2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

Full-Duplex MIMO Relaying Powered by Wireless

Energy Transfer
Mohammadali Mohammadi , Himal A. Suraweera , Gan Zheng , Caijun Zhong , and Ioannis Krikidis ,
Faculty of Engineering, Shahrekord University, Iran (e-mail: m.a.mohammadi@eng.sku.ac.ir)
Department of Electrical and Electronic Engineering, University of Peradeniya, Sri Lanka (e-mail: himal@ee.pdn.ac.lk)
School of Computer Science and Electronic Engineering, University of Essex, UK (e-mail: ganzheng@essex.ac.uk)
Institute of Information and Communication Engineering, Zhejiang University, China (e-mail: caijunzhong@zju.edu.cn)
 Department of Electrical and Computer Engineering, University of Cyprus, Cyprus (e-mail: krikidis@ucy.ac.cy)

AbstractWe consider a full-duplex decode-and-forward sys-
tem, where the wirelessly powered relay employs the time-
switching protocol to receive power from the source and then
transmit information to the destination. It is assumed that the
relay node is equipped with two sets of antennas to enable full-
duplex communications. Three different interference mitigation
schemes are studied, namely, 1) optimal 2) zero-forcing and
3) maximum ratio combining/maximum ratio transmission. We
develop new outage probability expressions to investigate delay-
constrained transmission throughput of these schemes. Our anal-
ysis show interesting performance comparisons of the considered
precoding schemes for different system and link parameters.
I. I NTRODUCTION
Most wireless radios so far have adopted half-duplex (HD)
communications due to the challenge of handling loopback
interference (LI) generated from simultaneous transmit/receive
operation. However, thanks to the progress made on LI
suppression recently, full-duplex (FD) communications have
emerged as a viable option [1][3]. In theory, FD operation
can double the HD capacity, hence is a key enabling technique
for 5G systems.
On the other hand energy harvesting communications is a
new paradigm that can power wireless devices by scavenging
energy from external resources such as solar, wind, ambient
RF power etc. Energy harvesting from such sources are not
without challenges due the unpredictable nature of these
energy sources. To this end, wireless energy transfer has been
proposed as a promising technique for a variety of wireless
networking applications [4].
RF signals can carry both information and energy and
pioneering contributions quantifying this fundamental tradeoff
have been reported. In order to remedy practical issues (same
signal can not be used for both decoding and rectifying)
associated with simultaneous information and energy transfer,
two practical techniques, i.e., time-switching (TS) and power-
splitting (PS) were proposed in [5]. Both TS and PS apply in
different network topologies and integration of the RF energy
transfer into cooperative relay networks is an interesting
research topic. Different HD relay networks have been studied
The work of C. Zhong was partially supported by the Zhejiang Provincial
Natural Science Foundation of China (LR15F010001) and the Fundamental
Research Funds for Central Universities (2014QNA5019).
The work of I. Krikidis was supported by the Research Promotion Founda-
tion, Cyprus under the project KOYLTOYRA/BP-NE/0613/04 Full-Duplex
Radio: Modeling, Analysis and Design (FD-RD).
considering amplify-and-forward (AF) relaying [5], [6], relay
selection [7] and multiple antenna relay systems [8]. In [9],
the achievable throughput of FD AF and DF relaying systems
with TS has been studied. However, [9] only assumed single
transmit/receive antenna at the relay.
Inspired by the FD approach, in this paper we consider a
source-relay-destination scenario where the multiple antenna
FD relay is powered via wireless energy transfer from the
source. The reason for the adoption of multiple antennas at the
relay is two-fold: (1) employment of an antenna array helps
the relay to accumulate more energy (2) spatial LI cancellation
techniques can be deployed. Specically, we investigate the
outage probability and the delay constrained throughput by
considering several precoding schemes at the relay. Our results
are general in the sense that we consider arbitrary number of
receive/transmit antennas at the FD relay input/output.
In summary, the contributions of this work are as follows:
1) Assuming different precoding schemes at the relay,
namely, optimal, zero-forcing (ZF) and maximum ratio
combining (MRC)/maximum ratio transmission (MRT),
we develop new expressions for the systems outage
probability, which are helpful to investigate the effect of
key system parameters on performance metrics such as
the outage probability and delay-constrained throughput.
2) In the case of ZF, we present simple high signal-to-
noise ratio (SNR) expressions for the outage probability
which enable the characterization of the systems diver-
sity order and array gain. Moreover, we compare the
performance of FD and HD modes to show the benets
of FD operation.
II. S YSTEM M ODEL
We consider a DF relaying network consisting of one source
S, one relay R, and one destination D. Both S and D are
equipped with a single antenna, while R is equipped with MR
receive (input) antennas and MT transmit (output) antennas to
enable FD operation. We assume that the S to D link does
not exist, due to severe shadowing and path loss effect.
It is also assumed that the relay has no external power
supply, and is powered through wireless energy transfer from
the source. We adopt the time-sharing protocol [5], hence
the entire communication process is divided into two phases,
i.e., for a transmission block time T , fraction of the block
time is devoted for energy harvesting and the remaining time,

978-1-4799-1931-4/15/$31.00 2015 IEEE 296

 2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

(1 )T , is used for information transmission. It is also end SINR in (5) is maximized. Specically, for a xed value
assumed that the channels experience Rayleigh fading and re- of , the SINR maximization problem can be formulated as
main constant over the block time T and varies independently max (in Eq. (5)) (6)
wt ,wr
and identically from one block to the other.
During the energy harvesting phase, the received signal re s.t. wr  = wt  = 1.

at the relay can be expressed as In order to solve the problem in (6), we rst x wt and
PS optimize wr to maximize . Therefore, the optimization
re = hSR xe + nR , (1) problem can be re-formulated as
d1
|wr hSR |2
where PS is the source transmit power, d1 is the distance max 2 2
, (7)
d hSR  |wr HRR wt | + 1
wr PS
between the source and relay, is the path loss exponent, 1

hSR is the MR 1 channel vector for the S-R link, i.e., s.t. wr  = 1,
input antennas at R are connected to the rectifying antenna which is a generalized Rayleigh ratio problem. It is well known
(rectenna), xe is the energy symbol with unit energy, and nR that the objective function in (7) is globally maximized when
is the zero mean additive white Gaussian noise (AWGN) with  1
unit variance. We assume that the energy collected during the hSR P
d1
S
h SR 2
H w w
RR t t RRH + I
rst phase is fully consumed by the relay to forward the source wr =  P 1
. (8)
signal to the destination. Hence, the relay transmit power can h S
h 2H w w
H + I
SR d1 SR RR t t RR
be computed as Pr = d PS hSR 2 , where  1
with
1
denoting the energy conversion efciency.
Accordingly, by substituting wr into the objective function,
Now, let us consider the information transmission phase.
the optimizationproblem in (6) can be re-expressed as
 at R can be written as
The received signal 2 2
d1 hSR  hSR HRR wt 
PS
PS 1 2
r[n] = hSR xS [n] + HRR xR [n] + nR [n], (2) max PS min hSR  2 2
,
d1 wt d1 1 + P d1 hSR  HRR wt 
S


where xS [n] is the source information symbol with unit PS 2 2
hSR  |hRD wt |
energy,
 and xR[n] is the transmitted relay signal satisfying d1 d2
E xR [n]xR [n] = Pr , and HRR denotes the MR MT LI s.t. wt  = 1, (9)
channel. Upon receiving the signal, R rst applies a 1 MR which is still difcult to solve. Therefore, instead of (9)
linear combining vector wr on r[n] to obtain an estimate we solve the following problem by introducing an auxiliary
of xS , then forwards signal to the destination D using the variable t, as
MT 1 transmit beamforming vector wt . It is assumed that f (t)  max |hRD wt |2 (10)
wt  = wr  = 1. wt =1
 estimate xS [n] = wr r[n] can be expressed as
The relays 2 2
d1 hSR  hSR HRR wt 
PS
PS s.t. = t.
xS [n] = wr hSR xS [n]+wr HRR xR [n]+wr nR [n]. (3) 1 + P 2 2
d1 hSR  HRR wt 
S
d1
This is a nonconvex quadratic optimization problem with
The relay transmit signal is given by xR [n] = quadratic equality constraint. To solve the problem in (10), we
Pr wt xS [n ] where accounts for the time delay caused
by relay processing. Finally, the received signal at D is apply a similar approach as in [10] to convert the optimization
 problem to
1
yD [n] = hRD xR [n] + nD [n]. (4) max tr(hRD Wt hRD ) (11)
d2 Wt

where hRD is the 1 MT channel vector of the R D link, td1

d2 is the distance between the relay and destination, and nD
is the zero mean AWGN with unit variance.
With the DF protocol, end-to-end signal-to-interference-
plus-noise ratio (SINR) can be written as
= (5)
 
PS |wr hSR |2 PS 2 2
min , hSR  |hRD wt | .
PS hSR 2 |wr HRR wt |2 +d1 d1 d2
III. J OINT P RECODING /D ECODING D ESIGNS
In this section, we consider several precoder/decoding de-
signs to suppress/cancel the effect of LI at the relay, each of
which offers different performance-complexity tradeoff.
A. The Optimal Scheme
In this subsection, our main objective is to jointly design the
precoder and the decoder at the FD relay so that the end-to-
s.t. tr(Wt (HRR hSR hSR HRR HRR HRR ))=
PS hSR 2
tr(Wt ) = 1,
where Wt = wt wt is a symmetric, positive semi-denite
matrix. In order to solve (11), we can resort to the widely
used semidenite relaxation approach. By dropping the rank-
1 constraint, the resulting problem becomes a semidenite
program, whose solution Wt can be found by using the
method provided in [10, Appendix B].
Denoting the optimal objective value of (11) as f (t), the
SINR maximization
  problem canbe formulated as 
1 PS
max min PS hSR 2 t , hSR 2 f (t) . (12)
t0 d1 d1 d2
Therefore, in order to solve (6), it remains to perform a one-
dimensional optimization with respect to the variable t.

297
 2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

B. Transmit ZF (TZF) Scheme where Q(a, x) = (a, x)/(a) and (a, x) is the upper
In the transmit ZF scheme, the FD relay takes advantage of incomplete Gamma function [13, Eq. (8.350.2)], and 1 = PS
the multiple transmit antennas to completely cancel the LI. To is the SNR of the rst hop.
ensure this is feasible, the number of the transmit antennas at To the best of the authorss knowledge, the integral in (16)
R should be greater than one, i.e., MT > 1. In this case wr = does not admit a closed-form expression. However, this single
hSR integral expression can be efciently evaluated numerically
hSR  and wt is the solution of the following optimization using software such as Matlab or Mathematica.
problem:
To gain further insights, we now look into the high SNR
max |hRD wt |2 (13) regime and derive a simple approximation for the outage prob-
wt
ability, which enables the characterization of the achievable
s.t. wt  = 1, hSR HRR wt = 0.
diversity order of the TZF scheme.
From the ZF constraint, we know that wt lies in the null space
H hSR hSR HRR Proposition 1. In the high SNR regime, i.e., 1 , the
of hSR HRR . Denoting B  I RR h H 2
, we have outage probability of FD relaying with the TZF scheme can
SR RR
BhRD
wtZF = BhRD 
. be approximated as (17) at the top of the next page, where
(x) is the digamma function [12, Eq. (6.3.1)].

C. Receive ZF (RZF) Scheme Proof: Due to limited space proof is omitted.

By inspecting (17), we see that the TZF scheme achieves a
As an alternative solution, the transmit beamforming vector diversity order of min(MR , MT 1). Moreover, we notice that
h MR
can be set using the MRT principle, i.e., wt = hRD RD 
, and for MR + 1 = MT , FTZF (z) decays as
1 ln(1 ) rather
wr based on the ZF criterion. To ensure feasibility of ZF, R than 1MR
as in the conventional case which implies that in
should equipped with MR > 1 receive antennas. the energy harvesting case the slope of FTZF (z) converges

By using similar procedure as shown for the transmit ZF much slower compared with that in the constant power case.
scheme, the combining vector wr can be obtained as wrZF =
DhSR HRR hRD hRD HRR B. RZF Scheme
DhSR  with D  I HRR hRD 2 .
Invoking (5), and using wrZF and wtMRT , the end-to-end SNR
RZF can be expressed as  
D. MRC/MRT Scheme
RZF = ZPS min X1 , hRD 2 , (18)
For the MRC/MRT scheme, wr and wt are set to match d2
the rst hop and second hop channel, respectively. Hence,  
2
h h where Z  d1 hSR 2 + |h1 |2 and X1 = h hSR .
wrMRC = hSR
SR 
and wtMRT = hRD
RD 
. 1 SR
2 +|h |2
1
It is well known that Z follows central chi-square distribution
with 2MR degrees-of-freedom, denoted as Z 22MR and that
IV. O UTAGE P ROBABILITY X1 follows a beta distribution with shape parameters MR 1
In this section, we investigate the outage probability of and 1, denoted as X1 Beta(MR 1, 1) with FX1 (x) =
the considered FD relay system assuming TZF, RZF, and xMR 1 , 0 < x < 1, [14, p. 138].
MRC/MRT schemes. In case of the optimal scheme, derivation Moreover, by denoting Y1 = d hRD 2 , we have FY1 (y) =
  2
of the outage is difcult and we use simulations in Section V. d
P MT , 2 y , where P (a, x) = (a, x)/(a) is the regu-
The outage probability is an important performance metric,
larized lower incomplete Gamma function [12, Eq. (6.5.1)].
which is dened as the probability that the instantaneous SINR
With FX1 (x) and FY1 (y) in hand, the cdf of RZF can be
falls below a predened threshold, th . Mathematically, it can
expressed as (19). Although Eq. (19) does not admit a closed-
be written as
form solution, it can be efciently evaluated. Now, we look
Pout = Pr( < th ) = F (th ). (14) into the high SNR regime, and investigate the diversity order
achieved by this scheme.
A. TZF Scheme Proposition 2. In the high SNR region, i.e., 1 , the
By substituting the wtZF and wrMRC into (5), the end-to-end outage probability of the FD relaying system with the RZF
SNR TZF can be expressed as  scheme can be approximated as
  MR 1
PS 2 2
TZF = hSR  min 1 , hRD  , (15)

1 d1 z
, MR < MT +1,
(1 )d1 (MR )  1
d2

 M T   M T
1 1 d2 d1 z
where hRD is an (MT  1) 1 vector with  hRD 2 (z)
(MR ) 1+ (MT ) 1 ,
FRZF
2 2
2(MT 1) . Let Y = min 1 , d hRD  . Then invoking


MR = MT +1,
2
the cumulative density function (cdf) of Y presented in [11,
 MT MT

(M M ) d d z
Appendix II], the cdf of TZF can be obtained as
R T
(MR )(MT +1)
2 1
1 , MR > MT +1.

 Q MT 1, 1 2d
d z
Proof: The proof is omitted due to space limit.
1 x
FTZF (z) = 1 d z xMR1 ex dx. (16) Proposition 2 indicates that the RZF scheme achieves a
1 (M R )
1 diversity order of min(MR 1, MT ).

298
 2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

   M +k1  M R

1
+ 1 (1)k+1
d2 T
1 d1 z
, MT > MR +1,

(MR +1) (MT 1)(MR ) k!(k+MT ) MR MT k+1 1

k=0
(z)
 k=MR MT +1
 M R   M R
FTZF (17)

1
1 + 1
(ln ( ) ln (d
z) + (1))
d2 d1 z
, MT = MR +1

(M +1) (M ) 1 1 1

R R

   
(MR MT +1) d2 MT 1 d1 z MT 1 , MT < MR +1
(MT )(MR ) 1

   
   MR 1  
d1 z 1 d1 d2 z MR 1 x d1 z d1 d2 z x
FRZF (z) = 1Q MR , + P MT , x e dx+ Q MT , e dx .
1 (MR ) d
1z
1
1 x 1 d
1z
1
1 x
(19)

C. The MRC/MRT Scheme simplies to

Substituting wrMRC and wtMRT into (5), analysis of the Y1 = (hRR t diag{1, 0, , 0}t hRR ) = |hRR,1 |2 , (24)
MRC/MRT scheme for arbitrary MT and MR appears to be where hRR = hRR t . Hence, MRC can be written as
cumbersome. Therefore, in the sequel we consider two special  
c1 |hSR |2 2 2
cases as follows: MRC = min , c3 |hSR | hRD  .
|hSR HRR hRD |2 c2 |hSR |2 |hRR,1 |2 + 1
Case-1) MT = 1, MR 1: In this case h SR hRD 
2 is
given by Let us dene X = c1 X1 / (c2 X1 Y1 + 1), and Y = c3 X1 Y2
| hSR,1 |2 where X1 = |hSR |2 , Y2 = hRD 2 . Note that conditioned on
X1  |wrMRC hRR |2 = hRR 2 . (20) X1 , the RVs, X and Y are independent and hence we have
hSR 2

P 2   
For notational convenience, we dene c1 = PdS , c2 = S d RR , FMRC (z) = 1 1FX|X1 (z) 1FY |X1 (z) fX1 (x)dx,
1 1 z
c3 = dPSd and write the end-to-end SINR as    
c1
1 2   
= 1 1e c2 x (
1 c1 x
z 1 ) Q M , z ex dx. (25)
c1 hSR 2 2 2 T
MRC = min , c3 hSR  |hRD | . z c3 x
c2 |hSR,1 |2 hRR 2 +1
c1

 
Let us denote X = c1 / c2 X1 + Y11 and Y = c3 Y1 Y2 , with
Y1 = hSR 2 and Y2 = |hRD |2 . Conditioned on Y1 , the
random variables (RVs) X and Y are independent and hence
D. Half-Duplex Scheme
We now present the outage probability of the HD relaying
system, which serves as a benchmark for performance com-
parison. The energy harvesting phase of the HD relaying is
   
1 c1 1 the same as that of the FD relaying system. However, in the
F MRC (z) = 1 F X1
z c2 z y information transmission phase, the remaining (1 ) portion
c1
   of block time is equally partitioned into two time slots for
z
1 FY2 fY1 (y)dy. (21) source and relay transmissions [5]. The end-to-end SNR of
c3 y the HD relaying scheme can be computed as
In order to evaluate (21) we require the cdf of the RV, X1 .  
|h |2 HD = hSR 2 min c1 , 2c3 hRD 2 . (26)
Note that hRR 2 22MR , Z  hSR,1 SR 
2 is distributed as

Z Beta(1, MR 1), and the cdf of X1 can be evaluated as The required cdf can be obtained by replacing MT 1 with
   MT in (16), and is given by
21  1, MR  
FX1 (t) = G23 t  , (22)  Q MT , d1 d2 z
1, MR , 0 21 x
  FHD (z) = 1 d z xMR 1 ex dx. (27)
where Gmnpq z | ab11 a
bq
p
is the Meijer G-function [13, Eq. P
1 (M R)
S
(9.301)]. Now, using the cdf of RV Y2 , and substituting (22)
into (21) we obtain V. N UMERICAL AND S IMULATION R ESULTS
    
1 1 c1 1  1, MR We now present numerical and simulation results to investi-
FMRC (z) = 1 G21 
(MR ) cz 23 c2 z y 1, MR , 0 gate the impact of key system parameters on the performance.
1
  In all cases, we have set = 1.
MR 1 y+ c2 y
z
y e dy. (23) Fig. 1 compares the outage probability of the considered
schemes with different antenna congurations and for = 0.5,
To the best of the authors knowledge, the integral in (23) 2
d1 = d2 = 1, RR = 0.1, and th = 0 dB. When the transmit
does not admit a closed-form solution. However, (23) can be
power is high and remains xed, an excessive amount of
evaluated numerically.
energy will be collected at the relay, which is detrimental for
|hSR HRR hRD |2
Case-2) MR = 1, MT 1: In this case hSR hRD 2 the MRC/MRT scheme since it results in strong LI. Therefore,

299
 2015 IEEE 16th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC)

MRC/MRT, (MR=1, MT =3) As expected, the optimal scheme exhibits the best throughput
0
10 TZF, (M =2, M =2)
R T
out of all precoding schemes studied. The superior perfor-
MRC/MRT, (M =3, M =1)
R T mance of the optimal scheme is more pronounced especially
between 0.4 and 0.8 values of . The highest throughput
1 RZF, (M =3, M =1)
10 R T
RZF, (M =3, M =3)
Outage Probability

2
R T
TZF, (MR=3, MT =3)
with optimized for the optimal, RZF, MRC/MRT and
10
MRC/MRT, (MR=3, MT =3) TZF schemes are given by 0.382, 0.374, 0.358 and 0.315,
3
Asymptotic respectively. Moreover, we see that each one of TZF, RZF and
10
MRC/MRT precoder designs can surpass the others depending
4 on the value of . Finally, all schemes achieve signicant
10
throughput gains as compared to the HD mode.
5
10
VI. C ONCLUSION
10
6
In this paper, we have studied the outage probability and
0 10 20 30 40 50 60
(dB)
1
throughput of FD MIMO relaying with RF energy transfer. We
Fig. 1. Outage probability versus 1 of the TZF, RZF and MRC/MRT
designed the optimal precoder/decoder as well as investigated
schemes for different antenna congurations. (th = 0 dB, and = 0.5). several sub-optimal low complexity precoding schemes. The
0.4
MRC/MRT scheme provides a better outage performance at
Optimal
RZF
low-to-medium SNR, while the ZF precoders outperform the
0.35 MRC/MRT former at high SNR. Further, it is demonstrated that all
TZF
HD proposed FD precoding schemes attain signicant throughput
Delayconstrained Throughput

0.3
gains compared to the HD mode. Therefore, FD relaying
0.25 is a promising solution for implementing future RF energy
harvesting cooperative communication systems.
0.2

0.15
0.1
0.05
0 vol. 59, pp. 5983-5993, Dec. 2011.
0.1 0.2 0.3
Fig. 2. Delay-constrained throughput of the proposed precoding schemes
2 Commun., vol. 10, pp. 3074-3085, Sept. 2011.
(MT = MR = 4, 1 = 10 dB, RR = 0.3, d1 = d2 = 2, = 3.1 and
Rc = 1 bit/sec/Hz). Results for the MRC/MRT scheme are from simulations.
the outage performance of the MRC/MRT scheme exhibits
an outage oor at high SNRs. On the contrary, the outage
probability of the TZF and RZF precoding schemes decays
proportional to the diversity orders provided in Proposition 1
and 2, respectively since the relay is capable of canceling LI.
However, a proper choice of can improve the outage of the
MRC/MRT scheme to some level at high SNR. On the other
hand observation of different curves in the low SNR region
reveal that the MRC/MRT scheme outperforms the TZF and
RZF schemes even at high LI level. This is because at low
SNR, overall interference can be treated as noise and therefore
MRC ltering helps to maximize the SNR. Comparing the
TZF and RZF schemes with the same diversity orders and
different receive antenna numbers (i.e., TZF, with MR = 2
and MT = 2, and RZF with MR = 3 and MT = 1) we see
that the additional receive antenna could harvest more energy
to facilitate information transfer. Moreover, for the case where
MR = MT , TZF achieves a higher array gain.
Fig. 2 shows the impact of optimal time split on the
delay-constrained transmission throughput dened as: R() =
(1 Pout )Rc (1 ), where = 1 for FD and = 0.5 for
HD and the optimal time portion can be obtained from
= arg max R(), 0 < < 1.
R EFERENCES
[1] A. Sabharwal et al., In-band full-duplex wireless: Challenges and
opportunities IEEE J. Sel. Areas Commun., vol. 32, pp. 1637-1652,
Sept. 2014.
[2] T. Riihonen, S. Werner, and R. Wichman, Mitigation of loopback self-
interference in full-duplex MIMO relays, IEEE Trans. Signal Process.,
0.4 0.5 0.6 0.7 0.8 0.9
[3] T. Riihonen, S. Werner, and R. Wichman, Hybrid full-duplex/half-
duplex relaying with transmit power adaptation, IEEE Trans. Wireless
[4] Z. Ding et al., Application of smart antenna technologies in simultaneous
wireless information and power transfer, IEEE Commun. Mag., vol. 53,
pp. 86-93, Apr. 2015.
[5] A. A. Nasir, X. Zhou, S. Durrani, and R. Kennedy, Relaying protocols
for wireless energy harvesting and information processing, IEEE Trans.
Wireless Commun., vol. 12, pp. 3622-3636, Jul. 2013.
[6] H. Chen, Y. Li, J. L. Rebelatto, B. F. Uchoa-Filho, and B. Vucetic,
Harvest-Then-Cooperate: Wireless-Powered Cooperative Communica-
tions, IEEE Trans. Signal Process., vol. 63, pp. 1700-1711 Apr. 2015.
[7] D. S. Michalopoulos, H. A. Suraweera, and R. Schober, Relay selection
for simultaneous information transmission and wireless energy transfer:
A tradeoff perspective, IEEE J. Select. Areas Commun., vol. 33, Sept.
2015.
[8] I. Krikidis, S. Sasaki, S. Timotheou, and Z. Ding, A low complexity
antenna switching for joint wireless information and energy transfer in
MIMO relay channels, IEEE Trans. Commun., vol. 62, pp. 1577-1587,
May 2014.
[9] C. Zhong, H. A. Suraweera, G. Zheng, I. Krikidis, and Z. Zhang,
Wireless information and power transfer With full duplex relaying,
IEEE Trans. Commun., vol. 62, pp. 3447-3461, Oct. 2014.
[10] G. Zheng, I. Krikidis, J. Li, A. P. Petropulu, and B. E. Ottersten,
Improving physical layer secrecy using full-duplex jamming receivers,
IEEE Trans. Signal Process., vol. 61, pp. 4962-4974, Oct. 2013.
[11] G. Zhu, C. Zhong, H. A. Suraweera, G. K. Karagiannidis, Z. Zhang, and
T. A. Tsiftsis,Wireless information and power transfer in relay systems
with multiple antennas and interference, IEEE Trans. Commun., vol. 63,
pp. 1400-1418, Apr. 2015.
[12] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions
With Formulas, Graphs, and Mathematical Tables. 9th ed. New York:
Dover, 1970.
[13] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and
Products. 7th ed. Academic Press, 2007.
[14] R. V. Hogg and A. T. Craig, Introduction to Mathematical Statistics. 4th
ed. Macmillan, New York, 1978.

300