Performance assessment of cooperative relay networks with advanced radio transmission techniques

In the first part, performance analysis of cooperativesingle and multiple relay networks using MIMO and OSTBC trans-mission is presented wherein the diversity gain, coding gain, outagepr

Abstract

In the past decade, cooperative communications has been emerging

as a pertinent technology for the current and upcoming generations ofmobile communication infrastructure The indispensable benefits of thistechnology have motivated numerous studies from both academia andindustry on this area In particular, cooperative communications hasbeen developed as a means of alleviating the effect of fading and henceimprove the reliability of wireless communications The key idea behindthis technique is that communication between the source and destina-tion can be assisted by several intermediate nodes, so-called relay nodes

As a result, cooperative communication networks can enhance the liability of wireless communications where the transmitted signals areseverely impaired because of fading In addition, through relaying trans-mission, communication range can be extended and transmit power ofeach radio terminal can be reduced as well

re-The objective of this thesis is to analyze the system performance

of cooperative relay networks integrating advanced radio transmissiontechniques and using the two major relaying protocols, i.e., decode-and-forward (DF) and amplify-and-forward (AF) In particular, the ra-dio transmission techniques that are considered in this thesis includemultiple-input multiple-output (MIMO) systems and orthogonal space-time block coding (OSTBC) transmission, adaptive transmission, beam-forming transmission, coded cooperation, and cognitive radio transmis-sion

The thesis is divided into an introduction section and six parts based

on peer-reviewed journal articles and conference papers The tion provides the readers with some fundamental background on co-operative communications along with several key concepts of cognitiveradio systems In the first part, performance analysis of cooperativesingle and multiple relay networks using MIMO and OSTBC trans-mission is presented wherein the diversity gain, coding gain, outageprobability, symbol error rate, and channel capacity are assessed It

introduc-is shown that integrating MIMO and OSTBC transmintroduc-ission into rative relay networks provides full diversity gain In the second part,the performance benefits of MIMO relay networks with OSTBC andadaptive transmission strategies are investigated In the third part, theperformance improvement with respect to outage probability of codedcooperation applied to opportunistic DF relay networks over conven-tional cooperative networks is shown In the fourth part, the effects ofdelay of channel state information feedback from the destination to thesource and co-channel interference on system performance is analyzedfor beamforming AF relay networks In the fifth part, cooperative di-versity is investigated in the context of an underlay cognitive AF relaynetwork with beamforming In the sixth part, finally, the impact of theinterference power constraint on the system performance of multi-hop

cognitive AF relay networks is investigated

Preface

This thesis summarizes my research within the fields of cooperative nications and cognitive radio networks The work has been carried out at theSchool of Engineering and School of Computing, Blekinge Institute of Tech-nology, Karlskrona, Sweden The thesis comprises an introduction sectionfollowed by six publication parts, as follows

commu-Part I

MIMO Cooperative Relay Networks with OSTBCs

A Performance Analysis of Decouple-and-Forward MIMO Relaying

Many special thanks to my previous co-advisor Professor Mats Petterssonand my co-advisor Dr Patrik Arlos for their valuable support and advices.Also, many thanks go to the Vietnam International Education Development(VIED) for funding this research.

Of many of my colleagues and friends, I like to thank Duong Quang Trung,Lei Shu, Maged Elkashlan, Tran Hung, Chu Thi My Chinh, Hoang Le Nam,Erik ¨Ostlin, and Ngo Quoc Hien for their cooperation I am also grateful to

my friends, Muhammad Imran Iqbal, Charles Kabiri, Louis Sibomana, UlrichEngelke, Thomas Sj¨ogren, and Vu Viet Thuy, for their friendship Manythanks also go to all my colleagues and friends at the Blekinge Institute ofTechnology for their help and support

Many profound thanks go to Staffan Andersson and David Erman, forteaching me Win Tsun Freely attending their class, I had the opportunity

to enjoy excellent time out of study with wonderful practice, knowledge, andphilosophy of martial art Special thanks go to my father, my mother, mybrothers and my sisters for their support, love, and encouragement, whichhave meant to me much more than what I can ever express

Phan HocKarlskrona, December 2012

Publication List

Part I is published as:

H Phan, T Q Duong, and H.-J Zepernick, “Performance analysis of and-forward MIMO relaying in Nakagami-m fading,” IEICE Transactions onCommunications, vol E95-B, no 09, pp 3003–3006, Sep 2012

decouple-H Phan, T Q Duong, and decouple-H.-J Zepernick, “MIMO cooperative relay networks with OSTBCs over Nakagami-m fading,” in Proc IEEE Wire-less Communications and Networking Conference, Paris, France, Apr 2012

multiple-Part II is published as:

H Phan, T Q Duong, H.-J Zepernick, and L Shu, “Adaptive sion in MIMO AF relay networks with orthogonal space-time block codesover Nakagami-m fading,” EURASIP Journal on Wireless Communicationsand Networking, vol 2012, pp 1-13, Jan 2012, DOI:10.1186/1687-1499-2012-11

transmis-Part III is published as:

H Phan, T Q Duong, and H.-J Zepernick, “Outage performance for tunistic decode-and-forward relaying coded cooperation networks over Naka-gami−m fading,” in Proc International Symposium on Wireless Communi-cation Systems, Aachen, Germany, Nov 2011

oppor-Part IV is published as:

H Phan, T Q Duong, M Elkashlan, and H.-J Zepernick, “Beamformingamplify-and-forward relay networks with feedback delay and interference,”IEEE Signal Process Lett., vol 19, no 1, pp 16–19, Jan 2012

Part V is published as:

H Phan, H.-J Zepernick, T Q Duong, H Tran, and T M C Chu, tive AF relay networks with beamforming under primary user power constraintover Nakagami-m fading channels,” Wireless Communications and MobileComputing, Nov 2012, DOI: 10.1002/wcm.2317

“Cogni-Part VI is published as:

H Phan, H.-J Zepernick, and H Tran, “Impact of interference power constraint

on multi-hop cognitive AF relay networks over Nakagami-m fading,” IETCommunications, 2013, accepted for publication with minor revision

Publications in conjunction with this thesis:

T M C Chu, H Phan, and H.-J Zepernick, “Opportunistic spectrum accessfor cognitive amplify-and-forward relay networks,” in Proc IEEE VehicularTechnology Conference, Dresden, Germany, Jun 2013, accepted for publica-tion

T M C Chu, H Phan, and H.-J Zepernick, “On the performance of derlay cognitive radio networks using M/G/1/K queueing model,” in IEEECommun Lett., Jan 2013, accepted for publication

un-H Phan, T M C Chu, un-H.-J Zepernick, and P Arlos, “Queueing sis of opportunistic decode-and-forward relay networks,” in Proc Internatio-nal Conference on Computing, Management and Telecommunications, Ho ChiMinh City, Vietnam, Jan 2013

analy-T M C Chu, H Phan, analy-T Q Duong, M Elkashlan, and H.-J Zepernick,

“Beamforming transmission in cognitive AF relay networks with feedbackdelay,” in Proc International Conference on Computing, Management andTelecommunications, Ho Chi Minh City, Vietnam, Jan 2013

T M C Chu, H Phan, and H.-J Zepernick, “Amplify-and-forward relayassisting both primary and secondary transmissions in cognitive radio net-works over Nakagami-m fading,” in Proc IEEE International Symposium onPersonal, Indoor and Mobile Radio Communications, Sydney, Australia, Sep.2012

H Tran, H.-J Zepernick, M Fiedler, and H Phan, “Outage probability, rage transmission time and quality of experience for cognitive radio networksover general fading channels,” in Euro-NF Conference on Next GenerationInternet, Karlskrona, Sweden, Jun 2012

ave-H Tran, ave-H.-J Zepernick, and ave-H Phan, “Impact of the number of antennasand distances among users on cognitive radio networks,” in Proc AdvancedTechnologies for Communications, Ha Noi, Vietnam, Oct 2012

H Phan, T Q Duong, and H.-J Zepernick, ”MIMO AF semi-blind relaynetworks with OSTBC transmission over Nakagami-m fading,” in Proc In-ternational Conference on Signal Processing and Communication Systems,Honolulu, Hawaii, USA, Dec 2011

H Phan, T Q Duong, and H.-J Zepernick, ”SER of amplify-and-forwardcooperative networks with OSTBC transmission in Nakagami-m fading,” inProc IEEE Vehicular Technology Conference, San Francisco, USA, Sep.2011

H Phan, T Q Duong, and H.-J Zepernick, “Full-rate distributed space-timecoding for bi-directional cooperative communications,” in Proc InternationalSymposium on Wireless Pervasive Computing, Modena, Italy, May 2010

Contents

Abstract v

Preface vii

Acknowledgements ix

Publication list xi

Contents xv

Introduction 1

Part I MIMO Cooperative Relay Networks with OSTBCs A Performance Analysis of Decouple-and-Forward MIMO Relaying in Nakagami-m Fading 43

B MIMO Cooperative Multiple-Relay Networks with OSTBCs over Nakagami-m Fading 59

Part II Adaptive Transmission in MIMO AF Relay Networks with Orthogonal Space-time Block Codes over Nakagami-m Fading 79

Part III Outage Performance for Opportunistic Decode-and-Forward Relaying Coded Cooperation Networks over Nakagami-m Fading 111

Part IV Beamforming Amplify-and-Forward Relay Networks with Feedback Delay and Interference 129

Part V Cognitive AF Relay Networks with Beamforming under Primary User Power Constraint over Nakagami-m Fading Channels 145

Part VI Impact of Interference Power Constraint on Multi-hop Cognitive AF Relay Networks over Nakagami-m Fading 177

The past decades have seen a tremendous growth of both wireless nications technology as well as research activities in this area This trendhas led to a significant change in many aspects of our daily life Especially,cellular mobile communications has experienced significant progress and hasnow become an essential part of human life all over the world with billions

commu-of users The era commu-of wireless communication revolution has been supported

by the large advances in very large-scale integration (VLSI) circuit logy, which enables to produce wireless terminals with small size, low powerconsumption, and complex processing In addition, the successful implemen-tations of the third generation (3G) and the forth generation (4G) cellularmobile networks, support considerably faster communications than their pre-decessors These new networks provide a wide range of high speed servicessuch as video streaming, games, and Internet access However, providing bothhigh speed and reliable communications over wireless channels subject to limi-ted radio spectrum is still a challenging problem There are two main aspects

techno-of wireless transmission which make it complicated and unreliable, that arefading and interference Firstly, the small-scale effect of multi-path fadingand the large-scale effects of path-loss and shadowing cause fluctuation andattenuation of the channel powers which seriously impair the wireless signals.Secondly, in contrast to wireline communication where the transmission chan-nel can be considered free of interference, wireless communications is subject

to considerable interference, i.e, caused by other neighboring terminals Thesetwo aspects of wireless communications cause unreliable signal transmission,leading to a serious performance degradation [1–3] As such, to assure a satis-factory level of quality-of-service (QoS), the design of wireless communicationnetworks should alleviate the impairments caused by fading and interference.Diversity systems have been developed as an efficient way to mitigate theimpairments caused by fading to wireless signals The key idea behind thistechnique is that it enables the destination to receive and process multipleindependently faded replicas of the source signal by using an appropriate

1

algorithm The probability that these multiple independent signal paths multaneously experience deep fades is much lower than that a single signalpath does As a result, the signal reliability can be improved substantially.Time diversity, frequency diversity, and spatial diversity are the most commonapproaches which have been standardized and implemented in many mobilecommunication systems In the context of cooperative communications andour reported studies, we focus on spatial diversity in this thesis Specifi-cally, spatial diversity can be achieved through the implementation of trans-mit and/or receive antenna arrays in each wireless transceiver, for example,multiple-input multiple-output (MIMO) systems It has been shown thatutilizing MIMO transmission may provide spatial diversity gain, which signi-ficantly improves the system performance such as outage probability, errorprobability, and channel capacity [4, 5] This benefit comes from the factthat MIMO systems can exploit the randomness feature of fading channels

si-In particular, by using multiple antenna terminals for communication, theprobability that the information is lost decreases significantly As mentionedbefore, this benefit is resulted from the fact that when some fading channelsare in poor conditions, the remaining channels with favorable conditions maytake care of carrying the information

Although providing spatial diversity gain, the increased size of wirelessterminals and cost associated with multiple antennas constitute major short-comings of MIMO systems To overcome such challenges, cooperative com-munications has been introduced Cooperative communications exploits thebroadcast nature of wireless transmission for providing spatial diversity Theconcept of cooperative communications can be traced back to the study of [6]

in the 1970’s During the past decades, research on the cooperative sity paradigm has attracted significant attention from the research commu-nity [7–17] The basic idea behind cooperative communications is that thedirect communication between the source and destination can be supported

diver-by the introduction of a relay channel [10, 11, 13, 14] Upon receiving thesignal from the source, the relay then processes and forwards the result tothe destination In particular, each single antenna terminal can assist thecommunication of its neighboring transceivers by forwarding their signals tothe destination, resulting in the effect of a virtual antenna array Throughrelaying channels, the destination can receive multiple independent copies ofthe source signal and hence can provide spatial diversity gain As a result,the obtained cooperative diversity gain can mitigate the effect of fading, en-hance the signal reliability, extend the radio coverage, and reduce the powerconsumption These advantages make cooperative communications a promi-sing approach for upcoming technologies of mobile networks

Furthermore, the growth of wireless services has led to a considerable

crease in the demand of radio frequency spectrum Nevertheless, the tional policies of radio frequency spectrum allocation, under which each spe-cific wireless system is allocated a fixed frequency band, have faced the realitythat the radio spectrum bands are under-utilized considerably Specifically,research campaigns have shown that an extensive portion of licensed radiospectrum is heavily under-utilization, i.e., as low as 6% of utilization [18, 19]

conven-In light of this, the cognitive radio (CR) paradigm was originally introduced

by Mitola [20] which allows secondary users (SUs) to utilize a suitable trum access scheme to transmit their signals on the licensed spectrum bands.Realization of the CR paradigm requires the functions of spectrum sensing,spectrum access, spectrum assignment, coordination scheme among the SUs,

spec-as well spec-as a reconfigurable hardware [20, 21] Recently, cognitive radios hspec-asbeen integrated into cooperative communication networks as a way of exploi-ting the combined advantages of these two technologies A more detaileddiscussion for this approach will be presented later, following the subsequentparagraph

Regarding to the interference imposed by the SUs to the primary users(PUs), cognitive radio transmission can be classified into two categories, that

is, interference avoidance and interference management [21] The interferenceavoidance is also referred to as interweave spectrum access, or opportunisticspectrum access With interweave spectrum access, before transmitting a si-gnal over licensed primary bands, an SU first identifies spectrum holes, i.e.,vacant radio spectrum, by sensing the presence of a PU transmission on thesespectrum bands Then, the SU will transmit its signal over these vacant bandswherein the transmit power is not limited by the interference power constraint

of the PUs Using such an approach, the SUs, in principle, guarantee thattheir signal transmission causes no interference to the PUs Nevertheless, anypresence of the PU transmission during the period of the secondary commu-nication has to face serious interference from the secondary signal This isdue to the fact that the SU is unable to perform spectrum sensing to detectthe presence of the PUs during the period of its data transmission To obtain

an accurate sensing outcome, when the SU performs the spectrum sensing, all

of the other transceivers should remain silent This may result in heavy rhead for the network In addition, interweave spectrum access leads to largereduction of the spectrum utilization Apart from the interference avoidance,interference management, which is further categorized as overlay and under-lay spectrum access, enables concurrent transmission of both the PUs andSUs and hence provides high spectrum utilization for the SUs This results

ove-in ove-interference of the SU transmission to the PU transmission In lar, with overlay spectrum access, the SU both assists the PU transmissionand transmits its own signal based on the knowledge of the PU transmission

Therefore, the impairment caused by interference from the SU signal to the

PU receiver can be compensated by the improvement of the PU signal ned through cooperative communications [22] In underlay spectrum access,

obtai-SU transmission is also allowed to coexist with PU transmission on the sed spectrum of the PU However, it is required that the SU transmit powermust be maintained such that the associated interference incurred at the PUreceiver does not exceed the interference power threshold of the PU As aconsequence, underlay spectrum access usually degrades system performance

licen-of the secondary transmission significantly

As mentioned above, cooperative communications has been known as aprominent technology to provide reliable communication, and to extend com-munication range by allowing terminals to relay each other’s signal Recently,cooperative communications has been incorporated into cognitive radio sys-tems that can provide the combined advantages from these two forms of com-munications [23–30] In fact, together with exploiting the spectrum bands ofthe primary users, cognitive users can assist the primary users by forwardingtheir signals to the primary receivers in order to improve the primary perfor-mance Also, due to the broadcast nature of wireless communications, it isfeasible for all cognitive users in the same vicinity to involve in cooperationsuch that the transmit power can be reduced and the interference incurred tothe rest of the network is limited

The remaining of this introduction is organized as follows In Section 2,some fundamentals of cooperative communications are provided In Section 3,background on cognitive radios along with several cognitive spectrum accessschemes are presented Finally, in Section 4, an overview of this thesis isgiven

The purpose of this section is to introduce some of the concepts of tive communications Specifically, the impulse response model of a wirelesschannel, cooperative diversity, relaying protocols, and typical topologies ofcooperative relay networks are presented

Multiple paths along with other channel impairments such as path-loss andshadowing cause channel power fluctuations and lead to unreliable receivedsignals Wireless channel fluctuations can be categorized into two types [1, 2,Chapter 2]:

Figure 1: Example of a wireless channel between the transmitter and receiver

• Large scale fading, which is a consequence of path-loss and shadowing,takes place over relatively long propagation distances Particularly,channel fluctuations caused by path-loss occur over quite long propa-gation distances (100-1000 meters) Path-loss measures the power atte-nuation of the transmitted signal with respect to the propagation dis-tance On the other hand, while channel fluctuations resulted fromshadowing occur over propagation distances proportional to the length

of the obstacles (usually, 10-100 meters) such as buildings and hills.These obstacles attenuate the signal power through absorption, reflec-tion, scattering, and diffraction

• Small scale fading is caused by constructive and destructive addition ofmultipath signals occurring over very short distances, on the order ofthe carrier wavelength

2.1.1 Input-output model of a wireless channel

A wireless signal is usually propagated through multiple paths before ching the receiver which therefore will receive multiple copies of the transmitsignal These multiple signal paths, which are caused by reflections, scat-tering, and diffractions from objects in the transmission environment, may

rea-be illustrated as in Fig 1 The phenomenon of multiple propagation pathscan be characterized by identifying the impulse channel response Usually,the continuous change of the physical characteristics of the communicationenvironment evidently results in a time-variant channel impulse response Tomodel a time-variant multipath channel, we assume that the signal with abandwidth limited to W transmitted over this multipath channel is given in

the general form as [31]

s(t) =√2Re[sb(t) exp(j2πfct)] (1)where Re[·] denotes the real part of a complex number, sb(t) is the basebandwaveform of the transmitted signal, and fc is the carrier frequency In view of[31, Chapter 13], a fading channel with multipath propagation can be modeled

as a linear time-varying filter The received bandpass signal at the receiverwhen ignoring the noise and interference from other users can be expressed inthe form

R (τ , τ ,∆t) = E [h∗(τ , t)h(τ , t+ ∆t)] (7)

where E[·] denotes the expectation operator, τ1 and τ2 are the propagationdelays, and ∆t is the difference in observation time In uncorrelated scatte-ring transmission media, where the attenuations and phase shifts of differentpropagation paths are uncorrelated, we have [31, Chapter 13]

RH(∆f, 0) = RH(∆f ) The range of values of ∆f which makes |RH(∆f )|essentially nonzero is called the coherence bandwidth Bc of the channel Ac-cording to [31, Chapter 13], we have the following relationship

Bc≈1

Tm

(11)When the signal bandwidth is greater than the coherence bandwidth W > Bc,the channel is called frequency selective Otherwise, if the signal bandwidth issmaller than the coherence bandwidth W < Bc, the channel is called frequencynon-selective

In addition, the Doppler power spectrum of the channel is given by [31,Chapter 13]

Tc≈1

2.1.2 Statistical Models of Fading Channels

Usually, statistical models are utilized to characterize a multipath fading nel and analyze the system performance of wireless communication networks.There exist various statistical models of fading channels, including Rayleigh,Rician, and Nakagami-m fading

chan-The Rayleigh fading characterizes a multipath channel with a large number

of signal propagation paths and without any line-of-sight (LOS) propagationpath In view of the central limit theorem, a Gaussian random process can beadopted to model the channel impulse response As a consequence, the fadingchannel coefficient can be modeled as a symmetric circular complex Gaussianrandom variable (RV) As such, the envelope of the channel impulse responsehas a Rayleigh distribution of which the probability density function (PDF)

fg(x) = 1

Ωexp



−xΩ



Fg(x) = 1 − exp−

xΩ



In the scenario when there is an LOS propagation path from the mitter to receiver of which the power is very high compared to the remaining

paths, the multipath fading channel is modeled as Rician fading The PDF

of the magnitude |h| of a Rician fading channel can be expressed as

!, x ≥ 0(18)where I0(·) is the modified Bessel function of the first kind [32, eq (8.431)] andthe Rician K factor expresses the ratio of the power of the LOS component

to the power of the non-LOS (NLOS) components The PDF of the channelpower gain is given by [33]

!, x ≥ 0 (19)

Note that the Rayleigh distribution is a special case of the Rician distributionwhen K = 0

In a fading environment where the size of clusters of the scatters is rable to the wavelength of the carrier, the multipath channel can be modeled

compa-as Nakagami-m fading In this ccompa-ase, the PDF of the channel magnitude |h|can be written as

In addition, the PDF and CDF of the channel power gain g = |h|2 are,respectively, given by



Fg(x) = 1 −Γ(m, mx/Ω)

where Γ(·, ·) is the incomplete gamma function defined as in [32, eq (8.350.2)]

In fact, the Nakagami-m fading model comprises several other kinds of fadingchannels depending on the particular value of m For instance, this modelbecomes a one-sided Gaussian and Rayleigh fading when m = 0.5 and m = 1,respectively

The system performance of a conventional wireless communication systemconsisting of a pair of single antenna terminals can be expressed through thefading channel gain of the associated link In this case, there exists a highprobability that the fading channel gain is deeply faded, i.e., the channel ma-gnitude attenuates significantly because of destructive addition of multipathsignal components at the receiver This phenomenon leads to significant re-duction in system performance, making conventional wireless communicationsunreliable Intuitively, because of the random nature of fading environments,

if the signal is transmitted on multiple independent paths, then there is a muchlower probability that all these paths simultaneously experience deep fades

At the receiver, these independent replicas of the transmit signal are ned by using an appropriate processing algorithm such that the signal powerincreases and system performance improves This performance improvement

combi-is a result of spatial diversity, which efficiently mitigates the impairment offading channels To get insight into this concept, some related definitions aresummarized in what follows

Definition 1 The exponential order n of an RV X with a non negative port ρ is defined as [34]

P∞ out = (Gcρ)−Gd

+ o(ρ−Gd

where o(g(x)), which satisfies limx→x 0

o(g(x)) g(x) = 0, denotes the higher orderterms of g(x) as x → x0 The term Gcdenotes the coding gain of the networkdefined as the SNR advantage of the asymptotic outage probability curve ascompared to the reference curve ρ−Gd In addition, Gd denotes the diversitygain which expresses the decrease exponent order of the outage probabilitywith respect to the average SNR

Definition 2 Assume that a wireless network operates at an average SNR =

ρand rate R(ρ) bits per channel use and P (R(ρ)) is the outage probability

at rate R(ρ) Diversity gain is defined as the exponent decrease order of

Pout(R(ρ)) with respect to the average SNR in the high SNR regime [34–36]:

Before discussing background on cooperative communications, let us considerthe conventional communication between a single pair of transmitter-receiver,so-called direct communication For direct communication from a source to adestination, the maximum mutual information is represented as

ID= log(1 + ρ|h0|2) (27)where ρ = P/N0is the average SNR, P is the transmit power of the source, h0

is the channel coefficient of the link from the transmitter to receiver, and N0isthe noise power at the destination Therefore, the system outage corresponds

to the event ID< R, where R is the target transmission rate, that results in

Figure 2: Topology of a conventional cooperative relay network

The factor ρ−1 indicates that the diversity gain is unity This implies thatdirect communication of a pair of single-antenna terminals does not providespatial diversity

Conventional diversity systems utilize multiple antenna transceivers to achievediversity gain even though implementing multiple transmit and/or receiveantennas may be not feasible because of cost and hardware limitations Asdiscussed earlier, cooperative communications has recently developed as anefficient means of providing diversity gain while overcoming the aforementio-ned limitations [13, 14] According to the cooperative diversity paradigm, thetransmit signal from a source can be forwarded by the neighboring terminals,called relay nodes, by making use of the broadcast nature of wireless trans-mission As a result, multiple copies of the signal, which are transmitted overindependent fading paths through relay nodes, are received and processed atthe destination, resulting in diversity gain The basic model of a cooperativerelay network is illustrated in Fig 2 In this figure, let us denote h0, h1, and

h2, respectively, the channel coefficients from source S to destination D, fromsource S to relay R, and from relay R to destination D Each cycle of signaltransmission from the source to the destination is divided into two phases Inthe first phase, the source broadcasts its signal to both the relay and desti-nation In the second phase, the relay processes the received signal from thesource and forwards the result to the destination The two received versions

of the source signal are combined at the destination by using an appropriatealgorithm, resulting in diversity gain Depending on the method that the re-lays utilize to process and forward the received signal from the source, thereare three major cooperative protocols, namely, amplify-and-forward (AF),decode-and-forward (DF), and compress-and-forward (CF) [10] In this the-sis, we concentrate on AF and DF because these two techniques are widely

where P1 = P2, ρ = P1/N0 For a given transmission rate R, system outage

is equivalent to the event IAF < R such that the outage probability is givenby

Assuming that the channel power gain |hi|2, i ∈ {0, 1, 2}, follows an pendent exponential distribution, i.e., Rayleigh fading environment, the ou-tage probability can be expressed in the high SNR regime as [11]

inde-PAF out = 12Ω0

Ω1+ Ω2

Ω1Ω2

 22R

− 1ρ

2

where Ωi denotes the channel mean power of the link corresponding to thechannel coefficient hi Clearly, factor ρ−2 in (35) implies that the systemachieves a diversity gain of two Although the AF protocol suffers from noiseaccumulation at the relays, this protocol still gains great interests because itimposes less signal processing burden to the relays as compared to the DF or

CF protocols

2.4.2 Fixed Decode-and-Forward Protocol

Different to AF relaying, DF relaying wherein the relay is able to decodethe source signal, to re-encode and forward the result to the destination, wasstudied in [10] It is not feasible to obtain cooperative diversity directly in

a DF relay network since it may suffer from forwarding erroneously decodedsignals at the relays [10, 11, 40, 41] In what follows, a detailed description ofthe operation of the DF protocol is presented In general, the communicationprocess of a DF relay network occurs over two time slots In the first timeslot, the source broadcasts the signal while the relay as well as the destinationlisten to this transmission Subsequently, if the relay decodes the source signalsuccessfully, it will cooperatively forward the result to the destination in thesecond time slot Otherwise, the relay keeps silent Note that the sourceremains silent during this time slot At the end of this transmission, thedestination tries to decode the received signal The maximum average mutualinformation of the DF relaying can be expressed as [11]

To guarantee that both the relay and the destination successfully decode thesource signal, it is needed to apply the minimum function of the two mutualinformation measures

Given a target rate R, system outage is equivalent to the event IDF < R.Therefore, outage probability of the DF relay network can be formulated in

In the high SNR regime, it has been shown that [11]

As a consequence, fixed DF relaying does not provide diversity gain because

of the ρ−1 behavior at high SNR This can be inferred from the fact thatrequiring both the relay and destination to fully decode the source signaldegrades the system performance to that of the direct transmission from thesource to the destination

2.4.3 Adaptive Decode-and-Forward Protocol

To overcome the shortcoming that the performance of a fixed DF relayingsystem is limited to the direct transmission from the source to the destination,another protocol, namely adaptive DF relaying, has been developed [10, 11,40–43] Instead of forwarding the decoded signal without considering whetherthe relay decodes the signal error-freely or not, in adaptive DF relaying, onlyerror-free decoded signals are forwarded to the destination The instantaneouschannel state information (CSI) h1 of the channel from the source to relay

is utilized to regulate the relaying transmission In particular, if the channelpower gain |h1|2is less than a predefined threshold, the source will retransmitthe signal and the relay keeps silent Otherwise, if the channel power gain |h1|2

is greater than the threshold, the relay decodes the received signal from thesource and forwards the result to the destination An adaptive DF relayingsystem is in outage as long as both the channel from the source to relay andthe direct channel from the source to destination are in outage To formalizethis protocol, we use the maximum mutual information for the adaptive DF

this case is attributed to the two repeated transmissions over the direct nel from the source to destination In the second case, the relay is able todecode the source signal and then it forwards the source signal to the destina-tion Therefore, the associated maximum mutual information is constitutedfrom both the direct transmission and relaying transmission

chan-At a given target R, outage occurs if IADF < R, that is,



{|h1|2

<(22R− 1)/ρ}\{2|h0|2

<(22R− 1)/ρ}[ 

{|h1|2

≥ (22R

− 1)/ρ}

\{|h0|2

+ |h2|2

<(22R− 1)/ρ} (40)Due to the mutually exclusive property of the events in (40), the outageprobability can be rewritten as [11]

2.4.4 Incremental Relaying Protocol

In the aforementioned relaying protocols, multiple transmissions of the sourcesignal are performed by the relay and/or the source that reduces the utili-zation efficiency of the degree of freedom of the channels To overcome thisshortcoming, the incremental relaying protocol has been proposed whereinthe relay takes advantage of the feedback of the acknowledgement from thedestination for making a decision of whether forwarding the source signal ornot [11,44] Specifically, if the direct transmission is successful, then the relayremains idle; otherwise, it cooperates with the source to forward the sourcesignal to the destination in an attempt to offer spatial diversity

The incremental relaying protocol operates with variable rate That is,

if the direct transmission is successful, the network offers transmission rate

R, and if the relay cooperates with the source to forward the source signal,

(45)

Clearly, the incremental AF relaying protocol provides spatial diversity gainwithout suffering from the loss in spectral efficiency as this mechanism makesefficient use of the degrees of freedom of the channels

2.4.5 Estimate-and-Forward Protocol

Apart from the above relaying protocols, the key feature of the forward (EF) (also called CF or quantize-and-forward (QF)) is that the relayretransmits a quantized and compressed version of its output signals to thedestination [45–47] Unlike DF relaying, the relay does not decode the sourcesignal, but it deploys different kinds of observations on the received signal.The dedicated relay quantizes (and possibly compresses) the received signalfrom the source and then forwards the result to the destination The pro-cessing at the EF relay comprises of a conditional Karhunen-Love transformfollowed by a separate Wyner-Ziv coding of each output data stream at dif-ferent transmit rates Scalar quantization under entropy constraint [46] orunder minimum mean square error (MMSE) [47] can be adopted to estimatethe received signal At the destination, the quantized/compressed versionfrom the relay and the signal from the source are combined by using a specificrelay combining technique EF relaying operates most effectively in the casethat the source-relay and source-destination channels are in similar qualityand the relay-destination channels are of good quality It is unlikely for therelay to decode the source signal; however, the independent received versions

estimate-and-at the destinestimate-and-ation can help to decode the source signal

2.4.6 Coded Cooperation Protocol

Different to the above mentioned relaying protocols, coded cooperation is other mechanism to offer spatial diversity wherein each terminal cooperateswith each other in the perspective of channel coding [8,48,49] Extensive stu-dies of this idea have been reported in the research literature (see, e.g., [48–53]and the references therein) In [50], for improving the diversity gain, Janani et

an-al have proposed two extensions to coded cooperation systems by ting turbo codes and space-time transmission In [51], a coded cooperationnetwork has been proved to achieve considerable diversity and coding gain, ascompared to the non-cooperation network The basic idea behind this concept

implemen-is that instead of relaying the source signal the relay, also called the partner,makes an attempt in transmitting the incremental redundancy for the source

At the destination, this incremental redundancy can be combined with thecodeword sent by the source to generate a codeword with higher redundancy.The redundancy in the received codeword is exploited to provide a betterpossibility to recover the original information from the erroneously receivedsignal In many coding mechanisms, the encoding and decoding processes can

be carried out in such a way that redundancy can be added or removed fromthe codeword easily and independently This will help the partner to computethe redundancy of the source codeword and hence these coding mechanismscan be applied in coded cooperation By cooperatively transmitting the incre-mental redundancy for the source, a coded cooperation system is able to offerhigher spectral efficiency as compared to the conventional relaying schemes.For instance, Fig 3 depicts the process of transmitting a codeword incoded cooperation networks using an error correcting code [54, p 131] In thisnetwork, the partner will assist the source by transmitting the redundancy

of the source codeword to the destination Each codeword of the source istransmitted in two phases In the first phase, at the source, a data stream

of Nsinformation symbols is encoded by the cyclic redundancy check (CRC)encoder and is then further encoded by the forward error correcting (FEC)encoder which generates a codeword of N1 symbols with a code rate R1 =

Ns/N1 This codeword is sent by the source to the destination and also isoverheard by the partner Then, the partner attempts to decode the sourcecodeword based on both the associated FEC and CRC codes In the casethat the CRC exposes no error in the decoded symbols, the recovered Ns

information symbols at the partner are encoded again by the CRC encoderand then by the FEC encoder, generating a codeword of N symbols, N > N1.Therefore, the code rate at the partner is less than that at the source, i.e.,

R= Ns/N < R1

The codeword at the partner is generated in the same way as the codewordtransmitted by the source but only N2= N − N1 extra parity symbols will be

Figure 3: Coded cooperation [54, p 132]

transmitted to the destination To do so, these extra symbols are separatedfrom the remaining symbols and hence can be extracted independently as anew redundancy for the source In the second phase, the partner transmits the

N2 redundancy symbols to the destination The N2 received symbols at thedestination will be combined with the N1symbols from the source to constructthe codeword of N symbols with a code rate R The original information can

be recovered by decoding this codeword which has larger redundancy andhence is stronger as compared to the source codeword

From the receiving end, the combined codeword at the destination can beseen as it is encoded and transmitted at a code rate R by the source Toconstruct the codeword with N1 symbols, the source will puncture N2 sym-bols from the overall codeword with N symbols and leave these N2puncturedsymbols for the partner to transmit Clearly, the received codeword at thedestination is strengthened by combining the N2 extra parity symbols recei-ved from the partner and the N1 symbol codeword received from the source.Note that the N1 symbol codeword of the source must be a valid codewordand actually belongs to a code which is weaker than the code applied at thedestination To perform coded cooperation, it is required that the partnermust decode the source codeword successfully which can be confirmed by

is also noted that various channel coding schemes can be applied for codedcooperation networks such as block or convolutional codes, or a combination

of both Puncturing, product codes, or different forms of concatenation can

be employed to extract the redundancy of a codeword

2.5 Relaying Strategies

There exist three fundamental relaying strategies, namely, repetition-basedrelaying, distributed space-time coding (DSTC), and relay selection The keyfeature behind repetition-based relaying is that the relays forward the sourcesignal over orthogonal channels Even though, repetition-based relaying issimple for implementation, occupying the orthogonal channels leads to signi-ficant bandwidth inefficiency [11] On the contrary, in a DSTC network, allthe relays forward the source signal in the same time slot The orthogonalproperty of the codeword received at the destination is constituted in boththe time and space domains Although DSTC requires synchronization atthe symbol level as well as signalling overheads, it is more effective than otherschemes in using the network resource [10,42,55] Full spatial diversity can beobtained under repetition-based relaying and DSTC networks Relay selection

is another approach to obtain cooperative diversity gain while improving thenetwork throughput considerably [40, 56] In a relay selection network, onlyone relay is selected to forward the source signal to the destination instead oftaking all the relays involved in the relaying transmission

2.5.1 Repetition-Based Relaying

Fig 4 depicts the operation of repetition-based relaying In a repetition-basedrelaying system, each relay takes its turn transmitting the source signal in aseparate time slot while the other terminals keep idle in this time slot [11] Inparticular, in the broadcast phase, the source sends its signal to all the relaysand the destination In the relaying phase, each relay forwards the sourcesignal to the destination in sequence Increase in the number of allocatedtime slots results in bandwidth inefficiency To overcome this drawback, theincremental relaying protocol has been proposed That is, the destination tries

to decode the source signal in the broadcast phase, and an acknowledgement(ACK) bit, which notifies the success or failure of the direct transmission, issent back to the relay If the destination decodes the source signal successfully,

Figure 4: Repetition-based relaying

the relays do not forward the source signal Otherwise, the relays forward thesource signal to the destination in the relaying phase

2.5.2 Distributed Space-time Coding

Although offering spatial diversity gain and implementation simplicity, tition based relaying suffers from decreased bandwidth efficiency [10] This isdue to the fact that each relay requires its own channel to forward the sourcesignal To enhance the bandwidth efficiency, DSTC has been developed whe-rein multiple forwarding transmissions of the relays occur simultaneously onthe same channel [10, 42, 55, 57] Communication in a DSTC relay networkoccurs over two phases In the first phase, the source sends signals to therelays, while in the second phase, the relays cooperatively encode the signalsreceived from the source into a space-time code and then forward the encodedsignals to the destination For a DSTC relaying system, the relays encodethe source signals by applying certain space-time codes such as space-timetrellis codes (STTCs) The received codeword at the destination is generated

repe-in both the space and time dimensions by usrepe-ing a specific space-time codrepe-ingscheme; thus, spatial diversity is obtained Even though, DSTC relaying pro-vides higher bandwidth efficiency as compared to repetition-based relaying,

it requires synchronization at symbol level and complicated signaling amongthe transceivers

2.5.3 Opportunistic Relaying

As mentioned earlier, repetition-based relaying may result in decreased width efficiency together with other network resources since multiple relaysare required to forward the source signal over orthogonal channels To enlargebandwidth efficiency, opportunistic relaying has been proposed which suffersfrom no performance loss as compared to repetition-based relaying [41] Inopportunistic relaying, a single relay out of the set of K relays, which offers

the best end-to-end path between the source and the destination, is selected

to forward the source signal [40, 41, 56, 58–60] In particular, opportunisticrelaying achieves diversity gain on the order of the number of relays in thenetwork by choosing the relay path contributing the most to the output SNR

at the destination To realize opportunistic relaying, local CSI from the source

to each relay and from the relay to the destination is required to be known

to the relay Therefore, each relay has to monitor the CSI towards the sourceand the destination and makes a decision of which relay has the strongestpath for signal relaying in a distributed manner To obtain such CSI, eachrelay overhears a single transmission of a ready-to-send (RTS) packet fromthe source and a clear-to-send (CTS) packet from the destination and esti-mates the power strength of the corresponding relaying channels [40] Based

on the measured CSI, each relay calculates the end-to-end performance of theassociated relaying path To reduce communication among all relays, a me-thod based on time is adopted when selecting the best relay That is, at thetime each relay successfully receives the CTS packet, an associated timer isstarted with the time to be inversely proportional to end-to-end performancemeasure of the corresponding relaying path [40] A relay with the best end-to-end channel conditions will be selected to forward the source signal sinceits timer will expire first Denoting h1i and h2i as, respectively, the channelpower gains from the i-th relay to the source and to the destination, selectingthe best relay can be carried out by the following two policies [40] UnderPolicy I, the minimum of the two channel power gains from each relay to thesource and each relay to the destination is utilized to decide the best relay,i.e.,

Under Policy II, the harmonic mean of the two channel power gains from eachrelay to the source and each relay to the destination is employed to select thebest relay as follows:

hi= 2h1ih2i

The relay Rb which maximizes the measure hi is selected as the best relay

to forward the source signal This relay provides the best end-to-end pathcondition between the source and the destination, as follows:

There exists a variety of relay network topologies to be studied in the context

of cooperative communications

2.6.1 Single Relay Networks

As mentioned before, Fig 2 depicts the topology of a conventional tive relay network [11, 15, 61, 62] The communication can be divided intotwo phases: the broadcast phase and the multiple-access phase During thebroadcast phase, the source broadcasts the signal s to both the relay and thedestination The received signals at the relay R and the destination D are,respectively, given by

where h0 and h1 denote the channel coefficients of the S → D and S → Rlinks, respectively In addition, nR and nD1 denote the AWGN at the relay

R and the destination D in the broadcast phase, respectively

During the second phase, the relay processes the received signal by using

a specific relaying protocol (AF, DF, or EF) and then forwards the resultingsignal to the destination The received signal at the destination D can beexpressed in the form

yD2= h2f(s, nR) + nD2 (51)where f (·, ·) represents the utilized relaying protocol, h2 denotes the channelcoefficient of the R → D link, and nD2 denotes the AWGN at the destination

D in the multiple-access phase The two received signals at the destination

D in the two phases are processed by utilizing MRC or selection combining(SC), providing spatial diversity gain

2.6.2 Multiple Relay Networks

Multiple relay networks have been extensively studied in the literature [40,63–65] Fig 5 shows a general relay network topology where the source S com-municates with the destination D through the assistance of multiple potentialrelay nodes, R1, ,RK All the relays forward the received signals from thesource to the destination At the destination, the received signals are proces-sed by using an appropriate algorithm In particular, the communication isstretched over two phases In the first phase, the source broadcasts the signal

coef-In the second phase, each relay processes the received signal according to

a specific protocol and forwards the result to the destination using transmitpower P2 over its own channel Without loss of generality, it is assumed thatthe source and the relay use the same transmit power, P1= P2 The receivedsignal at destination D corresponding to the S → Rk→ D relay path can beexpressed as

yD k = h2kf(s, nR k) + nD k (53)where f (·, ·) represents the employed relaying protocol, h2kdenotes the chan-nel coefficient of the Rk→ D link, andnD k denotes AWGN at destination D.The instantaneous SNR at the destination can be formulated as

γ= ξ(γ1, , γK) (54)where γk is the SNR for the transmission path S → Rk → D andξ(·, , ·)

is a function depending on the algorithm at the destination which is utilized

to process the received signals from all the relays For example, ξ will be thesummation function in the case MRC is applied at the destination

On the other hand, a relay network with opportunistic relaying, as depicted

in Fig 6, operates in a different manner That is, only the best relay is chosen

Figure 6: Topology of a multiple relay network with opportunistic relaying

to forward the source signal to the destination In particular, the networkobtains spatial diversity gain by selecting the relaying path contributing themost to the output instantaneous SNR at the destination Accordingly, theinstantaneous SNR at the destination can be written as

2.6.3 Multi-hop Relay Networks

Topology of a multi-hop relay network can be considered as a generalizedmodel of dual-hop relay systems wherein the source signal can be traversedthrough multiple intermediate relay nodes It has been shown that the use ofmulti-hop relaying transmission can expand the radio coverage significantly[66–72] Fig 7 depicts a general (K +1)-hop relay network consisting of source

S, destination D, and K relay nodes wherein communication spans over K + 1hops In the first hop, the source transmits its signal s to the first relay R1.During the (k + 1)-th hop, the k-th relay employs DF or AF protocols toforward the received signal to the next relay The instantaneous SNR at thedestination is given by

Figure 7: Topology for a multi-hop relay network

where γk = ρk|hk|2 denotes the instantaneous SNR over the k-th hop, ρk =

Pk−1/N0 is the average SNR of the k-th hop, and function ψ represents therelaying protocol applied by the relays

In has been shown that the conventional fixed spectrum allocation policieslead to the consequence that a large portion of the licensed spectrum is si-gnificantly under-utilized To overcome such a disadvantage, cognitive radioswere proposed by Mitola [20] as a means of enabling cognitive radio devices,also called secondary users, to access the licensed spectrum more efficiently.Different to conventional radio devices, a cognitive radio device must havecognitive capacity and reconfigurability [21, 73] Cognitive capacity enables

a secondary user to sense and acquire status of its surrounding environment,e.g., transmit power, carrier frequency, and modulation scheme By doing

so, each cognitive radio device can specify the best vacant spectrum to cess for its communication The latter capability is realized based upon aplatform, known as software-defined radio, which helps the SUs to flexiblyadapt their operation parameters in accordance with the sensing outcomes.More specifically, according to IEEE 1900.1, cognitive radios have four de-tailed functions [74] The first function is spectrum sensing which is mainlyresponsible for detecting the vacant licensed spectrum as well as the relatedoperation parameters Through this process, a secondary user is able to se-lect and access the best available spectrum for its communication Anotherduty of spectrum sensing is to detect a transmission of the primary user onthe vacant spectrum in order to help a secondary user releases it as a way ofprotecting the PU from being interfered The second function guarantees thatthe SU can access the vacant licensed channels to deliver a signal by using

ac-a specific spectrum ac-access mechac-anism There exist interweac-ave, underlac-ay, ac-andoverlay spectrum access which have been discussed in [75] The third functionprovides fairness among multiple or dissimilar SUs The forth function de-fines an appropriate radio interface which helps an SU to adjust its operationparameters to the variations of the cognitive radio environment in order to

utilize the available spectrum most efficiently

With interweave spectrum access, so-called opportunistic spectrum access, an

SU opportunistically accesses the licensed spectrum as long as this spectrum

is sensed to be vacant [20] In fact, opportunistically accessing the vacantspectrum was the original idea of cognitive radio to improve the spectrumutilization efficiency In particular, to protect the PUs from any harmful in-terference from the secondary transmission, it is required that the SUs mustacquire occupancy status of the PUs on the licensed spectrum bands The-refore, accurate spectrum sensing to gather knowledge of the PU activities

on the licensed channels is very important in interweave cognitive systems.Based on the outcomes of spectrum sensing, an SU is able to recognize thevacant spectrum, i.e., radio frequency bands which are not occupied by theSUs, and uses these spectrum bands for its communication

In underlay spectrum access, concurrent spectrum access by the SU and the

PU is allowed given that the interference from the SU imposed to the PUreceiver is kept below a given threshold That is, the power of interferencefrom the secondary transmission at the primary receiver plus its noise powermust be maintained below an acceptable level [76, 77] From an information-theoretic perspective, due to the advantage of high spectrum utilization effi-ciency, the research community pays much interest to the underlay cognitiveaccess (see [78–83] and the references therein) Taking different fading channelmodels into account, channel capacity of underlay cognitive radio networkshas been investigated based on the fact that the interference power constraint

at the primary receiver can be transformed to the transmit power constraint

of the secondary transmitter [78, 79] Considering different types of fadingchannels, channel capacity of underlay spectrum access systems subject tothe average and peak interference power constraints of the primary receiverwas addressed in [80] In [81], the channel capacity of an underlay cognitivechannel has been analyzed where the channels from the secondary transmitter

to the secondary receiver and primary receiver have different fading ments The problem of spectrum shaping of the SU signals has been studied

environ-as a way of maximizing the achievable rate under the constraint of estimationerror on the secondary signals [82]

Typically, since a secondary user in underlay spectrum access is strictlyconstrained by a predefined interference threshold, channel capacity and radiocoverage of the secondary transmission may be severely degraded As such,

channel capacity optimization is very crucial in underlay cognitive systemsand has recently attracted much attention [84–88] Particularly, to maximizechannel capacity of a cognitive radio network, it is required that at most one

SU utilizes a transmit power between zero and its peak transmit power reas all the other SUs adopt transmit powers of either zero or their peak trans-mit powers [84] Subject to the signal-to-interference plus noise ratio (SINR)constraint at the primary receiver and the transmit power limit constraint ofthe secondary transmitter, power allocation to maximize the channel capacity

whe-of a spectrum sharing system over fading channels has been developed in [85].Considering the minimum SINR constraints at the secondary transmitters andinterference power constraint at the primary receiver, power allocation for anunderlay cognitive system has been addressed in [86, 87] In [88], the opti-mization problem of maximizing the achievable capacity of underlay, overlay,hybrid underlay-overlay cognitive radio networks has been analyzed

Incorporating cooperative diversity techniques into cognitive radios can beperformed through cooperative spectrum sensing [89–91] or cognitive relaynetworks [23–30] In particular, cooperative spectrum sensing can improvethe accuracy of the spectrum sensing process as well as reduce the risk of hid-den terminal problems [91] In addition, the idea that letting the secondaryusers forward the primary signals has been shown to decrease the transmissiondelay of the primary users as well as enhance the spectrum utilization of thesecondary users [23, 24] In [25], a two-phase protocol for DF relay networkswherein the SUs are given spectrum access with the PUs and also assist thePUs by relaying their signals has been reported Therein, an optimal thre-shold value for the power fraction of the primary signal over the total transmitpower at the SU can be achieved If the power fraction of the primary signalover the total transmit power at the SU is above this threshold, then theoutage probability of the primary transmission will be equal to or lower thanthat of the primary transmission in the case without spectrum sharing Thework of [26] proposed an adaptive cooperation diversity scheme for cognitivemultiple relay networks with best relay selection as a means of both improvingthe performance of secondary transmission and guaranteeing the QoS of theprimary transmission This study has shown that, with a threshold of theprimary outage probability, full diversity of the cognitive secondary network

is achieved In addition, a novel cooperative transmission for cognitive radionetworks wherein the relay assists both the primary and secondary users toforward their signals by using DF mode has been developed in [27] It hasbeen shown that this cognitive cooperative mechanism improves the outage