You are on page 1of 5

Relay Selection and Power Allocation in Amplify-and-Forward Cognitive

Radio Systems
Krishna Ram Budhathoki, Mehdi Maleki, and Hamid Reza Bahrami
Department of Electrical and Computer Engineering
The University of Akron, Akron, Ohio, 44325
Emails: {kb80, mm158}@zips.uakron.edu, hrb@uakron.edu
AbstractIn this paper, we consider the problem of joint relay
selection and power allocation at the source and relay the nodes
in a cognitive radio (CR) system in which nodes are allowed to
cooperate with each other. In particular, we consider an amplify-
and-forward (AF) relaying secondary system that operates in the
presence of a primary system. Joint relay selection and power
allocation are aimed to achieve the maximum possible throughput
for the secondary system subject to power constraints at the
secondary source and relay nodes and by considering interference
constraints from the secondary to the primary system. Using
constrained optimization techniques, we derive an optimal joint
power allocation and relay selection scheme for the secondary
system. The proposed optimal scheme requires exhaustive search
over all the candidate relay nodes and to reduce its complexity, a
non-exhaustive sub-optimal search method is also proposed. By
simulation, we show that both optimal and sub-optimal schemes
provide a signicant throughput gain when compared to random
relay selection with optimal power allocation (OPA) and equal
power allocations (EPA).
Index TermsCognitive radio; cooperative communication;
relay selection; power allocation; throughput; maximum ratio
combining.
I. INTRODUCTION
Cognitive radio (CR) is an emerging technology dealing
with the scarcity of the radio spectrum [1]. In this paradigm,
the wireless nodes can sense the spectrum, adapt the trans-
mission/reception to the available spectrum, and dynamically
share the spectrum with other applications. In an underlay CR
system, it is essential that the transmitted signals of the CR
terminals, also known as the secondary users, satisfy some
maximum interference constraints on the primary system, i.e.
the system that has the right to use a licensed band.
Cooperative relay communication can improve power ef-
ciency in the wireless networks by increasing the spatial
diversity. By the use of relays, the transmitted power from the
source terminal can be signicantly reduced. This is especially
appealing in the underlay CR systems to keep the interference
to the primary system [1]- [4]. Also, relaying can help when
there is no common band between the secondary source and
the destination as it can work as a bridge between the source
and the destination [3]. The CR relays not only can provide
reliable links between the secondary nodes, they also can
assist in improving the throughput of the primary network by
amplifying the primary signal when possible [5]. Therefore,
cooperative communications is a viable choice for cognitive
radio systems.
In this paper, we consider the problem of relay selection
and power allocation at the source and relay nodes in a
multi-relay amplify-and-forward (AF) cognitive radio system.
It is assumed that there are multiple relay nodes available
in the system and the best relay is selected to maximize
the throughput of the secondary system subject to maximum
power constraints at source and relay nodes and maximum
interference constraints from the source and the relay nodes
to the primary system. We assume a more general problem
in which there are different source-relay and relay-destination
common bands, i.e. the common band between the source and
the relay nodes is different from that of the source and the
destination [3], [6].
In [3], the problem of optimal power allocation to maximize
the throughput in a cooperative secondary system with single
relay has been studied; however, [3] does not study the
problem of relay selection in a multi-relay secondary system.
In [8], relay selection and power allocation for a CR system
are performed in order to maximize the throughput considering
only the relay link. The secondary system throughput is
improved by taking into account the interference from the
primary user to the CR relay and the secondary destination
and by applying beamforming. Also in [9], the relay selection
and power allocation problem for a three-node cognitive radio
system without the direct link has been studied. It has been
shown that the optimization problem is a mixed-integer pro-
gramming problem and optimal solution is based on exhaustive
search over all the relay links.
In [10], distributed relay selection and optimal power al-
location in the CR networks are studied with the aim of
making a tradeoff between the achievable rate and the power
consumption. In [11], relay selection and power allocation
have been done for a multi-hop cognitive network considering
energy constraints at the secondary source, destination and
each secondary relay with the objective of maximizing the
secondary network capacity ensuring the interference to the
primary is within a threshold limit. Relay selection and power
allocation to improve the throughput when there are multiple
CR relays has been discussed in [6]. The proposed scheme
uses dual-domain and sub-gradient numerical optimization
methods to nd the optimal solutions which is complex and
requires several iterations. On the contrary, we propose a
simple yet effective optimal power allocation for each relay
and through exhaustive search, the best relay node is selected
in order to have maximum throughput value in a three node
relay system.
The rest of this paper is organized as follows. Section II,
describes a model for a cooperative cognitive radio system
operating in AF mode. We assume that the transmission from
the source to the destination and from the source to the
relay are done over different (uncommon) bands. We also
assume that the destination uses maximum ratio combining
(MRC) to combine the received signals from the source
and the relay nodes. With these assumptions, we come up
with an equation to express the throughput of the system.
In Section III, assuming a preselected optimal relay node,
we set up an optimization problem to nd out the set of
transmit powers at the source and relay nodes to maximize
the throughput obtained in Section II assuming maximum
power and interference constraints. It is shown that the optimal
solution always lies in the boundary and by examining the
boundary conditions, we nd out the optimal solution. The
best relay node is then found by an exhaustive search over
all available relay nodes (optimal approach). This exhaustive
search is costly and, therefore in Section IV, we propose an al-
ternative low-complexity suboptimal approach to jointly select
the best relay as well as the set of optimal powers (suboptimal
approach). In Section V, using the simulation results, we show
that the optimal approach provides a signicant throughput
gain compared to random relay selection while the proposed
suboptimal approach can reach the throughput gain close to
that of the optimal approach. Finally, Section VI concludes
the paper.
II. SYSTEM DESCRIPTION
We consider a general CR system as shown in Fig. 1. It
consists of a secondary source and a secondary destination
as well as a total of n relay nodes operating in AF mode.
For the sake of simplicity, we assume that the i
th
relay has
been selected as the best relay. The problem of relay selection
will be addressed later. In Fig. 1, CH0, CH1, and CH1 are
the channels from the secondary source to the destination,
from the secondary source to the i
th
relay node, and from
the i
th
relay node to the secondary destination, respectively.
We consider the channels CH0 and CH1 are operating in two
orthogonal bands, i.e. the source-relay and source-destination
common bands are different. In our model, we consider full-
duplex terminals; however, our analysis can also be easily
applied to the case of half-duplex terminals.
Assume P
1
and P
2
be the power transmitted from the
secondary source to the secondary destination and to the
i
th
secondary relay node, respectively; and P
3
be the power
transmitted from the i
th
relay node. We consider the following
power and interference constraints in our problem
P
1
+P
2
P
total
(1)
P
3
P
3max
(2)
P
1
|h
s,p,1
|
2
I
1
(3)
P
2
|h
s,p,2
|
2
I
2
(4)
P
3
|h
i,p
|
2
I
3
(5)
where P
total
and P
3max
are the maximum available power
at the secondary source and the relay nodes, respectively. I
1
,
I
2
and I
3
are the maximum allowed interference power to
the primary system over the channel bands CH0, CH1 and
CH1 respectively. Also, h
s,p,1
is the channel gain from the
secondary source to the primary user over the CH0 band, h
s,p,2
is the channel gain from the secondary source to the primary
user over the CH1 band, and h
i,p
is the channel gain from
the i
th
secondary relay node to the primary user over CH1
band. We assume that these channel state information (CSI)
are available to the CR system prior to transmission.
Let x be the signal transmitted from the source node, y
s,d
be
the signal received at the destination directly from the source,
y
s,r
be the signal received at the relay and y
r,d
be the signal
received at the destination from the relay. We can write
y
s,d
=
_
P
1
h
s,d
x +n
s,d
(6)
y
s,r
=
_
P
2
h
s,r
x +n
s,r
(7)
y
r,d
=
r
h
r,d
y
s,r
+n
r,d
(8)
Fig. 1. A general structure of a cooperative cognitive radio system
where h
s,d
is the channel gain from the secondary source
to the destination, h
s,r
is the channel gain from the source
to the relay, and h
r,d
is the channel gain from the relay
to the destination. We assume that n
s,d
, n
s,r
and n
r,d
are
independent additive white Gaussian noises with zero mean
and variance N
0
. For the relay operating in AF mode,
r
is
the amplication factor and is dened as

r
=

P
3
_
P
2
|h
s,r
|
2
+N
o
(9)
By replacing (9) in (8), we have
y
r,d
=

P
3
_
P
2
|h
s,r
|
2
+N
o
h
r,d
y
s,r
+n
r,d
(10)
Then, one can rewrite (7) as
y
r,d
=

P
2
P
3
h
r,d
h
s,r
x
_
P
2
|h
s,r
|
2
+N
o
+

P
3
h
r,d
n
s,r
_
P
2
|h
s,r
|
2
+N
o
+n
r,d
(11)
or
y
r,d
=

P
2
P
3
h
r,d
h
s,r
x
_
P
2
|h
s,r
|
2
+N
o
+n

r,d
(12)
where n

r,d
=

P3h
r,d
ns,r

P2|hs,r|
2
+No
+ n
r,d
. Since n
s,r
and n
r,d
are
independent Gaussian noises, n

r,d
is also Gaussian with zero
mean and variance
N

0
=
_
P
3
P
2
|h
s,r
|
2
+N
o
|h
r,d
|
2
+ 1
_
N
0
(13)
From (6), the signal-to-noise ratio (SNR) at the destination
due to the direct link is

1
=
P
1
|h
s,d
|
2
N
0
(14)
and using (12), the SNR at the destination due to the relay
link can also be written as

2
=
P
2
P
3
|h
r,d
|
2
|h
s,r
|
2
_
P
3
|h
r,d
|
2
+P
2
|h
s,r
|
2
+N
o
_
N
o
(15)
It has been shown that applying the maximum ratio combining
at the destination in an AF relay system maximizes the SNR
[7]. Therefore, we use the maximum ratio combining (MRC) at
the secondary destination to combine the two received signals
Fig. 2. Graphical method to obtain P

1
and P

2
from the direct and relay links. We assume that the secondary
destination has all the CSIs i.e. h
s,d
, h
s,r
and h
r,d
. As the
instantaneous SNR at the output of combiner equals the sum
of the SNRs of the incoming signals [7], one can write the
instantaneous SNR after applying the MRC as
=
1
+
2
(16)
where
1
and
2
are obtained from (14) and (15), respectively.
Therefore, assuming the i
th
relay has been selected, one can
write the instantaneous throughput of the CR system with the
MRC at the destination as
T
i
(P
1
, P
2
, P
3
) = (1 ) log
2
(1 +
1
+
2
) (17)
where is the probability of miss-detection of the spectrum
holes in the spectrum sensing process [6]. In the next section,
we propose a relay selection and power allocation scheme to
maximize the instantaneous throughput of the CR system in
(17).
III. OPTIMAL APPROACH
In the optimal approach, the optimal power allocation to
maximize the sum-rate for each relay is calculated and then,
through an exhaustive search amongst all relays, the best
relay is selected such that the throughput as dened in (17)
is maximized. Therefore, in the following, we set up an
optimization problem to nd the optimum set of transmit
powers P
1
, P
2
and P
3
for a typical relay node. The Power
allocation problem for each candidate relay can be formulated
as the following optimization
(P

1i
, P

2i
, P

3i
) = arg max
P1,P2,P3
T
i
(P
1
, P
2
, P
3
) (18)
subject to (1) (5)
where P

1i
, P

2i
, and P

3i
are the optimal values of P
1
, P
2
and
P
3
for the i
th
secondary relay. The optimal relay selection
is then can be formulated as an exhaustive search over all
candidate relays; i.e.

i = arg max
i
T
i
(P

1i
, P

2i
, P

3i
) (19)
where i ranges from 1 to n and

i represents the optimal relay.
The optimal power allocation for each candidate relay can
be obtained using a graphical method as shown in Fig. 2.
Since the throughput in (17) is an increasing function of , to
maximize the throughput, it is sufcient to maximize . Also,
note that
2
is an increasing function of P
3
. Using (2) and
(5), the optimal value for P
3
would be
P

3
= min
_
I
3
|h
i,p
|
2
, P
3max
_
(20)
Therefore, one can conclude that the optimal value of P
3
is
independent of P
1
and P
2
. Also, note that for a xed value
P
2
, is an increasing function of P
1
. The same argument
is valid for P
2
and, therefore, the throughput is maximized
when constraint (1) is satised with equality; i.e. when P
1
+
P
2
= P
total
. Therefore, one can conclude that the optimal
solution lies on the line P
1
+ P
2
= P
total
. Substituting P
1
with P
total
P
2
, one can write as a function of P
2
as
(P2) =
(P
total
P2) |h
s,d
|
2
N0
+
P2P

3
|h
r,d
|
2
|hs,r|
2

3
|h
r,d
|
2
+P2 |hs,r|
2
+N0

N0
(21)
Differentiating (21) with respect to P
2
and setting to zero will
result in

P
2
=
|h
r,d
|
|h
s,d
| |h
s,r
|
_
_
P

3
|h
r,d
|
2
+N
0
_
P

3

P

3
|h
r,d
|
2
+N
0
|h
s,r
|
2
(22)
which is clearly an extremum point of . It is easy to show
that the second order derivative of (P
2
) with respect to P
2
is
always negative and is a convex function of P
2
. Therefore,
(

P
2
) is a maximum for (P
2
). As shown in Fig. 2,

P
2
would
be an acceptable solution to the optimization problem if it also
satises other constraints in (1)-(5). In particular referring to
equations (3) and (4), let P
1C
and P
2C
be dened as
P
1C
=
I
1
|h
s,p,1
|
2
(23)
P
2C
=
I
2
|h
s,p,2
|
2
(24)
Then,

P
2
should be on line P
1
+P
2
= P
total
that is inside
the rectangle shown in Fig. 2. Note that in Fig. 2 for the
purpose of illustration and without loss of generality, we have
assumed P
2C
P
1C
. Fig. 2 shows different possible cases
for the location of the line P
1
+ P
2
= P
total
. Based on that,
we can consider three different cases.
Case I. P
total
P
1C
This is the case when line P
1
+P
2
= P
total
is represented by
line S1 in Fig. 2. In this case, the optimal values for P

2
and
P

1
are obtained as
P

2
=
_
_
_
0 if

P
2
< 0
P
total
if

P
2
> P
total

P
2
otherwise
(25)
P

1
= P
total
P

2
(26)
Case II. P
1C
P
total
P
1C
+P
2C
In this case, one of the S2 or S3 lines represent the P
total
line. In this case, the optimal values for P

2
and P

1
can be
formulated as
P

2
=
_
_
_
P
total
P
1C
if

P
2
< P
total
P
1C
min {P
total
, P
2C
} if

P
2
> min {P
total
, P
2C
}

P
2
otherwise
(27)
P

1
= P
total
P

2
(28)
Case III. P
total
P
1C
+P
2C
This is the case when P
total
line is represented by line S4
and, in this case, the optimal set of powers is
P

2
= P
2C
(29)
P

1
= P
1C
(30)
The optimal power allocation for each relay node can be
obtained from one of the above three cases. The best secondary
relay node is then found by an exhaustive search amongst all
possible candidate relays and nding the relay that maximizes
the throughput using (19).
IV. SUB-OPTIMAL APPROACH
In the optimal approach, the power allocation is carried out
for each candidate relay node and the relay node with the
largest throughput is selected. Therefore, the optimal relay
selection requires an exhaustive search over all the relays. To
reduce the complexity of the relay selection algorithm, we also
propose a reduced-complexity sub-optimal relay selection. In
the sub-optimal approach, we assume that the power allocation
problem is mainly dominated by the interference constraints
rather than the power constraints in (1)-(5). By considering
only the interference constraints in (3), (4) and (5), it is
possible to reduce the complexity of the optimal approach.
Note that under this assumption,
1
will be constant for all
the relay nodes. Therefore, maximizing
2
also maximizes
in (16). Now, replacing P
2
with
I2
|hs,p,2|
2
and P
3
with P

3
in
(15), we get

2
=
I
2
P

3
|h
r,d
|
2
|h
s,r
|
2
_
P

3
|h
s,p,2
|
2
|h
r,d
|
2
+I
2
|h
s,r
|
2
+|h
s,p,2
|
2
N
0
_
N
0
(31)
Notice that
2
in (31) is not a function of P
1
and P
2
. The
relay node that maximizes (31) will be selected as the best
relay. After the selection of the sub-optimum relay using (31),
we use the discussed power allocation algorithm in Section
III to nd the optimal power allocation for the selected relay
node. Therefore, the sub-optimal approach removes the need
to calculate the optimal powers for each candidate relay node
before proceeding to the relay selection and thereby reduces
the complexity. With simulation, we will show that the sub-
optimal approach has a throughput performance close to that
of the optimal approach especially at high received SNRs.
V. SIMULATION RESULTS
We consider a cooperative CR network with n CR relay
nodes operating in AF mode. Throughout the simulations, all
the channels are assumed to be Rayleigh fading with unity
bandwidth. We also consider I
1
= 0.5mW, I
2
= 0.8mW,
and I
3
= 1mW. The miss-detection probability is chosen
to be .01. Throughput the simulation, we also assume N
0
=
0.8mW/Hz.
First, we simulate the system throughput versus the total
source power, P
total
, assuming that P
3max
= 0.6W. The
10
3
10
2
10
1
10
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
P
total
(W)
T
h
r
o
u
g
h
p
u
t
(
b
p
s
/
H
z
)


Optimal approach
Random relay with EPA
Random relay with OPA
Suboptimal approach
Fig. 3. System throughput versus maximum transmit power limit of CR
source
10
3
10
2
10
1
10
0
2.5
3
3.5
4
4.5
5
5.5
P
3max
(W)
T
h
r
o
u
g
h
p
u
t
(
b
p
s
/
H
z
)


Optimal approach
Random relay with EPA
Random relay with OPA
Suboptimal approach
Fig. 4. System throughput versus maximum transmit power limit at CR relay
number of available CR relay nodes are kept xed at n = 20.
We apply our proposed optimal and suboptimal relay selection
and power allocation schemes and simulate the throughput
versus P
total
in Fig. 3. As a benchmark, we have also
simulated the cases of random relay selection with equal
power allocation (EPA) and optimal power allocation (OPA).
From Fig. 3, we see that the throughput increases signicantly
when our proposed relay selection scheme is applied compared
to the random relay selections (EPA and OPA). In fact by
applying our proposed optimal and sub-optimal approaches
throughput gains of 2.3 bps/Hz and 1.3 bps/Hz compared to
the EPA and the OPA schemes are obtained, respectively, at
sufciently large SNRs. The at portion of the throughput
curves is due to the fact that at sufciently large P
total
, the
dominant constraints are, in fact, the interference constrains,
and therefore, the total transmitted source power is limited
by the interference constraints. This means that increasing
P
total
does not increase the transmitted power and thereby
the throughput does not increase. Also note that especially
at high SNRs, the proposed sub-optimal approach almost
provides the same throughput as the optimal approach. This
means that the sub-optimal approach is an effective way to
achieve the maximum throughput of the CR system. The
small difference in the throughputs offered by the optimal and
the sub-optimal approaches is due to the omission of P
total
constraint in deriving the sub-optimal approach. In fact, at
low SNRs, the P
total
constraint is the dominant constraint.
Omitting it from the optimization problem is not wise and
will result in a hit in the throughput as observed from Fig. 3.
Fig. 4 shows the throughput performance of different power
allocation approaches versus the maximum transmit power at
the relay nodes, P
3max
, while P
total
is kept constant at 1mW.
As before, the optimal and sub-optimal approaches provide 2.1
bps/Hz and 1.3 bps/Hz throughput gains compared to the EPA
and the OPA, respectively, especially at high power regimes.
At lower P
3max
, optimal approach performs better than the
sub-optimal approach due to the fact that at low power regimes
the power constraints are dominant and removing them from
the optimization problem (as we did for the sub-optimal case)
results in a hit in the throughput as seen from Fig. 4.
Fig. 5 shows the performance of the optimal and the sub-
optimal approaches versus the number of CR relays when
P
total
= 1W and P
3max
= 0.6W and compared the
throughputs with those of the EPA and the OPA. As seen,
the proposed method provides an appealing sum-rate increase
by increasing the number of relays while the EPA and the OPA
do not provide any throughput increase. This is because the
relay is selected randomly in the later schemes. Finally, Fig.
6 illustrates the throughput behavior of our proposed schemes
versus the number of candidate relays for different values of
P
total
and P
3max
. As seen, the throughputs offered by the
sub-optimal and optimal approaches are close in all the cases.
VI. CONCLUSION
In this paper, we presented joint relay selection and power
allocation and relay selection schemes to improve the through-
put in a cooperative CR system with interference and power
limitations. First, an optimal relay selection and power alloca-
tion approach based on a constrained optimization problem to
maximize the throughput of the CR system was developed.
The proposed approach requires to calculate the optimum
power allocation for all the candidate relay nodes which
means an exhaustive search among all the relays. To reduce
the complexity, a sub-optimal approach was also developed.
Assuming that the interference constraints are dominant, the
sub-optimal approach is able to pick the best relay before
applying the power allocation. The optimal power allocation
is then calculated for the selected relay.This removes the need
to calculate the optimal power allocation for all the candidate
relays and thereby reduces the complexity of the proposed
optimal approach. By simulation, it was shown that the sup-
optimal scheme provides a throughput close to that of the
optimal approach especially at high SNR regimes.
REFERENCES
[1] K. B. Letaief and W. Zhang, Cooperative communication for cognitive
radio networks, Proc. IEEE, vol. 97, no. 5, pp. 878-893, May 2009.
[2] N. Devroye, P. Mitran and V. Tarokh, Limits on communications in
a cognitive radio channel, IEEE Commun. Mag., vol. 44, no. 6, pp.
44-49, June 2006.
[3] Z. Guodong, Y. Chenyang, G. Y. Li, L. Dongdong and A. C. K. Soong,
Power and channel allocation for cooperative relay in cognitive radio
networks, IEEE J. Sel. Topics Signal Process., vol. 5, no. 1, pp. 151-
159, Feb. 2011.
[4] A. Nosratinia, T. E. Hunter and A. Hedayat, Cooperative communica-
tion in wireless networks, IEEE Commun. Mag., vol. 42, no. 10, pp.
74-80, Oct. 2004.
2 4 6 8 10 12 14 16 18 20
3
3.5
4
4.5
5
5.5
Number of CR relays
T
h
r
o
u
g
h
p
u
t
(
b
p
s
/
H
z
)


Optimal approach
Suboptimal approach
Random relay with OPA
Random relay with EPA
P
total
= 1W
P
3max
= 0.6W
Fig. 5. System throughput versus number of candidate CR relays
2 4 6 8 10 12 14 16 18 20
2
2.5
3
3.5
4
4.5
5
5.5
Number of CR relays
T
h
r
o
u
g
h
p
u
t
(
b
p
s
/
H
z
)


Optimal approach
Suboptimal approach
P
total
= 1W
P
3max
= 0.6W
P
total
= 30mW
P
3max
= 10mW
P
total
= 5mW
P
3max
= 2mW
Fig. 6. System throughput versus number of candidate CR relays with
different P
total
and P
3max
[5] P. Gong, J. H. Park, J. M. Yoo, B. S. Yu and D. K. Kim, Throughput
maximization with multiuser non-selsh cognitive relaying in CR net-
works, in Proc. 4th Int. Symp. Wireless Pervasive Computing (ISWPC),
Feb. 2009, pp. 1-5.
[6] L. Li, X. Zhou, H. Xu, G. Y. Li, D. Wang and A. Soong, Simplied
relay selection and power allocation in cooperative cognitive radio
systems, IEEE Trans. Wireless Commun., vol. 10, no. 1, pp. 33-36,
Jan. 2011.
[7] K. J. R. Liu, A. K. Sadek, W. Su and A. Kwasinski, Cooperative
Communications and Networking, Cambridge Univ. Press, 2009.
[8] C. Sun and K. B. Letaief, User cooperation in heterogeneous cognitive
radio networks with interference reduction, in Proc. IEEE Int. Conf. on
Commun. (ICC), May 2008, pp. 3193-3197.
[9] D. Bharadia, G. Bansal, P. Kaligineedi and V. K. Bhargava, Relay and
power allocation schemes for OFDM-based cognitive radio systems,
IEEE Trans. Wireless Commun., vol. 10, no. 9, pp. 2812-2817, Sept.
2011.
[10] C. Luo, F. R. Yu, H. Ji and V. C. M. Leung, Distributed Relay Selection
and Power Control in Cognitive Radio Networks with Cooperative
Transmission, in Proc. IEEE Int. Conf. on Commun. (ICC), May 2010,
pp. 1-5.
[11] Y. Chen, Z. Feng and X. Chen, Joint relay selection and power
allocation for energy-constrained multi-hop cognitive networks, in Proc.
IEEE Veh. Tech. Conf. (VTC), May 2011, pp. 1-5.

You might also like