21. Beacon-Assisted Spectrum Access With Cooperative Cognitive Transmitter and Receiver

7/31/2019 21. Beacon-Assisted Spectrum Access With Cooperative Cognitive Transmitter and Receiver

1/15

Beacon-Assisted SpectrumAccess with Cooperative Cognitive

Transmitter and ReceiverAli Tajer, Student Member, IEEE, and Xiaodong Wang, Fellow, IEEE

AbstractSpectrum access is an important function of cognitive radios for detecting and utilizing spectrum holes without harming the

legacy systems. In this paper, we propose novel cooperative communication models and show how deploying such cooperations

between a pair of secondary transmitter and receiver assists them in identifying spectrum opportunities more reliably. These

cooperations are facilitated by dynamicallyand opportunistically assigning one of the secondary users as a relay to assist the other

one, which results in more efficient spectrum hole detection. Also, we investigate the impact of erroneous detection of spectrum holes

and thereof missing communication opportunities on the capacity of the secondary channel. The capacity of the secondary users with

interference-avoiding spectrum access is affected by 1) how effectively the availability of vacant spectrum is sensed by the secondary

transmitter-receiver pair, and 2) how correlated are the perceptions of the secondary ransmitter-receiver pair about network spectral

activity. We show that both factors are improved by using the proposed cooperative protocols. One of the proposed protocols requires

explicit information exchange in the network. Such information exchange in practice is prone to wireless channel errors (i.e., is

imperfect) and costs bandwidth loss. We analyze the effects of such imperfect information exchange on the capacity as well as theeffect of bandwidth cost on the achievable throughput. The protocols are also extended to multiuser secondary networks.

Index TermsCognitive radio, spectrum access, cooperative diversity, opportunistic communication, channel capacity.

1 INTRODUCTION

COGNITIVE radios have been introduced as a potentialsolution for alleviating the scarcity of frequency spec-trum in overly crowded environments [1], [2], [3], [4].According to the Federal Communication Commission(FCC), a vast portion of the frequency bands is used onlysporadically and furthermore, the usage varies geographi-cally and temporally. Such inefficient utilization of spectrum,as well as the increasing demand for frequency bands by theexisting and emerging wireless applications, motivatesopportunistic access to licensed bands by unlicensed(secondary) users. The notion of cognitive radios makes itpossible to accommodate self-configuring ad hoc linkswithin currently established wireless communication infra-structures. For accessing vacant spectrum bands assigned tolicensed users, the secondary userscontinuously and activelymonitor the spectrum in order to efficiently make use of thespectrum hole while avoiding interference with licensedusers (interference-avoiding [5] or interweave paradigm [6])or having controlled level of interference temperature(interference-controlled [5] or underlay paradigm [6]).

Related studies on spectrum access can be broadlycategorized as those proposing efficient methods forlocating the holes in the spectrum [7], [8], [9] and thosediscussing how to optimally allocate the available unusedfrequency bands to secondary users [10], [11], [12], [13]. Theworks in the former category aim at improving the

delectability of the unused portions of spectrum by usingphysical layer techniques, while those in the latter one,develop media access control (MAC) protocols seeking tomaximize secondary users network throughput. Thechannel access protocols we propose in this paper fallwithin the first category mentioned above. These protocols

allow the secondary transmitter and receiver to coopera-tively listen to the primary user instead of havingindependent observations.

The gains yielded for the detection of unused spectrumare due to the diversity gains introduced by the coopera-tion. The increase in the channel capacity is also partly dueto the same diversity gain, but is mostly shaped by reducingthe uncertainty at the secondary transmitter and receivernodes about each others observation of the spectralactivity, which is a by-product of the cooperation betweenthem. As discussed in [14], discrepancy in spectral activityawareness at the secondary transmitter and receiver, whichis due to their spatial separation and/or imperfect channelsensing, incurs a loss in channel capacity.

The underlying idea of the cooperation models devel-oped is to have a one-time broadcast of a beacon messageand a one-time relaying of this message by one of thesecondary users such that both enjoy a second-orderdiversity gain in detecting the beacon message. This beaconmessage is a codeword within the codebook of the primaryuser, reserved specifically for the purpose of informing thesecondary users of the vacant channel.

Besides seeking spectrum opportunities and utilizingthem, another major function of secondary users is todetect the return of primary users and agilely vacate thechannel. The idea of deploying cooperative diversityschemes and their merits in detecting the return of theprimary users has also been investigated as an independent

problem in [15], [16], [17].

112 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 9, NO. 1, JANUARY 2010

. The authors are with the Department of Electrical Engineering, ColumbiaUniversity, New York, NY 10027. E-mail: {tajer, wangx}@ee.columbia.edu.

Manuscript received 2 Oct. 2008; revised 3 Mar. 2009; accepted 6 May 2009;published online 27 May 2009.For information on obtaining reprints of this article, please send e-mail to:[email protected], and reference IEEECS Log Number TMC-2008-10-0393.

Digital Object Identifier no. 10.1109/TMC.2009.103.1536-1233/10/$26.00 2010 IEEE Published by the IEEE CS, CASS, ComSoc, IES, & SPS

Authorized licensed use limited to: Columbia University. Downloaded on November 28, 2009 at 23:20 from IEEE Xplore. Restrictions apply.


2/15

2 SYSTEM DESCRIPTIONS

We consider a primary transmitter Tp and a pair ofsecondary transmitter Tt and receiver Tr users. Thesecondary users continuously monitor the channel used bythe primary user, seeking an opportunity to take it overwhen it is unused. We assume that all channels between theprimary user and the secondary users and those between the

secondary users are quasi-static wireless flat fading channelswhich remain unchanged during the transmission of a blockof symbols and change to independent states afterwards.

We denote the channels between the primary user andthe secondary transmitter and receiver by p;t and p;r,respectively. The channel between the secondary nodes aredenoted by t;r and r;t. The physical channel between nodesi 2 fp; t; rg and j 2 ft; rg has an instantaneous realization

i;j ffiffiffiffiffiffiffi

i;jp hi;j; 1

where fading coefficients hi;j are assumed to be independent,circularly symmetric complex Gaussian random variablesCN 0; 1. The term i;j accounts for path loss and shadowingand is given by i;j Si;jd

i;j , where Si;j represents the log-normal shadowing effect, di;j is the physical distancebetween nodes i and j, and > 0 is the path loss exponent.

In [7], it is shown that compared with energy detectionmethods, the detection of vacant spectrum can be signifi-cantly improved through transmitting pilot signals by theprimary user and deploying coherent detection at thesecondary users. Motivated by this fact, in this paper weassume that whenever a channel is released by the primaryuser, it is announced to the secondary users by having theprimary user broadcast a beacon message. It is noteworthythat as eventuallywe areevaluating thenetworkthroughput,it is valid to assume that the secondary users are alwayswilling to access the channel and are continuously monitor-

ing the channel for the beacon message by the primary user.We consider Nconsecutive channel uses for the transmis-

sion of the beacon message. The beacon message, sent by theprimary user, is denoted by xxb xb1; . . . ; xbNT, and thereceived signals by the secondary transmitter and receiverare denoted by yyt yt1; . . . ; ytNT and yyr yr1; . . . ;yrNT, respectively. For the direct transmission from theprimary user to the secondary users, as the baseline in ourcomparisons, we have

ytn p;txbn ztn; 2and yrn p;rxbn zrn; n 1 . . . ; N; 3

where zzt and zzr denote the zero mean additive white

Gaussian noise terms with variance N0. Also we assume thatall transmitted signals have the same average power, i.e.,IEjxxbj2 Pp, and denote 4 PpN0 as the SNR without fading,pathloss, and shadowing. Therefore, the instantaneous SNRis given by

SNRi;j i;j jhi;jj2: 4Throughout the paper we say that two functions fx andgx are exponentially equal, denoted by fx : gx if

limx1

log fxlog gx 1:

The ordering operators_

and_

! are defined accordingly.

3 COOPERATION PROTOCOLS

We assume that the secondary users are informed of thevacancy of the channels by having the primary userbroadcast a beacon message a priori known to the secondaryusers. The beacon message is a reserved codeword in thecodebook of the primary user and is dedicated to announ-cing the availability of the channel for being accessed by thesecondary users. We intend to devise a cooperation modelsuch that both secondary transmitter and receiver decode thebeacon message with a second-order diversity gain.

Cooperative diversity has been studied and developedextensively as a means for making communication overwireless links more reliable [18], [19], [20], [21]. Incooperative communication, a point-to-point link is assistedby an intermediate node (relay) aiming at providing theintended receiver by additional diversity gains while theoverall resources (frequency bandwidth and power) remainunchanged compared to the noncooperative schemes.

In this paper, we first develop a novel cooperation modelappropriate for multicast transmissions and then show how

to adopt it for building cooperative spectrum accessschemes. Unlike the conventional three node cooperativemodels where the objective is to achieve a second-orderdiversity gain at only the intended receiver, we considerbroadcasting the same message to two receivers such thatboth enjoy second-order diversity gains and yet use the sameamount of resources. This cooperation scheme is a combina-tion of opportunistic relay assignment and bit forwarding.While this protocol, like most other existing cooperationmodels, guarantees performance enhancement in only highenough SNR regimes, we further modify this protocol suchthat it exerts cooperation only if it is beneficial. Thismodified protocol, compared to the noncooperative trans-mission, will ensure performance improvement over all

SNR regimes and all channel realizations.

3.1 Cooperative Spectrum Access (CSA)

Assume that the beacon message consists of K informa-tion bits and N K parity bits giving rise to the codingrate KN and requires N channel uses in a noncooperativetransmission. We utilize the regenerative scheme of [21]and divide the beacon message into the smaller segmentsof lengths K N1 < N and N2 4 N N1, and define thelevel of cooperation as 4 N1N . The transmission of thebeacon message is accomplished in two steps:

1. The primary user uses only the first N1 channel uses(as opposed to noncooperative that uses all the

N channel uses) to broadcast a reserved codebook inits codebook (xx0b) when it is willing to release thechannel. This reserved codeword is a weaker code-word compared to the original beacon. Meanwhile,the secondary transmitter and receiver try to decodethe first segment.

2. During the remaining N N1 channel uses, anysecondary user who has successfully decoded thefirst segment, immediately constructs N2 additionalextra bits and forwards them to the other secondaryuser. If both happen to decode successfully, bothwill try to transmit in the second phase and theirtransmissions collide. Such collision is insignificant

and incurs no loss as it only happens when both

TAJER AND WANG: BEACON-ASSISTED SPECTRUM ACCESS WITH COOPERATIVE COGNITIVE TRANSMITTER AND RECEIVER 113



3/15

secondary users have already successfully decodedthe beacon. If neither of the secondary users issuccessful in decoding the first segment of thebeacon message, there will be no transmission in thesecond phase and hence 1 portion of the timeslot is wasted. Despite such possible waste, asanalyses reveal, by cooperation a better overall

performance can be achieved. As shown in Section 4,the combination of opportunistic relay assignmentand bit forwarding uses essentially the same amountof resources (power and bandwidth) as in thenoncooperative approach, and yet provides a sec-ond-order diversity gain for both secondary users,whereas if one of the nodes is selected a priori andindependently of the channel conditions, this nodewill achieve only a first-order diversity gain.

It is also noteworthy that the CSA protocol can be extended

beyond the scope of cognitive networks and can be adopted

in any application in which a message is intended to be

multicast to several nodes.

3.2 Opportunistic Cooperative Spectrum Access(OCSA)

Attaining a higher diversity order, which is a high SNRmeasure, regime does not necessarily guarantee perfor-mance enhancement in the low or even moderate SNRregimes. For instance, when either the source-relay or the

relay-destination link is very weak, cooperation may not be

beneficial. When exchanging certain information between

the primary and the secondary users is possible, we canovercome this problem by modifying the CSA protocol

such that cooperation is deployed only when deemed

beneficial. Therefore, the protocol becomes opportunistic in

the sense it dynamically decides whether to allow coopera-tion or not. The cost of this improvement is that the primary

and the secondary users need to acquire the quality of theiroutgoing channels. Acquiring such channel gains is not

necessarily feasible in all networks, but those capable of it

can benefit from deploying the OCSA protocol instead of

the CSA protocol. The discussions on how the required

information exchange should take place are provided inSection 6.1. The cooperation is carried out in two steps:

1. The first step is similar to that of the CSA protocol.2. All the three nodes (the primary user and the pair of

secondary transmitter-receiver) step in a competition

for broadcasting additionalpartybits for xx0b during thenext N2 channel uses. Thecompetitionis carried outasfollows: We first assign the metrics tp 4 jp;tj2 jp;rj2,tt 4 jp;tj2 jt;rj2, and tr 4 jp;rj2 jr;tj2 to theprimaryuser, the secondary transmitter, and the secondaryreceiver, respectively. Then the N2 channel uses is

allocated to

a. the secondary transmitter if it successfullydecodes xx0b and tt > maxftp; trg,

b. the secondary receiver if it successfully decodesxx0b and tr > maxftp; ttg, and

c. to theprimary transmitter ifeither tp > maxftt; trgor both secondary users fail to decode xx0

b

.

As shown in the next sections, including the primarytransmitter in the competition has the advantage of guaran-teeing that cooperation is deployed only when it is helpful.

A simple technique for identifying the winner in thesecond phase in a distributed way (without having a centralcontroller) is to equip each primary and secondary userwith a backoff timer, with its initial value set inversely

proportional to its corresponding metric t. Therefore, thetimer of the user with the largest metric t goes off soonerand will start relaying. The idea of encoding the backofftimer by channel gain has also been used in [22], [23].

In multiuser primary networks, the central authority ofthe network responsible for spectrum coordinations shouldbe in charge of transmitting the beacon message. The reasonis that individual users are not fully aware of the networkspectral activity and if they act independently, there existsthe possibility that some user mistakenly announces that aspectrum is vacant while some other primary user isutilizing it. Such strategy will also avoid having multipleprimary users transmit beacon messages concurrently, in

which case these messages collide and are lost.

4 DIVERSITY ANALYSIS

In this section, we characterize the performance of theproposed protocols in terms of the probability of erroneousdetection of the beacon message, which we denote by Pe,at the signal-to-noise ratio . The achievable diversity gainwhich measures how rapidly the error probability decayswith increasing SNR is given by

diversity order lim1

log Pelog

:

4.1 Noncooperative Scheme

As a baseline for performance comparisons, we firstconsider the direct transmission (noncooperative) by theprimary user over the channels given by (2) and (3). Wedenote the events that the secondary transmitter andreceiver do not decode the beacon message successfullyby et and er, respectively. For a coded transmission withcoherent detection, the pairwise error probability (PEP) thatthe secondary transmitter erroneously detects the beaconcodeword cc in favor of the codeword cc for the channelrealization p;t is given by [24, (12.13)]

Pet j p;t 4 Prcc cc j p;t Q ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2djp;tj2q ; 5where d is the Hamming distance between cc and cc.Throughout the diversity order analyses, we will frequentlyuse the following two results:

Remark 1. For a real value a > 0 and the random variableu > 0 with probability density function fUu, we have

IEuQffiffiffiffiffiffiau

p Z1

0

Z1ffiffiffiffiau

p1ffiffiffiffiffiffi2

p ev2=2dv fUu du

Z1

0

Pr u v2

a

1

ffiffiffiffiffiffi2

p ev2 =2 dv:6




4/15

Lemma 1. For integer M > 0 and real ki > 0,Z10

YMi1

1 ekiv2= 1ffiffiffiffiffiffi

2p ev2=2 dv : M: 7

Proof. See Appendix A. tu

As a result, from (5), (6), and (7) we can show that fornoncooperative transmission, we have

PNC et IEp;t Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi

2djp;tj2q !

1;

and PNC er 1.4.2 CSA Protocol

The transmissions are accomplished in N N1 N2 con-secutive channel uses, where N1 uses is for the direct

transmission, and N2 uses for the cooperation phase. We

denote d1 and d2 as the Hamming distances of the codewords

sent in the first and second phase, and we have d1

d2

d.

In the case only one node is successful in the first phase, we

denote it by TR 2 fTt; Trg and also define TR 4 fTt; TrgnTR.Therefore, we get

PC e R j p; R; R; R Qffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffi

2d1jp; Rj2 2d2jR; Rj2q

; 8

and for the case that there is no relaying, we have

PC et j p;t Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi

2d1jp;tj2q

; 9

and PC er j p;r Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi

2d1jp;rj2q

: 10

Theorem 1. For sufficiently large SNR values, we have

PC et: PC er: 12 , PC et< PNC et, and PC er< PNC er.Proof. We denote the success and failure of the secondary

transmitter in the first phase by Tt : S and Tt : F,respectively. We also define Tr : Sand Tr : F similarly.By using (8)-(10) and by expansion over all differentcombinations of the statuses of the secondary users, forthe secondary transmitter we get

PC et PC et j Tt : S; Tr : S

|fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflffl}0PTt : S; Tr : S

PC et j Tt : S; Tr : F|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflffl}0

PTt : S; Tr : F

PC et j Tt : F; Tr : SPTt : F; Tr : S PC et j Tt : F; Tr : F|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflffl}

1

PTt : F; Tr : F

11

Q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2d1jp;tj2 2d2jr;tj2q

Q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2d1jp;tj2q

1 Q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2d1jp;rj2q

Q

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2d1jp;tj2

q

Q

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2d1jp;rj2

q 12

Z1

0

P

d1jp;tj2 d2jr;tj2 v

2

2

|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflffl}

P

d1jp;tj2v22

P

d1jp;rj2v22

ev2=2ffiffiffiffiffiffi

2p dv

Z1

0

P

d1jp;tj2 v

2

2

ev

2=2

ffiffiffiffiffiffi2

p dv

1 Z10

Pd1jp;rj2 v22 ev2 =2ffiffiffiffiffiffi2p dv" #

Z1

0

P

d1jp;tj2 v

2

2

ev

2=2ffiffiffiffiffiffi2

p dv

Z1

0

P

d1jp;rj2 v

2

2

ev

2=2ffiffiffiffiffiffi2

p dv

13

Z1

0

1 e v

2

2d1p;t

1 e v2

2d1p;r ev2=2ffiffiffiffiffiffi

2p dv|fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflffl fflfflfflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflffl fflfflfflffl}

: 12

from Lemma 1 and Remark 1

Z1

0 1 ev2

2d1p;t ev2=2

ffiffiffiffiffiffi2p dv|fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl}: 1 from Lemma 1 and Remark 1

Z1

0

1 e v

2

2d1p;t ev2=2ffiffiffiffiffiffi

2p dv|fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl}

: 1 from Lemma 1 and Remark 1

Z1

0

1 e v

2

2d1p;r ev2=2ffiffiffiffiffiffi

2p dv|fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflffl}

: 1 from Lemma 1 and Remark 1

: 12

13

: 12

;

where the transition from (12) to (13) follows fromRemark 1. Now, by comparing to the noncooperative

case, where we have PNC et

: 1 , we conclude that for

sufficiently large SNR, PC et < PNC et. The same lineof argument holds for the secondary receiver. tuThe CSA protocol ensures that for sufficiently high SNR

values, the probability of detecting unused spectrum holesis boosted. However, the performance in the low SNRregimes depends on the instantaneous channel conditionsand it might happen that noncooperative spectrum accessoutperforms the cooperative scheme.

4.3 OCSA Protocol

Like in the CSA protocol, the first N1 channel uses areallocated for direct transmission from the primary user tothe secondary users. During the remaining time, dependingon the instantaneous channel conditions, the primary usermight transmit the additional parity bits itself or one of thesecondary users might take over such transmission. Wedenote the transmitting user in the second phase byTR 2 fTp; Tt; Trg. Therefore, the transmission in the secondphase is given by

ytn R;txRn ztn; if TR 6 Tt;0; if TR Tt;

&and yrn

R;rxRn zrn; if TR 6 Tr;0; if TR Tr:

&As specified by the protocol, if neither of the secondary

users decodes the beacon message successfully, the primary




5/15

user will transmit the additional parity bits itself. In thiscase, transmission of the beacon message is similar to (2)-(3)for the entire channel uses.

Theorem 2. For all values ofSNR, channel realizations and level

of cooperation, we have POC et < PNC et, POC er tp, which

provides that jr;tj > jp;tj. Therefore, from (14) we getPOC et

IEp;t Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi

2d1jp;tj2 2d2jp;tj2q !

POC TR Tp

IEp;t;r;t Q ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi2d1jp;tj2 2d2jr;tj2q ! POC TR Tr

< IEp;t Q

ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2djp;tj2

q ! POC TR Tp POC TR Tr < PNC et:

15

d1 and d2 are the Hamming distances of the codewords

sent in the first and second phase, and we have

d1 d2 d. By following the same line of argument,we can accordingly show that POC er < PNC er. Inorder to assess the diversity gain, note that

POC et POC et j Tt : FPTt : F POC et j Tt : S|fflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflffl fflfflffl}

0

PTt : S: 16

On the other hand,

POC et j Tt : F POC et j Tt : F; Tr : FPTr : F

POC et j Tt : F; Tr : SPTr : S PTr : F POC et j Tt : F; Tr : S

IEp;tQ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi2d1jp;rj2

q ! IEp;t

Q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2djp;rj2q !

: 1

1: 1

:

17

Recalling that PTt : F : 1 and using (16)-(17) pro-vides that POC et : 12 . A similar argument holds forPOC er. tuNumerical evaluations of the probabilities of erroneous

spectrum hole detection in different schemes are providedin Fig. 1. These error probabilities for the noncooperative,the CSA, and the OCSA protocols are provided in (5), (13),

and (15), respectively. This figure compares the detection

performance of the proposed cooperative schemes with

that of the noncooperative scheme, where it demonstratesthat the CSA and the OCSA protocols achieve a second-order diversity gain. As expected theoretically, at low SNRregimes the CSA protocol might not outperform thenoncooperative scheme, whereas the OCSA protocol out-performs the noncooperative scheme over all SNR re-gimes. In the evaluations above, we have set 0:5 andhave considered channel realizations with parametersp;t; p;r; t;r 1; 2; 3. In Fig. 1, the error probability ofthe user Ti for i 2 ft; rg in the CSA and OCSA protocols isidentified by CSA-Ti and OCSA-Ti, respectively.

4.4 False Alarm

Thus far we have examined the improvements attained indetecting the vacant spectrum holes. In such detectionproblems, however, another aspect that should also betaken into consideration is the false alarm probability. Inour proposed spectrum access models, false alarm is theevent that the secondary users erroneously consider aspectrum band to be idle, i.e., they detect the beaconmessage while the channel is still used by the primary user.This leads to concurrent transmissions by the primary andsecondary users which can potentially harm the transmis-sion of the primary user. Due to stringent constraints onavoiding concurrent transmissions, it is imperative to studythe probability of false alarm as well.

For most detection problems, there exists a tension

between the probability of successful detection and theprobability of false alarm, and it is crucial to maintain agood balance between these two probabilities such thatneither of them is sacrificed in favor of the other one.

In this paper, we have translated a detection probleminto a communication problem rested on the basis ofexchanging a beacon message. Therefore, assessing theprobability of erroneously decoding a nonbeacon messagein favor of the beacon message has the same nature as thatof missing a transmitted beacon and both are quantified interms of decoding errors. More specifically, the probabilitythat for a channel realization p;t, a nonbeacon codeword ccis mistakenly decoded as the beacon message denoted by cc

(which is the false alarm probability), is given by


Fig. 1. Detection error probabilities and diversity gains of noncoopera-tive, CSA, and OCSA schemes.



6/15

P0et j p;t 4 Prcc cc j p;t Q ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

2djp;tj2q

; 18

which is equal to the probability in (5). Therefore, byfollowing the same lines as in Sections 4.2 and 4.3, it is seenthat the proposed cooperation models will also improve theprobability of false alarm which diminishes as fast as theprobability of missing the beacon, enjoying a second-orderdiversity gains.

As a result, deploying the proposed cooperation schemesnot only incurs no loss in the probability of false alarm, butalso improves it. Hence, by translating the detectionproblem into a communication problem, the tensionbetween the two aforementioned probabilities is resolvedand they can be improved simultaneously.

5 CAPACITY ANALYSIS

In this section, we analyze the effect of erroneous detectionof unused channel and thereof missing communicationopportunities on the capacity of the secondary users and

will show how exploiting the proposed cooperative diver-sity protocols enhances the capacity. In the OCSA protocol,each user requires to know the gains of its outgoing channelsto the other users. Acquiring such channel gains requiresinformation exchange over the wireless channel. Therefore,the acquired channel gain estimates are not perfect inpractice, which can affect the capacity. In this section, weassume that all channel estimates for the OCSA protocol areperfect and defer analyzing the effect of estimationinaccuracy on the channel capacity to Section 6.2.

For fading channels, there is always a nonzero prob-ability that a target rate R cannot be supported and there isno meaningful notion of capacity as a rate of arbitrarily

reliable communication. Therefore, we resort to looking intothe notions of outage capacity for slow fading channels andergodic capacity for fast fading channels.

Capacity of the secondary link is influenced by thespectral activity of the primary users and the efficiency ofthe secondary users in detecting the unused channels. In anearlier study in [14], it is also demonstrated how thesecondary user capacity is affected by dissimilar perceptionof secondary users of primary users spectral activity intheir vicinities, where it has been shown that morecorrelated perceptions lead to higher channel capacity forthe secondary link. We will show that the cooperativeprotocols are also effective in increasing such correlation.Lower and upper bounds on the capacity of the secondary

channel, when the received power at the secondary receiveris P, are given by [14]

CUP PrSt Sr 1 log

1 PPrSt Sr 1

19

and

CLP PrSt Sr 1 log

1 PPrSt 1

1

Tc: 20

where the states St; Sr 2 f0; 1g, assigned to the secondarytransmitter and receiver,respectively, indicate the perceptionof the secondary users about the activity of the primary user.

St 0 and St 1 mean that the secondary transmitter has

sensed thechannel to be busy and idle, respectively, and Sr is

defined accordingly. These state variables retain their states

for a period of Tc channel uses and vary to an i.i.d. state

afterwardsandasseenabove, Tc onlyaffectsthelowerbound.For further analysis we assume that all channels (i.e.,

hp;t; hp;r; ht;r) follow the same fading model and therefore

the states St and Sr have the same time variations as the

secondary channel ht;r, which means that all remainunchanged for Tc channel uses and change to indepen-

dent states afterwards. Now, corresponding to different

values of Tc we will have slow and fast fading processes

and need to look into meaningful notions of capacity foreach of them.

1. Fast fading: Small values of Tc correspond to fastfading for which a meaningful notion of capacity isgiven by ergodic capacity and is obtained byaveraging over all channel fluctuations

CUerg 4 IEt;r CUt;r; CLerg 4 IEt;r CLt;r:21

2. Slow fading: Corresponding to large values ofTc ) 1,we consider the -outage capacity C as t heperformance measure for which the bounds aregiven by

Pr

CULt;r < CUL : 22

As the ergodic and outage capacities provided in (21)-(22)

depend on the probability terms PrSt 1 and PrSt Sr 1, we need to assess these terms and their variationsunder different cooperation models. The states St and Srdepend not only on the spectral activity in the vicinity of the

secondary transmitter and receiver, but also on how

successfully these users sense this spectral activity.We introduce the random variables t; r 2 f0; 1g to

account for modeling the spectral activities. t 1 (r 1)states that no primary user is using the channel in the

vicinity of the secondary transmitter (receiver) and the

channel can be used by the secondary users. t 0 andr 0 are defined accordingly for busy channels. Based onthe definitions above and those of St and Sr, which indicate

the perception of secondary users of the activity of the

primary user, we have

PrSt 1 Prt 11 Pet; 23

PrSt Sr 1 Prt r 1Pet; er; 24where et and er are the erroneous detection events. Note

that the terms Prt 1 and Prt r 1 for thecooperative and noncooperative schemes are identical andthe effect of cooperation reflects in the terms Pet andPet; er. In the following two lemmas, we show how theCSA and the OCSA protocols improve the probability

PrSt Sr 1.Lemma 2. For sufficiently large SNR, we have PrCSt

Sr 1 ! PrNC

St Sr 1.




7/15

Proof. We first show that for sufficiently large SNR,

PC er j et ! PNC er as follows:PC er j et PC er j et; Tr : S|fflfflfflfflfflfflfflfflfflffl fflfflfflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflffl}

1

PTr : S

PC er j et; Tr : F

|fflfflfflfflfflffl{zfflfflfflfflfflffl}Tt:S;Tr :FPTr : F

PTr : S PC er j Tt : S; Tr : FPTr : F! PNC erPTr : S

PC er j Tt : S; Tr : F|fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl fflfflffl}: 1 1

2

PTr : F

! PNC erPTr : S PNC er|fflfflfflffl{zfflfflfflffl}: 11

PTr : F

PNC er:

25

On the other hand, from Theorem 1 we know that forsufficiently large SNR, PC

et

> PNC

et

, which in

conjunction with (25) provides that POC et; er PC er j etPC et ! PNC erPNC et, which is the desiredresult. tu

Lemma 3. For all values of SNR, we have

PrOCSt Sr 1 !maxfPrCSt Sr 1; PrNCSt Sr 1g:

P r o o f . W e e q ui v al e nt l y s h ow t h at POC et; er !maxfPC et; er; PNC et; erg. By expanding POC et; erover all different combinations of the statuses of the

secondary users (success/failure), we get

POC et; er POC et; er j Tt : S; Tr : S|fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflffl}

1

PTt : S; Tr : S

POC et; er j Tt : S; Tr : F|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflffl}POC er j Tt:S; Tr :F

PTt : S; Tr : F

POC et; er j Tt : F; Tr : S|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflffl}POC et j Tt:F; Tr :S

PTt : F; Tr : S

POC et; er j Tt : F; Tr : F|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl{zfflffl fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflffl}PNC etPNC erPTt : F; Tr : F

PTt : S; Tr : S PTt : S; Tr : F

1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffi

2d1jp;rj2 2d2 maxfp;rj2; jt;rj2gq !

PTt : F; Tr : S

1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffi

2d1jp;tj2 2d2 maxfp;tj2; jt;rj2gq !

PTt : F; Tr : F

1 Q

ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2djp;tj2

q !

1 Q

ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2djp;rj2

q !

:

26

By following similar steps, we can show that

PC et; er PTt : S; Tr : S

PTt : S; Tr : F

1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi

2d1jp;rj2 2d2jt;rj2q !

PTt : F; Tr : S

1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi

2d1jp;tj2 2d2jt;rj2q !

:

27

Comparing (26) and (27) establishes that POC et; er !PC et; er. Also from (26), we get

POC et; er

! PTt : S; Tr : S PTt : F; Tr : F PTt : S; Tr : F 1 Q

ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffi2djp;rj2

q ! PTt : F; Tr : S 1 Q

ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffi2djp;tj2

q !! PTt : S; Tr : S PTt : F; Tr : F

PTt : S; Tr : F

1 Qffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi

2djp;rj2q !

1 Qffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffi

2djp;tj2q !

PTt : F; Tr : S

1

Q ffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi2djp;tj2q ! 1 Q ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2djp;rj2q !

! 1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi

2djp;tj2q !

1 Qffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi

2djp;rj2q !

PNC et; er;which completes the proof. tu

Fig. 2 illustrates numerical evaluations comparing the

term Pet; er for different schemes given in (25)-(27). Asexpected, for very low SNR regimes, CSA exhibits no gain

over noncooperative scheme, while OCSA outperforms both

CSA and noncooperative schemes in all SNR regimes. As

seen in the figure, the CSA protocol achieves considerable


Fig. 2. Comparing Pet; er for different schemes.



8/15

gain in moderate SNR regimes. For the numerical evalua-

tions, we have assumed the same setup as in Fig. 1.

Theorem 3. For large enough SNR values, we have

CU;Cerg > CU;NCerg ; and CL;Cerg > CL;NCerg ;CU;C > CU;NC ; and CL;C > CL;NC :

Proof. See Appendix B. tuTheorem 4. For all SNR regimes, we have

CU;OCerg > max

CU;Cerg ; CU;NCerg

;

CL;OCerg > max

CL;Cerg ; CL;NCerg

;

CU;OC

> maxCU;C ; CU;NC ;and CL;OC > maxCL;C ; CL;NC :

Proof. See Appendix C. tu

Fig. 3 demonstrates numerical evaluations of the outage

capacity achieved under different schemes. For these

evaluations, we have used the capacity lower and upper

bounds given in (22) and have assumed 12

. For

evaluating PrSt Sr 1 given in (24), we have assumedPrt r 1 0:7 in both cooperative and noncoopera-tive schemes. We also have assumed SNR 10 dB, and forthe lower bounds on the outage capacity, we have set

Tc 10. It is observed that outage capacity bound increasesas we allow higher outage probability. Also, since the gap

between the lower and upper bounds for the cooperative

schemes is small, the bounds seem to be tight.

6 DISCUSSIONS

6.1 Information Exchange

In this section, we discuss how the primary and thesecondary users acquire the channel state information thatthey need in the OCSA protocol. The users need to acquirechannel gains, which can be facilitated via employingtraining-based channel estimation schemes and feedback

communication. Such estimations and feedback can be

carried out reliably via a very low-rate communication (bytransmitting few pilot symbols). The information about

channel gains is used only once for initializing the backofftimers, which is during the transition of the channel accessfrom the primary user to the secondary users, and is notneeded afterwards.

It is noteworthy that acquiring such channel state

information is not guaranteed to be feasible in all networks,in which case the secondary users can deploy the CSAprotocol that does not require any information exchange.

In the OCSA protocol, we need the following channelstate information.

1. The secondary transmitter should know jp;tj and thesecondary receiver should know jp;rj: We assume thatthe primary transmitter is periodically transmittingtraining symbols to its designated receiver fortraining purposes. The secondary users can overhearthese pilots due to the broadcast nature of thewireless channel and periodically acquire theirdesired channel states. Obtaining these channel gainsis only possible when the primary users transmittraining symbols to its designated receiver and thesesymbols are known to the secondary users a priori.

2. The secondary users should know jt;rj jr;tj: Thischannel gain can be acquired by having thesecondary users exchange pilot symbols. This ex-change should be performed while the primary useris still active. To avoid harming the communicationof primary users, the pilots can be exchanged usingultrawideband (UWB) communication which allowsthe secondary users to communicate the low-ratepilots well below the noise level of the primary user.As the secondary users do not have any prior

information about the time of the availability of thechannel, they should keep exchanging the pilotsaccording to how static the channel t;r is. Sinceexchanging pilot sequences requires a very low-ratecommunication, the process of channel gain estima-tion can be performed reliably. This level ofinformation exchange can be carried out in allnetworks. We also assume that the secondary usersalso exchange the estimates ofjp;t and jp;rj that theyhave obtained in Step 1.

3. The primary user should know jp;tj and jp;rj: We devisetwo time slots between the time that primary userfinishes the transmission of the beacon and the timethat all users run their backoff timers. During these

two time slots, the secondary users feed backthe value of jp;tj2 jp;rj2, which they have alreadyobtained during Steps 1 and 2. Note that eachsecondary user will have feedback transmission onlyif it has successfully decoded the beacon, otherwiseits dedicated feedback slot will be wasted. Also, theprimary user will not receive any feedback only ifboth secondary users fail to decode the beacon, inwhich case its lack of knowledge about the valuejp;tj2 jp;rj2 does not harm as the secondary userswill not step in the competition phase and anyrandom initializing of the primary userss backofftimer suffices to ensure that the primary user will be

transmitting in the second phase.


Fig. 3. Lower and upper bounds on outage capacity for differentschemes.



9/15


10/15

and, similarly,

IEtp;tt;tr PTR Tp j tr > tp; tt IEtp;tt;tr PTR Tr j tr > tp; tt

< IEtp;tt P~tp > ~tr j tr > tp:33

By noting that ~ti ti wi, where wi $ N0; 2,and denoting the probability density functions ofti by fti, (29)-(33) provide thatIEtp;tt;tr Pwrtp; tt; tr

tp tt j tp > tt

13

IEtp;tt Pwp wt > tt tp j tt > tp

13

IEtp;tr Pwr wp > tp tr j tp > tr

13

IEtp;tr Pwp wr > tr tp j tr > tp

1

3Z1

0 Z1

0

Qjtp ttjffiffiffiffiffiffiffiffi22p ftp dtp ftt dtt

13

Z10

Z10

Qjtp trjffiffiffiffiffiffiffiffi

22p

ftp dtp ftr dtr

13

IEtp;tt;tr Qjtp ttjffiffiffiffiffiffiffiffi

22p

Q jtp trjffiffiffiffiffiffiffiffi

22p

!:

34Now, from (28) and (34), we get

IEt;r ;p;t;p;rCOCt;r ~COCt;r

COCt;r

" #

< IEtp;tt;tr Pwrtp; tt; tr

< 13

IEtp;tt;tr Q jtp ttjffiffiffiffiffiffiffiffi22

p Q jtp trjffiffiffiffiffiffiffiffi22

p !;35

which is the desired result.2. We start by showing that

lim0

log IEtp ;tt Qjtpttjffiffiffiffiffiffi

22p

h ilog

1:

By some simplifications, we get

IEtp;tt Qjtp ttjffiffiffiffiffiffiffiffi

22p

!

IEp;r;t;r Q kp;r

j2

jt;r

j2

jffiffiffiffiffiffiffiffi22p " #;36

where we have assumed that fp;t; p;r; t;rg arei.i.d. distributed as N0; where we havedefined 4 p;t p;r t;r. Therefore, jp;rj2and jt;rj2 are distributed exponentially withmean 2. It can be readily shown that the random

variable X4 kp;rj2 jt;rj2j is also distributedexponentially with mean 2. Therefore, from (36),

we get

IEtp;tt Qjtp ttj

ffiffiffiffiffiffiffiffi22p

! IEX Q X

ffiffiffiffiffiffiffiffi22p !

: 37

On the other hand, we know that 8x > 01ffiffiffiffiffiffi2

px

1 1x2

ex

2=2 Qx 1ffiffiffiffiffiffi2

px

ex2 =2;

which provides

limx1

log 1ffiffiffiffi2p x 1 1x2 ex2=2 log ex2=2|fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflffl}

1

limx1

log Qxlog ex2 =2

limx1

log 1ffiffiffiffi2

px

ex2=2 log ex2 =2|fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl fflffl}

1

;

or equivalently, for asymptotically large values

of x,

Qx : ex2=2:For any given value of X, by setting x X

ffiffiffiffiffiffi22p ,

for asymptotically large values of x (or small

values of )

QXffiffiffiffiffiffiffiffi22

p

: eX2

42 :

Hence, for asymptotically small values of ,

IEX QXffiffiffiffiffiffiffiffi22

p !

: IEX eX2

42

h i 1

2

Z10

ex2 =42x=2dx

1

ffiffiffi

pe

2=42

2Q

ffiffiffi2p |fflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflffl}: 1

: 1

;

which, in conjunction with (37), gives rise to

IEtp ;tt Qjtp ttjffiffiffiffiffiffiffiffi

22p

! IEX Q Xffiffiffiffiffiffiffiffi

22p !

: 1

: 38

Similarly, we can show that

IEtp ;tt Qjtp trjffiffiffiffiffiffiffiffi

22p

!: 1

: 39

Hence,

IEtp;tt;tr Q jtp ttjffiffiffiffiffiffiffiffi22

p Q jtp trjffiffiffiffiffiffiffiffi22

p !: 1

: 40

Equations (35) and (40) provide that

lim0

log IEt;r ;p;t;p;rCOCt;r ~COCt;r

COCt;rh ilog

< lim0

log IEtp;tt;tr Qjtptt jffiffiffiffiffiffi

22p

Q jtptr jffiffiffiffiffiffi

22p

h i3log

13

;

which is the desired result. tuThe numerical evaluation of IEtp;tt;tr Qjtpttj

ffiffiffiffiffiffi22

p Qjtptrj

ffiffiffiffiffiffi22

p

versus the variance of noise estimation (

2

is depicted in




11/15

Fig. 4 for three different SNR values. We have considered

the setup p;t p;r t;r 1, where we can see that whenthe noise variance is 2 0:01, the capacity loss will be lessthan 1 percent at 0 dB, which is a negligible loss. Also, itis observed that for any fixed value of 2, increasing the

SNR results in less capacity loss. This is justified by noting

that for fixed noise level, more powerful signals are less

prone to estimation errors.

6.3 Throughput Analysis

Channel capacity, being an intrinsic characteristic of thewireless channel, is not influenced by how effectively thechannel is utilized or by how much the MAC-levelcoordinations (e.g., backoff timers and information exchange

in the OCSA protocol) cost. Therefore, in order to furnish afair comparison between the OCSA protocol and thenoncooperative scheme and to incorporate the MAC-levelcosts of this protocol, we assess the achievable throughputs.

We have developed a throughput analysis, which rests onthe basis of our capacity analysis, to account for two types ofthroughput losses incurred by the OCSA protocol. One lossis due to the time waste imposed by the backoff timers, andthe other loss is due to the required feedback from thesecondary users to the primary user.

As running the backoff timers and exchanging informa-tion take place only one time, if the secondary users accessthe channel for a period sufficiently larger than those of thebackoff timers and the feedback communications, the

throughput will be very close to the capacity and thesethroughput losses will have a negligible effect.

In order to formulate the throughput, which we denoteby R, we define TCR as the duration that the secondary linkwill use the channel and we denote the time dedicated tothe feedback communication by TFB. Finally, we denote theinitial value of the backoff timer of user Ti by T

iBT, which is

set as

TiBT 4

ti; for i 2 fp; t; rg;

where is a constant and its unit depends on the unit of ti.Since ftig are scalars, has the units of time. Therefore, forany channel realization fp;t; p;r; t;rg, when the relay is

user Ti, the time delay due to backoff timers is TiBT and the

throughput is found as

Rp;t; p;r; t;r TCRTCR TFB TiBT

~COCt;r

TCRTCR

TFB

ti

~COCt;r;41

where we have taken into account that there is a one-timefeedback transmission and backoff timer activation.Equation (41) suggests that when the duration that thesecondary users access the channel is considerably longerthan the time required for feedback, i.e., TCR ) TFB, thethroughput loss due to feedback will be negligible.However, the same argument does not apply to the effectof backoff timers as for very weak channels (small ti),

ti

can become a nonnegligible factor compared to TCR. In thefollowing theorem, we assess the average throughput lossover the possible channel realizations.

Theorem 6. The average throughput loss due to information

exchange and backoff timers is upper bounded by

IEt;r ;p;t;p;r

~COCt;r Rp;t; p;r; t;r~COCt;r

" # w1

1 w1 1 w1 ew2

1w1 1

;

where

w1 4 TFBTCR

and w2 4 TCRp;t

:

Proof. First note that for the node Ti selected as the relay,

ti !tp jt;pj2jt;rj2 > jt;pj2. Therefore, from (41), we getRp;t; p;r; t;r

~COCt;r! 1

1 w1 w2 1jhp;tj2:

Therefore,

IEt;r ;p;t;p;rRp;t; p;r; t;r

~COCt;r

" #

! IEp;tTCR

TCR TFB jp;tj2

24 35 IEhp;t

1

1

w1

w2

1

jhp;tj2

24

35

1 w1et w2e

w21w1 Ei w21w1t

1w1

1 w12

24 35t1t0

11 w1 w2e

w21w1

Z1w2

1w1

et

tdt

! 11 w1 w2e

w21w1

1 ew2

1w1 1 w1

w2

11 w1 1 w1

e

w21w1 1;

which establishes the desired result. tuFrom the above result, it is consequently concluded that

for the appropriate choice of , i.e., by setting


Fig. 4. Relative capacity loss versus channel estimation noise accuracy.



12/15

( TCRp;t w2 ( 1;we can make the effect of the backoff timers very negligiblewhich provides that w2 0 and

IEt;r ;p;t;p;r

~COCt;r Rp;t; p;r; t;r~COCt;r

" # w1

1 w1 :

As for the effect of feedback, there will be a loss which isupper bounded by TCRTCRTFB . This loss will be also marginal ifthe period that the cognitive link is active is sufficientlylarger than the small period of the feedback communica-tion, i.e., TCR ) TFB.

Fig. 5 depicts the achievable throughput by the OCSAunder different assumptions on feedback loads andbackoff timer settings and compares them to the through-

put achievable via the noncooperative scheme and theCSA protocol. The solid curves represent the lowerbounds on the capacity found in Section 5. For thenoncooperative and the CSA schemes, due to having noMAC-layer bandwidth loss, the throughput is essentiallyequivalent to the capacity, whereas for the OCSA protocol,the throughput becomes equal to the capacity when we setw1 w2 0. The dashed curves illustrate the throughputof the OCSA protocol for the different choices of w1; w2.We have considered a similar setup as in the numericalevaluations of Fig. 3. The results provide that for thechoice of w1 < 0:15 and w2 < 0:15, the throughput of theOCSA scheme is higher than that of the CSA scheme, and

as w1 and w2 exceed these levels, the throughput of theOCSA protocol falls below that of the CSA protocol. Notethat in practical scenarios, it is reasonable to assume thatw1; w2 < 0:15 as w1 TFBTCR , which is the ratio of the lengthof the one-time feedback packet to the length of theinformation packets and is close to zero. Also, w2 dependson the constant factor of the timers which can be chosensuch that keeps the value of w2 arbitrarily small.

7 MULTIUSER NETWORK

7.1 Multiuser CSA Protocol (MU-CSA)

The cooperation protocols proposed in Section 3 and the

subsequent analyses in Section 4 consider a cognitive

network with a single pair of secondary transmitter and

receiver. In this section, we provide a direction for

generalizing the CSA protocol to a cognitive network

consisting of multiple secondary transmitter-receiver pairs.

The objective is to provide all secondary users with the

diversity gain 2M for decoding the beacon message.

We consider a multiuser network consisting of M pairsof secondary transmitters and receivers denoted by

T1t ; T1r ; . . . ; TMt ; TMr . The physical channels between theprimary user and the mth secondary transmitter and

receiver are represented by mp;t and m

p;r, respectively, and

mt;r denotes the channel between Tmt ; Tmr for m 1; . . . ; M.As in Section 2 for the channel realizations, we have

mi;j ffiffiffiffiffiffiffi

mi;j

qhmi;j; 42

where hmi;j, the fading coefficients, are independent complex

Gaussian CN 0; 1 random variables and mi;j representspathloss and shadowing effects. In the CSA protocol, a

secondary user acts as relay based on its success in

decoding the first segment of the beacon. The MU-CSA

protocol consists of two steps:

1. The primary user broadcasts the first segment of the

beacon message during the initial 0 < < 1 portion

of the time slot as in the CSA and OCSA protocols

and all the secondary transmitters and receivers

listen to the message and at the end of the

transmission try to decode the message.2. During the remaining 1 portion of the time slot,

all secondary users who have successfully decoded

the first segment of the beacon construct the addi-tional parity bits and broadcast them. Note that there

is the possibility that multiple secondary users have

been successful in the first phase. Since all these users

broadcast identical information packets, their concur-

rent transmissions are not considered as collision and

rather can be deemed as a distributed multiple-

antenna transmission. For maintaining fairness in

terms of the amount of resources consumed, the

transmission power of individual secondary users

acting as relay should not exceed Pp2M

in order to keep

the aggregate transmission power below Pp. There is

the issue of synchronization between the successful

relays, which they can ensure by using the beacon

they have received as the synchronization reference.

7.2 Diversity Analysis

Next, we show that in a multiusercognitive network with the

MU-CSA protocol, all secondary users enjoy the diversity

gain of2M in detecting the beacon message. We define the

set T f ~T1; . . . ; ~T2Mg fT1t ; . . . ; TMt gSfT1r ; . . . ; TMr g. Also,

we define M 4 f1; ; 2Mg and denote the channel betweenthe secondary users ~Tm and ~Tn by ~m;n, and the channel

between Tp and ~Tm by ~p;m. The probability of missing the

beacon message by the secondary user~

Tm is


Fig. 5. Throughput comparison in different schemes.



13/15

PC e ~Tm X

AMnfmgA6;

PC e ~Tm j 8i 2 A; ~Ti : S; 8j 62 A; ~Tj : F

P8i 2 A; ~Ti : S; 8j 62 A; ~Tj : F PC e ~Tm j 8i; ~Ti : F

|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflffl}1

P8i; ~Ti : F

XAMnfmg

A6;

Q ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffi2d1j~p;mj2 2d2 2MXi2A

j~i;mj2s0@ 1A|fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflfflffl fflfflffl}: 1

jAj1from Lemma 1

Y

j62AQ

ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi2d1j~p;jj2

q |fflfflfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflffl fflfflfflfflfflfflfflfflffl}

: 12MjAj from Lemma 1

Yi2A

1 Q

ffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffi2d1j~p;ij2

q |fflfflfflfflfflfflfflfflfflfflffl fflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}

1 from Lemma 1

!

Yi2f1;...;2Mg Q ffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffi2d1j~p;jj2q |fflfflfflfflfflfflfflfflfflfflfflffl fflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}: 1 from Lemma 1

: 12M1

12M

: 12M

:

Therefore, all the secondary users enjoy a diversity order2M in decoding the beacon message.

The channel capacity can be obtained similarly as inSection 5 for each secondary pair when only that pairaccesses the channel. For finding the channel capacity of aspecific pair of secondary users, we need to find 1) theprobability of detecting the beacon by the secondarytransmitter and 2) the probability that both of secondary

transmitter and receiver are successful in decoding tobeacon. The lower and upper bounds on the channel capacityare given in (19) and (20), and the capacity for the linkbetween Tmt ; Tmr can be found by plugging the values of

PrSTmt 1 PrTmt 1PC eTmt and

PrSTmt STmr 1 PrTmt Tmr 1PC

eTmt ; eTmr

:

Numerical evaluations for the lower and upper bounds onthe secondary user capacity in a cognitive networkconsisting of five pairs of users are shown in Fig. 6, whichdepicts the outage capacity. For this evaluation, we have

used the same setup as that in Section 4.3. By comparingthese results with those of a network with one pair ofsecondary users, it is seen that more gain in terms ofchannel capacity is achievable for the multiuser networkwhich is due to opportunistic selection of the cognitive pair.As the number of secondary users pairs increases, thequality of the best secondary user is expected to becomebetter which justifies the additional gains observed in themultiuser cognitive networks.

8 CONCLUSIONS

In this paper, we have proposed opportunistic cooperation

protocols for supervised spectrum access in cognitive

networks where the primary and secondary users are only

allowed to have orthogonal access to the channel and

cannot coexist concurrently. The primary user broadcasts a

beacon message upon releasing the channel and a pair of

secondary transmitter and receiver cooperatively detect the

beacon message. The cooperation protocols are devised

such that by a single-time relaying both secondary

transmitter and receiver enjoy second-order diversity gains

in detecting the beacon message. We have also quantified

the effect of erroneous detection of vacant channels on the

secondary channel capacity and show the advantage of the

proposed protocols for achieving higher secondary channel

capacity. One of the proposed protocols imposes informa-

tion exchange and bandwidth costs, where we have

assessed the effect of imperfect information exchange on

the capacity and the loss due to the bandwidth cost on theachievable throughput. The cooperation models are also

extended to multiuser cognitive networks as well.

APPENDIX A

PROOF OF LEMMA 1

F ir st w e s ho w t ha t f or t he f un ct io n AM; n 4PMm0

Mm

mn1m, AM; n 0 fo r 8M > n ! 0, a nd

AM; M 6 0. We provide the proof by induction:1. F or n 0 an d 8M > 0, w e h av e AM; 0

PMm0

Mm

1m 1 1M 0.2. Assumption: 8M > n > 0, we have AM; n 0.3. Claim: 8M > n 1, we show that AM; n 1 0.

AM; n 1 XMm0

M

m

mn11m

MXM1m0

M 1m

m 1n1m

MXnk0

n

k

XM1m0

M 1m

mk1m

! M

Xnk0

n

k

AM 1; k

|fflfflfflfflffl{zfflfflfflfflffl}M1>n!k 0:


Fig. 6. Outage capacity in a cognitive network with five pairs of users.



14/15

Now, we show that AM; M 6 0. Again, by induction wehave:

1. For M 1, A1; 1 1.2. Assumption: AM; M 6 0.3. Claim: AM 1; M 1 6 0.

A

M

1; M

1

XM1m0

M 1m

mM11m

M 1XMm0

M

m

m 1M1m

M 1XMk0

M

k

XMm0

M

m

mk1m

!

M 1

AM; M|fflfflfflfflfflffl{zfflfflfflfflfflffl}60

XM1k0

M

k

AM; k|fflfflfflfflffl{zfflfflfflfflffl}

0

!6 0:

Knowing the above result and considering the followingTaylor series expansion

1 kx12 X1

n0

2n!22nn!2 kx

n;

and noting that 1ffiffiffiffi2

pR1

0e

av2

2 dv 12

a12, we getZ1

0

1 ekv2=M 1ffiffiffiffiffiffi

2p ev2=2 dv

XMm0

M

m

1m 1

ffiffiffiffiffiffi2

pZ1

0

ev2

212km=dv

12

XMm0

Mm

1m1 2km1=2

12

XMm0

M

m

1m

X1n0

2n!22nn!2 1

n

2km

n 1

2

X1n0

n2n!

2nn!2 knXMm0

M

m

mn1m

12

X1n0

n2n!

2nn!2 knAM; n

12

X1

nMn

2n!2nn!2 k

nAM; n : M;

therefore,

M :Z1

0

YMi1

1 eminikiv2= 1ffiffiffiffiffiffi

2p ev2 =2 dv

Z1

0

YMi1

1 ekiv2= 1ffiffiffiffiffiffi

2p ev2=2 dv

Z1

0

YMi1

1 eminikiv2 = 1ffiffiffiffiffiffi

2p ev2=2 dv

: M;

which concludes the lemma.

APPENDIX B

PROOF OF THEOREM 3

We show that for any P > 0, CU;CP > CU;NCP andCL;CP > CL;NCP, which consequently proves the theo-rem. First, we remark that p log1 ap is nondecreasing in psince

@

@pCUa @

@pp log

1 a

p

! 0:

The inequality above holds because for the function fu 4log u 1 1u, f1 0 and for u ! 1, @fu@u 1u 1u2 ! 0, andtherefore, fu ! 0 for u ! 1. Therefore, CUP as given in(19) is nondecreasing in the probability PrSt Sr 1. Byusing Lemma 2, it is concluded that for sufficiently large

SNRCU;CP > CU;NCP, which, in turn, establishes theresult for the upper bound. For the lower bound, as shown

in (25), for sufficiently large SNR, PC er j et ! PNC er asshown

PC et; erPNC etPNC er

>PC et

PNC etor, equivalently,

PrCSt Sr 1PrNCSt Sr 1

>PrCSt 1

PrNCSt 1: 43

Therefore, for the lower bound, we get that, for sufficiently

large SNR,

CL;CP PrCSt Sr 1 log

1 PPrC

St

1

1

Tc

> PrCSt Sr 1 log

1 P

PrCStSr1PrNCStSr1 Pr

NCSt 1

1

Tc

44

> PrNCSt Sr 1 log

1 P

PrNCSt 1

1Tc

CL;NCP;

45

where (44) follows from (43) and the transition from where

(44) to (45) follows from the fact that p log1 ap isincreasing in p.

APPENDIX C

PROOF OF THEOREM 4

As shown in Appendix B, CUP is nondecreasing inPrSt Sr 1. By using Lemma 3, it is readily verifiedthat for the function CUP

CU;OCP > maxfCU;CP; CU;NCPg:For the lower bound, we deploy the same approach as in

the proof of Theorem 3 and first show that




15/15

PrOCSt Sr 1PrCSt Sr 1

! PrOCSt 1

PrCSt 1; 46

andPrOCSt Sr 1PrNCSt Sr 1

! PrOCSt 1

PrNCSt 1: 47

The above inequalities can be proven by substituting and

manipulating the expansions of PrCet and PrOCet givenin Theorems 1 and 2 and the expansions of PrCet; er andPrOCet; er in (26) and (27). After establishing the aboveinequalities, the last step is similar to that of Appendix B.

REFERENCES[1] J. Mitola, Cognitive Radio: An Integrated Agent Architecture for

Software Defined Radio, PhD dissertation, KTHRoyal Instituteof Technology, Dec. 2000.

[2] S. Haykin, Cognitive Radio: Brain-Empowered Wireless Com-munications, IEEE J. Selected Areas in Comm., vol. 23, no. 2,pp. 201-220, Feb. 2005.

[3] N. Devroye, P. Mitran, and V. Tarokh, Achievable Rates in

Cognitive Radio Channels, IEEE Trans. Information Theory, vol. 52,no. 2, pp. 1813-1827, May 2006.[4] A. Jovicic and P. Viswanath, Cognitive Radio: An Information-

Theoretic Perspective, Proc. IEEE Intl Symp. Information Theory,pp. 2413-2417, June 2006.

[5] N. Devroye and V. Tarokh, Fundamental Limits of CognitiveRadio Networks, Cognitive Wireless Networks, Springer, 2007.

[6] A. Goldsmith, S.A. Jafar, I. Maric, and S. Srinivasa, BreakingSpectrum Gridlock with Cognitive Radios: An InformationTheoretic Perspective, Proc. IEEE, vol. 97, no. 5, May 2009.

[7] A. Sahai, N. Hoven, and R. Tandra, Some Fundamental Limits onCognitive Radio, Proc. Allerton Conf. Comm., http://www.eecs.

berkeley.edu/~ sahai/Pap ers/cog nitive_radio_p reliminary.pdf ,2004.

[8] D. Slepian, Some Comments on Detection of Gaussian Signalsin Gaussian Noise, IEEE Trans. Information Theory, vol. 4, no. 2,pp. 65-68, June 1958.

[9] N.A.R. Price, Detection Theory, IEEE Trans. Information Theory,vol. 7, no. 3, pp. 135-139, July 1961.[10] R. Etkin, A. Parekh, and D. Tse, Spectrum Sharing for Unlicensed

Bands, IEEE J. Selected Areas in Comm., vol. 25, no. 3, pp. 517-528,Apr. 2007.

[11] Q. Zhao, L. Tong, A. Swami, and Y. Chen, DecentralizedCognitive MAC for Opportunistic Spectrum Access in Ad HocNetworks: A PODMP Framework, IEEE J. Selected Areas inComm., vol. 25, no. 3, pp. 589-599, Apr. 2007.

[12] M. Agarwal and M.L. Honig, Spectrum Sharing on a WidebandFading Channel with Limited Feedback, Proc. IEEE Intl Conf.Cognitive Radio Oriented Wireless Networks & Comm. (CrownCom),pp. 72-76, Aug. 2007.

[13] L. Lai, H. El-Gamal, H. Jiang, and H.V. Poor, CognitiveMedium Access: Exploration, Exploitation and Competition,IEEE/ACM Trans. Networking, submitted.

[14] S.A. Jafar and S. Srinivasa, Capacity Limits of Cognitive Radio

with Distributed and Dynamic Spectral Activity, IEEE J. SelectedAreas in Comm., vol. 25, no. 3, pp. 529-537, Apr. 2007.[15] G. Ganesan and Y. Li, Cooperative Spectrum Sensing in

Cognitive Radio, Part I: Two User Networks, IEEE Trans. WirelessComm., vol. 7, no. 7, pp. 2204-2213, June 2007.

[16] G. Ganesan and Y. Li, Cooperative Spectrum Sensing inCognitive Radio, Part II: Two User Networks, IEEE Trans.Wireless Comm., vol. 7, no. 7, pp. 2214-2222, June 2007.

[17] G. Ganesan, Y. Li, B. Bing, and S. Li, Spatiotemporal Sensingin Cognitive Radio Networks, IEEE J. Selected Areas in Comm.,vol. 26, no. 1, pp. 5-12, Jan. 2008.

[18] A. Sendanoris, E. Erkip, and B. Aazhang, User CooperationDiversity Part I: System Description, IEEE Trans. Comm., vol. 51,no. 11, pp. 1927-1938, Nov. 2003.

[19] A. Sendanoris, E. Erkip, and B. Aazhang, User CooperationDiversity Part II: Implementation Aspects and PerformanceAnalysis, IEEE Trans. Comm., vol. 51, no. 11, pp. 1939-1948,

[20] J.N. Laneman, D.N.C. Tse, and G.W. Wornell, CooperativeDiversity in Wireless Networks: Efficient Protocols and OutageBehavior, IEEE Trans. Information Theory, vol. 50, no. 12, pp. 3062-3080, Dec. 2004.

[21] T. Hunter and A. Nosratinia, Diversity through Coded Coopera-tion, IEEE Trans. Wireless Comm., vol. 5, no. 2, pp. 283-289, Feb.2006.

[22] A. Bletsas, A. Khisti, D.P. Reed, and A. Lippman, A SimpleCooperative Diversity Method Based on Network Path Selection,

IEEE J. Selected Areas in Comm., vol. 24, no. 3, pp. 659-672, Mar.2006.[23] A. Tajer and A. Nosratinia, Opportunistic Cooperation via Relay

Selection with Minimal Information Exchange, Proc. IEEE IntlSymp. Information Theory (ISIT 07), June 2007.

[24] M.K. Simon and M.-S. Alouini, Digital Communication over FadingChannels: A Unified Approach to Performance Analysis. Wiley, 2000.

Ali Tajer received the BSc and MSc degreesin electrical engineering from Sharif Universityof Technology, Tehran, Iran, in 2002 and2004, respectively. Since June 2007, he hasbeen working toward the PhD degree inelectrical engineering at Columbia University,New York. In the summers of 2008 and 2009,he was a research assistant with NECLaboratories America, Princeton, New Jersey.

His research interests lie in the general areasof network information theory, cognitive radios, multiuser communica-tion, relay networks, and statistical signal processing. He is a studentmember of the IEEE.

Xiaodong Wang received the PhD degree inelectrical engineering from Princeton University.He is a professor of electrical engineering atColumbia University in New York. His researchinterests fall in the general areas of computing,signal processing, and communications, and hehas published extensively in these areas.Among his publications is a recent book entitledWireless Communication Systems: AdvancedTechniques for Signal Reception (Prentice Hall,

2003). His current research interests include wireless communications,statistical signal processing, and genomic signal processing. He

received the 1999 US National Science Foundation CAREER Awardand the 2001 IEEE Communications Society and Information TheorySociety Joint Paper Award. He has served as an associate editor for theIEEE Transactions on Communications, the IEEE Transactions onWireless Communications, the IEEE Transactions on Signal Proces-sing, and the IEEE Transactions on Information Theory. He is a fellow ofthe IEEE.

. For more information on this or any other computing topic,please visit our Digital Library at www.computer.org/publications/dlib.


Documents

21. Beacon-Assisted Spectrum Access With Cooperative Cognitive Transmitter and Receiver