Priority Queuing Based Spectrum Sensing

Methodology in Cognitive Radio Network

Thesis By

Name: Mujeeb Abdullah

PNo: 8401079531

Name: Sajid Mahmood

PNo: 8411296133

This thesis is submitted to the school of Engineering as part of

partial fulfillment for the Degree of Master Science in Electrical
Engineering.

Thesis Supervisor: Dr. Jorgen Nordberg

Thesis Examiner: Dr. Jorgen Nordberg
School of Engineering
Blekinge Institute of Technology
 Page |3

Abstract
Radio spectrum is becoming scarce resource due to increase in the usage of wireless
communication devices. However studies have revealed that most of the allotted spectrum is not
used effectively. Given the demand for more bandwidth and the amount of underutilized
spectrum, DSA (Dynamic Spectrum Access) networks employing cognitive radios are a solution
that can revolutionize the telecommunications industry, significantly changing the way we use
spectrum resources, and design wireless systems and services. Cognitive radio has improve the
spectral efficiency of licensed radio frequency bands by accessing unused part of the band
opportunistically without interfering with a license primary user PU. In this thesis we investigate
the effects on the quality of service (QoS) performance of spectrum management techniques for
the connection-based channel usage behavior for Secondary user (SU). This study also consider
other factors such as spectrum sensing time, spectrum handoff and generally distributed service
time and channel contention for different SUs. The preemptive resume priority M/G/1 queuing
theory is used to characterize the above mentioned effects. The proposed structure of the
model can integrate various system parameters such spectrum sensing, spectrum decision,
spectrum sharing and spectrum handoff.
 Page |4

DEDICATION

To Parents.
 Page |5

Acknowledgements

We would like to thank Prof Dr.Jorgen Nordberg, my thesis supervisor, for all his support,
encouragement and advice since we start our thesis, for their guidance and support throughout
the course of this research. Thanks also go to my friends and colleagues and the department
faculty and staff for making my time at a great experience. Finally, thanks to our families for
their encouragement.
 Page |6

Table of Contents
Abstract ......................................................................................................................................3
DEDICATION ............................................................................................................................4
Acknowledgements .....................................................................................................................5
List of Figure ..............................................................................................................................8
Chapter 1 : Background............................................................................................................. 10
1.1 Introduction ..................................................................................................................... 10
1.2 Thesis Outline ................................................................................................................. 11
Chapter 2 :Overview of Cognitive Radio Technology ............................................................... 12
2.1 Spectrum scarcity issue.................................................................................................... 12
2.2 The Cognition Cycle ........................................................................................................ 13
2.3 Network Architecture for Cognitive Radio Network ........................................................ 15
2.4 Standard and Regulation ................................................................................................ 16
2.4.1. IEEE standardization on cognitive radio .................................................................. 16
2.5 Industrial Applications .................................................................................................... 17
2.5.1 Adapt4's XG1 ........................................................................................................... 17
2.5.2 Cognichip ................................................................................................................. 18
2.5.3. Rockwell chip .......................................................................................................... 18
2.6 xMax Commercial Cognitive radio Network ................................................................... 18
Chapter 3 : Queuing Models ...................................................................................................... 20
3.1 Overview of the Queuing Model ...................................................................................... 20
3.2 Basic Queuing Model ...................................................................................................... 20
3.2.1FIFO Queuing: .......................................................................................................... 21
3.2.2 Priority Queuing ....................................................................................................... 22
3.2.3 Fair Queuing ............................................................................................................. 22
3.3 Operation characteristic for Basic Queuing Model ........................................................... 23
3.4 Priority queuing using M/G/1 Model ............................................................................... 25
3.5 Selection of the Queuing Model for Thesis ...................................................................... 27
Chapter 4 : Spectrum Management Frame Work for cognitive radio network ............................ 31
4.1 Radio Spectrum as a Resource ......................................................................................... 31
4.2 Radio Spectrum Management Design Issues .................................................................... 32
4.2.1 Spectrum Sensing ................................................................................................... 34
 Page |7

4.2.2 Spectrum Decision .................................................................................................... 34

4.2.3 Spectrum Sharing.................................................................................................... 35
4.2.4 Spectrum Mobility .................................................................................................... 36
4.3 PRP M/G/1 Queuing Network Model ............................................................................. 37
Chapter 5 : System Model And Simulation ................................................................................ 39
5.1 PRP M/G/1 System Model for calculating average waiting time of Secondary user ......... 39
5.1.1 Multiple Primary User and Single Secondary User.................................................... 39
5.1.2 Single Primary User and Single Secondary User ....................................................... 43
5.1.3 Simulation Results .................................................................................................... 47
5.2 Special Case M/D/1 ........................................................................................................ 49
5.2.1 Scenario A (Arrival rate of Primary and Secondary user is same):............................. 50
5.2.2 Scenario B: (Arrival rate of Primary is fixed and secondary user is varied) ............... 50
5.3 Channel contention and spectrum sharing issues in CRN ................................................ 51
5.3.1 PBSSS Distribution Vector: ...................................................................................... 52
5.3.2 Analytical Model for Spectrum Selection Schemes ................................................... 52
5.3.3 Analysis of Overall Transmission Time: ................................................................... 53
5.3.4 Overall transmission time for SU using Probability based spectrum sensing ............. 54
5.3.5 Overall transmission time for SU Instantaneously Sensing based spectrum selection 56
5.3.6 Simulation Results .................................................................................................. 56
5.4 Spectrum handoff based on Reactive-Sensing and Proactive-Sensing .............................. 61
5.4.1 Greedy target channel selection Algorithm: .............................................................. 62
5.4.2 System Model of PRP M/G/1 Queuing Network: ...................................................... 68
5.4.3 Simulation Results: ................................................................................................... 70
6.1 Summary And Conclusion .............................................................................................. 77
6.2 Future Extension ............................................................................................................. 77
 Page |8

List of Figure

Figure 1.1Measurement form 0-6 GHz spectrum utilization at BWRC [29]. .............................. 10
Figure 2.1: Schematic representation of division of the radio spectrum and radio ranges. .......... 12
Figure 2.2 : Radio spectrum measurements [3]. ......................................................................... 13
Figure 2.3: Cognitive Radio Cycle ............................................................................................ 14
Figure 2.4: Cognitive Radio Network Model ............................................................................. 15
Figure 2.5 : XG1™ Cognitive Radio ......................................................................................... 17
Figure 2.6 : Network Architecture of xMax cognitive radio network. ........................................ 18
Figure 3.1 Basic Queuing Model. .............................................................................................. 20
Figure 3.2 : FIFO Queuing Model. ............................................................................................ 21
Figure 3.3 : Priority Queuing Model. ......................................................................................... 22
Figure 3.4 : Fair Queuing Model. .............................................................................................. 23
Figure 3.5 : M/M/1 Queuing Model. .......................................................................................... 24
Figure 3.6 : Exponential traffic arrival distribution. ................................................................... 24
Figure 3.7 : M/G/1 Priority Queuing Model .............................................................................. 25
Figure 3.8 : Graphical representation of Call loss ratio experienced by users with different
Queuing Distribution systems.................................................................................................... 29
Figure 3.9: Simulation Strategies for Model selection. .............................................................. 30
Figure 4.1 Cognitive radio Network Architecture. ..................................................................... 33
Figure 4.2 : Relationship between spectrum sensing, spectrum decision, and spectrum sharing
and spectrum mobility. .............................................................................................................. 37
Figure 5.1 : Cognitive radio network using TDMA slot for communication. .............................. 40
Figure 5.2 Queuing model for multiple PUs and single SU. ....................................................... 41
Figure 5.3:Single PU and single SU channel access. ................................................................. 43
Figure 5.4: Queuing model for single PU and single SU. ........................................................... 44
Figure 5.5 : Multiple PU and single SU Vs PU arrival rate. ...................................................... 48
Figure 5.6 : Single PU and Single SU Vs PU arrival rate. .......................................................... 49
Figure 5.7 : The average waiting time for packets arriving to each priority queue with equal rate
for primary and secondary users. ............................................................................................... 50
Figure 5.8 : The average waiting time for packets arriving to each priority queue with varying
rate of secondary user and fixed rate for primary users. ............................................................. 51
Figure 5.9 : The analytical model for spectrum selection scheme using PRP-M/G/1 queuing. .... 53
Figure 5.10 : The overall transmission time of the secondary user in the cognitive radio network.
................................................................................................................................................. 54
Figure 5.11: Comparison of the Overall(Average) transmission time (T) for three different
channel selection schemes. ........................................................................................................ 57
Figure 5.12 : Overall (Average) transmission time T is plotted against SU traffic arrival rate for
three different schemes. Traditional based sensing scheme is used as a reference for comparison
and optimal sensing performs better in terms of Overall (average) transmission time. ............... 58
Figure 5.13 : : Probability based approach vs Instantaneous sensing based approach. ............... 59
Figure 5.14 : Stacked bar representation. Probability based spectrum selection scheme based
optimal distribution vector is varied with SU traffic arrival rate. Probability of each bar is unity.
................................................................................................................................................. 59
 Page |9

Figure 5.15 : Optimal distribution vector representation varying with SU traffic arrival rate.
Distribution fitting is shown in red. ........................................................................................... 60
Figure 5.16 : Channel utilization (channel busy period) and Busy period distribution is
represented with SU traffic arrival rate. ..................................................................................... 61
Figure 5.17 : The target channel selection by secondary user by computing six permutations
with basic aim of shortest handoff delay. ................................................................................... 63
Figure 5.18 : State transition graph of three channel interruption caused by PU during SU
spectrum handoff. ..................................................................................................................... 66
Figure 5.19 : Overview of a channel selection of the two-channel system. The occupation
duration of primary and secondary user is denoted by “H/LPC”. ............................................... 67
Figure 5.20 : The PRP M/G/1 queuing network model for two channel system.......................... 69
Figure 5.21 : Transmission Latency for three spectrum handoff strategy with respect to PU
Traffic arrival rate. .................................................................................................................... 71
Figure 5.22 : Reactive Based spectrum handoff vs Proactive-based spectrum handoff. .............. 72
Figure 5.23 : Total service time vs Primary user traffic arrival rate. Optimal in terms of SU
service time for comparison purpose. ........................................................................................ 73
Figure 5.24: Graphical representation of PU busy period and system utilization with PU traffic
arrival rate. ................................................................................................................................ 74
Figure 5.25 : Comparison of two target channel for spectrum handoff. Greedy algorithm is the
best to select .Assume channel sensing time is 0, Ts = 0. ........................................................... 75
Figure 5.26 : Comparison of two target channel for spectrum handoff. Greedy algorithm is the
best to select .Assume channel sensing time is 0, Ts = 0. ........................................................... 76
 P a g e | 10

Chapter 1 : Background
1.1 Introduction
Cognitive radio network (CRN) is a novel idea for efficient spectrum utilization of radio
spectrum resources. The main reason behind this idea was the survey conducted by the U.S
Federal Communication Commission (USA regulator body for spectrum management) in [1]
shows the spectrum under utilization from 3 KHz to 300 GHz. The studies conducted by the
spectrum task force and other spectrum regulatory bodies around the globe reveal low
utilization of the spectrum in some frequency bands [8] resulting in spectrum holes. The
spectrum occupancy measurements conducted by Shared Spectrum Company on the band
between 30 MHz and 3 GHz at six locations in the U.S.A (2004) reveal the average occupancy is
only 5 in urban area. Further, Fixed allocation assignment creates spectrum white holes,
eventually leads to spectrum underutilization with low duty cycle in real time wireless
communication system. Spectrum underutilization is the main challenge for spectrum
management and their control &regulation authorities. New technology is needed to remove and
overcome these problems and provides new methods to enhance spectrum utilization to greater
extent.
This problem is explained in [3] in detailed as “Survey on spectrum utilization in Europe:
Measurements, Analyses and observation”. This experiment was conducted in three different
locations such as in the cities of Czech Republic, and Paris in France in 2009.
Similar study conducted at Wireless research center Berkley (BWRC) showed that the spectrum
is underutilized from 3 GHz to 6 GHz as shown in the Figure 1.1 [29].

Figure 1.1Measurement form 0-6 GHz spectrum utilization at BWRC [29].

Cognitive radio provides a solution to cope with all these problems. To effectively utilize the
spectrum holes, the unused spectrum are shared among primary users PUs (licensed user) and
secondary users (unlicensed users) SUs dynamically. The cognitive radio is developed on
platform of Software Define Technology (SDR) and aware of the surrounding radio environment
so to respond changes occurring in the network.
 P a g e | 11

The main purpose of Cognitive radio technology is to increase Unlicensed Secondary User
(SU) throughput in terms of various traffic parameters. This issue is investigated in this thesis.
The SU perform the spectrum sensing or information gathering process in which it selects the
best vacant channel for transmissions of its data. The selection of a channel is based on various
criteria consist of the shortest waiting time [27], the largest idle period [28] and the maximum
throughput.
For smooth operation (SU) and to avoid interference with primary user (PU), the SU has to
stop its transmission when the PU enter the network and request for vacant channel from the
Base station.

In this thesis study is conducted on various channels assignment strategies for SU to

enhance the spectrum flexible usage. SU Grade of Service (GoS) is the main performance
measure which is thoroughly researched and analyzed in terms of call blocking, call dropping
and call handoffs. Priority based M/G/1 and Markov methodology is used to explore these issues
and challenges.

1.2 Thesis Outline

Chapter 2: Systematic literature overview is conducted regarding cognitive radio network and
its basic operation is discussed.

Chapter 3: Provide overview of Queuing system and its application in cognitive radio network.

Chapter 4: Discussing spectrum management frame work for cognitive radio network and
covering issues regarding spectrum sensing, spectrum decision, and spectrum sharing and
spectrum handoff.

Chapter 5: Provide system models consisting of Priority Queuing using M/G/1 to characterize
the performance of Secondary User in the network and simulation result can be used to grasp
the operation parameter characteristics for secondary user.

Chapter 6: Conclusion and suggestions which can further extend this work for future work.
 P a g e | 12

Chapter 2 :Overview of Cognitive Radio

Technology
2.1 Spectrum scarcity issue
Radio spectrum is a valuable resources shared by different wireless service providers. It can be
repeatedly use by frequency reuse techniques in global system of mobile communication etc.
Radio Spectrum management and utilization are the key research paradigms on which the most
researcher are working on. The figure 2.1 has been used over 100 years for spectrum
management.
The technical and policy aspects of spectrum management are extensively studied. The
technical aspect of spectrum management is concern with technology and physical world
phenomena that affects the spectrum utilization and policy aspect takes in to account of the
economical and political factors. The International Telecommunication Union radio
communication Sector has to manage radio frequency allocation for different emerging
technologies and devise new standards for it. The spectrum is divided into several segments for
specific wireless services. The licensed user has full right to divide the allocated spectrum into
fixed number of frequency channels within a specific geographical area.

Figure 2.1: Schematic representation of division of the radio spectrum and

radio ranges.

FCC is an independent organization of United State of America established by

communications Act of 1934. Its main task is to regulate interstate and international
communications by radio, television, wired, satellite and cable [2]. In 2002 FCC published
 P a g e | 13

spectrum policy report that aims to adopt new technologies and create new technological
resources. Radio spectrum is a limited resource by nature. The report concluded that Cognitive
radio can enable better spectrum usage and improved network efficiency [28]. Thus cognitive
radio can increase secondary market and effective spectrum utilization. FCC then published a
report called Notice of proposed rulemaking (NPRM) using cognitive radio Technology [4].
Similar model is proposed by Ofcom in United Kingdom which defined the “command and
control strategy” for spectrum management.

Figure 2.2 : Radio spectrum measurements [3].

Figure 2.2 shows, that most of the licensed spectrum is unused and free. This reflects that license
owner is not using the spectrum all the time and there are white holes/ white spaces/spectrum
opportunity. The given spectrum is severely underutilized in the middle of radio spectrum. To
solve the spectrum scarcity, to increase spectrum utilization and to fulfill the white spaces
cognitive radio technology is used to achieve this goal. Similar study is conducted at Berkley
wireless research center (BWRC) of Berkley University [29].

The measurement consist of a power spectral density of a radio signal between 0 and 6 GHz and
samples were collected at 20 GHz for the duration of 50 seconds as shown in Figure 2.6. The
radio spectrum is highly underutilized form 3 GHz to 6GHz resulting in wastage of the radio
spectrum. Such problem creates spectrum scarcity for modern wireless communication system

2.2 The Cognition Cycle

The cognitive radio cycle first proposed by J.Mitola[51] and described by different authors in
various paradigm covering multi features of cognitive radio. The CR is combination of
 P a g e | 14

intelligent signal processing and RF flexibility and as a result the cognitive radio cycle was
implanted in radio equipments.
According to Simon Haykin [1] cognitive radio can be defined as:
“ Cognitive radio is an intelligent wireless communication system that is aware of its
surrounding environment (i.e., outside world), and uses the methodology of understanding-by-
building to learn from the environment and adapt its internal states to statistical variations in
the incoming RF stimuli by making corresponding changes in certain operating parameters
(e.g,transmit-power, carrier-frequency, and modulation strategy) in real-time, with two primary
objectives in mind:

• Highly reliable communications whenever and wherever needed;

• Efficient utilization of the radio spectrum.”

The different actions taken by CR in cognition cycle are given in figure 2.3 [1]. To define it, a
state machine embedded in cognitive radio to learn, adjusts and reacts to surrounding radio
environment changes.

Figure 2.3: Cognitive Radio Cycle

The radio will collect information about present operating environment as illustrated by outside
world through direct observation. The importance of collected information is determined at
Orient. According to that the cognitive radio will use either Plan or Decide alternative for
operation. Wireless communication channels are considered to be unstable therefore cognitive
user will adjust its resources through Act alternative. This process is repeated to improve its
operations Learn and in creating new model states, seeking new alternatives.
 P a g e | 15

2.3 Network Architecture for Cognitive Radio Network

According to [16] the architecture of a cognitive radio and its functional framework can be
presented. The basic cognitive radio network model is shown in Figure 2.4. The cognitive radio
can be classified into two types; the primary network (licensed network) is the commercial
network in the operation. There are two kinds of users, licensed (Primary User) and Unlicensed
(Secondary user) in cognitive radio network. Licensed user is the owner of a radio spectrum. A
customer pay monthly/yearly basis to government authorities. Licensed user is also called
Primary user has the right and priority to use the spectrum anytime, anywhere in the world. The
primary base station controls various operating parameters for PU. The operation of PU has not
to be affected by unlicensed users. Unlicensed user or Secondary user has no authority to use any
radio spectrum without permission. Cognitive radio provides an opportunity for Secondary user
to use the spectrum, whenever Primary user is not using the radio spectrum.

Figure 2.4: Cognitive Radio Network Model

Cognitive radio using advance wideband access technology can access both the licensed and
unlicensed bands of communication channels. The Primary network uses licensed spectrum for
operation. The SU has to monitor the presence of PU in the network. In absence of PU the
spectrum band can be used by SU. Hence a complicated frame work of spectrum management is
developed for SU to utilize the vacant band of the PU.

Cognitive radio designed to provide better QoS for secondary user applications. Spectrum
management policies are devised to meet the QoS requirements of Secondary Users. Spectrum
management techniques are given below

 Spectrum Sensing (Monitoring the presence of PU)

 P a g e | 16

 Spectrum decision (Allocation of vacant channel)

 Spectrum Sharing (Avoiding Multiple User Colliding)
 Spectrum Handoff/Mobility (To move SU due to presence of PU to other vacant
channel)

2.4 Standard and Regulation

Cognitive radio standards are under careful investigation. The IEEE standards coordinating
committee (SCCI 41) on dynamic spectrum accessing work is related to cognitive radio network
[5]. The responsibilities of IEEE standard coordinating committee 41 (Dynamic Spectrum
Access Networks) consist of maintaining the standards developed by the committee in
accordance with IEEE-SA standard board operating Manual. Further, to cooperate with other
standard developing organizations. For next generation radio network and spectrum management
IEEE initiated 1900 Standards committee. The IEEE P.1900 committee was formed in 2005 to
formulate standards for new emerging radio networks.

2.4.1. IEEE standardization on cognitive radio

IEEE standardization related to cognitive radio include IEEE 802.22 Working group for the
development of physical (PHY) and Medium Access Control (MAC) for Wireless Regional Area
Network (WRAN) operating in TV broadcast band .WRAN's operate in unused channels in the
VHF/UHF TV bands between 54 and 862 MHz[7].
The motivation of developing IEEE 802.22 standard was to utilize the unused Digital T.V
spectrum to provide broadband services in rural areas. The accessibility to broadband services is
not that critical in suburban areas as compared to rural area where establishment of DSL network
is expensive for few customers. The operating TV bands chosen to provide broadband service
have high-quality propagation characteristics covering larger area. In U.S.A many T.V channels
highly unoccupied [8]. The 802.22 network will provide high quality voice and data services
with suitable QoS. The cognitive radio networks for the first time deployed in U.S.A for
commercial operation. The IEEE 802.22 standard is devising rules and regulation which can
accommodate system specification consisting of frequency range of operation from 54 -862 MHz
and suggestions are circulated to extend to 900 MHz [11].

2.4.2. ITU-R standardization on cognitive radios

ITU-R is conducting standardization activities related to cognitive radio networks. The

standardization activities are carried out in the Study Group 1 (SG 1) which is responsible for
spectrum management [6]. The IEEE is important sector member and contributes to ITU-R
standardization activities. The ITU Radio communication Assembly in 2010 sent Question ITU-
 P a g e | 17

R 241-1/5 on cognitive radio system in the mobile service. ITU-R 241/8 lists the following issues
to be studied in ITU-R WP 5A [6].

2.5 Industrial Applications

There is rapid transition from analog to digital technology in TV broadcasting. Many countries
around globe currently broadcast both analog TV and digital TV (DTV) simultaneously. Most of
analog services will be switched-off at different times in countries. There will be vacant channel
or white spaces in the TV bands. The FCC has proposed in U.S to allow unlicensed working in
the white spaces in the TV spectrum: 76-88 MHz, 174-216 MHz, and 470-608 MHz’s. The main
issue in the implementation of cognitive radios on TV bands is to reduce interference between
licensed and unlicensed user [8].

2.5.1 Adapt4's XG1

Adapt4 developed first commercial CR product available in the market with XG1 label in 2004
[8]. XG1 is an OFDM-based system operating in licensed frequency band in 217-220 MHz as
secondary user. The throughput rate of Adapt's XG1 is 192 Kbps. The CR network deployed
with XG1 supports point-to-point and point-to-multipoint with multi-hopping architectures. The
network identifies vacant channels and uses them for transmission and avoid channel when other
activity is detected.
Adapt4's XG1 consist of following capabilities [Adapt4]:

 Frequency hopping,
 Dynamic Power Management,
 Automatic Configuration
 Dynamic Frequency Selection Figure 2.5 : XG1™ Cognitive
Radio

The XG1 cognitive radio network constantly observes the activity of the license user in the band
and identifies unused channel. In CR OFDM systems a set of sub-carriers of 45 is created, each
of width 6.25 KHz and rapidly hops among them using each channel for 10ms. When another
licensed user is sensed, the network stops using the carries until they become vacant again. The
XG1 equipments detect the licensed users and forward this information to a central station.
 P a g e | 18

2.5.2 Cognichip
France Telecom developed simple low cost radio device called Cognichip. It has to be operated
in the ultra high frequency (UHF) spectrum of 470-870 MHz described in [9]. The Cognichip
can detect spectrum holes in the TV spectrum and can support centralized architecture. The
network consists of base station and User devices. The user device with transceiver chip
measures of radio signal strength indicator (RSSI) for spectrum occupancy. The sensing time for
one channel is 24 ms. The user device cannot receive data during sensing the spectrum.

2.5.3. Rockwell chip

The Rockwell Collins has developed a broadband spectrum sensor for frequency band 30
MHz -2.5 GHz [10]. It is power efficient with power consumption is below 2.5 W and scan the
spectrum at 18 GHz.

2.6 xMax Commercial Cognitive radio Network

The xG Technology developed first industrial commercial xMax cognitive radio cellular
network and presently it is deploying in Florida for U.S Army training center. xMax will
provide mobile communication service with larger coverage area. The network deployed will
consist of base station of BSN-250, a central switching center xMSC and special handset of
TX70 handsets. A reference model of xMax cognitive radio network supporting data and voice
service is given in figure 2.6 [21].

Figure 2.6 : Network Architecture of xMax cognitive radio network.

 P a g e | 19

Features of xMax Network:

 Base station (BSN) is 18 channel mobile VOIP transceiver device .It channelize the spectrum
from 902-928 MHz into 18 discrete channels, and is only in used when there is traffic for a
mobile equipment registered with a particular channel.

 Access Network Gateway (ANG), called the xMSC is the central switching center that
provides service of call processing, IP packet delivery services, mobility and other signaling
related functions.

 The xMax system provide soft handoffs features with capability of (make-before-break) time
slots are acquired before breaking a connection resulting in reliable roaming and a seamless
user experience.

 Ethernet switch aggregate BSN links in use.

 Firewall provides NAT (network address translation) services.

 SIP proxy server provides SIP call control, xG’s SIP message compression technology.

 Network Monitor handle end to end network management and monitoring services.

 Proxy DHCP server is used for IP addresses services.

 P a g e | 20

Chapter 3 : Queuing Models

3.1 Overview of the Queuing Model
Queuing or waiting lines can be found everywhere. Queuing for buying food at supermarket, to
withdraw money from bank or a telephone waiting to be placed through to receiver are day to
day examples. Queuing happen because the demand for a particular service is higher than the
server can cope with it. The Queuing theory developed from roots of studies conducted by A.K
Erlang and A.Markov on the stochastic processes. It is used to model broad variety of systems
including computer networks, telephone systems, and military operations. In early 1900 A.K
Erlang, conducted experiment with variable demand of telephone traffic. Later he published a
report in 1917 elaborating delays in automatic dialing equipment .

The Queuing theory [28] is the study of queues or waiting lines. It has been extensively used in
different system covering fields such as:

 Supermarket has to decide the number of cash outlets to be operative.

 Gasoline station has to decide the number of pumps and attendants on duty.
 Bank must have to decide the number of teller windows for customer service.

3.2 Basic Queuing Model

The basic queuing system can be illustrated as customers arriving for service, waiting for service
if the server is busy and leaving the system after service being completed as shown figure 3.1.

Figure 3.1 Basic Queuing Model.

 P a g e | 21

The basic queuing model depicted in figure 3.1 can be identified some basic elements of the
system as [29]:

 Input process: describe the input process in terms random variables representing either
the number of arriving during a time interval or the time interval between successive
arrivals. Furthermore, the distribution can also be used to determine the arrival of
customer to the system. If the arrivals of customers and the offer of service are according
to plan, a queue can be avoided. If customer arrives to the system and leave it without
service is referred as Balked.
 Service Mechanism: it involves the number of servers, the number of customers being
served at any time, and the duration of service and its modes. In network of queues more
than one servers arranged in series or parallel combinations. Random variables are used
to characterize the service times, and the number of servers. The processing time is
represented by appropriate distribution function.
 Queuing: the number of customer waiting for service is important point of consideration.
The waiting room or queue length can be considered infinite. The realization of such
queue is hard in real network such as telecommunication networks.
 Queue discipline: representing the way in which queue is organized. The rules which are
devised consist of inserting or removing customers from the queue.

3.2.1FIFO Queuing:
Main characteristics of FIFO given in figure 3.2 queuing discipline:

 All arriving packets are placed in a queues

 Transmission of packets in order of arrival
 Packets are discarded when buffer is full
 Delay and loss of packets depends on inter-arrival times and packet lengths
 Flow can interfere with each other
 Random drops due to malicious monopolization : as one flow sends packets at a high rate
and fills the buffer
 Fairness among customers is not achieved with packets from higher priority class are
buffered as long as there is space discarding the packets from lower priority packets.

Figure 3.2 : FIFO Queuing Model.

 P a g e | 22

3.2.2 Priority Queuing: following are the characteristics of the priority queuing is
presented in figure 3.3:

 Packets are served according to priorities set by system.

 Packets are separated and placed in different buffers according to the priority classes.
 High priority is set for packets for low delay requirement applications
 The preemptive selection of packets from highest priority queue can be disruptive for
the operation of queue of low priority class.
 Randomly selection of packets from low priority class of queue can be helpful for
smooth operation of both queues.

Figure 3.3 : Priority Queuing Model.

3.2.3 Fair Queuing: The fairness drawback of FIFO queuing is effectively solved by fair
queuing algorithm.
 To provide equitable access to the transmission bandwidth with each packet flow has its
own flow.
 Bandwidth is equally divided between flows; each flow is transmitted in turn round robin
fashion.
 At some instant time some queues will be periodically be empty with irregular traffic
giving other flow better share of bandwidth.
 In case of a fair packet switching network given in figure 3.4 packet lengths into account.
 For fair share bandwidth the transmission capacity is divided between different numbers
of flows. These different priority flows are serviced by server in a way that reduces
overall waiting time, overall service time to achieve fairness among users.

The introduction of fairness among users results in improvement of system utilization.

 P a g e | 23

Figure 3.4 : Fair Queuing Model.

3.3 Operation characteristic for Basic Queuing Model

David G. Kendall proposed notation for queuing models later named after him called Kendall’s
notation with expression A/B/n K/T.
Where:
A = inter-arrival time distribution
B= service time distribution
n= number of servers
K = System total capacity or number of waiting customers
T= queuing discipline

For A and B the following abbreviations have been developed representing various probability
distribution describing arrival and departures to a system under consideration.

M (Markov): denote exponential distribution with markov property (memoryless)

D (Deterministic): Constant time intervals
Ek :Erlang-k distributed time intervals
Hn :Hyper-exponential of order n distributed time intervals
COX : Cox-distributed time interval
Ph: phase – type distrusted time intervals
GI : General. Arbitrary distribution of time intervals

 = arrival rate of customers in system

 = service rate of customers in system
N = number of customers in system
N q = average number of customers in queue of the system
T = Mean time spent in the system by customers
W = mean waiting time
 P a g e | 24

    = offered traffic

The M/M/1 is the simple queuing system (with FIFO service) can be described as follow: a
single server with infinite queue length with customer inter-arrival rate is exponentially
distributed with parameter  and the customer service times are exponentially distributed with
parameter  is shown in figure 3.5.

Figure 3.5 : M/M/1 Queuing Model.

The M/M/1 model depicted in figure 3.5, assumes that the number of arrival of customers or
packets (data network) for a given interval of time t follows a Poisson distribution with
parameter of product of  t. let for n arrivals within a time interval t, the probability distribution
function P (n) is given as:

( t) n e   t (3.1)
P ( n) 
n
t
The probability density function of the inter-arrival times is given by e for t >=0.
This is called the negative exponential distribution function with parameter shown in figure 3.6.

Exponential distribution of interarrival traffic

3
arrival rate = 3
arrival rate = 2.5
2.5

2
Distribution of arrival

1.5

0.5

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time

Figure 3.6 : Exponential traffic arrival distribution.

 P a g e | 25

The property of stationarity and lack memory can be applied to M/M/1 systems such that a
Poisson input arrives to the server is independent of each other or the state of system. The
probability of arrival of a unit to the system depends only on the length of time s without
considering the initial state or the specific history of arrival preceding it. The queuing system
with Poisson input as Markovian process (denoted by M) is considered.

Examples of Queuing Systems

M/M/m: m stands for servers rather one server (a bank queue with multiple teller)
M/G/1: Similar to M/M/1 queuing model stands for servers rather one server (a bank queue with
multiple teller with general service time distribution).

M/G/1 Queuing Model

The M/G/1 queuing system is a single server queuing system in which the arrival of customer
follows Poisson distribution with rate of  and service time S for each customer follows general
distribution with mean E[S] [31].

Let to observing the queuing system for very long time T for a system . During this long period
of time there have T customers arriving to the queuing system, each customer on average take
E[S] minutes to be served and  (fraction of time server is busy).

T  E[ S ] (The time server is busy) (3.5)

 
T

3.4 Priority queuing using M/G/1 Model

The practical extension to the classical M/G/1 queue is the preemptive resume priority M/G/1
queue. The queues consist of two or more classes of services. Each customer belongs to a single
class of service. A predetermine priority ranking among the classes of service is set shown in
figure 3.7. Each class of service has its own distinct arrival rate (either exponentially distributed
arrival time) and service time distribution.

1 2 3 p

Class 1 Lowest Priority Class

Figure 3.7 : M/G/1 Priority Queuing Model

 P a g e | 26

Mathematical Notation for M/G/1

 Xi =random variable for service time of priority i customers

 Ri= random variable for residual service time of priority of i customers
 i =arrival rate for priority i customers (Poisson arrivals)
 i = i X i =utilization of the channel by customers of priority of i

 i = utilization of the channel by customers of priority 1 to j

i
 i = j 1

 Wi = random variable for the waiting time of a customer of priority i waiting until service
begins
 N i = random variable for the number of customers in queue i (not for customers in the
channel)

Ti = random variable for the time spend by i customer in the systems from arrival until the
completion.

The average waiting time for customers of different class of services and average number of
customers for each class of service in the queue is calculated. This will depend whether priority
discipline in queue is Preemptive or Non-Preemptive.
Non-Preemptive discipline: A non-preemptive denoted by head-of-line-NP queuing discipline
require customer that begin service to complete its service without pause, even if customers of
higher priority arrive in the mean time. The average queuing time of a customer depends on the
customer’s arrival rate of lower priority classes.

Mathematical Notation for Head of Line non preemptive (HOL-NP)

1 R
W0  
2 i 1
i E ( Si2 ) Expected residual service time of a job in service
R = number of priority classes

W0 Mean time in the system for class 1 job

W1  E ( S1 ) 
1  1
W0 Mean time in the system for class j job (j=2......R)
W1  E ( S1 )  j 1 j 1
(1   i ) (1   i )
i 1 i 1
 P a g e | 27

Preemptive Resume Priority discipline: A preemptive resume priority is denoted by head of

line preemptive (HOL-PR) involve a customer arriving at the queue find a customer of lower
priority in service, the arriving customer preempt the customer being served and begin service
immediately. A preempted customer will resume service at the point at which its service was
suspended as soon as there no higher priority customers. This all process is also called the
preemptive resume.

Mathematical Notation for Head of Line Non Preemptive (HOL-PR)

1 Expected residual service time of a class j job (j=1….r)

W0 j   j E ( S 2j )
2
W01 Mean time in the system for class 1 job
W1  E ( S1 ) 
1  1
 j



W0i 
 1 Mean time in the system for class j job (j=2….R)
W1   E ( S j )  i 1 j  j 1
 1   i  1 

 i 1 


i 1
i

Preemptive None Resume Priority discipline: In this scenario the customer with higher
priority being served after arrival to the system (presently serving low priority customer)
preempting the job of lower priority class. After finishing the higher priority class job system
will start afresh without remembering the service that has already been provided.

3.5 Selection of the Queuing Model for Thesis

Working with priority based (Queuing) for cooperative communication (CRN) in cognitive radio
is a challenging task. Several research questions related to the cognitive radio environment as
follow.
1) Which Priority model should be used and how to work with?
2) Which model will follow the real time feasibility and implementation?
3) How to prioritize different users with different priority classes?
4) Which type of (Queuing Distribution system) will be used for input and outgoing link in real
time?
5) How to deal with Queuing management issue with respect to incoming users? Etc…
 P a g e | 28

All such research questions need explanation with an acceptable accuracy. To find out the most
feasible Queuing system model to answer theses research questions. An Engineering and
mathematical literature contains 3 models, which 8 possible combination are made. The
simulation for all possible queuing system classified in terms of call loss ratio for the following
data as listed below. The simulation tool (TOOLS_V1.3) is developed by Markus Fiedler at the
department of Telecommunication systems at BTH, extensively used for queuing system
analysis.
1) Mean Inter-arrival time of users = 5 min
2) Mean service time = 4 min
3) Number of batches = n = 100
4) Number of arrival Per batch = k = 1000
5) Radom numer generator seed value = 0
6) Simulation warm-up period = True = 1
7) Queuing length is variable, dynamic

The simulation run for 8 possible distribution, and collected observations which are being listed
in Table No.3.1. The call loss ratio and observed the statistical confidence level signify results

\
SIMULATION OBSERVATIONS:

Distribution Type Call loss ratio Confidence interval

D/D/1/K7 0 0
D/G/1/K7 0 0
G/D/1/K7 0.000155 ± 3.62559e-005

G/G/1/K7 0.00173 ± 0.00013151

D/M/1/K7 0.005312 ± 0.000276096

M/D/1/K7 0.008698 ± 0.000355648

G/M/1/K7 0.015034 ± 0.000460548

M/G/1/K7 0.017377 ± 0.000570848

M/M/1/K7 0.038568 ± 0.000859249

Table No.3.1: Different Queuing Systems Distribution Types are tabulated with respect to
call loss ratio and confidence interval. Call loss ratio is in increasing order which is
inherently related t QOS and GOS of system users. * M/G/D - Poisson/General
/Deterministic Distribution respectively.
 P a g e | 29

Figure 3.8 : Graphical representation of Call loss ratio experienced by users

with different Queuing Distribution systems.

ANALYSIS: For D/D/1 and D/G/1 queuing system the input distribution is deterministic and
service time distribution is deterministic. These models are called as Deterministic System
showing constant and non-varying behavior with call loss ratio equal to zero. Confidence interval
in this case is negligible.
For M/M/1 as in table, inter-arrival rate and service rate are Poisson (Exponential) processes,
showing pure random and stochastic nature. The errors and call loss ratio is maximum for M/M/1
queuing system. Call loss ratio for different queuing distribution system is plotted in Figure 3.8.

Best Queuing Systems: D/D/1 and D/G/1 Queuing model with deterministic input distribution
is unrealistic consideration in real time Cooperative communication (CRN). So they are Real
time unrealistic.
Worst Queuing System: M/M/1 Queuing model are considered less accurate and more
problematic in terms of GOS (No Real time feasibility).

Feasible Queuing System Requirement: In Priority based Cooperative communication

(PBQCRN) each user (PU and SU) is treated by server with different priorities. Different Priority
classes have different service time distribution, a new theory is needed to characterize General
Service time Distribution (G-Distribution) for different users.
REQUIREMENTS:

a) Input arrival rate of users follows Poisson (Exponential) distribution

b) Priority based output service time will follow General distribution.
 P a g e | 30

This will fulfill the need and requirements for real time implementation of cognitive radio
networks. M/G/1/K is one such model which fulfill over requirements and shows real time
implementation accuracy and feasible results. The M/G/1 model is selected to carry out our
research in this master thesis. Figure 3.9 depicts the motivation strategy for the thesis.

(Markov)

Figure 3.9: Simulation Strategies for Model selection.

 P a g e | 31

Chapter 4 : Spectrum Management Frame

Work for cognitive radio network
4.1 Radio Spectrum as a Resource
The radio spectrum is valuable resource of nature and essential part of society’s
infrastructure. The radio waves have wide range of applications covering sectors such as social,
cultural, scientific and developmental proposes. It consists of emergency services, defense forces
and air traffic control, broadcasting scientific research; so on. The demand for the radio spectrum
allocation for the viewers and listener to radio and television and new developing technologies is
high. The technology for implementing radio wave communication can easily be developed and
deployed for commercial use. The radio wave spectrum is defined in technical term as portion of
electromagnetic spectrum that carries radio waves. The ranges of frequencies which identify the
boundary of radio spectrum are from 9 kHz to 3000 GHz as illustrated in table 4.1.

Table 4.1: Radio spectrum bands and its propagation characteristics

The other important characteristics of the spectrum consist of the amount of information it
carries and it propagation characteristics. The propagation characteristics describe the behavior
of radio waves in the spectrum. In different radio frequency bands depicted in table 4.1, waves
 P a g e | 32

act differently in term of its ability to hop, spread and penetrate.The lower frequency has less
bandwidth capacity to carry information then higher frequency. It can be used in for radio
navigation in the range of 3 kHz to 30kHz.The most of modern communication technologies
such as GSM, CDMA and LTE utilize higher frequency bands due to availability of more
bandwidth then lower frequency bands.

4.2 Radio Spectrum Management Design Issues

The developments in information and communication technologies have opened up new range of
useable radio spectrum, enabling greater access to new allocations and assignments. The
spectrum being an economic resource is both non-storable and non exhaustible and unlike oil it
will never run out. It may become congested. The spectrum regulator authority has to effectively
meet the demand of radio resources requested by public and commercial wireless communication
companies an effective strategy in the form of spectrum management which can be helpful
utilizing this scare resource among different masses of society [26].

The dynamic spectrum access is defined according to standard of I.E.E.E 1900.1 “as method by
which radio system dynamically adapts to select operating spectrum to use available in the form
spectrum holes with limited spectrum use right”. It requires advance technology for
opportunistically accessing the unutilized bandwidth for a service. A DSA radio has property of
agility and flexibility such that it can operate in different spectrum bands at time and support
various many different transmission standards.

The DSA technology also include cognitive radio technology which may be define according to
IEEE 1900.1 standard as “A type of radio in which communication systems are aware of their
environment and internal state and can make decisions about their radio operating behavior
based on that information” The secondary user based on software radio define technology or
cognitive radio technology consist of unique features for detection of the spectrum hole or white
spaces (unutilized portion of bandwidth). It makes use of this spectrum or hole for their
connectivity and transmission of data. The secondary also monitor the communication
environment and independently adapt the parameters accordingly to communication scheme to
maximize the quality of service for secondary user.
The two main characteristics of secondary user which summarize its operation in dynamic
spectrum access network are as follows [28]

 Cognitive Capability: Refers to the ability of radio technology to collect information or

sense in radio environment. The monitoring of the radio environment consist of not only
observing the power in frequency band of interest , but additional techniques of learning
 P a g e | 33

and action decision are required to gather information consisting of modulation schemes
or transmitter power.
 Reconfigurability: With advance hardware and software ability, the secondary user can
be reconfiguring to adapt to changes occurring in dynamic radio environment.

The cognitive capability and reconfigurability enables a secondary user to utilize the unused
spectrum in the network. In figure 4.1 a typical scenario is depicted, the secondary user network
architecture consists of Primary network and the secondary user network [3].The primary user
activity is controlled by primary base station. With centralized secondary user network a CR
base station control the operation of a cognitive user. The spectrum broker manages spectrum
resource among different CR networks.

Figure 4.1 Cognitive radio Network Architecture.

The cognitive radio operating in heterogonous network environment faces many technical
challenges. The challenges can be classified into four main topics as follow:

 Spectrum Sensing
 Spectrum Decision
 Spectrum Sharing
 Spectrum Mobility
 P a g e | 34

4.2.1 Spectrum Sensing

Spectrum sensing is the key feature of cognitive radio. The spectrum sensing gives SU to
observe the available radio environment for spectrum holes. It is closely related to other
spectrum management functions also to layering protocols to provide information on spectrum
availability. Nowadays research focused on the interference avoidance problem. The cognitive
radio cannot perform spectrum sensing and transmission operation due to hardware complexity
and propagation environment [16].

The spectrum sensing consists of PU detection by SU. When the spectrum sensing is
performing on wide frequency range for their transmission is referred as out of band sensing).
When the detection process is executed during transmission of data, the cognitive radio user
avoid interference is referred as (in-band sensing). In most of application devices energy and
feature detection methods are commonly used for PU presence in the network [12]. In energy
detector, the energy in the spectrum over an observation time window is collected and decision
is made regarding presence and absence of PU with respect to pre define threshold [13]. It is
simple and efficient spectrum sensing technique operates without the prior information of
Primary user (PU). Furthermore its performance worsens in fading channels as it cannot
differentiate between noise and primary signal. The cognitive radio network devised
functionalities for spectrum sensing are as follow [32]:

 Primary User Detection: The secondary users observe the surrounding environment for
detecting the presence of primary user transmission and accordingly identify spectrum
availability.
 Cooperative spectrum sensing: Spectrum sensing is performed by as SU and a group of SU
in cooperative fashion. Each secondary user observes individually the presence of PU and
sends its result to central station. The central station fuse individual spectrum sensing result
and make final decision regarding the presence of PU. To cope with multipath fading and
shadowing effects different cooperation methods are proposed [15].

The other technique is feature detection technique is that of its robustness to the uncertainty in
noise power [14].

4.2.2 Spectrum Decision

The spectrum decision functionalities are similar to that of spectrum sharing such that in the
spectrum decision the resource allocation based on application service requirement. The
 P a g e | 35

researches had extensively studied the spectrum sharing and explored QoS issues for secondary
user [31]. The unique feature which distinguishes the spectrum decision from spectrum sharing is
that the spectrum sharing is performed for short period of time in middle of a communication call
within the spectrum band [30]. It is carried out for packet based or a time-slot scheduling on
other hand the spectrum decision is event or connection based operation. Hence comparing the
operation of spectrum sharing, the spectrum decision considers traffic statistics and channel
characteristics for long period of time.

The main functionalities of spectrum decision consist of:

Spectrum characterization: secondary user network or (CRN) characterizes the spectrum band by
taking into consideration the receive signal strength interference and number of users currently
residing in the Spectrum Selections: according to the observed spectrum availability CRN
allocate the best spectrum band to satisfy QoS requirements.

The spectrum sharing operates in intra-spectrum boundary with its operation confined to a
specific spectrum. The spectrum decision is inter-spectrum operation such that available
spectrum is distributed over a wide range and due to some technical requirement (fading or QoS
requirement) the shifting from one band to other band induces delay or transmission latency
leading to degradation in service quality.
The selection of channel for transmission of data consider various factors such as lightest traffic
load , the shortest expected waiting time , the largest idle probability .The two selection schemes
which are extensively studied for spectrum selection take account of the traffic statistics of
primary and secondary user are as follow:

Instantaneous sensing-based spectrum selection method: the operating channel selection is done
by the secondary user according to instant or short-term sensing. Probability-based spectrum
selection method: The operating channel is selected based probability measures of the long term
observation outcomes. This scheme result in shorter overall transmission time for the secondary
user as it had not to scan the huge spectrum. The secondary user had to avoid in selecting the
busy channel with high probability.

4.2.3 Spectrum Sharing

The necessary coordination is needed between secondary for utilizing the communication
channels with minimum interference. The spectrum sharing in this respect play key role. The
spectrum sharing with most of MAC protocol functionalities maintain the QoS requirement for
secondary user and coordinate activates of multiple access of the communication channel and its
allocation.
The secondary user operates in challenging environment as it coexists with primary user and
different fading environments. Thus we can summarize, the spectrum sharing focus on two basic
following functionalities:

 Spectrum Access: It enables secondary users to share the spectrum resources by accessing it
effectively according to certain rules.
 P a g e | 36

 Resource allocation: The secondary user selects the spectrum band which fulfills the QoS
requirement for certain application. The other factor which it take in account while allocating
the resource is transmission power to avoid interference.

The architecture for spectrum sharing in the Secondary user network is based on access
technology defined according to [33] evolved during course of time for resource allocation in
spectrum sharing can be classified as either as

 Centralized spectrum sharing: a central station control the spectrum allocation and access
procedure. Moreover the central station can grant the spectrum access to users for specific
amount of time for small area. This technique is coordinated effort carried out by secondary
users to reduce the collision or interference while accessing free vacant channels.

 Distributed spectrum sharing: each SU perform the spectrum allocation and access based on
(local and global) policies. A different network uses distributed solutions such that different
Spectrum Managers (base stations) compete for allocation of spectrum according to the QoS
requirement of Secondary users.

4.2.4 Spectrum Mobility

The secondary user is outsider or visitor in the network. When the primary user enters the
network request for a vacant channel then the secondary user has to stop its transmission. The
secondary user communication has to be continued in another vacant channel give a spectrum
mobility phenomena. The main functionalities carried out in mobility management consist of
spectrum handoff and connection management. In spectrum handoff, a secondary user shift from
occupied channel to free channel due to arrival of primary user in the network. The secondary
user using its reconfigurablity property adjusts its operating parameters (frequency, modulation
types)[34]. Where as in connection management the secondary user interact with different
protocol layers to maintain QoS for specific application. In addition to Primary User activity in
the network, spectrum handoff may occur due to flowing reasons:

 The secondary user drop its connection due to mobility of user

 The current spectrum band cannot provide the QoS requirement
 P a g e | 37

4.3 PRP M/G/1 Queuing Network Model

The secondary user faces many challenges in its operation in licensed network due to quality of
Service (QoS) requirement for applications [35].

For smooth operation each SU had to address challenges as follow:

 Availability of the spectrum (to avoid interference with PU)

 Best Available channel (QOS requirement for different applications)
 Vacate with entry of licensed user in the network
 Seamless Communication.

The challenges for secondary user operation are transformed into relationship between spectrum
management functionalities for cognitive radio is shown in figure 4.2. Consider a cognitive
radio network where primary and secondary user operates on M1 independent channels. The
secondary user uses spectrum sensing algorithm to select M2 from out M1 channels. The
secondary user on requirement of QoS for a application select M3 channels out from M2 suitable
for transmitting data. Then on the spectrum decision algorithm base, a secondary user can decide
the operating channel from M3 channels.

Figure 4.2 : Relationship between spectrum sensing, spectrum decision, and

spectrum sharing and spectrum mobility.
 P a g e | 38

Based on the spectrum management functionalities the secondary user can sharply adjust its
operating parameters to cope changes occurring in the environment. Thus secondary can increase
its data throughput at same time minimize interference with primary user.

The PRP M/G/1 queuing model is used to characterize the spectrum usage behavior with
multiple handoffs strategies between primary and secondary users. Based on this model, the
overall transmission time with multiple handoffs can be evaluated. The key features for PRP
M/G/1 queuing model are listed below:

 The primary users have the preemptive priority to interrupt the transmission of the
secondary user.
 The interrupted secondary user (due to arrival of primary user) can resume the unfinished
transmission of data.
 A secondary connection may be interrupted due to random access request of primary user
for vacant channel.
 P a g e | 39

Chapter 5 : System Model And Simulation

5.1 PRP M/G/1 System Model for calculating average

waiting time of Secondary user
In the universal mobile telecommunication system, the size of packets varies to carrying the
payloads for different types of service such as data and voice services. Since voice
communication is sensitive to delay so the sizes of packets are small. Whereas for short
messaging service the packet can be long in size due to its tolerance to delay.
It is assumed that the communication channel is divided into equal time slots. And each user in
the network had two types of packets for voice and data communication. In the
telecommunication network voice packets are prioritized most then data packets. These different
types of packets are stored in different types of queue of infinite length. Each queue is processed
in first-in-first-out (FIFO) order. The priorities are also set among PU and SU. With PU is
preferred of using communication channel. In this study it is assume that the size of voice
packets is shorter the size of data packets. To realize this condition for PU conditions are set such
that L1<L0 and L3<L2 where L1 is the sizes of voice packet and L0 is the size of data packets for
PU. While for SU L3 and L2 denotes the size of voice and data packets respectively.

The analysis of arrival of PUs on the average waiting time for SUs is examined with help of two
scenarios consists of:

 Multiple PU and Single SU

 Single PU and Single SU

A preemptive priority queuing model M/G/1 is employed to analyze the average waiting time of
packets for the SU in the light of queuing theory, when it operates with single or different
number of Primary users.

5.1.1 Multiple Primary User and Single Secondary User

The system considered in this scenario consists of multiple PUs and a single SU with a time
slotted [40] cognitive wireless networks where PU is the Licensed user. The transmission
channel is divided into time slots to be used by voice and data packets as shown in figure 5.1.
The PU being licensed user access time slot of resources is at highest priority. The access to the
transmission channel can be described as Poisson process with arrival rate λ.
 P a g e | 40

The PU starts its transmission always at the beginning of a time slot. As matter of fact the
transmission is not always occupied by PU. The SU can effectively utilize it. With lower
priority in the network the SU has to check the availability at the beginning of time slot. The SU
can utilize the free time slot in the absence of PU in transmission channel. When the presence of
PU is detected, the SU has to wait at least one time slot before it may start its transmission. Let
the SU perform perfect sensing for short duration as shown in figure 5.1.

P1 P2 P3

Sensing Period

Busy Slot

Free Slot

Figure 5.1 : Cognitive radio network using TDMA slot for communication.

During this short interval of spectrum sensing the SU collect all necessary information regarding
presence of PU in the particular time slot and then decides whether to transmit or not.

A cognitive radio network under study consist of three PUs denoted as P1, P2, and P3
occupying transmission channel in fixed time slotted manner as shown figure 5.1. The SU can
only utilize the vacant time slot of transmission channel after the sensing period. It cannot start
its transmission in middle of time slot after being free by PU sometime prior to the end of the
slot.
To model this behavior traffic from PUs is defined as a sequence of special packets that uses one
time slot each as shown figure 5.2. The SU can only transmit voice and data packets when the
queue of special packets is empty. Hence the obtained three priority classes are with special
packets from PUs given first priority over voice packets with second priority and data packets
third priority of SU.
 P a g e | 41

Figure 5.2 Queuing model for multiple PUs and single SU.

The SU generate voice and data packets following Poisson process with average arrival rates
being λ2 and λ3 respectively.
The average service time of voice and data packets of the SU can be calculated as:

L2 (5.1)
E [ X sv ] 
C
L (5.2)
E [ X sd ]  3
C
Where C is the bandwidth allocated to the system, Xsv and Xsd denotes the service time for voice
and data packets of SU respectively.

This type of system is known as preemptive priority queuing with service interruption as in [41].
The special packets produced by PUs are placed at highest priority and time to transmit one
special packet is fixed to one slot. The SU voice and data packets are given second and third
priority, respectively. The preemptive resume priority queuing system is considered due to fact
that at most one special packet to be transmitted per slot and there are more than one SU packets
in the queues.
 P a g e | 42

Using formulas from [22] average waiting time of packets for three priority queuing classes is
derived. For special packets of PUs, the average waiting time in queue can be given as

 p E[ X 2 ]
W  E[ X ]+
2 1   p 
(5.3)

 p =λp E[X] denotes the channel utilization by special packets of the primary user.

After the queue of the special packets created by PUs is empty, voice packets of the SU can be
transmitted on the channel. The average waiting time of voice packets of SU using formulas is
drive in [22] is as follow:

E[ X sv ]  E[ X 2 ]   2 E[ X sv2 ]
W2  
1   p  2(1   p )(1   p  sv ) (5.4)

 sv =λsv E[Xsv] and E[Xsv 2 ] denotes the channel utilization and second moment of service
time of voice packets of secondary user respectively.

Now at some interval of time if both higher class of queuing are empty then data packets from
SU can be sent on the transmission channel. The transmission of data packets had to be
suspended if any packets arrive in the higher priority classes. It will resume its transmission once
the queues of higher priority service are empty. The related average waiting time of data packets
of the SU can be calculated as:

E[ X sd ]  p E[ X p 2 ]  sv E[ X sv2 ]  sd E[ X sd2 ] (5.5)

W3  
 p sv  2(1   p  sv )(1   p  sv  sd )
1    

From (5.4) and (5.5) it is cleared that the averaging waiting time of voice and data packets of SU
does not depend on arrival rate and size of packets from the SU but also depends on the arrival
rate of special packets from the PU and the system bandwidth. As performing further
simplifications with setting conditions that the service time distribution of voice packets is
deterministic and that of data packets is exponential. The variance of voice packet transmission
time is zero. The second moment of service time of voice packets and data packets and further
specification is given [42].

sv L22  2 FL2  C 2  p (5.6)

W2 
2 F ( F  sv L 2)
 P a g e | 43

(5.7)
2 H sd L23  2 L3 (sd C  H )  G
W3 
2 H ( H  sd L3 )
After simplifications:

F  C (1   p )
(5.8)

H  F  L 2sv (5.9)

G  C2 p  L22 p (5.10)

The calculation carried out in [42] regarding second moment of voice and data packets and
putting the values in (5.6) and (5.7) it is cleared that the SU averaging waiting time of voice and
data packets is not depended only on its arrival rate and size of packet but also on the arrival of
special packets from the PU and system bandwidth.

5.1.2 Single Primary User and Single Secondary User

In this scenario a cognitive radio network is considered with one primary and one secondary
links. The SU is not restricted to sense for the resources at the beginning of each time slot but it
can sense any time during transmission. The SU identifying the vacant channel through sensing
can easily exploit it for communication. A SU had stopped its transmission after PU request for
radio resources. In figure 5.3 channel allocations are shown with shaded part indicate occupation
by the PU while the reaming white part may be utilized by the SU.

S S P P S S P P S P

Figure 5.3:Single PU and single SU channel access.

The model of an M/G/1 preemptive resume priority queuing systems with four priority classes
for PU and SU is implemented in Matlab is shown in figure 5.4. The traffic generated by packets
at PU and SU follow a Poisson process. As licensed user PU traffic is given priority over SU.
Whereas within traffic generated, voice packets are prioritized over data packets for
transmission. Once the queues of PU are empty then SU can start its transmission by send first
voice packets over communication link and then data packets.
 P a g e | 44

A theoretical model of an M/G/1 preemptive resume priority queuing systems with four priority
classes for PU and SU implemented shown in figure 5.4. The traffic generated by packets at PU
and SU follow a Poisson process. As licensed user PU traffic is given priority over SU.
Whereas within traffic generated, voice packets are prioritized over data packets for
transmission. Once the queues of PU are empty then SU can start its transmission by sending
first voice packets over communication link and then data packets.

Figure 5.4: Queuing model for single PU and single SU.

The average service time of voice and data packets for PU can be calculated as follow:

L0
E [ X pv ]  (5.11)
C

L1 (5.12)
E [ X pd ] 
C

The average waiting time for the voice packets of the PU are given the highest priority can be

expressed as :
 P a g e | 45

 pv E[ X pv
2
] (5.13)
Wpv  E[ X pv ]+
2 1   pv 

 pv   pv E[ X pv ] Denotes channel utilization of voice packets from the PU.

 pv E[ X pv
E[ X pd ] 2
]   pd E[ X pd
2
] (5.14)
Wpd  
1   pv  2(1   pv ) (1   pv   pd )

The average waiting time of packets in the lower priority is affected by serving time and arrival
of high priority packets. As result the average waiting times for lower priority classes is
composed of following three parts:

 For the first time the packet to be transmitted then Average waiting time in queue
 Without interruption from higher priority classes , average transmission time of one
packet
 Packets from high priority classes instigate average interruption time

The average waiting time of packets belonging to the three remaining lower priority classes
using the result reported in [22] for M/G/1 preemptive resume priority queuing model can be
calculated.
The third priority class of the system under study consists of voice packets from SU. The
average waiting time consist for it can be given as [22].

E[ X sv ]  pv E[ X 02 ]   pd E[ X pd
2
]  sv E[ X sv2 ] (5.15)
Wsv  
1   pv   pd  2(1   pv   pd ) (1   pv   pd  sv )

sv  sv E[ X sv ] denote the channel utilization of voice packets from the SU.
 P a g e | 46

Finally the average waiting time for SU data packets being the lowest priority class can be

calculated as

E[ X sd ] (5.16)
Wsd  
1   pv   pd   sv 
 pv E[ X pv
2
]   pd E[ X pd
2
]  sv E[ X sv
2
]  sd E[ X sd
2
]
2(1   pv   pd ) (1   pv   pd   sv   sd )

sd  sd E[ X sd ] Denote the channel utilization by data packets of the SU.

The SU and PU the service time for voice packets is deterministic and for data packets the
distribution is exponential. By further calculation and computing second moment as solved in
[42] and substituting values in (5.15) and (5.16) and further simplifications are as follow:

sv L22  2 L2O  S  E (5.17)

Wsv 
2O(O  sv L2 )
r

S  L20 pv  2CL1 pd (5.18)

O  C  L0 pv  L1 pd (5.19)

2sd L23  L3 ( P  2sd C )  E (5.20)

W3 
P( P  sd L3)

(5.21)
P  C  L0 pv  L1 pd  L2sv

E  L20 pv  2CL1 pd  L22sv (5.22)

 P a g e | 47

For stable transmission the waiting time of packets does not go to infinity, the sizes of data and
voice packets of the SU must fulfill the following conditions.

O
l2  L2 
sv (5.23)

D
l3  L3 
sd (5.24)

Where l2 and l3 are the minimum sizes for voice and data packets respectively.

5.1.3 Simulation Results

The Average waiting time of SU packets depends on the following defining variables.

1. System Bandwidth(Hz)
2. SU packets size(bits)
3. SU traffic arrival rate(packets/slot or packets/second)
4. PU packets size(bits)
5. PU traffic arrival rate (packets/slot/second)

The system is highly dependent on these useful parameters, carefully chosen to meet the
requirements. The parameters for simulations studies to be carried out in Matlab consist of
bandwidth of channel C = 2 Mbps, Time slot duration of 25msecond, with the size of voice and
data packets in PU are L0 = 280bits, and L1=360bits respectively and parameters for SU same.

5.1.3.1 Multiple Primary User and Single Secondary User Scenario

When the network is in operation the different numbers of PU are using the spectrum such that
arrival rate of voice and data packets from λ2=5 packets/slot, λ3= 3packets/slot.
 P a g e | 48

SU Average Waiting Time

4
voice packet analytical
3.5 Voice packets Simulation

SU Average Waiting time Of Packets(ms)

Data packets Analytical
Data packets Simulation
3
Voice packets optimal
Data packets optimal
2.5

1.5

0.5

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
PU Arrival Rate Per Slot

Figure 5.5 : Multiple PU and single SU Vs PU arrival rate.

The SU average waiting time of packets with respect to arrival of PU traffic is plotted in the
figure 5.5. From the graph it is clear that SU waiting is increasing with PU arrival rate. SU
packets waiting time is maximally high after λ = 0.3. The voice packets in case of SU will be
treated as a delay sensitive, so less waiting time will be experienced by voice packets.

5.1.3.2 Single Primary User and Single Secondary User

In this scenario the arrival rate of PU voice packets are considered to be varied from [0 to 5]
packets/ slot 10 times higher than scenario A. PU Voice packets have highest priority and SU
waiting time is analyzed. PU data arrival rate is λ1 = 2packets/slot. SU voice and data arrival rate
are considered to be 5 and 2 packets/slot respectively.
PU voice packets have higher priority, as a result directly serviced by the server without
waiting in the Queue. The negligible waiting time for PU voice packets and the allocated BW is
considered to be large is noted. PU data packets have still higher priority but less than Voice and
waiting time is negligible for low Arrival rate. At higher arrival rate of PU voice packets data
packets experienced little delay in the buffer. Figure 5.6 shows the SU average waiting time in
terms of PU voice traffic rate.
SU voice and data packets experienced higher delay and waiting time in the queue in order to
access the channel and waiting for PU to release the channel. SU data packets have lowest
priority and experienced higher queuing delay.
 P a g e | 49

10
PU-VP-Theor
9 PU-VP-Siml
PU-DP-Theo
8
PU-DP-Siml
SU-VP-Theo

Average Waiting Time (ms)

7
SU-VP-Siml
6 SU-DP-Theo
SU-DP-Siml
5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
PU Voice Packet Arrival Rate

Figure 5.6 : Single PU and Single SU Vs PU arrival rate.

Then two scenarios are compared. SU average waiting in scenario 5.1.3.1 is still less than
scenario 5.1.3.2. The arrival rate of PU voice packets in 5.1.3.2 is 10 times higher than scenario
5.1.3.1. Small size of the voice packets and large system Bandwidth, The service time of Voice
packets is small. According to relation above, this means that large number of voice packets are
serviced in relatively less time and the remaining time slot duration is used by PU data packets,
SU voice and data packets resulting in the reduction of average waiting time of lower priority
classes. Thus the scenario 5.1.3.2 effectively utilized the system BW and offers better spectrum
utilization for SU than scenario 5.1.3.1. The secondary user traffic characteristic consisting of
waiting time analysis while opportunistically accessing the vacant channel can be studied by
using M/D1 priority queuing scheme in cognitive radio network as explained in 5.2 section. The
primary user has higher priority as compared to the secondary for using radio channel resource.
Thus secondary user has to wait until the channel is free.

5.2 Special Case M/D/1

M/D/1 is a special case of M/G/1 queuing model where the service time distribution is fixed for
every class of arrival. PU and SU are considered as class1 and class2 customer for time slotted
cognitive radio network respectively, having fixed service time. Assume perfect sensing for SU
and time slotted structure and sensing is done at the beginning of each time slot by SU. The
Fixed service time is considered in this case. Performance parameters to study: Waiting time and
Queue length.
 P a g e | 50

5.2.1 Scenario A (Arrival rate of Primary and Secondary user is

same):
In this scenario the arrival rate of PU and SU are equal so λp = λs. It is clear from the figure 5.7,
that for low arrival rate the average waiting is comparable, but when the arrival rate is increasing
for PU and SU the gap between average waiting for both users is increased significantly. For
higher arrival rate the PU waiting time increased little but SU waiting time is increased
significantly.

Average waiting time

20 PU-AvgW-Theo
PU-AvgW-Siml
18
SU-AvgW-Theo
16 SU-AvgW-Siml
Average Waiting Time(time slot)

14

12

10

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Packet Arrival Rate(packets per slot)

Figure 5.7 : The average waiting time for packets arriving to each
priority queue with equal rate for primary and secondary users.

5.2.2 Scenario B: (Arrival rate of Primary is fixed and secondary

user is varied)

In this scenario the PU traffic arrival rate is fixed and SU arrival rate is varied. Average waiting
time is analyzed for three fixed arrival rate of PU. SU arrival rate is varied and a function of
average waiting time. When PU arrival rate λ = 0.1, less average delay is experienced by SU and
 P a g e | 51

as PU arrival is increased up to 0.3, SU packets are significantly delayed as shown in figure 5.8.
The performance of SU is highly dependent and affected by PU traffic characteristics.

Average waiting time

30
lam-p1 = 0.1
lam-p1-Sim = 0.1
25 lam-p2 = 0.2
lam-p2-Sim = 0.2
Average Waiting Time(time slot)

lam-p3 = 0.3
20 lam-p-Sim = 0.3

15

10

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
SU Packet Arrival Rate(packets per slot)

Figure 5.8 : The average waiting time for packets arriving to

each priority queue with varying rate of secondary user and
fixed rate for primary users.

5.3 Channel contention and spectrum sharing issues in CRN

The selection of the spectrum holes or vacant channels is important task performed by secondary
user considering various channel characteristics such as the shortest expected waiting time, the
largest idle probability, the largest expected remaining ideal period. Let if the certain number of
secondary users select the same channel meeting its basic requirement for transmission of data
result in channel contention. In this thesis, emphasis is how to resolve the effects of channel
contention and spectrum sharing issue between multiple secondary users.

The two spectrum selection schemes studied in this thesis are as follow in [44]:

 Probability based spectrum selection (PBSSS): The sensing outcome of spectrum

selection methods is highly dependent on PU and SU traffic characteristics. The operating
 P a g e | 52

channel is selected based on predetermined probability which is calculated due to long term
observation. PBSSS results in reduction of shorter overall transmission time because it will
not search the huge spectrum to find best operating channel. Due to the long term
observations it can select the channel interrupted by PU with lowest probability. PBSSS will
prevent the SU from accessing the busy channel with high probability. This will reduce the
overall SU service time. A challenge for PBSSS is to determine the best channel selection
probability to minimize the overall transmission time.
 Instantaneously Sensing based spectrum selection (ISSS): The secondary user select
the vacant channel according instantaneous or short-term outcomes from spectrum sensing.
The secondary user has to observe the vacant or idle channel through its spectrum sensing
capabilities. In a scenario where more than one idle channel is available the secondary user
uniformly selects one channel. And if all channels in the system are busy then the secondary
user has to perform the spectrum sensing on next sensing slot.

5.3.1 PBSSS Distribution Vector:

The SU select the free channel form the network according to a predetermined distribution
vector. Consider channels from set of channels {1…M}. SU will search for idle channel, if more
than one idle channel is free then SU will select uniformly one channel among all the free
channels. If all channels are busy then SU will search for free channel in next time slot. In
probability based spectrum selection method the SU select from group of operating channels
according to a predetermined distribution vector. In this section of thesis we investigate the
overall transmission time minimization

5.3.2 Analytical Model for Spectrum Selection Schemes

The analytical model for spectrum selection scheme using PRP-M/G/1 queuing is shown in
Figure 5.9 with arrival rate of SU and PU follow Poisson distribution denoted by  p( k ) and s
respectively on certain channel k. The d p( k ) (x) and d s (x) are the probability density function
for the transmission of primary user and secondary user on k channel respectively. As shown in
the figure 5.9 the secondary user has the choice to select one best channel among M independent
channel. The spectrum selection algorithm employed at secondary user end can either be
instantaneous based or the probability based method. When SU arrived to access to channel
spectrum selection algorithm will assign a probability to select the best operating channel. An
optimal distribution vector (P= p1, p2, p3…..pn) is used to describe the results of spectrum
selection. The probability distribution vector for the two sensing scheme is analyzed[43].
 P a g e | 53

Figure 5.13: Analytical model for spectrum sharing

Figure 5.9 : The analytical model for spectrum selection scheme using PRP-M/G/1
queuing.

5.3.3 Analysis of Overall Transmission Time:

The overall total service transmission time T is very important QoS performance measure of SU
in CRN, which is equal to:
T = W (waiting time) + S (transmission time)

The average number of interruptions for the SU is denoted by E[N(k)] and the average busy
(k )
period resulted for the PU transmission on k channel by C p . The figure 5.10 shows multiple
interruption of SU caused by PU will leads to multiple-handoffs. These multiple-handoffs
increase the overall transmission time of SU, which degrades SU QoS. To evaluate the overall
transmission time the spectrum selection scheme is integrated with PRP-M/G/1 Queuing network
which characterizes the spectrum usage behavior between Primary and Secondary users.
 P a g e | 54

Figure 5.10 : The overall transmission time of the secondary user in the
cognitive radio network.

The total service time for SU in scenario of multiple interruptions from PU can be defined as:

S ( k )  E[ X s ](average duration of SU call ) 

(5.25)
(k ) k
E[ N ] ( average number of int erruption) C ( SU busy period )
p

Then average service time and busy period for SU will be:

M (5.26)
E[ S ]   pk S (k )

k 1

E[ X p( k ) ] (5.27)
C (k )

1   E[ X
p (k ) (k )
p p ]

5.3.4 Overall transmission time for SU using Probability based

spectrum sensing
k
The average overall transmission time is denoted by E[T p] and W p be the average waiting time
for SUs of channel k for the probability-based spectrum selection scheme. The determination of
distribution vector need traffic statistics for each channel into account. According to the
definition of the overall transmission time for SU:
 P a g e | 55

Tp( k )  Wp( k )  S ( k ) (5.28)

The time spends by SU in the waiting queue by a SU for probability-based spectrum selection
scheme [45]:

(5.29)
E[ R ( k ) ]
W (k )

(1   (pk ) )(1   (pk )   s( k ) )
p

E[ R( k ) ] = average residual service time for PU and SU [45].

1 1
E[ R ( k ) ] =  p E[( X p( k ) )2 ]  pk s E[( X s2 ]
2 2 (5.30)

M M
E[Tp ]  pW
k 1
k
(k )
p   pk S ( k )
k 1
(5.31)

Given for set of M channels the optimal distribution vector is denoted by (L*).

L*  arg min E[Tp ( L)] (5.32)

L

M (5.33)


k M
pk   pk
k 1
0  pk  1  k

The other constraints for this optimization problem are to find the channel utilization for SU and

PU.

 ( k )   p E[ X p( k ) ]  pk s E[ X s ] (5.34)
 1,k  M
 P a g e | 56

E[ X p( k ) ] = average connection length for PU

E[ X s ] = average connection length for SU

5.3.5 Overall transmission time for SU Instantaneously Sensing

based spectrum selection

The average waiting time of the SU using its instantaneously sensing selection method to select
the vacant channel the overall transmission time is:

E[Ts ]  Ws  E[ S ] (5.35)

The waiting time for the SU is defined as duration of finding at least one idle channel through
instant spectrum sensing. The channel busy probabilities for different slots are independent, the
probability distribution function for the waiting time follow geometric distribution.


Ws    i  i 1 (1   ) (5.36)
i 1

Where τ is the spectrum sensing time and ρ is the probability that M channel are simultaneously
busy.

M
    (k ) (5.37)
k 1

5.3.6 Simulation Results

The transmission duration of PU and SU follows exponential distribution and Priority access
channel follows first come first served (FCFS) queuing discipline.
Figure 5.11 shows the effect of SU arrival rate λs on overall average transmission time for three
different spectrum selection schemes. In case of Traditional selection all SU prefer to select the
same operating channel which is the lowest network load operating channel. Thus it leads to
channel contention and congestion problem severely. For system under study, assume t is a
sensing time for SU to be nonzero. Both schemes significantly reduce the average overall
transmission time as compared to traditional case, especially for increasing arrival rate of SU.
 P a g e | 57

When SU arrival rate is less (less than 0.3) PBSSS performs better and leads to shorter average
transmission time. The figure 5.12 represents an optimal sensing in order to minimize the overall
transmission time dynamically.

Comparison of overall Transmission time

3
Traditional approach
sensing based approach
2.5 probability based approach
Overall(average)Transmission time T

1.5

0.5

0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Arrival rate of SU

Figure 5.11: Comparison of the Overall(Average) transmission time

(T) for three different channel selection schemes.

This is due to the fact that PBSSS selects the less interrupted channels by PU with high
probability. When SU arrival rate is increasing (greater than 0.3) IBSSS performs better and
leads to shorter overall average transmission time because sensing based selection significantly
reduces the waiting time through spectrum sensing.
 P a g e | 58

Comparison of overall Transmission time

3
Traditional approach
. sensing based approach
2.5 probability based approach

Overall(average)Transmission time T
optimal sensing(T*=min(Ts,Tp))

1.5

0.5

0
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Arrival rate of SU

Figure 5.12 : Overall (Average) transmission time T is plotted against SU

traffic arrival rate for three different schemes. Traditional based sensing
scheme is used as a reference for comparison and optimal sensing
performs better in terms of Overall (average) transmission time.

When SU arrival rate is low such as 0.1, all SU prefer to select channel 1 as their operating
channel because channel 1 has lightest load. As SU arrival rate is increasing spectrum selection
schemes finds probability vector to distribute SU to different channels in order to reduce channel
contention problem. This will balance the network load over all channels.

For two different SU arrivals rate two different probability vectors is calculated and percentage
of channels access and Probability based SU distribution is analyzed as shown in figure 5.13.

λ1 = 0.5, P*1 = [0.60, 0.25, 0.15, 0]

λ2 = 0.5, P*2 = [0.55, 0.27, 0.15, 0.03]

The figure 5.13 depicts the effect of arrival rate of secondary user on the optimal distribution
vector for the probability based spectrum selection scheme. Consider 4 channels with the loads
as (λ1, λ2, λ3, λ4 = 0.3 0.3 0.5 0.5).
 P a g e | 59

Figure 5.13 : : Probability based approach vs Instantaneous sensing

based approach.

The distributed vector is plotted in each bar and the summation of all probabilities in each bar is
equal to one. This shows that how SU is going to select various channels with increasing SU
arrival rate. The figure 5.14 represents the effect of SU arrival rate on the probability distribution
vector. As SU arrival rate is increasing the effect of channel contention is also increasing for the
best channel. To reduce this problem SUs are distributed among different channels according to
the probabilities assigned by Probability distribution vector. The probability assigned to SU to
access the channel 1 is decreasing as SU arrival rate is increasing and probability distribution of
other channels are increasing. The aim is to distribute SUs among different channels to overcome
channel contention and spectrum sharing.

optimal distribution vector

1
P*1
0.9 P*2
P*3
0.8
P*4
optimal distribution vector

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Arrival rate of secondary user

Figure 5.14 : Stacked bar representation. Probability based spectrum

selection scheme based optimal distribution vector is varied with SU
traffic arrival rate. Probability of each bar is unity.
 P a g e | 60

optimal distribution vector

1
optimal
0.9 distribution fitting

0.8
optimal distribution vector

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
0 5 10 15 20 25 30 35 40 45
Arrival rate of secondary user

Figure 5.15 : Optimal distribution vector representation varying with SU

traffic arrival rate. Distribution fitting is shown in red .

Probability based weight vector dynamically allocates the channels among SUs resulting SU
channel utilization. Increasing rate of SU, the probability of assigning the weight vector is
decreasing. This means that with different probabilities SUs are divided to access the channel in
oder to overcome channel congestion. This Probability based weight vector distribution curve
fitting is illustrated in figure 5.15.

The four channels with the following parameters of PU arrival rate as 0.1, 0.2, 0.5, and 0.7. As
the λs = 0, busy probability of each channel is less than but equal 0.2. When λ s is increasing the
channel utilization is also increasing. So channel 4 is mostly utilized among all four channels and
most of the SU will prefer it to be its operating channel because the waiting time for channel is
minimum as shown in figure 5.16. The probability of each bar is equal to one.
 P a g e | 61

Channel utilization
0.5
channel 1
0.45 channel 2
channel 3
0.4
channel 4

0.35

0.3
utilization

0.25

0.2

0.15

0.1

0.05

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Arrival rate of SU

Figure 5.16 : Channel utilization (channel busy period) and Busy period
distribution is represented with SU traffic arrival rate.

There exist a tradeoff of overall average transmission time between PBSSS and IBSSS.
Probability based selection performs well in scenario of lower traffic loads while the
instantaneously based sensing selection performs well at higher traffic load. The channel
selection scheme should be adaptive to the variation of traffic characteristics.

5.4 Spectrum handoff based on Reactive-Sensing and

Proactive-Sensing

The total service time for target channel is investigated for proactive sensing spectrum handoff.
A greedy target channel algorithm is investigated to reduce the complexity of finding the target
channels. The complexity of the greedy selection scheme is independent of total number of
channels. The objective is to minimize overall total service time by SUs, and focusing on to find
an optimal target channel for proactive-based decision spectrum handoff. Multiple spectrum
handoffs due to multiple interruptions caused by primary user will result in the increase of total
service time completion of SU.
 P a g e | 62

In proactive sensing spectrum handoffs the target channel for future handoffs is determined
before data connection is established, which is obtained by the long term traffic statistics
observations. Whenever SU is interrupted by PU, such predetermined channel is used by SU to
continue unfinished data transmission. No sensing time is required but the key challenge is that
the predetermined channel is available with lower probability.

Generally spectrum handoff delay in time slotted structure is given by

Spectrum Handoff delay = Packet (transmission) pausing time event – service resumption time
event.

5.4.1 Greedy target channel selection Algorithm:

Multiple interruption caused by PU leads to multiple spectrum handoffs for SU. These multiple
handoffs lead to increase the total service time of SU. In this thesis, the total service time
minimization problem related to SU is investigated and analyzed. The greedy channel selection
will search for minimum spectrum handoff delay (reducing the total service time) and also an
optimal channel sequence for spectrum handoff. The statement to find an optimal (in terms of
total service time) channel sequence that minimize the total service time denoted by:

  arg min S ()



Busy period of channel  U

Handoff delay of channel  waiting time at channel k + channel switching time

Handoff delay of channel  Ws (k) + ts  Dh

Default channel  channel α

The aim of SU to select a channel sequence which results shortest spectrum handoff delay at
each handoff. The above optimization problem can be analyzed as [47].

Theorem: let i = {1,2,…M} the number of channels available and Ws(k channel) be the
expected service of SU in channel k. the shortest handoff delay principle is selected to find the
best channel sequence in order to get feasible solution. The feasible solution set contains 6
possible states as shown in the figure 5.17.
 P a g e | 63

Figure 5.17 : The target channel selection by secondary

user by computing six permutations with basic aim of
shortest handoff delay.

Theorem Proof: Let SU transmit on channel α in a network. After some period of time the
arrival of primary user results in interruption in service of SU. In this scenario the expected
handoff delay for staying on the current channel equals to the busy period resulted from the
operation of primary user on α channel. The SU has to jump to other vacant channel or target
channel or stay on the same channel. Hence two possible cases for first interruption is revealed in
the figure 5.17. To stay on the same channel the handoff delay is the PU utilization is given by
Handoff delay = PU busy period at channel α
If SU changes its operating channel from channel α to channel β, then handoff delay is the sum
of channel switching time (ts) and Ws (k) SU waiting time at channel k i.e.

Handoff delay = channel switching time (ts) + Ws (k)

Case 1. In case 1

P0  min
k  
Wsk  ts  (5.38)


P0(α) is the busy probability resulted by arrival of the PU in channel α. Thus SU prefers to stay
on channel α because that will result minimal expected handoff delay and it will start using the
facility of the network after PUs already served.
 P a g e | 64

Case 2. In case 2 if some of the channels like channel β the expected delay may be even less as
result the SU will prefer to change the channel which in this case be β.


    Ws(  )  ts  min min Wsk  ts , P0
k  

 (5.39)

The Case 2 can be further portioned into three sub cases as depicted in the figure 5.17 given to
channel α, β, and γ. {γ ≠ α, β}.
Case 2-1.

In this case the SU prefer to stay on β channel until it complete it transmission.

P0  min Wsk  ts  (5.40)

k  


Case 2-2

SU will switch back to channel α because it will result minimal service time and handoff delay.
If:

 
(5.41)
Ws(  )  ts  min min
k  
Wsk  ts , P0
 ,  

Case 2-3

SU prefers to jump to channel γ to minimize the spectrum handoff delay. SU can select second
target channel sequence upon second interruption. When SU is interrupted for third time case 2-3
(channel γ) can be further divided into three sub cases as in the given figure 5.17.

   ,   Ws( )  ts  min
 min
k  
 ,  
Wsk  ts , P0  (5.42)

Case 2-3-1

Interrupted SU by moving of PU, will prefer to stay on channel γ.

 P a g e | 65

P0( )  min{Ws k   ts } (5.43)

k Ω
{ }

Else it will change to α or β channel

if

Ws( )  ts  min
 min
k  
 , 
Wsk  ts , P0  (5.44)

Or if
Ws(  )  ts  min
 k  
min
 ,  
Wsk  ts , P0  (5.45)

Interrupted SU will prefer to change to channel β that will minimize service time in auxiliary
case

if for some channel η = α, β, γ

then
if Ws(η) +ts < min { min { Ws(k)+ts ), Y1(γ)} (5.46)
then
Ws(η) + ts < Ws(k) + ts, for all k = γ,η (5.47)
Also
Ws (β) +ts < Ws (k) + ts, for all k = α, β (5.48)
This case contradicts, Not Possible.

The state transitions for α,β,γ channels depicted in figure 5.18 for the SU interruptions due to
PU.
 P a g e | 66

Figure 5.18 : State transition graph of three channel interruption caused by PU during SU
spectrum handoff.

Spectrum handoff delay can be illustrated in figure 5.19. When SU is interrupted by PU it will
change its operating channel (Ch1) to another free channel (Ch2). The remaining data
transmission of SU will be considered as a new secondary arrival to Ch2. There are two possible
situations in this case.

Case 1: if Ch2 is idle SU1 will immediately starts transmission and successfully depart.
Case 2: if Ch2 is busy, SU1 needs to wait until all other SUs in the waiting queue are serviced.
 P a g e | 67

Figure 5.19 : Overview of a channel selection of the two-channel system. The

occupation duration of primary and secondary user is denoted by “H/LPC”.

On the other hand Ch1 can be considered as a target channel. The SU1 will stay in the same Ch1
and wait until all PUs finishes their service. The similar procedure is repeated for SUs when
interrupted by PU. Multiple spectrum switching and handoff will increase the delay, service
time, transmission time, waiting time and ultimately QoS is highly affected.
 P a g e | 68

5.4.2 System Model of PRP M/G/1 Queuing Network:

The proposed analytical framework model of PRP M/G/1 queuing network structure
representing the cognitive radio network with two channels for services is shown in Figure 5.20.
This model is helpful to characterize the spectrum usage of the network between PUs and SUs.
With this model the total service time for various target channel selection can be calculated. In
this model, PUs is given higher priority and SUs are given lower priority as a standard tool in
cognitive radio. The other properties for this model consist: the arrival of customers or SUs (with
Poisson distribution) with arrival rates and the service time distribution is denoted by d s( k ) ( x)
with mean E[ X s( k ) ] . Furthermore, the PU arrival is ( 0( k ) ) follow Poisson process with service
(k )
time distribution d0 ( x) with mean of E[ X 0( k ) ] .

The capability of SU consists of to resume the unfinished transmission instead of retransmission

of whole packet. The first-come-first-served (FCFS) scheduling scheme is implemented in this
model to arrange the channel access among SUs.

When SUs are interrupted by PUs as shown in figure 5.19 (1.c), SU is temporarily suspended
from service. SU can stay in the same channel or can change to other operating channel to
resume its unfinished transmission. These two scenarios can be further explained as:

Case 1: In always stay case: if SU stay in the same channel, then this interrupted SU has higher
priority and will be inserted in to the head of lower priority queue in hope to resume unfinished
transmission.

Case 2: In always change case: if the interrupted SU change from current channel to another
target channel it has lower priority and will be queued in to tail of lower priority queue.

One of the key parameters is the effective transmission time which the time duration of frame is
transmitted or resumes the interrupted frame until the time instant the interruption event occurs.
If SU finishes transmission with no interruption, then the effective transmission time is the whole
frame length. In case of interrupted SU, the partial frame length is its effective transmission time.
 P a g e | 69

Figure 5.20 : The PRP M/G/1 queuing network model for two channel system.

The transmission latency or delay due to spectrum handoff based on reactive and proactive-
sensing have been derived in [47] and [48] respectively. The system under study consists of two
channels such that each channel has identical traffic patterns.
For reactive-sensing spectrum handoff the transmission latency can be given as:

[t p s  ( E[ X 0 ]2 0 s  E[ X 0 ](s  t p0 s )]

E[ Lrect ]  E[ X s ]  (5.49)
(1  E[ X 0 ]2 )0 ( s )2

Where tp= processing time for SU, t s= channel switch time, t f = channel sensing time.

Case 1: In always stay case:

 P a g e | 70

The average transmission latency can be expressed as:

E[ X 0 ] (5.50)
E[ Lalstay ]  E[ X s ]  0 E[ X s ]
(1  E[ X 0 ]0 )

Case 2: In always change case

s 2
(0 E[ X s ]) 2   0 E[ X 0 ]
(0  s ) s 1  0 (5.51)
E[ Lchange ]  E[ X s ]  E[ N ](  ts )
(1  0   s )

Hence for the proactive-sensing spectrum handoff scheme the optimal transmission latency can
be given as
L  min{E[ X ], E[ X ]} (5.52)
proact alstay change

5.4.3 Simulation Results:

The Transmission latency of SUs is plotted as a function of PU traffic arrival rate both in always
stays and always change cases. From the figure 5.21 it is clear that transmission latency is low
for always change case for lower PU traffic arrival. At higher arrival rate always stay case
provides lower transmission latency. Thus lower rate of PU, SUs prefer to change from current
channel to another free channel. In case of increasing arrival rate of PUs, SUs will prefer to
switch from always change case to always stay in order to reduce the transmission latency.
Optimal case is switching from always change case to always stay case to enhance SU
transmission efficiency is shown in figure 5.21.
 P a g e | 71

Transmission Latency
4.5
always-stay-case-analytical
4 always-stay-case-Sim
always-change-case-analytical
3.5 always-change-case-Sim
optimal-case
Transmission Latency

2.5

1.5

0.5

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
PU Traffic Arrival Rate

Figure 5.21 : Transmission Latency for three spectrum handoff strategy

with respect to PU Traffic arrival rate.

The comparison between reactive and proactive sensing schemes is depicted in figure 5.22.
When sensing time t f = 0, then Reactive based sensing spectrum handoff has lowest transmission
latency. As sensing time is going to increase the transmission latency for reactive based sensing
is also increasing. This means that SUs consumes most of the time in search idle channel due to
PUs interruption events. This includes the waiting time of interrupted SUs.

As proactive based sensing spectrum handoff has lower Transmission latency for low PU
arrival rate. As sensing time for proactive based scheme is zero, so it performs better in case of
low PU traffic. This means that it can find target channel to be idle channel with higher
probability. Transmission latency is increasing after increasing rate of PU traffic as depicted in
figure 5.22.
 P a g e | 72

Reactive-Based Spectrum Sensing Handoff

2.25
Reactive-based(tf =0)
2.2 Reactive-based(tf = 0.7)
Reactive-based(tf =2)
Proactive-based(optimal case)
2.15
Transmission Latency

2.1

2.05

1.95

1.9

1.85
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
PU Traffic Arrival Rate

Figure 5.22 : Reactive Based spectrum handoff vs Proactive-

based spectrum handoff.

The figure 5.23 shows the comparison of total service time of SU for always stay (secondary
user stay on channel) and always change (secondary user jump from one channel to another
channel) cases. When the arrival rate of PU is low most of SUs prefers to change the operating
channel because it will minimize the waiting time that leads to minimize total service time. From
figure 5.23, it is clear that at λp =0.17, SUs prefers to stay on the current channel in order to
reduce the total service time of SU. An optimal greedy approach can intelligently decide when to
change and select the operating channel. At lower rate it prefers to select always change the
channel but at higher rate of PUs it switches to always stay on the same channel to reduce the
waiting time.
 P a g e | 73

comparison of total service time

4.5
always-stay case
always-change case
4
always-stay case (simulation)
always-change case(simulation)
optimal-case(simulation)
Total service time(seconds)
3.5

2.5

1.5

1
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Primary user arrival rate

Figure 5.23 : Total service time vs Primary user traffic arrival rate.
Optimal in terms of SU service time for comparison purpose.

The given figure 5.24 illustrates the relationship between (PU busy period and utilization) vs PU
traffic arrival rate behavior. The beginning of transmission PU traffic arrival rate is lower. This
means that the busy period resulted by PU is low. This again means that busy period is the period
when PU is utilizing the spectrum, so busy period is the PU spectrum utilization. Most of
spectrum is freely available during beginning of transmission time that is frequently occupied by
SU.

As the arrival rate of PU in increasing with respect to time then the PU spectrum utilization is
increasing. But the increase of spectrum utilization is slow, that signifies that most of the time
spectrum holes are available for SU to efficiently utilize the spectrum.
 P a g e | 74

PU Buzy period vs utilization

3.5
PU Buzy period
utilization
3

PU buzy perid and utilization

2.5

1.5

0.5

0
0.1 0.11 0.12 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.2
Primary user arrival traffic

Figure 5.24: Graphical representation of PU busy period and system utilization

with PU traffic arrival rate.

In order to reduce or to minimize the SUs total service time completion, a greedy algorithm is
proposed. Using Greedy algorithm SU service time is minimized. The figure 5.25 and figure
5.26, shows the comparison between random channel selection and greedy selection approach. In
case of random selection SU does not consider any shortest possible handoff delay that leads to
the increase in total service time of SU. For lower arrival rate of PU, greedy selection has less
spectrum handoff delay and total service time as compared to random channel selection approach
but the difference is small.

As the rate of PU is increasing, busy period is also increasing that leads to reduction of free
spectrum holes, SU must wait to search for free channel. This will increase the spectrum handoff
delay and also total service time of SU. Greedy selection approach performs better in case of
high PU arrival rate.
 P a g e | 75

comparison of total service time

4
SU-SR = 0.5
SU-SR = 0.6
3.5 SU-SR = 0.7
optimal Total service time(S* seconds)
SU-SR = 0.8
SU-SR = 0.9
3

2.5

1.5

1
0.1 0.15 0.2 0.25
Primary user arrival rate

Figure 5.25 : Comparison of two target channel for spectrum

handoff. Greedy algorithm is the best to select .Assume channel
sensing time is 0, Ts = 0.

The figure 5.26 shows the effect of service rate μs of SU on the total service time. As service rate
of SU increasing, this means that more SUs are serviced and Spectrum opportunities are available.
During this time if interruption is resulted by PU, then SU will prefer to stay on the same channel
because the handoff delay and waiting time in another channel may be longer. On the other hand if
PU arrival rate is increasing some of SU prefers to change the operating to obtain less handoff
delay and minimize service time.
 P a g e | 76

comparison of total service time

6.5
Random channel Selection spectrum handoff
6 Total service time greedy(optimal) algorithm
Total service time random selection simulation
5.5 optimal simulation
Total service time(seconds)

4.5

3.5

2.5

2
0.1 0.15 0.2 0.25
. Primary user arrival rate

Figure 5.26 : Comparison of two target channel for spectrum handoff.

Greedy algorithm is the best to select .Assume channel sensing time is
0, Ts = 0.
 P a g e | 77

CHAPTER 6

6.1 Summary And Conclusion

Wireless spectrum is the most demanding entity in today’s modern communication technology.
Wireless spectrum is a highly scare resourse, if properly not managed can cause severe
communication constraints. Study reveals that most of the time the spectrum is idle and free
which is the main problem to spectrum scarcity. In this thesis work various spectrum
management techniques are investigated and analyzed such as spectrum sensing, spectrum
sharing , spectrum handoff. Cognitive radio technology is on the way to resolve the spectrum
mangement issues regarding licensed user (PU) and unlicensed user (SU). PU has full access
right to utilize the spectrum anywhere and SU has secondary priority to access the spectrum if
any spectrum hole is available. Cognitive radio technology provides a meaningfull solution to
overcome the spectrum under-utilization and improving the SUs system capacity. Various
spectrum sensing methods are investigated in this thesis which provides a good understanding of
theoritical work. In this thesis Preemptive resume priority M/G/1 (PRP-M/G/1) model is used to
investigate the issues involved in spectrum management.
Markov chain is studied as a good representation model of cognitive radio. The system capacity
performance parameters such as CBP and CDP are investigated. In thesis the effects of traffic
characteristics of PU & SU on SUs system capacity, utilization, CDP, CBP are thoroughly
analyzed as the performance metrics for cognitive radio. This thesis provides good theoritical
analysis on Radio spectrum management techniques and thorough analytical reasoning for the
PRP-M/G/1 priority model.

6.2 Future Extension

The main goal of CRT is to efficiently use & manage the radio spectrum in order to overcome
the spectrum scarcity. Efficient techniques are needed to reduce the sensing period, sensing
power for SU to improve the system utilization. New real time methods are needed to reduce the
CBP and CDP for unlicensed users. Power control mechanism in CRT is on the hot research.
Introducing Power control in priority queueing in order to reduce the sensing duration and to
avoid interference. New sensing techniques with low computational intensity are needed to
efficiently find a spectrum hole and cleverly release the spectrum upon PU arrival to avoid
interference.
 P a g e | 78

References
[1] S. Haykin. “Cognitive Radio: Brain-Empowered wireless communications” IEEE
Journal on selected areas in communications, vol. 23, no 2 pp201-220, February 2005.

[2] Federal communication Commission.[Online]Available www.fcc.gov/aboutus.html

[3] Alenta, V. Marsalek, R.; Baudoin, G.Villegas, M.; Suarez, M. Robert, F , "Survey on
spectrum utilization in Europe: Measurements, analyses and observations," Fifth
Proceedings of the International Conference on Cognitive Radio Oriented Wireless
Networks & Communications (CROWNCOM), 2010,pp.1-5, 9-11 June 2010.

[4] Federal Communications Commission, “Notice of proposed rule making and order:
Facilitating opportunities for flexible, efficient, and reliable spectrum use employing
cognitive radio technologies,” Feb. 2005.

[5] IEEE Standards Coordinating Committee 41 on Dynamic Spectrum Access Networks

(Dyspan). [Online] Available: http://www.scc41.org/

[6] Question ITU-R 241/8. Cognitive radio systems in the mobile service. [Online]
Available: http://www.itu.int/pub/r-que-sg08.241/en.

[7] M. A. Sturza and F. Ghazvinian, “White spaces engineering study: can cognitive
radio technology operating in the TV white spaces completely protect licensed
Tv broadcasting,” New America Foundation: wireless future program: working
paper #16, Jan. 2007.

[8] Adapt 4. [Online] Available: http://www.adapt4.com/adapt4-products.php.

[9] F.Germani, M. Ghozzi , J-P. Laval, O. Lobry,T.Jarboui and F.Marx, “The

Cognichip: A Flexible, Lightweight Spectrum Monitior ”,Cognitive Systems with
Interactive Sensors 2006,COGGIS 2006,PARIS,FRANCE,MARCH 15-17,2006.

[10] R. Newgard, C. Conway, T. Journot and J. Jiacinto, “Broadband spectral Sensing for
Dynamic Spectrum Allocation”, in proc of the International Symposium on Advanced
Radio Technologies, pp. 137-140, 7-9 March 2006.
 P a g e | 79

[11] Cordeiro, C.Challapali, K. Birru, D. Sai Shankar, N , "IEEE 802.22 the first worldwide
wireless standard based on cognitive radios," First IEEE International Symposium on
New Frontiers in Dynamic Spectrum Access Networks, 2005. DySPAN 2005. pp.328-
337, 8-11 Nov. 2005

[12] Digham, F.F, Alouini, M.-S, Simon, M.K., "On the energy detection of unknown
signals over fading channels," in Proc ICC '03, vol.5, pp. 3575- 3579 vol.5, 11-15
May 2003

[13] A.Sahai.N.Hoven and R.Tandra ,“Some Fundamental Limits on Cognitive Radio”

Allerton Conference on Communication ,control and Computing,2003.

[14] S. Enserink, and D. Cochran. “A Cyclostationary feature detector”. In 28 Conference

on signals, systems and computers, 1994. 199, vol 2, 1995.

[15] Pasi Lassila and Aleksi Penttinen, “Survey on performance analysis of cognitive radio
networks," COMNET Department, Helsinki University of Technology.

[16] H.L.Van Trees, Detection Estimation, and Modulation Theory, John Wiley & Sons,
1968.

[17] A. Sahai, N. Hoven, and R. Tandra, “Some Fundamental limits on Cognitive Radio,”
in proc of Allerton Conference, October 2004.

[18] D.Cabric, S.M Mishra, R.W Brodersen , "Implementation issues in spectrum sensing
for cognitive radios," Conference on Signals, Systems and Computers, pp. 772- 776
Vol.1, 7-10 Nov. 2004.

[19] G, Mohamed, F.Marx, M.Dohler, J.Palicot,; , "Cyclostatilonarilty-Based Test for

Detection of Vacant Frequency Bands," 1st International Conference on Cognitive
Radio Oriented Wireless Networks and Communications, 2006, pp.1-5, 8-10 June
2006.
 P a g e | 80

[20] IEEE 802.22 Wireless Regional area Network Developed standards for Cognitive
wireless RAN medium access control and physical layer (PHY) specifications: policies
and operation in the T.V bands.

[21] xMax cognitive radio, Available: http://www.xgtechnology.com/Technology/network-

architecture.html.

[22] Leonard Kleinrock's, "Queueing Systems Theory" John Wiley and Sons, 1975.

[23] Chapter 1, Queuing Theory, Available: www. faculty.smu.edu/nbhat/chapter1.pdf

[24] Chapter1, Analysis of a M/G/1/K Queue without Vacations Available:

www.iitg.ac.in/skbose/qbook/MG1K_Queue.PDF

[25] MS&E 121, Introduction to Stochastic Modeling Available:

http://www.stanford.edu/class/msande121/materials.html

[26] Radio spectrum module, Available: www.ictregulationtoolkit.org.

[27] Akyildiz, I.F.Won-Yeol Lee; Vuran, M.C. Mohanty, S., "A survey on spectrum
management in cognitive radio networks," Communications Magazine, IEEE , vol.46,
no.4, pp.40-48, April 2008.

[28] Won Yeol Lee,” Spectrum Management in Cognitive Radio Networks” Doctor of
Philosophy in the School of Electrical and Computer Engineering, August, 2009.

[29] Wang, Li-Chun; Wang, Chung-Wei; , "Spectrum management techniques with QoS
provisioning in cognitive radio networks," 5th IEEE International Symposium on
Wireless Pervasive Computing (ISWPC), pp.116-121, 5-7 May 2010.
 P a g e | 81

[30] Won-Yeol Lee, Akyldiz, I.F., "A Spectrum Decision Framework for Cognitive Radio
Networks," , IEEE Transactions on Mobile Computing, vol.10, no.2, pp.161-174, Feb.
2011.

[31] Canberk, B.; Akyildiz, I.F.; Oktug, S."A QoS-aware framework for available spectrum
characterization and decision in Cognitive Radio networks," IEEE 21st International
Symposium on Personal Indoor and Mobile Radio Communications (PIMRC), 2010
pp.1533-1538, 26-30 Sept. 2010

[32] Kaplan, M.; Buzluca, F. "A dynamic spectrum decision scheme for heterogeneous
cognitive radio networks," International Symposium on Computer and Information
Sciences, 2009. ISCIS 2009. 24th, pp.697-702, 14-16 Sept. 2009.

[33] Huseyin Arslan., Cognitive Radio, Software Define Radio 2005., pp. 78- 85, 13-16 June
2005.

[34] Dubey, R.; Sharma, S. "Distributed Shared Spectrum Techniques for Cognitive
Wireless Radio Networks," International Conference on Computational Intelligence
and Communication Networks (CICN), 2010, pp.259-264, 26-28 Nov. 2010

[35] Chung-Wang Wang; Li-Chung Wang; , "Modeling and Analysis for Proactive-Decision
Spectrum Handoff in Cognitive Radio Networks," IEEE International Conference on
Communications, 2009. ICC '09., vol., no., pp.1-6, 14-18 June 2009

[36] Tran, Hung; Duong, Trung Q. Zepernick, Hans-Jurgen , "Average waiting time of
packets with different priorities in cognitive radio networks," 5th IEEE International
Symposium on Wireless Pervasive Computing (ISWPC), 2010, pp.122-127, 5-7 May
2010.

[37] Maseng, T.; Ulversoy, T , "Dynamic frequency broker and Cognitive Radio," IET
Seminar on Cognitive Radio and Software Defined Radios: Technologies and
Techniques, 2008,pp.1-5, 18-18 Sept. 2008.
 P a g e | 82

[38] Moayeri, N.; Hui Guo; , "How Often and How Long Should a Cognitive Radio Sense
the Spectrum?," 2010 IEEE Symposium on New Frontiers in Dynamic Spectrum, ,
pp.1-10, 6-9 April 2010.

[39] Cabric, D.;Mishra, S.M.; Brodersen, R.W."Implementation issues in spectrum sensing

for cognitive radios," Conference Record of the Thirty-Eighth Asilomar Conference on
Signals, Systems and Computers, 2004 ,pp. 772- 776, 7-10 Nov. 2004.

[40] Hsien-Po Shiang, van der Schaar, M. ,”Queuing-based dynamic channel selection for
heterogeneous multimedia applications over cognitive radio networks” IEEE Trans.
Multimed. (USA), vol.10, no.5, pp.896-909, 14-18 August 2008.

[41] Awi Federgruen , Linda V Green, “Queuing sytems with service interruptions”
Institute for Operations Research and the Management Sciences (INFORMS),
Linthicum, Maryland, USA,vol.34,pp 752-768,1986.

[42] Tran, Hung , Duong, Trung Q.,Zepernick, Hans-Jürgen “Average waiting time of
packets with different priorities in cognitive radio networks” IEEE 5th International
Symposium on Wireless Pervasive Computing ,pp.122-127,2010

[43] Zhu, Ping; Li, Jinglong; Han, Kun; Wang, Xufa; , "A New Channel Parameter For
Cognitive Radio," Cognitive Radio Oriented Wireless Networks and Communications,
2007. CrownCom 2007. 2nd International Conference on , vol., no., pp.482-486, 1-3
Aug. 2007.

[44] Chung-Wei Wang; Li-Chun Wang; Adachi, F.; , "Modeling and analysis of multi-user
spectrum selection schemes in cognitive radio networks," Personal, Indoor and Mobile
Radio Communications, 2009 IEEE 20th International Symposium on , vol., no.,
pp.828-832, 13-16 Sept. 2009.

[45] Chee-Hock Ng, Professor Soong Boon-Hee,” Queueing Modelling Fundamentals with
Application in communication in communication Networks,2nd. Johon Wiley and Sons.
Inc.2008.
 P a g e | 83

[46] Li-Chun Wang; Chung-Wei Wang; , "Spectrum Handoff for Cognitive Radio
Networks: Reactive-Sensing or Proactive-Sensins?," Performance, Computing and
Communications Conference, 2008. IPCCC 2008. IEEE International , vol., no.,
pp.343-348, 7-9 Dec. 2008.

[47] Chung-Wang Wang; Li-Chung Wang; , "Modeling and Analysis for Proactive-Decision
Spectrum Handoff in Cognitive Radio Networks," Communications, 2009. ICC '09.
IEEE International Conference on , vol., no., pp.1-6, 14-18 June 2009

[48] Chung-Wei Wang; Li-Chun Wang; Adachi, F.; , "Modeling and Analysis for Reactive-
Decision Spectrum Handoff in Cognitive Radio Networks," GLOBECOM 2010, 2010
IEEE Global Telecommunications Conference , vol., no., pp.1-6, 6-10 Dec. 2010

[49] Tang, P.K.; Chew, Y.H.; Ong, L.C.; Haldar, M.K.; , "Performance of Secondary Radios
in Spectrum Sharing with Prioritized Primary Access," Military Communications
Conference, 2006. MILCOM 2006. IEEE , vol., no., pp.1-7, 23-25 Oct. 2006.

[50] Qiming Tian; Chuan Ma; Guanding Yu; Aiping Huang; , "Analysis of cognitive radio
spectrum access with finite primary users and infinite secondary users," Wireless
Communications and Signal Processing (WCSP), 2010 International Conference on ,
vol., no., pp.1-5, 21-23 Oct. 2010.

[51] Joseph Mitola III, "Cognitive Radio Model-Based Competencefor Software Radios,"
P.Hd DISSERTATION,Royal Institute of Technology, Stockholm, Sweden, 1997.