以排隊理論為基礎對感知無線網路頻譜管理技術之研究

(1)

國立交通大學

電信工程研究所

博士論文

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

研究生：王中瑋

指導教授：王蒞君

(2)

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

研究生：王中瑋

Student:

Chung-Wei

Wang

指導教授：王蒞君博士 Advisor:

Dr.

Li-Chun

Wang

國立交通大學

電信工程研究所

博士論文

A Dissertation

Submitted to Institute of Communication Engineering

College of Electrical and Computer Engineering

National Chiao Tung University

in Partial Fulfillment of the Requirements

for the Degree of Doctor of Philosophy

in

Communication Engineering

Hsinchu, Taiwan

(3)

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

研究生：王中瑋指導教授：王蒞君博士

國立交通大學

電信工程研究所

摘要

本論文探討感知無線網路的頻譜管理問題。在此網路中，來自主要使用者的『多次中斷』將大大地影響次要使用者的通訊效能。每當次要使用者被主要使用者中斷時，次要使用者必須選擇一個適合的通道進行頻譜切換，以便繼續未完成的傳輸。很明顯地，『多次中斷』將造成多次的頻譜切換，並且增加次要使用者連線的傳輸延遲。為了從一個宏觀的角度來分析感知無線網路下『多次中斷』行為對『次要使用者連線』所造成的傳輸延遲，本論文提出一個優先權排隊理論的分析模型替感知無線網路的頻譜使用行為進行建模。藉由此模型，我們分析次要使用者的一個重要服務品質參數：『完整系統時間』。在此論文中，基於排隊理論分析模型，我們發展具服務品質考量的頻譜管理機制，其中包括 (1) 頻譜選擇機制、(2) 頻譜切換機制、和 (3) 頻譜分享機制的設計與討論。針對這些機制的具體研究成果敘述如下：(1) 針對頻譜選擇問題，我們提出一個具有負載平衡功效的頻譜選擇機制來優化次要使用者的『完整系統時間』；(2) 針對頻譜切換問題，我們量化在多通道下多次頻譜切換對次要使用者所造成的『完整系統時間』增加量；(3) 針對頻譜分享問題，我們提出一個允入控制機制來避免主要使用者被次要使用者干擾並優化次要使用者的『完整系統時間』。我們完整探討這三種頻譜管理機制對次要使用者所造成的傳輸延遲。基於這些分析結果，在不同資料到達率與服務時間分佈下，我們可以設計相對應的頻譜管理機制來增強次要使用者連線的傳輸品質。總而言之，本論文的主要貢獻是提出一個以排隊理論為基礎的分析模型並用多樣化的角度與觀點來對感知無線網路效能進行分析。本論文所建議之模型

(4)

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

Student: Chung-Wei Wang Advisor: Dr. Li-Chun Wang

Department of Electrical Engineering

National Chiao Tung University

Abstract

In this dissertation, we investigate spectrum management techniques in cognitive radio (CR) networks with quality of service (QoS) provisioning. One fundamental issue in enhancing QoS performance for the secondary users is the multiple interruptions from the primary users during each secondary user’s connection. These interruptions from the primary users result in the phenomenon of multiple spectrum handoffs within one secondary user’s con-nection. Thus, a set of target channels for spectrum handoffs are needed to be selected sequentially. In order to characterize the general channel usage behaviors with multiple handoffs from a macroscopic viewpoint, an analyti-cal framework based on the preemptive resumption priority (PRP) M/G/1 queueing theory is introduced. Based on the PRP M/G/1 queueing network model, we can evaluate the effects of multiple handoffs on the overall system time, which is an important QoS performance measure for the secondary connections in CR networks.

(5)

The proposed analytical framework can provide important insights into the design of spectrum management techniques in CR networks. In order to demonstrate the effectiveness of this analytical model, we discuss various spectrum management techniques, consisting of spectrum decision, spectrum sharing, and spectrum mobility. For the spectrum decision issue, we show how to determine which channels are required to probe and transmit. For the spectrum mobility issue, we illustrate how to characterize the effects of multiple handoffs, where the secondary users can have different operating channels before and after spectrum handoff. For the spectrum sharing issue, we explore how to determine the optimal admission probability to avoid the interference between primary and secondary users in the presence of false alarm and missed detection. From numerical results, we can develop traffic-adaptive spectrum management policies to enhance the QoS performance of the secondary users in CR networks with various traffic arrival rates and service distributions.

To summarize, the main contribution of this dissertation is to investi-gate the modeling techniques for CR networks from a macroscopic viewpoint based on the queueing theory. The proposed analytical framework can help analyze the performances of CR networks and provide important insights into the design of various spectrum management techniques with enhanced QoS performances.

(6)

Acknowledgements

First of all, I want to express my deeply gratitude to my advisor, Prof. Li-Chun Wang. Not only the important insights to research problems, encour-agement, and support, he also shows me a way of being optimistic to face difficulties. Without his advice, guidance, comments, and all that, this work could not have been done. He indeed opened a door to the future for me.

Special thanks to my mates of Wireless System Laboratory in National Chiao Tung University. They gave me kindly help in many aspects in my study. Drs. Chih-Wen Chang, Anderson Chen, Wei-Cheng Liu, and Jane-Hwa Huang gave me many valuable suggestions and ideas in my research. Messrs. Chu-Jung Yeh, Samer Talat, and Ang-Hsun Tsai encouraged me every time when I felt frustrated. I was so lucky to have all these lab mates. Most importantly, I would like to thank Prof. Fumiyuki Adachi. During the study in Tohoku University from April, 2009 to March, 2010, he give me many valuable comments. Furthermore, I also wish to thank my mates of Wireless Signal Processing and Networking Laboratory in Tohoku University, especially for Prof. Wei Peng, Mr. Takeda, and Mr. Guan Gui.

Finally, my thanks would go to my wife, Hui-Cheng Chuang, and beloved family for their loving considerations and great confidence in me all through these years. I also want to thank Hsing Tian Kong Culture and Education Development Foundation. Because of their scholarship supports, I can

(7)

con-centrate our attention on my research. I also owe my sincere gratitude to my friends who gave me their help and time in listening to me and helping me work out my problems during the difficult course of the thesis.

(8)

List of Tables

2.1 Comparison of Various Analytical Models for CR Networks. . 15

2.2 Comparison of Various Load-balancing Spectrum Decision Schemes for Cognitive Radio Networks, where PP, GT, and LA stand for the packet-wise probabilistic, game-theoretic, and learning automata approaches, respectively. . . 17

2.3 Comparison of Various Proactive Handoff Models. . . 21

2.4 Comparison of Various Channel Usage Models. . . 26

2.5 Comparison of Various Objective Functions. . . 29

(16)

List of Figures

1.1 An illustrative example of CR network, which consists a mary network and a secondary network. There are three pri-mary users (PUs) and one secondary user (SU) in the pripri-mary

and secondary networks, respectively. . . 2

1.2 During the transmission period of secondary user (SU), it

ex-periences multiple handoffs. . . 3

1.3 Relationships between spectrum sensing, spectrum decision,

spectrum mobility, and spectrum sharing functionalities. . . . 5

1.4 Outline of this dissertation. . . 12

3.1 Illustration of transmission procedures in a two-channel sys-tem. The gray areas indicate that the channels are occupied by the existing primary users’ connections (PCs) or the other

secondary users’ connections (SCs). . . 35

3.2 The PRP M/G/1 queueing network model with three

chan-nels. λ(k)p , λ(k)s , and ωn(k) are the arrival rates of the

pri-mary connections, the secondary connections, and the

type-n secotype-ndary cotype-ntype-nectiotype-ns (type-n ≥ 1) at chatype-ntype-nel k. Note that

ω₀(k) = λ(k)s . Furthermore, fp(k)(x) and f_i(k)(φ) are the pmfs of

(17)

3.3 Illustration of the physical meaning of random variable Φ(k)_i .

For example, Φ(1)₂ is one of the third segments (gray areas) of

the first, the third, and the fourth secondary connections. . . . 45

4.1 Spectrum decision behavior model. . . 50 4.2 Example of the overall system time of the secondary

connec-tion SCA. The white areas indicate that channel is occupied

by SCA. Furthermore, the gray areas indicate that channel is

occupied by the primary connections (PCs) and its duration is the busy period resulting from transmissions of the primary

connections. Here, SCAencounters two interruptions from the

primary connections during its transmission period. . . 52 4.3 Performance model for the probability-based channel selection

scheme where the channel usage behaviors are characterized by the PRP M/G/1 queueing systems. . . 55 4.4 Performance model for the sensing-based channel selection

scheme where the channel usage behaviors are characterized by the PRP M/G/1 queueing systems. . . 56 4.5 Optimal distribution probability vector for the

probability-based spectrum decision with various arrival rates of the

sec-ondary connections, where PF = 0.1, PM = 0.1, and E[Xs] = 10. 68

4.6 Channel busy probability for the probability-based spectrum decision with various arrival rates of the secondary

connec-tions, where PF = 0.1, PM = 0.1, and E[Xs] = 10. . . 69

4.7 Channel busy probability for the probability-based spectrum decision with various arrival rates of the secondary

(18)

4.8 Optimal distribution probability vector for the probability-based spectrum decision with various arrival rates of the

sec-ondary connections, where PM = 0.1, λs= 0.03, and E[Xs] = 15. 72

4.9 Overall system time for the sensing-based spectrum decision

with various numbers of candidate channels n, where PF = 0.1,

PM = 0.1, τ = 2, and E[Xp] = 20. . . 73

4.10 Overall system time for the sensing-based spectrum decision

with various numbers of candidate channels n, where PM =

0.1, τ = 2, E[Xp] = 20, and E[Xs] = 5. . . 74

4.11 Comparison of the overall system time for three considered

spectrum decision schemes, where PF = 0.1, PM = 0.1, and

E[Xs] = 10. . . 76

5.1 An example of transmission process for the secondary

connec-tion SCA, where ts is the channel switching time, T is the

ex-tended data delivery time of SCA, and Di is the handoff delay

of the ith _{interruption. The gray areas indicate that the}

chan-nels are occupied by the existing primary connections (PCs)

or secondary connections (SCs). Because SCA is interrupted

three times in total, the overall data connection is divided into four segments. . . 83 5.2 The PRP M/G/1 queueing network model with three

chan-nels where λ(k)p , λ(k)s , and ωn(k) are the arrival rates of the

ω₀(k) = λ(k)s . Furthermore, fp(k)(x) and f_i(k)(φ) are the pdfs of

(19)

5.3 Effects of Pareto and exponential service time distributions for primary connections on the extended data delivery time (E[Tchange]) of the secondary connections when the

always-changing spectrum handoff sequence is adopted, where ts = 1

(slot), λs = 0.01 (arrivals/slot), E[Xs] = 10 (slots/arrival),

and E[Xp] = 20 (slots/arrival). . . 99

5.4 Effects of Pareto and exponential service time distributions for primary connections on the extended data delivery time

(E[Tstay]) of the secondary connections when the always-staying

spectrum handoff sequence is adopted, where ts = 1 (slot),

λs = 0.01 (arrivals/slot), E[Xs] = 10 (slots/arrival), and

E[Xp] = 20 (slots/arrival). . . 100

5.5 Comparison of the extended data delivery time for the always-staying and always-changing spectrum handoff sequences as well as the traffic-adaptive channel selection approach, where

ts = 1 (slot), λs= 0.01 (arrivals/slot), E[Xp] = 20 (slots/arrival),

and E[Xs] = 10 (slots/arrival). . . 101

5.6 Effects of secondary connections’ service time E[Xs] on the

cross-point for the traffic-adaptive channel selection approach,

where ts = 1 (slot), E[Xp] = 20 (slots/arrival), and λs = 0.01

(arrivals/slot). . . 102

5.7 Admissible region for the normalized traffic workloads (ρp, ρs),

where the average cumulative delay constraint can be satisfied

when ts = 0 (slot), E[Xp] = 20 (slots/arrival) and E[Xs] = 10

(slots/arrival). . . 104 5.8 Comparison of average extended data delivery time for

(20)

6.1 An example of state diagram of the target channel sequences for a newly arriving secondary user, where the default channel

η = 1, the number of total channels M = 3, and the required

length of the target channel sequence L = 4. Furthermore, (k, i) stands for the state of operating at the channel k with

the ith _{interruption. . . 113}

6.2 Six kinds of candidate sequences for the Cumulative Hand-off Delay Minimization Problem when the greedy shortest-handoff-delay target channel selection strategy is adopted. . . 118 6.3 Effects of the newly arriving secondary user’s average service

time E[χs] on the cumulative handoff delay for λ(k)p = 0.02 and

λ(k)s = 0.01 when 1 ≤ k ≤ 4. . . 127

6.4 Effect of the average service time E[Xs] and the arrival rate λs

of the secondary users’ connections on the cumulative handoff delay of the newly arriving secondary user’s connection for

(λ(1)p , λ(2)p , λ(3)p , λ(4)p ) = (0.019, 0.02, 0.02, 0.02) and E[Xp(k)] =

15 when 1 ≤ k ≤ 4. . . 130

6.5 Effect of the average service time E[Xp] and the arrival rate λp

of the primary users’ connections on the cumulative handoff delay of the newly arriving secondary user’s connection for

(21)

7.1 An example of transmission process for the secondary

connec-tion SCA, where T is the extended data delivery time of SCA

and Di is the handoff delay of the ith interruption. The gray

areas indicate that the channels are occupied by the existing primary connections (PCs) or secondary connections (SCs).

Because SCA is interrupted three times in total, the

over-all data connection is divided into four segments. Note that

D1 = δc, D2 = δs, and D3 = δc. . . 138

7.2 The PRP M/G/1 queueing network model with three

chan-nels where λ(k)p , λ(k)s , and ωn(k) are the arrival rates of the

λ(k)s = ω(k)₀ . Furthermore, fp(k)(x) and f_i(k)(φ) are the

proba-bility density functions (pdfs) of Xp(k) and Φ(k)i , respectively. . 140

7.3 Illustration of the physical meaning of random variable Φ(k)_i .

For example, Φ(1)₂ is one of the third segments (gray areas) of

the first and the third secondary connections in (a) as well as the second secondary connection in (b). Note that the third secondary connection in (b) does not have the third segment because it is interrupted only once. . . 144 7.4 State diagram of target channel sequence for a secondary

con-nection, where default channel η = 1. . . 150 7.5 Tree-structured representations of the proposed state diagram

where the grounding symbols represent the ending of state transition. Note that this figure considers the secondary con-nections whose default channels are Chk. . . 154

(22)

7.6 Effects of the arrival rate of the primary connections (λp) on

the channel utilizations at the channels 1 and 2, where δs = 1

and δc= 2. . . 156

7.7 Effects of the arrival rate of the primary connections (λp) on

the extended data delivery time of the secondary connections

whose default channels are channels 1 and 2, where δs = 1 and

δc= 2. . . 157

7.8 Effects of the initial arrival rate of the secondary connections

(λs) on the channel utilizations at the channels 1 and 2, where

δs= 1 and δc= 2. . . 159

7.9 Effects of the initial arrival rate of the secondary connections

(λs) on the extended data delivery time of the secondary

con-nections whose default channels are the channels 1 and 2,

where δs= 1 and δc = 2. . . 160

7.10 Comparison of average extended data delivery time for differ-ent target channel selection schemes. . . 162

7.11 Admissible region (λp, λs), where the average extended data

delivery time constraint can be satisfied when τ = 0. . . 165 7.12 Comparison of the average extended data delivery time for

different spectrum handoff schemes, where E[Xs] = 10, ts = 0,

and th = 0 . . . 167

8.1 Interference ratio (Θp) for various arrival rates of the

sec-ondary connections, where PM = 0.1. . . 182

8.2 Average overall system time (E[Ss]) for various arrival rates

of the secondary connections, where PF = 0.1. . . 183

8.3 Optimal traffic admission probability for various arrival rates

(23)

Glossary of Symbols

• type-i secondary connection: the secondary user’s connection that has

experienced i interruptions.

• λ(k)p : the arrival rate of the primary users’ connections whose default

channels are channel k.

• λs: the arrival rate of the secondary users’ connections.

• λ(k)s : the initial arrival rate of the secondary users’ connections whose

initial channels are channel k.

• ω_i(k): the arrival rate of the type-i secondary connections at channel k.

• Xp(k): the service time of the primary users’ connections whose default

• Xs: the service time of the secondary users’ connections.

• Xs(k): the service time of the secondary users’ connections whose default

• Φ(k)_i is the effective service time for the ith _{interruption at channel k.}

It is the transmission duration of a secondary connection between the

(24)

• eXs: the actual service time of the secondary users’ connections when

the effects of sensing errors are considered.

• fp(k)(x): probability density function of Xp(k).

• fs(x): probability density function of Xs.

• fs(k)(x): probability density function of Xs(k).

• f_i(k)(φ): probability density function of Φ(k)_i .

• PM: missed detection probability.

• PF: false alarm probability.

• πp: outage probability for the primary connections.

(25)

Chapter 1 Introduction

Recent measurements show that the licensed spectrum is under-utilized [1]. In order to solve this spectrum waste issue, many technologies have been pro-posed. Cognitive radio (CR) is one of the promising approaches to improve spectrum utilization [2–8]. A CR network consists of the primary and the secondary networks as shown in Fig. 1.1. The primary networks are defined as the systems that own the licensed spectrum, such as the cellular mobile networks or the TV broadcast networks. By contrast, the secondary net-works do not have any licensed frequency. By allowing the secondary users to temporarily access the primary user’s under-utilized licensed spectrum, CR can significantly improve spectrum efficiency and enhance the quality of service (QoS) performance of the secondary users.

One fundamental issue for enhancing QoS performance of the secondary users in CR networks is the spectrum handoff issue. When the high-priority primary user appears at its licensed channel being occupied by the low-priority secondary users, these secondary users must vacate the occupied channel. In order to vacate the occupied channel to the primary user and discover the suitable target channel to resume the unfinished transmission,

(26)

SU

Secondary

network

Primary

network

PU

Primary

Coordinator

Secondary

coordinator

Figure 1.1: An illustrative example of CR network, which consists a primary network and a secondary network. There are three primary users (PUs) and one secondary user (SU) in the primary and secondary networks, respectively.

(27)

Frequency

Power

Time

PU

SU

Spectrum Handoff

Transmission time of SU connection

Figure 1.2: During the transmission period of secondary user (SU), it expe-riences multiple handoffs.

the spectrum handoff procedures are initiated for the secondary users [9,

10]1_{. During the transmission period of a secondary connection, multiple}

interruptions from the primary users result in multiple spectrum handoffs as show in Fig. 1.2. These spectrum handoffs will degrade the QoS performance of the secondary users.

In order to overcome the performance degradation issue due to multiple

1_{Spectrum handoff in CR networks is different from the conventional handoff} mecha-nisms in cellular mobile networks. Spectrum handoff considers two types of users with different priorities, where the high-priority primary users have the right to interrupt the transmission of the low-priority secondary users. When the interruption event occurs, the secondary user must stop using the current channel even though the received signal strength is still acceptable. In contrast, all users in the conventional handoff mechanisms have the same priority to access channels and they change their operating channels mainly due to deterioration of signal quality [11].

(28)

spectrum handoffs for the secondary users, various spectrum management techniques in CR networks are re-examined from a link connection quality perspective. There are four spectrum management functionalities in CR networks [12]:

1. Spectrum sensing: The secondary users should monitor all channels in order to capture channel characteristic and detect spectrum holes. Based on sensing results, the secondary users can find some candi-date channels to transmit data. In this dissertation, we consider a fully-connected CR network. Hence, the transmitter and receiver of a secondary connection can have the same consensus on sensing results.

2. Spectrum decision: The secondary users can select the best channel from many candidate channels to transmit data. This decision should take the traffic statistics of the primary users as well as the secondary users into account.

3. Spectrum mobility: The secondary users must vacate the occupied channel when the primary user appears because the primary users have the preemptive priority to access channels. In order to return the occupied channel to the primary users and resume the unfinished transmission at the suitable channel, the spectrum handoff procedures are initiated for the interrupted secondary user.

4. Spectrum sharing: The secondary users must coordinate their trans-missions and avoid interfering with transmission of the primary users.

Referring to [13], the relationships of these four spectrum management functionalities are shown in Fig. 1.3. In the beginning, the traffic requests of secondary users arrive at the CR network. With the spectrum decision

(29)

Spectrum

Decision

Spectrum Sharing

Channel 1 Channel 2

Spectrum Sensing

Spectrum Mobility

Spectrum Handoff

Channel M

Arrivals of secondary

users

Figure 1.3: Relationships between spectrum sensing, spectrum decision, spec-trum mobility, and specspec-trum sharing functionalities.

functionality, they can determine their initial operating channels from all M candidate channels, which can be found by the spectrum sensing functional-ity. In order to alleviative the channel contention when multiple secondary users select the same channel and the interference on the primary users when missed detection occur, the spectrum sharing functionality must be imple-mented. Furthermore, if the primary user appears at the occupied channel, the spectrum handoff procedures in the spectrum mobility functionality must be initiated. Based on this dynamic spectrum management, spectrum effi-ciency can be improved. Note that the multiple handoff issue should be taken into account when designing these spectrum management functionalities.

(30)

In this dissertation, we focus on the spectrum decision, spectrum mobility, and spectrum sharing issues. In order to evaluate the system performance of the proposed spectrum management techniques, an analytical framework based on the preemptive resumption priority (PRP) M/G/1 queueing the-ory is developed to characterize the connection-based channel usage behav-iors with multiple handoffs. We investigate the effects of multiple handoffs on the QoS performance and study the performance limitation of various spectrum management techniques in different traffic loads. Different from the traditional work which investigated the effects of multiple handoffs on the network throughput, this dissertation concentrates on the effects of la-tency performance of the secondary users. Based on the proposed analytical framework, some useful insights into the design of the spectrum management techniques can be provided and the traffic-adaptive spectrum management schemes can be developed according to traffic conditions such as traffic arrival rates and service time distributions.

1.1 Problems and Solutions

In this section, we will briefly describe our problem formulations and the cor-responding solutions, including modeling technique for CR networks, traffic-adaptive spectrum mobility issues, load-balancing spectrum decision issues, and interference-avoiding spectrum sharing issues.

1.1.1 Modeling Techniques for Cognitive Radio

Net-works

In this part, we outline the fundamental modeling issues of opportunistic spectrum access in cognitive radio (CR) networks. In particular, we identify

(31)

the effects of the general behaviors for the connection-based channel usage on the quality of service (QoS) performances of spectrum management tech-niques. During the transmission period of a secondary user’s connection, the phenomenon of multiple spectrum handoffs resulted from the interruptions of the primary users arises quite often. In addition to multiple interruptions, the connection-based channel usage behaviors are also affected by other fac-tors, including spectrum sensing time, channel switching between different channels, generally distributed service time, and channel contention between multiple secondary users. An analytical framework based on the preemptive resumption priority M/G/1 queueing theory is introduced to characterize these effects simultaneously. The proposed analytical framework can pro-vide important insights into the design of spectrum management techniques in CR networks and can be adapted more flexibly for various traffic arrival rates and service time distributions.

1.1.2 Load-Balancing Spectrum Decision

In this part, we present an analytical framework to design system param-eters for load-balancing multiuser spectrum decision schemes in cognitive radio (CR) networks. Unlike the non-load-balancing methods that multi-ple secondary users may contend for the same channel, the considered load-balancing schemes can evenly distribute the traffic loads of secondary users to multiple channels. Based on the preemptive resume priority (PRP) M/G/1 queueing theory, a spectrum decision analytical model is proposed to evaluate the effects of multiple interruptions from the primary user during each link connection and the sensing errors of the secondary users. With the objective of minimizing the overall system time (i.e., waiting time plus data delivery time) of the secondary users, we derive the optimal number of candidate

(32)

channels and the optimal channel selection probability for the sensing-based and the probability-based spectrum decision schemes, respectively. We find that the probability-based scheme can yield a shorter overall system time compared to the sensing-based scheme when the traffic loads of the sec-ondary users is light, whereas the sensing-based scheme performs better in the condition of heavy traffic loads. If the secondary users can intelligently adopt the best spectrum decision scheme according to sensing time and traf-fic parameters, the overall system time can be improved by 50% compared to the existing methods. Furthermore, the proposed analytical model also takes into account of the probability of missed detection and false alarm for the appearance of the primary users, and can help evaluate the impacts of imperfect sensing on the spectrum decision schemes for CR networks.

1.1.3 Proactive Spectrum Handoff

In this part, we present an analytical framework to evaluate the latency performance of connection-based spectrum handoffs in cognitive radio (CR) networks. During the transmission period of a secondary connection, multi-ple interruptions from the primary users result in multimulti-ple spectrum handoffs and the need of predetermining a set of target channels for spectrum hand-offs. To quantify the effects of channel obsolete issue on the target channel predetermination, we should consider the three key design features: (1) gen-erally distributed service time of the primary and secondary connections; (2) different operating channels in multiple handoffs; and (3) queueing delay due to channel contention from multiple interrupted secondary connections. To this end, we propose the preemptive resume priority (PRP) M/G/1 queue-ing network model to characterize the spectrum usage behaviors with all the three design features. This model aims to analyze the extended data delivery

(33)

time of the secondary connections with proactively designed target channel sequences under various traffic arrival rates and service time distributions. These analytical results are applied to evaluate the latency performance of the connection-based spectrum handoff based on target channel sequences used in the IEEE 802.22 wireless regional area networks standard. Then, to reduce the extended data delivery time, a traffic-adaptive spectrum handoff is proposed, which changes the target channel sequence of spectrum handoffs based on traffic conditions. Compared to the existing target channel selection methods, this traffic-adaptive target channel selection approach can reduce the extended data transmission time by 35%, especially for the heavy traffic loads of the primary users.

1.1.4 Optimal Proactive Spectrum Handoff

In this part, we investigate how to determine an optimal target channel se-quence for multiple spectrum handoffs with the minimum cumulative handoff delay for the secondary users in cognitive radio networks. In addition to mul-tiple interruptions from the high-priority primary users, the optimal sequence for spectrum handoffs incorporates the effects of various traffic statistics of both the primary and the secondary users. Compared to the exhaustive

search with time complexity of O(ML_{), where L is the total number of}

ele-ments in the target channel sequence and M is the total number of candidate channels for spectrum handoffs, a dynamic programming algorithm with the

complexity of O(LM2_{) is proposed to determine the optimal target channel}

sequence for spectrum handoffs. Furthermore, we propose a greedy algorithm with time complexity of O(M) for spectrum handoffs and prove that it only requires to compare six permutations of the target channel sequences. Nu-merical results show that the cumulative handoff delay of the low-complexity

(34)

greedy algorithm can approach that of the optimal solution.

1.1.5 Reactive Spectrum Handoff

Spectrum handoff is an important functionality in cognitive radio (CR) net-works. Whenever a primary user appears, transmission of the secondary users must be interrupted. In this case, spectrum handoff procedures are initiated for the secondary users in order to search the idle channel to resume the un-finished transmission. Although this dynamic spectrum access scheme can enhance channel utilization, multiple interruptions from the primary users will result in multiple handoffs and thereby increase the transmission latency of the secondary users. Hence, two fundamental issues in CR networks are how much channel utilization can be improved and how long transmission latency is extended for the secondary users due to multiple spectrum han-odffs. To solve the first problem, we introduce the preemptive resume priority (PRP) M/G/1 queueing network to characterize the channel usage behaviors of CR networks. Based on this queueing network, channel utilization under various traffic arrival rates and service time distributions can be evaluated. Furthermore, on top of the proposed queueing network, a state diagram is developed to characterize the effects of multiple handoff delay on the trans-mission latency of the secondary users. The analytical results can provide a helpful insight to study the effects of traffic arrival rates and service time on the transmission latency and then facilitate the designs of admission control rules for the secondary users subject to their latency requirements.

1.1.6 Interference-Avoiding Spectrum Sharing

In this part, we present an analytical framework to design key system param-eters for an interference-avoiding admission control mechanism to enhance

(35)

channel utilization, while maintaining the quality of service (QoS) require-ments for both the primary users and secondary users in cognitive radio (CR) networks. Intuitively, a larger admission probability for the secondary users can increase channel utilization, but it leads to more contention be-tween the secondary users and thus affects the latency performance of the secondary users. More importantly, if the missed detection for the presence of the primary user happen, the larger the admission probability of the sec-ondary user, the more the interference to the primary user. In order to find the optimal traffic admission probability, a cross-layer optimization problem is formulated. Our cross-layer design can incorporate the following effects: (1) false alarm and missed detection, power outage in the physical layer; (2) admission probability in the medium access control (MAC) layer; and (3) the traffic statistics as well as the QoS constraints of both the primary and the secondary users in the application layer. The analytical results proposed in this part can calculate the optimal traffic admission probability under various cross-layer parameters and provide useful insights into the tradeoff design of channel utilization and the QoS performance for both the primary and the secondary users.

1.2 Dissertation Outlines

This dissertation consists of four themes as shown in Fig. 1.4. The first part is to outlines the fundamental modeling issues of various spectrum manage-ment techniques in CR networks. Then, an analytical framework based on the preemptive resumption priority M/G/1 queueing theory is introduced to characterize these modeling issues simultaneously. In order to demonstrate the effectiveness of this model, three illustrative examples are presented in

(36)

Spectrum Mobility

(Part III) Spectrum_Decision (Part II) Spectrum Sharing (Part IV) Proactive Spectrum Handoff Optimal Proactive Spectrum Handoff Reactive Spectrum Handoff Load-Balancing Spectrum Decision Interference-Avoiding Spectrum Sharing Modeling Technique for

Cognitive Radio Networks

PRP M/G/1 Queueing Network Model (Part I)

Figure 1.4: Outline of this dissertation.

the second. third, and fourth parts as follows. The second part investi-gates the spectrum decision issue. We determine which channels to probe and transmit in a load-balancing manner. The third part focuses on the spectrum mobility issue. We illustrate how to model the effects of multiple handoffs, where the secondary users can have different operating channels before and after spectrum handoff. The final part considers the spectrum sharing issue. The optimal admission probability for the secondary users is determined to satisfy the interference constraint to the primary users.

The remaining chapters of this dissertation are organized as follows. In Chapter 2, we first give a literature survey of the state-of-the-art techniques. Chapter 3 provides an analytical framework to characterize the general chan-nel usage behaviors with multiple handoffs from a macroscopic viewpoint. Based on the proposed analytical framework, Chapter 4 designs system pa-rameters for load-balancing multiuser spectrum decision schemes in CR net-works. Furthermore, Chapters 5 and 6 evaluate the latency performance and determine the optimal target channel sequence for the proactive spectrum handoff, respectively. Next, the effect of sensing time on the latency

(37)

per-formance of the reactive spectrum handoff is investigated in Chapter 7. In Chapter 8, an admission control mechanism for the secondary users’ spectrum sharing is discussed. Finally, the concluding remarks and some suggestions for future research topics are given in Chapter 9.

(38)

Chapter 2 Background and Literature

Survey

In this chapter, we firstly survey related work to the modeling techniques of the connection-based channel usage behaviors with multiple handoffs. Then, the existing spectrum management techniques, which consist of spectrum decision, spectrum mobility, and spectrum sharing, also are discussed.

2.1 Modeling Techniques for Cognitive Radio

Networks

Most of the modeling techniques of channel usage behaviors in CR networks can be classified into three categories: the partially observable Markov de-cision process (POMDP), the two-dimensional Markov chain (TDMC), and the PRP M/G/1 queueing model (QM). However, these models have not simultaneously considered all of the five design features. Table 2.1 classifies the existing modeling techniques, where the signs “ ◦ ” and “ × ” indicate that the proposed model “does” and “does not” consider the corresponding

(39)

Table 2.1: Comparison of Various Analytical Models for CR Networks.

Multiple Spectrum Various General Multiple Model Name Spectrum Sensing Operating Service Secondary

Handoff Time Channels Time Connections

POMDP [14] ◦ × ◦ × × TDMC [15] ◦ × ◦ × ◦ QM [16–18] ◦ × × ◦ ◦ Proposed ◦ ◦ ◦ ◦ ◦ Model feature, respectively.

In [14], the evolutions of the channel usage of the primary network is char-acterized by a discrete-time Markov chain which has two occupancy states (the busy and the idle states). The framework of partially observable Markov decision process (POMDP) was developed to preselect the best action (target channel) to maximize the immediate reward (expected per-slot throughput) of the decision maker (secondary user) at the next time slot [14]. Unlike [14] considered only the effects of the traffic loads of the primary network, the authors in [15] considered the effects of the traffic loads of both the primary and the secondary users on the statistics of channel occupancy. In [15], the channel usage behaviors of a CR network is modeled by a two-dimensional Markov chain where the two dimensions represent the total numbers of the primary and the secondary users in a CR network, respectively. When the secondary users are interrupted, it is assumed that they can immediately find the idle channel if at least one idle channel exists. Hence, the spectrum sens-ing time is neglected in this model. Note that the two Markov chain models are suitable for the exponentially distributed service time, and how to extend

(40)

them to the case with generally distributed service time is not clear.

Some researchers used the PRP M/G/1 queueing model to character-ize the spectrum usage behaviors of each channel. For example, the effects of multi-user contention and multiple interruptions on the latency perfor-mance of the secondary users’ connections were studied in [16–18]. This PRP M/G/1 queueing model assumed that the secondary user must stay on its current operating channel to resume its unfinished transmission when it is interrupted. That is, there is one candidate channel for spectrum handofff and thus the sensing time issue has not been addressed.

2.2 Load-Balancing Spectrum Decision

The load-balancing spectrum decision schemes can be categorized into two methods: the sensing-based spectrum decision and the probability-based spectrum decision. Table 2.2 compares the existing load-balancing spectrum decision schemes. In the following, we discuss the features of these spectrum decision methods in more details.

2.2.1 Probability-based Spectrum Decision

In the literature, many probability-based spectrum decision schemes were proposed to balance the traffic loads of secondary users in multi-channel CR networks, which can be categorized into three types: (1) packet-wise probabilistic (PP) approach [19–23]; (2) game-theoretic (GT) approach [24– 26]; and (3) learning automata (LA) approach [27].

• Packet-wise probabilistic spectrum decision approaches [19–23] aim at

maximizing the expected throughput of the secondary users at each slot by determining the probability of selecting each channel from the

(41)

Table 2.2: Comparison of Various Load-balancing Spectrum Decision Schemes for Cognitive Radio Networks, where PP, GT, and LA stand for the packet-wise probabilistic, game-theoretic, and learning automata approaches, respectively.

Channel Occupancy Model Multiple Sensing

of a Primary Network Interruptions Errors

PP Bernoulli Process [19] × ×

Probability- Bernoulli Process [20–23] × ◦

based

GT Deterministic Process [24] × ×

Methods M/M/1 [25] or M/G/1 [26] ◦ ×

LA General Distribution [27] ◦ ×

Sensing- Deterministic Process [28] × ×

based Two-state Markov Chain [29] × ×

Methods Bernoulli Process [30–32] × ×

(42)

pool of candidate channel. Based on busy probability and capacity of each channel, [19] suggested a method to determine the probabil-ity for selecting channels on top of p-persistent carrier sense multiple access (CSMA) medium access control (MAC) protocol in a decentral-ized manner. They claimed that their proposed sub-optimal channel probability assignment can achieve the Nash equilibrium as the number of secondary users tends to infinity. Furthermore, [20, 21] considered the effects of sensing errors in terms of false alarm and missed detec-tion on the throughput of the secondary users in a two-channel system, and proposed a probabilistic channel selection approach to maximize the throughput of the secondary users in each slot while maintaining the latency constraint of the primary users. Moreover, [22] formulated an optimization problem for channel selection probability to maximize the throughput of the secondary users in each slot while maintaining the interference constraint of the primary users when the primary and secondary networks are asynchronous. Unlike [20–22] considered only the case that one single secondary user can select the channel at each time instant, [23] further extended the probabilistic channel selection approach of [20, 21] to the case that multiple users can simultaneously select their operating channels from the pool of candidate channels, and analyzed the throughput of the secondary users based on the proba-bilistic channel selection approach taking into account of the effects of channel contention as well as sensing errors. Note that the packet-wise probabilistic spectrum decision approaches in [19–23] were executed in a slot-by-slot manner, which may lead to many channel-switching behaviors during each secondary user’s link connection. Moreover, it is assumed that the traffic loads of the secondary users are saturated.

(43)

Further, the channel occupancy model of a primary network is modeled as a Bernoulli process and thus the length of busy and idle periods are exponentially distributed.

• Game-theoretic approaches were proposed to solve the spectrum

deci-sion problem in CR networks [24–26]. Based on the game theory model, each player (secondary user) can decide the best strategy (channel se-lection probability) to maximize its utility function. [24] proposed a game-theoretic load-balancing approach to find a set of channel selec-tion probabilities so that no secondary user has incentive to unilaterally change his/her action. To converge to such the Nash equilibrium, a best-reply algorithm was designed for each user to calculate each chan-nel’s selection probability as well as its transmission duration based on a utility function related to the load-balancing channel selection. Be-side the load-balancing issue, [25] suggested that the utility function in the game-theoretic spectrum decision should also incorporate the channel bandwidth and its idle period as well as the cost of spectrum handoff because the spectrum decision procedure must be executed many times due to multiple interruptions. They emphasized that the channel selection game shall be repeated many times to capture the scenario when primary users stochastically activate or deactivate at each epoch. Unlike the pervious work that considered the homoge-neous secondary users, [26] assumed that the secondary users can have different priorities. They proposed a dynamic strategy learning algo-rithm to determine the channel selection strategies that can converge to the Nash equilibrium. Noteworthily, the Nash equilibrium solution of the game-theoretic approach is not necessary the globally optimal solution from the viewpoint of the overall network [33].

(44)

• In [27], a learning automata (LA) approach was suggested to determine

the channel selection probabilities by exploring the uncertainty of traffic patterns in CR networks. After a huge number of trials, the secondary users can estimate the optimal channel selection probability. However, the problem for this method is its converging speed, especially for a large number of users.

2.2.2 Sensing-based Spectrum Decision

The sensing-based spectrum decision scheme requires scanning all the candi-date channels to determine the most suitable operating channel. Thus, the total number of candidate channels significantly affects the overall system time in the sensing-based spectrum decision scheme. In [28–30], the opti-mal number of candidate channels to maximize the spectrum accessibility and the procedures to determine the optimal set of candidate channels were investigated. Furthermore, the authors in [31, 32] formulated the sequential channel sensing problem as an optimal stopping problem with the objective of maximizing the throughput of the secondary users. They studied when the secondary users shall stop sensing and start transmitting data. Never-theless, the effects of multiple interruptions from the primary user and the sensing errors for the primary user’s occurrence on the overall system time of the secondary users in the CR networks have not been addressed in these existing sensing-based spectrum decision methods.

2.3 Proactive Spectrum Handoff

In order to characterize the multiple handoff behaviors in CR networks, we should consider the three key design features, consisting of (1) generally

(45)

Table 2.3: Comparison of Various Proactive Handoff Models.

General Various Multiple

Model Name Service Operating Secondary

Time Channels Connections

TMC [29, 34–39] × ◦ × OORP [40–42] ◦ ◦ × BRP [43] × ◦ × MMC [44] × × ◦ QM [16–18, 26, 45–49] ◦ × ◦ Proposed Model ◦ ◦ ◦

distributed service time; (2) various operating channels; and (3) queueing delay due to channel contention from multiple secondary connections. Based on these three features, Table 2.3 classifies the existing modeling techniques for the proactive spectrum handoff. In the literature, the modeling techniques for spectrum handoff behaviors can be categorized into the following five types: (1) the two-state Markov chain; (2) the Bernoulli random process; (3) the arbitrary ON/OFF random process; (4) the birth-death process with multi-dimensional Markov chain; and (5) the PRP M/G/1 queueing model. One can observe that the current modeling techniques have not considered all the aforementioned three design features. In the following, we briefly discuss the features of these analytical models for spectrum handoff behaviors.

• Two-state Markov chain (TMC): In [29, 34–39], the evolutions of

the channel usage of the primary networks at each channel were char-acterized by a discrete-time Markov chain which has two occupancy states: busy (ON) and idle (OFF) states. The idle (OFF) state can be regarded as a potential spectrum opportunity for the secondary users.

(46)

Note that the Markov chain model is suitable for the exponentially dis-tributed service time, and it is not clear how to extend it to the case with generally distributed service time. In this model, the target chan-nel selection problem in every time slot is modeled as a Markov decision process. According to the channel occupancy state at the current time slot, a decision maker (secondary user) can preselect the best action (target channel) to maximize its immediate reward at the next time slot such as expected per-slot throughput [29,34–37], expected idle pe-riod [38], or expected waiting time [39]. Note that this model belongs to the slot-based modeling technique because the secondary user shall decide its target channel at each time slot. In this scheme, even though the primary users do not appear at the current operating channel, the secondary user may still need to change its target channel, resulting in frequent spectrum handoffs.

• Arbitrary ON/OFF random process (OORP): Unlike [29,34–39]

assumed that the channel usage behaviors of the primary networks have the Markov property, the authors in [40–42] used the ON/OFF random process with arbitrary distributed ON/OFF period to characterize the channel usage behaviors of the primary networks at each channel. It was assumed that the secondary user can estimate the distributions of the ON period and the OFF period based on long-term observations. In each time slot, the secondary user must calculate the expected re-maining idle periods of all channels and then will immediately switch to the channel with the longest remaining idle period. This model also belongs to the slot-based modeling technique because the target channel is decided in each time slot.

(47)

• Bernoulli random process (BRP): The authors in [43] examined

the effects of multiple interruptions from the primary users on the con-nection maintenance probability in a concon-nection-based environment, where the spectrum usage behaviors of the primary networks on each channel were characterized by a Bernoulli random process. Because both the busy and idle periods of the considered primary networks follow the geometrical distributions, it is more difficult to extend this modeling technique to the cases with other generally distributed service time.

• Multi-dimensional Markov chain (MMC): In [44], the spectrum

usage behaviors of both the primary and secondary networks were mod-eled by the multi-dimensional Markov chain. Each state in the Markov chain indicates the identity number for the serving users and the wait-ing users for the channel. It was assumed that the secondary user must stay on its current operating channel after the primary user’s interrup-tion. This analytical model is suitable for the single channel network, and the issue of different operating channels in multiple handoffs has not been addressed.

• M/G/1 queueing model (QM): Some researchers used the

preemp-tive resume priority (PRP) M/G/1 queueing model to characterize the spectrum usage behaviors in a single-channel CR network. The effects of multi-user sharing and multiple interruptions on the extended data delivery time of the secondary users were studied in [16–18, 26, 45–49]. Note that the authors in [16–18, 26, 45–49] also assumed that the sec-ondary users must stay on the current operating channel to resume their unfinished transmissions when they are interrupted.

(48)

To summarize, the first three analytical models, two-state Markov chain, arbitrary ON/OFF random process, and bernoulli random process, did not incorporate the effects of the traffic loads of the secondary users on the statis-tics of channel occupancy. How to extend these models to consider the queue-ing delay due to channel contention from multiple secondary connections is unclear. The last two models, multi-dimensional Markov chain and M/G/1 queueing model, can characterize the effects of spectrum sharing between multiple secondary users. However, these two models assumed that the in-terrupted secondary user must stay on the current operating channel. and have not dealt with the handoff interaction issue among different channels.

2.4 Optimal Proactive Spectrum Handoff

In the literature, some predetermined target channel selection methods for spectrum handoffs have been proposed and can be categorized into two kinds: probability-based channel selection methods and Markov decision process.

• In [20,21,23], the probability-based channel selection methods were

de-veloped to predetermine the probability that each channel is selected to the target channel. Based on the predetermine probabilities, the optimal channel hopping sequence can be decided in packet-by-packet or slot-by-slot manners. The work in [20, 21] designed of the optimal channel hopping sequence in the single-user case, while [23] extended the similar problem to the multiple user case. The above approaches for channel hopping sequence design are optimal in the sense of maxi-mizing the per-slot throughput. However, the latency issue in spectrum handoff has not been considered yet. Clearly, the cumulative delay in one connection due to multiple spectrum handoffs is an important QoS

(49)

performance measure for CR networks.

• Besides the probability-based channel selection methods, another kind

of target channel selection approach is to apply the theory of Markov decision process. In [29, 34–39], the target channel selection problem in every time slot is modeled as a Markov decision process. According to the channel occupancy state at the current time slot, a decision maker (secondary user) can preselect the best action (target channel) to maximize its immediate reward at the next time slot. The considered reward or objective function includes the expected per-slot throughput [29, 34–37], expected idle period [38], and expected waiting time [39]. However, only the effects of channel usage behaviors of the primary users are considered on the channel occupancy. In fact, the traffic loads of the secondary users are also needed to be considered in channel selection.

2.5 Reactive Spectrum Handoff

A key property of reactive spectrum handoff is that the interrupted secondary user can actually find the idle if at least one idle channel exists at the moment of link transition. In order to characterize the channel usage behaviors with this property, we should consider the three key design features, consisting of (1) heterogeneous arrival rates of the primary users (PUs); (2) various arrival rates of the secondary users (SUs); (3) handoff processing time. Based on these three features, Table 2.4 classifies the existing modeling techniques for the reactive spectrum handoff. In the literature, the modeling techniques for spectrum handoff behaviors can be categorized into the following four types: (1) ON/OFF random process; (2) M/M/m queueing Model; (3)

(50)

two-Table 2.4: Comparison of Various Channel Usage Models.

Heterogeneous Various Handoff

Model Name Arrival Rates Arrival Rates Processing

of PUs of SUs Time

OORP [50, 51] × × ×

OORP [43] × × ◦

M/M/m [52] × × ×

MDMC [15, 53–66] × × ×

M/G/1 [67, 68] ◦ ◦ ×

Proposed Unifying Model ◦ ◦ ◦

dimensional Markov chain; and (4) M/G/1 queueing model. One can observe that the current modeling techniques have not considered all the aforemen-tioned three design features. In the following, we briefly discuss the features of these analytical models for spectrum handoff behaviors.

• ON/OFF random process (OORP): In [43, 50, 51], the ON/OFF

random process was used to characterize the channel usage behaviors of the primary networks at each channel, where the distributions of ON (busy) period- and OFF (idle) period at each channel are geometrical distributed. The OFF state can be regarded as a potential spectrum opportunity for the secondary users. The authors in [50] and [51] inves-tigated the channel utilization factors and the extended data delivery time of the secondary users, respectively. Unlike [51] that did not ad-dress the effects of spectrum sensing time, the authors in [43] examined the effects of spectrum sensing time on the extended data delivery time of the secondary users. However, [43] assumed that at least one chan-nel is certainly available after spectrum sensing, and the case that all

(51)

channels are busy after spectrum sensing did not been considered.

• M/M/m queueing model: In [52], the channel usage behaviors of

the primary users are characterized by the M/M/m queueing model, where m is the total number of channels in the CR network. The author in [52] calculated the handoff delay of the secondary users. However, it is assumed that the handoff delay only results from the waiting time which is the duration from the instant that interruption event occurs until the instant that one idle channel is found. The sensing time had not been considered when calculating handoff delay.

• Multiple-dimensional Markov chain (MDMC): In [15,53–63], the

spectrum usage behaviors of both the primary and secondary networks were modeled by a two-dimensional Markov chain, where the two di-mensions represent the total numbers of the primary and the secondary users in a CR network, respectively. The blocking probability and forced termination probability for the secondary users’ connections in the CR network without and with queue are studied in [15, 53–58] and [59–61], respectively. Different from [15, 53–61] that considered infinite user population, [62, 63] derived the blocking probability in a CR network with finite user population. Furthermore, the authors in [64–66] further extended the two-dimensional Markov chain model to the multiple-dimensional Markov chain, where the new dimension is used to describe the channel state or queue length. Note that these analytical models are suitable for the CR network with homogeneous traffic loads, and the issues of heterogeneous arrival rates of the primary and the secondary users has not been addressed.

(52)

characterize the channel usage behaviors of a secondary network, where each secondary user can simultaneously use all idle channels to transmit its data. Because the total number of idle channels depends on how many channels are occupied by the primary users, the service rates of the secondary users are related to the traffic statistics of the primary users, which results in a non-trivially distributed service time. Thus, the authors suggested using the M/G/1 queueing system to characterize this system. However, authors did not show how to obtain this non-trivial service time distribution.

2.6 Interference-Avoiding Spectrum Sharing

In order to determine the optimal admission probability, we should con-sider four key design features: (1) interference on the primary users (PUs), where the transmission of the primary users may be stained by the secondary users due to missed detection; (2) channel contention between multiple sec-ondary users (SUs), where channel contention will increase waiting time of the secondary users; (3) multiple handoffs, a secondary user may have mul-tiple handoffs due to mulmul-tiple interruptions from the primary users during its transmission period; and (4) generally distributed service time, where the probability mass functions (pmfs) of service time of the primary and sec-ondary connections can be any distributions. Based on these four design features, Table 2.5 classifies the existing admission control techniques.

2.6.1 Admission Control with Perfect Sensing

• Network-throughput-oriented approach: The authors in [6,19,44]

以排隊理論為基礎對感知無線網路頻譜管理技術之研究

國 立 交 通 大 學

電信工程研究所

博 士 論 文

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

研 究 生： 王 中 瑋

指導教授： 王 蒞 君

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

研究生：王中瑋

Student:

Chung-Wei

Wang

指導教授：王蒞君 博士 Advisor:

Dr.

Li-Chun

Wang

國立交通大學

電信工程研究所

博士論文

A Dissertation

Submitted to Institute of Communication Engineering

College of Electrical and Computer Engineering

National Chiao Tung University

in Partial Fulfillment of the Requirements

for the Degree of Doctor of Philosophy

in

Communication Engineering

Hsinchu, Taiwan

以排隊理論為基礎對

感知無線網路頻譜管理技術之研究

研究生：王中瑋 指導教授：王蒞君 博士

國立交通大學

電信工程研究所

摘要

Queueing-Theoretical Spectrum Management

Techniques for Cognitive Radio Networks

Department of Electrical Engineering

National Chiao Tung University

Abstract

Acknowledgements

Contents

List of Tables

List of Figures

Glossary of Symbols

Chapter 1

Introduction

SU

Secondary

network

Primary

network

PU

PU

PU

Primary

Coordinator

Secondary

coordinator

Frequency

Power

Time

PU

PU

PU

PU

PU

PU

SU

Spectrum Handoff

Spectrum

Decision

Spectrum Sharing

Spectrum Sensing

Spectrum Mobility

國立交通大學

博士論文

研究生：王中瑋

指導教授：王蒞君

指導教授：王蒞君博士 Advisor:

研究生：王中瑋指導教授：王蒞君博士