each service is considered with respect to the patient health status.
Figure 3.22 presents the importance of each service per patient health status, as it is obtained
84 Mobility Management
using the TF-AANP method. As it can be observed, the importance of the MED services
depends on the patient health status. Indicatively, when the patient health status becomes
immediate, the MED services obtain higher importance than the DA and the PEnI services.
Accordingly, when the patient health status becomes Non-Urgent, the relative importance of
the services is quite similar. Subsequently, the TF-AANP estimates the decision weights we
per service type and patient health status, using the ANP network model proposed in [171].
The criteria weights per SLA for the DA and the PEnI services are presented in Figures 3.23
and 3.24, respectively. Also, for each possible health status, the criteria weights per healthcare
service for the MED services are presented in Figure 3.25. As illustrated the weights are
proportional to the constraints of each service as well as to the health status of each patient.
In particular, the weight of the price criterion is low for Immediate health status, resulting in
a weight value which is very close to 0. Accordingly, when the health status is evaluated as
Non-Urgent, the medical risk for the patient is very low and the price criterion becomes more
important.
Considering the relative importance ω̃ p̃s of each service and the criteria weights we for the
DA, PEnI and MED services, the final criteria weights are estimated for each vehicle with
respect to the health status of onboard patients, as well as to the SLA of each vehicle (Figure
3.26).
The ranking of the network alternatives is performed using the TFT-ACW algorithm using
the afforementioned criteria weights for each vehicle.
Figure 3.22 The Importance of each Service per Patient Health Status.
Subsequently, the experimental results of the TFT-ACW method are compared with the
ones obtained using the TFT [171] and the FSAW [179] algorithms (Table 3.20). When the
3.1 Related Methodologies 85
TF-AANP weight
0,3
0,25
0,2
0,15
0,1
0,05
0
SLA 1 SLA 2 SLA 3 SLA 1 SLA 2 SLA 3
Navigation Assistance Parking Assistance
Throughput Delay Jitter Packet loss Service Reliability Security Price
Figure 3.23 The TF-AANP Criteria Weights for the DA Services per SLA.
Criteria Weights for Passengers Entertainment and Information services per SLA
0,35
TF-AANP weight
0,3
0,25
0,2
0,15
0,1
0,05
0
SLA 1 SLA 1 SLA 2 SLA 1 SLA 2 SLA 1 SLA 2 SLA 3
VoIP Conversational Video Buffered Streaming Web
Throughput Delay Jitter Packet loss Service Reliability Security Price
Figure 3.24 The TF-AANP Criteria Weights for the PeNI Services per SLA.
patient health status is Non-Urgent (vehicles 2 and 5) or Standard (vehicles 3 and 8) the results
of TFT-ACW and TFT are similar, due to the similar relative importance considered by the
TFT-ACW for each service. However, when the patient health status gets worse, the TFT-ACW
assigns higher importance to the MED services and selects the most appropriate network to
satisfy their strict constraints. Indicatively, in the case of vehicle 1, TFT-ACW selects the
WAVE 3 network, which provides VG for service reliability, as well as AG for throughput,
delay, jitter, packet loss and security, for the LHVideo medical service. On the contrary, the
results of both TFT and FSAW are negatively affected by the existence of non medical services
in the vehicle 1 ignoring the immediate health status of the patient. Specifically, TFT selects
the LTE Femto 2 network, which provides worse specifications for the LHVideo service (e.g. G
for throughput and VG for delay), while FSAW selects the LTE Femto 1 network, which also
provides worse specifications for the aforementioned medical service (e.g. MP for throughput
and MG for delay).
86 Mobility Management
Figure 3.25 The TF-AANP Criteria Weights for the MED Services per SLA and Patient Health
Status.
0,25
0,2
0,15
0,1
0,05
0
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
Throughput Delay Jitter Packet loss Service Reliability Security Price
Femtocells, respectively. Both of these thresholds are evaluated, considering the Reference
Signal Received Power (RSRP) threshold defined in LTE [14], as well as network load infor-
mation. When RSRP drops below the corresponding γth threshold, a Time-to-Trigger (TTT)
timer is activated, which is initialized to a certain value T, considering parameters such as the
cell transmission power, the distances between the available cells, the path loss, the carrier
frequency, the network traffic load and the user velocity. During the countdown, if RSRP
returns to values above the corresponding γth threshold, the timer is deactivated. Otherwise,
when the timer becomes equal to zero, the handover is initiated and the user is transferred to
the network with the highest RSRP.
The Context Aware Load Balancing (HO-CALB) [185] is another HO initiation scheme. It
considers the scenario where both WiFi and WiMAX networks coexist. A network load aware
algorithm distributes the traffic load to the WiFi and the WiMAX networks considering the
available bandwidth of each network. Each user has a set of services with strict QoS constraints.
If the observed QoS drops below a threshold, a handover policy instructs the users to perform a
handover.
Several schemes implement only the network selection subprocess. These algorithms aim
at the selection of the most appropriate access network for each vehicular user among a set
of network alternatives. Similar to the HO initiation algorithms, several parameters can be
considered, including Quality of Service (QoS) or Quality of Experience (QoE) factors, user
preferences and operators’ policies. In these cases, the HO initiation and the HO execution
must be performed using third party algorithms.
The Mobility Aware Handover in Ultra Dense Vehicular Environment (MAH-UDVE) [186]
scheme considers the direction of a vehicle’s mobility to select the most appropriate network.
The vehicle movement is statistically analyzed in order to create a set of possible next cells.
Subsequently, four different approaches can be applied to select the most appropriate cell from
the aforementioned set. The first approach selects the nearest cell, while the second approach
selects the next cell that exists in the same street in the same direction with the direction of
vehicle movement. The third approach selects the next cell that exists in the same street, but in
the opposite direction of the vehicle’s movement. Finally, the fourth approach selects the cell
with the lowest load.
The Intelligent Network Selection (INS) [187] scheme implements the network selection
process for supporting audio and video streaming vehicular services in heterogeneous network
access environments. A utility function which considers the SNR, the channel capacity and the
connection life time parameters is used. For the estimation of the connection lifetime parameter
the vehicle’s velocity and the cell radius are also considered. The candidate network which
obtains the higher score from the proposed utility function is selected.
3.2 Mobility Management Schemes for 5G Wireless Networks 89
The Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) [188][189]
algorithm performs the selection of the best network access technology by considering contra-
dictory selection criteria. The main concept of the TOPSIS algorithm is that the best alternative
network should have the shortest distance from the positive ideal solution and the longest
distance from the negative ideal solution. The criteria that could be considered include the
network throughput, the delay, the jitter, the packet loss ratio, the network usage price, the
security and the service reliability. Also, the importance of these criteria can be estimated either
using the Analytic Hierarchy Process (AHP) [190] or the Analytic Network Process (ANP)
[191] method.
The Simple Additive Weighting (SAW) [188] considers multiple criteria to perform the
network selection. Firstly, the criteria weights are estimated using a method such as AHP or
ANP. Subsequently, for each network alternative, the criteria values are normalized. Then, each
value is multiplied with the corresponding weight in order the weighted criteria values to be
calculated. Finally, the weighted criteria values are summarized in order to estimate the overall
rating of the network alternative. The network with the highest rate is finally selected.
The Multiplicative Exponent Weighting (MEW) [188] method is similar to SAW. However,
in MEW the weighted criteria values are multiplied with each other to estimate the overall
rating of the network alternative.
The Gray Relational Analysis (GRA) [192] analyzes the relationship grade between the
network alternatives and the ideal solution to perform the network selection. The ideal solution
is calculated first. Subsequently, the similarity of each alternative network with the ideal
solution is estimated using the Grey Relational Coefficient (GRC) [193]. The network with the
highest similarity to the ideal solution is selected.
The concept of the Distance to Ideal Alternative (DIA)[188] algorithm is similar to the
concept of TOPSIS. Firstly, DIA estimates the positive and the negative ideal solutions. Then,
the Manhattan distance [194] of each network from the positive and from the negative solution
is calculated. The network that has the shortest distance from the positive ideal solution and the
longest distance from the negative ideal solution is selected.
The Weighted Product Method (WPM) [195] for each alternative network raises the nor-
malized criteria values to the corresponding criteria weights values. Subsequently, the method
calculates the product of the weighted values. The network with the highest product is finally
selected.
In Quantum-inspired Immune Clonal Algorithm for Network Selection (QICA-NS) [196]
two utility functions are used for network selection, one for the user and one for the network.
The user utility function considers user related parameters, such as the user preferences, his
moving velocity and his monetary budget. Accordingly, the network utility function considers
90 Mobility Management
network related parameters, such as the network load, the price and the network QoS which is
estimated using TOPSIS.
Sometimes many users have similar trajectories as they move. This situation becomes
more frequent in the case of vehicular networks, since each vehicle usually serves multiple
passengers. Consequently, many users with similar positions and requirements need to perform
a network selection, usually considering the same network alternatives. In such cases, increased
computational overhead arises from the decision process, since this process runs independently
for each user. To address this issue, algorithms for performing group handover have been
proposed in the research literature. An algorithm of this category groups users with similar
trajectories in order to perform the network selection once for the entire user group, thus
releasing usable system resources.
The Group Vertical Handover using Time Window (GVHO-TW) [197] distributes the MTs
HO requests in a time sequence. Thus, each MT performs network selection using a utility
function when its turn comes, considering the decision of all the previous MTs in order to
select the most appropriate network not only to satisfy its requirements, but also to achieve a
satisfactory load distribution to the available access networks.
Although GVHO-TW avoids the calculation of simultaneous decisions of multiple MTs, ad-
ditional handover delays occur since each MT waits for its turn in order to perform the network
selection. To address this issue, the Group Vertical Handover with Probability Distribution
(GVHO-PD) [197] scheme uses a probability distribution vector instead of a time window. This
vector contains a value for each available network, which determines the probability of the
network to be selected from the MTs, considering its performance. Subsequently, each MT
generates a random MT probability value and compares it with the values of the aforementioned
probability vector. The network with the closest probability to the random MT probability
value is selected. Thus, the MTs are distributed to the available networks.
Although the GVHO-PD scheme distributes probabilistic the workload to the available
networks, it does not optimize the entire system performance. To address this issue the Network
Assisted Group Vertical Handover (NA-GVHO) [197] scheme does not use a probability vector,
but it collects information from both the networks and the MTs. Then, each MT performs
the network selection using a Multi Attribute Decision Making (MADM) algorithm which
considers both the MT’s requirements and the current workload of each network, in order to
distribute the workload to the available networks.
In general, there is a rate of uncertainty in characterizing performance measurements as well
as rates of influence of performance metrics. Therefore, fuzzy MADM (FMADM) methods
expressing uncertain quantities by fuzzy numbers have received the interest of many researchers
in decision theory. In particular several FMADM network selection methods are suggested
3.2 Mobility Management Schemes for 5G Wireless Networks 91
utilizing linguistic variables, triangular fuzzy numbers, trapezoidal fuzzy numbers etc. to model
network attributes and their respective weights. The use of fuzzy logic for network selection
requires the definition of logic rules from specialists with thorough knowledge of the behavior
of the available access networks in various conditions.
The Trapezoidal Fuzzy TOPSIS (TFT) [171] deals with the network selection process. It
is a fuzzy version of the TOPSIS algorithm. TFT uses linguistic values for evaluating the
criteria attributes, while each linguistic value is represented by an Interval Valued Trapezoidal
Fuzzy Number (IVTFN) [198]. The criteria considered for the candidate networks ranking
include throughput, delay, jitter, packet loss, price, security and service reliability. The Analytic
Network Process (ANP) [191] is used for the estimation of the criteria weights indicating the
importance of each criterion. The network that accomplishes the higher TFT rank is selected.
The Trapezoidal Fuzzy TOPSIS with Adaptive Criteria Weights (TFT-ACW) [66] is an
improved version of the TFT algorithm. It deals with the satisfaction of the constraints of
multiple vehicular services, in cases where MED services coexist with other service types
in the vehicular environment. Specifically, DA services, PEnI services and MED services
are considered that provided to vehicular users. The presence of MED services affects the
importance of other services in situations where patients with immediate health status exist
within the vehicle. The criteria used for network evaluation include throughput, delay, jitter,
packet loss, service reliability, security and price. The Trapezoidal Fuzzy Adaptive Analytic
Network Process (TF-AANP) is also proposed for the estimation of the services importance as
well as of the corresponding criteria weights, which are adapted considering the health status
of onboard patients and the Service Layer Agreement (SLA) of each vehicle.
In Intelligent Access Selection using Fuzzy Neural Network (IAS-FNN) [199] the concepts
of fuzzy logic, neural networks and utility functions are combined to perform network selection.
The proposed method makes use of a fuzzy neural network which obtains network, user
and terminal related input criteria and evaluates the performance of each access network.
The attributes of the criteria are defined through utility functions and processed through the
fuzzification, interference and defuzzification layers of the neural network.
It should also be noted that some schemes implement only the HO execution. These
algorithms determine the signaling that should be performed in order the HO to be successfully
completed. A critical parameter that should be satisfied is the HO delay which is directly
affected by the aforementioned signaling. Also, in the case of such algorithms, the HO
initiation and the network selection must be performed using third party algorithms.
In Bulk Fast PMIPv6-based Network Mobilty (BFP-NEMO) [200], one or more RSUs are
controlled by a Mobility Management Entity (MME) [201], which is called Mobile Access
Gateway (MAG). The vehicles that are inserted in a MAG domain are grouped. Then, the
92 Mobility Management
corresponding MAG pre-establishes a bidirectional tunnel between the vehicles’ group and
the candidate neighboring MAGs. Each neighboring MAG caches the data for each vehicle of
the aforementioned group. Subsequently, when a vehicle of the group is connected to the next
RSU, the corresponding MAG forwards the cached data to the vehicle.
The Enhanced Proxy Fast-handover Mobile IPv6 for Vehicular Networks (ePFMIPv6)
[202] scheme implements a proactive Layer 3 handover mechanism. The concept of Mobile
Access Gateway (MAG) is considered in this scheme as well. The issue of the simple PFMIPv6
protocol, where the next MAG waits to receive a Handover Acknowledge (HAck) [203] message
from the serving MAG (although the vehicle has arrived to it) to start the data forwarding
to vehicles, is resolved. Specifically, when the serving MAG services the vehicle, it also
establishes a tunnel with each neighboring MAG which is considered as a candidate next
MAG for the vehicle according to its RSS. When the vehicle arrives to its next MAG, its data
are immediately forwarded to it through the aforementioned tunnel, without waiting for the
HAcK message. Finally, when the HAcK message is received by the next MAG, the tunnel is
abolished and the data which follows are forwarded to the vehicle by the next MAG without the
tunnel. In this case, the "next MAG" which is now considered as the serving MAG establishes
a tunnel with each neighboring candidate next MAG and so on.
Schemes that implement more than one mobility management subprocesses have also been
proposed in the research literature. Many of these schemes implement both HO initiation and
network selection.
The Predictive Handover Mechanism for Video Streaming in Cloud Based Urban VANET
(PHMVV) [204] scheme defines a list of available networks is created for each vehicle con-
sidering its velocity, its distance from each network, as well as its moving direction. The
assumption that the vehicle knows its geographical position, as well as the positions of the
available networks, is made. Each vehicle monitors the link quality of the candidate networks,
considering their Signal to Noise Ratio (SNR), Bit Error Rate (BER) and Packet Error Rate
(PER) parameters. If a network alternative has higher link quality than the vehicle’s current
network, the vehicle performs a handover to this alternative.
Another scheme that implements both HO initiation and network selection is the Mobility
Aware Handover for Smart Cities Environment (MAH-SCE) [205]. MAH-SCE uses traffic
information that is gathered by sensors that are deployed in a Smart City [206] environment, to
perform the HO initiation and the network selection. The scheme decides when to initiate a
handover as a function of the aforementioned sensor data about the road traffic, as well as the
perceived SNR. Subsequently, the vehicles with velocity up to 16 m/s consider both Macrocells
and Femtocells as alternatives for performing a handover to them. In this case, if Femtocells
exist inside the vehicle’s area, the scheme selects the Femtocell which has the higher estimated
3.2 Mobility Management Schemes for 5G Wireless Networks 93
time that the vehicle will remain connected without a need to perform a handover to another
cell. On the contrary, if the vehicle velocity is higher than 16 m/s only the Macrocells are
considered as alternatives and the one that maximizes the time that the vehicle will remain
connected to it, is selected.
The Speed-based Vertical Handover (S-VHO) [207] scheme also implements the HO
initiation and the network selection subprocesses of the HO management. During the HO
initiation, the algorithm considers the vehicle’s velocity, the ingress time of the vehicle to the
current cell and the geographical position of the vehicle (using its GPS) to calculate the cell
crossing time. Subsequently, considering both the estimated cell crossing time and the observed
throughput, the algorithm decides when a HO must be initiated. Afterwards, the alternative
network that maximizes the throughput is selected and, then, the algorithm is deactivated for
a time period T in order the ping-pong effect to be reduced. Finally, when the algorithm is
reactivated it estimates when the next HO must be initiated.
The Media Independent Handover using Predictive Geographical Information (MIH-PGI)
[208] defines a MIH-oriented HO mechanism for ensuring the continuity of the vehicular
services and the minimal handover latency. During the HO initiation the vehicle estimates the
appropriate time to perform a handover by predicting the behavior of its current network link
quality based on the geographic information received from the Media Independent Information
Services (MIIS) component of the MIH protocol, about the network access point position
as well as the velocity of the vehicle. Subsequently, the vehicle lists the available candidate
networks, while the next network is selected considering the estimated time duration that the
vehicle will remail inside the coverage area of each candidate network. Thus, the network
which maximizes this time is selected in order to minimize the number of future handovers.
The Cognitive Vertical Handover for Vehicular Communication (CVHO) [209] scheme
considers multiple criteria to perform mobility management in a cognitive radio vehicular
environment. The handover initiation considers parameters such as the RSS, the available
bandwidth, the delay, the network load, the usage cost, the vehicle velocity and the service
type, to evaluate the necessity of performing a VHO. Subsequently, the network selection is
performed using the Analytic Hierarchy Process (AHP) [190] method using the aforementioned
criteria. An Artificial Neural Network (ANN) [210] is used to train the scheme to adapt its
handover threshold in order to take the reflex action to avoid either unnecessary handovers
or link-down situations. The ANN is trained considering the success or the failure of all the
previous handover initiation and network selection decisions.
The Auction Approach for Mobility Prediction (AAMP) [211] scheme defines a group
based HO management mechanism which aims to maximize system throughput, while at
the same time to minimize the network load when multiple vehicles perform simultaneous
94 Mobility Management
handovers. It assumed that each vehicle is equipped with a GPS device. When a vehicle reaches
the edge of its current network’s coverage, it triggers a VHO, while at the same time it reports
its position to a Vertical Handover Decision (VHD) controller, which could be considered as an
SDN controller for HO functionalities implementation. When a VHD controller receives a HO
request from a vehicle, it creates a list with the candidate networks for the vehicle considering
the vehicle’s position. Thus, considering the vehicle’s velocity and position, the VHD controller
predicts the entrance time and the exit time of the vehicle from each candidate network. When
multiple vehicles of similar positions trigger a handover, the VHD controller groups them.
Subsequently, an auction based [212] algorithm selects the most appropriate network for the
aforementioned group of vehicles in order the total throughput of the group to be maximized.
Thus, the network selection is performed once for the entire group of vehicles and not once per
vehicle, thus minimizing the network load occurring from the decision process.
The Vertical Handover for Vehicular Cloud Computing Systems (VHO-VCC) [67] scheme
takes into account the vehicle’s velocity as well as its current connection type. A two step HO
algorithm that reduces operation costs and optimizes mobility management is applied. Initially,
a HO initiation process evaluates the necessity to perform handover using a Mamdani inference
system [213] which evaluates the user satisfaction grade. When the user satisfaction drops
below a predefined threshold, which depends on the vehicle’s SLA, the network selection is
performed using TFT in order to select the most appropriate network alternative considering
both the vehicular service requirements and the operators’ policies.
The Geolocation-based Multi ACcess network Handover algorithm for vehicUlar envi-
ronments (GEO-MACHU) [214] algorithm implements three tasks, namely the Networking,
the Neighborhooding and the Decision Making task. The Networking task performs the HO
initiation considering context-aware information about the current network provided by the
Media Independent Event Service (MIES) and by the Media Independent Command Service
(MICS) [215] of the IEEE 802.21 MIH protocol (e.g. information about link-down events).
The Neighborhooding task collects information about the available networks in the vehicle’s
location using the Media Independent Information Service (MIIS) [216] of the MIH protocol
. The gathered information is considered by the Decision Making task, which performs the
network selection using the TOPSIS algorithm.
The Hybrid Communication Approach (HCA) [217] scheme defines two network interface
types, namely the IEEE 802.11p network access technology which is considered as the primary
interface, while the 3GPP LTE is considered as the secondary. By default, a vehicle is connected
to the primary interface. A QoS aware HO initiation algorithm is applied, which instructs the
vehicle to perform handover to the secondary interface when the observed packet loss of the
provided services exceeds a maximum acceptable threshold. Thereafter, a timer is activated
3.2 Mobility Management Schemes for 5G Wireless Networks 95
specifying the time interval that the user remains connected to a secondary interface. When
the timer expires and the estimated packet loss ratio of the primary interface is lower than the
maximum acceptable threshold, the vehicle performs a handover back to the primary interface.
The Velocity Aware Handover (VAH) [218] HO management scheme defines two network
tiers, namely the tier-1 consisting of Macrocells and the tier-2 consisting of Femtocells. Four
vertical handover strategies are considered, namely the Best Connected (BC), the Femto
Skipping (FS), the Femto Disregard (FD) and the Macro Skipping (MS). The BC strategy is
applied to static users and to users with Low mobility, while both Macrocells and Femtocells
are considered as alternatives. The user connects to a Macrocell if P1 · B1 · R−η 1 > P2 · B2 · R2
−η
selection process, a location tracking mechanism is used in order to estimate the vehicle’s
moving direction. Thus, the most probable next RSU that the vehicle will pass through its
coverage area is selected. Subsequently, the scheme proactively performs a Layer 3 handover
where the appropriate signaling is exchanged in order to start the forwarding of the next data
packets. Afterwards, a Layer 2 handover is performed where the vehicle is attached to the
selected RSU.
The Hidden Markov Model - Kalman Filter (HMM-KF) [222] scheme deals with the
network selection and the HO execution subprocesses of the mobility management. The
mobility of the vehicles is modeled by tracking their movement considering information from
their GPS devices. Also, each network maintains a Hidden Markov Model (HMM) indicating
the possibilities that each vehicle connected to the network has to handover to each neighboring
network, which are estimated using the Kalman Filter (KF). Subsequently, considering the
HMM each vehicle selects its next network. Afterwards, the appropriate signaling is exchanged
between the vehicle and the selected network in order the handover to be performed. Finally,
the HMM probabilities of the previous and the new vehicle’s network are updated.
The QoS Aware Handover for Session Initiation Protocol services in Vehicular Networks
(QAH-SIP) [223] scheme is implemented in the Application Layer of the OSI stack. QAH-SIP
defines a HO execution process for supporting vehicular SIP services. Initially, the vehicle
performs the network selection considering QoS aware information (such as the maximum
acceptable delay and the packet loss, as well as the minimum required RSS) required to satisfy
the SIP service constraints. Then, it performs a pre-registration to the selected RSU where the
vehicle also sends information about its position and velocity. Then, the selected RSU reserves
the required resources in order to be available to serve the incoming vehicle. Subsequently, the
SIP signaling process [224] is performed in order the HO to be completed.
It should also be noted that some schemes implement the entire mobility management
subprocesses, including the HO initiation, network selection and HO execution.
The Vertical Handover for VANET using Multiple Parameters (VHOMP) [225] algorithm
considers multiple criteria to perform the mobility management. During the HO initiation the
Received Signal Strength (RSS) parameter is turned into account. When it becomes lower than
a predefined threshold the HO is initiated. Subsequently, the network selection is performed
considering the network bandwidth and the vehicle’s velocity. Finally, the HO is executed
using the appropriate signaling in order the vehicle to be connected to the selected network.
The Navigation Assisted Seamless Handover (NASH) [226] uses a HO initiation algorithm
which considers the vehicle velocity, as well as RSRP information, to evaluate the necessity to
perform a handover. The navigation information about the vehicle is used in order the most
probable trajectory to be estimated, while the next cell that exists in this trajectory is selected.
3.2 Mobility Management Schemes for 5G Wireless Networks 97
Finally, considering an LTE-A network infrastructure the scheme proposes the appropriate
signaling that should be performed between the current RSU/BS, the next RSU/BS the MME
and the SGW, in order seamless handover to be operated ensuring the continuity of the vehicle’s
services.
The Robust Handover in Vehicular Networks (RHVN) [227] scheme defines that in Layer 2
the HO initiation is performed when the RSS drops below a predefined threshold. Subsequently,
the vehicle selects the RSU with the higher RSS and performs the HO execution. In Layer 3
the handover occurs when a Layer 2 handover is completed. In this case, the vehicle uses its
Original Care of Address (OCoA) [228] for fast IP connectivity and then its data are forwarded
from its new RSU by applying an improved version of MIPv6 protocol, which considers the
permanent vehicle’s OCoA.
The Intelligent Handover for Vehicular Networks (IHVN) [229] scheme aims at the sat-
isfaction of the real-time services’ constraints during the vehicle’s movement across differ-
ent networks. The FMIPv6 mobility management protocol is considered, while the use of
802.21 MIH ensures the interoperability between different network access technologies. Dur-
ing the HO initiation process, when the link quality drops below a predefined threshold, a
MIH_Link_Going_Down message is transmitted to the vehicle. This message triggers the
vehicle to perform a VHO. Subsequently, a list of existing nearby networks is obtained, while
link quality information (e.g. bandwidth) is obtained from each candidate network using
MIH_MN/N2N_HO_Candidate_Query messages. Then, the link quality parameter is consid-
ered in order the network with the higher link quality to be selected. Finally, during the HO
execution the appropriate signaling is exchanged between the vehicle and the selected network,
while the HO is completed using a MIH_Link_U p message indicating that now the vehicle has
been connected to the selected network.
In the VANET Backup Communication (VANBA) [230] scheme, the concept of the Com-
munication Mediator (CM) is introduced. A CM is a vehicle that provides access to vehicular
services through V2V communication. However, in a 5G-VCC systems, the CM could also be
a RSU/BS or a pedestrian, that provides access to the services through V2I or V2P communi-
cation, respectively. Two processes for accomplishing the mobility management are defined,
namely the Link Monitoring and the HO process. The Link Monitoring process, implemented
at OSI Layers 6 and 7, monitors which CMs in the vehicle’s area are available for providing
access to vehicular applications (e.g. considering the coordinates of their positions). The HO
process, implemented at OSI Layer 3, monitors the IP reachability to the services’ providers.
When the reachability is lost or the link quality drops below a predefined QoS threshold, a
handover should be performed. Subsequently, the candidate CMs are sorted using a utility
function where the bandwidth and the monetary cost parameters are considered. Then, the first
98 Mobility Management
CM in the sorted list is selected and the HO is executed. If the HO execution fails (e.g. due to
topology changes) then the next CM in the aforementioned list is selected and so on.
In Vehicular Fast Handover for Mobile IPv6 (VFMIPv6) [231] when the vehicle approaches
the boundary of its current RSU (considering the RSS), it sends to the current RSU a request
in order handover to be initiated. Along with the request, Handover Assist Information (HAI)
is also sent from the vehicle to its current RSU, containing information about the candidate
RSUs for the vehicle, considering the RSS of each RSU. The current RSU implements a utility
function based mechanism where the HAI information is considered, in order to select the most
appropriate next RSU for the vehicle, which is the one that minimizes the handover latency,
the packet loss that occurs during the handover and the total communication overhead that
is observed due to the handover latency and the corresponding packet loss. Thereafter, the
appropriate signaling is exchanged between the current RSU, the selected next RSU, the Mobile
IPv6 (MIP) Home Agent (HA) [232][233] and the MIP Correspondent Node (CN) [232] in
order the vehicle to be connected to the selected next RSU. Table 3.21 summarizes the main
functionalities of the discussed schemes.
3.2 Mobility Management Schemes for 5G Wireless Networks 99
Message Algorithm
Control Entities
Exchange Models Categories
Neural Network
based
MIH based
Probabilistic
Group based
Auction based
Considering the entities that control the mobility management the entire process could be
categorized as Vehicle Controlled or as Network Controlled. Furthermore, some solutions
combine both Vehicle and Network control. In these cases, the mobility management process
is called as Hybrid Controlled. The following paragraphs describe the available solutions,
considering the aforementioned control entities.
Considering the underling network architecture, the mobility management task can be im-
plemented from a Mobility Management Entity (MME) or an SDN controller installed in a
Cloud or a Fog infrastructure. Alternatively, a Mobile Access Gateway (MAG) could be used,
namely a BS or a RSU enhanced with mobility management capabilities. In general, the entity
that performs the mobility management collects the necessary information for evaluating the
necessity to perform a HO initiation. Also, after each HO initiation this entity selects the most
appropriate network for the vehicle, using the methods that each mobility scheme implements
for the network selection process. Subsequently, it orchestrates the HO execution through the
appropriate signaling in order the vehicle to be connected to the selected network.
Compared to the vehicle controlled mobility management, the network controlled solutions
succeed better performance since their functionalities are implemented to the network equip-
ment. As a result, the mobility management workload is minimized for the vehicles, while at
the same time their energy requirements are decreased.
management workload that the vehicles OBUs undertake, while at the same time their energy
requirements are decreased. Furthermore, in cases where an SDN controller is used, the results
of these scheme can be improved since the controller supervises the manipulation of the entire
system. Finally, the BFP-NEMO [200] scheme implements only network controlled mobility
management.
Message exchange models determine how the entities communicate each other in a VCC
architecture. Two models are available in VCC, the host centric and the information centric.
3.2.2.2.1 Information Centric : The information centric model proposes messages’ de-
livery considering the content semantics. The information centric model is also called as
publish-subscribe model. Considering its operating principles, two entities interact each other,
the publisher and the subscriber. Firstly, the publisher publishes the set of items. Then, the sub-
scriber expresses its interests for items through subscriptions. When an item that is contained
into the subscriber’s interests become available to the publisher, it is being sent to the subscriber.
The use of the information centric model facilitates the vehicles of obtaining easily the required
information, due to the fact that vehicles are usually interested for the information itself, ignor-
ing its origin. The authors of [42] strengthen this opinion by describing two examples. In the
first one, an autonomous vehicle travels in high speed on a highway. The vehicle must obtain
contiguous information of surrounding vehicles that exist in short distance, in order to provide
a safe course to its passengers. Using the information centric model, the vehicle regularly
expresses its interests to the surrounding vehicles. Then, it periodically receives information
about their position, speed and direction. Correspondingly, in the second example, the case of
an accident ahead is considered. More specifically, the vehicle expresses the interest for retriev-
ing information about the accident to the surrounding vehicles and RSUs. Vehicles or RSUs
that become nearest to the accident provide the required information to the vehicle. Then, the
vehicle’s driver is immediately alerted in order to be able to manipulate efficiently the situation
(e.g. by selecting an alternative route to avoid the accident area and save time). Furthermore, in
[43] the authors show that the information centric model enables the development of routing
methods with enhanced scalability characteristics. More specifically, they demonstrate that this
model provides many advantages to the routing functionality, due to the fact that it enables
the network architecture manipulation in terms of “information connectivity” considering the
context semantics, instead of considering the physical nodes connectivity.
3.2.2.2.2 Host Centric : The host centric model is also called as request-response or client-
server model [234–236], where a client machine requests to receive items offered from a host
(or server) machine. Specifically, when a client machine needs an item that belongs to this set, it
sends a request to the host machine. The host machine receives the clients request, retrieves the
required item and sends it to the client. The communication is performed using the IP protocol,
while in large scale network architectures, thousand connections could exist. Furthermore, as
described in [237], the use of the term “host centric” was started when the “information-centric”
104 Mobility Management
term was introduced. Specifically, since the “information centric” has been proposed as an
alternative message exchange model, the “host centric” term is used to represent the existing
model for messages exchange in current network infrastructures, in order the fundamental
difference between Information Centric Networking (ICN) [238] and Host Centric Networking
(HCN) [239] to be clearly identified. Specifically, in order to explain the difference between
the two models, the authors of [240] refer that in the ICN model the information itself obtains a
specific identifier (or address), while in the traditional HCN model each host machine has an IP
address assigned. Indicatively, regarding the host centric model operating principle, the authors
of [241] have proposed the implementation of a Service Oriented Architecture (SOA) [242] for
providing both information and entertainment services for vehicular clients. Specifically, each
client interacts with the implemented web services to obtain access to real time information
(e.g. about road traffic conditions) as well as to multimedia content.
Table 3.23 The Applicability of each Scheme to the Available Message Exchange Models.
Several HO initiation and network selection approaches have been described considering
network and user related parameters and can be classified into the following categories:
3.2.2.3.1 Context Aware : The decision is based on signal quality parameters. In this case,
RSS, SNR or SINR are usually considered in order the signal quality of each network to be
evaluated. Context aware HO initiation is performed when the signal quality of the current
network drops below a threshold, while in case of context aware network selection the network
with the best signal quality is selected. It should be noted that although several algorithms
consider the signal quality to determine the set of network alternatives, in this category belong
only the algorithms that make the final decision (e.g. the network ranking) considering such
information.
3.2.2.3.2 Cost Function based : The cost of each network is estimated using a function.
In the case of HO initiation, if the cost of the current network exceeds a maximum acceptable
threshold, HO is initiated. Accordingly, in the case of a cost function network selection
algorithm the cost of each candidate network is calculated and the network with the lowest
cost is selected. The cost should refer to monetary cost, energy consumption aware cost,
bandwidth etc. Each HO initiation or network selection algorithm could belong to more than
one categories.
3.2.2.3.3 Multi Attribute Decision Making (MADM) : The decision is made considering
multiple and conflicting network and user criteria. In the case of HO initiation, the current
network is ranked, while HO is initiated when this rank drops below a predefined threshold.
Also, in the case of a MADM network selection algorithm the entire network alternatives are
ranked and the network with the higher rank is selected.
3.2.2.3.4 User Centric : The decision considers user preferences, which could be refer
to parameters related to QoS, QoE, monetary costs and so on. In case of HO initiation the
user satisfaction is evaluated considering his preferences. If the user satisfaction grade drops
below a threshold the HO is initiated. Accordingly, in case of user centric network selection the
utility of each network alternative is evaluated considering the user preferences and the network
with the highest utility is selected. As it could be observed, this category can be considered as
similar to the cost function based one. Specifically, the user centric schemes estimate the utility
of each network alternative which is a positive parameter, while the cost based ones estimate
3.2 Mobility Management Schemes for 5G Wireless Networks 107
the cost of each alternative which is a negative parameter. Also, the schemes of this category
are usually combined with MADM algorithms.
3.2.2.3.5 Fuzzy Logic based : They use fuzzy logic in order the decision uncertainty to be
addressed, considering the related criteria to perform HO initiation or network selection. A
fuzzy number is represented by a set of real values representing an uncertain quantity and a
convex normalized continuous function which estimates the degree of membership for each
value in the subset. Triangular, trapezoidal or pentagonal fuzzy numbers are frequently used
to represent uncertain information. These algorithms are usually combined with MADM
algorithms, while the HO initiation and the network selection are performed according to the
MADM based networks ranking.
3.2.2.3.7 MIH based : The algorithms of this category apply the operating principles of
Media Indepentent Handover (MIH) []. In some cases, they are combined with algorithms of
other categories, including context aware and MADM.
3.2.2.3.8 Probabilistic : They consider several probabilities to perform the mobility man-
agement, including probabilities about user trajectory or velocity. Indicatively, considering the
most probable user trajectory, the distribution of the signal strength of each network can be
predicted. Thus, in case of HO initiation the most appropriate time to perform a HO can be
estimated, or in case of network selection the most appropriate network for the user is selected,
considering the aforementioned prediction.
3.2.2.3.9 Group based : They consider the fact that sometimes many users with similar
trajectories and requirements need to perform a HO simultaneously. Thus, increased computa-
tion requirements for performing the mobility management arise. In this case, the algorithms
of this category group users with similar trajectories and perform the mobility management
once for the entire group, decreasing the computational requirements of the entire process.
3.2.2.3.10 Auction based : The algorithms of this category perform the network selection
using auctions. Each user makes an offer (e.g. according to his Service Level Agreement).
Subsequently, users with higher offers obtain connectivity to better networks.
108 Mobility Management
Scheme Cost Function User MADM Fuzzy Logic Neural Net- Context MIH Probabilistic Group Auction
based Centric based work based aware based based based
VHOMP [225] X X
SSHVeT [180] X X
PHMVV [204] X X
MAH-SCE [205] X X
NASH [226] X X
MAH-UDVE [186] X
RHVN [227] X
S-VHO [207] X
BFP-NEMO [200] X
IINH [182] X
IHVN [229] X
VANBA [230] X
iMAG-IDH [221] X
VFMIPv6 [231] X X
MIH-PGI [208] X X
CVHO [209] X X X
DPVHO [183] X X X
GVHO-TW [197] X X
GVHO-PD [197] X X
NA-GVHO [197] X X
ePFMIPv6 [202] X
AAMP [211] X X
INS [187] X
HMM-KF [222] X
QAH-SIP [223] X X
CAHP [184] X
HO-CALB [185] X
TOPSIS [188][189] X
TFT [171] X X
TFT-ACW [66] X X
VHO-VCC [67] X X X X
GEO-MACHU [214] X X X
SAW [188] X
MEW [188] X
GRA [192] X
DIA [188] X
WPM [195] X
QICA-NS [196] X X X
IAS-FNN [199] X X X
FLB-VHO [219] X X
FMVHO [220] X X
HCA [217] X
VAH [218] X
110 Mobility Management
The applicability and the performance of each mobility management scheme is affected of
factors related to the implemented 5G-VCC architecture. Table 3.25 presents the factors that
each 5G-VCC architecture satisfies, while table 3.26 presents the applicability of the discussed
mobility management schemes to each 5G-VCC architecture.
The applicability of each scheme depends on the existence of a backbone network infras-
tructure. Specifically, the VC architecture as already mentioned has no backbone network
infrastructure. It determines a complete ad-hoc topology, which consists of vehicles construct-
ing the Cloud. Thus, only the VANBA [230] and the HCA [217] schemes can be applied to
VC architectures, since in these cases the mobility management concerns the connectivity
between the participating vehicles instead of the interaction between vehicles and a network
infrastructure. This fact reinforces the need for further research since in some cases the VC
architectures could support disaster management services [245] where the availability of access
network infrastructures could be decreased. On the other hand, in the VuC, VuF, SDN-V and
HVA architectures, a communication infrastructure providing access to cloud/fog resources,
is implemented. Hence, the entire schemes can be applied to VuC, VuF, SDN-V and HVA
architectures, since they have been designed to manipulate the user mobility in relation with an
existing network infrastructure.
Furthermore, architectural factors such as the network control type applied (centralized or
decentralized) and which architectural components are virtualized, affect the performance of
each mobility management scheme.
Regarding the network control type applied, the discussed mobility management schemes
could be adapted to perform either centralized or distributed control. Centralized control is
optional in VC, SDN-V and HVA architectures. Such a configuration uses a central entity
performing the mobility management, such as a single SDN controller or a central vehicle with
increased responsibilities. Accordingly, decentralized control is performed when the network
control process is distributed to more than one control entities. Decentralized control can be
applied to the entire 5G-VCC architectures. Furthermore, both centralized and decentralized
control can be combined. For instance, in some VC architectures clustering of vehicles is
performed, where a vehicle is elected as the Cluster Head (CH) [246] implementing the
3.2 Mobility Management Schemes for 5G Wireless Networks 111
functionalities of a single SDN controller. In this case, centralized control is applied for intra-
cluster communication, while distributed control is applied for inter-cluster communication.
Finally, the considered mobility management schemes can be deployed to several virtualized
resources of a 5G-VCC architecture. Specifically, in a VC architecture virtualization of
resources is performed by the end-user devices. Also, in a VuC architecture virtualization
is applied in the Cloud, while in a VuF architecture virtualization is applied in the Fog.
Network as a Service (NaaS) [247–249] enables the configuration of network resources in
either the Cloud or the Fog. Furthermore, in HVA architectures, virtualization is simultaneously
performed in two or more architectural components. Indicatively, in the architecture discussed
in [73] virtualization is applied to both Cloud and Fog nodes. Also, a fully virtualized HVA
architecture [31][75], optimizes the performance of mobility management schemes, since it
provides increased system reliability, resource utilization and performance isolation.
Table 3.26 The Applicability of each Scheme to the Available 5G-VCC Architectures.
Each communication model could be applied to specific 5G-VCC architectures (Table 3.27). In
general, V2V and V2P models can be applied in each 5G-VCC architecture, since the entire
topologies can include both vehicular and pedestrian users. However, the V2I communication
model requires a network infrastructure. Consequently, it cannot be applied to a VC architecture
which is completely ad-hoc. On the other hand, V2I could be used in VuC and VuF topologies,
since they provide access to network infrastructure through RSUs. Also, HVC could be applied
to the entire topologies, since it combines multiple communication models.
Table 3.28 presents the applicability of each mobility management scheme to the available
communication models. As could be observed the entire mobility management schemes support
V2I communications, since these schemes could be applied to 5G-VCC topologies where an
access network infrastructure exists. However, only the HCA [217] scheme supports V2V and
V2P communications since it could also be applied to VC architectures where no access network
infrastructure exists. Thus, the design of mobility management algorithms for supporting V2V
and V2P communications could be considered as a scientific field for further future research,
since these communications are necessary for supporting several vehicular services in smart
cities environments [95][250].
114 Mobility Management
• Seamless mobility During the vehicles movement, their ability to remain connected while
roaming across different access networks must be ensured, in order to avoid interruption
of user’s services.
• Minimization of handovers count Even in cases where the handover latency has been
reduced enough, more handovers will always occur to higher total latencies and network
signaling overhead. In the modern 5G network access environment, where multiple cells
coexist in the same place, a vehicle could perform a large number of handovers reducing
the perceived Quality of Service (QoS) of vehicular services.
116 A Proposed Mobility Management Approach for 5G-VCC Systems
• Optimal network selection for satisfying QoS constraints of vehicular services Connec-
tivity to the most appropriate network supporting the constraints of vehicular services
must be ensured. Accordingly, the available network alternatives must be continuously
evaluated considering vehicular services requirements, as well as changes in the radio
environment and operators policies.
number of the parameters considered and K the number of the available services.
K
1 1 1
Qu,i = ∑ (wth,k · thu,i,k + wd,k · du,i,k + w j,k · ju,i,k + w pl,k · plu,i,k )/N /K (4.1)
k=1
The user’s equipment continuously monitors its perceived SINRu,i and Qu,i values. When
either SINRu,i or Qu,i becomes less than a predefined threshold, the user equipment sends
the obtained values to the Fog infrastructure. Subsequently, the Fog infrastructure uses the
Mamdani satisfaction chart presented in Figure 4.2, in order to determine the Su,i satisfaction
grade. If the satisfaction grade is less than a predefined Sth threshold value, then the network
selection process is executed.
The satisfaction indicators chart presents the Su,i values obtained for each possible SINRu,i
and Qu,i combination. Indicatively, when the SINRu,i and Qu,i values are too low, the produced
Su,i value is too low as well. On the contrary, when the SINRu,i and Qu,i values are close to
1, the produced Su,i value is also high, indicating that the user is fully satisfied. Furthermore,
when only one of the SINRu,i or the Qu,i values is close to 0, the user satisfaction is in quite
low levels.
At this point, it has to be noted that since the user’s satisfaction is obtained at the Fog by
performing a lookup at the satisfaction indicators chart, the overhead introduced at the Fog is
minimal. Also, the method does not impose any significant overhead at the user equipment due
to the monitoring of the SINRu,i and Qu,i parameters.
The Mamdani satisfaction chart is evaluated once during the instantiation of the Fog services.
Each satisfaction indicator Su,i of Figure 4.2 is obtained using the MPFIS method, considering
the SINRu,i and the Qu,i as input parameters. Both SINRu,i and Qu,i are normalized in order
to have values within the range [0, 1]. Also, the MFSINR , MFQ , MFS membership functions
(MF) are defined using the EUM method, indicating the linguistic terms and the corresponding
Interval Valued Pentagonal Fuzzy Numbers (I-VPFN) for the fuzzy representation of the
SINRu,i , Qu,i and Su,i respectively. Thus, for each crisp value, two membership degrees are
determined in the corresponding MF, one for the upper pentagon and one for the lower pentagon.
Table 4.1 represents the linguistic terms and the corresponding interval-valued pentagonal fuzzy
numbers of MFSINR , MFQ and MFS membership functions, which are equally distributed inside
the domain [Umin ,Umax ] = [0, 1] as illustrated in Figures 4.3, 4.4 and 4.5, respectively. Table
4.2 shows the considered fuzzy rule base.
118 A Proposed Mobility Management Approach for 5G-VCC Systems
Table 4.1 Linguistic Terms and the corresponding Interval-Valued Pentagonal Fuzzy Numbers
used for MFSINR , MFQ and MFS .
formula 4.2 defined in [252], where r f is the radius of femtocell i and vu is the velocity of the
vehicle u.
residence π · ri
tu,i = (4.2)
2 · vu
Then, the vehicle velocity is obtained by formula 4.3 defined in [253].
π · NCRu
vu = √ (4.3)
4 · TCRmax · λu
The λu parameter, obtained by the SDN controller, denotes the cell’s density in the location of
vehicle u.
are introduced by other handover algorithms proposed in the research literature. The novelty
in our approach is that most of the computational complexity is performed at the Cloud/Fog
infrastructure.
Network specifications are expressed directly using linguistic terms (Appendix B). In
particular, the crisp values of the network QoS characteristics are converted into linguistic
terms which correspond to specific ranges of values per service type. Table 4.3 illustrates the
mapping between ranges of network QoS characteristics values and the respective linguistic
terms for the VoIP service. Table 4.4 summarizes the simulation parameters.
During the HO initiation process, the user satisfaction Su,i is obtained using the MPFIS
Satisfaction Chart, considering the estimated Qu,i and SINRu,i values. It should be noted that
the services weights used for the Qu,i estimation are calculated using the PF-ANP method. The
criteria used include throughput, delay, jitter and packet loss. Figure 4.7 depicts the estimated
HO initiation weights for each service. Furthermore, the Sth,SLA thresholds given in Table
4.6 are estimated considering the minimum acceptable QSLA and SINRSLA values per SLA. If
the Su,i becomes less than the respective Sth,SLA threshold then the vehicle should perform a
handover to a new access network. The HO initiation results for each vehicle are presented in
Table 4.7. As it can be observed, vehicle 8 will remain to its current network i, while the rest of
the vehicles will procceed to handover.
For each one of the monitored vehicles, Table 4.8 presents its SLA, the services used, its
geographical position and its estimated velocity.
During the network selection process the PF-ANP method is applied in order to estimate the
weights of the network selection criteria per service type and SLA. Figure 4.8 depicts the
PF-ANP network model. The criteria are classified into two groups, namely the Network
QoS and the Network Policy characteristics. The Network QoS characteristics group contains
122 A Proposed Mobility Management Approach for 5G-VCC Systems
network performance related criteria including throughput, delay, jitter and packet loss. The
Network Policy characteristics group contains operator defined rules such as price, security and
service reliability. Pairwise comparison decision matrices are created based on relations among
the eight selection criteria depicted in Figure 4.9. Then, these pairwise comparison decision
matrices are used to evaluate the priority vectors of the criteria and to form the supermatrix
per service type and SLA. Subsequently, the weighted supermatrices and, finally, the limited
supermatrices are obtained. Indicatively, for the SLA1 NAV service, the initial, the weighted
and the limited supermatrices are presented in Tables 4.9, 4.10 and 4.11 respectively.
The criteria weights per service and SLA obtained by the limit supermatrices are presented
in Figures 4.10 - 4.13. As illustrated, the weights are proportional to the constraints of each
service as well as to the agreements of each SLA. In particular, the weight of the price criterion
is low for SLA1, in which the service reliability and the network QoS characteristics are
considered as the most important factors. In SLA2 the price criterion is more important than in
SLA1, thus the respective weight is greater than that of SLA1. Consequently, the weights of the
service reliability and QoS characteristics criteria in SLA2 are lower compared to the relative
weights of SLA1. In SLA3 the weights of price and service reliability criteria are balanced
as they are almost of equivalent importance. Finally, in SLA4 the price is the most important
criterion resulting in a high estimated weight value.
The PFT algorithm selects the best network for each vehicle considering the vehicle service
requirements. Figure 4.14 compares the results of the proposed scheme with the ones obtained
using VAH [218], FAT [255], FAS [255], FAM [255] and FAE [172]. In this Figure, for each
vehicle the PFT ranking of the current network as well as of the target network connection
estimated by each of the six schemes are presented. From the obtained results it is clear that
the suggested algorithm outperforms the existing schemes since it selects as target networks for
vehicles the ones with the best ranks. Also, in special cases where the velocity of the vehicles
is high (eg. for vehicle 7) the proposed scheme considers only the wide coverage candidate
networks as alternatives avoiding the handovers to femtocell networks.
Additionally, for a further evaluation of the proposed scheme, the user’s satisfaction grade
Su,i after the HO completion is considered. Figure 4.15 presents the estimated Su,i after
the execution of each HO scheme. As it can be observed, the results are similar to the
aforementioned PFT rankings, while the proposed approach achieves higher satisfaction grades
than the existing schemes.
4.3 Simulation Setup 123
Sc = 2Sn − 1 (4.4)
The average PFT rankings of each HO scheme are presented in Figure 4.16, where all the
39807 vehicles are considered. As it can be observed, the proposed scheme accomplishes
higher ranking than the other schemes. On the contrary, the VAH scheme accomplishes the
lower ranking, due to the fact that it does not consider neither QoS related parameters, nor
operator policy related parameters. The other schemes accomplish intermediate results. The
FAT results are slightly better than the ones achieved from the rest of the schemes. Also,
Figure 4.17 presents the average satisfaction grade Su,i estimated for each HO scheme. The
proposed approach accomplishes the highest average values, compared to the rest of the HO
schemes. It should be noted that the number of HOs performed should be considered as a
critical parameter for the evaluation of the proposed HO scheme efficiency, since a smaller
number of HOs implies a lower network signaling overhead in both the user and the operator
sides. Figure 4.18 presents the total number of HOs performed in each scheme, during the
entire 24 hours simulation. As it is shown, the proposed scheme outperforms the other schemes,
since it performs a smaller number of HOs. Also, the VAH scheme accomplishes an acceptable
number of HOs, since it also considers the vehicle velocity for the handover process. However,
it does not avoid useless HOs to femtocells providing worse results than the proposed scheme.
The rest of the schemes accomplish similar HO numbers, since they are based on the perceived
RSS per user for the HO initiation process.
124 A Proposed Mobility Management Approach for 5G-VCC Systems
Fog MMaaS
Vehicle u Cloud
Current SDN Controller
Network i
A. VHO initiation
Monitoring of
Qu,i and SINR u,i
When Qu,i < Qth or
SINRu,i < SINR th:
C. Network selection
ObtainVelocity u(NCRu, TCRmax, λu)
If velocity = High:
Obtain offered characteristics of
Alternatives_Set_A(Macrocells)
Execute PFT for
Alternatives_Set_A(Macrocells)
Result=Selected_Network
If velocity = Medium:
Obtain Femtocells_subset
considering tresidence
B. Velocity Monitoring
Obtain offered characteristics of
Alternatives_Set_A
(Macrocells, Femtocells_subset)
Execute PFT for
Alternatives_Set_A
(Macrocells, Femtocells_subset)
Result=Selected_Network
If velocity = Low:
Obtain offered characteristics of
Alternatives_Set_A
(Macrocells, Femtocells)
Execute PFT for Alternatives_Set_A
(Macrocells, Femtocells)
Result=Selected_Network
Else:
Result=Handover_not_required
If Result=Selected_Network:
Fog
Selected_Network
D. VHO execution
Cloud
...
VM ...
VM Services
Services
... VM
... Services
...
VM Fog
Services
... a b c d e f g h i j k l m n o p q r s t u v w x y
... SDN controller 1 1
...
MMaaS
2 2
3 3
4 4
5 5
Fog 6 6 LTE Macrocell
7 7
LTE Femtocell
8 8
9 9 WiMAX Macrocell
WiMAX Macrocells
LTE Macrocells
10 10 WiMAX Femtocell
11 11
WAVE RSU
WiMAX Femtocells
12 12
Vehicle
LTE Femtocells 13 13
WAVE RSUs
14 14
15 15
16 16
17 17
18 18
19 19
a b c d e f g h i j k l m n o p q r s t u v w x y
Figure 4.6 The Simulated Topology for the Evaluation of the proposed Mobility Management
Scheme.
Table 4.3 Relation of the Network QoS Characteristics and Linguistic Terms for VoIP.
0,3
0,2
0,1
0
Navigation Assistance VoIP Conversational Video Buffered Streaming Web
Table 4.4 The Simulation Parameters for the Evaluation of the proposed Mobility Management
Scheme.
Parameter Value
Simulation duration 86400 seconds (24 hours)
LTE Macrocell BSs: 20
Networks count LTE Femtocell BSs: 37
WiMAX Macrocell BSs: 17
WiMAX Femtocell BSs: 25
WAVE RSUs: 22
Total: 121
LTE/WiMAX Macrocells: 400 meters
Cells radius
LTE/WiMAX Femtocells: 30 meters
WAVE RSUs: 150 meters
According to the Hellenic Telecommu-
Networks positions nications and Post Commission (EETT)
[254] data (see Appendix A)
Networks frequencies See Appendix A
Service Layer Agreements
4
(SLA) count
Vehicles count 39807
Average vehicles arrival rate 0,460729167 vehicles per second
Average vehicles departure rate 0,46 vehicles per second
SLA 1: 9952 vehicles (25,00062803 %)
Vehicles per SLA SLA 2: 9952 vehicles (25,00062803 %)
SLA 3: 9952 vehicles (25,00062803 %)
SLA 4: 9951 vehicles (24,99811591 %)
Available networks per SLA See Appendix B
Navigation Assistance (NAV)
Services Voice over IP (VoIP)
Conversational Video (CV)
Buffered Streaming (BS)
Web Browsing (WB)
Distribution of services to See Appendix C
vehicles
Normal: 13348 (33.53 %)
Vehicles per velocity
Medium: 13348 (33.53 %)
High: 13111 (32.94 %)
Table 4.6 The QSLA , SINRSLA and Sth,SLA thresholds per SLA.
Vehicle Current Network i Qu,i SINRu,i (SINRdBu,i ) Su,i Sth,SLAu Handover Required
1 WAVE 14 0.871 0.14 (-5.1 dB) 0.18157 0.81843 Yes
2 LTE Macro 11 0.912 0.17 (-4.05 dB) 0.18212 0.81843 Yes
3 WiMAX Macro 8 0.948 0.29 (0.15 dB) 0.32523 0.81843 Yes
4 LTE Macro 7 0.644 0.23 (-1.95 dB) 0.12122 0.66980 Yes
5 WAVE 9 0.990 0.14 (-5.1 dB) 0.18156 0.66980 Yes
6 WAVE 11 0.827 0.04 (-8.6 dB) 0.02539 0.66980 Yes
7 WiMAX Macro 7 0.999 0.21 (-2.65 dB) 0.18965 0.57767 Yes
8 WiMAX Macro 11 0.704 0.71 (14.85 dB) 0.61247 0.57767 No
9 WiMAX Macro 5 0.988 0.25 (-1.25 dB) 0.27297 0.45457 Yes
10 LTE Macro 18 0.990 0.29 (0.15 dB) 0.35622 0.45457 Yes
Criteria Groups
Criteria
Troughput Service Reliability
Delay Security
Jitter Price
Packet Loss
Throughput
Price
Delay Jitter
Service
Security
Reliability
Packet loss
Figure 4.9 The Relations of the Criteria considered by the PF-ANP Network Model.
0,15
0,1
0,05
0
Navigation Assistance VoIP Conversational Video Buffered Streaming Web
Figure 4.10 The Network Selection Weights per Service for SLA 1.
0,15
0,1
0,05
0
Conversational Video Buffered Streaming Web
Figure 4.11 The Network Selection Weights per Service for SLA 2.
0,15
0,1
0,05
0
Buffered Streaming Web
Figure 4.12 The Network Selection Weights per Service for SLA 3.
0,15
0,1
0,05
0
Web
Figure 4.13 The Network Selection Weights per Service for SLA 4.
4.3 Simulation Setup 131
Table 4.9 The PF-ANP Initial Supermatrix for the SLA 1 NAV Service.
Table 4.10 The PF-ANP Weighted Supermatrix for the SLA 1 NAV Service.
Table 4.11 The PF-ANP Limit Supermatrix for the SLA 1 NAV Service.
0,5
0,4
PFT rank
0,3
0,2
0,1
0
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
v=Medium v=Normal v=Medium v=Normal v=High v=Normal v=High v=Medium v=High v=Normal
Current Network Proposed Scheme VAH FAT FAS FAM FAE
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
Vehicle 1 Vehicle 2 Vehicle 3 Vehicle 4 Vehicle 5 Vehicle 6 Vehicle 7 Vehicle 8 Vehicle 9 Vehicle 10
0,47
0,46
0,45
0,44
0,43
Proposed VAH FAT FAS FAM FAE
VHO scheme
Figure 4.16 The Average PFT Ranking of each HO Scheme during the 24 Hours Simulation.
0,74
0,72
0,7
0,68
0,66
0,64
0,62
0,6
Proposed VAH FAT FAS FAM FAE
VHO scheme
Figure 4.17 The Average Satisfaction Grade obtained from each HO Scheme during the 24
Hours Simulation.
400000
VHOs count
350000
300000
250000
200000
Proposed VAH FAT FAS FAM FAE
VHO scheme
Figure 4.18 The Total HOs Count of each HO Scheme during the 24 Hours Simulation.
Chapter 5
Conclusion
In this doctoral thesis, two key aspects of the 5G-VCC systems were studied, namely the
allocation of network resources and the handling of the vehicles’ mobility.
Regarding the allocation of the network resources, an overview of the available MAC
schemes for vehicular networks has been performed [257]. These schemes can be applied in
5G-VCC systems in order optimal manipulation of communication resources to be performed.
Some schemes organize the vehicles into clusters, while GPS receivers are used in some
cases in order the vehicles to be synchronized with each other. The discussed schemes apply
either distributed or centralized control. In a 5G-VCC system the centralized control could be
enhanced using an SDN controller, while the distributed control could be easily configured
using a Fog computing infrastructure. Furthermore, V2V communication is supported from
most schemes, while some schemes support V2I communication. Finally, during the medium
access, CRN could be applied to take advantage of free white spaces into the available spectrum.
Subsequently, a QoS aware scheduler called FLSA [258][259] was proposed in order to
improve the resource allocation for real time services in the LTE downlink. FLSA has been built
on three distinct levels which cooperate with each other to allocate the network resources to
users. In every TTI the upper level estimates the quota of the data that each real-time flow should
transmit to satisfy its QoS constraints. Then, the middle level uses the M-LWDF algorithm
to allocate RBs to real time flows in each TTI for transmitting their quota of data obtained by
the upper level. Thereafter, if there are free RBs available in the current TTI, the lower level
allocates them both to real time and to best effort flows, using the M-LWDF algorithm. The
performance of FLSA for real time flows was compared against other scheduling algorithms in
terms of PLR, attainable throughput and fairness. The simulation results showed that the FLSA
scheduler outperforms the rest of the schemes by achieving better resource allocation for real
time services.
134 Conclusion
Furthermore, FLSA-CC [260], QoS aware cross carrier downlink scheduler has also been
proposed as an extended version of the FLSA. FLSA-CC operates in an LTE-A network
with relay nodes in a CA mode. The proposed scheduler has been built upon three distinct
levels which cooperate with each other to allocate the network resources to users in a manner
that the requirements of strict real times services are satisfied while a starvation of the best
effort traffic is avoided. The performance of FLSA-CC is compared against other scheduling
algorithms in terms of PLR, throughput and fairness index, in a cloud assisted software defined
architecture. The simulation results showed that the FLSA-CC scheduler outperforms the rest
of the scheduling schemes by achieving better resource allocation for both real time and best
effort services.
An overview of existing mobility management schemes has been made. These schemes
can be applied to 5G-VCC systems in order an optimal manipulation of vehicles mobility to be
performed. Each scheme implements one or more mobility management subprocesses, namely
HO initiation, network selection and HO execution. The 5G-VCC topologies (e.g. VC, VuC,
VuF, SDN-V and HVT) that each scheme supports were mentioned, while the communication
models that could be applied in each scheme (V2V, V2I, V2P and HVC) were introduced. The
control type of each scheme (vehicle or network controlled) was also described, while at the
same time the OSI layer where each scheme was implemented indicate its mobility management
capabilities. In addition, the supported message exchange mechanisms were considered, while
the evaluation of each scheme considering the type of the implemented algorithms provided
useful knowledge about its design.
Novel mobility management algorithms for 5G networks were proposed. Specifically,
firstly, we described a network selection method that takes into account the network QoS
characteristics policies, application requirements and different types of user SLAs to select the
optimal network that will satisfy simultaneously all the applications’ requirements and the users’
preferences running on a mobile user’s device [171]. The proposed method employed two
MADM algorithms: the ANP for criteria weights calculation and the TFT for accomplishing
the overall rating of the network technologies. ANP was selected to determine the relative
importance and the dependence of the criteria. As selection criteria, we considered the network
QoS parameters, service constraints, user requirements and provider policies. These criteria
were represented by interval-valued trapezoidal fuzzy numbers. Then, the TFT algorithm was
applied to calculate the overall rating of the available networks. The performance evaluation of
TFT showed that when a user has only one service, it provides similar results to the original
TOPSIS and FAE methods. However, when a user requires multiple services, TFT performs
better by satisfying multiple groups of criteria per user since the original TOPSIS and FAE
methods cannot support more than one services. Furthermore, according to the sensitivity
135
analysis of results it was shown that the described method doesn’t suffer from the ranking
abnormality problem, thus, the results are normally adjusted to the heterogeneous network
environment changes.
Furthermore, we proposed a network selection scheme for supporting medical services in
5G-VCC systems [68][261]. The discussed scheme consists of the TF-ANP method to calculate
the relative importance of the selection criteria and the TFT to accomplish the ranking of the
candidate networks. The health status of each patient is considered, while the criteria used for
network evaluation included throughput, delay, jitter, packet loss, service reliability, security
and price. The performance evaluation showed that the proposed scheme outperforms existing
network selection methods by satisfying multiple groups of criteria and medical services per
user. Also, we proposed an improved version of this scheme [66]. In this case, the discussed
scheme consists of two FMADM algorithms, namely the TF-AANP to calculate the relative
importance of each service, as well as the weights of the selection criteria and the TFT-ACW to
accomplish the ranking of the candidate networks. The health status of onboard patients and
the SLA of each vehicle were considered, while the criteria used for network evaluation include
throughput, delay, jitter, packet loss, service reliability, security and price. The performance
evaluation showed that the proposed scheme outperforms existing network selection methods
by satisfying the strict constraints of medical services, when the patient’s health status becomes
immediate and multiple types of non-medical services coexist with medical services to the
vehicle.
Finally, we proposed a HO management scheme for 5G-VCC systems that takes into
account the vehicle velocity, network’s QoS and policy characteristics, the vehicular services’
requirements and the various vehicle SLAs to select the optimal network that will satisfy
simultaneously all the service requirements and onboard user preferences [67][262]. The
HO management services are executed at the Cloud or the Fog infrastructure alleviating the
vehicle’s processing load and reducing HO latency. IVPFNs are used for the representation
of both HO Initiation and Network Selection criteria values, while the PF-ANP algorithm is
used for the estimation of the criteria weights for both HO Initiation and Network Selection.
During the HO Initiation, MPFIS evaluates the necessity to perform a handover, while during
the Network Selection, the PFT algorithm selects the most appropriate network considering
the vehicle’s velocity, the constraints of used vehicular services and the SLA of the vehicle.
The performance evaluation showed that the Always Best Connected (ABC) principle is fully
ensured, while at the same time the proposed scheme outperforms existing HO management
algorithms in terms of the PFT ranks of the selected networks, the user satisfaction grade, as
well as the VHOs count.
136 Conclusion
11. Emmanouil Skondras, Angelos Michalas, Nikolaos Tsolis, Aggeliki Sgora, Dimitrios D.
Vergados, “A Network Selection Scheme for Healthcare Vehicular Cloud Computing Sys-
tems”, International Conference on Information, Intelligence, Systems and Applications
(IISA), Larnaka, Cyprus, August 28-30, 2017
13. Emmanouil Skondras, Angelos Michalas, Aggeliki Sgora, Dimitrios D. Vergados, “QoS-
aware scheduling in LTE-A networks with SDN control”, International Conference on
Information, Intelligence, Systems and Applications (IISA), Chalkidiki, Greece, July
13-15, 2016
16. Emmanouil Skondras, Aggeliki Sgora, Angelos Michalas, Dimitrios D. Vergados, “An
analytic network process and trapezoidal interval-valued fuzzy technique for order
preference by similarity to ideal solution network access selection method”, International
Journal of Communication Systems (IJCS), Wiley, September 2014
17. Emmanouil Skondras, Angelos Michalas, Malamati Louta, George Kouzas, “Personal-
ized Multimedia Web Services in Peer to Peer Networks Using MPEG-7 and MPEG-21
Standards”, Studies in Computational Intelligence – Semantic Hyper/Multimedia Adapta-
tion, Volume 418/2013 (book chapter), pp. 361-383, Springer-Verlag Berlin Heidelberg,
2013
Bibliography
[1] Ricard Vilalta, Victor Lopez, Alessio Giorgetti, Shuping Peng, Vittorio Orsini, Luis
Velasco, Rene Serral-Gracia, Donal Morris, Silvia De Fina, Filippo Cugini, et al. Tel-
cofog: A unified flexible fog and cloud computing architecture for 5g networks. IEEE
Communications Magazine, 55(8):36–43, 2017.
[2] Faqir Zarrar Yousaf, Michael Bredel, Sibylle Schaller, and Fabian Schneider. Nfv
and sdn-key technology enablers for 5g networks. IEEE Journal on Selected Areas in
Communications, 2017.
[3] Hojjat Salehinejad, Hossein Nezamabadi-pour, Saeid Saryazdi, and Fereydoun Farrahi-
Moghaddam. Combined a*-ants algorithm: a new multi-parameter vehicle navigation
scheme. arXiv preprint arXiv:1504.07329, 2015.
[4] Pietro Edoardo Carnelli, Joy Yeh, Mahesh Sooriyabandara, and Aftab Khan. Parkus: A
novel vehicle parking detection system. In AAAI, pages 4650–4656, 2017.
[5] Yang Li, Xiaoming Tao, and Jianhua Lu. Hybrid model-and-object-based real-time
conversational video coding. Signal Processing: Image Communication, 35:9–19, 2015.
[6] Toon De Pessemier, Isabelle Stevens, Lieven De Marez, Luc Martens, and Wout Joseph.
Quality assessment and usage behavior of a mobile voice-over-ip service. Telecommuni-
cation Systems, 61(3):417–432, 2016.
[7] Henrik Klessig and Gerhard Fettweis. Impact of inter-cell interference on buffered video
streaming startup delays. In Vehicular Technology Conference (VTC Fall), 2015 IEEE
82nd, pages 1–2. IEEE, 2015.
[8] Verdes Pedras, Marco Sousa, Pedro Vieira, Maria-Paula Queluz, and Antonio Rodrigues.
A no-reference user centric qoe model for voice and web browsing based on 3g/4g radio
measurements. In Wireless Communications and Networking Conference (WCNC), 2018
IEEE, pages 1–6. IEEE, 2018.
[9] Zinonas C Antoniou, Andreas S Panayides, Marios Pantziaris, Anthony G Constantinides,
Constantinos S Pattichis, and Marios S Pattichis. Dynamic network adaptation for real-
time medical video communication. In XIV Mediterranean Conference on Medical and
Biological Engineering and Computing 2016, pages 1099–1104. Springer, 2016.
[10] Luís FR Lucas, Nuno MM Rodrigues, Luis A da Silva Cruz, and Sérgio MM de Faria.
Lossless compression of medical images using 3-d predictors. IEEE transactions on
medical imaging, 36(11):2250–2260, 2017.
142 Bibliography
[11] Gopi Krishna Garge, Chitra Balakrishna, and Soumya Kanti Datta. Consumer health care:
Current trends in consumer health monitoring. IEEE Consumer Electronics Magazine,
7(1):38–46, 2018.
[12] Qinghan Xue and Mooi Choo Chuah. Incentivising high quality crowdsourcing medical
data for disease diagnose & survival prediction. Smart Health, 2017.
[13] Tugce Bilen, Berk Canberk, and Kaushik R Chowdhury. Handover management in
software-defined ultra-dense 5g networks. IEEE Network, 31(4):49–55, 2017.
[14] TS 36.213 version 14.2.0: Evolved Universal Terrestrial Radio Access Network (E-
UTRAN) (Release 14). Technical Specification, 3GPP, 2017.
[15] P802.16/d4 - ieee draft standard for air interface for broadband wireless access systems
(revision of ieee std 802.16-2012). IEEE Standard, 2017.
[16] 1609.12-2016 - ieee standard for wireless access in vehicular environments (wave) –
networking services. IEEE Standard, 2016.
[17] Hucheng Wang, Shanzhi Chen, Ming Ai, and Hui Xu. Localized mobility management
for 5g ultra dense network. IEEE Transactions on Vehicular Technology, 66(9):8535–
8552, 2017.
[18] Daniel Calabuig, Sokratis Barmpounakis, Sonia Gimenez, Apostolos Kousaridas, Tilak R
Lakshmana, Javier Lorca, Petteri Lunden, Zhe Ren, Pawel Sroka, Emmanuel Ternon,
et al. Resource and mobility management in the network layer of 5g cellular ultra-dense
networks. IEEE Communications Magazine, 55(6):162–169, 2017.
[19] Tarik Taleb, Adlen Ksentini, and Riku Jantti. " anything as a service" for 5g mobile
systems. IEEE Network, 30(6):84–91, 2016.
[20] Rakibul Islam Rony, Akshay Jain, Elena Lopez-Aguilera, Eduard Garcia-Villegas, and
Ilker Demirkol. Joint access-backhaul perspective on mobility management in 5g
networks. In Standards for Communications and Networking (CSCN), 2017 IEEE
Conference on, pages 115–120. IEEE, 2017.
[21] Akshay Jain, Elena Lopez-Aguilera, and Ilker Demirkol. Mobility management as a
service for 5g networks. arXiv preprint arXiv:1705.09101, 2017.
[22] Ke Zhang, Yuming Mao, Supeng Leng, Yejun He, and Yan Zhang. Mobile-edge
computing for vehicular networks: A promising network paradigm with predictive
off-loading. IEEE Vehicular Technology Magazine, 12(2):36–44, 2017.
[23] Binxu Yang, Wei Koong Chai, Zichuan Xu, Konstantinos V Katsaros, and George Pavlou.
Cost-efficient nfv-enabled mobile edge-cloud for low latency mobile applications. IEEE
Transactions on Network and Service Management, 2018.
[24] Xumin Huang, Rong Yu, Jiawen Kang, Yejun He, and Yan Zhang. Exploring mobile
edge computing for 5g-enabled software defined vehicular networks. IEEE Wireless
Communications, 24(6):55–63, 2017.
Bibliography 143
[25] Mehdi Sookhak, F Richard Yu, Ying He, Hamid Talebian, Nader Sohrabi Safa, Nan Zhao,
Muhammad Khurram Khan, and Neeraj Kumar. Fog vehicular computing: Augmenta-
tion of fog computing using vehicular cloud computing. IEEE Vehicular Technology
Magazine, 12(3):55–64, 2017.
[26] Cheng Huang, Rongxing Lu, and Kim-Kwang Raymond Choo. Vehicular fog computing:
architecture, use case, and security and forensic challenges. IEEE Communications
Magazine, 55(11):105–111, 2017.
[27] Jiafu Wan, Daqiang Zhang, Shengjie Zhao, Laurence T Yang, and Jaime Lloret. Context-
aware vehicular cyber-physical systems with cloud support: architecture, challenges,
and solutions. IEEE Communications Magazine, 52(8):106–113, 2014.
[28] A.R. Deepti G.Sasikala. Real time services for future cloud computing enabled vehicle
networks. In IOSR Journal of Computer Engineering (IOSR-JCE), volume 11, pages
8–14, 2013.
[29] Dipal Vashi Priyank Sharma, Sandip Vaniya. Communication as a service based cloud
computing. IEEE International Conference on Emerging Technology Trends (ICETECT),
Nagercoil, India, pages 15–17, 2011.
[30] Quang Hieu Vu, Tran-Vu Pham, Hong-Linh Truong, Schahram Dustdar, and Rasool Asal.
Demods: A description model for data-as-a-service. In 2012 IEEE 26th International
Conference on Advanced Information Networking and Applications, pages 605–612.
IEEE, 2012.
[31] Fekri M Abduljalil. Video capture service in the intelligent transportation system based
on cloud computing. International Journal of Computer Applications, 97(5), 2014.
[32] Rasheed Hussain, Fizza Abbas, Junggab Son, and Heekuck Oh. Tiaas: Secure cloud-
assisted traffic information dissemination in vehicular ad hoc networks. In Cluster, Cloud
and Grid Computing (CCGrid), 2013 13th IEEE/ACM International Symposium on,
pages 178–179. IEEE, 2013.
[33] Mario Gerla, Jui-Ting Weng, and Giovanni Pau. Pics-on-wheels: Photo surveillance
in the vehicular cloud. In Computing, Networking and Communications (ICNC), 2013
International Conference on, pages 1123–1127. IEEE, 2013.
[34] Stefano Ferretti and Gabriele D’Angelo. Smart shires: The revenge of countrysides.
In Computers and Communication (ISCC), 2016 IEEE Symposium on, pages 756–759.
IEEE, 2016.
[35] Rasheed Hussain and Heekuck Oh. Cooperation-aware vanet clouds: Providing secure
cloud services to vehicular ad hoc networks. JIPS, 10(1):103–118, 2014.
[36] Moustafa AbdelBaky, Manish Parashar, Kirk Jordan, Hyunjoo Kim, Hani Jamjoom,
Zon-Yin Shae, Gergina Pencheva, Vipin Sachdeva, James Sexton, Mary Wheeler, et al.
Enabling high-performance computing as a service. Computer, 45(10):72–80, 2012.
144 Bibliography
[37] Steffen Kächele, Christian Spann, Franz J Hauck, and Jörg Domaschka. Beyond iaas
and paas: An extended cloud taxonomy for computation, storage and networking. In
Proceedings of the 2013 IEEE/ACM 6th International Conference on Utility and Cloud
Computing, pages 75–82. IEEE Computer Society, 2013.
[38] Khaleel Mershad and Hassan Artail. Finding a star in a vehicular cloud. IEEE Intelligent
transportation systems magazine, 5(2):55–68, 2013.
[39] Rasheed Hussain, Junggab Son, Hasoo Eun, Sangjin Kim, and Heekuck Oh. Rethinking
vehicular communications: Merging vanet with cloud computing. In Cloud Computing
Technology and Science (CloudCom), 2012 IEEE 4th International Conference on, pages
606–609. IEEE, 2012.
[40] Alexander Stanik, Matthias Hovestadt, and Odej Kao. Hardware as a service (haas):
Physical and virtual hardware on demand. In Cloud Computing Technology and Science
(CloudCom), 2012 IEEE 4th International Conference on, pages 149–154. IEEE, 2012.
[41] Monica Mandal, Chaitrali Landge, Pramila Gaikwad, Uma Nagaraj, and Ashwini Abhale.
Implementing storage as a service in vanet using cloud environment. International
Journal of Advance Foundation and Research in Computer (IJAFRC), 1, 2014.
[42] Mario Gerla, Eun-Kyu Lee, Giovanni Pau, and Uichin Lee. Internet of vehicles: From
intelligent grid to autonomous cars and vehicular clouds. In Internet of Things (WF-IoT),
2014 IEEE World Forum on, pages 241–246. IEEE, 2014.
[43] Peyman Talebifard and Victor CM Leung. Towards a content-centric approach to
crowd-sensing in vehicular clouds. Journal of Systems Architecture, 59(10):976–984,
2013.
[44] Hajar Mousannif, Ismail Khalil, and Hassan Al Moatassime. Cooperation as a service in
vanets. J. UCS, 17(8):1202–1218, 2011.
[45] K Rajbhojsupriya and Pankaj R Ch. Implementation of enhanced security on vehic-
ular cloud computing. International Journal of Computer Science and Information
Technologies (IJCSIT), 5:1315–1318, 2014.
[46] Jithrendra H. N. Jyoti Metan, Narasimha K. N. Murthy. Moving vanet to vehicular
cloud. International Journal of Innovative Research in Computer and Communication
Engineering (IJIRCCE), 2, 2014.
[47] Mehmet Ali Ertürk, Muhammed Ali Aydin, Luca Vollero, and Roberto Setola. Ieee
802.11 s mesh network analysis for post disaster communication. In International
Telecommunications Conference, pages 53–59. Springer, 2019.
[48] Mehdi Sookhak, F Richard Yu, and Helen Tang. Secure data sharing for vehicular ad-hoc
networks using cloud computing. In Ad Hoc Networks, pages 306–315. Springer, 2017.
[49] Md Whaiduzzaman, Mehdi Sookhak, Abdullah Gani, and Rajkumar Buyya. A survey on
vehicular cloud computing. Journal of Network and Computer Applications, 40:325–344,
2014.
Bibliography 145
[50] Stephan Olariu, Tihomir Hristov, and Gongjun Yan. The next paradigm shift: from
vehicular networks to vehicular clouds, 2013.
[51] Benjamin Baron, Miguel Campista, Prométhée Spathis, Luís Henrique MK Costa,
Marcelo Dias de Amorim, Otto Carlos MB Duarte, Guy Pujolle, and Yannis Viniotis.
Virtualizing vehicular node resources: Feasibility study of virtual machine migration.
Vehicular Communications, 4:39–46, 2016.
[52] Tarek K Refaat, Burak Kantarci, and Hussein T Mouftah. Virtual machine migration
and management for vehicular clouds. Vehicular Communications, 4:47–56, 2016.
[53] Salim Bitam, Abdelhamid Mellouk, and Sherali Zeadally. Vanet-cloud: a generic cloud
computing model for vehicular ad hoc networks. IEEE Wireless Communications,
22(1):96–102, 2015.
[54] Neeraj Kumar, Rahat Iqbal, Sudip Misra, and Joel JPC Rodrigues. Bayesian coalition
game for contention-aware reliable data forwarding in vehicular mobile cloud. Future
Generation Computer Systems, 48:60–72, 2015.
[55] Yen-Wen Lin, Jie-Min Shen, and Hao-Chun Weng. Gateway discovery in vanet cloud.
In High Performance Computing and Communications (HPCC), 2011 IEEE 13th Inter-
national Conference on, pages 951–954. IEEE, 2011.
[56] Yen-Wen Lin, Jie-Min Shen, and Hao-Jun Weng. Cloud-assisted gateway discovery for
vehicular ad hoc networks. In Information Science and Service Science (NISS), 2011 5th
International Conference on New Trends in, volume 2, pages 237–240. IEEE, 2011.
[57] Rasheed Hussain, Zeinab Rezaeifar, Yong-Hwan Lee, and Heekuck Oh. Secure and
privacy-aware traffic information as a service in vanet-based clouds. Pervasive and
Mobile Computing, 24:194–209, 2015.
[58] Priya. V. Cloud service for best gateway in vanet. International Journal of Advanced
Research in Computer Science and Software Engineering, 4, 2014.
[59] Ms Madhuri H Badole and Mr Pritam Nikam. Performance evaluation of an efficient
cloud-assisted gateway discovery for vehicular ad hoc network. Performance Evaluation,
1(7), 2014.
[60] C Shravanthi and HS Guruprasad. Vanet using cloud [vuc]. 2:1–7, 2014.
[61] China Mobile. C-ran: the road towards green ran. White Paper, ver, 2, 2011.
[62] Diego Kreutz, Fernando MV Ramos, Paulo Esteves Verissimo, Christian Esteve Rothen-
berg, Siamak Azodolmolky, and Steve Uhlig. Software-defined networking: A compre-
hensive survey. Proceedings of the IEEE, 103(1):14–76, 2015.
[63] Ievgen Volvach and Larysa Globa. Mobile networks disaster recovery using sdn-nfv. In
Radio Electronics & Info Communications (UkrMiCo), 2016 International Conference,
pages 1–3. IEEE, 2016.
[64] Rashid Mijumbi, Joan Serrat, Juan-Luis Gorricho, Niels Bouten, Filip De Turck, and
Raouf Boutaba. Network function virtualization: State-of-the-art and research challenges.
IEEE Communications Surveys & Tutorials, 18(1):236–262, 2015.
146 Bibliography
[65] Sherif Abdelwahab, Bechir Hamdaoui, Mohsen Guizani, and Taieb Znati. Network
function virtualization in 5g. IEEE Communications Magazine, 54(4):84–91, 2016.
[66] Emmanouil Skondras, Angelos Michalas, Nikolaos Tsolis, and Dimitrios D Vergados.
A network selection scheme with adaptive criteria weights for 5g vehicular systems.
In Information, Intelligence, Systems & Applications (IISA), 2018 9th International
Conference on. IEEE, 2018.
[67] Emmanouil Skondras, Angelos Michalas, Aggeliki Sgora, and Dimitrios D Vergados. A
vertical handover management scheme for vanet cloud computing systems. In Computers
and Communications (ISCC), 2017 IEEE Symposium on, pages 371–376. IEEE, 2017.
[68] Emmanouil Skondras, Angelos Michalas, Nikolaos Tsolis, and Dimitrios D Vergados. A
vho scheme for supporting healthcare services in 5g vehicular cloud computing systems.
In Wireless Telecommunications Symposium (WTS), 2018, pages 1–6. IEEE, 2018.
[69] Mohammad A Salahuddin, Ala Al-Fuqaha, Mohsen Guizani, and Soumaya Cherkaoui.
Rsu cloud and its resource management in support of enhanced vehicular applications.
In 2014 IEEE Globecom Workshops (GC Wkshps), pages 127–132. IEEE, 2014.
[70] Mohammad Ali Salahuddin. Opportunistic service differentiation and cloud resource
management in support of enhanced vehicular applications. Thesis at Western Michigan
University, 2014.
[71] Md Faizul Bari, Arup Raton Roy, Shihabur Rahman Chowdhury, Qi Zhang, Mo-
hamed Faten Zhani, Reaz Ahmed, and Raouf Boutaba. Dynamic controller provisioning
in software defined networks. In Proceedings of the 9th International Conference on
Network and Service Management (CNSM 2013), pages 18–25. IEEE, 2013.
[72] Nguyen B Truong, Gyu Myoung Lee, and Yacine Ghamri-Doudane. Software defined
networking-based vehicular adhoc network with fog computing. In 2015 IFIP/IEEE
International Symposium on Integrated Network Management (IM), pages 1202–1207.
IEEE, 2015.
[73] Nicola Cordeschi, Danilo Amendola, Mohammad Shojafar, and Enzo Baccarelli. Dis-
tributed and adaptive resource management in cloud-assisted cognitive radio vehicular
networks with hard reliability guarantees. Vehicular Communications, 2(1):1–12, 2015.
[74] Joy Eze, Sijing Zhang, Enjie Liu, and Elias Eze. Cognitive radio-enabled internet of
vehicles: a cooperative spectrum sensing and allocation for vehicular communication.
IET Networks, 2018.
[75] Rong Yu, Yan Zhang, Stein Gjessing, Wenlong Xia, and Kun Yang. Toward cloud-based
vehicular networks with efficient resource management. IEEE Network, 27(5):48–55,
2013.
[76] Md Ali Al Mamun, Khairul Anam, Md Fakhrul Alam Onik, and AM Esfar-E-Alam.
Deployment of cloud computing into vanet to create ad hoc cloud network architecture.
In Proceedings of the World Congress on Engineering and Computer Science, volume 1,
pages 24–26, 2012.
Bibliography 147
[77] Carl Bergenhem, Erik Coelingh, Rolf Johansson, and Ali Tehrani. V2v communica-
tion quality: measurements in a cooperative automotive platooning application. SAE
International Journal of Passenger Cars-Electronic and Electrical Systems, 7(2014-01-
0302):462–470, 2014.
[78] Mohammed Abdulhakim Al-Absi, Ahmed Abdulhakim Al-Absi, Young-Jin Kang, and
Hoon Jae Lee. Obstacles effects on signal attenuation in line of sight for different envi-
ronments in v2v communication. In 2018 20th International Conference on Advanced
Communication Technology (ICACT), pages 17–20. IEEE, 2018.
[79] Juinn-Horng Deng, Su-Hua Chen, Chia-Fang Lee, Chorng-Ren Sheu, Hua-Lung Tsai,
Shubhranshu Singh, and Jen-Shun Yang. Joint time-frequency dmrs design for high
mobility lte-a v2v communication systems. In Microwaves, Antennas, Communications
and Electronic Systems (COMCAS), 2017 IEEE International Conference on, pages 1–6.
IEEE, 2017.
[80] Hazem M Fahmy, Gerd Baumann, Mohamed A Abd El Ghany, and Hassan Mostafa.
V2v-based vehicle risk assessment and control for lane-keeping and collision avoidance.
In Microelectronics (ICM), 2017 29th International Conference on, pages 1–5. IEEE,
2017.
[81] Wanlu Sun, Erik G Ström, Fredrik Brännström, Yutao Sui, and Kin Cheong Sou. D2d-
based v2v communications with latency and reliability constraints. In 2014 IEEE
Globecom Workshops (GC Wkshps), pages 1414–1419. IEEE, 2014.
[82] Hao Ye and Geoffrey Ye Li. Deep reinforcement learning for resource allocation in v2v
communications. In 2018 IEEE International Conference on Communications (ICC),
pages 1–6. IEEE, 2018.
[83] Yanbing Liu, Yuhang Wang, and Guanghui Chang. Efficient privacy-preserving dual
authentication and key agreement scheme for secure v2v communications in an iov
paradigm. IEEE Transactions on Intelligent Transportation Systems, 18(10):2740–2749,
2017.
[84] Arindam Ghosh, Vishnu Vardhan Paranthaman, Glenford Mapp, Orhan Gemikonakli,
and Jonathan Loo. Enabling seamless v2i communications: toward developing coop-
erative automotive applications in vanet systems. IEEE Communications Magazine,
53(12):80–86, 2015.
[85] Yuanguo Bi, Haibo Zhou, Weihua Zhuang, and Hai Zhao. Safety message dissemination
in v2i communication networks. In Safety Message Broadcast in Vehicular Networks,
pages 83–101. Springer, 2017.
[86] Xiao-Feng Xie and Zun-Jing Wang. Siv-dss: Smart in-vehicle decision support system
for driving at signalized intersections with v2i communication. Transportation Research
Part C: Emerging Technologies, 90:181–197, 2018.
[87] Michał Hoeft and Jacek Rak. How to provide fair service for v2i communications in
vanets? Ad Hoc Networks, 37:283–294, 2016.
148 Bibliography
[88] Gerard Aguilar Ubiergo and Wen-Long Jin. Mobility and environment improvement
of signalized networks through vehicle-to-infrastructure (v2i) communications. Trans-
portation Research Part C: Emerging Technologies, 68:70–82, 2016.
[89] Ribal Atallah, Maurice Khabbaz, and Chadi Assi. Multihop v2i communications: A
feasibility study, modeling, and performance analysis. IEEE Transactions on Vehicular
Technology, 66(3):2801–2810, 2017.
[90] Otilia Popescu, Sarwar Sha-Mohammad, Hussein Abdel-Wahab, Dimitrie C Popescu,
and Samy El-Tawab. Automatic incident detection in intelligent transportation systems
using aggregation of traffic parameters collected through v2i communications. IEEE
Intelligent Transportation Systems Magazine, 9(2):64–75, 2017.
[91] I Rashdan, F de Ponte Muller, Wei Wang, M Schmidhammer, and S Sand. Vehicle-to-
pedestrian channel characterization: Wideband measurement campaign and first results.
2018.
[92] Pooya Rahimian, Elizabeth E O’Neal, Shiwen Zhou, Junghum Paul Yon, Luke Franzen,
Jodie M Plumert, and Joseph K Kearney. Vehicle-to-pedestrian (v2p) communications
technology: Do cell phone warnings improve road-crossing safety for texting pedes-
trians? AND30 Standing Committee on Simulation and Measurement of Vehicle and
Operator Performance, 2018.
[93] José Javier Anaya, Pierre Merdrignac, Oyunchimeg Shagdar, Fawzi Nashashibi, and
José E Naranjo. Vehicle to pedestrian communications for protection of vulnerable road
users. In Intelligent Vehicles Symposium Proceedings, 2014 IEEE, pages 1037–1042.
IEEE, 2014.
[94] Pooya Rahimian, Elizabeth E O’Neal, Shiwen Zhou, Jodie M Plumert, and Joseph K
Kearney. Harnessing vehicle-to-pedestrian (v2p) communication technology: Sending
traffic warnings to texting pedestrians. Human Factors, 2018.
[95] Pierre Merdrignac, Oyunchimeg Shagdar, and Fawzi Nashashibi. Fusion of percep-
tion and v2p communication systems for the safety of vulnerable road users. IEEE
Transactions on Intelligent Transportation Systems, 18(7):1740–1751, 2017.
[96] Ahmed Hussein, Fernando García, José María Armingol, and Cristina Olaverri-Monreal.
P2v and v2p communication for pedestrian warning on the basis of autonomous vehicles.
In Intelligent Transportation Systems (ITSC), 2016 IEEE 19th International Conference
on, pages 2034–2039. IEEE, 2016.
[97] Mehrdad Bagheri, Matti Siekkinen, and Jukka K Nurminen. Cellular-based vehicle to
pedestrian (v2p) adaptive communication for collision avoidance. In Connected Vehicles
and Expo (ICCVE), 2014 International Conference on, pages 450–456. IEEE, 2014.
[98] Kai Liu, Joseph KY Ng, Victor CS Lee, Sang H Son, and Ivan Stojmenovic. Cooperative
data scheduling in hybrid vehicular ad hoc networks: Vanet as a software defined network.
2015.
[99] Yuanzhi Ni, Jianping He, Lin Cai, and Yuming Bo. Data uploading in hybrid v2v/v2i
vehicular networks: Modeling and cooperative strategy. IEEE Transactions on Vehicular
Technology, 67(5):4602–4614, 2018.
Bibliography 149
[100] Norbert Bissmeyer, Jan-Felix van Dam, Christian Zimmermann, and Kurt Eckert. Secu-
rity in hybrid vehicular communication based on its-g5, lte-v, and mobile edge computing.
In AmE 2018-Automotive meets Electronics; 9th GMM-Symposium, pages 1–6. VDE,
2018.
[101] Susumu Ishihara, Vince Rabsatt, and Mario Gerla. Secure autonomous platooning with
hybrid vehicular communication. ACM HotMobile, 2015.
[102] Kai Liu, Joseph KY Ng, Victor Lee, Sang H Son, and Ivan Stojmenovic. Cooperative
data scheduling in hybrid vehicular ad hoc networks: Vanet as a software defined network.
IEEE/ACM Transactions on Networking (TON), 24(3):1759–1773, 2016.
[103] Nils Dreyer, Andreas Moller, Zeeshan Hameed Mir, Fethi Filali, and Thomas Kurner.
A data traffic steering algorithm for ieee 802.11 p/lte hybrid vehicular networks. In
Vehicular Technology Conference (VTC-Fall), 2016 IEEE 84th, pages 1–6. IEEE, 2016.
[104] Nils Dreyer, Andreas Moeller, Johannes Baumgarten, Zeeshan Hameed Mir, Thomas
Kuerner, and Fethi Filali. On building realistic reference scenarios for ieee 802.11
p/lte-based vehicular network evaluations. In 2018 IEEE 87th Vehicular Technology
Conference (VTC Spring), pages 1–7. IEEE, 2018.
[105] Wei-dong YANG, LI Pan, Hong-song ZHU, et al. Adaptive tdma slot assignment
protocol for vehicular ad-hoc networks. The Journal of China Universities of Posts and
Telecommunications, 20(1):11–25, 2013.
[106] Tsang-Ling Sheu and Yu-Hung Lin. A cluster-based tdma system for inter-vehicle
communications. J. Inf. Sci. Eng., 30(1):213–231, 2014.
[107] Kazi Atiqur Rahman and Kemal E Tepe. Towards a cross-layer based mac for smooth
v2v and v2i communications for safety applications in dsrc/wave based systems. In 2014
IEEE Intelligent Vehicles Symposium Proceedings, pages 969–973. IEEE, 2014.
[108] Rui Zou, Zishan Liu, Lin Zhang, and Muhammad Kamil. A near collision free reservation
based mac protocol for vanets. In 2014 IEEE Wireless Communications and Networking
Conference (WCNC), pages 1538–1543. IEEE, 2014.
[109] Hassan Aboubakr Omar, Weihua Zhuang, and Li Li. Vemac: A tdma-based mac
protocol for reliable broadcast in vanets. IEEE Transactions on Mobile Computing,
12(9):1724–1736, 2013.
[110] VanDung Nguyen, Duc Ngoc Minh Dang, Sungman Jang, and Choong Seon Hong. e-
vemac: an enhanced vehicular mac protocol to mitigate the exposed terminal problem. In
Network Operations and Management Symposium (APNOMS), 2014 16th Asia-Pacific,
pages 1–4. IEEE, 2014.
[111] Nurullah Shahin and Young-Tak Kim. An enhanced tdma cluster-based mac (etcm) for
multichannel vehicular networks. In Selected Topics in Mobile & Wireless Networking
(MoWNeT), 2016 International Conference on, pages 1–8. IEEE, 2016.
[112] Xiaoxiao Jiang and David Du. Ptmac: A prediction-based tdma mac protocol for
reducing packet collisions in vanet. 2015.
150 Bibliography
[113] Rongqing Zhang, Xiang Cheng, Liuqing Yang, Xia Shen, and Bingli Jiao. A novel
centralized tdma-based scheduling protocol for vehicular networks. IEEE Transactions
on Intelligent Transportation Systems, 16(1):411–416, 2015.
[114] Sailesh Bharati and Weihua Zhuang. Cah-mac: Cooperative adhoc mac for vehicular
networks. IEEE Journal on Selected Areas in Communications, 31(9):470–479, 2013.
[115] Lili Zheng, Yi Wu, Zhexin Xu, and Xiao Lin. A novel mac protocol for vanet based
on improved generalized prime sequence. In Advanced Infocomm Technology (ICAIT),
2014 IEEE 7th International Conference on, pages 1–7. IEEE, 2014.
[116] Hang Yu, Zhiqiang He, and Kai Niu. Stdma for vehicle-to-vehicle communication in
a highway scenario. In Microwave, Antenna, Propagation and EMC Technologies for
Wireless Communications (MAPE), 2013 IEEE 5th International Symposium on, pages
133–138. IEEE, 2013.
[117] Mohammad S Almalag, Stephan Olariu, and Michele C Weigle. Tdma cluster-based mac
for vanets (tc-mac). In World of Wireless, Mobile and Multimedia Networks (WoWMoM),
2012 IEEE International Symposium on a, pages 1–6. IEEE, 2012.
[118] Shanqiang Yi, Wuwen Lai, Di Tang, and Hua Wang. A context-aware mac protocol
in vanet based on bayesian networks. In Communications and Networking in China
(CHINACOM), 2014 9th International Conference on, pages 191–196. IEEE, 2014.
[119] Xin Zhou, Changwen Zheng, Xiaoxin He, et al. Adaptive contention window tuning
for ieee 802.11. In 22nd International Conference on Telecommunications: ICT 2015,
page 74. Engineers Australia, 2015.
[120] Peppino Fazio, Floriano De Rango, and Cesare Sottile. A predictive cross-layered
interference management in a multichannel mac with reactive routing in vanet. 2015.
[121] Hikmat El Ajaltouni, Richard W Pazzi, and Azzedine Boukerche. An efficient qos
mac for ieee 802.11p over cognitive multichannel vehicular networks. In 2012 IEEE
International Conference on Communications (ICC), pages 413–417. IEEE, 2012.
[122] Celimuge Wu, Satoshi Ohzahata, Yusheng Ji, and Toshihiko Kato. A mac protocol for
delay-sensitive vanet applications with self-learning contention scheme. In 2014 IEEE
11th Consumer Communications and Networking Conference (CCNC), pages 438–443.
IEEE, 2014.
[123] Weijie Guo, Liusheng Huang, Long Chen, Hongli Xu, and Jietao Xie. An adaptive
collision-free mac protocol based on tdma for inter-vehicular communication. In Wireless
Communications & Signal Processing (WCSP), 2012 International Conference on, pages
1–6. IEEE, 2012.
[124] Rongqing Zhang, Jinsung Lee, Xia Shen, Xiang Cheng, Liuqing Yang, and Bingli Jiao. A
unified tdma-based scheduling protocol for vehicle-to-infrastructure communications. In
Wireless Communications & Signal Processing (WCSP), 2013 International Conference
on, pages 1–6. IEEE, 2013.
Bibliography 151
[125] Duc Ngoc Minh Dang, Hanh Ngoc Dang, Vandung Nguyen, Zaw Htike, and
Choong Seon Hong. Her-mac: A hybrid efficient and reliable mac for vehicular ad
hoc networks. In 2014 IEEE 28th International Conference on Advanced Information
Networking and Applications, pages 186–193. IEEE, 2014.
[126] Weijie Guo, Liusheng Huang, Long Chen, Hongli Xu, and Chenglin Miao. R-mac:
Risk-aware dynamic mac protocol for vehicular cooperative collision avoidance system.
International Journal of Distributed Sensor Networks, 2013, 2013.
[127] Lin Zhang, Zishan Liu, Rui Zou, Jinjie Guo, and Yu Liu. A scalable csma and self-
organizing tdma mac for ieee 802.11 p/1609. x in vanets. Wireless Personal Communi-
cations, 74(4):1197–1212, 2014.
[128] Alessandro Bazzi, Alberto Zanella, and Barbara Maví Masini. An ofdma-based mac
protocol for next-generation vanets. IEEE Transactions on Vehicular Technology,
64(9):4088–4100, 2015.
[129] Baozhu Li, Xuhui Zhao, Shiyuan Han, and Zhenxiang Chen. New sdn-based architecture
for integrated vehicular cloud computing networking. In 2018 International Conference
on Selected Topics in Mobile and Wireless Networking (MoWNeT), pages 1–4. IEEE,
2018.
[130] Redowan Mahmud, Ramamohanarao Kotagiri, and Rajkumar Buyya. Fog computing:
A taxonomy, survey and future directions. In Internet of everything, pages 103–130.
Springer, 2018.
[131] Behnam Khodapanah, Ahmad Awada, Ingo Viering, David Oehmann, Meryem Sim-
sek, and Gerhard P Fettweis. Fulfillment of service level agreements via slice-aware
radio resource management in 5g networks. In 2018 IEEE 87th Vehicular Technology
Conference (VTC Spring), pages 1–6. IEEE, 2018.
[132] Dizhi Zhou, Nicola Baldo, and Marco Miozzo. Implementation and Validation of LTE
Downlink Schedulers for ns-3. In Simulation Tools and Techniques (SimuTools ’13), 6th
International ICST Conference on, pages 211–218, 2013.
[133] F. Capozzi, G. Piro, L. A. Grieco, G. Boggia, and P. Camarda. Downlink packet
scheduling in LTE cellular networks: Key design issues and a survey. Communications
Surveys and Tutorials, IEEE, 99, 2012.
[134] Akhilesh Pokhariyal, Klaus I Pedersen, Guillaume Monghal, Istvan Z Kovacs, Claudio
Rosa, Troels E Kolding, and Preben E Mogensen. HARQ aware frequency domain packet
scheduler with different degrees of fairness for the UTRAN long term evolution. In
Vehicular Technology Conference, 2007. VTC2007-Spring. IEEE 65th, pages 2761–2765.
IEEE, 2007.
[135] Dizhi Zhou, Wei Song, Nicola Baldo, and Marco Miozzo. Evaluation of TCP perfor-
mance with LTE downlink schedulers in a vehicular environment. In Wireless Communi-
cations and Mobile Computing Conference (IWCMC), 2013 9th International, pages
1064–1069. IEEE, 2013.
152 Bibliography
[136] Huda Adibah Mohd Ramli, Kumbesan Sandrasegaran, Riyaj Basukala, Rachod Patacha-
ianand, and Toyoba Sohana Afrin. Video Streaming Performance Under Well-Known
Packet Scheduling Algorithms. International Journal of Wireless & Mobile Networks
(IJWMN)l, 3:25–38, 2011.
[137] Yasir Zaki, Thushara Weerawardane, Carmelita Gorg, and Andreas Timm-Giel. Multi-
QoS-aware fair scheduling for LTE. In Vehicular technology conference (VTC spring),
2011 IEEE 73rd, pages 1–5. IEEE, 2011.
[138] Li Chen, Bin Wang, Xiaohang Chen, Xin Zhang, and Dacheng Yang. Utility-based
resource allocation for mixed traffic in wireless networks. In Computer Communications
Workshops (INFOCOM WKSHPS), 2011 IEEE Conference on, pages 91–96. IEEE, 2011.
[139] M Andrews, K. Kumaran, K. Ramanan, A. Stolyar, P. Whiting, and R. Vijayakumar.
Providing quality of service over a shared wireless link. Communications Magazine,
IEEE, 39(2):150–154, 2001.
[140] K. Sandrasegaran R. Basukala, H. Mohd Ramli. Performance analysis of EXP/PF
and M-LWDF in downlink 3GPP LTE system. In AH-ICI 2009. 1st Asian Himalayas
International Conference on Internet, pages 1–5. IEEE, 2009.
[141] Bilal Sadiq, Ritesh Madan, and Ashwin Sampath. Downlink scheduling for multi-
class traffic in LTE. EURASIP Journal on Wireless Communications and Networking,
2009(14), 2009.
[142] Giuseppe Piro, Luigi Alfredo Grieco, Gennaro Boggia, Rossella Fortuna, and Pietro
Camarda. Two-level downlink scheduling for real-time multimedia services in LTE
networks. Multimedia, IEEE Transactions on, 13(5):1052–1065, 2011.
[143] Huda Adibah Mohd Ramli, Riyaj Basukala, Kumbesan Sandrasegaran, and Rachod
Patachaianand. Performance of well known packet scheduling algorithms in the downlink
3GPP LTE system. In Communications (MICC), 2009 IEEE 9th Malaysia International
Conference on, pages 815–820. IEEE, 2009.
[144] Sadiq Bilal, Madan Ritesh, and Sampath Ashwin. Downlink scheduling for multiclass
traffic in LTE. EURASIP Journal on Wireless Communications and Networking, 2009,
2009.
[145] Giuseppe Piro, Luigi Alfredo Grieco, Gennaro Boggia, Francesco Capozzi, and Pietro
Camarda. Simulating LTE cellular systems: An open-source framework. Vehicular
Technology, IEEE Transactions on, 60(2):498–513, 2011.
[146] TS 36.322 (V8.8.0): Evolved Universal Terrestrial Radio Access (E-UTRA), Radio Link
Control (RLC) protocol specification(Release 8). Technical Specification, 3GPP, 2010.
[147] Mohamed Hadi Habaebi Mohammed Abdul Jawad M. Al-Shibly and Md. Rafiqul Islam.
Radio resource scheduling in LTE-Advanced system with Carrier Aggregation. ARPN
Journal of Engineering and Applied Sciences, 10(22):17281–17285, 2015.
Bibliography 153
[148] Sangchul Oh, JeeHyeon Na, and Dongseung Kwon. Performance Analysis of Cross
Component Carrier Scheduling in LTE Small Cell Access Point System. In The Second
International Conference on Electrical, Electronics, Computer Engineering and their
Applications (EECEA2015), page 146, 2015.
[149] Alberto Núñez, Jose L Vázquez-Poletti, Agustin C Caminero, Gabriel G Castañé,
Jesus Carretero, and Ignacio M Llorente. iCanCloud: A flexible and scalable cloud
infrastructure simulator. Journal of Grid Computing, 10(1):185–209, 2012.
[150] Dominik Klein and Michael Jarschel. An OpenFlow extension for the OMNeT++ INET
framework. In Proceedings of the 6th International ICST Conference on Simulation
Tools and Techniques, pages 322–329. ICST (Institute for Computer Sciences, Social-
Informatics and Telecommunications Engineering), 2013.
[151] András Varga and Rudolf Hornig. An overview of the OMNeT++ simulation envi-
ronment. In Proceedings of the 1st international conference on Simulation tools and
techniques for communications, networks and systems & workshops, page 60. ICST (In-
stitute for Computer Sciences, Social-Informatics and Telecommunications Engineering),
2008.
[152] TR 36.814 (V9.0.0): Further advancements for E-UTRA physical layer aspects (Release
9). Technical Specification Group Radio Access Network, 3GPP, 2010.
[153] TR 36.942 (V10.2.0): Radio Frequency (RF) system scenarios (Release 10). Technical
Specification, 3GPP, 2011.
[154] Lotfi Asker Zadeh. Fuzzy sets. Information and control, 8(3):338–353, 1965.
[155] Roland Sambuc. Fonctions and floues: application a l’aide au diagnostic en pathologie
thyroidienne. Faculté de Médecine de Marseille, 1975.
[156] Peide Liu and Fang Jin. A multi-attribute group decision-making method based on
weighted geometric aggregation operators of interval-valued trapezoidal fuzzy numbers.
Applied Mathematical Modelling, 36(6):2498–2509, 2012.
[157] Chris Cornelis, Glad Deschrijver, and EE Kerre. Advances and challenges in interval-
valued fuzzy logic. Fuzzy sets and systems, 157(5):622–627, 2006.
[158] Behzad Ashtiani, Farzad Haghighirad, Ahmad Makui, et al. Extension of fuzzy topsis
method based on interval-valued fuzzy sets. Applied Soft Computing, 9(2):457–461,
2009.
[159] Shih-Hua Wei and Shyi-Ming Chen. Fuzzy risk analysis based on interval-valued fuzzy
numbers. Expert Systems with Applications, 36(2):2285–2299, 2009.
[160] Apurba Panda and Madhumangal Pal. A study on pentagonal fuzzy number and its
corresponding matrices. Pacific Science Review B: Humanities and Social Sciences,
Elsevier, 1(3):131–139, 2015.
[161] Mu-Song Chen and Shinn-Wen Wang. Fuzzy clustering analysis for optimizing fuzzy
membership functions. Fuzzy sets and systems, Elsevier, 103(2):239–254, 1999.
154 Bibliography
[162] Marcos E Cintra, Heloisa A Camargo, and Maria Carolina Monard. Genetic generation of
fuzzy systems with rule extraction using formal concept analysis. Information Sciences,
Elsevier, 349:199–215, 2016.
[163] Jin-Shyan Lee and Chih-Lin Teng. An enhanced hierarchical clustering approach for
mobile sensor networks using fuzzy inference systems. IEEE Internet of Things Journal,
4(4):1095–1103, 2017.
[164] Javier Andreu-Perez, Fan Cao, Hani Hagras, and Guang-Zhong Yang. A self-adaptive
online brain machine interface of a humanoid robot through a general type-2 fuzzy
inference system. IEEE Transactions on Fuzzy Systems, 2016.
[165] Chengdong Li, Junlong Gao, Jianqiang Yi, and Guiqing Zhang. Analysis and design
of functionally weighted single-input-rule-modules connected fuzzy inference systems.
IEEE Transactions on Fuzzy Systems, 2016.
[166] Thomas L Saaty. Decision making with dependence and feedback: The analytic network
process. 1996.
[167] İhsan Yüksel and Metin Dagdeviren. Using the analytic network process (anp) in a swot
analysis–a case study for a textile firm. Information Sciences, 177(16):3364–3382, 2007.
[168] Thomas L Saaty and Luis G Vargas. Diagnosis with dependent symptoms: Bayes
theorem and the analytic hierarchy process. Operations Research, 46(4):491–502, 1998.
[169] Tong Wu, Xin-Wang Liu, and Shu-Li Liu. A fuzzy anp with interval type-2 fuzzy
sets approach to evaluate enterprise technological innovation ability. In Fuzzy Systems
(FUZZ-IEEE), 2015 IEEE International Conference on, pages 1–8. IEEE, 2015.
[170] Ching-Lai Hwang and Kwangsun Yoon. Multiple attribute decision making. Springer,
1981.
[171] Emmanouil Skondras, Aggeliki Sgora, Angelos Michalas, and Dimitrios D Vergados.
An analytic network process and trapezoidal interval-valued fuzzy technique for order
preference by similarity to ideal solution network access selection method. International
Journal of Communication Systems, 29(2):307–329, 2016.
[172] Dimitris E Charilas, Ourania I Markaki, John Psarras, and Philip Constantinou. Appli-
cation of fuzzy ahp and electre to network selection. In Mobile Lightweight Wireless
Systems, pages 63–73. Springer, 2009.
[173] Thereza Raquel Machado Azeredo, Helisamara Mota Guedes, Ricardo Alexandre Rebelo
de Almeida, Tânia Couto Machado Chianca, and José Carlos Amado Martins. Efficacy
of the manchester triage system: a systematic review. International emergency nursing,
Elsevier, 23(2):47–52, 2015.
[174] Hoe Tung Yew, Eko Supriyanto, Muhammad H Satria, and YW Hau. Adaptive network
selection mechanism for telecardiology system in developing countries. In Biomedical
and Health Informatics (BHI), 2016 IEEE-EMBS International Conference on, pages
94–97. IEEE, 2016.
Bibliography 155
[175] Driss Aboutajdine Maroua Drissi, Mohammed Oumsis. A fuzzy ahp approach to network
selection improvement in heterogeneous wireless networks. Networked Systems, pages
169–182.
[176] Open street map (osm). https://www.openstreetmap.org, 2018. Accessed: 2018.
[177] Michael Behrisch, Laura Bieker, Jakob Erdmann, and Daniel Krajzewicz. Sumo–
simulation of urban mobility: an overview. In Proceedings of SIMUL 2011, The Third
International Conference on Advances in System Simulation. ThinkMind, 2011.
[178] Network simulator 3 (ns3). https://www.nsnam.org/, 2018. Accessed: 2018.
[179] E Roszkowska and D Kacprzak. The fuzzy saw and fuzzy topsis procedures based on
ordered fuzzy numbers. Information Sciences, 369:564–584, 2016.
[180] Zhe Ren, Peter Fertl, Qi Liao, Federico Penna, and Slawomir Stanczak. Street-specific
handover optimization for vehicular terminals in future cellular networks. In Vehicular
Technology Conference (VTC Spring), 2013 IEEE 77th, pages 1–5. IEEE, 2013.
[181] David González, Mario García-Lozano, Silvia Ruiz, and Dong Seop Lee. A
metaheuristic-based downlink power allocation for lte/lte-a cellular deployments. Wire-
less Networks, 20(6):1369–1386, 2014.
[182] Soumaya Cherkaoui, Tarik Taleb, and Eugene David Ngangue Ndih. Improved inter-
network handover for highly mobile users and vehicular networks. In Vehicular Technol-
ogy Conference (VTC Spring), 2011 IEEE 73rd, pages 1–5. IEEE, 2011.
[183] Suganthi Evangeline and Vinoth Babu Kumaravelu. Decision process for vertical
handover in vehicular adhoc networks. In Microelectronic Devices, Circuits and Systems
(ICMDCS), 2017 International conference on, pages 1–5. IEEE, 2017.
[184] Francesco Guidolin, Irene Pappalardo, Andrea Zanella, and Michele Zorzi. Context-
aware handover policies in hetnets. IEEE Transactions on Wireless Communications,
15(3):1895–1906, 2016.
[185] Abhijit Sarma, Sandip Chakraborty, and Sukumar Nandi. Deciding handover points
based on context-aware load balancing in a wifi-wimax heterogeneous network environ-
ment. IEEE Transactions on Vehicular Technology, 65(1):348–357, 2016.
[186] Shipra Kapoor, David Grace, and Tim Clarke. A base station selection scheme for
handover in a mobility-aware ultra-dense small cell urban vehicular environment. In
Personal, Indoor, and Mobile Radio Communications (PIMRC), 2017 IEEE 28th Annual
International Symposium on, pages 1–5. IEEE, 2017.
[187] Ali Safa Sadiq, Kamalrulnizam Abu Bakar, Kayhan Zrar Ghafoor, Jaime Lloret, and
Rashid Khokhar. An intelligent vertical handover scheme for audio and video streaming
in heterogeneous vehicular networks. Mobile Networks and Applications, 18(6):879–895,
2013.
[188] Mohamed Lahby, Cherkaoui Leghris, and Abdellah Adib. New multi access selection
method based on mahalanobis distance. Applied Mathematical Sciences, 6(53-56):2745–
2760, 2012.
156 Bibliography
[189] Vishal Gupta. Network selection in 3g-wlan interworking environment using topsis.
In Industrial and Information Systems (ICIIS), 2016 11th International Conference on,
pages 512–517. IEEE, 2016.
[190] Werner Toth and Harald Vacik. A comprehensive uncertainty analysis of the analytic
hierarchy process methodology applied in the context of environmental decision making.
Journal of Multi-Criteria Decision Analysis, 2018.
[191] Mohamed Lahby, Cherkaoui Leghris, and Abdellah Adib. New multi access selection
method using differentiated weight of access interface. In Communications and Infor-
mation Technology (ICCIT), 2012 International Conference on, pages 237–242. IEEE,
2012.
[192] KSS Anupama, S Sri Gowri, and B Prabakara Rao. Application of grey relational
analysis to network selection: A case study. 2017.
[193] Abudukeremu Kadier, Peyman Abdeshahian, Yibadatihan Simayi, Manal Ismail,
Aidil Abdul Hamid, and Mohd Sahaid Kalil. Grey relational analysis for compara-
tive assessment of different cathode materials in microbial electrolysis cells. Energy,
90:1556–1562, 2015.
[194] Wei-Yu Chiu, Gary G Yen, and Teng-Kuei Juan. Minimum manhattan distance approach
to multiple criteria decision making in multiobjective optimization problems. IEEE
Transactions on Evolutionary Computation, 20(6):972–985, 2016.
[195] Novi Sofia Fitriasari, Syifa Afifah Fitriani, and Rosa Ariani Sukamto. Comparison
of weighted product method and technique for order preference by similarity to ideal
solution method: Complexity and accuracy. In Science in Information Technology
(ICSITech), 2017 3rd International Conference on, pages 453–458. IEEE, 2017.
[196] Xingwei Wang, Dapeng Qu, Keqin Li, Hui Cheng, Sajal K Das, Min Huang, Renzheng
Wang, and Shuliu Chen. A flexible and generalized framework for access network
selection in heterogeneous wireless networks. Pervasive and Mobile Computing, 40:556–
576, 2017.
[197] Lei Sun, Hui Tian, and Ping Zhang. Decision-making models for group vertical handover
in vehicular communications. Telecommunication Systems, 50(4):257–266, 2012.
[198] B Farhadinia. Sensitivity analysis in interval-valued trapezoidal fuzzy number linear
programming problems. Applied Mathematical Modelling, 38(1):50–62, 2014.
[199] Huiqiang Wang, Zhendong Wang, Guangsheng Feng, Hongwu LV, Xiaoming Chen, and
Qiang Zhu. Intelligent access selection in cognitive networks: A fuzzy neural network
approach. Journal of Computational Information Systems, 8(21):8877–8884, 2012.
[200] Mun-Suk Kim and SuKyoung Lee. Group-based fast handover for pmipv6-based network
mobility in vehicular networks. In Computer Communications Workshops (INFOCOM
WKSHPS), 2015 IEEE Conference on, pages 113–114. IEEE, 2015.
Bibliography 157
[213] Faisal Riaz, Rehana Asif, Hina Sajid, and Muaz A Niazi. Augmenting autonomous vehic-
ular communication using the appreciation emotion: A mamdani fuzzy inference system
model. In Frontiers of Information Technology (FIT), 13th International Conference on,
pages 178–184. IEEE, 2015.
[214] Johann Marquez-Barja, Carlos T Calafate, Juan-Carlos Cano, and Pietro Manzoni. A
geolocation-based vertical handover decision algorithm for vehicular networks. In Local
Computer Networks (LCN), 2012 IEEE 37th Conference on, pages 360–367. IEEE,
2012.
[215] Johann M Marquez-Barja, Hamed Ahmadi, Sergio M Tornell, Carlos T Calafate, Juan-
Carlos Cano, Pietro Manzoni, and Luiz A DaSilva. Breaking the vehicular wireless
communications barriers: Vertical handover techniques for heterogeneous networks.
IEEE Transactions on Vehicular Technology, 64(12):5878–5890, 2015.
[216] Wei-Kuo Chiang and Shih-Chieh Chien. Deploying the media independent information
service at the edge of next-generation network. In Electronics Information and Emer-
gency Communication (ICEIEC), 2015 5th International Conference on, pages 182–185.
IEEE, 2015.
[217] Mohamed Ben Brahim, Zeeshan Hameed Mir, Wassim Znaidi, Fethi Filali, and Noured-
dine Hamdi. Qos-aware video transmission over hybrid wireless network for connected
vehicles. IEEE Access, 5:8313–8323, 2017.
[218] Rabe Arshad, Hesham ElSawy, Sameh Sorour, Tareq Y Al-Naffouri, and Mohamed-Slim
Alouini. Velocity-aware handover management in two-tier cellular networks. IEEE
Transactions on Wireless Communications, 16(3):1851–1867, 2017.
[219] Lu Zhang, Lu Ge, Xin Su, and Jie Zeng. Fuzzy logic based vertical handover algorithm
for trunking system. In Wireless and Optical Communication Conference (WOCC), 2017
26th, pages 1–5. IEEE, 2017.
[220] Aymen Ben Zineb, Mohamed Ayadi, and Sami Tabbane. Fuzzy madm based vertical
handover algorithm for enhancing network performances. In Software, Telecommuni-
cations and Computer Networks (SoftCOM), 2015 23rd International Conference on,
pages 153–159. IEEE, 2015.
[221] Kang-Won Lee, Won-Kyeong Seo, You-Ze Cho, Jong-Woo Kim, Jin-Soo Park, and
Byoung-Sub Moon. Inter-domain handover scheme using an intermediate mobile
access gateway for seamless service in vehicular networks. International Journal of
Communication Systems, 23(9-10):1127–1144, 2010.
[222] Alexander Magnano, Xin Fei, Azzedine Boukerche, and Antonio AF Loureiro. A novel
predictive handover protocol for mobile ip in vehicular networks. IEEE Transactions on
Vehicular Technology, 65(10):8476–8495, 2016.
[223] Bayrem Triki, Slim Rekhis, and Noureddine Boudriga. Secure and qos-aware sip han-
dover for voip communication in vehicular adhoc networks. In Wireless Communications
and Mobile Computing Conference (IWCMC), 2011 7th International, pages 695–700.
IEEE, 2011.
Bibliography 159
[224] Ashraf A Ali and Khalid Al-Begain. Ip multimedia subsystem and sip signaling perfor-
mance metrics. In Multimedia Services and Applications in Mission Critical Communi-
cation Systems, pages 19–35. IGI Global, 2017.
[225] U Kumaran and RS Shaji. Vertical handover in vehicular ad-hoc network using mul-
tiple parameters. In Control, Instrumentation, Communication and Computational
Technologies (ICCICCT), 2014 International Conference on, pages 1059–1064. IEEE,
2014.
[226] Ming-Chin Chuang and Meng Chang Chen. Nash: Navigation-assisted seamless han-
dover scheme for smart car in ultradense networks. IEEE Transactions on Vehicular
Technology, 67(2):1649–1659, 2018.
[227] Hayoung Oh and Chong-kwon Kim. A robust handover under analysis of unexpected
vehicle behaviors in vehicular ad-hoc network. In Vehicular Technology Conference
(VTC 2010-Spring), 2010 IEEE 71st, pages 1–7. IEEE, 2010.
[228] Hayoung Oh. Mobility-aware video streaming in mimo-capable heterogeneous wireless
networks. Mathematical Problems in Engineering, 2016, 2016.
[229] André Almeida, Nuno Lopes, and Alexandre Santos. Intelligent handover for vehicular
networks. In Software, Telecommunications and Computer Networks (SoftCOM), 2014
22nd International Conference on, pages 298–304. IEEE, 2014.
[230] José Antonio Olivera, Ismael Cortázar, Carolina Pinart, Alberto Los Santos, and Iván
Lequerica. Vanba: a simple handover mechanism for transparent, always-on v2v
communications. In Vehicular Technology Conference, 2009. VTC Spring 2009. IEEE
69th, pages 1–5. IEEE, 2009.
[231] Laurence Banda, Mjumo Mzyece, and Guillaume Nóel. Fast handover management in
ip-based vehicular networks. In Industrial Technology (ICIT), 2013 IEEE International
Conference on, pages 1279–1284. IEEE, 2013.
[232] Jasmine P Valera and Sunguk Lee. Security measures in overcoming mobile ipv6 security
issues. International Journal of Database Theory and Application, 9(7):297–304, 2016.
[233] H. Takahashi and T. Minohara. Enhancing location privacy in mobile ipv6 by using
redundant home agents. In 2012 IEEE International Conference on Pervasive Computing
and Communications Workshops, pages 451–454, March 2012.
[234] Yuqing Zhu, Weili Wu, and Deying Li. Efficient client assignment for client-server
systems. IEEE Transactions on Network and Service Management, 13(4):835–847,
2016.
[235] Luis Diego Briceno, Howard Jay Siegel, Anthony A Maciejewski, Ye Hong, Brad
Lock, Charles Panaccione, Fadi Wedyan, Mohammad Nayeem Teli, and Chen Zhang.
Resource allocation in a client/server system for massive multi-player online games.
IEEE Transactions on Computers, 63(12):3127–3142, 2014.
[236] Hiroshi Nishida and Thinh Nguyen. Optimal client-server assignment for internet
distributed systems. IEEE Transactions on Parallel and Distributed Systems, 24(3):565–
575, 2013.
160 Bibliography
[237] Farkhana Muchtar, Abdul Hanan Abdullah, Suhaidi Hassan, and Farhan Masud. Energy
conservation strategies in host centric networking based manet: A review. Journal of
Network and Computer Applications, 2018.
[238] Christian Dannewitz, Dirk Kutscher, BöRje Ohlman, Stephen Farrell, Bengt Ahlgren,
and Holger Karl. Network of information (netinf)–an information-centric networking
architecture. Computer Communications, Elsevier, 36(7):721–735, 2013.
[239] Hongbin Luo, Hongke Zhang, Moshe Zukerman, and Chunming Qiao. An incrementally
deployable network architecture to support both data-centric and host-centric services.
IEEE Network, 28(4):58–65, 2014.
[240] Bengt Ahlgren, Christian Dannewitz, Claudio Imbrenda, Dirk Kutscher, and Borje
Ohlman. A survey of information-centric networking. IEEE Communications Magazine,
50(7), 2012.
[241] Haolin Guo, Dewan Tanvir Ahmed, and Abdulmotaleb El Saddik. Web services for
vanet: a service oriented architecturefor infotainment system based on mashup using
open apis. In Proceedings of the third ACM international symposium on Design and
analysis of intelligent vehicular networks and applications, pages 61–68. ACM, 2013.
[242] Wei-Tek Tsai, Xin Sun, and Janaka Balasooriya. Service-oriented cloud computing
architecture. In Information Technology: New Generations (ITNG), 2010 Seventh
International Conference on, pages 684–689. IEEE, 2010.
[243] Xiang Sun and Nirwan Ansari. Dynamic resource caching in the iot application layer
for smart cities. IEEE Internet of Things Journal, 5(2):606–613, 2018.
[244] Jyoti Deogirikar and Amarsinh Vidhate. An improved publish-subscribe method in appli-
cation layer protocol for iot. In 2017 International Conference On Smart Technologies
For Smart Nation (SmartTechCon), pages 1070–1075. IEEE, 2017.
[245] Zubaida Alazawi, Saleh Altowaijri, Rashid Mehmood, and Mohmmad B Abdljabar.
Intelligent disaster management system based on cloud-enabled vehicular networks. In
ITS Telecommunications (ITST), 2011 11th International Conference on, pages 361–368.
IEEE, 2011.
[246] Weijing Qi, Qingyang Song, Xiaojie Wang, Lei Guo, and Zhaolong Ning. Sdn-enabled
social-aware clustering in 5g-vanet systems. IEEE Access, 6:28213–28224, 2018.
[247] Hamada Alshaer. An overview of network virtualization and cloud network as a service.
International Journal of Network Management, 25(1):1–30, 2015.
[248] Navid Nikaein, Raymond Knopp, Lionel Gauthier, Eryk Schiller, Torsten Braun, Do-
minique Pichon, Christian Bonnet, Florian Kaltenberger, and Dominique Nussbaum.
Closer to cloud-ran: Ran as a service. In Proceedings of the 21st Annual International
Conference on Mobile Computing and Networking, pages 193–195. ACM, 2015.
[249] Rastin Pries, Hans-Jochen Morper, Nandor Galambosi, and Michael Jarschel. Network
as a service-a demo on 5g network slicing. In Teletraffic Congress (ITC 28), 2016 28th
International, volume 1, pages 209–211. IEEE, 2016.
Bibliography 161
[250] Hiroki Nishiyama, Thuan Ngo, Shoki Oiyama, and Nei Kato. Relay by smart de-
vice: Innovative communications for efficient information sharing among vehicles and
pedestrians. IEEE Vehicular Technology Magazine, 10(4):54–62, 2015.
[251] TS 36.839 version 11.1.0: Mobility enhancements in heterogeneous networks (Release
11). Technical Specification, 3GPP, 2012.
[252] Jae-Wook Lee and Sang-Jo Yoo. Probabilistic path and data capacity based handover
decision for hierarchical macro-and femtocell networks. Mobile Information Systems,
2016, 2016.
[253] Arvind Merwaday and Ismail Güvenç. Handover count based velocity estimation and mo-
bility state detection in dense hetnets. IEEE Transactions on Wireless Communications,
15(7):4673–4688, 2016.
[254] Hellenic telecommunications and post commission (eett). http://keraies.eett.gr/, 2018.
Accessed: 2018.
[255] Raman Kumar Goyal, Sakshi Kaushal, and Arun Kumar Sangaiah. The utility based
non-linear fuzzy ahp optimization model for network selection in heterogeneous wireless
networks. Applied Soft Computing, 2017.
[256] John Riordan. Introduction to combinatorial analysis. Courier Corporation, 2012.
[257] Emmanouil Skondras, Angelos Michalas, and Dimitrios D Vergados. A survey on
medium access control schemes for 5g vehicular cloud computing systems. In 2018
Global Information Infrastructure and Networking Symposium (GIIS), pages 1–5. IEEE,
2018.
[258] Emmanouil Skondras, Angelos Michalas, Aggeliki Sgora, and Dimitrios D Vergados. A
downlink scheduler supporting real time services in lte cellular networks. In 2015 6th
International Conference on Information, Intelligence, Systems and Applications (IISA),
pages 1–6. IEEE, 2015.
[259] Emmanouil Skondras, Angelos Michalas, Angeliki Sgora, and Dimitrios D Vergados. A
qos aware three level scheduler for the lte downlink. In 2015 Wireless Telecommunica-
tions Symposium (WTS), Poster Session, pages 1–2. IEEE, 2015.
[260] Emmanouil Skondras, Angelos Michalas, Aggeliki Sgora, and Dimitrios D Vergados.
Qos-aware scheduling in lte-a networks with sdn control. In 2016 7th International
Conference on Information, Intelligence, Systems & Applications (IISA), pages 1–6.
IEEE, 2016.
[261] Emmanouil Skondras, Angelos Michalas, Nikolaos Tsolis, Aggeliki Sgora, and Dim-
itrios D Vergados. A network selection scheme for healthcare vehicular cloud computing
systems. In 2017 8th International Conference on Information, Intelligence, Systems &
Applications (IISA), pages 1–6. IEEE, 2017.
[262] Emmanouil Skondras, Angelos Michalas, and Dimitrios D Vergados. Mobility manage-
ment on 5g vehicular cloud computing systems. Vehicular Communications, 16:15–44,
2019.
Appendix A
Network Position Geographic Latitude Geographic Longitude Downlink & Uplink Spectrum in MHz (WiMAX 2 Band)
WiMAX Macro 1 1e 37.949444 23.642500 3500-3510 & 3400-3410 (5L)
WiMAX Macro 2 3e 37.947778 23.643333 3510-3520 & 3410-3420 (5L)
WiMAX Macro 3 3j 37.947222 23.647500 3530-3540 & 3430-3440 (5L)
WiMAX Macro 4 3o 37.947222 23.652778 3540-3550 & 3440-3450 (5L)
WiMAX Macro 5 6r 37.944722 23.655000 3590-3600 & 3490-3600 (5L)
WiMAX Macro 6 8g 37.943333 23.645278 3590-3600 & 3490-3600 (5L)
WiMAX Macro 7 11d 37.940833 23.641944 3500-3510 & 3400-3410 (5L)
WiMAX Macro 8 11l 37.940556 23.649722 3530-3540 & 3430-3440 (5L)
WiMAX Macro 9 11n 37.940000 23.651389 3540-3550 & 3440-3450 (5L)
WiMAX Macro 10 11r 37.940278 23.655278 3550-3560 & 3450-3460 (5L)
WiMAX Macro 11 12j 37.939494 23.648333 3510-3520 & 3410-3420 (5L)
WiMAX Macro 12 12u 37.939444 23.657778 3500-3510 & 3400-3410 (5L)
WiMAX Macro 13 14e 37.938056 23.643333 3580-3590 & 3480-3490 (5L)
WiMAX Macro 14 14v 37.937222 23.658611 3510-3520 & 3410-3420 (5L)
WiMAX Macro 15 17b 37.935833 23.639722 3540-3550 & 3440-3450 (5L)
WiMAX Macro 16 18c 37.934722 23.640833 3530-3540 & 3430-3440 (5L)
WiMAX Macro 17 18i 37.934444 23.645278 2305-2315 & 2345-2355 (2)
WiMAX Femto 1 3f 37.947500 23.644444 3520-3530 & 3420-3430 (5L)
WiMAX Femto 2 5g 37.945278 23.644167 3550-3560 & 3450-3460 (5L)
WiMAX Femto 3 6e 37.945000 23.642222 3560-3570 & 3460-3470 (5L)
WiMAX Femto 4 6g 37.944444 23.645278 3570-3580 & 3470-3480 (5L)
WiMAX Femto 5 6j 37.944444 23.647222 3580-3500 & 3480-3490 (5L)
WiMAX Femto 6 7j 37.943889 23.647778 3520-3530 & 3420-3430 (5L)
WiMAX Femto 7 8h 37.942222 23.646111 2305-2315 & 2345-2355 (2)
WiMAX Femto 8 8k 37.942500 23.648056 3550-3560 & 3450-3460 (5L)
WiMAX Femto 9 8n 37.942778 23.651389 3520-3530 & 3420-3430 (5L)
WiMAX Femto 10 9g 37.942222 23.645278 2305-2315 & 2345-2355 (2)
WiMAX Femto 11 9i 37.941667 23.646389 3550-3560 & 3450-3460 (5L)
WiMAX Femto 12 9j 37.941944 23.647500 2305-2315 & 2345-2355 (2)
WiMAX Femto 13 9m 37.942222 23.651111 2305-2315 & 2345-2355 (2)
WiMAX Femto 14 10f 37.941389 23.644167 3520-3530 & 3420-3430 (5L)
WiMAX Femto 15 10h 37.941389 23.645833 3520-3530 & 3420-3430 (5L)
WiMAX Femto 16 10j 37.941389 23.647500 3550-3560 & 3450-3460 (5L)
WiMAX Femto 17 10m 37.941389 23.650000 3520-3530 & 3420-3430 (5L)
WiMAX Femto 18 10s 37.941389 23.655556 3520-3530 & 3420-3430 (5L)
WiMAX Femto 19 11f 37.940278 23.643333 2305-2315 & 2345-2355 (2)
WiMAX Femto 20 11h 37.940556 23.645556 2305-2315 & 2345-2355 (2)
WiMAX Femto 21 11k 37.939722 23.648611 2305-2315 & 2345-2355 (2)
WiMAX Femto 22 12k 37.929167 23.648056 3520-3530 & 3420-3430 (5L)
WiMAX Femto 23 15b 37.937222 23.640000 2305-2315 & 2345-2355 (2)
WiMAX Femto 24 15d 37.936944 23.641667 3520-3530 & 3420-3430 (5L)
WiMAX Femto 25 15u 37.936944 23.657778 2305-2315 & 2345-2355 (2)
Network Position Geographic Latitude Geographic Longitude Spectrum in MHz (DSRC Europe Band)
WAVE 1 2f 37.948056 23.644444 5915-5925 (SCH4)
WAVE 2 2h 37.947500 23.646389 5905-5915 (SCH3)
WAVE 3 3m 37.946667 23.650278 5895-5905 (SCH2)
WAVE 4 4e 37.946111 23.642500 5895-5905 (SCH2)
WAVE 5 4u 37.945556 23.658056 5875-5885 (SCH1)
WAVE 6 5i 37.945000 23.646944 5915-5925 (SCH4)
WAVE 7 6k 37.944167 23.648056 5905-5915 (SCH3)
WAVE 8 7e 37.943056 23.643333 5875-5885 (SCH1)
WAVE 9 7i 37.943889 23.646667 5895-5905 (SCH2)
WAVE 10 7n 34.943889 23.651111 5875-5885 (SCH1)
WAVE 11 9h 37.942222 23.646389 5915-5925 (SCH4)
WAVE 12 10e 37.941667 23.643333 5895-5905 (SCH2)
WAVE 13 10j 37.941667 23.647500 5875-5885 (SCH1)
WAVE 14 10n 37.941111 23.651667 5905-5915 (SCH3)
WAVE 15 11h 37.940556 23.645556 5905-5915 (SCH3)
WAVE 16 12c 37.939444 23.641389 5875-5885 (SCH1)
WAVE 17 12q 37.938889 23.654444 5875-5885 (SCH1)
WAVE 18 14b 37.938056 23.640556 5915-5925 (SCH4)
WAVE 19 14f 37.937778 23.644167 5905-5915 (SCH3)
WAVE 20 17c 37.935000 23.641389 5895-5905 (SCH2)
WAVE 21 17n 37.935278 23.651389 5895-5905 (SCH2)
WAVE 22 18i 37.934722 23.645278 5875-5885 (SCH1)
Appendix B
SLA 1
NAV VoIP CV BS Web 166
Network
Price
Price
Price
Price
Price
Jitter
Jitter
Jitter
Jitter
Jitter
Delay
Delay
Delay
Delay
Delay
Security
Security
Security
Security
Security
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Throughput
Throughput
Throughput
Throughput
Throughput
Service Reliability
Service Reliability
Service Reliability
Service Reliability
Service Reliability
LTE Macro 1 AG AG G VG MG G AP AG AG G VG G VG VP G VG G MG MG M AP M MG AG M G AG P AG AG G VG VG AG AP
LTE Macro 2 VG AG G G AG AG AP M MG MG M G M P AG AG M VG M MG AP MG G MG M MG AG VP VG AG G G G AG P
LTE Macro 3 VG AG M G M G VP M MG MG M MG VG VP MG G G M VG MG VP G VG MG MG VG M P VG AG G G M AG VP
LTE Macro 4 M MG M M VG AG P M MG MG M G G P VG AG AG G VG AG P AG AG M VG MG G P MG G MG M G G VP
LTE Macro 5 AG AG VG VG VG AG VP AG AG VG VG M VG AP G VG AG MG AG M AP MG G VG MG AG G VP G VG AG MG M G VP
LTE Macro 6 M MG G M VG G P MG G MG M VG MG AP VG AG AG G G MG VP G G VG MG VG VG P AG AG VG VG VG AG VP
LTE Macro 7 MG G MG M AG G VP MG G G M M VG VP MG G G M M MG P VG AG AG G G MG P AG AG G VG G VG P
LTE Macro 8 MG G AG MG AG M P AG AG G VG AG AG AP AG AG AG VG AG G P G G M MG AG AG P G VG MG MG VG AG AP
LTE Macro 9 M MG VG M AG M VP M MG AG M MG M AP G VG G MG MG VG P VG AG VG G VG VG P M MG VG M AG AG P
LTE Macro 10 MG G AG MG MG MG AP AG AG M VG VG M P G VG G MG M AG VP G VG G MG AG AG P G VG MG MG G G VP
LTE Macro 11 M MG VG M VG MG VP M MG MG M AG MG VP G VG G MG M VG AP M MG MG M G G P G G MG MG G VG AP
LTE Macro 12 M MG M M VG G AP VG AG AG G G MG AP VG AG G G AG G VP AG AG MG VG G VG AP AG AG VG VG MG MG P
LTE Macro 13 VG AG MG G G MG AP M MG M M MG MG P AG AG VG VG G M P AG AG VG VG MG G AP VG AG VG G VG AG AP
LTE Macro 14 VG AG AG G MG M AP AG AG VG VG VG G P VG AG G G G AG VP VG AG G G G M AP G VG G MG G AG AP
LTE Macro 15 M MG MG M G VG P AG AG G VG G M P M MG G M VG VG AP M MG AG M VG AG VP AG AG M VG VG G AP
LTE Macro 16 AG AG MG VG AG VG VP VG AG VG G MG MG P VG AG AG G AG MG VP G G MG MG AG M P G VG MG MG G VG P
LTE Macro 17 G VG MG MG MG M VP AG AG AG VG AG M P G VG VG MG MG AG P G VG VG MG G AG AP VG AG VG G G MG AP
LTE Macro 18 MG G MG M G G VP VG AG G G VG VG VP G VG G MG AG MG AP MG G MG M VG MG AP M MG AG M G G VP
LTE Macro 19 AG AG MG VG AG VG VP MG G MG M G M VP G VG AG MG G G P G VG M MG VG MG AP VG AG VG G VG VG P
LTE Macro 20 M MG MG M AG AG VP MG G G M M G P MG G G M AG MG VP G VG G MG MG M AP MG G AG M VG VG P
LTE Femto 1 MG G AG M MG M P VG AG M G VG M VP MG G VG M VG AG AP VG AG M G G VG VP G VG MG MG AG G VP
LTE Femto 2 VG AG M G M M AP AG AG G VG M M AP MG G M M VG MG AP MG G M M VG MG P G VG MG MG G MG P
LTE Femto 3 AG AG MG VG VG MG P AG AG G VG AG VG VP G VG G MG VG G P AG AG VG VG G G VP VG AG MG G MG VG AP
LTE Femto 4 MG G AG M VG AG VP G VG VG MG VG AG VP G VG G MG AG MG VP MG G G M M AG AP AG AG G VG MG G VP
LTE Femto 5 MG G M M G G VP M MG MG M AG M P VG AG AG G VG AG AP VG AG AG G G VG P G VG VG MG AG VG P
LTE Femto 6 AG AG M VG VG VG AP AG AG AG VG MG M P M MG AG M G VG AP M MG MG M VG AG AP MG G G M MG AG P
LTE Femto 7 MG G M M M MG AP VG AG VG G VG M VP AG AG G VG MG MG AP MG G MG M VG VG P AG AG G VG AG G P
LTE Femto 8 M MG AG M MG M VP AG AG M VG M M VP AG AG AG VG G AG P G VG G MG MG M P AG AG AG VG VG G AP
LTE Femto 9 G VG M MG AG VG VP MG G G M M MG AP G VG MG MG MG MG AP MG G G M MG AG P M MG MG M MG AG AP
LTE Femto 10 VG AG AG G G M AP VG AG VG G VG VG VP AG AG VG VG MG AG AP VG AG M G AG MG AP VG AG VG G G AG AP
LTE Femto 11 AG AG G VG VG AG P AG AG G VG M MG AP G G M MG M M P AG AG G VG MG M P AG AG G VG MG VG VP
LTE Femto 12 MG G M M VG MG P M MG VG M VG AG AP MG G AG M G G AP AG AG MG VG G G AP M MG G M VG VG P
LTE Femto 13 MG G MG M M M P VG AG M G M M VP AG AG M VG AG G VP MG G AG M G MG VP AG AG G VG G MG AP
LTE Femto 14 G VG M MG M AG VP MG G G M AG G VP VG AG VG G VG MG VP AG AG M VG MG G VP G VG MG MG G AG AP
LTE Femto 15 MG G G M G G AP MG G G M AG MG P M MG AG M AG M VP MG G MG M G AG AP VG AG MG G MG MG VP
LTE Femto 16 VG AG M G VG MG VP AG AG M VG AG G VP G VG MG MG G MG P G G MG G AG VG VP AG AG M VG M AG AP
LTE Femto 17 VG AG M G AG VG P VG AG AG G VG MG VP MG G AG M VG VG P AG AG VG VG G M AP G G MG MG G VG AP
LTE Femto 18 MG G M M VG G P G VG M MG MG G VP VG AG MG G AG MG P G VG MG MG VG MG AP G VG VG MG VG AG VP
LTE Femto 19 VG AG G G AG M P VG AG AG G G AG AP AG AG M VG VG M AP VG AG G G G G P VG AG MG G MG G P
LTE Femto 20 G VG G MG MG AG AP M MG MG M MG M VP VG AG AG G MG MG AP VG AG VG G G G VP MG G G M G MG VP
LTE Femto 21 M MG G M VG AG VP M MG M M MG AG AP G VG G MG MG VG AP M MG G M MG M AP VG AG MG G VG MG P
LTE Femto 22 AG AG MG VG VG AG VP G VG MG MG AG VG VP AG AG AG VG VG AG P M MG VG M G G P MG G VG M M M VP
LTE Femto 23 M MG G M M AG P AG AG M VG MG AG AP AG AG MG VG G MG P G VG AG MG VG G AP M MG M M AG G P
LTE Femto 24 MG G MG M AG AG AP AG AG MG VG M M VP MG G MG M G G AP M MG VG M M G P VG AG MG G AG G P
LTE Femto 25 G VG AG MG G MG AP G VG G MG MG MG VP AG AG AG VG VG VG P VG AG AG G G G AP AG AG AG VG VG AG VP
LTE Femto 26 G VG G MG M G VP MG G AG M G G VP AG AG M VG G G VP G VG G MG VG G VP M MG VG M G VG AP
LTE Femto 27 MG G G M G M AP VG AG G G VG AG P MG G G MG AG AG VP G VG AG MG G VG P AG AG AG VG VG G AP
LTE Femto 28 M MG G M MG G VP MG G G M G G VP MG G AG M G G AP G G VG MG G MG VP G VG VG MG G MG P
LTE Femto 29 VG AG M G M MG AP AG AG VG VG VG AG P G VG MG MG MG M VP AG AG AG VG MG MG VP M MG VG M MG G AP
LTE Femto 30 AG AG AG VG M AG VP G VG M MG MG MG AP MG G G M G G VP AG AG G VG G AG P M MG M M G G VP
LTE Femto 31 VG AG AG G AG MG P AG AG VG VG VG AG AP MG G G M G M AP M MG MG M MG M AP AG AG G VG AG G AP
LTE Femto 32 VG AG AG G VG MG AP G VG M MG G MG VP VG AG VG G M MG VP MG G M M MG G AP M MG M M VG MG VP
The available networks per SLA
LTE Femto 33 AG AG MG VG AG MG P G VG M MG M VG VP VG AG AG G G AG P VG AG G G VG VG VP G VG VG MG G VG P
LTE Femto 34 AG AG G VG MG MG VP VG AG MG G MG MG P VG AG M G G MG VP M MG VG M AG AG VP MG G VG M G M P
LTE Femto 35 VG AG G G MG M VP G VG M MG AG VG P VG AG VG G MG MG P VG AG G G MG AG P G VG G MG VG M P
LTE Femto 36 G VG M MG MG G AP MG G VG M MG AG VP VG AG G G AG MG AP MG G VG M AG MG P G VG MG MG M G AP
LTE Femto 37 G VG AG MG M MG P AG AG VG VG VG AG AP AG AG VG VG M G AP AG AG AG VG MG MG AP G VG VG MG VG MG P
Table B2 The available WiMAX and WAVE networks of SLA 1.
SLA 1
NAV VoIP CV BS Web
Network
Price
Price
Price
Price
Price
Jitter
Jitter
Jitter
Jitter
Jitter
Delay
Delay
Delay
Delay
Delay
Security
Security
Security
Security
Security
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Throughput
Throughput
Throughput
Throughput
Throughput
Service Reliability
Service Reliability
Service Reliability
Service Reliability
Service Reliability
WiMAX Macro 1 MG G G M G VG P G VG VG MG VG G P M MG M M VG M P G G VG MG G MG AP MG G M M AG G AP
WiMAX Macro 2 MG G M M MG G P MG G VG M MG G P AG AG G VG AG MG AP AG AG VG VG MG M AP G VG AG MG AG M P
WiMAX Macro 3 G VG MG MG G G VP MG G M M VG M P VG AG MG G G M P M MG G M MG M VP VG AG MG G AG M VP
WiMAX Macro 4 MG G M M AG VG VP MG G M M AG MG AP VG AG MG G VG M VP AG AG AG VG MG G P G VG M MG G AG AP
WiMAX Macro 5 AG AG AG VG MG M VP VG AG G G MG AG P VG AG M G AG M AP MG G VG M G G P AG AG G VG M G VP
WiMAX Macro 6 M MG MG M G M P MG G VG M VG VG AP VG AG G G VG M P G VG MG MG VG VG AP G VG G MG VG VG VP
WiMAX Macro 7 AG AG G VG M VG AP AG AG MG VG VG G P G VG MG MG VG M AP G VG M MG AG G AP MG G M M AG G AP
WiMAX Macro 8 M MG VG M AG MG VP VG AG MG G MG VG AP M MG VG M G G VP MG G G M G AG VP G VG MG MG M MG P
WiMAX Macro 9 AG AG M VG M MG P M MG M M AG VG AP AG AG G VG G M VP MG G MG M G M AP M MG MG M AG G VP
WiMAX Macro 10 MG G M M VG G P AG AG AG VG MG G VP AG AG G VG G MG VP G VG M MG M VG AP VG AG AG G G VG VP
WiMAX Macro 11 M MG G M VG MG AP MG G AG M G MG P G VG M MG G M AP G VG VG MG VG G P M MG VG M VG G VP
WiMAX Macro 12 G VG AG MG G M P AG AG VG VG G G P VG AG M G G AG AP VG AG G G G MG P VG AG G G M MG P
WiMAX Macro 13 G VG MG MG AG M VP M MG VG M M G P AG AG VG VG VG G VP M MG AG M AG G AP G VG G MG AG AG VP
WiMAX Macro 14 AG AG AG VG G M P M MG M M M VG P VG AG G G AG VG P MG G G M M VG AP MG G MG M G MG VP
WiMAX Macro 15 VG AG VG G MG G AP AG AG G VG G M VP MG G VG M VG MG P M MG M M AG G P G VG G MG VG MG AP
WiMAX Macro 16 M MG M M VG MG AP VG AG MG G M VG AP M MG M M MG MG AP G G VG MG VG G P AG AG AG VG G AG AP
WiMAX Macro 17 M MG G M G AG AP M MG MG M AG AG VP MG G G M AG VG VP AG AG VG VG G MG AP VG AG VG G VG MG VP
WiMAX Femto 1 VG AG AG G G MG AP M MG M M VG MG P MG G MG M G MG AP VG AG G G AG G AP G VG VG MG VG G P
WiMAX Femto 2 M MG G M MG AG VP G VG AG MG AG AG P M MG MG M VG VG VP G VG M MG M VG VP MG G VG M G VG P
WiMAX Femto 3 M MG G M AG G AP M MG AG M G M AP VG AG VG G G G AP MG G VG M M AG VP M MG MG M AG AG AP
WiMAX Femto 4 MG G VG M MG VG P AG AG AG VG VG MG P G VG G MG MG MG VP M MG M M M MG P G G G MG G M AP
WiMAX Femto 5 AG AG G VG G M VP VG AG MG G MG AG VP VG AG AG G AG VG VP VG AG M G VG VG P MG G M M MG M AP
WiMAX Femto 6 M MG MG M G VG AP MG G AG M M MG P VG AG G G MG M AP MG G VG M G G VP AG AG G VG VG VG VP
WiMAX Femto 7 MG G G M MG M VP M MG M M VG VG P G VG VG MG G VG P M MG G M G VG P AG AG VG VG G AG P
WiMAX Femto 8 G VG G MG G MG AP VG AG M G G VG P AG AG MG VG AG G P VG AG AG G MG AG VP AG AG VG VG VG MG AP
WiMAX Femto 9 MG G MG M VG M P M MG M M AG G AP MG G G M G MG VP M MG VG M AG M AP MG G MG M G G AP
WiMAX Femto 10 AG AG AG VG M AG AP G VG AG MG M MG AP G VG M MG G AG VP VG AG G G VG G P AG AG AG VG MG MG P
WiMAX Femto 11 MG G G M G AG P VG AG AG G VG G P AG AG AG VG MG G P AG AG VG VG G VG AP AG AG AG VG VG AG VP
WiMAX Femto 12 G VG G MG AG VG P M MG MG M M MG AP MG G MG M VG G P G VG G MG VG AG VP G VG MG MG G M P
WiMAX Femto 13 AG AG MG VG VG VG VP M MG AG M G M AP G VG VG MG G VG P AG AG VG VG G VG VP VG AG MG G MG M AP
WiMAX Femto 14 AG AG VG VG M G AP AG AG M VG M M AP VG AG MG G MG G VP AG AG VG VG VG G AP MG G G M VG G AP
WiMAX Femto 15 AG AG G VG G MG VP MG G VG M G MG VP G G MG MG G AG AP AG AG M VG MG G P G VG M MG MG MG P
WiMAX Femto 16 M MG M M MG M VP G VG M MG G M AP VG G MG G G AG P VG AG MG G AG VG P AG AG AG VG AG G P
WiMAX Femto 17 M MG G M MG M AP G VG VG MG M M AP G VG AG MG AG VG P MG G VG M VG AG VP VG AG AG G AG MG P
WiMAX Femto 18 G VG AG MG M VG VP AG AG VG VG VG M VP VG AG G G MG G P M MG AG M G VG P VG AG VG G AG M VP
WiMAX Femto 19 M MG M M G M VP AG AG M VG AG M VP AG AG G VG G M AP G VG MG MG G MG AP VG AG M G G VG P
WiMAX Femto 20 G VG AG MG G AG VP M MG M M M MG AP G VG MG MG M AG P MG G G M VG G P M MG AG M VG VG AP
WiMAX Femto 21 MG G M M MG AG P M MG M M M AG P M MG AG M M AG P G VG M MG VG G AP VG AG G G G G AP
WiMAX Femto 22 VG AG AG G MG M AP G VG M MG M AG VP G VG G MG VG M AP M MG VG M M VG VP MG G M M G MG AP
WiMAX Femto 23 MG G G M MG M AP VG AG AG G AG M VP MG G MG M G MG AP MG G G M G AG AP VG AG G G VG AG VP
WiMAX Femto 24 G VG M MG VG G VP AG AG AG VG MG MG AP G VG MG MG MG M VP G VG AG MG G MG AP VG AG VG G G G P
WiMAX Femto 25 MG G VG M MG VG AP VG AG G G VG G P MG G G M AG MG P VG AG AG G MG MG AP MG G AG M VG MG AP
WAVE 1 AG AG G VG VG VG VP G VG VG MG AG AG P MG G M M VG G P MG G G M VG AG P VG AG G G AG M VP
WAVE 2 VG AG MG G AG MG VP MG G MG M G M P VG AG AG G G M VP M MG VG M G VG AP M MG VG M VG MG P
WAVE 3 VG AG M G M VG AP M MG G M VG VG VP AG AG MG VG VG MG VP MG G M M VG G P MG G G G M VG VP
WAVE 4 AG AG AG VG AG AG P G VG M MG MG MG VP G G M MG AG G VP AG AG G VG AG MG VP G VG MG MG G VG AP
WAVE 5 M MG M M AG VG AP MG G VG M VG AG AP G VG VG MG MG MG AP AG AG G VG MG MG P M MG AG M M G AP
WAVE 6 M MG VG MG MG M VP G VG AG MG AG VG AP M MG G M MG G VP AG AG M VG G VG AP MG G G M G MG AP
WAVE 7 M MG AG M MG VG VP VG AG MG G M G P VG AG VG G AG G VP AG AG VG VG G G P VG AG G G AG VG AP
WAVE 8 MG G G MG MG MG VP M MG M M AG MG P VG AG AG G M MG AP AG AG G VG M MG P VG AG AG G G VG AP
WAVE 9 MG G G MG G M P M MG M M VG MG VP M MG M M VG VG P AG AG M VG VG G AP M MG G M AG G P
WAVE 10 MG G G M AG VG P M MG VG M AG M VP AG AG M VG MG G VP M MG G M MG G VP MG G VG M G AG AP
WAVE 11 M MG G M G AG AP VG AG G G M M VP G VG G MG MG MG VP AG AG G VG AG MG VP VG AG G G VG M P
WAVE 12 AG AG MG VG M VG VP AG AG VG VG VG G VP M MG VG M G VG VP VG AG MG G MG VG AP AG AG M VG G AG AP
WAVE 13 M MG G M G M VP MG G VG M VG MG P AG AG G VG VG G VP VG AG MG G AG G AP AG AG VG VG VG VG P
167
WAVE 14 AG AG AG VG G VG AP G VG VG MG G M AP G G G MG VG AG P M MG G M G MG P MG G MG M AG G VP
WAVE 15 G VG G MG M MG VP AG AG VG VG G MG AP AG AG G VG AG M VP M MG VG M G MG P MG G G M G AG VP
WAVE 16 VG AG MG M VG G VP G VG MG MG VG AG VP G VG AG MG G G AP VG AG AG G MG G AP G VG G MG AG G AP
WAVE 17 M MG VG M VG VG P VG AG AG G M MG VP M MG G M AG MG VP VG AG MG G MG M AP VG AG G G AG VG P
WAVE 18 G VG G MG MG AG AP MG G M M VG VG AP MG G M M VG G VP AG AG G VG AG G P M MG AG M VG G P
WAVE 19 VG AG MG G MG M P G VG AG MG VG G AP VG AG G G AG G P MG G AG M MG VG AP AG AG AG VG M AG AP
WAVE 20 M MG M M G G AP MG G AG M G MG VP VG AG MG G AG AG AP M MG MG M VG AG P M MG VG M AG AG VP
WAVE 21 M MG VG M VG AG P G VG AG MG MG G AP G VG M MG MG MG P G VG G MG AG VG VP G VG M MG MG G P
WAVE 22 MG G G MG MG G VP MG G G M G AG AP AG AG MG VG MG AG VP VG AG G G G MG VP AG AG G VG G AG AP
Table B3 The available LTE networks of SLA 2, SLA 3 and SLA 4.
Price
Price
Price
Price
Price
Price
Jitter
Jitter
Jitter
Jitter
Jitter
Jitter
Delay
Delay
Delay
Delay
Delay
Delay
Secutiy
Security
Security
Security
Security
Security
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Throughput
Throughput
Throughput
Throughput
Throughput
Throughput
Service Reliability
Service Reliability
Service Reliability
Service Reliability
Service Reliability
LTE Macro 1 G G G MG MG P P MP M MG P G MP P G G G MG G MP P MP M MG P M MP M P MP MG P MG MP M VP P VP AP Service Reliability
VP P VG
LTE Macro 2 M MG P MP M MP M MG G M M P MG M G G MP MG MP MG P P MP MP P P MP M M MG MP MP MP MG MG M M AP MP MP P G
LTE Macro 3 MG G G M G MG MP P MP M P P P M MG G P M M P M P MP P P P P M MG MG P M M P G AP VP VP AP M VP AG
LTE Macro 4 M MG P MP P MP MP G G P MG P G P MG G M M G M P MP M P P P P M MG MG MP M MG P M M M MP MP VP P AG
LTE Macro 5 MP M M P P M M MG G G M MG G M MG G P M M G M MG MG MG M P MP M M MG P MP M MP M AP VP P AP AP P AG
LTE Macro 6 MP M MP P G MG M G G G MG MP G M M MG G MP MG MP MP MG MG MP M MP P M M MG MP MP MP P G P MP MP VP VP P G
LTE Macro 7 MG G G M P M P M MG MG MP M P M M MG P MP G M P MP M P P MP P MG MP M P P P P M AP VP VP AP AP VP VG
LTE Macro 8 G G P MG P MP M G G P MG MG G MP M MG M MP G M P P MP P MP P M G M MG P MP MP MP M P MP AP VP VP MP AG
LTE Macro 9 MG G P M M G MP MP M MG P G MP P M MG MG MP MG MG P P MP MP VP M P G P MP M VP P MP G P MP P VP P VP AG
LTE Macro 10 G G G MG M M M MG G G M MP MG M P MP P P G M MP MP M P P MP P G P MP P P M MP MG P MP VP P AP MP VG
LTE Macro 11 M MG MP MP P MG MP P MP MG P G P M G G P MG G G M P MP MG P P P MG M MG P MP MP P G P MP VP VP MP P AG
LTE Macro 12 MP M G P MP G M MP M MG P M MG MP MP M P P MG MG P MP M MG P M MP MG MP M P P MP M M MP M VP P VP AP VG
LTE Macro 13 G G G MG G MP MP P MP MG P MG G M P MP P MP P P MP P MP MG P MP M M P MP P VP P P G P MP P VP P VP VG
LTE Macro 14 M MG MG MP G P M G G G MG P M MP P MP G P G G P M MG M MP P MP M P MP MG P MP P G P MP M VP MP AP G
LTE Macro 15 M MG MG MP G MP P P MP M P MG MP MP MG G P M P P P P MP M P P P M MP M P P P VP G MP M P P P VP VG
LTE Macro 16 MG G MP M P MG P G G MG MG MG P M M MG MG MP M MP M P MP M P P P M M MG P MP M P M M M P MP MP P G
LTE Macro 17 MP M P P P G M P MP P P P MG P M MG P MP MP MG P P MP P P P M G P MP P P MP M MG P MP AP VP VP VP G
LTE Macro 18 MG G M M MP M MP MG G MG M P MP M M MG G MP M G P P MP P P P MP G P MP P VP MP P G AP VP P AP P AP VG
LTE Macro 19 MP M MP P M P M G G MP MG MG P M MP M MG P MP MG M MP M MP MP MP P M MP M MP P MP MP G MP M P P P MP AG
LTE Macro 20 MG G MG M M MG M M MG G MP MG M MP MG G G M G M M MP M P P P M G MP M M P P M MG P MP AP P AP AP AG
LTE Femto 1 P MP G P G M MP G G M MG MG P P MP M P P M P P P MP P P MG P MG MP M P P M P M MP M VP P AP VP G
LTE Femto 2 P MP MP P M MG MP M MG P MP P MG MP MP M P MP G M M M MG P MP P MP G P MP P VP P P G P MP AP VP AP VP G
LTE Femto 3 G G MG MG M G M MG G G M MP P P MP M MP P P M M P MP P P MP P MG MP M MP P P P M VP P MP AP P AP VG
LTE Femto 4 G G P MG G M P MG G MG M P MP P MG G M M P G MP MP M MG P P P G MP M MP P P P G MP M MP P VP P VG
LTE Femto 5 G G M MG P MP M MG G MG M MG P M P MP G P G M P P MP P P MP P G P MP MP P MG P G P MP MP VP P VP VG
LTE Femto 6 P MP M P G P P P MP MG P M G MP MP M G P P M MP P MP MG P P P MG MP M MG P P MP MG MP M AP P VP MP G
LTE Femto 7 M MG P MP M M MP MG G MG M MP P P MG G G M MP M M M MG MP MP P P M P MP P P P P G P MP P VP VP P VG
LTE Femto 8 G G MP MG MP M P P MP G P MG MP P G G MP MG G M P P MP P P MP P G MG MG MP M P P G M M P MP P VP VG
LTE Femto 9 MG G M M MG M P MG G M M MG MP M M MG M MP MP MG M P MP M P MP MP G M MG M MP MP M G MP M MP P MP MP G
LTE Femto 10 P MP MG P MP MG M G G P MG MP MP MP MP M G P G G P MP M P P MP MP M MP M P P M MG G P MP P P P P AG
LTE Femto 11 G G P MG P MP M MP M G P MG M P MG G M M M MG P MP M P P P M M P MP M P P P MG P P P P VP VP AG
LTE Femto 12 P MP M P MP P MP MP M MG P G P MP MP M G P MG M M P MP M P M P MG P MP P VP MP MP M P MP VP VP VP MP G
LTE Femto 13 M MG P MP MP M MP MG G G M MP P MP MP M MG P G MG P MG MG MG M MP P G MP M M P MG MP M P MP VP VP VP VP VG
LTE Femto 14 MG G G M MG MP MP MG G P M MG M P P MP MP P MP G P P MP P P M MP G P MP MP P MP MP M VP P VP P VP MP VG
LTE Femto 15 P MP G P G P MP MG G M M MG M P MG G MG M MP M M MG MG P M MP MP MG P MP MP P MP M M P MP VP VP VP M VG
LTE Femto 16 M MG P MP G MG P MG G MG M M G M G G M MG MP M P P MP MP P P P M MG MG MP M MP MP M MP M MP P MP AP VG
LTE Femto 17 MG G MG M G MG M MG G P M MG MG P P MP P P M G MP MG MG P M MG MP MG P MP P VP MP MG G P MP AP VP MP M G
LTE Femto 18 MP M P P G MG M P MP MG P MG MP M MG G M M MG M M P MP MP P M MP G MG MG M M MG P MG M M AP MP M VP AG
LTE Femto 19 G G M MG G P MP MP M MP P G MG MP G G M MG P MG MP MP M P P M MG MG MG MG M M P MP M M M M MP AP MP AG
LTE Femto 20 M MG M MP MG P MP MP M M P G MP MP MG G MG M MG MG M MP M M P P MP G P MP M P M P M VP P AP AP M VP G
LTE Femto 21 P MP P P P M P P MP G P MG MP MP MG G M M G M M P MP MP VP MP MP MG MG MG M M M M M M M MP MP P AP G
LTE Femto 22 MG G G M MP MG P P MP MG P G MP MP MG G G M M M MP P MP MP P P P MG MG MG M M M M G AP VP M AP AP P AG
LTE Femto 23 MG G M M M M MP MG G G M M M M M MG M MP G P P MP M M P M P G M MG P MP M P MG MP M VP P P P AG
LTE Femto 24 M MG M MP G G M MP M M P P MG M MG G P M MP P M MP M M P P P G M MG P MP P P M M M VP MP P VP G
LTE Femto 25 G G G MG G MG MP P MP MG P MG MG MP M MG MG MP G G P P MP MP VP MG P G M MG M MP MP MG M P MP M VP VP VP VG
LTE Femto 26 MG G M M MG P M MG G MG M MP MP MP G G MP MG M M M P MP MG P P MP M M MG M MP P MP G M M M MP P P G
LTE Femto 27 MG G G M G G MP P MP P P MG P MP G G M MG P P M VP P P VP M P M M MG M MP P P MG MP M AP P VP VP AG
LTE Femto 28 P MP MP P M P P G G MP MG MP P MP M MG P MP MG M MP MG MG P M P P M M MG P MP P M M P MP AP VP AP AP G
LTE Femto 29 G G P MG P P MP G G G MG M MP MP MP M P P M M MP MG MG M M MG M M MP M P P M M M MP M P P P VP AG
LTE Femto 30 G G M MG MG P P P MP M P MG M M P MP G P M MP M MP M MG P M MP MG M MG MG MP MP MP MG MP M M P MP MP VG
LTE Femto 31 M MG G MP G P P M MG P MP P MP MP MP M G P MP MP P MP M P P P MP MG MP M P P MP P M VP P AP VP VP P G
LTE Femto 32 M MG G MP MP MG M MP M MP P P MG MP M MG MP MP MG P P MP M MP P P MP MG MP M P P M P G P MP AP VP MP AP AG
LTE Femto 33 MG G P M MP MP P M MG P MP MG P M MP M G P M G MP MP M P P MP P M MP M M P M M G MP M M P VP M AG
The available networks per SLA
LTE Femto 34 P MP P P P M MP M MG G MP M MP MP MP M MG P G MP MP MP M M MP M P MG P MP P VP M MP G VP P P AP M P VG
LTE Femto 35 P MP P P MP MP P G G G MG P MP P MG G G M G MP M P MP MG MP P P M P MP MP P MP P G P MP MP VP MP P AG
LTE Femto 36 M MG G MP MG P P M MG MP MP M P M P MP MP MP P G M P MP P P M P M VP P P MP P M M VP P VP P P AP VG
LTE Femto 37 P MP G P M M P MG G MP M M MG P P MP MP P MP MG M M MG MP MP M MG G P MP MP P MP M MG P MP VP P AP VP VG
Table B4 The available WiMAX and WAVE networks of SLA 2, SLA 3 and SLA 4.
SLA 2 SLA 3 SLA 4
CV BS Web BS Web Web
Network
Price
Price
Price
Price
Price
Price
Jitter
Jitter
Jitter
Jitter
Jitter
Jitter
Delay
Delay
Delay
Delay
Delay
Delay
Secutiy
Security
Security
Security
Security
Security
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Packet Loss
Throughput
Throughput
Throughput
Throughput
Throughput
Throughput
Service Reliability
Service Reliability
Service Reliability
Service Reliability
Service Reliability
Service Reliability
WiMAX Macro 1 MP M M P P P P G G G MG G M P M MG MP MP MG MG P M MG MG MP P MP G M MG MP MP M P G AP VP AP AP M VP G
WiMAX Macro 2 G G MG MG G MG P P MP P P M MP P M MG MP MP M P MP P MP P P P MP M M MG MP MP M P M M M AP MP P P AG
WiMAX Macro 3 MG G MG M G MP P P MP P P P P M P MP P P M M M P MP P P P P MG P MP P P MP M G VP P VP AP AP MP G
WiMAX Macro 4 M MG P MP P P MP MP M MG P MG G MP MG G MP M MP G MP MP M MG P MP M M MG MG MP M P M G AP VP VP AP P MP G
WiMAX Macro 5 MP M MP P G P MP MP M P P MG G MP MG G M M M MP P MP M P P MP MP MG P MP M P MP P M AP VP P AP AP VP AG
WiMAX Macro 6 MP M G P G M M G G M MG P MP M M MG MP MP MG P MP MG M M M P MP MG M MG MP MP P P M VP P MP AP P AP VG
WiMAX Macro 7 M MG MG MP MG MP P MG G M M MG MG MP MP M P P M G MP MP M P P MP MP M MP M P P P P MG P MP AP VP MP AP AG
WiMAX Macro 8 M MG M MP G MP MP M MG G MP M P MP MP M MP P P MG P M M M MP MP P M MP M MP P P P MG P MP MP VP P VP AG
WiMAX Macro 9 M MG M MP G M M MP M P P MP P P MP M MP P MP G P MP M P P MP P G MP M P P P MP MG MP M VP P AP AP VG
WiMAX Macro 10 P MP MG P MG M MP MP M M P MP MP P P MP MG P G MP MP MP M M P MP MP M P MP MG P P P MG P MP AP VP P P G
WiMAX Macro 11 MP M P P G M MP P MP G P G MP M M MG MG MP P MG MP M MG M MP MP MP G P MP MP P MP P M VP P VP AP MP P AG
WiMAX Macro 12 MG G M M P MG M P MP G P G MG P P MP G P P MP M P MP MP P M P MG VP P P P P MP M AP VP P AP AP VP G
WiMAX Macro 13 MP M MP P G M P M MG G MP P MG P MP M MP P G MG M P MP M P P MG G P MP MP P M P MG P MP MP VP M VP VG
WiMAX Macro 14 MP M MG P M G M P MP G P M G MP MG G MP M G MP M VP P MG P M P M M MG P MP MP MP MG P MP VP VP MP AP G
WiMAX Macro 15 MG G P M P P M P MP P P MG M MP P MP G P M MP MP P MP P P P MP G P MP M P P MP G VP P VP AP P AP AG
WiMAX Macro 16 MP M P P MG MG MP G G MP MG G G M MG G M M MP M MP M MG MP MP MP MP M MG MG M M P M G P MP AP VP P MP VG
WiMAX Macro 17 MG G MP M P MP M G G MG MG M MP MP M MG MG MP MG MG MP M MG MP MP M MP G P MP P VP MG MG MG P MP AP VP P VP G
WiMAX Femto 1 MP M MG P G MG P M MG M MP MG MG MP G G M MG G MG P M MG MP MP P MG M MP M M P MP MP MG AP VP M AP VP P AG
WiMAX Femto 2 M MG MP MP M P MP MG G M M MP MP MP MG G P M P MG P P MP MP P MP MP M MP M P P P P G MP M AP P VP AP VG
WiMAX Femto 3 P MP MP P G M M MG G M M P MP MP MP M M P M M MP P MP MP P P P G MP M M P M M M VP P P AP AP P AG
WiMAX Femto 4 G G P MG MP P MP P MP MP P P MG P G G G MG P MP P P MP P P P MP M M MG MP MP P MP M MP M P P AP P VG
WiMAX Femto 5 G G MP MG M G M M MG P MP MP P P MP M MP P P M MP M MG P MP P P M MP M MP P P M G P MP AP P AP VP G
WiMAX Femto 6 M MG M MP MP M P M MG MG MP P MG M MP M G P M MP M M MG MP MP P M MG MP M MP P M P G AP VP VP AP MP P VG
WiMAX Femto 7 M MG MG MP MG M M P MP P P MG P M MG G MP M MG M M P MP P P P P G MP M MP P M M G VP P MP AP P VP VG
WiMAX Femto 8 G G M MG P MP MP MG G M M MP G MP P MP M P MP M MP P MP P P MP P M P MP P P MP M MG P MP VP VP MP P AG
WiMAX Femto 9 MG G G M P MP P MP M MG P MG P P P MP M P G G P MP M P P MP P G P MP M VP MG P M AP VP M AP MP VP G
WiMAX Femto 10 MG G MP M MP MP M MP M G P M MP MP P MP P P MG MG M MP M M P M MP G P MP P P MG M M VP P AP AP P P AG
WiMAX Femto 11 M MG M MP M G MP G G P MG MP MP P MP M M P M MP MP MG MG P M MP MP MG MP M M P MP MP G AP VP M AP VP VP AG
WiMAX Femto 12 MP M MG P G G P M MG G MP G MG P MP M M P M MP MP MP M P P MP MG G MP M M P MP P M AP VP MP AP MP P AG
WiMAX Femto 13 G G MG MG G P MP G G P MG G MG P MP M MG P MP MP P M MG P MP P MG G MP M P P MP MP M P MP P VP MP AP VG
WiMAX Femto 14 MP M P P M G P M MG MP MP MP G MP MP M M P P G P M MG MP MP P MG MG P MP M P P P M VP P AP VP AP AP AG
WiMAX Femto 15 G G M MG MP MP M MP M P P MG MP P G G MP MG MG P M MP M P P MG MP G M MG MP MP MP P G M M MP MP MP VP G
WiMAX Femto 16 G G MG MG G MG MP P MP M P MG P P M MG MG MP M MP M P MP M P P P M M MG MG MP M MP MG VP P MP AP AP AP VG
WiMAX Femto 17 MG G P M G MP P MP M P P P P MP G G MP MG MP P M MP M P P VP P G P MP MP P MP P M VP P AP AP P VP G
WiMAX Femto 18 G G G MG P MP P M MG MP MP P MP M G G MG MG M M MP P MP VP P P P M P MP MP P M P MG P MP VP P VP P AG
WiMAX Femto 19 MG G MP M MG M P P MP MG P M MP M MP M MP P G MP M P MP MG P M MP M P MP P P P P M P MP P VP P P VG
WiMAX Femto 20 MG G MG M M MG P MP M MP P G P MP MP M P P P MP MP MP M MP P MG P M P MP P P P MP MG P MP P P VP AP G
WiMAX Femto 21 P MP P P M P M G G M MG P M P M MG MP MP M G MP MG MG P M P MP G M MG MP MP MP M M M M VP MP MP P AG
WiMAX Femto 22 P MP P P G P M P MP G P P G MP M MG P MP P P MP P MP M P P MG MG P MP P P P P MG P MP AP VP VP AP AG
WiMAX Femto 23 MG G MG M G MG P MG G MP M M M P MG G M M MG MG M M MG MP MP M P MG M MG MP MP M MG MG P MP P VP AP AP VG
WiMAX Femto 24 MG G MP M M MP M MP M MG P P P M M MG P MP G M M MP M MG P P P MG MP M P P MG M G MP M P P AP MP AG
WiMAX Femto 25 MG G M M G P MP MG G M M MG MG P MP M M P G MG M MG M P M MG MP M P MP M P MG M MG VP P P AP AP MP AG
WAVE 1 MG G MP M MG G MP P MP MP P MP M P MP M M P MG P M P MP MP P MP M MG P MP M VP MP P G P MP M VP VP P G
WAVE 2 M MG P MP P M M P MP G P G G P M MG MG MP P MP MP P MP MP VP MP MP G MP M M P P MP MG VP P MP AP AP MP AG
WAVE 3 M MG MP MP G MP MP MP M M P MG G P MP M G P MP MG P MP M MP P MP M MG MP M M P P MG M AP VP M AP VP P AG
WAVE 4 G G M MG M MG P MP M G P G P MP M MG M MP G G P MP MP M P P P M P MP M P MP MP M VP P M AP AP AP AG
WAVE 5 MG G MP M MG P P P MP G P MP MG M MP M MG P M P P P MP MP P MP P G MP M M P P P G VP P P AP P AP VG
WAVE 6 M MG G MP M G P M MG M MP MG G MP P MP G P MP P M MP M P P MG M G P MP MG P MP P G P MP P VP VP VP AG
WAVE 7 P MP MG P P MP P MG G MG M M MP MP MP M MG P MG MP P M MG MG MP P MP G P MP P VP MG MP M VP P VP VP MP MP G
WAVE 8 G G P MG P M MP MG G M M P M MP G G G MG P MP MP M M MP MP P M G P MP M VP P MP G P MP P VP AP MP G
WAVE 9 MP M P P MP M M G G P MG MP G P P MP MP P P MG M MG M P M MP M M P MP MP P P P G VP P AP P VP P AG
WAVE 10 G G MP MG MG MP M MP M MP P MG G P P MP G P G MP P MP M MP P MG MP G P MP MP P P MP MG P MP VP VP P AP VG
WAVE 11 MP M P P MP MG M MP M MG P M P M G G M MG G M P MP M M P P P MG MG MG M M MP MP G M M M MP P VP G
WAVE 12 M MG MG MP G P MP M MG MG MP MG G P MG G MP M G G M M M MP MP M MG G M MG MP MP MG MP M M M MP MP VP MP VG
WAVE 13 G G G MG MP G P G G MG MG MG G P G G M MG MG MP P MG MG P M MG MP G P MP M P MP P M VP P M AP MP P AG
169
WAVE 14 G G MP MG M MG MP P MP G P G P P M MG MP MP P G M P MP P P MG P MG MP M MP P P P MG VP P VP AP AP VP AG
WAVE 15 P MP G P M P M MP M G P P MP MP P MP MG P G MP MP MP M M P P MP MG P MP P P MP MP M P MP VP VP AP VP AG
WAVE 16 G G P MG MG MG M MP M M P MG P P M MG MP MP P G M MP M P P MG P G MP M MP P P M M MP M VP P AP VP G
WAVE 17 P MP P P MG M M M MG MG MP M MP M MG G MG M MP MP MP MP M MP P M MP MG MP M M P MP MP G P MP P VP AP AP AG
WAVE 18 MG G MP M G G MP MG G MG M MP MP M P MP M P MP G M P MP MG P P MP M P MP P P MP MP M AP VP P AP AP MP AG
WAVE 19 M MG M MP MG MP M MP M MP P M P P MG G G M P MP P MP M P P MP P M MP M MP P P MP MG AP VP MP AP AP VP G
WAVE 20 M MG P MP MP P P MP M P P G M MP M MG G MP G MP P P MP P P MG MP MG M MG P MP MP MP M P MP AP VP VP P AG
WAVE 21 G G M MG MP MG MP M MG P MP G G P MP M P P M M M P P P P M M M MP M P P M M G MP M VP P VP P VG
WAVE 22 G G M MG MG MG M MG G G M G M M MP M MG P G MG P P MP P P MG MP M MP M M P M M MG VP P AP AP P M VG
Appendix C
Table C1 Number of vehicles that start from each LTE network and their corresponding
velocities. Network Number of vehi- Normal Medium High
cles (%) velocity velocity velocity
LTE Macro 1 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 2 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 3 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 4 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 5 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 6 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 7 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 8 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 9 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 10 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 11 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 12 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 13 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 14 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 15 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 16 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 17 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 18 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 19 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Macro 20 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 1 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 2 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 3 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 4 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 5 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 6 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 7 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 8 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 9 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 10 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 11 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 12 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 13 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 14 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 15 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 16 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 17 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 18 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 19 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 20 329 (0,82648%) 109 (0,27382%) 111 (0,27884%) 109 (0,27382%)
LTE Femto 21 329 (0,82648%) 110 (0,27633%) 110 (0,27633%) 109 (0,27382%)
LTE Femto 22 329 (0,82648%) 110 (0,27633%) 110 (0,27633%) 109 (0,27382%)
LTE Femto 23 329 (0,82648%) 110 (0,27633%) 110 (0,27633%) 109 (0,27382%)
LTE Femto 24 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 25 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 26 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 27 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 28 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 29 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 30 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 31 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 32 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 33 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 34 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 35 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 36 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
LTE Femto 37 329 (0,82648%) 111 (0,27884%) 110 (0,27633%) 108 (0,27130%)
173
Table C2 Number of vehicles that start from each WiMAX network and their corresponding
velocities.
Table C3 Number of vehicles that start from each WAVE network and their corresponding
velocities.
This work may be used in accordance with the terms of the Creative Commons license
or other rights statement, as indicated in the copyright statement or in the metadata
associated with this work. Unless otherwise specified in the copyright statement
or the metadata, all rights are reserved by the copyright holder.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346 USA