MASTER’S THESIS UNDERSTANDING THE IMPACT OF SPATIAL...

DEGREE PROGRAM IN WIRELESS COMMUNICATION ENGINEERING

MASTER’S THESIS

UNDERSTANDING THE IMPACT OF SPATIAL

REUSE ON AUTONOMOUS SENSING ORDER

CHANNEL SELECTION

Author Akmal Sultan

Supervisor Adjunct Professor (Docent) Janne Lehtomaki

Second Examiner D.Sc. (Tech.) Zaheer Khan

November, 2016

A. Sultan (2016) Understanding the Impact of Spatial Reuse on Autonomous

Sensing Order Channel Selection. University of Oulu, Department of Communica-

tions Engineering, Degree Program in Wireless Communications Engineering. Mas-

ter’s thesis, 44 p.

ABSTRACT

In wireless communication systems, there is a need to design efficient schemes in

order to overcome the problem of spectrum scarcity. One technology to address

the problem of spectrum scarcity is cognitive radio (CR), in which a network

entity is able to adapt intelligently to the environment through observation, ex-

ploration and learning.

When multiple autonomous cognitive radios are searching for spectrum oppor-

tunities, they face competition from each other in order to access the available

free channel. This will result in reduced throughput which occurs due to collision

between cognitive radios, when they try to transmit in the same channel.

The purpose of this thesis is to study a smart adaptation scheme for efficient

channel access which enable autonomous cognitive radios to improve their overall

bandwidth efficiency in a distributed cognitive radio network with the help of

spatial reuse.

An adaptive persistent strategy with efficient collision detection has been stud-

ied in this work for autonomous channel sensing order selection which enable dis-

tributed CRs to avoid collision and allow them to improve their overall system ef-

ficiency by increasing the average number of successful transmissions, especially,

when number of available channels are less than the number of CRs competing

to access these free channels.

The performance of the studied strategy is compared with random selection of

sensing orders. Simulation results are presented, which indicate that the studied

strategy with spatial reuse achieves the highest number of successful transmis-

sions in a given time slot as compared to other strategies. Simulation results are

also compared for the case with no spatial reuse and the results indicate that

it degrades the system efficiency by reducing the average number of successful

transmissions in a given time slot.

Keywords: Autonomous cognitive radios, channel sensing, distributed cognitive

radio network, opportunistic spectrum access, spatial reuse.

TABLE OF CONTENTS

ABSTRACT

TABLE OF CONTENTS

FOREWORD

LIST OF ABBREVIATIONS AND SYMBOLS

1. INTRODUCTION 7

1.1. Cognitive Radio Technology . . . . . . . . . . . . . . . . . . . . . . 7

1.1.1. Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.2. Benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.1.3. Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2. Autonomous devices in a multichannel CRN . . . . . . . . . . . . . . 9

1.2.1. Distributed Coordination in a multichannel CRN . . . . . . . 9

1.2.2. OSA for autonomous devices in a multichannel CRN . . . . . 9

1.2.3. OSA radio rendezvous problem in multichannel cognitive ra-

dio network . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2.4. Challenges in designing efficient OSA techniques for a multi-

channel CRN . . . . . . . . . . . . . . . . . . . . . . . . . . 11

1.2.5. Solution: an adaptive persistent strategy with efficient colli-

sion detection mechanism . . . . . . . . . . . . . . . . . . . 11

2. RELATED STUDIES 14

3. SYSTEM MODEL 16

3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.2. Studied methods under different PU activity models . . . . . . . . . . 16

3.3. Channel Sensing & Data transmission . . . . . . . . . . . . . . . . . 16

3.3.1. Sequential Channel sensing . . . . . . . . . . . . . . . . . . 18

3.3.2. Sequential Channel Sensing Scenarios . . . . . . . . . . . . . 18

4. CHANNEL SENSING ORDER SELECTION 21

4.1. Studied Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4.2. γ-Persistent Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.2.1. Initialization . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.2.2. Weighted Coin Toss . . . . . . . . . . . . . . . . . . . . . . 22

4.2.3. Possible Outcomes . . . . . . . . . . . . . . . . . . . . . . . 22

4.2.4. Restart Again . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.3. Collision Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.3.1. Spatial Reuse . . . . . . . . . . . . . . . . . . . . . . . . . . 24

4.3.2. Proposed Efficient Collision Detection with spatial reuse . . . 24

5. PERFORMANCE ANALYSIS OF THE PROPOSED STRATEGY 27

5.1. Randomize after every collision strategy . . . . . . . . . . . . . . . . 27

5.2. Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

5.2.1. Scenario A . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

5.2.2. Scenario B . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

5.2.3. Scenario C . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

5.2.4. Scenario D . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

5.2.5. Scenario E . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

6. SUMMARY 40

7. REFERENCES 41

FOREWORD

This thesis has been written in completion of Masters in Wireless Communication En-

gineering at University of Oulu, Finland. This thesis is funded by Center for Wireless

Communication as a part of its Master’s Thesis grant program. The main theme of

this research work is to study the impact of spatial reuse on autonomous sensing order

channel selection.

In the beginning, it was very challenging for me to start my thesis but D.Sc. (Tech.)

Zaheer Khan always keep on encouraging me that I can finish it. I would like to

thank my supervisor Adjunct Professor (Docent) Janne Lehtomaki for his continuous

support, incomparable supervision and mentoring that I was able to finish my thesis.

I would also like to thank my co-supervisor D.Sc. (Tech.) Zaheer Khan for being an

amazing mentor and friend throughout my thesis.

I have also been blessed with incredible friends especially Abdul Moiz, Ikram

Ashraf and Saad Bin Liaqat who really motivated me whenever I was feeling down

or having any problem related to my thesis.

Finally, I would like to dedicate my thesis to my parents Safdar Sultan and Zaid

Akhter for being the light of my life and always supporting me in whatever I have

done in my life.

Oulu, Finland November 24, 2016

Akmal Sultan

LIST OF ABBREVIATIONS AND SYMBOLS

CR Cognitive Radio

CDR Constant Detection Rate

CRN Cognitive Radio Network

DSA Dynamic Spectrum Access

FCC Federal Communications Commission

MAC Medium Access Control

NTIA National Telecommunications and Information Administration

OSA Opportunistic Spectrum Access

OSP Optimal Stopping Problem

PUs Primary Users

RSM Radio Resource Management

RSOP Random Sensing Order Policy

SUs Secondary Users

M total number of cognitive transmitter-receiver pairs

N set of available channels

Pd,i detection probability

Pd desired target value

θi probability of PU being present in each time slot

αi PU transition probability from the state of being occupied to free

βi PU transition probability from the state of being free to occupied

iTs entire duration of the sensing stage

T − Ts time required for the data transmission stage

Ts duration of single sensing step

T entire duration of the given time slot

IP sensing order

L set of available sensing orders

SC success counter

b binary flag

γ persistence factor

Pfa probability of False alarm

p∣

∣L∣

∣-element probability vector

pi probability of detecting the ith sensing order

1. INTRODUCTION

In recent years, wireless services have been growing at a very rapid rate, yielding

a huge demand on the radio spectrum. Resource scarcity has been a very common

drawback for wireless communication systems [1, 2, 3]. Thus, there is an urgent need

for finding intelligent ways of managing the scarce spectrum resources in order to

accommodate the explosive growth of wireless services.

Federal Communications Commission (FCC) is the regulatory authority responsible

for managing the usage of radio spectrum resources. It assigns radio spectrum on long

term basis to permanent users or licensed holders which are also known as primary

users (PUs) for usually a large geographical area. Generally, the available radio spec-

trum resources are not intelligently used by the PUs because the assigned spectrum

resources are not always used by the PUs which eventually results in under utilization

of large portions of available spectrum resources. This inefficient usage of bandwidth

limited resources leads us in finding dynamic spectrum access techniques which can be

employed for users having no allocated licensed spectrum by allowing them to utilize

temporarily the unused licensed spectrum, when PUs are not active. These users are

known as secondary users (SUs).

In recent years, FCC has been searching for more reliable, comprehensive and flexi-

ble ways for the usage of available scarce spectrum resources [4] by employing cogni-

tive radio technologies [5].

The overall network performance can be improved by the deployment of smart ra-

dios (cognitive radios) which can efficiently coordinate and adopt to their environment.

Cognitive radio (CR) is considered to be a very important technology in enabling com-

munication networks to opportunistically utilize the unused spectrum which is also

known as Opportunistic Spectrum Access (OSA) [5].

In this thesis, smart adaptation and coordination methods have been studied for wire-

less communication which allows wireless devices to perform in a very effective fash-

ion to enhance the overall bandwidth efficiency of a wireless network. The introduction

of dynamic spectrum access (DSA) and radio resource management (RSM) in wireless

networks enable wireless devices to employ their available radio spectrum efficiently.

The main theme of this research work is to study a smart adaptation scheme for

efficient channel access which helps to find out a way to improve the overall bandwidth

efficiency for distributed CRs in a multichannel Cognitive radio network (CRN).

An adaptive persistent strategy with efficient collision detection and spatial reuse

has been studied in this work for autonomous channel sensing order selection. This

will enable distributed CRs to avoid collision and allow them to improve their overall

system efficiency by increasing the average number of successful transmissions, espe-

cially, when number of available channels are less than the number of CRs competing

to access these channels.

1.1. Cognitive Radio Technology

In wireless communication, spectrum scarcity is a very common issue. There has been

a lot of research going on through out the world to overcome the problem of spectrum

scarcity.Cognitive radio technology has been considered a very important technology

8

in overcoming spectrum scarcity by employing smart radios to utilize the spectrum

efficiently.

1.1.1. Definitions

The term ”cognitive radio” was used for the first time by Joseph Mitola in [6] and

ever since it has gained a great interest from a very wider section of wireless commu-

nity. Some of the most popular definitions of cognitive radio from the literature are as

follows.

• Mitola [6] defines cognitive radio as: ”A radio that employs model based rea-

soning to achieve a specified level of competence in radio-related domains.”

• Simon Haykin [7] defines cognitive radio as: ”An intelligent wireless communi-

cation system that is aware of its surrounding environment (i.e., outside world),

and uses the methodology of understanding-by-building to learn from the envi-

ronment 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; and efficient utilization of the radio spectrum.”

• The National Telecommunications and Information Administration (NTIA) [8]

has defined cognitive radio as:”A radio or system that senses its operational elec-

tromagnetic environment and can dynamically and autonomously adjust its radio

operating parameters to modify system operation, such as maximize throughput,

mitigate interference, facilitate interoperability, and access secondary markets.”

In this thesis, a cognitive network is defined as a group of autonomous wireless

devices which have the tendency to act intelligently in making decisions related to their

operating parameters and enhance coordination with each other efficiently in order to

achieve desired goals.

1.1.2. Benefits

Some of the common benefit of Cognitive Radios discussed in [9] are as follows:

• CR helps to achieve greater spectrum efficiency through improved access.

• Interoperability and coexistence

• CR simplifies and reduces the tasks needed to setup and use a radio.

• Enhanced interface for applications related to communication tasks.

• A CR can improve its performance based on the information known to CR re-

lated to its internal affairs.

9

1.1.3. Challenges

There are certain hurdles and complexities in designing and implementing a CR net-

work. Some of them are discussed in [10, 11, 12, 13] which are as follows:

• Sometimes CRs have to deal with or articulate complex queries which are des-

tined from one radio to another and in some cases it has to execute certain com-

mands sent by another radio.

• CRs are supposed to use frequency spectrum efficiently and this is only possible

when CRs have the ability to switch among different frequency channels with

minimum delay.

• Adaptive decision making ability of CRs require them to make use of past out-

comes in order to efficiently utilize the frequency spectrum.

1.2. Autonomous devices in a multichannel CRN

Licensed frequency spectrum is generally a scarce resource in wireless networks. CR

networks utilize this scarce resource in a very efficient manner by exploiting it oppor-

tunistically in order to access unused spectrum bands for the time period until there

is no PU activity. There are many OSA techniques in the literature but the most com-

monly used technique is sensing based OSA because it does not requires PUs to change

their structure or way of working [14].

Autonomous devices in a multichannel cognitive radio network are capable of mak-

ing decisions independently based on information related to spectrum sensing and

feedback in a wireless network. Sometimes, these autonomous devices have to deals

with lack of computational resources and information related to their surrounding en-

vironment which can eventually degrade the overall wireless system performance.

1.2.1. Distributed Coordination in a multichannel CRN

The main idea behind distributed coordination in a multichannel CRN for autonomous

devices is to build a very adaptable wireless network which enables these independent

devices to act independently and to ensure efficient utilization of resources by using

low complexity methods.

1.2.2. OSA for autonomous devices in a multichannel CRN

Periodic spectrum sensing is performed for autonomous devices in sensing based OSA

so that they can free the channel when a PU becomes active in a channel in order to

avoid interruption in PU activity [15]. Time slotted multiple access is widely utilized

for OSA in multichannel CRN when multiple frequency channels are available [16,

17, 18, 19, 20]. Autonomous devices uses first portion of each time slot for spectrum

sensing and the second portion is used for accessing the channel, if found free.

13

are certain performance degrading factors in random sensing order selection which are

discussed in chapter 5.

The likelihood of collision reduces when autonomous CRs choose those sensing

orders which allows them to avoid one another, but still there are chances that two or

more autonomous CRs will choose the same sensing order.If more then one CR finds

the channel free and they transmit in the same channel at the same then a collision will

occur. There should be an efficient collision detection mechanism which can reduce

the likelihood of collision.

Spatial reuse has been utilized in this thesis by taking advantage of the transmitting

and receiving ranges of the autonomous CR’s transmitter-receiver node pairs for effi-

cient collision detection in order to increase the overall throughout of the system which

is explained in detail in Chapter 4.

14

2. RELATED STUDIES

Recently, the problem of designing efficient channel sensing selection strategies have

gained a lot of attention. In sensing based OSA, periodic spectrum sensing is achieved

when a CR vacates the channel free as a result of PU being active in a channel [15].

A time slotted periodic sensing model is divided in two categories: Periodic sensing

for a single potentially available primary user band and Periodic sensing for a multiple

potentially available primary user bands. In case of single primary user band, single

licensed spectrum band is explored by the CRs [23]. CRs use the first portion of each

time slot for the purpose of sensing the licensed band and if it is free then the second

portion of each time slot is used for accessing the band [24].

Multiple licensed spectrum bands are explored by CRs in case of periodic sensing for

multiple potentially available primary user bands [21] and from here it can be further

divided in to single channel sensing and sequential channel sensing [23]. In single

channel sensing, the CR initially selects a channel for sensing and transmitting only

when it is free in a given time slot, apart from that it remains quite for the whole

duration of the slot. Whereas in sequential channel sensing, a CR can sense more than

one channel for the entire duration of the time slot.

Sequential channel sensing for a single user CR system has been discussed in [25,

26, 27] which utilizes the optimal stopping problem (OSP) models. In multi-user CR

systems, some initial results were reported in literature for OSP models, such as in [28]

a heuristic solution was proposed and in [17] a centralized solution for two-user system

was proposed. In this work, sequential channel sensing has been used in which two

or more autonomous CRs sense the channels sequentially for spectrum opportunities

in some sensing order. There have been several optimal policies in the literature for

the selection of channel sensing order. Optimal policies for the selection of channel

sensing order is explained in [21, 18, 29, 30]. Problem of multi-user sequential channel

sensing and access in dynamic CR networks have been studied in [30], which involves

active user set to be changed randomly from slot to slot and also there is no information

exchange among users. The goal of the users in [30] is to determine the channel order

for sensing and access, it also addresses the overlapping of channel sensing order by

introducing a generalized interference metric. A coordinator for a two-user CRN has

been used by Fan [17] to determine the optimal sensing order selection which is based

on channel availability statistics.

Impact of cooperative sensing has been considered in [31] along with the sensing of

channels in multi-channel DSA networks for decision making, whereas, we consider

channel occupancy for decision making in channel sensing. Several cooperative sens-

ing strategies, i.e., parallel, sequential, and parallel-sequential has been proposed and

compared in [31] to schedule all users to sense multiple channels in order to attain op-

timal throughput. In [32], the author has studied cooperative sensing in heterogeneous

CR network, where the number of receive antennas may be different for each SU and

have different signal processing capabilities.

A simple channel sensing order has been proposed in [20] for SUs in a distributed

CRN, where there is no prior knowledge of PU’s activity and it assumes that CRs

have the knowledge of channel gains. Whereas, we do not consider the knowledge of

channel gains in our work and we have proposed an adaptive sensing order selection

strategies for channel sensing.

15

The effect of imperfect information on the performance of the autonomous SUs has

been discussed in [14] which will access the spectrum resources opportunistically. The

results in [33] shows that the rate of convergence has been reduced due to the impact

of the imperfect information and the amount of competition between SUs and the level

of PU activity which determines this performance loss. The author in [34] analyzes the

impact of imperfect channel sensing on casual channel estimation methods in corre-

lated CR channels. We have also considered the impact of imperfect information which

includes the effect of false alarms and channel errors on making adaptation decisions.

A two stage process for allocating fungible channel sets to multiple CRs for op-

portunistic spectrum access is the topic of [35], which eliminates the possibility of

collision among CRs and allows CRs to focus more on avoiding collisions with the

primary user (PU). A learning based dynamic channel selection algorithm on the set

of channels is used in [35] whose performance is accurately estimated by the proposed

neural network, and it’s done by using the duty cycle and the complexity of the PU’s

behavior on channels.

The modelling and performance analysis of a random sensing order policy (RSOP)

has been studied in [36] for a distributed CRN. A novel markov process has been used

for modeling the behaviors of SUs and an optimization problem is defined to keep the

interference level bounded and to maximize the average throughput in [36].In addition,

two efficient algorithms are proposed which improves the performance of RSOP by

adjusting the sensing-access parameters adaptively.

Collision free schedules have been achieved in [37, 38, 39] using learning based

medium access control (MAC) methods. The transmitter in the above methods must

do sensing before transmission in order to determine available channels because these

schemes do not consider channel availability due to the presence of PUs.

Spatial reuse in wireless mesh networks has been studied in [40] to increase the

network throughput. It also optimizes the efficiency of spectrum utilization by taking

use of the interplay between spatial reuse and network coding. A joint spatial-temporal

spectrum-sensing scheme has been proposed in [41] that improves the performance of

temporal sensing by exploiting the information from spatial sensing.

16

3. SYSTEM MODEL

3.1. Introduction

In this thesis, a distributed cognitive radio network has been studied which consists

of M cognitive transmitter-receiver pairs and a set N=1,2,3,....,N of channels. The

separation distance between transmitter and receiver node in an autonomous CR is

considered to be constant and it is same for all CR pairs in a distributed CRN. CRs

can make use of the available free channels, when these channels are not occupied by

primary users. A time slotted system for both PU and CR activity is adopted and it

assumes that the primary user can only be present for the entire time slot or it remains

absent for the entire time slot.

As we know, there are always hardware constraints in designing a distributed CRN,

so it is considered that a CR can either transmit or it can only sense during any time

slot but it is not allowed to perform sensing and transmitting at the same time. Each

CR is assumed to sense only one channel at a given time.

The detection probability (Pd,i) for an autonomous CR is set to a desired target value,

(Pd,i = Pd), for all i ∈ M . According to [15], (Pd) needs to be close to 1. When the

detection probability is fixed at a constant target value, it is defined as constant detec-

tion rate (CDR) requirement [42] and the false alarm probability for an autonomous CR

varies. The effect of varying probabilities has been considered in our work, whereas

we have neglected the impact of missed detection errors for the sake of simplicity.

3.2. Studied methods under different PU activity models

We have analyzed the PU activity model as follows:

1. In a given time slot, the probability of PU being present in each time slot is θi,i ∈ N . The PU activity in this model for a given time slot is considered to be

independent of PU activity in any other time slot and it also has no dependency

of PU activity in any other channel as well. Similar channel occupancy model

for the PU activity has been adopted by [20, 16, 43].

2. In consecutive time slots, the PU utilizes correlation in channel occupancy for

this model and a two-state markov chain describes the state of each channel. A

two-state markov chain contains αi which defines the PU transition probability

for the ith channel from the state of being occupied to free, and βi which defines

the PU transition probability for the ith channel from the state of being free to

occupied. H. Su [28] has also adopted the similar PU activity model.

3.3. Channel Sensing & Data transmission

In a distributed CRN, channel sensing and data transmission using opportunistic trans-

mission is explained as follows by the help of Figure 4:

18

3.3.1. Sequential Channel sensing

When multiple autonomous CRs search for available free channels for spectrum oppor-

tunities, the following events might appear in a single sensing step from the viewpoint

of a single CR:

1. The CR arrives at a given channel and finds it free and is the only one to find it

free. The CR will transmit in that channel and it will keep control of this channel

until the remainder of the time slot.

2. The CR arrives at a given channel and finds it busy which means it is either

occupied by PU activity or any other CR activity. In that case, CR continues its

searching process in the next sensing step until it finds a free channel.

3. The CR arrives at a given channel and finds it free which means it will transmits

in that channel. In the meanwhile, another CR during the same time finds the

same channel free and it also transmits in that free channel. Ideally, a collision

will occur when both CRs will transmit in the same channel at the same time,

but in this thesis an efficient collision detection mechanism has been studied

which employs spatial reuse by taking advantage of the transmitting and receiv-

ing ranges of the transmitter-receiver node pairs of each CR in order to make

the decision whether collision will occur or not (details are presented in the next

chapter). If both the CRs are out of their transmitting and receiving ranges which

means that they won’t interrupt their data transmission, then there will be no col-

lision even if they are transmitting in the same channel, otherwise, there will be

a collision.

3.3.2. Sequential Channel Sensing Scenarios

Some times, CR thinks that the channel is busy of either PU activity or any other CR

activity but in reality it is free. It happens due to False alarms which could effect our

goal of improving overall system efficiency by increasing the number of successful

transmissions in a distributed CRN. Different sequential channel sensing scenarios has

been explained below:

1. In this channel sensing scenario, CR1 finds channel 5 free in its third sensing step

as shown in Figure 6. CR1 will keep control of the channel until the remainder

of the time slot. Whereas, CR2 and CR3 will collide with each other because

both of them finds channel 2 free in their first sensing step and also they have the

same sensing order. When two or more CRs finds the same channel free then a

collision will occur because every CR will try to transmit in the same channel.

19

Figure 6: Scenario (1)

2. In this channel sensing scenario, CR1 finds channel 4 free in its 3rd sensing step,

CR2 finds channel 1 free in its first sensing step, and CR3 finds channel 2 free

in its fourth sensing step after finding out that channel 4 and channel 1 are busy

because both of them have already been found free in the other steps as shown

in Figure 7.


3. In this channel sensing scenario, CR1 and CR2 will collide with each other be-

cause both of them have chosen the same PU-free channel index in their first

sensing step and both of them have found channel 3 free in their first sensing

step as depicted in Figure 8. On the other hand, CR3 finds channel 5 free in its

fourth sensing step.

20


4. In this channel sensing scenario, CR1 and CR2 have chosen the same sensing

order as shown in the Figure 9. Ideally, collision will occur between CR1 and

CR2 but they will not collide because CR1 has found channel 2 free in its first

sensing step and CR2 generates a false alarm in its first sensing step which avoids

collision and it finds channel 1 free in step 3. On the other hand, CR3 finds

channel 5, 2, and 1 busy because other CRs may have found them free in the

earlier steps.


21

4. CHANNEL SENSING ORDER SELECTION

The main goal of this research work is to study a channel sensing order selection strat-

egy with spatial reuse which will help to improve the average rate of successful trans-

missions in a distributed CRN and by successful transmission, it is meant that CR is the

only sole user to transmit in that particular channel because it has found that channel

free from any PU activity or any other CR activity.

4.1. Studied Strategy

Generally, when two or more autonomous CRs search for available free channel for

spectrum opportunities, there is usually a competition between the CRs to access the

free channel because the purpose of this competition from the CR prospective is to

become the sole user of that free channel. As a result of this competition, collisions

will occur among CRs because they simultaneously transmit in the same channel and

this will lead to an overall reduced throughput.

The main theme of this work is to find out a way to improve the average rate of

successful transmissions in a distributed CRN which means that the chances of find-

ing a channel free simultaneously for two or more CRs is reduced by adopting smart

channel sensing and efficient collision detection techniques. In this thesis, an adaptive

persistent strategy [23] has been studied for channel sensing order selection which will

help distributed CRs to avoid collision and allow them to improve the average rate of

successful transmissions.

In adaptive persistent strategy, we start with set of available sensing orders by denot-

ing it with L. The sensing order can either be chosen from the space of all permutation

of channel N or from a Latin Square as shown in Figure 10.

Figure 10: Different channel sensing order illustration. Space of all permutations for

N=4 channels is shown in (a), and a predefined Latin Square is illustrated in (b).

22

Sensing orders from a predefined Latin square (Figure 10 (b)) which is a N by Nmatrix of N channel indices and each channel index does not repeat itself in any row

or column of the matrix which means it appears only once in any row or column of

the matrix [44, 45, 23]. The sensing order of an autonomous CR can be any row of

the Latin square (Figure 10 (b)) or it can be any row of the space of all permutations

(Figure 10 (a)) depending on which strategy has been selected for channel sensing.

4.2. γ-Persistent Strategy

A γ-persistent strategy [23] with efficient collision detection has been studied in this

thesis, which make use of the successes and failures of a CR in using the current

sensing order in the prior time slots. The successes and failures of a CR in the prior

time slots can be tracked by the help of a success counter (SC) and a binary flag b.

The persistence factor of this strategy is γ, which is utilized as follows:

1. A fixed value of the persistence factor (γ ∈ (0, 1)) has been used by the au-

tonomous CR.

2. Success counter (SC) and false alarm probability is considered by the au-

tonomous CR to employ the persistence factor γ = 1 − ( 1SC−log

2(Pfa)

). This

approach also assumes that the autonomous CR has the ability to estimate its

false alarm probability.

Both these approaches will be explained in detail at the end of this chapter.

We start with γ-Persistent Strategy by letting each CR maintain a∣

∣L∣

∣-element prob-

ability vector p and we donate pi as the probability of detecting the ith sensing order

which can be taken from space of all permutation of channel indices N , or it can be

taken from a Latin Square. The step by step explanation of γ-Persistent Strategy is

given below:

4.2.1. Initialization

We begin with∣

∣L∣

∣-element probability vector p=[

1∣

∣L

∣

∣

, 1∣

∣L

∣

∣

, 1∣

∣L

∣

∣

, . . . , 1∣

∣L

∣

∣

]

and by setting

the binary flag and the success counter to b=SC=0.

4.2.2. Weighted Coin Toss

A weighted coin is tossed to choose a sensing order with pi as the probability of se-

lecting the ith sensing order. After the weighted coin toss, the channels starts sensing

sequentially in the pattern mentioned in the chosen sensing order.

4.2.3. Possible Outcomes

Following are the possible outcomes after sensing the channels sequentially:

23

1. Successful Transmission: When CR finds a channel free and in the mean while

no other PU or CR finds that channel free, the CR will transmit in that channel.

The CR updates pi and pj as pi = 1 and pj = 1, ∀j 6= i using the current sensing

order i. Thus, it will use the same sensing order to visit the channels in the next

slot and the CR sets SC=SC+1.

2. All Channels found busy: In this scenario, CR finds all channels busy which

means that in the current slot the CR using the sensing order i find all the chan-

nels occupied by either a PU or any other CR. The CR updates pi and pj as

pi = 1 and pj = 0, ∀j 6= i using the current sensing order i.Thus, it will use the

same sensing order to visit the channels in the next slot and CR then sets b=1.

3. Collision between CRs: When a CR collides with any other CR in the current

slot using sensing order i, the CR updates pi as follows:

pi =

{

1/∣

∣L∣

∣ SC=0 and b=1

γpi, otherwise

and the CR updates pj as follows

pj =

{

1/∣

∣L∣

∣ if SC=0 and b=1

γpi +1−γ∣

∣L

∣

∣

−1, otherwise

In the current slot, when a collision occurs the CR randomly selects a sensing

order in case of SC=0 and b=1; else ways the probability of selecting sensing or-

der i reduces multiplicatively as the CR evenly distributes the probability across

the other sensing orders. Now, the CR sets SC=b=0. This is further explained by

going through all the different states of the success counter and binary flag b on

experiencing the collision as follows:

When SC > 0 and b = 0 or b = 1, an autonomous CR is not sure whether the

sensing order it has selected was not selected by any other CR, whereas SC > 0

indicates that there are high chances that the CR is the only user of that sensing

order in time slot n. Once a collision occur, a CR does not change its sensing

order, whereas it continues with the same sensing order i with probability γpi,and the probability (1 − γpi) is reassigned uniformly across the other sensing

orders. This will enhance the average rate of successful transmissions because

the successful CRs will continue with the same sensing order which will directly

reduce the number of CRs randomly selecting a sensing order.

When SC = 0 and b = 0, it means that the CR was neither successful nor found

all channels busy in the current time slot using the sensing order i before expe-

riencing the collision. Thus, the probability of choosing the same sensing order

in the next time slot decreases because there is a possibility that any other CR

may have been successful in the same sensing order and it may continue with the

same sensing order, and the probability (1− γpi) is reassigned uniformly across

the other sensing orders.

When SC = 0 and b = 1, it indicates that all channel were found busy by CR t

in time slot n using the sensing order i. As CR t stays quiet because it is not

24

sure that whether CR t is the only user of sensing order in time slot n because

it may happen that any other CR have chosen the same sensing order and it is

quite possible that it may have found it free and thus transmitted or it may have

found all channels busy. Therefore, after experiencing the collision the CR t will

choose the sensing order independently and randomly with uniform probability

because SC = 0 and it cannot be sure that it is the sole user of sensing order in

time slot n.

4.2.4. Restart Again

Once any of the possible outcomes in section 4.2.3 happens, the process starts again

by returning to section 4.2.2 and the same process continues again.

4.3. Collision Detection

A collision occurs in a distributed CRN whenever CR fails to receive an acknowl-

edgment (ACK) for a transmitted data frame. As in our case, we have a common

predefined Latin Square from which the CR selects the sensing order. Two or more

CRs will only collide with each other once they try to transmit simultaneously after

choosing the same sensing order.

4.3.1. Spatial Reuse

Spatial reuse gives the freedom to use the same band which is at far off location simul-

taneously with out any interference, whereas, a collision will occur if the same band is

simultaneously used in a neighbouring location [46]. Thus, spatial reuse provides an

opportunity to SUs to use the same band simultaneously at a far off location with out

any interference.

Spatial reuse is generally achieved with careful planning of the interference parame-

ters which are set for SUs to improve the overall wireless system bandwidth efficiency.

For every configuration of the network, there exists a minimum distance which must be

separating SUs from each other in order to transmit simultaneously in the same band

[47]. As a result, maximum no of simultaneous transmissions are achieved which al-

lows to attain the maximum network throughput.

In this thesis, we have utilized spatial reuse in order to enhance the efficiency of

collision detection of the studied adaptive persistent strategy. The details are explained

in the section below.

4.3.2. Proposed Efficient Collision Detection with spatial reuse

Ideally, a collision occurs when two or more CRs transmits at the same time and they

have the same sensing order but there are various other factors which could be em-

ployed to reduce the likelihood of collisions and to improve the overall system per-

27

5. PERFORMANCE ANALYSIS OF THE PROPOSED

STRATEGY

In this chapter, we analyze the performance of our studied adaptive scheme with ef-

ficient collision detection mechanism by comparing it with a randomize after every

collision adaptive (RAND) strategy [23] with the help of numerical results in terms

of average number of successful transmissions in a network. A comparison of the

proposed strategy is also done in case when there is no utilization of spatial reuse.

In the first section, RAND strategy is discussed and in the next section simulation

results are presented for the performance analysis of the studied adaptive persistent

strategy.

5.1. Randomize after every collision strategy

In randomize after every collision strategy, adaptive randomization is utilized which is

based on feedback for CRs to reduce the likelihood of collision and thus increasing the

average number of successful transmissions in a network.

In this strategy, a CR selects independently and randomly a sensing order which

may come from the space of all permutations of N channels or from a Latin-square

with equal probability. A CR can only select the other sensing order in a time slot,

when a collision has occurred in the previous time slot, otherwise, it sticks with its

sensing order. The main motive is to reduce the likelihood of collision by randomizing

the sensing orders of CRs, only when there is a collision in the previous time slot.

5.2. Simulation results

In this section, numerical results are presented to illustrate the performance of adap-

tive persistent strategy with efficient collision detection mechanism. These results are

utilized to analyze the performance of our studied strategy in terms of average number

of successful transmission in a given time slot considering different parameters and

scenarios which are explained in detail in this section.

For simulation purposes, a distributed cognitive radio network has been consid-

ered which consists of M cognitive transmitter-receiver pairs and a set N=1,2,3,....,N

of channels. M cognitive transmitter/receiver pairs are randomly distributed over

a 150m*150m square area. The separation distance between a cognitive transmit-

ter/receiver pair is kept constant for all M cognitive transmitter/receiver pairs as ex-

plained in system model.

In adaptive persistent strategy, spatial reuse is utilized in which a collision will only

occur when two CRs finds the same channel free and the receiver of one CR falls in

the transmitting range of another CR, otherwise there will be no collision even when

two CRs are transmitting in the same channel simultaneously. We have analyzed the

impact of spatial reuse by varying the transmitting and receiving ranges of cognitive

transmitter/receiver pairs and observed its impact on the average number of successful

transmission in a given time slot. In our scenarios, we have varied the transmitting

28

and receiving ranges from 70 meters to 200 meters over a 150m*150m square area in

which M cognitive transmitter/receiver pairs are randomly distributed.

Different scenarios are considered in this section in which number of channels are

kept constant (N=16), but number of M cognitive transmitter/receiver pairs are varied

in order to determine the impact of average number of successful transmission in a

given time slot when we vary M. We are more interested in the cases when M>N in

order to determine the efficiency of our studied scheme when it has to face the problem

of spectrum scarcity (M>N).

The different scenarios are explained as follows:

5.2.1. Scenario A

In this scenario, a distributed cognitive radio network of M=16 cognitive transmit-

ter/receiver pairs and a set N=1,2,3,....,16 of channels is considered. M cognitive

transmitter/receiver pairs are randomly distributed over a 150m*150m square area.

The separation distance between a cognitive transmitter/receiver pair is kept constant

for all M cognitive transmitter/receiver pairs.

In Figure 13, for N=16 channels and M=16 cognitive transmitter/receiver pairs, the

average number of successful transmission for different strategies in a given time slot

is depicted. There are five different plots shown in Figure 13 which gives us simu-

lation results based on different transmission and reception ranges for the cognitive

transmitter/receiver pair.

It can be seen from Figure 13 that our proposed adaptive persistent strategy with

efficient collision detection achieves the highest average number of successful trans-

mission in all five different plots.

When we keep on increasing the transmission and reception ranges for the cognitive

transmitter/receiver pair over a 150m*150m square area, it will increase the probability

of collision because now there are more chances that a receiver of one CR falls in the

transmitting range of another CR due to increase in the transmission range. This will

reduce the average number of successful transmissions in a given time slot which can

be seen in Figure 13. In this case, our strategy with high persistence value γ=0.9

performs slightly better, as compared to when γ is used as a function of false alarm

probability and SC as shown in Figure 13(e). This happens due to the fact that high

persistence value allows CR to continue with its sensing order with high probability.

Efficient collision detection plays a pivotal role in increasing the average number

of successful transmission by reducing the likelihood of collision with help of spatial

reuse. Figure 14 gives us simulation results for the adaptive persistent strategy pro-

posed in [23] for the case when M=N=16. There is no spatial reuse in this case when

two CRs selects the same sensing order and they wants to transmit simultaneously.

29

(a) Range=70 meters

-

(b) Range=90 meters

(c) Range=110 meters

-

(d) Range=130 meters

(e) Range=200 meters

Figure 13: Simulation results for average number of successful transmissions in a given

time slot for different strategies when M=N=16, false alarm probability of each CR is

set to 0.2, and θk = 0,3, ∀k ∈ N

30


time slot for different strategies when M=N=16, false alarm probability of each CR is

set to 0.2, and θk = 0,3, ∀k ∈ N

By comparing the simulation result in Figure 13 and Figure 14. It is quite clear

that our strategy has the highest average number of successful transmissions in case

when we have M=N. This happens due to the fact that spatial reuse helps to reduce

the likelihood of collision by allowing CRs to transmit in the same channel when the

receiver of one CR does not falls in the transmitting range of another CR.

5.2.2. Scenario B











31

(a) Range=70 meters

-

(b) Range=90 meters


-




time slot for different strategies when M=20 and N=16, false alarm probability of each

CR is set to 0.2, and θk = 0,3, ∀k ∈ N

It can be seen from Figure 15 that adaptive persistent strategy with efficient collision

detection achieves the highest average number of successful transmission in all five

different plots. In case of M>N, the likelihood of collision increases which happens

due to spectrum scarcity as we have now more CRs then number of available channels.




transmitting range of another CR due to increase in the transmission range.This will

32


be seen in Figure 15. In this case, our strategy with high persistence value γ=0.9 and

spatial reuse performs slightly better, as compared to when γ is used as a function

of false alarm probability and SC as shown in Figure 15(e). This happens due to the

fact that high persistence value allows CR to continue with its sensing order with high

probability.

Spatial reuse plays a pivotal role in increasing the average number of successful

transmission by reducing the likelihood of collision. Figure 14 gives us simulation

results for the adaptive persistent strategy proposed in [23] for the case when M =20

and N=16. There is no spatial reuse in this case when two CRs selects the same sensing

order and they wants to transmit simultaneously.




By comparing the simulation result in Figure 15 and Figure 16. It is quite clear that

our strategy with spatial reuse is performing significantly better than the one proposed

in [23] in terms of average number of successful transmissions for the case when we

have M=20 and N=16. This happens due to the fact that spatial reuse helps to reduce


receiver of one CR do not falls in the transmitting range of another CR.

33

5.2.3. Scenario C






(a) Range=70 meters

-

(b) Range=90 meters


-






34

In Figure 17, for N=16 channels and M=24cognitive transmitter/receiver pairs, the





It can be seen from Figure 17 that adaptive persistent strategy with efficient collision

detection achieves the highest average number of successful transmissions when we

have high persistence value γ=0.9 in all five different plots. As in case of M>N, the

likelihood of collision increases which happens due to spectrum scarcity, as there are

more CRs then number of available channels.




transmitting range of another CR due to increase in the transmission range. This will


be seen in Figure 17. In this case, high persistence value γ=0.9 performs slightly

better, as compared to when γ is used as a function of false alarm probability and SC.

This happens due to the fact that high persistence value allows CR to continue with its

sensing order with high probability.




35


transmissions by reducing the likelihood of collision. Figure 18 gives us simulation

results for the adaptive persistent strategy proposed in [23] for the case when M =24

and N=16. There is no spatial reuse in this case when two CRs selects the same sensing



our strategy is performing significantly better than the one proposed in [23] in terms

of average number of successful transmissions for the case when we have M=24 and

N=16. This happens due to the fact that our proposed efficient collision detection helps

to reduce the likelihood of collision by allowing CRs to transmit in the same channel

when the receiver of one CR do not falls in the transmitting range of another CR.

5.2.4. Scenario D











It can be seen from Figure 19 that results are quite random because M»N in all five

different plots. This will increase the likelihood of collision and now there are more

chances a collision will occur which will result in reduce overall network throughput.

Figure 20 gives us simulation results for the adaptive persistent strategy proposed

in [23] for the case when M =32 and N=16. There is no smart and efficient collision

detection utilized in this case when two CRs selects the same sensing order and they

wants to transmit simultaneously.

36

(a) Range=70 meters

-

(b) Range=90 meters


-






37




By comparing the simulation result in Figure 19 and Figure 20. It can be seen that

the simulations results are quite similar because of the randomness in the plots due to

M >> N .

5.2.5. Scenario E






In Figure 21, for N=10 channels and M=16 cognitive transmitter/receiver pairs, it

compares average number of successful transmission achieved in timeslots, for n=200,

as a function of number of CRs for different adaptive strategies. There are six dif-

ferent plots shown in Figure 21 which gives us simulation results based on different

transmission and reception ranges for the cognitive transmitter/receiver pair.

38

(a) Caption1

-

(b) Caption 2

(c) Caption12

-

(d) Caption 22

(e) Caption 212

-

(f) Caption 212

Figure 21: Caption for this figure with two images

It can be seen from Figure 21 that adaptive persistent strategy with spatial reuse

achieves the highest average number of successful transmissions in all six different

plots.



of collision because now there are more chances that a receiver of one CR falls in

the transmitting range of another CR due to increase in the transmission range. This

will reduce the average number of successful transmissions in a given time slot which

can be seen in Figure 21. In case of M>N, our strategy with high persistence value

gamme=0.9 and spatial reuse performs slightly better, as compared to when gamma

39

is used as a function of false alarm probability and SC as shown in Figure 21. This

happens due to the fact that high persistence value allows CR to continue with its

sensing order with high probability.


transmission by reducing the likelihood of collision. Figure 14 gives us simulation

results for the adaptive persistent strategy proposed in [23] for the case when M=16

N=10. There is no spatial reuse in this case when two CRs selects the same sensing


Figure 22: Efficient Collision Detection Scenario (a)


our strategy has the highest average number of successful transmissions in case when

we have M=N and M>N. This happens due to the fact that spatial reuse helps to reduce


receiver of one CR do not falls in the transmitting range of another CR.

40

6. SUMMARY

Generally, when two or more autonomous CRs search for available free channel for

spectrum opportunities, there is usually a competition between the CRs to access the

free channel. The purpose of this competition from the CR prospective is to become

the sole user of that free channel. As a result of this competition, collisions will occur

among CRs because they simultaneously transmit in the same channel which will lead

to an overall reduced throughput.

The main theme of this research work is to find out a way to improve the average

rate of successful transmissions in a distributed CRN when number of available free

channels are less then the number of autonomous CRs competing to access these free

channels. This can be achieved by reducing the probability of finding a channel free

simultaneously for two or more CRs by adopting smart channel sensing and efficient

collision detection techniques.

A adaptive persistent strategy with spatial reuse has been studied in this thesis for

efficient channel access in a distributed CRN. We found out that the efficiency of the

wireless system reduces when channels are randomly accessed. An adaptive persistent

strategy with efficient spatial reuse is the answer for the reduced efficiency caused by

random selection of sensing orders.

The likelihood of collision between CRs that are transmitting in the same chan-

nel can be reduced by utilizing the transmitting and receiving ranges of the CR’s

transmitter-receiver node pair with the help of spatial reuse. Simulation results have

validated that adaptive persistent strategy with spatial reuse achieves the highest num-

ber of average successful transmissions in a given time slot as compared to RAND

strategy and also for the case when there is no spatial reuse for the adaptive persistent

strategy.

In simulation results, when we increase the transmitting and receiving ranges of

the CR’s transmitter-receiver node pair, it reduces the average number of successful

transmissions in a given time slot because now there are more chances that a transmitter

of one CR falls in the receiving range of another CR which will lead to a collision when

both CRs try to transmit at the same time. High persistence value γ=0.9 and spatial

reuse performs slightly better, as compared to when γ is used as a function of false

alarm probability and SC, for the case when probability of collision is high.

We also found out that when adaptations are employed, it will increase the average

number of successful transmissions when channel sensing orders are selected from a

predefined Latin-square. We also investigated the impact of false alarms and channels

errors on system efficiency. The average number of successful transmissions increases

even when there are false alarms and channels errors in our studied strategy.

41

7. REFERENCES

[1] Heinzelman W.B. (2000) Application Specific Protocol Architectures for

Wireless Networks. Ph.D. thesis. URL: http://www-mtl.mit.edu/

researchgroups/icsystems/pubs/theses/wendi_phd_2000.

pdf.

[2] Liu S., Wu J., Koh C.H. & Lau V.K. (2011) A 25 gb/s (/km 2) urban wireless

network beyond imt-advanced. Communications Magazine, IEEE 49, pp. 122–

129.

[3] Kolios P., Friderikos V. & Papadaki K. (2011) Future wireless mobile networks.

Vehicular Technology Magazine, IEEE 6, pp. 24–30.

[4] FCC, Facilitating opportunities for flexible, efficient and reliable spectrum

use employing cognitive radio technologies: Notice of proposed rule mak-

ing and order (Dec. 2003). URL: www.eecs.berkeley.edu/~dtse/3r_

Cognitive_radio_FCC-03-322A1.doc.

[5] Mitola III J. (2000) Cognitive Radio: An Integrated Agent Architecture for Soft-

ware Defined Radio. Ph.D. thesis. URL: http://www.diva-portal.org/

smash/get/diva2:8730/FULLTEXT01.pdf.

[6] Mitola III J. (1999) Cognitive radio for flexible mobile multimedia communica-

tions. In: Mobile Multimedia Communications, 1999.(MoMuC’99) 1999 IEEE

International Workshop on, IEEE, pp. 3–10.

[7] Haykin S. (2005) Cognitive radio: brain-empowered wireless communications.

Selected Areas in Communications, IEEE Journal on 23, pp. 201–220.

[8] NTIA, Facilitating opportunities for flexible, efficient, and reliable

spectrum use employing cognitive radio technologies (2005). URL:

https://www.ntia.doc.gov/legacy/ntiahome/fccfilings/

2005/cogradio/ETDocket03-108_02152005.htm.

[9] SDRF-06-P-0009-V1.0.0 S.F.A.D., Cognitive radio definitions and

nomenclature (2008). URL: http://www.sdrforum.org/pages/

documentLibrary/documents/SDRF-06-P-0009-V1_0_0_CRWG_

Defs.pdf.

[10] Thomas R.W., Friend D.H., Dasilva L.A. & Mackenzie A.B. (2006) Cognitive

networks: adaptation and learning to achieve end-to-end performance objectives.

Communications Magazine, IEEE 44, pp. 51–57.

[11] Mangold S., Jarosch A. & Monney C. (2005) Cognitive radio-trends and research

challenges. Swisscom Comtec Magazine 3, pp. 6–9.

[12] Slingerland A.M., Pawelczak P., Prasad R.V., Lo A. & Hekmat R. (2007) Per-

formance of transport control protocol over dynamic spectrum access links. In:

New Frontiers in Dynamic Spectrum Access Networks, 2007. DySPAN 2007.

2nd IEEE International Symposium on, IEEE, pp. 486–495.

42

[13] Mitola III J. (2007) Cognitive Networks. Ph.D. thesis. URL:

https://theses.lib.vt.edu/theses/available/

etd-07172007-150149/unrestricted/Thomas_

CognitiveNetworksDiss4.pdf.

[14] Huang S., Liu X. & Ding Z. (2008) Opportunistic spectrum access in cognitive

radio networks. In: INFOCOM 2008. The 27th Conference on Computer Com-

munications. IEEE, IEEE, pp. 1427–1435.

[15] Zhao Q., Geirhofer S., Tong L. & Sadler B.M. (2008) Opportunistic spectrum

access via periodic channel sensing. Signal Processing, IEEE Transactions on

56, pp. 785–796.

[16] Fan R. & Jiang H. (2010) Optimal multi-channel cooperative sensing in cognitive

radio networks. Wireless Communications, IEEE Transactions on 9, pp. 1128–

1138.

[17] Fan R. & Jiang H. (2009) Channel sensing-order setting in cognitive radio net-

works: A two-user case. Vehicular Technology, IEEE Transactions on 58, pp.

4997–5008.

[18] Chang N.B. & Liu M. (2008) Competitive analysis of opportunistic spectrum

access strategies. In: INFOCOM 2008. The 27th Conference on Computer Com-

munications. IEEE, IEEE, pp. 1535–1542.

[19] Liu K. & Zhao Q. (2010) Distributed learning in multi-armed bandit with multiple

players. Signal Processing, IEEE Transactions on 58, pp. 5667–5681.

[20] Cheng H.T. & Zhuang W. (2011) Simple channel sensing order in cognitive radio

networks. Selected Areas in Communications, IEEE Journal on 29, pp. 676–688.

[21] Theis N.C., Thomas R.W. & DaSilva L.A. (2011) Rendezvous for cognitive ra-

dios. Mobile Computing, IEEE Transactions on 10, pp. 216–227.

[22] DaSilva L.A. & Guerreiro I. (2008) Sequence-based rendezvous for dynamic

spectrum access. In: New Frontiers in Dynamic Spectrum Access Networks,

2008. DySPAN 2008. 3rd IEEE Symposium on, IEEE, pp. 1–7.

[23] Khan Z., Lehtomaki J.J., DaSilva L.A. & Latva-aho M. (2013) Autonomous

sensing order selection strategies exploiting channel access information. Mobile

Computing, IEEE Transactions on 12, pp. 274–288.

[24] Liang Y.C., Zeng Y., Peh E.C. & Hoang A.T. (2008) Sensing-throughput tradeoff

for cognitive radio networks. Wireless Communications, IEEE Transactions on

7, pp. 1326–1337.

[25] Jia J., Zhang Q. & Shen X. (2008) Hc-mac: A hardware-constrained cogni-

tive mac for efficient spectrum management. Selected Areas in Communications,

IEEE Journal on 26, pp. 106–117.

43

[26] Kim S.J. & Giannakis G.B. (2010) Sequential and cooperative sensing for multi-

channel cognitive radios. Signal Processing, IEEE Transactions on 58, pp. 4239–

4253.

[27] Pei Y., Liang Y.C., Teh K.C. & Li K.H. (2011) Energy-efficient design of sequen-

tial channel sensing in cognitive radio networks: optimal sensing strategy, power

allocation, and sensing order. Selected Areas in Communications, IEEE Journal

on 29, pp. 1648–1659.

[28] Xu Y., Gao Z., Wang J. & Wu Q. (2011) Multichannel opportunistic spectrum

access in fading environment using optimal stopping rule. In: Wireless Commu-

nications and Applications, Springer, pp. 275–286.

[29] Chang N.B. & Liu M. (2007) Optimal channel probing and transmission schedul-

ing for opportunistic spectrum access. In: Proceedings of the 13th annual ACM

international conference on Mobile computing and networking, ACM, pp. 27–38.

[30] Xu Y., Wu Q., Wang J., Shen L. & Anpalagan A. (2015) Robust multiuser se-

quential channel sensing and access in dynamic cognitive radio networks: Poten-

tial games and stochastic learning. Vehicular Technology, IEEE Transactions on

64, pp. 3594–3607.

[31] Azarfar A., Liu C.H., Frigon J.F., Sanso B. & Cabric D. (2014) Cooperative

spectrum sensing scheduling optimization in multi-channel dynamic spectrum

access networks. In: Global Communications Conference (GLOBECOM), 2014

IEEE, IEEE, pp. 810–815.

[32] Yang G., Wang J., Luo J., Wen O., Li H., Qiang L. & Li S. (2013) Cooperative

spectrum sensing in heterogeneous cognitive radio networks based on normalized

energy detection .

[33] Wang J., Macaluso I. & DaSilva L.A. (2014) Perfect versus imperfect monitor-

ing in multi-agent opportunistic channel access. In: Cognitive Radio Oriented

Wireless Networks and Communications (CROWNCOM), 2014 9th International

Conference on, IEEE, pp. 480–485.

[34] Akin S. & Gursoy M.C. (2015) Performance analysis of cognitive radio systems

with imperfect channel sensing and estimation. Communications, IEEE Transac-

tions on 63, pp. 1554–1566.

[35] Macaluso I., Ahmadi H. & DaSilva L.A. (2015) Fungible orthogonal channel

sets for multi-user exploitation of spectrum. Wireless Communications, IEEE

Transactions on 14, pp. 2281–2293.

[36] Shokri-Ghadikolaei H. & Fischione C. (2014) Analysis and optimization of dis-

tributed random sensing order in cognitive radio networks. arXiv preprint arXiv

1401.

44

[37] Barcelo J., Bellalta B., Cano C., Sfairopoulou A. & Oliver M. (2009) Carrier

sense multiple access with enhanced collision avoidance: A performance analy-

sis. In: Proceedings of the 2009 International Conference on Wireless Commu-

nications and Mobile Computing: Connecting the World Wirelessly, ACM, pp.

733–738.

[38] Barcelo J., Bellalta B., Cano C. & Oliver M. (2009) Learning-beb: avoiding colli-

sions in wlan. Carrier Sense Multiple Access with Enhanced Collision Avoidance

, p. 23.

[39] Fang M., Malone D., Duffy K.R. & Leith D.J. (2013) Decentralised learning

macs for collision-free access in wlans. Wireless networks 19, pp. 83–98.

[40] Ning Z., Song Q., Guo L. & Jamalipour A. (2013) Throughput improvement

by network coding and spatial reuse in wireless mesh networks. In: 2013 IEEE

Global Communications Conference (GLOBECOM), IEEE, pp. 4572–4577.

[41] Do T. & Mark B.L. (2010) Joint spatial–temporal spectrum sensing for cognitive

radio networks. IEEE Transactions on Vehicular Technology 59, pp. 3480–3490.

[42] Peh E. & Liang Y.C. (2007) Optimization for cooperative sensing in cognitive ra-

dio networks. In: Wireless Communications and Networking Conference, 2007.

WCNC 2007. IEEE, IEEE, pp. 27–32.

[43] Anandkumar A., Michael N., Tang A. & Swami A. (2011) Distributed learning

and allocation of cognitive users with logarithmic regret. IEEE JSAC on Ad-

vances in Cognitive Radio Networking and Communications 29, pp. 731–745.

[44] Laywine C.F. & Mullen G.L. (1998) Discrete mathematics using Latin squares,

vol. 49. John Wiley & Sons.

[45] Davis P.J. (2012) Circulant matrices. American Mathematical Soc.

[46] Kasbekar G.S. & Sarkar S. (2010) Spectrum pricing games with bandwidth un-

certainty and spatial reuse in cognitive radio networks. In: Proceedings of the

eleventh ACM international symposium on Mobile ad hoc networking and com-

puting, ACM, pp. 251–260.

[47] Zhu J., Guo X., Yang L.L. & Conner W.S. (2004) Leveraging spatial reuse in

802.11 mesh networks with enhanced physical carrier sensing. In: Communica-

tions, 2004 IEEE International Conference on, vol. 7, IEEE, vol. 7, pp. 4004–

4011.

MASTER’S THESIS UNDERSTANDING THE IMPACT OF SPATIAL...

Documents

Transcript of MASTER’S THESIS UNDERSTANDING THE IMPACT OF SPATIAL...