See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328845536

Full-Duplex User Relaying for NOMA System with Self-Energy Recycling

Article  in  IEEE Access · November 2018

DOI: 10.1109/ACCESS.2018.2876859

CITATIONS READS

0 212

3 authors, including:

Xinwei Yue
Beijing Information Science and Technology University (BISTU)
21 PUBLICATIONS   163 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Non-Orthogonal Multiple Access View project

All content following this page was uploaded by Xinwei Yue on 12 November 2018.

The user has requested enhancement of the downloaded file.

Received September 4, 2018, accepted October 10, 2018. Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2018.2876859

Full-Duplex User Relaying for NOMA System

With Self-Energy Recycling
ZHENLING WANG 1, XINWEI YUE2 , AND ZHANGYOU PENG1
1 Key Laboratory of Specialty Fiber Optics and Optical Access Networks, Shanghai University, Shanghai 200444, China
2 School of Information and Communication Engineering, Beijing Information Science and Technology University, Beijing 100101, China

Corresponding author: Zhangyou Peng (zypeng@mail.shu.edu.cn)

ABSTRACT This paper proposes a self-energy recycling full-duplex cooperative non-orthogonal multiple
access (SRF-NOMA) system, where a nearby user can be employed as a decode-and-forward, amplify-and-
forward, and quantize-map-forward relay with self-energy recycling protocol to assist a distant user. The
relay harvests energy from dedicated energy signal sent by a base station, while it reuses energy from loop
self-interference (LI) signal. To characterize the performance of the system proposed, new expressions of
exact and asymptotic outage probabilities for two users are derived. Based on analytical results, the diversity
orders achieved by the nearby and distant user are one and two, respectively. We confirm that the diversity
orders of the distant user are obtained with direct link and relaying link, since the energy from LI is harvested
for information transmission at the relay. Numerical results are presented to validate the accuracy of the
derived results and demonstrate that: 1) when the DF or AF protocol is employed, the outage behavior of
the SRF-NOMA system is superior to orthogonal multiple access; 2) in the high signal-to-noise ratio region,
DF relay in the SRF-NOMA system is capable of enhancing the outage performance compared to AF relay;
and 3) the distant user in the SRF-NOMA system with QMF relay is able to obtain a better performance than
that of DF and AF relays.

INDEX TERMS Full-duplex, non-orthogonal multiple access, self-energy recycling, user relaying.

I. INTRODUCTION
With the rapid increase in demand for the Internet of Things,
the fifth generation (5G) networks have been designed
to increase spectrum efficiency and system throughput in
large-scale heterogeneous data traffic. Non-orthogonal mul-
tiple access (NOMA) is one of the promising techniques,
which is capable of enhancing the spectrum efficiency for
the 5G networks [1]–[3]. Compared to orthogonal multiple
access (OMA), the key characteristic of NOMA is that mul-
tiple users are multiplexed in the same time/frequency/code
domain with different power domain. By using superposition
coding at the transmitter and successive interference cancel-
lation (SIC) at the receivers [4], the NOMA scheme is able to
achieve better performance than OMA.
At present, many contributions have investigated the down-
link NOMA systems extensively [5]–[7]. In [5], the outage
behavior and ergodic rate of NOMA system with randomly
deployed users was investigated. By assuming imperfect
channel state information (CSI) and second order statistics,
Yang et al. [6] analyzed the performance of a downlink
single-cell NOMA network. The outage performance of a
NOMA system was researched, where each user only feed-
back one bit of its CSI to a base station (BS) [7]. Compared
with conventional opportunistic one-bit feedback, NOMA is
capable of improving the fairness for multiple users. From
the perspective of security scenario, the authors investigated
the physical layer security of NOMA networks under single-
antenna and multiple-antenna aided transmission scenarios
in [8]. A new framework named NOMA-assisted multi-
region geocast was introduced [9], in which the NOMA tech-
nology was exploited to realize the synchronous transmission
of different information to different user groups.
NOMA has been extended to cooperative communication
systems, as the higher diversity and extended coverage can
be obtained in wireless networks [10]. Cooperative NOMA
scheme was first proposed to enhance the system reliability
by transiting information to distant users with the assistant of
nearby users [11]. NOMA was employed in cooperative net-
works with amplify-and-forward (AF) relay [12], [13], where
the performance was characterized in terms of outage proba-
bility and ergodic sum capacity. In [14], the outage behaviors
of cooperative NOMA system with partial CSI were studied

2169-3536
2018 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 6, 2018 Personal use is also permitted, but republication/redistribution requires IEEE permission. 1
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling
when both decode-and-forward (DF) and AF protocols were
adopted at the relay. In addition to AF and DF relay protocols,
quantize-map-forward (QMF) protocol was considered for
relaying and the outage behaviors of cooperative NOMA was
analyzed in [15]. However, the above works mainly based on
half-duplex (HD) relay technology, which are inefficient in
spectrum utilization. To overcome this issue, full-duplex (FD)
relay technique has been employed in wireless communica-
tion systems to obtain higher spectral efficiency [16], since
it allows simultaneous reception and transmission of signals
over the same frequency band. The outage performance of
the FD device-to-device-aided cooperative NOMA system
was investigated in [17]. The analysis results were shown
that the proposed FD cooperative NOMA system outper-
formed the conventional NOMA and OMA systems. In [18],
the performance of user relaying for FD/HD NOMA systems
was analyzed in consideration of self-interference (LI). Addi-
tionally, the FD NOMA system with a dedicated relay was
proposed and outage behaviors of the system was analyzed
over Nakagami-m fading in [19]. Furthermore, a FD NOMA
scheme in [20] was presented for a cooperative relay sharing
network, i.e., a dedicated FD relay was shared by two source-
destination pairs.
In addition to enhance spectral efficiency, it is significant
to improve the energy efficiency for future 5G networks.
In [21], simultaneous wireless information and power trans-
fer (SWIPT) has been first proposed, which encourages a
receiver to decode signal and harvest energy at the same time.
To achieve SWIPT in practical scenarios, time switching
(TS) and power splitting (PS) receiver architectures were
proposed in [22]. By exploiting both the TS and PS receiver
architectures, most researches on cooperative SWIPT net-
works adopted either TS based relaying or PS-based relaying
protocols [23]. In [24], the achievable throughput and optimal
time split of the FD SWIPT system with TS relaying protocol
were studied under three different communication modes.
Besides, a cooperative SWIPT system based on PS relay-
ing protocol was investigated in [25], where the error rate
performance was analyzed. At present, the researches on
cooperative NOMA with SWIPT system were almost based
on the TS and PS relaying protocols. However, the exit-
ing TS and PS relaying protocols have the disadvantage
of interrupted information transmission, since TS or PS is
needed at the relay for energy harvesting. To avoid this
issue, a new self-energy recycling protocol was proposed
and employed in the FD SWIPT system [26]. Since NOMA
and SWIPT is capable of improving the spectral efficiency
and energy efficiency of cooperative networks, respectively,
it is natural to combine NOMA and SWIPT in cooperative
networks. SWIPT was introduced to NOMA networks [27],
in which the closed-form expressions of the outage probabil-
ity and system throughput were derived. In [28], the outage
performance of the cooperative SWIPT-NOMA system was
investigated under fixed power allocation and cognitive radio
inspired allocation based NOMA. With utilizing SWIPT in
2 VOLUME 6, 2018
a cooperative FD NOMA system [29], the impact of the
LI signal in FD communications was mitigated.
A. MOTIVATION AND CONTRIBUTIONS
While the aforementioned research contributions laid a solid
foundation for the understanding the cooperative NOMA
and SWIPT techniques, the SWIPT technique for cooper-
ative NOMA is still under exploration. Some researches
on cooperative NOMA and SWIPT have been investigated
in [18] and [26], respectively. In [26], a new self-energy recy-
cling protocol was proposed in the FD SWIPT system, where
the self-energy recycling relay not only harvests energy from
the dedicated energy signal sent by the BS, but also reuses
the energy from LI signal. As the harvested energy is used
to support relay for information transmission, the harmful
LI signal in FD communications becomes beneficial. How-
ever, the self-energy recycling protocols has not been intro-
duced in NOMA system. In [18], the performance of user
relaying for FD/HD NOMA systems had been investigated
in detail, but the user relaying with self-energy recycling
for NOMA system was not considered. To the best of our
knowledge, there are no existing works to investigate the self-
energy recycling for FD cooperative NOMA (SRF-NOMA)
networks. Driven by these, we specifically consider
a SRF-NOMA system, where a nearby user is employed as a
self-energy recycling relay with FD mode to assist a distant
user. More specifically, the performance of our proposed
system is characterized when implementing DF, AF and QMF
protocols at the relay, respectively. The main contributions of
our paper are summarized as follows:
1) We derive new expressions of exact outage probability
at the nearby user and distant user, when AF, DF and
QMF protocols are carried out in proposed system.
We study the system throughput in delay-limited trans-
mission based on the probability derived of the two
users.
2) We derive new expressions of asymptotic outage prob-
ability at nearby user and distant user. Based on the
analytical results, we acquire diversity orders of two
users, which are one and two, respectively. We confirm
that the diversity orders of distant user are obtained
with direct link and relaying link.
3) We conclude that the outage behavior of SRF-NOMA
system with DF or AF relay is superior to OMA-based
system. The SRF-NOMA system with DF relay is
capable of achieving better performance compare to
AF relay in the high signal-to-noise ratio (SNR).
In addition, the distant user with QMF relay is able
to obtain better outage performance than that of
DF and AF relays. Furthermore, we also confirm that
the outage performance of the system becomes better
with increasing the values of energy conversion effi-
ciency and LI. The LI becomes beneficial, as the energy
from LI signal is harvested for information transmis-
sion at the relay.
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

where a1 +a2 = 1 and a2 > a1 .1 x1 and x2 are the normalized

signal with unit power of D1 and D2 , respectively. Hence the
observation at D1 and D2 are given by
p p 
y1,D1 = h1 a1 Ps x1 + a2 Ps x2 + n1,D1 , (1)

and
p p 
y1,D2 = h2 a1 Ps x1 + a2 Ps x2 + n1,D2 , (2)

respectively, where n1,D1 ∼ CN (0, N0 ) and n2,D2 ∼

FIGURE 1. A self-energy recycling FD cooperative NOMA system model.
CN (0, N0 ) are the additive white Gaussian noise (AWGN)
at D1 and D2 , respectively.
The SIC scheme is first employed at D1 for detecting
B. ORGANIZATION and decoding D2 ’s information. Hence the received signal-
The rest of this paper is organized as follows. In Section II, to-interference-plus-noise ratio (SINR) at D1 to detect x2 is
the system model of a FD user relaying with self-energy given by
recycling NOMA network is presented. In Section III, the per- |h1 |2 a2 ρ
formance of the system is evaluated based on AF, DF and γD2 →D1 = , (3)
|h1 |2 a1 ρ + 1
QMF relaying protocols, which includes the outage prob-
ability of users, diversity analysis and system throughput. where ρ = NP0s is transmit SNR. After decoding D2 ’s message
Numerical results are provided in Section IV and conclusions and subtracting it, D1 is further to detect its own information
of this paper is presented in Section V. with the following SINR
γ1,D1 = |h1 |2 a1 ρ. (4)
II. SYSTEM MODEL
Consider a SRF-NOMA system, where one BS transmits The received SINR at D2 to detect its own information x2
information to the distant user D2 by the virtue of nearby with the direct link is given by
user D1 . As shown in Fig. 1, D1 can be employed as a DF,
|h2 |2 a2 ρ
AF and QMF relay, which operates in FD mode with self- γ1,D2 = . (5)
energy recycling. D1 is equipped with one transmit antenna |h2 |2 a1 ρ + 1
and one receive antenna. The transmit antenna can be acti- Different from the general situation, when D1 employs the
vated for information transmission and the receive antenna QMF protocol, D1 first decodes its own information x1 with
for information/energy reception. In addition, the direct link the following SINR
between the BS and D2 exists. The BS and D2 are equipped
one antenna, respectively. The complex channel coefficient of QMF |h1 |2 a1 ρ
γ1,D1 = , (6)
BS → D1 , BS → D2 and D1 → D2 links are denoted as h1 , |h1 |2 a2 ρ + 1
h2 and h0 , respectively. The corresponding channel coeffi- and then subtracting it from the superposed signal. Hence the
cient are assumed to be independent Rayleigh fading channel SINR at D1 to detect the signal x2 is given by
with average powers of 1 , 2 and 0 , respectively. Without
loss of generality, the LI channel is assumed to be Rayleigh γ1,D
QMF
2
= |h1 |2 a2 ρ. (7)
fading, i.e., hLI ∼ CN (0, LI ). Note that the total commu-
nication process consists of two phases and the duration of B. THE SECOND PHASE
each phase is T /2, where T denotes the time of the whole In the second phase, D1 is wireless powered by the dedicated
communication process. In the first phase, the BS transmits energy signal from the BS while the LI signal can also be
the two users’s superposed information to D1 and D2 , respec- reused as the transmitting power. Based on the above descrip-
tively. In the second phase, D1 transmit information to D2 , tion, the observation at D1 is given by
and concurrently, D1 not only harvests energy from dedicated p p
energy signal sent by the BS, but also recycles energy from y2,D1 = h1 Ps xe + hLI Pr xLI + n2,D1 , (8)
LI signal. The harvested energy is capable of supporting D1 where xe is normalized signal with unit power transmitted by
to transmit the information. the BS, Pr is the power transmitted of D1 , xLI denotes the
LI signal which is supposed to be normalized unity
A. THE FIRST PHASE power signal. n2,D1 ∼ CN (0, N0 ) is the AWGN at D1 .
On the basis √
of NOMA protocol,
√ the BS transmits the super-
1 In this paper, we assume that the power allocation coefficients are fixed
posed signal a1 Ps x1 + a2 Ps x2 to D1 and D2 , respectively,
values. More efficient power allocation algorithms [28], [30] will be inves-
where Ps is the transmission power for the BS. a1 and a2 are tigated to improve system performance and energy efficiency in our future
the power allocation coefficients of D1 and D2 , respectively, work.

VOLUME 6, 2018 3
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

Assuming that the energy harvested from the noise can be 1) OUTAGE PROBABILITY OF D1
ignored [26] and the harvest energy is given by Based on NOMA protocol, the outage events at D1 would
not occur when D1 can detect D2 ’s information and also can
T  
E= η Ps |h1 |2 + Pr |hLI |2 , (9) detect its own information in the first phase. The D1 ’s outage
2 probability PDF can be obtained from [18, eq. (12)].
D 1
where η is the energy conversion coefficient. The transmitted
power TE/2 at D1 can be expressed as 2) OUTAGE PROBABILITY OF D2
In the proposed system, the signals from the relaying link and
ηPs |h1 |2 direct link are combined by selection combining at D2 . The
Pr = , (10) outage events of D2 can be described as follows. One of the
1 − η|hLI |2
events is that neither D1 nor D2 can detect x2 . Another event
where |hLI |2 < η1 . is when x2 can be detected at D1 , but cannot be detected at D2
Additionally, D1 forwards the information to D2 . Based with direct link and relaying link. The outage probability
on the different relay protocols employed at D1 , DF, AF and of D2 can be expressed as
QMF protocols are discussed in the following parts.
PDF = Pr γD2 →D1 < γth2 , γ1,D2 < γth2


D2

+ Pr γD2 →D1 > γth2 , γ1,D2 < γth2 , γ2,D2 < γth2 ,


1) DF RELAYING CASE
D1 decodes the information of D2 from the received super- (15)
posed signal and forwards it to the D2 . The observation at D2
is given by where γth1 = 22R1 − 1 with R1 being target rate at D1 to
p detect x1 and γth2 = 22R2 − 1 with R2 being target rate at D1
y2,D2 = h0 Pr x2 + n2,D2 , (11) to detect x2 .
As the complexity of (15), the closed-form expression can-
where n2,D2 ∼ CN (0, N0 ) is the AWGN at D2 . Hence, not be derived successfully, but it can be used for numerical
the SINR for D2 to detect x2 is given by simulation. Substituting (3), (5) and (12) into (15), the outage
probability of D2 can be given as
ηρ|h1 |2 |h0 |2
γ2,D2 = . (12) 
− τ

− 1 − 2
γth LI
1 − η|hLI |2 DF
PD = 1 − e 2 1 − e ηLI ρ0 1
2

γth2 LI τ γth2 LI

 
2) AF RELAYING CASE
× Ei − +
D1 is amplify and forward the superposed signal to D2 . Hence ρ0 1 1 ρ0 1
the received signal at D2 is given by γth
#
ρ0 z
Z ∞
− z − ηρz2
−
e 1 0 dz ,
√ p p 
τ 1 ρ0 z − γth2 LI
yAF
2,D2
= Gh0 h1 a1 Ps x1 + a2 Ps x2
√ (16)
+ Gh0 n1,D1 + n2,D2 , (13)
γth
where τ = a - γ 2 a
ρ with a2 > a1 γth2 . Ei(·) is the expo-
2 th2 1
ηρ|h 2

1|
where G =
is the amplification gain. The nential integral function [31, eq. (8.211.1)]. Note that (16) is
1−η|hLI |2 ρ|h1 |2 +1
SNR expression for D2 can be written as (14), as shown at the derived on the condition of a2 > a1 γth2 , otherwise PDF
D2
= 1.
top of the next page. Proof: See Appendix A.

3) QMF RELAYING CASE 3) DIVERSITY ANALYSIS

D1 first decodes its own information x1 and then subtracts it To gain more insights, the outage probability in high
from the superposed signal in the first phase. The remaining SNR region is investigated in this section, and the diversity
signal will be quantize-map-forwarded and sent to D2 in order achieved by the users can be obtained based on the
the second phase. The received signal at D2 and the SINR above analytical results. The diversity order [18] is defined
for D2 to detect x2 are the same as (11) and (12), respectively. as
log (P(ρ))
III. PERFORMANCE EVALUATION d = − lim . (17)
ρ→∞ log ρ
In this section, the performance of the SRF-NOMA system is
characterized in terms of outage probability. a: Diversity order of D1
When ρ → ∞, the asymptotic outage probability of D1 can
A. DF RELAYING CASE be obtained from [18, eq. (20)]. The diversity order of D1 is
In this subsection, the performance of the proposed system equal to one, which is the same as D1 in user relaying HD
with DF protocol is investigated. NOMA system.

4 VOLUME 6, 2018
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

|h0 |2 |h1 |4 a2 ηρ 2
γ2,D
AF
=
. (14)
|h0 |2 |h1 |2 ηρ a1 ρ|h1 |2 + 1 + 1 − η|hLI |2 ρ|h1 |2 + 1



2

b: Diversity order of D2 In our considered system, the BS transmits information at a

Based on (16), when ρ → ∞, the asymptotic outage proba- constant rate R, which is subject to the effect of outage proba-
bility of D2 can be given by bility due to wireless fading channels. The system throughput
for DF relaying case is given by
τ − η1 γth2 LI γth2 LI
  
DF,∞
PD = 1−e LI 1−
2 ρ0 1 ρ0 1
   
RDF = 1 − PDF D1 R1 + 1 − PD2 R2 ,
2 DF
(21)
τ γth2 LI
   
× ln − +C
1 ρ0 1 where PDF DF
D1 and PD2 are given in [18, eq. (12)] and (16)
γth
#
ρ0 z
Z ∞
− z − ηρz2 respectively.
−
e 1 0 dz ,
τ 1 ρ0 z − γth2 LI
B. AF RELAYING CASE
(18)
In this subsection, the performance of the proposed system
where (18) is derived by using e−x ≈ 1 − x and Ei (−x) ≈ with AF protocol is investigated.
ln (x) + C [31, eq. (8.212.1)], and the C is the Euler
constant. 1) OUTAGE PROBABILITY OF D1
To further obtain the theoretical results, assuming Based on NOMA protocol, D1 first performs SIC to
0  LI , the outage probability of D2 at high SNR can decode D2 ’s signal. After decoding D2 ’s signal successfully
be further approximated as and subtracting it, D1 will decodes its own signal in the first
phase. If the information of D1 and D2 is decoded success-
τ − η1 γth2 LI γth2 LI
  
PDF,∞ = −e LI 1 − fully, the outage event will not occur. Therefore, D1 ’s outage
D2 ,appro
2 ρ0 1 ρ  probability PAF
 0 1 is also the same as [18, eq. (12)].
τ γth2 LI D
   1
× ln − +C
1 ρ0 1 2) OUTAGE PROBABILITY OF D2
γth2 γth2
 
− ln + 2c0 , (19) The outage events will occur when the received SINR at D2
ηρ0 1 ηρ0 1 with direct link and relaying link is both less than its tar-
get SINR. The outage probability of D2 can be expressed as
where c0 = − ϕ(1) ϕ(2)
2 − 2 and ϕ (·) denotes psi function  
in [31]. PAF = Pr γ < γ , γ AF
< γ . (22)
D 1,D 2 th 2 2,D th2
Proof: See Appendix B. 2 2

Substituting (19) into (17), we can obtain dDDF = 2.

2 Similar to the DF relaying protocol case, the closed-form
Remark 1: The diversity gains of D1 and D2 for the
expression of (22) cannot be obtain. Substituting (5) and (14)
SRF-NOMA system in DF relaying case are one and two,
into (22), the outage probability can be written as
respectively, which are the same as the user relaying HD
NOMA system. AF

− τ
 
− η1

− τ
PD = 1 − e 2 1−e LI e 1
2
4) SYSTEM OUTAGE PROBABILITY
z2 − zτ
Z ∞
1
Based on the above analysis, the system outage events would +
1 τ z2 − zτ 1 + LI − LI τ
 
not occur when no outage events occur at D1 and D2 in 0 0 ρ
the whole transmission process. Therefore the system outage τ (z+1/ρ)
! #
z 1 −
× e 1 e ηLI − e η0 (z −zτ ) dz . (23)
− − 2
probability can be expressed as
  
PDF
sys = 1 − 1 − PD
DF
1 − PDF
D
, (20)
1 2
Note that (23) is derived on the condition of a2 < a1 γth2 ,
where D PDF and PDF are given in [18, eq. (12)], otherwise PAF
D2
= 1.
1 D2
respectively. Proof: See Appendix C.

5) THROUGHPUT ANALYSIS 3) DIVERSITY ANALYSIS

In this section, the throughput will be considered in a: Diversity order of D1
delay-limited transmission [24], [27] for DF relaying In AF relaying case, the diversity of D1 is one, which is the
case. same as in DF relaying case.

VOLUME 6, 2018 5
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

b: Diversity order of D2 Substituting (6) into (28), the closed-form expression for the
When ρ → ∞, the asymptotic outage probability of D2 can outage probability of D1 can be obtained as
be given by QMF − σ
PD1 =1−e 1 , (29)
τ τ τ LI τ LI
   
PDAF,∞
= 1− 1− 1− γth1
2 2    0 1 where σ = a1 −a2 γth1 ρ

and (29) is derived on the condition
 1 0 1 
− η1 τ LI of a1 > a2 γth1 , otherwise PD1
QMF
= 1.
×e LI ln +C
0 1
0 ε
Z ∞
− ε − τ 2) OUTAGE PROBABILITY OF D2

− e 1 η0 ε dε .
0 1 (0 ε − τ LI ) The outage events will occur as follows. The first is that
(24) D1 cannot decode its own information and D2 cannot decode
its signal with the direct link. The second is that D1 is able to
Proof: See Appendix D. detect its own signal, but D2 is unable to decode its signal with
To further obtain the theoretical results, assuming the direct link and relay link. Based on the above analysis,
0  LI , the outage probability of D2 at high SNR can the outage probability of D2 can be written as
be further approximated as  
QMF
PQMF = Pr γ1,D1 < γth1 , γ1,D2 < γth2
τ τ τ LI τ LI
    D2
AF,∞
PD ,appro = 1− − 1− 
QMF

2 2  0 1 0 1 + Pr γ1,D1 > γth1 , γ1,D2 < γth2 , γQMF < γth2 ,
 1
τ LI
 
− η1 (30)
×e LI ln +C
0 1
τ

τ
 where γQMF is the SINR for quantize-map-forward relay link,
− ln + 2c0 . (25) and according to [32, eq. 10] γQMF can be obtained as follows
η0 1 η0 1
γ1,D
QMF
γ
Substituting (25) into (17), we can obtain dDAF = 2. γQMF = 2 2,D2
. (31)
2
Remark 2: The diversity gains of D1 and D2 in AF relaying 1 + γ1,D
QMF
2
+ γ2,D2
case are also one and two, respectively, which is the same as
Assuming γ1,D QMF
2
+ γ2,D2  1 and substituting (7) and (12)
in DF relaying case.
into (31), the γQMF can be approximated as
4) SYSTEM OUTAGE PROBABILITY a2 ηρ|h1 |2 |h0 |2
γQMF ≈ . (32)
Similar to (20), the system outage probability for AF relaying a2 − ηa2 |hLI |2 + η|h0 |2
case can be expressed as Substituting (5), (6) and (32) into (30), after some algebraic
  
PAF = 1 − 1 − P AF
1 − P AF
, (26) manipulations the outage probability of D2 can be rewritten
sys D D
1 2 as
PAF PAF
 h  
where D and D2
are given in [18, eq. (12)] and (23), PQMF = Pr |h |2
< τ Pr |h |2
< σ
1 D2 2 1
respectively. 

a γ
2 th2 1 − η|h LI |2
5) THROUGHPUT ANALYSIS + Pr |h1 |2 > σ, |h0 |2 <  γth2
 ,
η |h1 | − a2 ρ
2
Similar to (21), the system throughput for AF relaying case
is given by (33)
where (33) is derived on the condition of a1 > a2 γth1 . Similar
   
RAF = 1 − PAF D1 R1 + 1 − PD2 R2 ,
AF
(27)
to the derivation process of (16), we can calculate (33) as
where PAF AF
D1 and PD2 are given in [18, eq. (12)] and (23)
follows.
respectively. When γth2 < a2 ρσ , after some algebraic manipulations,
(33) can be calculated as
γth
C. QMF RELAYING CASE  ϑ −ϑ

− τ − a ρ2

In this subsection, the system performance with QMF proto- PQMF = 1 − e 2 1 + e 2 1 − e 1
D2
1
col is investigated in detail. 
σ γth2 ϑ

− η1
×e LI Ei − + +
1 a2 ρ1 1
1) OUTAGE PROBABILITY OF D1
γth
#)
δ
Z ∞ δ 2
The outage events at D1 will be occur when the SINR of −
−  −  ηρδ
e 1 0 dδ , (34)
γth
D1 decoding itself is less than its target SINR. Therefore the σ − a ρ2 1 (δ − ϑ)
2
outage probability of D1 in this case can be expressed as
γth LI
QMF

QMF
 where ϑ = 2
ρ0 . Note that (34) is derived
PD1 = Pr γ1,D1 < γth1 . (28) on [31, eq. (3.352.2)].

6 VOLUME 6, 2018
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

When γth2 > a2 ρσ , similar to (34) we can obtain where the derivation process of (38) is similar
  γ to (19).
− τ − σ − th2
 
PQMF
D2
= 1 − e 2 1 − e 1 + e a2 ρ1 When ρ → ∞ and γth2 > a2 ρσ , the asymptotic outage
probability of D2 is given by
ϑ − η1 − ϑ ϑ
  
× 1− e LI 1 Ei
τ σ γth2
  
 1 PDQMF,∞
= + 1−
Z ∞ 1 γth 2 1 a2 ρ1
δ − δ −  ηρδ
2

2
− e 1 0 dδ . (35) 
ϑ

ϑ

− 1

ϑ

0 1 (δ − ϑ) × 1− 1− e ηLI ln +c
 1 
Remark 3: From the above analysis, it can be concluded Z ∞ 1 γth  1
δ − −δ 2
as follows: the outage performance of D2 with QMF relay is − e 1 0 ηρδ dδ , (39)
better than that of DF and AF relays when a2 < a1 γth2 and 0 1 (δ − ϑ)
a1 > a2 γth1 . This conclusion can be explained as follows. where (39) is derived on Ei (x) ≈ ln (x) + C
When a2 < a1 γth2 , the outage probabilities of D2 with DF
[31, eq. (8.214.2)]. Assuming 0  LI , the outage proba-
and AF relays are one. Similarly, the first item Pr |h2 |2 < τ bility of D2 at high SNR can be further approximated as
in (33) is one. However, the remaining part in (33) will be
γth τ σ γth2
  
equal to one only when |h1 |2 − a2 ρ2 ≤ 0 for all values QMF,∞
PD ,appro = + 1−
2 2 1 a2 ρ1
|h1 |2 > σ . Both conditions cannot be achieved at the same
ϑ ϑ ϑ
    
time, therefore the probability of D2 wtih QMF relay is less − 1
× − 1− e ηLI ln +c
than one. 1  1
 1
γth2 γth2

− ln + 2c0 , (40)
3) DIVERSITY ANALYSIS 0 1 ηρ ηρ0 1
a: Diversity order of D1 where the derivation process of (40) is similar to (25).
Based on the analytical result in (31), when ρ → ∞, with Substituting (38) and (40) into (17), respectively, we can
e−x ≈ 1 − x the asymptotic outage probability of D1 is given obtain dDQMF = 2.
2
by Remark 4: The diversity gains of D1 and D2 in QMF
QMF,∞ σ 1 relaying case are one and two, respectively. The diversity
PD1 = ∝ , (36)
1 ρ gains of D1 and D2 for the three relaying protocols are the
where ∝ represents ‘‘be proportional to’’. same.
Substituting (36) into (20), we can obtain dDQMF = 1. The
1 4) SYSTEM OUTAGE PROBABILITY
diversity of D1 is one in QMF relaying case, which is the same
as in the DF and AF relaying case. Similarly to III-A4, the system outage probability for the
QMF relaying case can be expressed as
b: Diversity order of D2
  
PQMF = 1 − 1 − PQMF
1 − P QMF
, (41)
When ρ → ∞ and γth2 < a2 ρσ . Similar to the derivation sys D 1 D 2

process of (18), the asymptotic outage probability of D2 at where

QMF
PD1 and
QMF
PD2 are given in (29) and (34)/(35),
high SNR is given by respectively.
τ γth2 ϑ ϑ
    
PQMF,∞ = 1 + 1 − − 1 −
D2
2 a ρ 1  5) THROUGHPUT ANALYSIS
  2 1  1
− 1 σ γth2 ϑ Similarly, the system throughput for QMF relaying case is
× e ηLI ln − − +C given by
1 a2 ρ1 1
γth
#)    
δ
Z ∞ QMF QMF
− δ − 2 RQMF = 1 − PD1 R1 + 1 − PD2 R2 , (42)
− γth e 1 0 ηρδ dδ . (37)
σ− a ρ 2  1 (δ − ϑ)
2 QMF QMF
where PD1 and PD2 are given in (29) and (34)/(35),
Assuming 0  LI , the outage probability of D2 at high respectively.
SNR can be further approximated as As shown in TABLE 1, the diversity orders of two users
τ γth2 γth2
  
QMF,∞ for the SRF-NOMA system with different relaying protocols
PD ,appro = + 1−
2  a2 ρ1 a2 ρ1 are summarized, in order to illustrate the comparison between
2
ϑ ϑ them. In TABLE 1, we assume that ‘‘D’’ represents the diver-
 
− 1
− 1− e ηLI sity order.
 1
 1
σ γth2 ϑ
 
× ln − − +C IV. NUMERICAL RESULTS
1 a ρ 1
 2 1 In this section, numerical results are presented to validate
γth2 γth2

− ln + 2c0 , (38) the performance of the SRF-NOMA system. Without loss
ηρ0 1 ηρ0 1 of generality, the distance between BS and D2 is assumed
VOLUME 6, 2018 7
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

TABLE 1. Diversity orders for SRF-NOMA system.

to be unity and d denotes the normalized distance from BS

to D1 . Hence assuming 2 = 1, 1 = d −α and 0 =
(1 − d)−α with d = 0.3 and α = 2, where α is the path loss
exponent. The power allocation coefficients of D1 and D2 for
NOMA system are a1 = 0.2 and a2 = 0.8, respectively. The
performance of OMA is selected to be as a benchmark for
comparison and the total communication process of OMA is
completed in three time slots. In the first slot the BS sends FIGURE 2. Outage probability versus SNR for DF and AF protocols. R1 = 3
and R2 = 0.3 bit per channel use (BPCU); E{|hLI |2 } = −10 dB and η = 0.8.
information x1 to D1 and in the second slot sends x2 to D2 .
In the last slot, D1 forwards information to D2 , meanwhile
D1 harvests the energy from dedicated energy signal xe and
LI signal xLI .

A. OUTAGE PERFORMANCE WITH FIXED VALUES

OF η AND LI
In this subsection, the outage probability of the two users for
the DF, AF and QMF protocols is demonstrate. In the follow-
ing simulation, η and LI are set to be η = 0.8, E{|hLI |2 } =
−10 dB, where E{·} denotes expectation operation.
Fig. 2 plots the outage probability of the two users ver-
sus SNR for DF and AF protocols. The black and red
curves represent the outage performance based on DF and
AF protocols, respectively. The exact theoretical curves of
D1 and D2 for the two protocols are plotted according
to [18, eq. (12)] and (16), (23), respectively. Obviously,
the Monte Carlo simulation verifies the accuracy of our
derivation. It is shown that the outage performance of NOMA FIGURE 3. Outage probability versus SNR for QMF protocol. R1 = 0.1 and
outperforms OMA. We can observed that D2 is able to achieve R2 = 0.3 BPCU; E{|hLI |2 } = −10 dB, η = 0.8.

better outage performance when AF relay is employed in the

low SNR region. The reason is that x2 cannot be correctly
decoded at DF relay in low SNR region and hence no for-
warding to D2 ; However the superposition signal is directly
amplified and forwarded to D2 by AF relay. Another obser-
vation is that the outage performance of D2 for DF protocol is
better than that of AF protocol in high SNR region. This is due
to the fact that the information of D1 has been severe inter-
ference when D2 decodes its own information with relaying
link in AF relaying case. Furthermore, the asymptotic outage
probability curves of D1 and D2 for the two protocols are
plotted according to [18, eq. (20)] and (18), (24), respectively.
As can be observed from the figure, the asymptotic curves
approximate the exact curves well in the high SNR region.
Fig. 3 plots the outage probability of the two users versus
SNR for QMF protocol. The exact theoretical curves are for
the outage probability of the two users corresponding to the
analytical results (29) and (34), respectively. The asymptotic
outage probability curves of two users are plotted correspond-
ing to the derivation (36) and (37), respectively. Similar to
Fig. 2, it is clear that the exact theoretical curves are consistent
with the Monte Carlo simulation and the asymptotic curves
approximate the exact curves well in the high SNR region.
Additionally, one can observe that the outage performance
of D1 for FD OMA exceeds FD NOMA. The main reason
is that in QMF protocol, D1 needs to decode itself first in
the first phase, hence the signal of D2 has been regarded
as interference, which causes the outage performance of D1
to deteriorate. We also observed that the outage behavior
of D2 for FD OMA is superior to FD NOMA in low SNR
region. This is because D1 cannot decode its own information
successfully in the low SNR region, and then nothing is
forwarded to D2 with the relaying link in the second phase.
Fig. 4 plots the outage probability versus SNR for DF and
AF protocols with different target rates. The black and red

8 VOLUME 6, 2018
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling
FIGURE 4. Outage probability versus SNR for DF and AF protocols with
different target rates. E{|hLI |2 } = −10 dB, η = 0.8.
FIGURE 5. Outage probability versus SNR for QMF protocol with different
target rates. E{|hLI |2 } = −10 dB, η = 0.8.
curves denote the outage performance for DF and AF pro-
tocols, respectively. The figure shows that the outage proba-
bility of D1 decreases as R1 decreases. This is due to the fact
that decreasing R1 will make the threshold of decoding lower,
which leads to less outage. Similarly, the outage probability
of D2 also decreases as R2 decreases. Besides, we see that
the outage performance of D2 is not affected by adjusting
D1 ’s target rates, which validates our analytical results (15),
(16) and (22), (23). An important observation is that all the
curves of D1 and D2 have the same slopes, respectively, which
indicates that the diversity gains of D1 and D2 for the two
protocols are the same, respectively. As shown in Remark 1
and Remark 2, the diversity gains of D1 and D2 for the two
protocols are one and two, respectively.
As shown in Fig. 4, Fig. 5 plots the outage probability
versus SNR for QMF protocol with different target rates.
It is observed that the outage performance of D2 deteriorates
VOLUME 6, 2018
FIGURE 6. Outage probability of D2 versus SNR for DF, AF and QMF
protocols, when a2 < a1 γt h and a1 > a2 γt h . R1 = 0.1 and
2 1
R2 = 2 BPCU. E{|hLI |2 } = −10 dB, η = 0.8.
as R1 and R2 increase, since increasing R1 and R2 will make
the threshold of decoding higher. The figure also demon-
strates that the outage probability of D1 increases as R1
increases, whilst outage performance of D1 is not affected by
adjusting D2 ’s target rate. As in the first phase D1 needs to
decode itself first for QMF protocol. It verifies the derivation
(28) and (29). Similar to Fig. 4, all the curves of D1 and D2
have the same slopes, respectively, i.e., the diversity gains
of D1 and D2 with different target rates are the same, respec-
tively. In Remark 4, we have concluded that the diversity
gains of D1 and D2 for QMF protocols are one and two,
respectively. Based on the analysis of Fig. 4 and Fig. 5, it is
necessary to set the appropriate target rate for NOMA users
in different application scenarios.
Fig. 6 plots the Outage probability of D2 versus SNR
for DF, AF and QMF protocols, when a2 < a1 γth2 and
a1 > a2 γth1 . The black, red and green curves denote the
outage probability of D2 for DF, AF and QMF protocols,
respectively. The black and red curves coincide, which shows
that the probability of D2 for DF and AF protocols is one.
Based on the above analysis, when a2 < a1 γth2 , the out-
age probability of D2 for DF and AF protocols is one. For
QMF protocol, the exact theoretical curve is for the outage
probability of D2 corresponding to the derivation (35), when
a1 > a2 γth1 , the outage probability of D2 is less than one.
This verifies the conclusion in Remark 3 that the outage
probability of D2 for QMF protocol is smaller than that of
DF and AF protocols when a2 < a1 γth2 and a1 > a2 γth1 .
B. OUTAGE PERFORMANCE WITH DIFFERENT
VALUES OF η AND LI
This subsection, the outage probability and system through-
put are analyzed with adjusting the values of η and LI. In the
following simulation, the target rates of D1 and D2 for DF and
AF protocols are both set to R1 = 3 and R2 = 0.3 BPCU,
9
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling
FIGURE 7. Outage probability of D2 versus SNR with different values of η,
E{|hLI |2 } = −10 dB.
FIGURE 8. Outage probability of D2 versus SNR with different values of
LI. η = 0.8.
while the target rates of D1 and D2 for QMF protocol are set
to R1 = 0.1 and R2 = 0.3 BPCU.
In the SRF-NOMA system, the information decoding and
the energy harvest at D1 are in the first and second phase,
respectively. Hence, as above analytical results the values
of η and LI have no effect on the outage probability of
D1 . We just analyze the outage probability of D2 with dif-
ferent values of η and LI. Fig. 7 plots the outage proba-
bility of D2 versus SNR with different values of η. The
blue, black and red curves represent the outage probability
of D2 for η = 0.05, η = 0.2 and η = 1, respectively.
Fig. 8 plots the outage probability of D2 versus SNR with
different values of LI. The black and red curves represent
the outage performance of D2 for E |hLI |2 = −10 dB
and E{|hLI |2 } = 0 dB. As can be seen from Fig. 7 and
Fig. 8, the outage performance of D2 is affected by η and LI.
10
FIGURE 9. System throughput versus SNR for DF protocol in delay-limited
transmission mode with different values of η and LI.
FIGURE 10. System throughput versus SNR for AF protocol in
delay-limited transmission mode with different values of η and LI.
With the value of η and LI increasing, the outage probability
of D2 decreases. This is because according to the deriva-
tion (10), the transmit power of relay becomes larger as η and
LI increasing. It is worth noting that all the curves
in Fig. 7 and Fig. 8 have the same slopes, which indicate
the conclusion in Remark 3 and the diversity gains of D2
for the three protocols are two. This phenomenon can be also
explained as follows. The diversity orders of D2 are obtained
with direct link and relaying link, since the energy from
LI can be reused for information transmission at the relay.
The system throughput in delay-limited transmission mode
with different values of η and LI is analyzed in the next
part. Fig. 9, Fig. 10 and Fig. 11 plot the system through-
put versus SNR in delay-limited transmission mode for DF,
AF and QMF protocols, respectively, which are plotted
according to (21), (27) and (42), respectively. We can observe
VOLUME 6, 2018
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

APPENDIX A
According to (15), the outage probability of D2 in DF relaying
case can be expressed as


PDF = Pr γ1,D2 < γth2 Pr γD2 →D1 < γth2



D2
| {z } | {z }
Q11 Q12


+ Pr γD2 →D1 > γth2 , γ2,D2 < γth2 .


| {z }
Q13
(A.1)

Substituting (3), (5)into Q11 , Q12 , we can obtain

− τ
 
FIGURE 11. System throughput versus SNR for QMF protocol in Q11 = 1 − e 2 , (A.2)
delay-limited transmission mode with different values of η and LI.
− τ
 
Q12 = 1 − e 1 , (A.3)

that the throughput of FD NOMA is higher than FD OMA γth

when employing DF and AF protocols, since the outage per- where τ = a - γ 2 a
ρ with a2 > a1 γth2 . Substituting (3),
2 th2 1
formance in FD NOMA is better than FD OMA. On the con- (13)into Q13 , after some algebraic, Q13 can be given as
trary, the throughput of FD OMA is higher than FD NOMA
when employing QMF protocol, as the outage performance γth2 1 − η|hLI |2

!
of D1 for FD NOMA is much more worse than FD OMA. Q13 = Pr |h1 |2 > τ, |h0 |2 <
One can observe that the proposed system achieves higher ηρ|h1 |2
throughput as increasing η and LI for the three protocols in ∞ 1/η
Z Z
1 − z 1 − y
= e 1 dz e LI dy
the low SNR region. This is due to the fact that the outage τ 1 0 LI
probability of D2 decreases as increasing η and LI. Another Z γth (1−ηy)
2
observation is that throughput ceilings exist in the proposed 1 − x
ηρz
× e 0 dx
system for the three protocols in the high SNR region. The 0  0
 
reason is that in the high SNR region, the outage probability − 1 − τ
= 1 − e ηLI e 1
tends zero and the throughput only depends on the target data
rate. ρ0 z
Z ∞
− 1 − z
+ e ηLI
e 1 dz
τ 1 ρ0 z − γth2 LI
V. CONCLUSION | {z }
J1
In this paper SRF-NOMA system has been investigated γth
∞ ρ0 z
Z
insightfully. The performance of the proposed system has − z − 2
−
e 1 ηρz0 dz,
been characterized based on DF, AF, and QMF protocols, τ 1 ρ0 z − γth2 LI
respectively. New expressions of outage probability for two (A.4)
users have been derived. Based on the analytical results,
the diversity orders implemented by the nearby and distant based on [31, eq. (3.352.2)], J1 can be given by
user were one and two, respectively. The diversity orders of
the distant user were obtained with direct link and relay- − η1
Z ∞ 1

γth2 LI

− z
ing link. Numerical results showed that the outage behavior J1 = e LI 1+ e 1 dz
τ 1 ρ0 z − γth2 LI
of SRF-NOMA system with DF or AF relay outperformed
− η1 − τ − η1 γth2 LI
OMA-based system. The outage performance of the pro- =e LI 1 −e LI

posed system with DF relay was superior to AF relay in ρ0 1

γth LI
τ γth LI
 
2
the high SNR region. Additionally, the distant user in the −
× e ρ0 1 Ei − + 2 . (A.5)
SRF-NOMA system with QMF relay was capable of achiev- 1 ρ0 1
ing better performance than that of DF and AF relays.
Furthermore, the system performance has become better Note that (A.2), (A.3), (A.4) and (A.5) are derived on the
when η and LI increase. The LI signal has become beneficial, condition of a2 > a1 γth2 . Combining (A.2), (A.3), (A.4)
since the energy from LI can be harvested for information and (A.5), we can obtain (16).
transmission at the relay. The proof is completed.

VOLUME 6, 2018 11
 Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling

APPENDIX B Note that (C.1), (C.2) and (C.3) are derived on the condition
Based on (18), we assume 0  LI , the outage probability of a2 > a1 γth2 . Substituting (C.2) and (C.3) into (C.1),
for D2 at high SNR can be further approximated as we can obtain (23).
 The proof is completed.


τ − 1 γth2 LI γth LI

  
DF,∞
PD ,appro = 1 − e ηLI 1− 2 APPENDIX D
2 2   ρ0 1 ρ0 1 When ρ → ∞, the approximated expression of (14) can be

obtained as


τ γth2 LI |h0 |2 |h1 |2 a2 ηρ

   
× ln − +C γ2,D
AF
≈ (D.1)
1 ρ0 1 |h0 |2 |h1 |4 a1 ηρ + η|h0 |2 + 1 − η|hLI |2

.
2


 Substituting (5) and (D.1) into (22), the expression of (22)
Z ∞ γth 
1 −  − ηρz 
z 2 at high SNR can be expressed as
− e 1 0 dz . (B.1)
0 1 

(
γth2 1 − η|hLI |2

)
AF,∞
= Pr |h0 | < 
2
| {z 
}
J2 PD
η |h1 |2 ρ a2 − γth2 a1 − γth2


2

Based on [31, eq. (3.324.1)], J2 can be calculated as !

γth2
γth2 γth2 × Pr |h2 | <
2
. (D.2)
r  r 
a2 − γth2 a1 ρ


J2 = 2 K1 2 , (B.2)
ηρ0 1 ηρ0 1
2
Similarly to (C.2) and (C.3), (D.2) can be calculated as
when x → 0, xK1 (x) ≈ 1 + x2 ln x2 + c0 , J2 can be


 
− τ − τ − η1
 
approximated as PAF,∞
D2
= 1 − e 2 e 1 1 − e LI

γth2 γth2
 
J2 = 1 + ln + 2c0 , (B.3)

− τ
Z ∞ 1 0 (z − τ )
ηρ0 1 ηρ0 1 + 1−e 2
τ  1 0 (z − τ ) − τLI
where c0 = − ϕ(1) ϕ(2)
2 − 2 and ϕ (·) denotes psi function [31].
− z − η (z−τ
×e 1 e 0
τ
)
τ
− 1
e 0 η(z−τ ) ηLI − 1 dz
Substituting (B.2) and (B.3) into (B.1), we can obtain (19).
  
The proof is completed. − τ − τ − 1 − τ
  
= 1 − e 2 e 1 1 − e ηLI + 1 − e 2
APPENDIX C 
Substituting (5), (14) into (22), after some algebraic, (22) can
− τ
Z ∞ 0 ε − ε − 1
be expressed as ×e e 1 ηLI dε
1

 0 1 (0 ε − τ LI )

(
)
γ η|h 2
ρ|h 2


th 1 − LI | 1 | + 1 | {z }
PAF = Pr |h0 |2 < 2
V1
ηρ|h1 |2 |h1 |2 ρ a2 − γth2 a1 − γth2
D2

 
| {z }
∞ 0 ε − ε − τ
Z
U1

− e 1 η0 ε dε .

1 (0 ε − τ LI )
!
γth2 0
× Pr |h2 |2 < .

(C.1)
a2 − γth2 a1 ρ


| {z } (D.3)
U2

Furthermore, U1 and U2 can be calculated as Based on [31, eq. (3.352.2)], V1 can be calculated as
τ LI
Z ∞  
Z ∞ Z 1/η 1 − ε − 1
1 − z 1 − y V1 = 1+ e 1 ηLI dε
U1 = e 1 dz e LI dy 1 0 ε − τ LI
τ 1 0 LI 0
τ LI − η1 − τ LI τ LI
 
1
−
Z h γth2 (1−ηy)(ρz+1) = e ηLI − e LI 0 1 Ei . (D.4)
1 − x 0 1 0 1
 i
ηρz zρ a2 −γth a1 −γth
× 2 2 e 0 dx
0 0 Note that (D.2), (D.3) and (D.4) are derived on the condition
 
− 1 − τ of a2 > a1 γth2 . Combining (D.3) and (D.4), and using e−x ≈
= 1 − e ηLI e 1
1 − x and Ei (x) ≈ ln (x) + C [31, eq. (8.214.2)], we can
Z ∞
1 z2 − zτ obtain (24).
+ The proof is completed.
1 z2 − zτ 1 + LI − LI τ
 
τ
0 0 ρ
τ (z+1/ρ) REFERENCES
!
z 1 −
× e 1 e ηLI − e η0 (z −zτ ) dz,
− − 2
(C.2) [1] Z. Ding, X. Lei, G. K. Karagiannidis, R. Schober, J. Yuan, and V. Bhargava,
‘‘A survey on non-orthogonal multiple access for 5G networks: Research
− τ challenges and future trends,’’ IEEE J. Sel. Areas Commun., vol. 35, no. 10,
U2 = 1 − e 2 . (C.3) pp. 2181–2195, Oct. 2017.

12 VOLUME 6, 2018
Z. Wang et al.: Full-Duplex User Relaying for NOMA System With Self-Energy Recycling
[2] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo,
‘‘Nonorthogonal multiple access for 5G and beyond,’’ Proc. IEEE,
vol. 105, no. 12, pp. 2347–2381, Dec. 2017.
[3] Z. Ding et al., ‘‘Application of non-orthogonal multiple access in LTE
and 5G networks,’’ IEEE Commun. Mag., vol. 55, no. 2, pp. 185–191,
Feb. 2017.
[4] T. M. Cover and J. A. Thomas, Elements of Information Theory. Hoboken,
NJ, USA: Wiley, 1991.
[5] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, ‘‘On the performance of non-
orthogonal multiple access in 5G systems with randomly deployed users,’’
IEEE Signal Process. Lett., vol. 21, no. 12, pp. 1501–1505, Dec. 2014.
[6] Z. Yang, Z. Ding, P. Fan, and G. K. Karagiannidis, ‘‘On the performance of
non-orthogonal multiple access systems with partial channel information,’’
IEEE Trans. Commun., vol. 64, no. 2, pp. 654–667, Feb. 2016.
[7] P. Xu, Y. Yuan, Z. Ding, X. Dai, and R. Schober, ‘‘On the outage per-
formance of non-orthogonal multiple access with 1-bit feedback,’’ IEEE
Trans. Wireless Commun., vol. 15, no. 10, pp. 6716–6730, Oct. 2016.
[8] Y. Liu, Z. Qin, M. Elkashlan, Y. Gao, and L. Hanzo, ‘‘Enhanc-
ing the physical layer security of non-orthogonal multiple access in
large-scale networks,’’ IEEE Trans. Wireless Commun., vol. 16, no. 3,
pp. 1656–1672, Mar. 2017.
[9] Y. Zhang, H. M. Wang, Z. Ding, and M. H. Lee, ‘‘Non-orthogonal multiple
access assisted multi-region geocast,’’ IEEE Access, vol. 6, pp. 2340–2355,
2018.
[10] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, ‘‘Cooperative diversity in
wireless networks: Efficient protocols and outage behavior,’’ IEEE Trans.
Inf. Theory, vol. 50, no. 12, pp. 3062–3080, Dec. 2004.
[11] Z. Ding, M. Peng, and H. V. Poor, ‘‘Cooperative non-orthogonal multiple
access in 5G systems,’’ IEEE Commun. Lett., vol. 19, no. 8, pp. 1462–1465,
Aug. 2015.
[12] J. Men, J. Ge, and C. Zhang, ‘‘Performance analysis of nonorthogonal
multiple access for relaying networks over Nakagami-m fading channels,’’
IEEE Trans. Veh. Technol., vol. 66, no. 2, pp. 1200–1208, Feb. 2017.
[13] X. Yue, Y. Liu, S. Kang, and A. Nallanathan, ‘‘Performance analysis of
NOMA with fixed gain relaying over Nakagami-m fading channels,’’ IEEE
Access, vol. 5, pp. 5445–5454, 2017.
[14] D. Wan, M. Wen, F. Ji, Y. Liu, and Y. Huang, ‘‘Cooperative NOMA systems
with partial channel state information over Nakagami-m fading channels,’’
IEEE Trans. Commun., vol. 66, no. 3, pp. 947–958, Mar. 2018.
[15] M. Ashraf, A. Shahid, J. W. Jang, and K. G. Lee, ‘‘Energy harvesting
non-orthogonal multiple access system with multi-antenna relay and base
station,’’ IEEE Access, vol. 5, pp. 17660–17670, 2017.
[16] L. Li, C. Dong, L. Wang, and L. Hanzo, ‘‘Spectral-efficient bidirec-
tional decode-and-forward relaying for full-duplex communication,’’ IEEE
Trans. Veh. Technol., vol. 65, no. 9, pp. 7010–7020, Sep. 2016.
[17] Z. Zhang, Z. Ma, M. Xiao, Z. Ding, and P. Fan, ‘‘Full-duplex device-
to-device-aided cooperative nonorthogonal multiple access,’’ IEEE Trans.
Veh. Technol., vol. 66, no. 5, pp. 4467–4471, May 2017.
[18] X. Yue, Y. Liu, S. Kang, A. Nallanathan, and Z. Ding, ‘‘Exploiting full/half-
duplex user relaying in NOMA systems,’’ IEEE Trans. Commun., vol. 66,
no. 2, pp. 560–575, Feb. 2018.
[19] T. M. C. Chu and H. J. Zepernick, ‘‘Performance of a non-orthogonal
multiple access system with full-duplex relaying,’’ IEEE Commun. Lett.,
vol. 22, no. 10, pp. 2084–2087, Oct. 2018.
[20] M. F. Kader, S. Y. Shin, and V. C. M. Leung, ‘‘Full-duplex non-orthogonal
multiple access in cooperative relay sharing for 5G systems,’’ IEEE Trans.
Veh. Technol., vol. 67, no. 7, pp. 5831–5840, Jul. 2018.
[21] L. R. Varshney, ‘‘Transporting information and energy simultaneously,’’
in Proc. IEEE Int. Symp. Inf. Theory, Toronto, ON, Canada, Jul. 2008,
pp. 1612–1616.
[22] A. A. Nasir, X. Zhou, S. Durrani, and R. A. Kennedy, ‘‘Relaying protocols
for wireless energy harvesting and information processing,’’ IEEE Trans.
Wireless Commun., vol. 12, no. 7, pp. 3622–3636, Jul. 2013.
[23] X. Zhou, R. Zhang, and C. K. Ho, ‘‘Wireless information and power trans-
fer: Architecture design and rate-energy tradeoff,’’ IEEE Trans. Commun.,
vol. 61, no. 11, pp. 4754–4767, Nov. 2013.
[24] 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, no. 10, pp. 3447–3461, Oct. 2014.
[25] Y. Lou, Y. Zheng, J. Cheng, and H. Zhao, ‘‘Performance of SWIPT-based
differential AF relaying over Nakagami-m fading channels with direct
link,’’ IEEE Wireless Commun. Lett., vol. 7, no. 1, pp. 106–109, Feb. 2018.
VOLUME 6, 2018
View publication stats
[26] Y. Zeng and R. Zhang, ‘‘Full-duplex wireless-powered relay with
self-energy recycling,’’ IEEE Wireless Commun. Lett., vol. 4, no. 2,
pp. 201–204, Apr. 2015.
[27] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor, ‘‘Cooperative non-
orthogonal multiple access with simultaneous wireless information
and power transfer,’’ IEEE J. Sel. Areas Commun., vol. 34, no. 4,
pp. 938–953, Apr. 2016.
[28] Z. Yang, Z. Ding, P. Fan, and N. Al-Dhahir, ‘‘The impact of power
allocation on cooperative non-orthogonal multiple access networks with
SWIPT,’’ IEEE Trans. Wireless Commun., vol. 16, no. 7, pp. 4332–4343,
Jul. 2017.
[29] Y. Alsaba, C. Y. Leow, and S. K. A. Rahim, ‘‘Full-duplex cooperative
non-orthogonal multiple access with beamforming and energy harvesting,’’
IEEE Access, vol. 6, pp. 19726–19738, 2018.
[30] H. Zhang et al., ‘‘Energy efficient dynamic resource optimization in
NOMA system,’’ IEEE Trans. Wireless Commun., vol. 17, no. 9,
pp. 5671–5683, Sep. 2018.
[31] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products,
6th ed. New York, NY, USA: Academic, 2000.
[32] M. Fang, J. Li, Z. Wang, and W. Li, ‘‘Performance of quantize-map-
forward based cooperative multicast systems with successive relaying,’’
IEEE Commun. Lett., vol. 20, no. 11, pp. 2257–2260, Nov. 2016.
ZHENLING WANG received the B.S. degree from
the Information Engineering University of Peo-
ple’s Liberation Army in 2009 and the M.S. degree
from Henan Normal University in 2012. She is cur-
rently pursuing the Ph.D. degree with the School
of Communication and Information Engineering,
Shanghai University.
Her research interests include 5G networks,
Internet of Things, cooperative networks, and non-
orthogonal multiple access.
XINWEI YUE graduated from the Yellow River
Conservancy Technical Institute in 2007. He
received the M.S. degree from Henan Normal
University in 2013 and the Ph.D. degree in com-
munication and information system from Beihang
University, Beijing, in 2018. He has been a
Lecturer with the School of Information and Com-
munication Engineering, Beijing Information Sci-
ence and Technology University, Beijing, since
2018.
His research interests include 5G networks, wireless communications
theory, non-orthogonal multiple access, cooperative networks, and Internet
of Things.
ZHANGYOU PENG received the M.S. degree in
communication engineering from Tsinghua Uni-
versity in 1992 and the Ph.D. degree in com-
munication engineering from Shanghai University
in 2009. Since 2002, he has been a Professor with
the School of Communication and Information
Engineering, Shanghai University.
His research interests include satellite signal
processing, radar signal processing, and wireless
communication.
13