1178

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 62, NO. 4, APRIL 2014

Rate Adaptation and Power Allocation for Cognitive

Radio Networks with HARQ-Based Primary System
Moon-Gun Song, Young-Jin Kim, Eun-Yeong Park, and Gi-Hong Im, Senior Member, IEEE

AbstractIn this paper, we propose a protocol for the coexistence of primary and secondary systems over block-fading
channels. In the protocol, the primary system employs a hybrid
automatic repeat request (HARQ). When the primary system
retransmits the data signal, the secondary system serves as a relay
for the retransmission of the primary system and simultaneously
transmits its data signal. To efficiently accomplish the protocol,
we analyze the average throughput of the primary and secondary
systems by using the long-term average throughput (LAT). We
formulate an optimization problem to maximize the LAT of the
secondary system. The constraint of the optimization problem
is that the LAT of the primary system with secondary system
is not less than that of the primary system alone. Through the
optimization problem, we obtain the closed-form solutions of the
transmission rate of the secondary system and the fraction of
the transmit power for relaying the primary systems data signal
and transmitting the secondary systems data signal. Numerical
results show that the primary system does not lose the average
throughput, and rather achieves an additional throughput gain
by adjusting the fraction of the transmit power of the secondary
system.
Index TermsCognitive radio, cooperative communication,
hybrid automatic repeat request, outage probability, long-term
average throughput.

I. I NTRODUCTION

OGNITIVE radio (CR) has attracted considerable attention because of its potential to efficiently utilize scarce
radio spectrum resources [1][7]. In a CR-based network, an
unlicensed system, referred to as a secondary system, can
share the licensed bands dedicated to a primary (licensed)
system. There are three basic operational models used to implement the CR networks: overlay, underlay, and cooperation
models [8]. In the overlay model, the secondary system senses
the temporal spectrum holes in the licensed bands, and then
uses these holes to avoid interference to the primary system.
Although this approach can improve the spectral efficiency,
the improvement depends on the accuracy of spectrum sensing
[2], [3]. In the underlay model, the secondary system coexists
with the primary system and can simultaneously use the
licensed bands while the primary system occupies the bands.
In this case, the secondary system works under the interference

Manuscript received February 18, 2013; revised July 30, 2013 and February
12, 2014. The editor coordinating the review of this paper and approving it
for publication was L. K. Rasmussen.
This work was supported by the Basic Science Research Program through
the National Research Foundation of Korea (NRF), funded by the Ministry
of Science, ICT and Future Planning (No. 2011-0013807).
The authors are with the Department of Electrical Engineering, Pohang
University of Science and Technology (POSTECH), Pohang 790-784, Korea
(e-mail: {knights, yjcom, pey0531, igh}@postech.ac.kr).
Digital Object Identifier 10.1109/TCOMM.2014.021714.130140

constraint of the primary system to preserve the qualityof-service (QoS) requirement of the primary system [5]. In
the cooperation model, the secondary system cooperates with
the primary system and improves the QoS of the primary
system by using cooperative diversity. As compensation for
cooperation, the secondary system can access the licensed
bands of the primary system [6], [7].
In wireless communication systems, the automatic repeat
request (ARQ) and hybrid ARQ (HARQ) have been considered to be principal techniques to provide high reliability to
users. The ARQ is a simple retransmission technique based
on acknowledgment (ACK) for reliability of transmission. The
HARQ is a retransmission technique that employs forwarderror-correction (FEC) code to improve reliability [28]. There
are two widely-used HARQ schemes inducing the buffer to
store and combine the received information: HARQ with
chase combining (HARQ-CC) and HARQ with incremental
redundancy (HARQ-IR) [9]. If a destination fails to decode
the coded data packet of the source, the destination requests
retransmission through a non-acknowledgment (NACK) signal. In the HARQ-CC scheme, the source sends the same
coded data packet for (re)transmissions. The destination stores
the coded data packets for the (re)transmission and combines
them before decoding. In the HARQ-IR scheme, the source
transmits different parity of the coded data packet for each
(re)transmission, and the destination accumulates the parities of the coded data packet during the (re)transmission.
HARQ-IR generally achieves higher performance and data
rates than HARQ-CC [10]. There are several prior studies
related to the outage probability and throughput analysis
of wireless communication systems with HARQ [10][13].
With the quasi-static channel assumption, [11] investigated the
delay-limited throughput and scheduling optimization of the
downlink transmission in a direct link. The authors of [12]
derived a closed-form expression of the outage probability for
the decode-and-forward (DF) cooperative system with HARQIR. On the other hand, utilizing the block-fading channel
assumption, the authors of [10] analyzed the long-term average
throughput (LAT) with a limited number of retransmissions. In
[13], the outage probability of a DF cooperative relay scheme
with HARQ-IR was derived, and the diversity gain caused by
the cooperation and HARQ-IR was described with a high SNR
assumption over Rayleigh fading channels.
Recently, studies have been conducted to investigate the
performance of the CR network with a retransmission-based
or HARQ-based primary system [8], [16][19], [24][27].
The authors of [16] proposed a framework where the secondary system shares the licensed band of the primary system

c 2014 IEEE
0090-6778/14$31.00 

SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM

and ensures a specified target rate for the primary system

by estimating the throughput loss from the ARQ feedback
through an information-theoretic approach. In [17], a transmission policy of the secondary system was proposed to
maximize its throughput under a limited throughput loss of
the retransmission-based primary system. In [8], under both
the peak and the average transmit power constraints in the
secondary system, the channel average rate of the secondary
system was analyzed with a constraint on the interference
power to the HARQ-based primary system when the channel state information (CSI) of the interference channel was
perfectly or imperfectly available at the secondary system.
The authors of [24] analyzed the throughput of the secondary
system with repetition time diversity and HARQ-IR protocols
under the interference constraint to the primary system where
the secondary system is provided with perfect CSI or imperfect
CSI. The authors of [25] solved the optimization problem
of joint power allocation and relay selection for a clusterbased multi-hop CR network to minimize the total power
consumption. The authors of [18] proposed a cooperate-andaccess spectrum-sharing protocol. In this protocol, if the
secondary system detects the NACK signal from the primary
destination (PD), the secondary source (SS) serves as a relay
for the primary system and provides cooperative diversity.
As a reward for cooperative diversity, the secondary system
accesses the spectrum band. The authors of [27] proposed
the backward interference cancellation method in which the
secondary system buffers the outage data signal caused by the
interference from the primary system until the data signal of
the primary system is available at the secondary system and
the interference can be canceled. The authors of [19] proposed
a protocol where the secondary system operates as a relay for
the retransmission of the primary system, and simultaneously
uses the spectrum band to transmit its data signal by exploiting
superposition coding. However, in [8], [17], [24], [25] and
[27], the coexistence of the primary and secondary systems
causes a throughput loss in the primary system. In [18], the
secondary system corrupts the ACK signal of the PD, which
causes a latency increase in the primary system. Furthermore,
the abovementioned studies considered quasi-static channels
that are not changed during the retransmission process. If the
channels suffer from block-fading, the quasi-static channel
models are not available and instantaneous CSI is not permitted for the SS. The authors of [26] proposed the spectrumleasing strategy with the fading temporal correlation and the
retransmission by exploiting locally available information.
However, the primary system should recognize the existence
of the secondary system and adjust the frame structure for
cooperation.
In this paper, we consider the coexistence of the HARQbased primary and secondary systems over block-fading channels. Because the SS cannot obtain the instantaneous CSI, we
propose the coexistence protocol based on Alamouti coding
[20] rather than the superposition coding of [19]. In the
protocol, if the PD fails to decode the data signal of the
primary source (PS), the SS transmits both its data signal and
the data signal of the PS, which was decoded in the initial
transmission of the PS. We also consider the optimization
problem to find the optimal transmission rate and the fraction

1179

of the transmit power for the PD and the secondary destination

(SD) to maximize the average throughput of the secondary
system, subject to the constraint that the average throughput of
the primary system with the secondary system is not less than
that of the primary system alone. To solve the optimization
problem, we asymptotically derive the outage probabilities of
the primary and secondary systems using only the statistics
of the channels. On the basis of the outage probabilities, the
average throughput of the primary and secondary systems is
analyzed through their LATs, which are widely employed to
measure the throughput performance of a system with HARQ
[10], [22]. From the LATs of the primary and secondary
systems, the closed-form solutions are derived with a simple
approximation approach introduced in [23]. By adjusting the
fraction of the transmit power, the secondary system can
provide additional throughput gain to the primary system.
A. Related Works
In [30], the secondary system coexists with the ARQbased primary system. The authors considered a block-fading
scenario, but each fading block contains more than two
transmission time slots. The secondary system overhears the
ACK/NACK signals of the primary to evaluate the channel
quality of the primary system. As the channel quality is
high, the secondary system have more opportunity to use the
spectrum resource. Although the secondary system takes into
account the channel quality of the primary system, the interference from the secondary system affects the performance of
the primary system. In [19], if the initial transmission of the
primary system fails and the secondary system successfully
decodes the data of the primary system, the secondary system
uses the spectrum resource of the primary system. Both the
underlay transmission scheme without CSI at the transmitter
(CSIT) and the overlay transmission scheme with CSIT are
proposed by using superposition coding in quasi-static fading
channels where the data signal experiences the same channel
gain during (re)transmissions. The authors of [31] extended
the work of [19], considering multiple relays for the secondary
system. The outage probability and the average throughput of
the primary and secondary systems are analyzed in quasi-static
fading channels. Due to the interference, the primary system
accumulates different mutual information for retransmission
whether the secondary system and its relays are active or not.
Hence, the average throughput of the primary system with
secondary system is lower than that of the primary system
without secondary system. We extend the overlay scheme of
[19] for block-fading channels. In block-fading channels, the
channel gain is static for each data block, but it independently
changes for retransmission of the data block. We also employ
Alamouti coding rather than superposition coding, and thus
the proposed scheme does not require CSIT. The authors
of [29] proposed the overlay transmission scheme where the
primary system without retransmission process coexists with
the secondary system employing two transmit antennas to
exploit Alamouti coding. The protocol consists of two phases,
and all channels are static during the two phases. In the phase
1, the secondary system transmits its own data through one
antenna and receives the data of the primary system through

1180

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 62, NO. 4, APRIL 2014

SS

SS

hp s s s ,1

PS

hps pd ,1

PD

hpss d ,1

SD

hpspd ,2

PS

hsssd ,2

hss pd ,2

PD

hpssd ,2

SD

Fig. 1. Protocol of the cognitive radio network with HARQ-based primary

system.

the other antenna. In the phase 2, if the secondary system

successfully decodes the data of the primary system for the
phase 1, the secondary system transmits the Alamouti pair of
its own data and the data of the primary system, i.e., Alamouti
coding is applied at block level. In our scheme, Alamouti
coding is applied at symbol level by the data of the primary
system during initial transmission and retransmission. We
exploits Alamouti coding to maintain the average throughput
of the primary system.
B. Organization
The remainder of this paper is organized as follows. In
Section II, we describe the protocol and the system model
of the CR network with the HARQ-based primary system.
In Section III, we derive the LATs of the primary and the
secondary systems. In Section IV, we formulate and solve an
optimization problem based on the LATs to maximize the LAT
of the secondary system. Section V provides the numerical
results and Section VI presents the conclusions of this study.
II. S YSTEM M ODEL AND P ROBLEM S TATEMENT
A. Protocol Description
We consider a cognitive network in which a secondary
system coexists with a HARQ-based primary system over
block-fading channels. The network consists of a pair of PS
and PD and a pair of SS and SD as shown in Fig. 1. The
protocol works as follows.
During initial transmission, the PS transmits a data signal;
the PD, SS and SD receive it.
If the PD fails to decode the data signal, it sends an
NACK signal to report the failure.
After the SS and SD overhear the NACK signal of the PD,
the SD decodes the data signal of the PS for the initial
transmission. If decoding is successful, the SD sends an
ACK signal to the SS.
If the SS successfully decodes the data signal, it is ready
to cooperate with the primary system.
During retransmission, while the PS transmits the additional parity of the data signal, the SS transmits the
Alamouti coded parity of the PS data signal and its own
data signal simultaneously.
In the protocol, we assume that the SS perfectly estimates the
statistics of the channels of the SS-PD, SS-SD and PS-PD
links from the reporting channels. We also assume that all
nodes perfectly decode the ACK and NACK signals and that
the ACK/NACK feedback delay is ignored. There are a variety
of channel estimation techniques [14], [15], and we assume

that all the receivers perfectly estimate the instantaneous CSIs

of the channels corresponding to them. We also assume that
the reporting signals are always successfully decoded.
B. Transmitted and Received Signals
All the links among the PS, PD, SS and SD are represented
as normalized Rayleigh channels with block-fading, i.e., the
channel gains in the initial transmission and retransmission
are independent. Let hab,l be the channel gain of the AB link in the lth (re)transmission. The probability density
function (PDF) and cumulative density function of |hab,l |2 are
respectively given by
P|hab,l |2 (t) = et , F|hab,l |2 (t) = 1 et .

(1)

The PS encodes the information bits to codeword vector x

with a coding rate based on the feedback channel quality. We
assume that a continuous modulation scheme and a capacity
achieving channel coding are used. Also, xp,l is the lth part of
codeword x for the lth (re)transmission, and the total number
of (re)transmissions of the PS is limited to 2. When the PS
transmits data signal vector xp,1 for the initial transmission,
the PD and the pair of the SS and SD receive the signal
vectors. The received signal vectors of the PD, SS and SD
are respectively given by

Pp hps pd ,1 xp,1 + npd ,1 ,
(2)
ypd ,1 = Pploss
s pd

yss ,1 = Pploss
Pp hps ss ,1 xp,1 + nss ,1 ,
(3)
s ss

Pp hps sd ,1 xp,1 + nsd ,1 ,
(4)
ysd ,1 = Pploss
s sd
,
where Pp denotes the transmit power of the PS, and Pploss
s pd
loss
and
P
are
the
path
losses
of
the
PS-PD,
PS-SS
and
Pploss
ps sd
s ss
PS-SD links, respectively. Also, npd ,1 , nss ,1 and nsd ,1 are the
complex additive white Gaussian noise (AWGN) vectors at the
PD, SS and SD, respectively, and their elements are distributed
as CN (0, 1). If the PD fails to decode the received signals,
the PS transmits another part of codeword x.
In the retransmission process, the SS can cooperate with
the PS when both the SS and SD successfully decode the data
signal of the PS. The transmit signal of the SS is given by

xp,2 + xs ,
(5)
xs,2 = 1
p,2 is
where xs is the data signal for the SD, and x
p,2 =
the Alamouti code pair of xp,2 , defined as x
1
4
3
i
[x2
p,2 , xp,2 , xp,2 , xp,2 , ], where xp,2 is the ith element
of xp,2 and () is the complex conjugate operator [20], [21].
We assume perfect synchronization between the primary and
secondary system for the distributed Alamouti code. We also
assume that the PD recognizes the activity of the SS to decode
the data signal applied by the distributed Alamouti code. And,
is the fraction of the transmit power allocated to the data
signal for the SD. The received signal vector of the PD with
cooperation in the retransmission is expressed as


P
h
x
+
Psloss
Ps hss pd ,2 xs,2 +npd ,2 , (6)
ypd ,2 = Pploss
p
p
p
,2
p,2
p
s pd
s d
s d
is the path loss of the SS-PD link, and npd ,2 is the
where Psloss
s pd
complex AWGN vector with CN (0, 1) in the retransmission.

SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM

1181


 
b
 p

 P Ips pd ,1 Rp

T P Ips pd ,1 Rp +2T P Ips pd ,1 <Rp
 
 



+ 1 P Ips ss ,1 Rp P Ips sd ,1 Rp P Ips pd ,1 < Rp Ips pd ,1 + Ips pd ,2


 
 
+P Ips ss ,1 Rp P Ips sd ,1 Rp P Ips pd ,1 < Rp Ips pd ,1 + Ips ss pd ,2






Rp
out
out
out
out
out
=
1P
(R
)
+
1

P
(R
)
1P
(R
)
P
(R
)
P
(,
R
)
.
p
p
p
p
p
p
,2
s
,1
s
,1
p
,2
p
,2
p
p
p
p
p
s d
s s
s d
s d
s d
1 + Ppout
(Rp )
s pd ,1
(13)

CP (, Rp ) =

The SD also receives the data signal and interference from the
SS and PS. The received signal vector of the SD is given by


Ps hss sd ,2 xs,2 + Pploss
Pp hps sd ,2 xp,2 +nsd ,2 , (7)
ysd ,2 = Psloss
s sd
s sd
is the path loss of the SS-SD link in the retransmiswhere Psloss
s sd
sion, and npd ,2 is the complex AWGN vector with CN (0, 1)
at the SD.
If the SS or SD fails to decode the data signal of the PS,
the SS cannot participate in the cooperation and the received
signal vector of the PD in the retransmission is given by

Pp hps pd ,2 xp,2 + npd ,2 .
(8)
ypd ,2 = Pploss
s pd
C. Problem Statement
To efficiently design the protocol, we find the optimal
and transmission rate Rs of the SS to maximize the average
throughput of the secondary system. Additionally, the average
throughput of the primary system with the secondary system
should be no less than that of the primary system alone.
III. L ONG - TERM AVERAGE T HROUGHPUT
To obtain the optimal and the transmission rate of the SS,
we should analyze the outage probability and the throughput
performance of the primary and secondary systems. LAT is
E[b]
defined as CLAT  E[T
] , where b is the number of successfully
decoded information bits and T is the number of channels
used. In this section, we derive the LAT of the primary and
secondary systems based on their outage probabilities.
A. Long-term Average Throughput of the Primary System
For comparison, we derive the LAT of the primary system
when the secondary system is absent. In the (re)transmission,
the mutual information is accumulated on the basis of the
HARQ-IR at the PD. The LAT of the primary system is given
by
C P,Conv (Rp )
 



bp P Ips pd ,1 Rp +P Ips pd ,1 <Rp Ips pd ,1 + Ips pd ,2




=
T P Ips pd ,1 Rp +2T P Ips pd ,1 <Rp


(R
)
Rp 1 Ppout
p
p
,2
s d
,
(9)
=
1 + Ppout
(Rp )
p
s d ,1
where Ips pd ,l is the mutual information of the PS-PD link
in the lth (re)transmission, which is defined as Ips pd ,l =
Pp |hps pd ,l |2 ). Also, Rp is the transmission rate
log2 (1 + Pploss
s pd
of the PS. The data signal of the PS in each (re)transmission

contains bp bits corresponding to T channel symbols, thus

b
(Rp ) denotes the outage probability of
Rp = Tp . And, Ppout
s pd ,1
the PS-PD link with respect to transmission rate Rp in the nth
(re)transmission. The outage probability is expressed as
n


out
Ips pd ,l < Rp .
(10)
Pps pd ,n (Rp ) = P
l=1

We now analyze the LAT of the primary system when

the secondary system coexists with the primary system. If
both the SS and the SD succeed in decoding the data signal
in the initial transmission of the PS, the SS transmits xs,2
and mutual information Ips pd ,2 is accumulated at the PD. The
mutual information is expressed as
Ips pd ,2


2

2
Pploss
Pp hps pd ,2  + (1) Psloss
Ps hss pd ,2 
s pd
s pd
= log2 1+
.

2
1+P loss Ps hs p ,2 
ss pd

s d

(11)
Thus, the outage probability of the PD in the retransmission
is given by



.
(12)
Ppout
(,
R
)
=
P
I
+
I
<
R
p
p
p
,1
p
p
,2
p
,2
p
s
d
s d
s d
If the SS or the SD fails to decode the data signal in the
initial transmission of the PS, mutual information Ips pd ,2 is
accumulated at the PD.
In the case of a primary system that coexists with a
secondary system, cooperation is considered when the SS and
the SD successfully decode the data signal of the PS in the
initial transmission. The LAT of the primary system is given
by (13), where Ips ss ,1 and Ips sd ,1 are the mutual information
of the PS-SS and PS-SD links in the initial transmission of
the PS, respectively, which are given by Ips ss ,1 = log2 (1 +
Pp |hps ss ,1 |2 ), and Ips sd ,1 = log2 (1+Pploss
Pp |hps sd ,l |2 ). And,
Pploss
s ss
s sd
out
out
Pps ss ,1 (Rp ) and Pps sd ,1 (Rp ) are the outage probabilities of these
links, which are expressed as




(Rp )=P Ips ss ,1 <Rp , Ppout
(Rp )=P Ips sd ,1 <Rp .
Ppout
s ss ,1
s sd ,1
(14)
B. Long-term Average Throughput of the Secondary System
To derive the LAT of the secondary system, we assume
that the interference from the data signal of the PS is perfectly
canceled at the SD. This assumption is allowed in the proposed
protocol because the SS transmits when the SD succeeds
in decoding the data signal of the PS during the initial

1182

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 62, NO. 4, APRIL 2014

 
 
 
 

bs P Ips pd ,1 < Rp P Ips ss ,1 Rp P Ips sd ,1 Rp P Iss sd ,2 Rs




CS,t (, Rs ) =
T P Ips pd ,1 Rp + 2T P Ips pd ,1 < Rp




out
out
(R
)
1

P
(R
)
1

P
(R
)
1 Psout
(, Rs )
Rs Ppout
p
p
p
p
,1
s
,1
s
,1
p
p
s sd ,2
s d
s s
s d
=
.
1 + Ppout
(Rp )
s pd ,1

(18)

transmission [19]. Hence, the effective received signal vector

of the SD is given by


ysd ,2 = Psloss
Ps hss sd ,2 xs + nsd ,2 .
(15)
s sd

A. Constraint Simplification
With Rp given, constraint (19c) depends only on parameter
. We now simplify the constraint as inequalities of .
The constraint is expressed as

From the effective received signal, we can derive the LAT of

the secondary system. In the case of the secondary system,
because the SS does not always transmit its data signal,
we consider two types of LATs. One is the LAT for the
transmission time of the SS, and the other is the LAT for
the total time duration. LAT CS,u of the secondary system for
the transmission time is simply expressed as




bs P Iss sd ,2 Rs
= Rs 1Psout
(, Rs )
CS,u (, Rs ) =
s sd ,2
T

CP (, Rp ) CP,Conv (Rp )



Rp 1 Ppout
(Rp ) 1 Ppout
(Rp )
s ss ,1
s sd ,1
=
1 + Ppout
(Rp )
s pd ,1


out
out
Pps pd ,2 (Rp ) Pps pd ,2 (, Rp ) 0.
(20)

=Rs e

Rs
2loss 1
Ps s Ps
s d

(16)

where Iss sd ,2 denotes the mutual information of the SS-SD link,

Ps |hss sd ,2 |2 ). The data signal of
and Iss sd ,2 = log2 (1+ Psloss
s sd
the SS for the SD contains bs bits corresponding to T channel
symbols, and Rs = Tbs . The outage probability Psout
(, Rs )
s sd ,2
of the SS-SD link is given by


Psout
(17)
(, Rs ) = P Iss sd ,2 < Rs .
s sd ,2
In the case of the LAT for the total time duration, we consider
the transmission opportunity of the SS. When both the SS and
SD decode the data signal of the PS in the initial transmission,
the SS can participate in the transmission. Thus, LAT CS,t of
the secondary system for the total time duration is given by
(18).
IV. R ATE A DAPTATION AND P OWER A LLOCATION
In the previous section, we analyzed the LATs of the
primary and secondary systems. On the basis of these LATs,
we now formulate the optimization problem to maximize
the LATs of the secondary system subject to the condition
that the LAT of the primary system with the secondary
system is not less than that of the primary system without
the secondary system. Because the primary system is not
responsible for maximizing the throughput of the secondary
system, we assume that Rp is given for the QoS of the primary
system during the optimization. To consider the transmission
opportunity of the secondary system, the objective of the
optimization problem should be based on CS,t . However, with
Rp given, we can discover that CS,t CS,u and the two LATs
have the same optimal and Rs . Thus, for simplicity we
formulate the optimization problem based on CS,u as follows.
arg maxCS,u (, Rs ) ,
,Rs

s.t. Rs 0, 0 1,
CP (, Rp ) CP,Conv (Rp ).

From the Appendix A, (20) is simplified as

l < u ,

(21)

where l and u are given by



Pp
Pploss
1
s pd
l = R C
1 loss
,
Pss pd Ps
2 p p,1




R C
p,1 1
2 p
Pploss
P
p
1
loss
p
P
P
s d
u = R C
1 loss
1 e ps pd p
,
Pss pd Ps
2 p p,1

(22)

(23)

and the expected value Cp,1 of Ips pd ,1 is expressed as

1
 



Pploss
Pp
s pd
e
1
2 Rp

Cp,1 =
E1
E1
ln 2
Pploss
Pp
Pploss
Pp
s pd
s pd

Rp
Rp e

2
Pploss
Pp
s pd

ln 2 ,

(24)

from the Appendix B. The

integral function E1 (x)
 exponential
et
is defined as E1 (x) = x t dt.
B. Solving the Optimization Problem
The optimization problem is rewritten as
( , Rs ) = arg max CS,u (, Rs ) ,

(25a)

s.t. Rs 0, 0 1,
l < u ,

(25b)
(25c)

,Rs

where and Rs are the optimal solutions of (25). Because

Rp > Cp,1 and Rp 1Cp,1 < 1, l < 1. Hence, the set of s
2
that satisfy (25b) and (25c) is not empty.
The first order partial derivative of CS,u (, Rs ) with respect
to is given by
Rs

2loss 1

2 Rs 1
Ps s Ps
s d
CS,u (, Rs ) = loss
R
e
.
s

Pss sd Ps 2

(26)

(19b)

CS,u (, Rs ) > 0 for all > 0, CS,u (, Rs ) is

Because
maximized when is the upper bound value, i.e., is
expressed as

(19c)

= min (1, u ) .

(19a)

(27)

SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM

The first and second order partial derivatives of CS,u (, Rs )

with respect to Rs are respectively obtained as

2.2

Rs

(28)

Rs
2loss 1
Ps s Ps
s d

2
2Rs ln 2e
CS,u (, Rs ) = 
2
2
Rs
Ps
Psloss
s sd



Rs ln 2eRs ln 2 Psloss
Ps Psloss
Ps (1 + Rs ln 2) .
s sd
s sd
(29)
W (Psloss
P )
s sd s
ln 2

for a given such

There is one solution Rs =

that R
C
(,
R
)
=
0,
where
W
(t)
denotes
the Lambert
S,u
s
s
W function defined as W (t) = x such that t = xex . When
Rs =
as

W (Psloss
P )
s sd s
ln 2

R2s CS,u (, Rs )

Rs =

< 0. Hence, Rs is obtained

Ps )
W (Psloss
s sd
,
ln 2

C. Reward to the Primary System

To provide a reward to the primary system for resource
sharing, the constraint of can be adjusted as
(31)

where r is the reserve parameter for the reward to the primary

system. For the feasibility of (25c), r is limited to
0 r < u l .

1.8
1.6

Conventional
Pr oposed (Pp
Conventional
Pr oposed (Pp
Conventional
Pr oposed (Pp

1.4
1.2

( Pp = 40 dB)
= 40 dB)
(Pp = 30 dB)
= 30 dB)
(Pp = 20 dB)
= 20 dB)

1
0.8

feasible

0.6
0.4

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fig. 2. Average throughput of the primary systems of the conventional and

proposed schemes versus for Pp = 20, 30 and 40 dB when Rp = 2 bpcu.

(30)

for a given .

l < u r ,

Average th roughput (bpcu)

2loss 1
1

CS,u (, Rs ) = loss
e Pss sd Ps
Rs
Pss sd Ps


Psloss
Ps Rs ln 2eRs ln 2 ,
s sd

1183

(32)

When r = 0, the average throughput of the primary system

with the secondary system is almost the same as that of the
primary system without the secondary system. If r increases,
the SS uses more transmit power to relay the data signal of the
primary system, and thus the primary system that coexists with
the secondary system can achieve a higher average throughput
than the primary system alone.
V. N UMERICAL R ESULTS
There exist the primary and secondary systems that consist
of PS, PD, SS, and SD. The number of (re)transmissions of the
PS is limited to two. Simplified path loss models are used, i.e.,
= (d0 /dps pd ) , Pploss
= (d0 /dps ss ) , Pploss
= (d0 /dps sd ) ,
Pploss
s pd
s ss
s sd
loss

loss
Pss sd = (d0 /dss sd ) and Pss pd = (d0 /dss pd ) , where the
reference distance d0 = 10 m and the path loss exponent
= 3. Also, dps pd , dps ss , dps sd , dss sd and dss pd denote the
distances of the PS-PD, PS-SS, PS-SD, SS-SD and SS-PD
links, respectively. The distance of the PS-PD link is 35 m
(dps pd = 35 m). The distances of the SS-SD and SS-PD links
are 10 m (dss sd = dss pd = 10 m), and the SS and SD are
located 26.81 m from the PS (dps ss = dps sd = 26.81 m).
The normalized block-fading Rayleigh channel is considered
for each link as the small-scale channel model. We assume
that the transmit power of the SS is 30 dB (Ps = 30 dB)
and the noise power of each receiver is 1. For comparison,
a genie-aided scheme is considered where the SS knows the

instantaneous CSIs of all the links for the (re)transmission.

For efficient evaluation of performance, throughput is defined
as the transmission rate over the number of (re)transmissions
until the data signal is successfully decoded. If the decoding
of the data signal fails for the retransmission, the throughput
is zero.
We investigate the average throughput for the proposed
protocol where the HARQ-based primary system coexists with
the secondary system. In Fig. 2, we compare the average
throughputs of the primary systems for the proposed scheme
and the conventional scheme where the HARQ-based primary
system alone exists for Pp = 20, 30 and 40 dB when Rp = 2
bpcu. The cross point of the average throughput lines for the
proposed scheme and the conventional scheme is the optimal
of (19), and the region below the cross point is the feasible
region of (19c). As increases, the average throughput of the
primary system for the proposed scheme decreases because the
fraction of the transmit power for the cooperation declines and
the interference to the PD increases. Because the slope of the
average throughput for the proposed scheme is more gradual
as Ps increases, the mismatch between and the optimal
of (19) does not seriously affect the average throughput of the
primary system when Ps is high.
In Fig. 3, we describe solutions and Rs of (25) versus
Pp when Rp = 2 bpcu. In the region of low Ps , and Rs
are almost matched with the optimal and Rs obtained by
an exhaustive search. Although the mismatch exists in the
region of high Ps , the average throughput of the primary and
secondary systems with and Rs is almost the same as
that of the primary and secondary systems with the optimal
and Rs as shown in Fig. 4. Furthermore, and Rs are
obtained from the closed-form solutions of (25) instead of an
exhaustive search. When Rp = 2 bpcu, the average throughput
of the secondary system declines as Pp increases. Note that the
opportunity for the secondary system to access the spectrum
band decreases because the outage probability of the primary
system in the initial transmission decreases, and because
and Rs decreases. Compared to the genie-aided scheme, the
secondary system with and Rs achieves approximately

1184

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 62, NO. 4, APRIL 2014

7
2.2
6.6

0.5

6.2

0.6

0.4

5.8

0.3

Average throughput (bpcu)

- exhaustive
- proposed
Rs - exhaustive
Rs - proposed

Rs (bpcu)

0.7

2
1.8
1.6

Conventional (Rp = 3 bpcu)

Proposed (Rp = 3 bpcu)
Conventional (Rp = 2 bpcu)
Proposed (Rp = 2 bpcu)
Conventional (Rp = 1 bpcu)
Proposed (Rp = 1 bpcu)

1.4
1.2
1

5.4

0.8
0.2
20

22

24

26

28

30

Pp (dB)

32

34

36

38

5
40

Fig. 3. Comparison of the and Rs parameters obtained by the optimization

problem and exhaustive search when Rp = 2 bpcu.

0.6

feasible
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fig. 5. Average throughput of the primary systems of the conventional and

proposed schemes versus for given Rp = 1, 2 and 3 bpcu when Pp = 30
dB.

2
1

1.8

1.4
1.2
1

0.8

7.2

0.6

6.4

0.8
0.6

0.4

5.6

0.2

4.8

Rs (bpcu)

Rs - exhaustive
Rs - proposed

Primary - conv., genie, exhaustive

Primary - proposed ( r = 0)
Primary - proposed ( r = 0.1)
Secondary - genie
Secondary - exhaustive
Secondary - proposed ( r = 0)
Secondary - proposed ( r = 0.1)

Average throughput (bpcu)

1.6

- exhaustive
- proposed

0.4
0.2
0
20

22

24

26

28

30

Pp (dB)

32

34

36

38

40

Fig. 4. Average throughput of the primary and secondary systems versus Pp

when Rp = 2 bpcu.

90% of the average throughput of the genie-aided scheme.

From Fig. 4, we also find out that the secondary system can
provide additional average throughput gains to the primary
system by using r in the region of low Pp when Rp = 2
bpcu.
We now investigate the average throughput of the primary
and secondary systems versus Rp when Pp = 30 dB. Figure
5 shows the average throughput of the primary systems for
the proposed scheme and the conventional scheme for Rp =
1, 2 and 3 bpcu when Pp = 30 dB. The cross point of the
average throughput lines is the optimal of (19), and the
region below the cross point is the feasible region. When Rp
is high, the primary system requires more mutual information
accumulated at the PD in the retransmission to succeed in
decoding. Thus, the mismatch of and the optimal of (19)
causes a decrease in the average throughput of the primary
system when Rp is high.
In Fig. 6, we compare and Rs and the optimal and
Rs obtained by an exhaustive search when Pp = 30 dB. In
the region of low Rp , and Rs are almost matched with
the optimal and Rs . On the other hand, in the region

0.5

1.5

Rp (bpcu)

2.5

3.5

Fig. 6. Comparison of the and Rs parameters obtained by the optimization

problem and exhaustive search when Pp = 30 dB.

of high Rp , and Rs are overestimated because of the

approximation. This causes the loss of the average throughput
of the primary system for the proposed scheme as compared
with the conventional scheme when Rp is high. However,
as shown in Fig. 7, the loss of average throughput can be
mitigated or the primary system can achieve the gain of
average throughput by using r . As r increases, the average
throughput of primary system is enhanced when Pp = 30 dB
and Rp = 4 bpcu. When r > 0.01, the average throughput
of the primary system of the proposed scheme is higher than
that of the primary system of the conventional system. In
Fig. 8, although both and Rs decline as Rs increases, the
average throughput of the secondary system increases. This is
due to the fact that the opportunity for the secondary system
to use the spectrum band increases because of an increase
in the outage probability of the primary system in the initial
transmission. The average throughput of the secondary system
with and Rs is approximately 90% of that of the genieaided scheme when Pp = 30 dB.

SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM

1185

A PPENDIX A
3

(20) can be rewritten as

Ppout
(Rp ) Ppout
(, Rp ) 0.
s pd ,2
s pd ,2

Average throughput (bpcu)

2.5

To simplify (33), we obtain the distributions of Ips pd ,1 , Ips pd ,2

and Ips pd ,2 . From the PDF and cumulative distribution function
(CDF) of |hps pd ,l |2 , the PDFs and CDFs of both Ips pd ,1 and
Ips pd ,2 are given by

Primary - conv., genie, exhaustive

Primary - proposed
Secondary - genie
Secondary - exhaustive
Secondary - proposed

1.5

2loss1
ln 2
PIps pd ,l (t) = loss 2t e Pps pd Pp ,
Pps pd Pp

0.5

FIps pd ,l (t) = 1 e
0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Fig. 7. Average throughput of the primary and secondary systems versus r

when Pp = 30 dB.

Primary - conv., genie, exhaustive

Primary - proposed ( r = 0)
Primary - proposed ( r = 0.1)
Secondary - genie
Secondary - exhaustive
Secondary - proposed ( r = 0)
Secondary - proposed ( r = 0.1)

2.5

Average throughput (bpcu)

(33)

2t 1
Pploss
Pp
s pd

, for l = 1, 2.

(34)

Ps  1, Ips pd ,2 is asymptotically
With the assumption of Psloss
s pd
given by
Ips pd ,2


2

2 
Pploss
Pp hps pd ,2  + (1) Psloss
Ps hss pd ,2 
s pd
s pd
log2 1+

2
Psloss
Ps hss pd ,2 
s pd


2 
loss
1 Pps pd Pp hps pd ,2 
1
+
= log2
.
(35)


Psloss
Ps hs p ,2 2
s pd
s d

This assumption is reasonable when the SS is located close to

the PD. The CDF of Ips pd ,2 is expressed as

2

1


t


h

p p ,2
FIpp ,2 (t) =P Ips pd ,2 < t = P  s d 2 < P loss

s d
hs p ,2 
1 ps pd Pp
s d
loss

1.5

Pss p Ps
d

0.5


=
0

0.5

1.5

Rp (bpcu)

2.5

3.5

Fig. 8. Average throughput of the primary and secondary systems versus

Rp when Pp = 30 dB.


=

VI. C ONCLUSIONS
We proposed a protocol for the coexistence of the HARQbased primary and secondary systems over block-fading channels. In the protocol, if the PD fails to decode the data signal
of the PS in the initial transmission, the SS serves as a
relay for the primary system by using Alamouti coding, and
simultaneously transmits its data signal in the retransmission
of the primary system. We analyzed the average throughput of
the primary and secondary systems through their LATs. We
also obtained the closed-form solutions of the transmission
rate and the fraction of the transmit power of the SS. The
closed-form solutions maximize the LAT of the secondary
system subject to the constraint that the LAT of the primary
system with the secondary system is not less than that of the
primary system alone. By adjusting the fraction of the transmit power, the secondary system can provide the additional
average throughput gain to the primary system.

2t 1

P loss Pp
1 ps pd
P loss Ps
ss pd

=1

1 e

ex ey dxdy

2t 1

P loss Pp
1 ps pd
P loss Ps
ss pd

loss
1 Pps pd Pp
Psloss
Ps 2t
s pd

y
e dy

1

1
1

Pploss
Pp
s pd
Psloss
Ps
sp

(36)

As shown in Fig. 9, the serious mismatch between the distribution obtained by simulation and the derived CDF of Ips pd ,2
exists at the tail part. This is because the domain of FIpp ,2 (t)
t

is limited such that 2 >

Pploss
Pp
1
s pd
(1

Psloss
Ps )
sp

s d

due to the denominator.

Utilizing the PDFs and CDFs of Ips pd ,1 and Ips pd ,2 , the first
term of (33), Ppout
(Rp ), is calculated as
s pd ,2


Ppout
(Rp ) =P Ips pd ,1 + Ips pd ,2 < Rp
s pd ,2
 Rp  Rp x
=
PIps pd ,1 (x) PIps pd ,2 (y) dydx
0
0
 Rp
=
FIps pd ,2 (Rp x) PIps pd ,1 (x) dx
0


Rp x

Rp

=
0

1e

1
Pploss
Pp
s pd

PIps pd ,1 (t) dx.

(37)

1186

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 62, NO. 4, APRIL 2014

where Cp,1 denotes the expected value of Ips pd ,1 .

From the limitation of the domain of FIpp ,2 (t), is
s d
bounded by


Pp
Pploss
1
s pd
1 loss
.
(42)
> R C
Pss pd Ps
2 p p,1

1
0.9
0.8
0.7

CDF

0.6

A PPENDIX B
0.5

The expected value Cp,1 is obtained as

 Rp
 Rp
t
ln 2 t P2plossp1Pp

Cp,1 =
tPIps pd ,1 (t) dt =
t loss 2 e s d dt
Pps pd Pp
0
0

0.4
0.3

Simulation (Pploss
P
s pd p
Approx. (Pploss
P =
s pd p
Simulation (Pploss
P
s pd p
Approx. (Pploss
P =
s pd p
loss
Simulation (Pps pd Pp
Approx. (Pploss
P =
s pd p

0.2
0.1
0
1.5

2.5

Ips pd ,2

Fig. 9. CDF of mutual information

and 40 dB, and Pploss
Pp = 20 dB.
s pd

= 20 dB, Psloss
Ps
s pd
20 dB, Psloss
Ps =
s pd
= 20 dB, Psloss
Ps
s pd
20 dB, Psloss
Ps =
s pd
loss
= 20 dB, Pss pd Ps
20 dB, Psloss
Ps =
s pd

3.5

Ipp ,2
s d

when

= 20 dB)
20 dB)
= 30 dB)
30 dB)
= 40 dB)
40 dB)

4.5

Psloss
P
s pd s


=
5

= 20 dB, 30 dB

R
2 p
Pploss
Pp
s pd
1
Pploss
Pp
s pd

R
2 p
 loss
Pp p Pp


 s d
e

P
x
ex ln Pploss
=
p
p

s d
ln 2
 1
1
Pploss
Pp
s pd

Pploss
Pp
s pd

The second term of (33), Ppout

(, Rp ), is calculated as
s pd ,2
P out
ps pd ,2

loss

e Pps pd Pp x  loss
e ln Pps pd Pp x dx
ln 2

(, Rp )



=P Ips pd ,1 + Ips pd ,2 < Rp
 Rp  Rp x
=
PIps pd ,1 (x) PIpp ,2 (y) dydx
s d
0
0
 Rp
FIpp ,2 (Rp x) PIps pd ,1 (x) dx
=
s d
0

 Rp
loss
1
1 Pps pd Pp

=


1
loss P
Pploss
Pp
p

P
1
s d
0
ss pd s 2Rp x
1 P loss Ps

1
Pploss
Pp
s pd

ln 2
1
loss

e Pps pd Pp
=
ln 2
Rp e

R
2 p
Pploss
Pp
s pd
1
Pploss
Pp
s pd

E1
R
2 p
Pploss
Pp
s pd

1 x
e dx
x

1
Pploss
Pp
s pd



E1

ln 2 ,

2 Rp
Pploss
Pp
s pd

(43)

where E1 () denotes the

integral function, which
 exponential
t
is defined as E1 (x) = x e t dt.

s s pd

PIps pd ,1 (x) dx,

(38)

where PIpp ,2 (y) denotes the PDF of Ips pd ,2 . Hence, (33) is

s d
expressed as
out
P out
ps pd ,2 (Rp ) Pps pd ,2 (, Rp )

 Rp
loss
1
1 Pps pd Pp
=


loss
Pss pd Ps 2Rp x 1 1
0

Pploss
Pp
s pd
Psloss
Ps
sp

Rp x
2 loss 1
Pp p Pp
s d

PIps pd ,1 (x) dx.

(39)

However, it is difficult to obtain the closed-form of (39).

By using an approximation approach introduced in [23], we
replace the random variable Ips pd ,1 with the expected value
over the fading as follows.
(Rp ) Ppout
(, Rp )
Ppout
s pd ,2
s pd ,2

R p C
p,1

loss
2
1 Pps pd Pp
1



e
loss
Psloss
Ps 2Rp Cp,1 1 1 Pps pd Pp
s pd

P loss Ps

Pploss
Pp
s pd

0.

ss pd

(40)
Thus, is bounded by




R C
p,1 1
2 p
Pploss
Pp
1
Pploss
Pp
s pd
p
s d
R C
1 loss
1e
,
Pss pd Ps
2 p p,1

(41)

R EFERENCES
[1] S. Haykin, Cognitive radio: brain-empowered wireless communications, IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 201220, Feb.
2005.
[2] H. J. Lim, D. Y. Seol, and G. H. Im, Joint sensing adaptation and
resource allocation for cognitive radio with imperfect sensing, IEEE
Trans. Commun., vol. 60, no. 4, pp. 10911100, Apr. 2012.
[3] D. Y. Seol, H. J. Lim, and G. H. Im, Optimal threshold adaptation with
radio environment map for cognitive radio networks, in Proc. 2009
IEEE International Symposium on Information Theory, pp. 25272531.
[4] K. M. Kang, J. C. Park, S. I. Cho, B. J. Jeong, Y. J. Kim, H. J. Lim,
and G. H. Im, Deployment and coverage of cognitive radio networks
in TV white space, IEEE Commun. Mag., vol. 50, no. 12, pp. 8894,
Dec. 2012.
[5] L. B. Le and E. Hossain, Resource allocation for spectrum underlay
in cognitive radio networks, IEEE Trans. Wireless Commun., vol. 7,
no. 12, pp. 53065315, Dec. 2008.
[6] Q. Zhang, J. Jia, and J. Zhang, Cooperative relay to improve diversity
in cognitive radio networks, IEEE Commun. Mag., vol. 47, no. 2, pp.
111117, Feb. 2009.
[7] H. J. Lim, D. Y. Seol, and G. H. Im, Power budget allocations for
auctions with multi-bands in cooperation-based spectrum leasing, in
Proc. 2011 IEEE International Conference on Communications.
[8] B. Makki and T. Eriksson, On the average rate of HARQ-based quasistatic spectrum sharing networks, IEEE Trans. Wireless Commun.,
vol. 11, no. 1, pp. 6577, Jan. 2012.
[9] J.-F. Cheng, Coding performance of hybrid ARQ schemes, IEEE
Trans. Commun., vol. 54, no. 6, pp. 10171029, Jun. 2006.
[10] P. Wu and N. Jindal, Performance of hybrid-ARQ in block-fading
channels: a fixed outage probability analysis, IEEE Trans. Commun.,
vol. 58, no. 4, pp. 11291141, Apr. 2010.
[11] R. Narasimhan, Throughput-delay performance of half-duplex hybridARQ relay channels, in Proc. 2008 IEEE International Conference on
Communications, pp. 986990.

SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM

[12] S. Tomasin, M. Levorato, and M. Zorzi, Steady state analysis of coded

cooperative networks with HARQ protocol, IEEE Trans. Commun.,
vol. 57, no. 8, pp. 23912401, Aug. 2009.
[13] S. Lee, W. Su, S. Batalama, and J. Matyjas, Cooperative decode-andforward ARQ relaying: performance analysis and power optimization,
IEEE Trans. Wireless Commun., vol. 9, no. 8, pp. 26322642, Aug.
2010.
[14] M. K. Ozdemir and H. Arslan, Channel estimation for wireless OFDM
systems, IEEE Commun. Surveys Tuts., vol. 9, no. 2, pp. 1848, Second
Quarter 2007.
[15] D. Kim, H. M. Kim, and G. H. Im, Iterative channel estimation with
frequency replacement for SC-FDMA systems, IEEE Trans. Commun.,
vol. 60, no. 7, pp. 18771888, Jul. 2012.
[16] K. Eswaran, M. Gastpar, and K. Ramchandran, Bits through ARQs:
spectrum sharing with a primary packet system, in Proc. 2007 IEEE
International Symposium on Information Theory, pp. 21712175.
[17] M. Levorato, U. Mitra, and M. Zorzi, Cognitive interference management in retransmission-based wireless networks, IEEE Trans. Inf.
Theory, vol. 58, no. 5, pp. 30233046, May 2012.
[18] Q. Li, S. H. Ting, A. Pandharipande, and M. Motani, Cooperate-andaccess spectrum sharing with ARQ-based primary systems, IEEE Trans.
Commun., vol. 60, no. 10, pp. 28612871, Oct. 2012.
[19] R. Tannious and A. Nosratinia, Cognitive radio protocols based on exploiting hybrid ARQ retransmissions, IEEE Trans. Wireless Commun.,
vol. 9, no. 9, pp. 28332841, Sep. 2010.
[20] S. Alamouti, A simple transmit diversity technique for wireless communications, IEEE J. Sel. Areas Commun., vol. 16, no. 8, pp. 14511458,
Oct. 1998.
[21] D. Y. Seol, U. K. Kwon, and G. H. Im, Performance of single carrier
transmission with cooperative diversity over fast fading channels, IEEE
Trans. Commun., vol. 57, no. 9, pp. 27992807, Sep. 2009.
[22] J. Shen, T. Jiang, S. Liu, and Z. Zhang, Maximum channel throughput
via cooperative spectrum sensing in cognitive radio networks, IEEE
Trans. Wireless Commun., vol. 8, no. 10, pp. 51665175, Oct. 2009.
[23] P. Jia, M. Vu, T. Le-Ngoc, S.-C. Hong, and V. Tarokh, Capacity- and
Bayesian-based cognitive sensing with location side information, IEEE
J. Sel. Areas Commun., vol. 29, no. 2, pp. 276289, Feb. 2011.
[24] B. Makki, A. Graell i Amat, and T. Eriksson, HARQ feedback in
spectrum sharing networks, IEEE Commun. Lett., vol. 16, no. 9,
pp. 13371340, Sep. 2012.
[25] P. Zhang, Y. Wang, Z. Feng, R. Li, Z. Wei, and S. Chen, Joint power
allocation and relay selection for multi-hop cognitive network with
ARQ, in Proc. 2012 IEEE International Symposium on Personal Indoor
and Mobile Radio Communications, pp. 12201225.
[26] M. Levorato, O. Simeone, U. Mitra, and M. Zorzi, Cooperation and
coordination in cognitive networks with packet retransmission, in Proc.
2009 IEEE Information Theory Workshop, pp. 495499.
[27] N. Michelusi, P. Popovski, M. Levorato, O. Simeone, and M. Zorzi,
Cognitive transmissions under a primary ARQ process via backward
interference cancellation, in Proc. 2011 Allerton Conference on Communication, Control, and Computing, pp. 727735.
[28] D. Kim, Y. Choi, S. Jin, K. Han, and S. Choi, A MAC/PHY crosslayer design for efficient ARQ protocols, IEEE Commun. Lett., vol. 12,
no. 12, pp. 909911, Dec. 2008.
[29] V. A. Bohara, S. H. Ting, Y. Han, and A. Pandharipande, Interferencefree overlay cognitive radio network based on cooperative space time
coding, in Proc. 2010 International Conference on Cognitive Radio
Oriented Wireless Networks and Communications, pp. 15.
[30] J. C. F. Li, W. Zhang, A. Nosratinia, and J. Yuan, Opportunistic spectrum sharing based on exploiting ARQ retransmission in cognitive radio
networks, in Proc. 2010 IEEE Global Telecommunications Conference.
[31] S. B. Mafra, R. D. Souza, J. L. Rebelatto, E. MG Fernandez, and
H. Alves, Cooperative overlay secondary transmissions exploiting primary retransmissions, EURASIP J. Wireless Commun. and Networking
Jul. 2013.

1187

Moon-Gun Song received the B.S. and M.S.

degrees in electronic and electrical engineering
from Pohang University of Science and Technology
(POSTECH), Gyeongbuk, Korea, in 2008 and 2010,
respectively. He is currently working toward the
Ph.D. degree at Communications Research Laboratory, POSTECH. His current research interests are
dynamic spectrum access, cognitive radio systems,
and MIMO systems.

Young-Jin Kim received the B.S. and Ph.D degrees

in electronic and electrical engineering from Pohang
University of Science and Technology (POSTECH),
Gyeongbuk, Korea, in 2007 and 2013, respectively.
He is currently a senior researcher at Samsung
Electronics Co., Ltd., Suwon, Gyeonggi, Korea.
His current research interests are MIMO systems,
spectrum sharing, and cognitive radio networks.

Eun-Yeong Park received the B.S. degree in electronic and electrical engineering from Pohang University of Science and Technology (POSTECH),
Gyeongbuk, Korea, in 2013. She is currently working toward the Ph.D. degree at the Communications
Research Laboratory, POSTECH. Her current research interests are cognitive radio systems.

Gi-Hong Im (M87-SM94) was with AT&T Bell

Laboratories, Holmdel, NJ, where he was responsible for the design and implementation of highspeed digital transmission systems for loop plant,
local area network, and broadband access applications (1990-1996). He authored or co-authored more
than twenty standards contributions to standards
organizations such as ANSI T1E1.4, ETSI, IEEE
802.9, ANSI X3T9.5, and the ATM Forum. These
contributions led to the adoption of three AT&T
proposals for new standards for high-speed LANs
and broadband access. In 1995, he was appointed as Distinguished Member
of Technical Staff at AT&T Bell Laboratories. Since 1996, he has been with
POSTECH as a professor. From 1996 to 2000, he was a Bell Laboratories
Technical Consultant. From 2002 to 2003, he was a visiting vice president
of Samsung Electronics, where he worked on 4G wireless communication
systems. His current research interests include signal processing and digital
communications with applications to high-speed digital transmission systems.
Dr. Im received the 1996 Leonard G. Abraham Prize Paper Award from
the IEEE Communications Society for the paper entitled Bandwidth-efficient
digital transmission over unshielded twisted pair wiring, published in the
IEEE J OURNAL ON S ELECTED A REAS IN C OMMUNICATIONS, the 2000 LG
Award from LG Electronics, and the 2005 National Scientist Award from the
Korean government. From 2004 to 2010, he served as an editor for the IEEE
T RANSACTIONS ON C OMMUNICATIONS and IEEE C OMMUNICATIONS L ETTERS . From 2010 to 2013, he served as a division editor for the Journal of
Communications and Networks. He has published over 100 papers in IEEE
Journals and Conferences, and has been granted 29 U.S. patents.