Beruflich Dokumente
Kultur Dokumente
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
1179
1180
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
(1)
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
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
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)
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
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)
s.t. Rs 0, 0 1,
CP (, Rp ) CP,Conv (Rp ).
(21)
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)
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)
(25a)
s.t. Rs 0, 0 1,
l < u ,
(25b)
(25c)
,Rs
2loss 1
2 Rs 1
Ps s Ps
s d
CS,u (, Rs ) = loss
R
e
.
s
Pss sd Ps 2
(26)
(19b)
(19c)
= min (1, u ) .
(19a)
(27)
SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM
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
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 =
Ps )
W (Psloss
s sd
,
ln 2
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
(30)
for a given .
l < u r ,
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)
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
1184
7
2.2
6.6
0.5
6.2
0.6
0.4
5.8
0.3
- exhaustive
- proposed
Rs - exhaustive
Rs - proposed
Rs (bpcu)
0.7
2
1.8
1.6
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
0.6
feasible
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
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
1.6
- exhaustive
- proposed
0.4
0.2
0
20
22
24
26
28
30
Pp (dB)
32
34
36
38
40
0.5
1.5
Rp (bpcu)
2.5
3.5
SONG et al.: RATE ADAPTATION AND POWER ALLOCATION FOR COGNITIVE RADIO NETWORKS WITH HARQ-BASED PRIMARY SYSTEM
1185
A PPENDIX A
3
2.5
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
2.5
(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
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
=
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
Pploss
Pp
1
s pd
(1
Psloss
Ps )
sp
s d
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
(37)
1186
1
0.9
0.8
0.7
CDF
0.6
A PPENDIX B
0.5
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
= 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
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)
s s pd
(38)
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
(39)
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
1187
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.