Physical Communication 46 (2021) 101335

Contents lists available at ScienceDirect

Physical Communication
journal homepage: www.elsevier.com/locate/phycom

Full length article

Full/half duplex cooperative relaying NOMA network under power

splitting based SWIPT: Performance analysis and optimization
∗
Aswathi V. , Babu A.V.
Department of Electronics and Communication Engineering, National Institute of Technology Calicut, India

article info a b s t r a c t

Article history: Non-orthogonal multiple access (NOMA) and full duplex (FD) relaying are two promising techniques to
Received 13 June 2020 enhance the spectral efficiency of the fifth generation (5G) wireless networks. Simultaneous wireless
Received in revised form 9 January 2021 information and power transfer (SWIPT) technique has recently emerged as an effective solution to
Accepted 25 March 2021
improve the energy efficiency of wireless networks. This paper investigates outage and throughput
Available online 31 March 2021
performance of SWIPT enabled FD cooperative NOMA network under power splitting relaying (PSR)
Keywords: protocol, i.e., SWIPT enabled FD-PSR-NOMA network. We assume a single cell network consisting of a
Full/half duplex base station (BS) and two pre-paired users, i.e., a near user and a far user, where the near user acts as
NOMA FD relay to assist the BS for information delivery to the far user, and power splitting (PS) technique
SWIPT is used at the relay for energy harvesting (EH). We derive analytical expressions for the outage
Outage
probabilities experienced by the downlink users, system outage probability and delay limited system
Throughput
throughput under imperfect successive interference cancellation (i-SIC). We also derive analytical
expressions for outage probabilities and throughput of SWIPT enabled half duplex (HD) cooperative
NOMA network under PSR, i.e., SWIPT enabled HD-PSR-NOMA network. The optimal power allocation
(OPA) factor at the BS and optimal power splitting (OPS) factor at the relay that independently
minimizes the asymptotic system outage probabilities of SWIPT enabled FD/HD-PSR-NOMA networks
are determined. Further, we explore the jointly optimal power allocation and power splitting factor
that minimize the system outage probability. Furthermore, we derive analytical expression for the
OPS factor that maximizes the asymptotic system throughput. We provide extensive numerical and
simulation results to establish that system outage and throughput improve significantly under the
proposed schemes.
© 2021 Elsevier B.V. All rights reserved.

1. Introduction well [2–4]. Owing to these advantages, NOMA has become a

preferred multiple access technology for the fifth generation (5G)
When conventional orthogonal multiple access (OMA) tech- wireless networks [1–4]. When power domain NOMA is used for
nique is used, the number of users that can be simultaneously downlink communications, the base station (BS) will combine
supported by the wireless network is limited by the number of messages for the users with distinct power levels and the com-
available orthogonal resources. Compared to OMA, the recently bined signal is transmitted utilizing the same frequency resource.
proposed non-orthogonal multiple access (NOMA) technique can Successive interference cancellation (SIC) is employed at the user
provide higher spectral efficiency (SE) and improved through- equipment terminals to recover the messages. In cooperative
put, since it allows multiple users to be served by employing NOMA, the users close to the BS (i.e., near users) act as relay nodes
the same frequency resources [1–4]. Since NOMA can support to assist the delivery of messages to the far users [5–7]. However,
more number of users within a given frequency resource, it if the relay node operates in half duplex (HD) mode, two distinct
improves the connection density as well. Accordingly, NOMA is non-overlapping time slots are needed to deliver messages to the
a preferred candidate technology for realizing massive machine far user, which halves the SE of the resulting HD-NOMA system.
type communications (mMTC) and ultra-reliable and low-latency Thus HD relaying will degrade the SE advantage provided by the
communications (URLLC). Further, since NOMA can serve multiple NOMA technique in HD-NOMA systems.
users experiencing distinct channel conditions simultaneously, In-band full-duplex (FD) wireless that supports simultaneous
it can provide improved user fairness and reduced latency as transmission and reception (STAR) in the same frequency band
has been identified as a strong radio technology candidate for
∗ Corresponding author. next generation wireless networks [8]. Realization of FD radio has
E-mail address: aswathi481@gmail.com (Aswathi V.). always been considered unfeasible owing to the fact that STAR

https://doi.org/10.1016/j.phycom.2021.101335
1874-4907/© 2021 Elsevier B.V. All rights reserved.
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

in the same frequency band leads to strong self-interference (SI). downlink with SWIPT, while [19] has investigated the secrecy
Hence, most of the previously available literatures have focused rate maximization problem for a MISO secrecy channel with
on the use of conventional HD radio or out-of-band FD radio tech- SWIPT. In [20], the authors have investigated robust secrecy
nology, where the nodes transmit and receive either over differ- rate maximization problem for a secure MISO SWIPT system.
ent time slots, or over distinct frequency bands. However, recent However, these studies are not applicable for SWIPT enabled
studies have shown that the SI can be suppressed substantially [9] FD-NOMA systems.
and thus FD radios can be efficiently implemented [10]. Since Recently, extensive research has been reported on the perfor-
then, a vast amount of literature has been published on different mance analysis of SWIPT enabled HD-NOMA systems, see [21–43]
aspects of FD radio technology, including its radio frequency and references therein. The outage and throughput performance
(RF) engineering, hardware developments, protocol design, e.g., of SWIPT enabled HD-NOMA system has been analyzed in [21]
[9–12] and references therein. When compared with conventional by modeling the locations of users using stochastic geometry
HD wireless radio, in-band FD wireless has the potential to dou- based approach. The outage performance of SWIPT enabled HD-
ble the SE by supporting simultaneous uplink/downlink via FD NOMA system under partial relay selection scheme has been an-
base stations, and simultaneous listening/forwarding via FD relay alyzed in [22]. Two power allocation policies, namely NOMA with
cooperation [13]. Since FD radios would be an integral part of fixed power allocation and cognitive radio inspired NOMA, have
the future generation wireless networks, FD relaying can be inte- been investigated in [23] for analyzing the outage performance
grated into NOMA systems; the resulting FD-NOMA system will of SWIPT enabled HD-NOMA system. The outage performance
exhibit very high SE and improved throughput than HD-NOMA of SWIPT based HD-NOMA system has been evaluated in [24],
systems. where multiple antennas are used at the source and the des-
Apart from providing higher data rate, massive connectivity tination terminals, while the relay node has been assumed to
and reduced latency, 5G wireless networks are required to be be a single-antenna device. The outage performance of SWIPT
highly energy efficient as well [2]. However, improvement of data enabled multi-user HD-NOMA system has been analyzed in [25],
rate usually causes higher energy consumption, which reduces while the work in [26] has considered maximizing the throughput
the lifetime of battery powered nodes. Recently, there has been of the secondary network in SWIPT based cognitive cooperative
growing interest to utilize energy harvesting (EH) techniques NOMA system. An efficient user-paring method has been pro-
to improve the energy efficiency (EE) and life time of wireless posed in [27] considering a SWIPT based HD-NOMA system. The
networks [14]. When wireless power transfer (WPT) is used as the outage probability and the diversity order of the cell edge user has
EH technique, the nodes in the network can harvest energy either been analyzed in [28], for a two-user MISO HD-NOMA network
from the ambient signals or from dedicated power sources. The
with SWIPT. The outage and ergodic rate performance of SWIPT
simultaneous wireless information and power transfer (SWIPT)
enabled HD-NOMA system has been analyzed in [29] considering
technique enables the relay node in a wireless network to harvest
the effects of imperfect SIC (i-SIC) conditions. The outage perfor-
power from the received RF signals and use the harvested power
mance of SWIPT enabled HD-NOMA system has been investigated
for relaying the message. Accordingly, SWIPT can provide notable
in [30] assuming Weibull fading channels, taking into account i-
gains in terms of power consumption and transmission delay,
CSI. The authors of [31] have explored the joint design of power
since it enables simultaneous transmission of both power and
allocation coefficients and the PS factor for a SWIPT enabled HD-
information. As a result, integrating SWIPT technology in NOMA
NOMA system, while the outage probability of the downlink users
based 5G wireless networks offers a promising solution to pro-
has been evaluated in [32]. The authors of [33] and [34] have
long the life time of wireless devices with high transmission rate.
independently analyzed the outage performance of SWIPT en-
Generally, two protocols are used for realizing SWIPT, i.e., power
abled HD-NOMA system under PSR protocol. In [35], joint power
splitting relaying (PSR) and time switching relaying (TSR) [14].
allocation and PS factor have been investigated to optimize the
When PSR is used, the received signal is split into two parts:
total transmission rate, while the authors of [36] have jointly
one for information detection and the other for energy harvesting
(EH). Under TSR, the total received RF signal is used for both optimized the PS ratios and the beamforming vectors, considering
information processing and EH, but in different time slots. While SWIPT enabled HD-NOMA systems. In [37], the authors have
integration of FD relaying improves the SE of NOMA system, proposed a new PS protocol for SWIPT enabled HD-NOMA system.
the use of SWIPT can enhance the EE of the network. Thus the The authors of [38] have analyzed the outage and throughput
integrated cooperative NOMA system consisting of FD relaying performance of both near and far users in SWIPT enabled incre-
and SWIPT, i.e., SWIPT enabled FD-NOMA system, can meet some mental relay based cooperative NOMA system with non-linear
of essential requirements of 5G wireless networks in terms of energy harvesting, while the work in [39] have evaluated the
SE and EE. Motivated by the above considerations, the objective performance of two cooperative spectrum sharing protocols that
of this paper is to investigate and optimize the performance of incorporate time-switching/ power-splitting based SWIPT, in the
SWIPT enabled FD-NOMA networks under PSR, i.e., we call it as context of relay aided NOMA systems. In [40], the authors have
SWIPT enabled FD-PSR-NOMA network. proposed improved TS based SWIPT for cooperative NOMA net-
works. In [41–43], the authors have independently considered EE
1.1. Literature survey and problem formulation optimization for SWIPT enabled HD-NOMA system.
The detailed literature survey presented above shows that
The application of SWIPT technique has been extensively stud- the performance of SWIPT enabled HD-NOMA system has been
ied in the context of OMA systems [15–20]. The authors of [15] investigated extensively. Even though such systems outperform
have proposed interference aided energy harvesting, integrated the conventional HD-OMA systems, use of HD relaying degrades
with best relay section scheme for cooperative OMA systems. the SE of cooperative NOMA system. An alternative strategy is
In [16], secure beamforming designs are investigated in multiple- to consider SWIPT enabled FD-NOMA system that can provide
input multiple-output (MIMO) secrecy channels with SWIPT, higher SE compared to the corresponding HD-NOMA system.
while [17] has considered the design of robust transmit beam- Higher SE is very essential to support data rate intensive applica-
forming vectors and the power splitting (PS) factors in the pres- tions in the 5G wireless networks. Very recently, a few research
ence of imperfect channel state information (CSI). The authors works have appeared on the performance analysis of FD-NOMA
of [18] have investigated robust weighted sum secrecy rate maxi- systems integrated with SWIPT [44–46]. The authors of [44] have
mization for multiuser multiple-input single-output (MISO) explored transmit power minimization; the authors of [45] have
2
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

considered EE maximization problem; while the authors of [46] Table 1

Important symbols and their meanings.
have attempted to maximize the data rate of the near user; all
Symbol Description
these works have appeared in the context of FD-NOMA systems
integrated with SWIPT. However, the determination of outage Ps , Pr Transmit power of BS, relay node U1

probabilities experienced by the users, system outage probability ρs , ρr Transmit SNR of BS, relay node U1

and throughout of SWIPT enabled FD-NOMA systems remain as α Power allocation factor for U1 at BS
IO,PSR Mean RSI power
open problems. Further, the problems involving determination
η Energy harvesting efficiency
of optimal power allocation factor at the BS and the optimal
β Imperfect SIC factor
power splitting factor at the relay for enhancing the performance
ρ Power splitting factor
of the FD-NOMA systems are still unexplored. Motivated by the
ζ,θ Parameters of SI cancellation technique
aforementioned facts, the major objectives of this paper are: (i)
R1 , R2 Target rates of user 1, user 2
to analyze the outage and throughput performance of SWIPT
hij Fading channel coefficient of link i-j
enabled FD/HD-PSR-NOMA network and (ii) to investigate the
λij Mean channel power gains of link i-j
optimization of system outage and throughput performance of
x1 , x2 Message symbols for user 1, user 2
the considered networks.
uFD HD
i , ui (i ∈ 1, 2) Target SINRs of user i in FD/HD-PSR-NOMA systems
FD,PSR
Pout ,i (i ∈ 1, 2) Outage probability of user i in FD-PSR-NOMA
1.2. Contributions HD,PSR
Pout ,i (i ∈ 1, 2) Outage probability of user i in HD-PSR-NOMA
αFD
∗
,PSR , ρFD,PSR
∗
OPA, OPS factors for minimum system outage in
The major contributions of our paper are significantly different FD-PSR-NOMA
all the above mentioned papers [21–46] and are described as αHD,PSR , ρHD,PSR
∗ ∗
OPA, OPS factors for minimum system outage in
follows: HD-PSR-NOMA
FD,PSR GD,PSR
ρopt , ρopt OPS factor for maximum throughput in
• First of all, we describe the protocol details of SWIPT en- FD/HD-PSR-NOMA
abled two-user downlink FD-PSR-NOMA network. Analytical
expressions are derived for the outage probabilities expe-
rienced by the users and the system outage probability
in the presence of i-SIC. To provide more insights on the 2. System model
outage performance of the users and the considered sys-
This section describes the system model, which include the
tem, we derive analytical expressions for the asymptotic
model for the FD-PSR-NOMA network and the details of the
outage probabilities as well, which are valid for high SNR
channel model, used for the analysis. Thereafter we introduce the
conditions. Further, we formulate expressions for the de-
signal model and write the SINR equations for various cases. Fig. 1
lay limited system throughput as well. The impact of i-
shows the SWIPT enabled FD-PSR-NOMA network considered for
SIC factor and other system related parameters on outage
the analysis, consisting of a BS communicating to two users U1
and throughput performance of the network are extensively
and U2 over the downlink. Due to distinct channel conditions,
analyzed. U1 is considered the near (i.e., cell-centric) user and U2 is the far
• Analytical expression for the optimal power allocation (OPA) (i.e., cell-edge) user. Although theoretically there is no limitation
factor at the BS that minimizes the asymptotic system out- on the number of downlink NOMA users that can be supported by
age probability of SWIPT enabled two-user downlink FD/ the system considered, a two user power domain NOMA scenario
HD-PSR-NOMA network is determined. We also derive an- is more practical. A large number of downlink receivers multi-
alytical expression for the optimal power splitting (OPS) plexed in the power domain might not be feasible in practice, due
factor at the relay that minimizes the system outage prob- to the complexity involved in the implementation of SIC at the
ability. Furthermore, we consider joint optimization, where receivers [2]. Therefore, we focus on a two-user FD-PSR-NOMA
the jointly optimal power allocation factor (at the BS) and scenario. Now U1 is configured as a FD decode-and-forward (DF)
power splitting factor (at the relay), are determined that relay to assist message delivery to U2 . Since the FD node U1
minimizes the system outage of SWIPT enabled FD/HD-PSR- has to implement STAR operation, we assume that it has two
NOMA network. antennas while BS and U2 are assumed to be single antenna
• In the next part of the paper, we derive analytical expres- HD nodes, as they do not implement STAR. Use of FD relaying
sions for the OPS factor that maximizes the delay limited at U1 improves the overall SE of the cooperative-NOMA system
throughput (asymptotic) of the SWIPT enabled FD/HD-PSR- considered. We assume the communication links to experience
NOMA network. Extensive numerical and simulation inves- frequency flat block Rayleigh fading with hij , i ∈ (s, 1), j ∈ (1, 2)
tigations are carried out to determine the sensitivity of the representing the fading channel coefficients corresponding to the
optimal quantities against various system parameters and link connecting nodes i and j. Thus |hij |2 have exponential PDF
their impacts on the outage and throughput performance of with mean λij . The considered communication scenario repre-
the considered systems under the proposed schemes. sents the downlink of single-cell wireless cellular network that
employs user assisted relaying, where the near user is energy
The rest of this paper is organized as follows: Section 2 describes constrained. Table 1 describes a summary of various important
the system model, where the protocol details, signal model and symbols used in this paper.
the signal-to-noise plus interference ratio (SINR) equations for At time t, BS transmits the power domain NOMA signal rep-
SWIPT enabled FD/HD-PSR-NOMA system are described. Section 3 resented as
describes the outage probability analysis. Section 4 considers √ √
x(t) = α Ps x1 (t) + (1 − α )Ps x2 (t) (1)
system outage probability minimization while Section 5 analyzes
the delay-limited throughput of the considered systems. The per- Here x1 and x2 are the messages for U1 and U2 respectively; α Ps
formance evaluation results are described in Section 6, while and (1 − α )Ps respectively are the transmit powers allocated for
Section 7 describes the conclusions. U1 and U2 at BS (0 < α < 1), where Ps is the total power of BS;
3
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

Fig. 2. Full duplex power splitting relaying (FD-PSR).

Fig. 1. Full duplex NOMA network under PSR.

PSR scheme, we can employ FD relaying with virtual harvest-

further (1 − α ) > α so that U2 is allocated higher power at BS, use model proposed in [58] as well, that deploys a battery group
following the basic principle of NOMA technique [1,2]. Now U1 consisting of two rechargeable batteries employed at the energy
receives y1 (t) i.e., harvester, which are activated for EH and power supplying (for
√ √ relaying) in a time-switching manner. Assuming that the block
y1 (t) = hs1 α Ps x1 (t) + hs1 (1 − α )Ps x2 (t) + s1 (t) + na (t) (2)
time duration T is divided into odd and even slots of T /2 each,
where na (t) is the receiver front-end noise, which is assumed to the relay transmission power is supplied by battery #1 (battery
be Gaussian and s1 (t) is the residual SI (RSI) generated at U1 due #2) in the odd (even) slot of a block, while the relay-harvested
to FD operation (i.e., portion of SI that remains at U1 after several energy during the odd (even) slot is stored in battery #2 (battery
stages of SI cancellation). Recall that FD relaying (FDR) allows the #1). Thus EH, ID and message relaying can be carried out in
relay to perform STAR operation using the same time–frequency the same block time T [58]. As mentioned before, we assume
that, with the use of sophisticated passive and active cancellation
resource block. However, it leads to strong SI at the receiver of the
techniques, SI at the FD node U1 has been suppressed to the
relay nodes, which is induced from its transmitter. This SI is or-
level of receiver’s noise floor [51–53]. Accordingly, the harvested
ders of magnitude above the noise floor and could be higher than
power corresponding to the RSI s1 (t) and that corresponding to
the desired signal from any distant node in the network. Hence,
the noise term na (t) are assumed to be negligible. Thus the total
for a node to operate successfully in FD mode, the node must harvested power Pr = ηρ Ps |hs1 |2 , where η is the harvesting
implement SI suppression. The suppression must be high enough θ
efficiency (0 < η < 1). Thus the mean RSI power IO,PSR = ζ P r =
to push the power of the SI down to the noise floor, so that it can θ
ζ (ηPs ρλs1 ) . The signal available for information decoding is
be treated in the same way as the usual additive noise. Recently, √ √ √
several studies have considered the problem of SI cancellation in y1,ID (t) = 1 − ρ[hs1 α Ps x1 (t) + hs1 (1 − α )Ps x2 (t)]
FD systems by investigating different system architectures and √
+ 1 − ρ s1 (t) + n1 (t) (3)
SI cancellation techniques to mitigate the SI signal, which are
√
mainly divided into two main categories: passive suppression, Here n1 (t) = 1 − ρ na (t) + np (t), where np (t) is the processing
and active cancellation. In passive suppression, e.g., [47,48], the noise introduced by the receiver, which affects the SINR at U1 . We
SI signal is suppressed in the propagation domain before it is pro- assume n1 (t) as AWGN of variance N0 .
cessed by the receiver circuitry. In active cancellation techniques,
e.g., [49,50], the SI signal is mitigated by subtracting a processed 2.1. SINR calculations
copy of the transmitted signal from the received signal. Active
cancellation techniques could be divided into digital and analog Notice that at the near user U1 , the symbol x2 is received with
(RF) cancellation techniques based on the signal domain (digital- higher power than symbol x1 . Hence the near user U1 has to
domain or analog-domain), where the SI is subtracted. Typical FD decode symbol x2 from y1 (t) firstly and then use SIC to decode its
own symbol x1 , by canceling the known x2 from y1 (t). Imperfect
systems deploy both passive suppression and active cancellation
SIC (i-SIC) will generate residual interference at U1 . Since U1
techniques to achieve significant mitigation of the SI. Extensive
operates as a DF relay in FD mode, it simultaneously forwards
research works have reported that the combination of passive
the symbol x2 to U2 . The SINR corresponding to the decoding of
and active (i.e., both analog and digital domain) techniques can
x2 and x1 (Γ12
PSR
and Γ11
PSR
respectively) are given by
reduce the SI to the level of noise floor at the receiver, e.g.,
[51–53]; the component that remains is the RSI. Similar to pre- (1 − ρ )|hs1 |2 ρs (1 − α )
vious works, see [54–56] and references therein, we assume RSI
Γ12
PSR
= (4)
(1 − ρ )|hs1 |2 ρs α + ω(1 − ρ )ρO,PSR + 1
to be zero mean Gaussian random variable with variance IO,PSR =
E [|s1 (t)|2 ] = ζ (Pr )θ , where P r is the average transmit power of (1 − ρ )|hs1 |2 ρs α
U1 ; ζ (0 < ζ < 1) and θ (0 < θ < 1) are related to the SI Γ11
PSR
= (5)
cancellation technique used at U1 . The assumed model for RSI β (1 − ρ )|hs1 |2 (1 − α )ρs + ω(1 − ρ )ρO,PSR + 1
I
justifies the experimental results reported in [10]. where ρs = NPs , ρO,PSR = ON,PSR , β (0 < β < 1) is the i-SIC factor
0 0
As shown in Fig. 2, the PSR scheme employed by U1 ensures and ω equals 1 for FD and 0 for HD based systems. In (4) and (5),
that both EH and information transmission happen during the (1 − ρ )ρO,PSR represents the RSI term. In (4), signal component
entire transmission cycle [57]. The PS receiver at the relay node corresponding to x1 is treated as interference, while (5) considers
splits the received signal y1 (t) into two power streams of different the component arising due to i-SIC as residual interference at
power levels, with a certain PS ratio before signal processing is U1 . Since U1 implements STAR, it simultaneously forwards x2 to
√
performed at the receiver. Here ρ percentage of the received U2 . The received signal at U2 is given by y2 (t) = Pr h12 x2 (t) +
power is sent to the energy harvester and (1 − ρ ) percentage n2 (t) where n2 (t) is the AWGN at U2 of variance N0 . Notice that,
is used for information decoding (ID), where ρ (0 < ρ < 1) is use of DF relaying at U1 prevents the propagation of noise and
the PS factor. Meanwhile the harvested energy is transferred to interference (i.e., RSI) from U1 to U2 . Accordingly, the SNR at U2
the battery for temporary storage and then used to power the corresponding to the decoding of x2 is given by, Γ22 PSR
= |h12 |2 ρr
transmission section at the relay [57]. To implement the FD based where ρr = N = ηρρs |hs1 | .
Pr 2
0

4
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

3. Outage probability analysis U2 in SWIPT enabled HD-PSR-NOMA system. For a fair comparison,
among FD and HD systems, we assume the target rates for U1 and U2
In this section, we derive analytical expressions for the outage in FD/HD systems to be the same. However, since HD systems require
probabilities experienced by the users and the system outage two distinct time slots to complete the transmissions of symbols x1
probability in FD-PSR-NOMA network. We provide corresponding and x2 , the achievable rate is halved. Thus to achieve the same target
expression for HD-PSR-NOMA system as well. To provide more rates as that of FD systems, the target SINRs for U1 and U2 are higher
insights on the outage performance, we derive asymptotic outage in HD systems.
probability expressions as well, which are applicable for high SNR
conditions. Let R1 and R2 (bits per channel use-bpcu) be the target
rates for U1 and U2 respectively; further, let uFD R1 3.2. Outage probability experienced by the far user U2
1 = 2 − 1 and
uFD
2 = 2R2
− 1 respectively be the target SINRs for the successful
decoding of x1 and x2 in SWIPT enabled FD-PSR-NOMA system. The far user U2 will not suffer outage, if x2 is successfully
decoded at U1 as well as U2 . Thus outage probability experienced
3.1. Outage probability experienced by the near user U1 by U2 in SWIPT enabled FD-PSR-NOMA system is given by
FD,PSR
The near user U1 has to successfully decode x2 and x1 ; other- Pout ,2 = 1 − Pr {Γ12
PSR
2 ; Γ22 ≥ u2 }
≥ uFD PSR FD
(10)
wise it will suffer outage. Thus the outage probability experienced
FD,PSR
by U1 in SWIPT enabled FD-PSR-NOMA system is determined as: Proposition 2. Assuming that 0 < α < 1
, Pout ,2 is given by
1+uFD
2
(11).
FD,PSR
Pout ,1 = 1 − Pr {Γ12
PSR
2 ; Γ11 ≥ u1 }
≥ uFD PSR FD
(6) [ uFD ((1−ρ )ρO,PSR +1)
FD,PSR
− 2
(1−ρ )ρs (1−α−α uFD )λs1
uFD
2
Pout ,2 =1 − e 2 n−
Proposition 1. Assuming that 0 < α < 1
and
β uFD
1
< α < 1, ηρs ρλ12 λs1
1+uFD
2
1+β uFD
1
FD,PSR
2 ((1 − ρ )ρO,PSR + 1)
uFD
Pout ,1 is given by the following expression:
( )
× Γ 0,
φ FD ((1−ρ )ρO,PSR +1) 2 )λs1
(1 − ρ )ρs (1 − α − α uFD
FD,PSR − (1−ρ )ρs λs1
Pout ,1 =1−e (7)
( uFD )n [ uFD ((1−ρ )ρO,PSR +1)
uFD uFD
∞
(−1)n 2
ηρs ρλ12
− 2
FD,PSR (1−ρ )ρs (1−α−α uFD )λs1
∑
where φ ,
FD
{ }
= max 2 1
. Otherwise, Pout ,1 + e 2
(1−α−α uFD
2
) (α−uFD
1
β+uFD
1
βα ) λs1 n!
n=2
becomes unity.
n−1
(m − 1)!(− λ1 )n−m−1
Proof. Setting ω = 1 in (4)–(5) and substituting for Γ12
PSR
∑
and × ( s1
)m
Γ11 in (6), we get
PSR
m=1 (n − 1)!
uFD
2
((1−ρ )ρO,PSR +1)
(1−ρ )ρ (1−α−α uFD )λ
s 2 s1
(1 − ρ )|hs1 |2 ρs (1 − α )
{
FD,PSR
Pout ,1 = 1 − Pr ≥ uFD ;
(− λ1 )n−1
2 )]]
(1 − ρ )αρs |hs1 |2 + (1 − ρ )ρO,PSR + 1 2 ((1 − ρ )ρO,PSR + 1)
uFD
(
s1
− Ei −
(n − 1)! (1 − ρ )ρs (1 − α − α uFD
2 )λs1
(1 − ρ )|hs1 |2 ρs α
}
≥ uFD
1 (11)
(1 − ρ )|hs1 |2 (1 − α )βρs + (1 − ρ )ρO,PSR + 1
(8a) where Γ (x, y) is the upper incomplete gamma function and Ei() is
FD,PSR
the exponential integral. When α > 1 FD , Pout ,2 will become unity.
uFD ((1 − ρ )ρO,PSR + 1)
{
1+u2
= 1 − Pr |hs1 |2 ≥ 2 ;
(1 − ρ )ρs (1 − α − α uFD
2 )
Proof. Setting ω = 1 in (4) and substituting for Γ12
PSR
and Γ22
PSR
in
1 ((1 − ρ )ρO,PSR + 1)
uFD
}
(10),
|hs1 |2 ≥ (8b)
(1 − ρ )ρs (α − β uFD
1 + αβ u1 )
FD {
(1 − ρ )|hs1 |2 ρs (1 − α )
FD,PSR
Pout ,2 = 1 − Pr ≥ uFD
2 ;
Notice that |hij | have exponential PDF with mean value λij .
2
(1 − ρ )αρs |hs1 |2 + (1 − ρ )ρO,PSR + 1
FD,PSR
Defining φ FD as done previously, Pout ,1 can be obtained as in (7) }
above. The probability term on the RHS of (8b) will become zero, ρr |h12 |2 ≥ uFD
2 (12a)
when either 1−α−α uFD 2 < 0 or α−β u1 +αβ u1 . This implies that
FD FD

the outage probability will become unity, when either α > 1 HD

uFD
{
1+u2
2
β uFD = 1 − Pr |h12 |2 ≥ ; |hs1 |2
or α < 1
. Proposition 1 is thus proved. ηρs ρ|hs1 |2
1+β uFD
1

uFD ((1 − ρ )ρO,PSR + 1)

}
Corollary 1. When 0 < α < 1
and
β uHD
1
< α < 1, outage ≥ 2 (12b)
1+uHD 1+β uHD (1 − ρ )ρs (1 − α − α uFD2 )
2 1

probability experienced by U1 in SWIPT enabled HD-PSR-NOMA 1

∫ ∞
b1
system is given by =1− e−( x +c1 x) dx (12c)
λs1 a1
φ HD
HD,PSR − (1−ρ )ρ λ
Pout ,1 =1−e s s1 (9) uFD
2
((1−ρ )ρO,PSR +1) uFD
2 1
where a1 = , b1 = ηρs ρλ12
and c1 = λs1
.
(1−ρ )ρs (1−α−α uFD
2
)
uHD uHD FD,PSR
where φ ,
{ }
HD
= max 2 1
, uHD = 2 2R1
−1 Now Pout ,2 can be determined as follows, by using Maclaurin
(1−α−α uHD
2
) (α−uHD
1
β+uHD
1
βα ) 1
b1
and uHD
2 = 2 2R2
− 1 respectively are the target SINRs for U1 and series expansion for the term e− x and further simplifying using
5
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

uFD
(3.352.2) and (3.353.1) in [59]. 1
}
.
(α−u1 β+uFD
FD
1
βα )
∞ ∞ ∞
e−c1 x (−1)n bn1
[∫ ∫
FD,PSR 1 ∑
Pout ,2 = 1 − e−c1 x dx − b1 dx + [
λs1 a1 a1 x n! FD,PSR −
φ FD ((1−ρ )ρO,PSR +1)
(1−ρ )ρs λs1
1 uFD
2
n=2 Pout ,sys =1− e −
∞
λs1 ηρs ρλ12
e−c1 x
∫ ]
× dx (12d) φ ((1 − ρ )ρO,PSR + 1)
FD
( )
1
a1 xn × Γ 0, +
(1 − ρ )ρs λs1 λs1
∞
e−a1 c1 (−1)n bn
[ ( uFD
1 ∑ )n
=1− − b1 Γ (0, a1 c1 ) + 1 ∞
(−1)n ηρ ρλ 2

λs1
∑ s 12
c1 n! ×
n=2 n!
n=2
n−1
(v − 1)!(−c1 )n−v−1
[ φFD ((1−ρ )ρ +1)
(−c1 )n−1
[ ∑ O,PSR
−
× e−a1 c1 v − × e (1−ρ )ρs λs1
(n − 1)!a1 (n − 1)!
v=1
]] n−1
∑ (m − 1)!(− λ1 )n−m−1
s1
× Ei(−a1 c1 ) (12e) × ( )m
m=1 (n − 1)!
φ FD ((1−ρ )ρ
O,PSR +1)
(1−ρ )ρs λs1
FD,PSR
Substituting the expressions for a1 , b1 and c1 in (12e), Pout ,2
(− λ1 )n−1
) ]]
φ FD ((1 − ρ )ρO,PSR + 1)
(
FD,PSR
given by (11) can be obtained. However, if α > 1
, Pout ,2 will − Ei − s1
(15)
1+uFD
2 (n − 1)! (1 − ρ )ρs λs1
become unity, since the probability term on the RHS of (12b) will
become zero. Proposition 2 is thus proved.
Proof. Substituting the expressions for Γ12
PSR
, Γ11
PSR
and Γ22
PSR
in
Corollary 2. When 0 < α < 1
, the outage probability (14), we get
1+uHD
2

experienced by U2 in SWIPT enabled HD-PSR-NOMA system is given {

(1 − ρ )|hs1 |2 ρs (1 − α )
FD,PSR
by (13). Pout ,sys = 1 − Pr ≥ uFD
2 ;
(1 − ρ )αρs |hs1 |2 + (1 − ρ )ρO,PSR + 1
uHD
(1 − ρ )|hs1 |2 ρs α
[
HD,PSR
− 2
(1−ρ )ρs (1−α−α uHD )λs1
uHD ≥ uFD
1 ;
2
Pout ,2 =1− e 2 − β (1 − ρ )|hs1 | (1 − α )ρs + (1 − ρ )ρO,PSR + 1
2
ηρs ρλ12 λs1 }
uHD ρr |h12 |2 ≥ uFD
( )
(16a)
× Γ 0, 2 2
(1 − ρ )ρs (1 − α − α uHD
2 ) s1 λ
uFD
{
2
∞
(−1) n
( uHD
2
)n = 1 − Pr |h12 |2 ≥ ; |hs1 |2
∑ ηρs ρλ12 ηρs ρ|hs1 |2
+
φ FD ((1 − ρ )ρO,PSR + 1)
}
n!
n=2 ≥ (16b)
[ uHD
2
(1 − ρ )ρs
− ∫ ∞
(1−ρ )ρs (1−α−α uHD )λs1 1 b1
× e 2
=1− e−( x +c1 x) dx (16c)
λs1 a2
n−1 1 n−m−1
∑ (m − 1)!(− λ )
s1
× )m φ FD ((1−ρ )ρO,PSR +1) FD,PSR
(
uHD where a2 = . Now Pout ,sys can be determined as
m=1 (n − 1)! 2 (1−ρ )ρs
(1−ρ )ρ (1−α−α uHD )λ s 2 s1 b1
follows, by using Maclaurin series expansion for the term e− x
1 n−1 )]]
(− λ ) uHD
(
− s1
Ei − 2 and further simplifying using (3.352.2) and (3.353.1) given in [59].
(n − 1)! (1 − ρ )ρs (1 − α − α uHD
2 )λs1
∞ ∞ ∞
e−c1 x (−1)n bn
[∫ ∫
(13) FD,PSR 1 ∑
1
Pout ,sys = 1 − e−c1 x dx − b1 dx +
λs1 a2 a2 x n!
n=2
∫ ∞
e−c1 x
]
3.3. System outage probability analysis × dx (16d)
a2 xn
∞
e−a2 c1 (−1)n bn
[
The system outage probability of the SWIPT enabled FD-PSR- 1 ∑
=1− − b1 Γ (0, a2 c1 ) + 1

NOMA network includes the probabilities of the following events: λs1 c1 n!

n=2
either U1 or U2 fail to successfully decode x1 , or both U1 and U2 n−1
(v − 1)!(−c1 )n−v−1 (−c1 )n−1
[ ∑
fail to successfully decode x1 and x2 respectively. Thus the system × e−a2 c1 −
FD,PSR
outage probability (Pout ,sys ) is determined as follows: (n − 1)!av2 (n − 1)!
v=1
]]
FD,PSR
Pout ,sys = 1 − Pr {Γ12
PSR
2 ; Γ11 ≥ u1 ; Γ22 ≥ u2 }
≥ uFD PSR FD PSR FD
(14) × Ei(−a2 c1 ) (16e)

β uFD
Proposition 3. Assuming that 0 < α < 1
and 1
< FD,PSR
1+uFD
2
1+β uFD
1 Substituting for a2 , b1 and c1 in (16e), Pout ,sys given by (15) can be
FD,PSR uFD
α < 1, is given by (15), where φ FD = max ,
{ 2
Pout ,sys obtained, where Ei() is the exponential integral function.
(1−α−α uFD
2
)

6
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

β uHD HD,PSR
Corollary 3. When 0 < α < 1
HD and 1
< α < 1, P out ,sys
4. System outage performance optimization
1+u2 1+β uHD1
uHD uHD
is given by (17), where φ HD = max ,
{ 2 1
}
. In this section, we determine the OPA factor at the BS and the
(1−α−α uHD
2
) ( α−u HD β+uHD βα )
1 1
OPS factor at the relay that independently minimizes the system
[
φ HD
1 uHD
(
φ HD
) outage probability. We also determine the jointly optimal power
FD,PSR − (1−ρ
Pout ,sys =1− e )ρs λs1
− 2
Γ 0, allocation and power splitting factors that minimizes the system
λs1 ηρs ρλ12 (1 − ρ )ρs λs1
outage probability.
∞ n
( uHD
2
)n [
1 ∑ (−1) φ HD
ηρs ρλ12 − (1−ρ )ρs λs1 4.1. OPA factor that minimizes system outage probability
+ × e
λs1 n!
n=2
Increase of power allocation factor (α ) at BS increases the out-
∑n−1
(m − 1)!(− λ1 )n−m−1 (− λ1 )n−1 age probability of U2 while lower α increases outage probability
× ( s1 )m − s1
of U1 ; the system outage probability increases in both these cases.
φ HD (n − 1)!
FD,PSR
m=1 (n − 1)!
(1−ρ )ρs λs1 We find the OPA coefficient αFD
∗
,PSR that minimizes Pout ,sys for fixed
) ]] ρ.
φ HD
(
× Ei − (17) Proposition 4. For a given ρ , the OPA factor αFD
∗
(1 − ρ )ρs λs1 ,PSR that minimizes
FD,PSR
Pout ,sys is given by
3.4. Asymptotic outage probability analysis
1 + u1 u2 β
uFD FD FD
αFD
∗
,PSR = (21)
1 + u2 + u1 u2 + u1 u2 β
uFD FD FD FD FD FD
Since the outage probability expressions given in the previous
subsections are intricate, it is difficult to get insights on the FD,PSR
Proof. Consider the expression for Pout ,sys given in (16c):
outage performance of the users in the system. Here we provide
asymptotic outage probability expressions, which are valid for
⎛ ⎞
uFD
high SNR conditions, i.e., ρs → ∞. We utilize the approximations,
∫ 2
∞ −⎝ ηρs ρλ + λx ⎠
k FD,PSR 1 12 x s1
− −x Pout ,sys = 1 − e dx (22)
limρs →∞ e ρs ≃ 1 − ρk and −Ei(x) ≥ 1e+x in (7) and (9). After λs1
s a2
simplifying, the asymptotic outage probability experienced by U1
FD,PSR
in FD-PSR-NOMA and HD-PSR-NOMA systems (Pout ,1 (∞) and Assuming high SNR condition (i.e., ρs → ∞) and utilizing the
HD,PSR approximation e−y = 1 − y for 0 < y ≪ 1, we can get an
Pout ,1 ( ∞) respectively) can be obtained as follows: FD,PSR
approximation for Pout ,sys as follows:
FD,PSR φ FD ((1 − ρ )ρO,PSR + 1) [ ∫ ∞ (
uFD
) ]
Pout ,1 (∞) ∼
= (18a) FD,PSR ∼ 1 2 − λx
(1 − ρ )ρs λs1 Pout ,sys =1− 1− e s1 dx
λs1 φ FD ((1−ρ )ρO,PSR +1)
(1−ρ )ρs
ηρs λ12 ρ x
(23)
HD,PSR φ HD
Pout ,1 (∞) ∼
= (18b) φ FD ((1−ρ )ρO,PSR +1)
(1 − ρ )ρs λs1 ∼ − (1−ρ )ρs λs1
=1−e
The asymptotic outage probability experienced by U2 in FD-PSR- [
1 uFD
[ (
φ FD ((1 − ρ )ρO,PSR + 1)
)]]
FD,PSR HD,PSR 2
NOMA and HD-PSR-NOMA systems (Pout ,2 (∞) and Pout ,2 (∞) + × −Ei −
respectively) can be determined by setting the above approxima- λs1 ηρs λ12 ρ (1 − ρ )ρs λs1
tions in (11) and (13). Further simplifying, the following expres- (24)
sions can be obtained: e−x
Using the relation, −Ei(−x) ≥ 1+x
, (24) becomes:
FD,PSR 2 ((1 − ρ )ρO,PSR + 1)
uFD uFD
Pout ,2 ( ∞) ∼
= + 2
(19a) φ FD ((1−ρ )ρO,PSR +1)
1 uFD
ρs (1 − ρ )(1 − α − α uFD
2 )λs1
ηρs ρλ12 λs1 FD,PSR
Pout ,sys ∼
=1−e
− (1−ρ )ρs λs1
+ 2
λs1 ηρs λ12 ρ
( )
HD,PSR uHD uHD φ FD ((1−ρ )ρO,PSR +1)
Pout ,2 ( ∞) ∼
= 2
+ 2
(19b) 1− (1−ρ )ρs λs1
ρs (1 − α − α uHD
2 )(1 − ρ )λs1
ηρs ρλs1 λ12 × ( ) (25)
φ FD ((1−ρ )ρO,PSR +1)
The asymptotic system outage probability of FD-PSR-NOMA and 1+ (1−ρ )ρs λs1
HD-PSR-NOMA networks can be determined by utilizing the
above approximations in (15) and (17) respectively. After sim- Substituting for ρO,PSR and setting θ = 1, (25) becomes
plifying the following expressions can be obtained:
FD,PSR φ FD uFD
φ FD ((1 − ρ )ρO,PSR + 1) 1 uFD = φ FD ζ ηρ +
Pout ,sys ∼ + 2
(26)
FD,PSR
Pout ,sys (∞) ∼
= + 2
(20a) (1 − ρ )ρs λs1 ηρs λ12 λs1 ρ
(1 − ρ )ρs λs1 λs1 ηρs λ12 ρ { uFD uFD
where φ FD ,
2 1
}
= max (1−α−α uFD ) (α−uFD β+uFD βα )
. We consider two
2 1 1
HD,PSR φ HD uHD distinct cases for (26) as given below:
Pout ,sys (∞) ∼
= + 2
(20b)
(1 − ρ )ρs λs1 ηρs ρλs1 λ12 uFD uFD
Case (i): 2
(1−α−α uFD )
> (α−uFD
1
β+uFD βα )
.
2 1 1
From (18)–(20), it is clear that the asymptotic outage probabilities uFD +uFD uFD β
decreases when ρs becomes larger; further higher mean channel The above implies
uFD + u
1 1 2
FD +uFD uFD +uFD uFD β < α < 1. From Propo-
1 2 1 2 1 2
gain values can reduce the outage probability. Moreover, higher β uFD
target rates increase the outage probability. Thus the asymptotic sition 3, we have 0 < α < 1
FD and 1
< α < 1. Combining
1+u2 1+β uFD1
( )
outage probability equations (18)–(20) can be used to easily uFD +uFD uFD β β uFD
describe the performance of the system in the high SNR region.
these conditions, we have max
uFD +u
1 1 2
FD +uFD uFD +uFD uFD β , 1+β1uFD
1 2 1 2 1 2 1

7
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

uFD +uFD uFD β β uFD

<α< 1
. Since 1 1 2 1
> , the previous Proposition 5. For a given α , the OPS factor ρFD
∗
,PSR that minimizes
1+uFD
2
u1 +u2 +u1 u2 +u1 u2 β
FD FD FD FD FD FD 1+β uFD1 FD,PSR
u1 +u1 u2 β
FD FD FD Pout ,sys is given by
condition is equivalent to
uFD +uFD +uFD uFD +uFD uFD β
< α < 1
1+uFD
. √
1 2 1 2 1 2 2 b 1 q
Substituting the value of φ FD corresponding to case (i) above, in ρFD
∗
,PSR = − −S+ −4S 2 − 2p + (30)
FD,PSR 4a 2 S
(26), Pout ,sys becomes,
where a = A, b = −2A, c = (A + B − C ), d = 2C , e = −C ,
2 ζ ηρ
uFD uFD φ FD uFD 8ac −3b2 b3 −4abc +8a2 d
FD,PSR
Pout ,sys ∼
= + 2 A = φ FD ηζ , B = ρs λs1
, C = 2
ηρs λs1 λ12
,p= 8a2
, q= 8a3
,
(1 − α − α uFD
2 ) (1 − α − α uFD
2 )(1 − ρ )ρs λs1 √ √
3 ∆1 + ∆2 −4∆3
√
∆
uFD 1 1
− 23 p + 3a , ∆0 = c 2 − 3bd +
1 0
2 S= 2
(Q + Q0 ), Q = 2
+ (27)
ηρs λ12 λs1 ρ 12ae, ∆1 = 2c − 9bcd + 27b e + 27ad − 72ace.
3 2 2

Differentiating (27) w.r.t. α and after conducting extensive nu- FD,PSR

FD,PSR
∂ Pout
Proof. To find ρFD∗
,PSR that minimizes Pout ,sys , differentiating (26)
,sys
merical investigations, we observe that ∂α
> 0, for the w.r.t ρ , we get the following expression:
FD,PSR
range of α considered, i.e., Pout ,sys monotonically increases, when ∂ FD,PSR ∼ B C
uFD +uFD uFD β P =A+ − 2 (31)
1 1 2
<α< 1
. ∂ρ out ,sys (1 − ρ )2 ρ
uFD
1
+uFD
2
+uFD
1 2
uFD +uFD
1 2
uFD β 1+uFD
2
φ FD uFD
uFD uFD where A = φ FD ζ η, B = and C = 2
. Further, for
Case (ii): 2
(1−α−α uFD
< (α−uFD
1
β+uFD βα )
. ρs λs1 ηρs λs1 λ12
) ∂ 2 FD,PSR FD,PSR
2 1 1
uFD +uFD uFD β
0 < ρ < 1, > 0, since B, C > 0. Thus Pout ,sys is a
P
∂ρ 2 out ,sys
In this case 0 < α < 1 1 2
However since 0 < convex function of ρ for 0 < ρ < 1. Accordingly, the solution for
uFD +uFD +uFD uFD +uFD uFD β
1 2 1 2 1 2 ∂ FD,PSR
β uFD ρFD
∗
,PSR can be obtained by setting ∂ρ Pout ,sys = 0 in (31). Since A,
α < 1
1+uFD
and < α < 1, the previous condition implies
1+β uFD
1 √
∂ FD,PSR
2 ( 1 ) B, C > 0, P
∂ρ out ,sys
< 0 for 0 < ρ < − 4a
b
− S + 12 −4S 2 − 2p + Sq
β uFD uFD +uFD uFD β
that 1
< α < min uFD +uFD1+uFD1uFD2+uFD uFD β , 1+1uFD . However ∂ FD,PSR
√
1+β uFD
1 1 2 1 2 1 2 2 and P
∂ρ out ,sys
> 0 for ρ > − 4a
b
− S + 12 −4S 2 − 2p + Sq . Thus,
uFD +uFD uFD β FD,PSR
since 1 1 2
< 1
, the previous condition is Pout ,sys monotonically decreases, when ρ is in the range 0 <
uFD
1
+uFD
2
+uFD
1 2
uFD +uFD
1 2
uFD β 1+uFD
2
√
FD,PSR
equivalent to
β uFD
1
< α <
uFD
1
+uFD uFD β
1 2
. Substituting
ρ < − 4a
b
− S + 21 −4S 2 − 2p + Sq ; further, Pout ,sys monotonically
1+β uFD
1
u1 +u2 +u1 u2 +uFD
FD FD FD FD uFD β
1 2
√
the value of φ FD corresponding to case (ii) in (26), Pout ,sys becomes
FD,PSR increases, when ρ satisfies ρ > − 4a b
− S + 12 −4S 2 − 2p + Sq .
FD,PSR
Accordingly, the OPS factor that minimizes Pout ,sys can be obtained
FD,PSR 1 ζ ηρ
uFD as in (30). Proposition 5 is thus proved.
Pout ,sys ∼
=
(α − uFD
1 β + u1 βα )
FD

uFD Corollary 5. For a given α , the OPS factor ρHD

∗
,PSR that minimizes
1 HD,PSR
+ Pout ,sys is given by
(α − uFD
1 β + u1 βα )(1 − ρ )ρs λs1
FD
√
uFD −2N + (2N)2 + 4(M − N)N
+ 2
(28) ρHD,PSR =
∗
(32)
ηρs λ12 λs1 ρ 2(M − N)
Differentiating (28) w.r.t. α , we observe, from numerical investi- where M =
φ HD
and N =
uHD
2
.
FD,PSR ρs λs1 ηρs λ12 λs1
∂ Pout ,sys FD,PSR
gations, that ∂α
< 0 for the range of α considered, i.e., Pout ,sys
β uFD 4.3. Joint optimization
monotonically decreases, when
1+β uFD
1
< α <
1
FD,PSR
uFD
1
+uFD
1 2
uFD β
. Thus, it can be seen that Pout ,sys is a mono-
FD,PSR Proposition 4 provides αFD
∗
,PSR at the BS that minimizes Pout ,sys ,
uFD +uFD +uFD uFD +uFD uFD β FD,PSR
1 2 1 2 1 2 while Proposition 5 gives ρFD,PSR that minimizes ∗
In this Pout ,sys .
β uFD
tonically decreasing function of α for 1
< α < section, we explore the jointly optimal parameters, i.e., (αFD
∗∗
,PSR ,
1+β uFD
uFD +uFD uFD β
1
ρFD
∗∗
,PSR ) that minimizes the System outage probability. The opti-
1 1 2
and monotonically increasing function of α mization problem is:
uFD
1
+uFD
2
+uFD
1 2
uFD +uFD
1 2
uFD β
uFD +uFD uFD β FD,PSR
for 1 1 2
<α < 1
. Thus the OPA factor that min Pout ,sys ; s.t . 0 < ρ < 1, 0<α<1 (33)
uFD
1
+uFD +uFD uFD +uFD uFD β
2 1 2 1 2
1+uFD
2 α,ρ
FD,PSR uFD +uFD uFD β
minimizes Pout ,sys is given by αFD
∗
,PSR =
1 1 2
. The
uFD +uFD +uFD uFD +uFD uFD β Proposition 6. The jointly optimal α and ρ i.e., (αFD ,PSR , ρFD,PSR )
∗∗ ∗∗
1 2 1 2 1 2
Proposition 4 is thus proved. FD,PSR
that minimizes Pout ,sys are given by αFD
∗∗
,PSR = αFD,PSR and ρFD,PSR =
∗ ∗∗
√
Corollary 4. For a given ρ , the OPA factor αHD
∗
,PSR that minimizes
− 4b̂â − Ŝ + 21 −4Ŝ 2 − 2p̂ + Ŝq̂ where b̂, â, Ŝ, p̂, and q̂ are calculated
HD,PSR
Pout ,sys is given by by plugging α = αFD ∗
,PSR for b, a, S, p and q defined in Proposition 5.

uHD 1 u2 β
+ uHD HD
αHD
∗ 1
(29) Proof. From Proposition 4, αFD ∗
,PSR is independent of ρ ; thus
,PSR =
uHD
1 + HD
u2 HD HD
1 u2 β
+ u1 u2 + uHD HD αFD
∗
,PSR is jointly optimal as well, i.e., αFD
∗∗
,PSR = αFD,PSR . Now
∗

the derivation of jointly optimal ρ (i.e., ρFD,PSR ) is similar to the

∗∗

derivation of ρFD ∗
,PSR given in the previous section. Accordingly,
4.2. OPS factor that minimizes system outage probability the jointly optimal ρFD ∗∗
,PSR is given by (30) itself, where quantities
b, a, S, p, q are replaced by b̂, â, Ŝ, p̂, q̂, respectively, which are
FD,PSR obtained by plugging α = αFD ∗
,PSR in the equations for b, a, S, p
Here we find ρFD,PSR that minimizes
∗
Pout ,sys . and q given in Proposition 5. Thus Proposition 6 is proved.
8
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

φ FD
( )
Corollary 6. The jointly optimal α and ρ i.e., (αHD
∗∗
,PSR , ρHD,PSR ) that
∗∗
HD,PSR
= R1 1 − φ FD ζ ηρ −
∼
minimizes Pout ,sys are given by αHD (1 − ρ )ρs λs1
,PSR = αHD,PSR and ρHD,PSR =
∗∗ ∗ ∗∗
√ FD
−2N + (2N)2 +4(M̂ −N)N uHD u ((1−ρ )ρO,PSR +1)
2
2
[
where N = ηρs λ12 λs1
and M̂ = M given in (29) −
ρs (1−ρ )(1−α−α uFD )λs1
2(M̂ −N) + R2 e 2
where α is replaced as αHD
∗
,PSR .
uFD
2
5. Analysis of system throughput −
ηρs ρλ12 λs1
2 ((1 − ρ )ρO,PSR + 1)
uFD
( ( )) ]
In the delay-limited transmission mode, BS serves the users
with constant rates R1 and R2 . Under this scenario, the system
× −Ei −
ρs (1 − ρ )(1 − α − α uFD
2 )λs1
throughput depends on outage probabilities. When x1 (or x2 ) is
delivered successfully to U1 (or U2 ), the achieved rate is R1 (or (39)
R2 ). Thus the throughput of SWIPT enabled FD-PSR-NOMA and e−x
Utilizing the tight lower bound for −Ei(−x), i.e., −Ei(−x) ≥ 1+x
,
HD-PSR-NOMA systems are determined as:
T FD,PSR becomes:
FD,PSR FD,PSR
T FD,PSR = R1 (1 − Pout ,1 ) + R2 (1 − Pout ,2 ) (34a) R1 φ FD
T FD,PSR ∼
= R1 − R1 φ FD ζ ηρ − + R2
(1 − ρ )ρs λs1
2 ((1 − ρ )ρO,PSR + 1)
uFD
HD,PSR HD,PSR
T HD,PSR = R1 (1 − Pout ,1 ) + R2 (1 − Pout ,2 ) (34b) − R2
ρs (1 − ρ )(1 − α − α uFD2 )λs1
5.1. OPS factor for maximizing the throughput of SWIPT enabled (
u2 ((1−ρ )ρO,PSR +1)
FD
)
FD/HD-PSR-NOMA systems 1 −
uFD ρs (1−ρ )(1−α−α uFD 2
)λs1
2
− R2 (40)
ηρs ρλ12 λs1
( )
The harvested power becomes smaller, when ρ is lowered, uFD
2
((1−ρ )ρO,PSR +1)
1+
which leads to reduction of T FD,PSR . Similarly, when ρ is increased, ρ (1−ρ )(1−α−α uFD )λ
s 2 s1

ID may fail at U1 triggering decrease of T FD,PSR . Here we find R1 φ 2 ζ ηρ

R2 uFD
FD
the OPS factor that maximizes throughput of SWIPT enabled = R1 − R1 φ FD ζ ηρ −
∼ + R2 −
FD/HD-PSR-NOMA system: (1 − ρ )ρs λs1 (1 − α − α uFD
2 )
R2 uFD R2 uFD
max T FD,PSR or T HD,PSR − 2
− 2
(41)
ρ (1 − ρ )(1 − α − α uFD ρλ
2 ) s s1
λs1 ηρs λ12 ρ
s.t . 0<ρ<1 (35) ∂ 2 FD,PSR
We observe that ∂ρ 2
T < 0 for 0 < ρ < 1 and 0 < α < 1
1+uFD
.
2
Proposition 7. The optimal ρ (ρopt
FD,PSR
) that maximizes T FD,PSR is Thus T FD,PSR is a concave function of ρ (0 < ρ < 1). Accordingly,
FD,PSR ∂ FD,PSR
given by the following expression: ρopt can be obtained by solving ∂ρ T = 0 and is given by
√ (36).
FD,PSR b 1 q
ρopt =− −S+ −4S 2 − 2p + (36) HD,PSR
4a 2 S Corollary 7. The optimal ρ (ρopt ) that maximizes the throughput
where a = (D + F ), b = −2(D + F ), c = (D + F + E + G − H), of SWIPT enabled HD-PSR-NOMA , T HD,PSR is given by
φ FD R1 uFD R ηζ
d = 2H, e = −H, D = R1 φ FD ηζ , E = 2 2
√
ρs λs1
, F = , −2L + (2L)2 + 4(J + K − L)L
(1−α−α uFD ) HD,PSR
uFD R uFD R
2 ρ opt = (42)
2 2 2 2 8ac −3b2 b3 −4abc +8a2 d 2(J + K − L)
G = ,H = ηρs λs1 λ12
,p = , q = ,
ρs λs1 (1−α−α uFD
2
) 8a2 8a3
√ φ HD R1 uHD
2
R2 uHD
2
R2
where J = , K = and L = .
√
3 ∆ + ∆2 −4∆3 ρs λs1 ηρs λ12 λs1
√
∆0 1 (1−α−α uHD )ρs λs1
1 2 1
, ∆0 = c 2 − 3bd +
1 0
S = 2 − 3 p + 3a (Q + Q ), Q = 2
2

12ae, ∆1 = 2c − 9bcd + 27b e + 27ad − 72ace.

3 2 2
6. Results and discussions
FD,PSR
Proof. Substituting the expressions for Pout ,1 given by (7) and This section describes the analytical and Monte-Carlo simu-
FD,PSR lation results. A two-dimensional topology has been considered,
Pout ,2 given by (12c) in (34a), we get
where the locations of BS, U1 and U2 are selected as (0,0), (1.5,0)
φ FD ((1−ρ )ρO,PSR +1)
and (2,0) respectively. The mean power gains λij = Gd− n
ij , where
T FD,PSR = R1 (e
− (1−ρ )ρs λs1
) −n
G = 1, dij is the distance between nodes i and j and n is the path
∞ uFD
( ∫ )
1 2
− ηρ ρλ − λx loss exponent (equals 3 for all the links).
+ R2 e s 12 x s1 dx (37)
λs1 uFD ((1−ρ )ρO,PSR +1)
2
(1−ρ )ρs (1−α−α uFD ) 6.1. Evaluation of outage probabilities
2

FD,PSR
The high SNR approximation (i.e., ρs → ∞) for T is given by
Fig. 3 compares the outage probabilities experienced by U1 and
φ FD
( )
FD,PSR
U2 in SWIPT enabled FD-PSR-NOMA and HD-PSR-NOMA systems,
T ∼
= R1 1 − φ ζ ηρ −
FD
against ρs . For a fair comparison, we set the target rates for U1 and
(1 − ρ )ρs λs1
( ∫ ∞ U2 in HD system to be twice the target rates in FD system. In the
1
+ R2 low ρs region, both U1 and U2 experience lower outage in FD-PSR-
λs1 uFD ((1−ρ )ρO,PSR +1)
2 NOMA system compared to HD-PSR-NOMA system. However, in
ρs (1−ρ )(1−α−α uFD
2
)
the high transmit power region, HD system outperforms FD sys-
− λx
uFD ∞
tem, owing to the fact that, increase of ρs leads to increase of ρr
∫ )
− λx 2 e s1
×e s1 dx − dx as well, which makes the RSI power at U1 to be higher, triggering
ηρs ρλ12 λs1 uFD ((1−ρ )ρO,PSR +1)
2 x
ρs (1−ρ )(1−α−α uFD
2
) increase of outage of both U1 and U2 in FD-PSR-NOMA system.
(38) Both U1 and U2 experience outage floor in FD–PSR-NOMA system,
9
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

Fig. 3. Outage probability vs. ρs (R1 = 0.5, R2 = 0.6, η = 0.5, ρ = 0.4, α = 0.32, Fig. 4. Outage probability vs. ρs : distinct β and θ (R1 = 0.5, R2 = 0.6, η = 0.5,
ζ = −3 dB, β = 0.2, θ = 0.9). ρ = 0.3, α = 0.3, ζ = −3 dB).

when ρs becomes larger, due to the fact that larger ρs makes the
mean RSI power at U1 to be higher. However, since RSI is absent
in HD system, outage floor is not observed in HD-PSR-NOMA
system.
Fig. 4 shows the outage probabilities experienced by U1 and U2
in FD-PSR-NOMA system for distinct values of β (i.e., i-SIC factor)
and θ (i.e., related to mean RSI power). As β is increased, outage
probability of U1 becomes larger, due to the residual interference
generated by i-SIC at U1 . With ρs = 30 dB and θ = 0.9, the outage
probability of U1 in FD-PSR-NOMA system increases by 25% as
β is increased from 0.2 to 0.5; the outage probability increases
by 50% as β is increased from 0.2 to 0.8. When η > 0.5, the
outage probability of U1 will increase further, owing to the higher
residual interference generated by i-SIC at U1 . However, outage
probability of U2 is insensitive to β , since SIC is not required
for the detection of the symbol x2 at U2 . As θ is increased, the
mean RSI power increases, causing significant increase of outage
probability of both U1 and U2 in FD system. For higher values of
ρs and θ , an outage floor appears, since increase of either ρs or θ
increases the RSI at U1 . The impact of β on the outage probability
of U1 and U2 in HD-PSR-NOMA system can also be observed in Fig. 5. System outage probability vs. ρs (R1 = 0.5, R2 = 0.6, η = 0.5, ρ = 0.4,
Fig. 4. The system outage probability results are shown in Fig. 5 α = 0.3, ζ = −3 dB, θ = 0.9).
against ρs . As β is increased, the system outage probabilities of
FD/HD-PSR-NOMA networks become higher due to the residual
interference generated by i-SIC at U1 . When θ is increased, the larger values of ρ affects successful decoding of both x1 and x2
mean RSI power at U1 becomes larger, which degrades the SINR at U1 . In the first case, the harvested power is reduced while
over BS-U1 link, triggering increase of system outage of FD-PSR- for the second case, the power available for ID at U1 reduces. It
NOMA network. Further, an outage floor is observed in FD system is observed that an optimal ρ exists that minimizes the system
for higher values of ρs , owing to the presence of higher RSI power, outage probability. Fig. 6 shows the impact of θ on optimal ρ
while the outage floor is absent in HD-PSR-NOMA system. (ρFD
∗
,PSR ). Larger value of θ increases the mean RSI power at U1 ,
which degrades the SINR over BS-U1 link, triggering increase of
6.2. Evaluation of OPA and OPS factors for minimizing system outage system outage. To minimize the system outage, ρFD ∗
,PSR shall be
probability increased so that more power is made available for ID at U1 ,
which improves the system outage probability. Fig. 7 shows the
In this section, we determine OPA and OPS factors that inde- system outage probability results for FD/HD-PSR-NOMA network
pendently minimize the system outage probabilities of SWIPT en- against ρ for distinct values of target rates R1 and R2 . The results
abled FD/HD-PSR-NOMA networks. We also determine the jointly show that there exists optimal values of ρ for which the system
optimal parameters that minimize the system outage probabili- outage is minimized, for both HD as well as FD systems. It can be
ties. Fig. 6 shows the system outage probability of SWIPT enabled observed that, when R1 is increased (keeping R2 constant), both
FD-PSR-NOMA network against ρ , the power splitting (PS) factor. ρFD
∗
,PSR as well as ρHD,PSR must be reduced for the corresponding
∗

Lower values of ρ affects successful decoding of x2 at U2 ; while systems. This is because, when R1 increases, the threshold SINR
10
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

Table 2
SWIPT enabled FD-PSR-NOMA: αFD ∗
,PSR , ρFD,PSR and jointly optimal (αFD,PSR , ρFD,PSR )
∗ ∗∗ ∗∗

for distinct β (R1 = 0.5, R2 = 0.6, ρs = 20 dB, θ = 1, ζ = −3 dB).

i-SIC factor OPA factor OPS factor Joint optimal
β αFD
∗
,PSR ρFD
∗
,PSR (αFD
∗∗
,PSR , ρFD,PSR0 )
∗∗

0.1 0.3739 0.2902 (0.3739,0.3275)

0.2 0.3852 0.2747 (0.3852,0.3212)
0.3 0.3961 0.2581 (0.3961,0.3150)
0.4 0.4066 0.2403 (0.4066,0.3089)
0.5 0.4167 0.2209 (0.4167,0.3028)

for U1 becomes higher. In this case, optimal ρ shall be reduced so

as to ensure that more power is made available to improve the
probability of successful decoding at U1 for ID. On the other hand,
when R2 is increased (keeping R1 constant), both ρFD ∗
,PSR as well
as ρFD,PSR must be increased for the corresponding systems.
∗

Figs. 8 and 9 respectively show the system outage probability

results for SWIPT enabled FD-PSR and HD-SPR systems against
the power allocation factor (α ). For both lower and higher values
of α , the system outage probability has been observed to increase
as can be seen in Figs. 8 and 9. When α is reduced, the probability
of successful decoding of x1 at U1 reduces, which leads to increase Fig. 6. System outage probability vs. ρ (R1 = 2.5, R2 = 0.5, η = 0.5, α = 0.2,
β = 0.3).
of the system outage probability. Optimal α values (αFD ∗
,PSR and
αHD,PSR ) exist that minimizes the system outage probability. Figs. 8
∗

and 9 respectively show the sensitivity of αFD ∗

,PSR and αHD,PSR
∗

against β and target rates R1 and R2 . When β is increased, it

affects the successful decoding of x1 at U1 due to the residual
interference generated by i-SIC, which ultimately increases the
system outage. To minimize the system outage, αFD ∗
,PSR and αHD,PSR
∗

shall be increased (for FD and HD systems respectively) so that

more power is allocated for U1 at BS. Increase of target rate R1
(for fixed R2 ), increases the target SINR for the decoding of x1
at U1 , which leads to higher system outage. To minimize the
system outage, αFD ∗
,PSR and αHD,PSR shall be increased (for FD and
∗

HD systems respectively), when R1 becomes larger. When R2

becomes larger (for fixed R1 ), the target SINR for the decoding
of x2 increases, which makes the system outage to be higher. To
minimize the system outage probability, αFD ∗
,PSR and αHD,PSR shall
∗

be decreased (for FD and HD systems respectively) as can be seen

in Figs. 8 and 9 respectively.
Table 2 depicts αFD ∗
,PSR , ρFD,PSR and jointly optimal (αFD,PSR ,
∗ ∗∗

ρFD,PSR ) for the considered SWIPT enabled FD-PSR-NOMA sys-

∗∗

tem, while Table 3 depicts the corresponding values for SWIPT

enabled HD-PSR-NOMA system. Fig. 10 plots the system out-
age for the following cases: (i) optimal α and random ρ , (ii)
optimal ρ and random α , (iii) jointly optimal (α , ρ ) and (iv) Fig. 7. System outage probability vs. ρ (η = 0.5, α = 0.2, ζ = −3 dB, β = 0.3,
non-optimal (i.e., random ρ and α ), considering FD-PSR-NOMA θ = 0.9).
systems. The results show that the system outage probability
reduces significantly for the jointly optimal case, since it pro- Table 3
SWIPT enabled HD-PSR-NOMA: αHD ∗
,PSR , ρHD,PSR and jointly optimal (αHD,PSR ,
∗ ∗∗
vides additional flexibility to choose the parameters to minimize
ρHD
∗∗
,PSR ) for distinct β (R1 = 0.5, R2 = 0.6, ρs = 20 dB, θ = 1, ζ = −3 dB).
the system outage probability. Under this approach, the system
i-SIC factor OPA factor OPS factor Joint optimal
outage probability reduces by 65% for FD, compared to the non-
β αHD
∗
ρHD
∗
(αHD
∗∗
,PSR , ρHD,PSR )
∗∗
optimal case, for ρs = 30 dB. Further, the system outage floor ,PSR ,PSR
0.1 0.3033 0.5866 (0.3033,0.5886)
behavior disappears for cases (ii) and (iii), which means that
0.2 0.3268 0.5421 (0.3268,0.5647)
either tuning ρ independently or tuning ρ and α jointly, can sig- 0.3 0.3487 0.4703 (0.3487,0.5368)
nificantly reduce the system outage probability of FD-PSR-NOMA 0.4 0.3692 0.2950 (0.3692,0.5030)
networks.

6.3. Evaluation of throughput of SWIPT enabled FD-PSR-NOMA sys-

tem to HD-PSR-NOMA system; however when ρs becomes higher,
the mean RSI power becomes larger, which leads to throughput
This section describes the results for the delay limited degradation in FD-PSR-NOMA system. Further, increase of β leads
throughput as well as the OPS factor that maximizes the
to degradation of the throughput of both the systems. Fig. 12
throughput. Fig. 11 shows the throughput of the considered
systems against ρs for distinct values of β . In the low ρs region, shows the throughput against power splitting factor ρ for both
the throughput of FD-PSR-NOMA system is higher compared FD-PSR-NOMA and HD-PSR-NOMA systems. It is observed that,
11
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

Fig. 8. System outage probability vs. α (η = 0.5, α = 0.2, ζ = −3 dB, β = 0.3, Fig. 10. System outage probability vs. ρs : optimal and non-optimal (R1 = 2.5,
θ = 0.9). R2 = 0.5, η = 0.5, α = 0.2, ζ = −3 dB, β = 0.3).

Fig. 11. Throughput vs. ρs (θ = 0.9, η = 0.5, ζ = −3 dB, α = 0.2, ρ = 0.3).

Fig. 9. System outage probability vs. α (η = 0.5, α = 0.2, ζ = −3 dB, β = 0.3,
θ = 0.9).

the SINR over the BS-UI link suffers, which leads to throughput
FD,PSR
reduction. To improve the throughput, ρopt shall be reduced so
both lower and higher values of ρ trigger decrease of the through-
that more power is made available for ID at U1 . Thus optimal ρ for
puts. When ρ becomes small, harvested power is reduced, which
FD-PSR-NOMA system decreases, when either ρs or θ is increased,
reduces the probability of successful decoding of the message at
as can be seen in Fig. 13. Tables 4 and 5 tabulate the optimal
U2 . When ρ is increased, it reduces the probability of successful
parameters that maximize the throughput for various cases. To
decoding of the message at U1 . An optimal ρ exists that maxi- summarize, the results presented in this paper have established
mizes the throughput. Fig. 12 shows the impact of target rates R1 that proper selection of system parameters can improve the
FD,PSR HD,PSR
and R2 on ρopt and ρopt . Increase of R1 (keeping R2 constant) performance of SWIPT enabled FD/HD-PSR-NOMA networks.
increases the threshold SINR for U1 . In this case, optimal ρ shall
be reduced so as to ensure that more power is made available to 7. Conclusion
increase the probability of successful decoding of message at U1 .
FD,PSR HD,PSR
This implies that both ρopt and ρopt must be reduced for the This paper considered a SWIPT enabled, full duplex/half du-
respective systems, when R1 becomes higher. Further, increase plex (FD/HD) cooperative NOMA system under power splitting re-
FD,PSR HD,PSR
of R2 (keeping R1 constant) will require both ρopt and ρopt laying (SWIPT enabled FD/HD-PSR-NOMA system) scheme, where
to be increased for the respective systems so as to maximize the a single cell scenario was considered and the BS served two down
FD,PSR
throughput. Fig. 13 shows the impact of ρs and θ on ρopt in link users, with distinct channel conditions (a strong user and
FD-PSR-NOMA system. As mentioned before, when either ρs or weak user). Under the assumption of imperfect SIC, analytical
θ is increased, the mean RSI power becomes very high so that expressions were derived for (i) outage probabilities experienced
12
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

Table 4
FD,PSR
SWIPT enabled FD-PSR-NOMA: ρopt for distinct R1 , R2 and β (ρs = 15 dB,
θ = 1, ζ = −3 dB).
β = 0.1 β = 0.3
R1 , R2 0.5,0.6 0.6,0.5 0.6,0.6 0.5,0.6 0.6,0.5 0.6,0.6
FD,PSR
ρopt 0.2109 0.1637 0.1875 0.1961 0.1446 0.1678

Table 5
HD,PSR
SWIPT enabled HD-PSR-NOMA: ρopt for distinct R1 , R2 and β (ρs = 15 dB,
θ = 1, ζ = −3 dB).
β = 0.1 β = 0.3
R1 , R2 0.5,0.6 0.6,0.5 0.6,0.6 0.5,0.6 0.6,0.5 0.6,0.6
HD,PSR
ρopt 0.1215 0.1320 0.1205 0.1194 0.1083 0.1074

in the 5G wireless communication systems. On the other hand,

massive MIMO is capable of further enhancing the system per-
formance, as a benefit of the large antenna array gain. NOMA is
expected to co-exist with MIMO and massive MIMO technolo-
gies for meeting the challenges of 5G and beyond 5G wireless
networks. In future, research work will be carried out on SWIPT
Fig. 12. Throughput vs. ρ (θ = 0.9, η = 0.5, ζ = −3 dB, α = 0.3, β = 0.3, enabled multiple-antenna aided NOMA systems.
ρs = 20 dB).

Declaration of competing interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared
to influence the work reported in this paper.

Acknowledgment

The authors would like to thank the Department of Science &

Technology and the Ministry of Education (MoE), Government of
India.

References

[1] B. Makki, K. Chitti, A. Behravan, M. Alouini, A survey of NOMA: current

status and open research challenges, IEEE Open J. Commun. Soc. 1 (2020)
179–189.
[2] S.R. Islam, N. Avazov, O.A. Dobre, K.S. Kwak, Power-domain non-orthogonal
multiple access (NOMA) in 5G systems: Potentials and challenges, IEEE
Commun. Surv. Tutor. 19 (2016) 721–742.
[3] M. Vaezi, R. Schober, Z. Ding, H.V. Poor, Non-orthogonal multiple access:
common myths and critical questions, IEEE Wirel. Commun. 26 (5) (2019)
Fig. 13. Throughput vs. ρ (R1 = 0.5, R2 = 0.6, η = 0.5, ζ = −3 dB, α = 0.3, 174–180.
β = 0.3). [4] Z. Wei, J. Guo, D.W.K. Ng, J. Yuan, Fairness comparison of uplink NOMA and
OMA, in: 2017 IEEE 85th Vehicular Technology Conference (VTC Spring),
Sydney, NSW, 2017, pp. 1–6.
by the users, (ii) system outage probability, and (iii) system
[5] A.A. Amin, S.Y. Shin, Channel capacity analysis of non-orthogonal multiple
throughput. Then the paper determined the OPA and OPS factors access with OAM-MIMO system, IEEE Wirel. Commun. Lett. 9 (9) (2020)
for independently minimizing the system outage probability and 1481–1485.
thereafter, the jointly optimal parameters were determined to [6] Md. Fazlul Kader, Mohammed Belal Uddin, Anik Islam, Soo Young Shin,
minimize the system outage probability. In the last part of the Cooperative non-orthogonal multiple access with SWIPT over Nakagami-m
paper, the OPS factors for maximizing the throughputs of SWIPT fading channels, Trans. Emerg. Telecommun. Technol. 30 (5) (2019) 3571.
enabled FD/HD-PSR-NOMA network were determined. Detailed [7] A.A. Amin, S.Y. Shin, Investigate the dominating factor of hybrid SWIPT
protocol by performance analysis of the far user of hybrid SWIPT based
investigations were carried out on the outage and the through-
CNOMA downlink transmission, in: 2019 International Conference on
put performance of SWIPT enabled FD/HD-PSR-NOMA networks. Electrical, Computer and Communication Engineering (ECCE), Cox’sBazar,
It was established that careful selection of OPA and OPS fac- Bangladesh, 2019, pp. 1–6.
tors can improve the performance of the networks considered. [8] D. Kim, H. Lee, D. Hong, A survey of in-band full duplex transmission:
The research work in this paper focused on single input single From the perspective of PHY and MAC layers, IEEE Commun. Surv. Tutor.
output (SISO) NOMA systems. To further improve the spectral 17 (4) (2015) 2017–2046.
[9] E. Everett, D. Dash, C. Dick, A. Sabharwal, Self-interference cancellation
efficiency, the researchers have recently combined NOMA with
in multi-hop full duplex networks via structured signaling, in: Proc. 49th
multi-antenna systems. The resulting MIMO-NOMA system com- AACCCC, IEEE, Monticello, 2011, pp. 1619–1626.
bines multiple-input multiple-output (MIMO) and NOMA tech- [10] M. Duarte, C. Dick, A. Sabharwal, Experiment-driven characterization of
niques to address heterogeneous challenges, such as massive full duplex wireless systems, IEEE Trans. Wirel. Commun. 11 (12) (2012)
connectivity, low latency, high throughput and high reliability 4296–4307.

13
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

[11] E. Everett, D. Dash, C. Dick, A. Sabharwal, Practical, real-time, full duplex [34] Van-Vinh Nguyen, Trong-Tuyen Tran, Vo Viet Tri, Van-Van Huynh, Hoang-
wireless, in: Proceedings of the 17th annual international conference on Sy Nguyen, Miroslav Voznak, Power-splitting protocol nonorthogonal
Mobile computing and networking, Las Vegas, Nevada, USA, Sep. 2011, pp. multiple access (NOMA) in 5G systems: outage performance, in: The Tenth
19–23. International Symposium on Information and Communication Technology
[12] O. Taghizadeh, A.C. Cirik, R. Mathar, Hardware impairments aware (SoICT 2019), Hanoi - Ha Long Bay, Vietnam, Dec. Dec. 4–6, 2019, pp.
transceiver design for full duplex amplify-and-forward MIMO relaying, 224–228.
IEEE Trans. Wirel. Commun. 17 (3) (2018) 1644–1659. [35] J. Tang, et al., Joint power allocation and splitting control for SWIPT-
[13] Recommendation ITU-R M.2083-0, IMT Vision Framework and Overall enabled NOMA systems, IEEE Trans. Wirel. Commun. 19 (1) (2020)
Objectives of the Future Development of IMT for 2020 and beyond, 120–133.
Recommendation ITU, 2083, 2015. [36] Y. Xu, et al., Joint beamforming and power-splitting control in downlink
[14] T.D. Ponnimbaduge Perera, D.N.K. Jayakody, S.K. Sharma, S. Chatzinotas, J. cooperative SWIPT NOMA systems, IEEE Trans. Signal Process. 65 (18)
Li, Simultaneous wireless information and power transfer (SWIPT): recent (2017) 4874–4886.
advances and future challenges, IEEE Commun. Surv. Tutor. 20 (1) (2018) [37] Y. Ye, Y. Li, D. Wang, G. Lu, Power splitting protocol design for the
264–302. cooperative NOMA with SWIPT, 2017 IEEE International Conference on
[15] M.F. Kader, O. Aminul, M.B. Uddin, S.Y. Shin, Interference aided energy Communications (ICC), Paris, 2017, pp. 1–5.
harvesting employing power splitting protocol under the best relay se- [38] Y. Liu, Y. Ye, H. Ding, F. Gao, H. Yang, Outage performance analysis for
lection scheme, in: 2019 International Conference on Information and SWIPT-based incremental cooperative NOMA Networks with Non-Linear
Communication Technology Convergence (ICTC), Jeju Island, Korea (South), Harvester, IEEE Commun. Lett. 24 (2) (2020) 287–291.
2019. [39] M.F. Kader, M.B. Shahab, S.Y. Shin, Cooperative spectrum sharing with en-
[16] Z. Zhu, Z. Chu, F. Zhou, H. Niu, Z. Wang, I. Lee, Secure beamforming designs ergy harvesting best secondary user selection and non-orthogonal multiple
for secrecy MIMO SWIPT systems, IEEE Wirel. Commun. Lett. 7 (3) (2018) access, in: 2017 International Conference on Computing, Networking and
424–427. Communications (ICNC), Santa Clara, CA, 2017.
[17] Z. Zhu, S. Huang, Z. Chu, F. Zhou, D. Zhang, I. Lee, Robust designs of [40] A.A. Amin, S.Y. Shin, Performance analysis of cooperative non-orthogonal
beamforming and power splitting for distributed antenna systems with multiple access with improved time switching simultaneous wireless infor-
wireless energy harvesting, IEEE Syst. J. 13 (1) (2019) 30–41. mation and power transfer protocol, Trans. Emerg. Telecommun. Technol.
[18] H. Niu, B. Zhang, Y. Huang, D. Guo, Z. Chu, Z. Zhu, Robust secrecy (2020) e4077.
beamforming and power-splitting design for multiuser MISO downlink [41] Y. Zhang, J. He, S. Guo, F. Wang, Energy efficiency maximization in wireless
with SWIPT, IEEE Syst. J. 13 (2) (2019) 1367–1375. powered networks with cooperative non-orthogonal multiple access, IET
[19] Z. Chu, Z. Zhu, M. Johnston, S.Y. Le Goff, Simultaneous wireless information Commun. 12 (18) (2018) 2374–2383.
power transfer for MISO secrecy channel, IEEE Trans. Veh. Technol. 65 (9) [42] Jie Tang, Jingci Luo, Mingqian Liu, Daniel K.C. So, Emad Alsusa, Gaojie Chen,
(2016) 6913–6925. Kai-Kit Wong, Jonathon A. Chambers, Energy efficiency optimization for
NOMA with SWIPT, J. Sel. Top. Signal Process. 13 (3) (2019) 452–466.
[20] Z. Chu, Z. Zhu, J. Hussein, Robust optimization for AN-aided transmission
[43] Tewodros A. Zewde, Mustafa C. Gursoy, NOMA-based energy-efficient
and power splitting for secure MISO SWIPT system, IEEE Commun. Lett.
wireless powered communications, IEEE Trans. Green Commun. and Netw.
20 (8) (2016) 1571–1574.
2 (3) (2018) 679–692.
[21] Liu Yuanwei, Ding Zhiguo, Elkashlan Maged, Poor H. Vincent, Cooperative
[44] J. Liu, K. Xiong, Y. Lu, P. Fan, Z. Zhong, K.B. Letaief, SWIPT-enabled
non-orthogonal multiple access with simultaneous wireless information
full-duplex NOMA networks with full and partial CSI, IEEE Trans. Green
and power transfer, IEEE J. Sel. Areas Commun. 34 (4) (2016) 938–953.
Commun. Netw. (2020) http://dx.doi.org/10.1109/TGCN.2020.2977611.
[22] L.T. Dung, et al., Analysis of partial relay selection in NOMA systems with
[45] H. Huang, M. Zhu, Energy efficiency maximization design for full du-
RF energy harvesting, in: 2nd Int. Conference on Recent Advances in Signal
plex cooperative NOMA systems with SWIPT, IEEE Access 7 (2019)
Process., Telecommun. & Comput. (SigTelCom), IEEE, Ho Chi Minh, Vietnam,
20442–20451.
2018, pp. 13–18.
[46] W. Wu, X. Yin, P. Deng, T. Guo, B. Wang, Transceiver design for downlink
[23] Z. Yang, Z. Ding, P. Fan, N. Al-Dhahir, The impact of power allocation on
SWIPT NOMA systems with cooperative full-duplex relaying, IEEE Access
cooperative non-orthogonal multiple access networks with SWIPT, IEEE
7 (2019) 33464–33472.
Trans. Commun. 16 (7) (2017) 4332–4343.
[47] E. Everett, M. Duarte, C. Dick, A. Sabharwal, Empowering full-duplex wire-
[24] Weiliang Han, Jianhua Ge, Jinjin Men, Performance analysis for NOMA
less communication by exploiting directional diversity, in: 2011 Asilomar
energy harvesting relaying networks with transmit antenna selection and
Conference on Signals, Systems and Computers, 2011, pp. 2002–2006.
maximal-ratio combining over Nakagami-m fading, IET Commun. 10 (18)
[48] E. Everett, A. Sahai, A. Sabharwal, Passive self-interference suppression for
(2016) 2687–2693.
full-duplex infrastructure nodes, IEEE Trans. Wirel. Commun. 13 (2) (2014)
[25] Yangyang Zhang1, Jianhua G, Performance analysis for nonorthogonal 680, 694.
multiple access in energy harvesting relaying networks, IET Commun. 11 [49] J.I. Choi, M. Jain, K. Srinivasan, P. Levis, S. Katti, Achieving single channel,
(11) (2017) 1768–1774. full duplex wireless communication, in: Proc. ACM MobiCom, Chicago, IL,
[26] Fudong Li, Hai Jiang, Rongfei Fan, Peng Tan, Cognitive nonorthogonal Sep. 2010.
multiple access with energy harvesting: an optimal resource allocation [50] T. Riihonen, S. Werner, R. Wichman, Mitigation of loopback self-
approach, IEEE Trans. Veh. Technol. 68 (7) (2019) 7080–7095. interference in full-duplex MIMO relays, IEEE Trans. Signal Process. 59 (12)
[27] N.T. Do, D.B. Da Costa, T.Q. Duong, B. An, A BNBF user selection scheme (2011) 5983–5993.
for NOMA-based cooperative relaying systems with SWIPT, IEEE Commun. [51] E. Ahmed, A.M. Eltawil, All-digital self-interference cancellation tech-
Lett. 21 (3) (2017) 664–667. nique for full-duplex systems, IEEE Trans. Wirel. Commun. 14 (7) (2015)
[28] T.N. Do, D.B. da Costa, T.Q. Duong, B. An, Improving the performance 3519–3532.
of cell-edge users in MISO-NOMA systems using TAS and SWIPT-based [52] M. Duarte, A. Sabharwal, V. Aggarwal, R. Jana, K.K. Ramakrishnan, C.W.
cooperative transmissions, IEEE Trans. Green Commun. Netw. 2 (1) (2018) Rice, N.K. Shankaranarayanan, Design and characterization of a full-duplex
49–62. multi-antenna system for wifi networks, IEEE Trans. Veh. Technol. 63 (3)
[29] X. Li, L. Meng, et al., Joint effects of residual hardware impairments and (2014) 1160–1177.
channel estimation errors on SWIPT assisted cooperative NOMA networks, [53] Z. Zhang, K. Long, A.V. Vasilakos, L. Hanzo, Full-duplex wireless commu-
IEEE Access 7 (2019) 135499-135513. nications: Challenges, solutions, and future research directions, Proc. IEEE
[30] Xingwang Li, Jingjing Li, Lihua Li, Performance analysis of impaired SWIPT 104 (7) (2016) 1369–1409.
NOMA relaying networks over imperfect Weibull channels, IEEE Syst. J. 14 [54] Guoling Liu, et al., Performance analysis and optimization of cooperative
(1) (2019) 1–4. full-duplex D2D communication underlaying cellular networks, IEEE Trans.
[31] Y. Liu, et al., Outage performance analysis for SWIPT-based cooperative Wirel. Commun. 18 (11) (2019) 5113–5127.
non-orthogonal multiple access systems, IEEE Commun. Lett. 23 (9) (2019) [55] S. Sohaib, M. Uppal, Full-duplex compress-and-forward relaying un-
1501–1505. der residual self-interference, IEEE Trans. Veh. Technol. 67 (3) (2018)
[32] Ha Dac-Binh, Sang Quang Nguyen, Outage performance of energy har- 2776–2780.
vesting DF relaying NOMA networks, Mobile Netw. Appl. 23 (6) (2018) [56] T.M.C. Chu, H.J. Zepernick, Performance of a non-orthogonal multiple
1572–1585. access system with full-duplex relaying, IEEE Commun. Lett. 22 (10) (2018)
[33] Huu Q. Tran, Tien-Tung Nguyen, Ca V. Phan, Quoc-Tuan Vien, On the 2084–2087.
performance of NOMA in SWIPT systems with power-splitting relaying, [57] Dexin Wang, et al., Capacity-enhancing full-duplex relay networks based
in: 19th Int. Symp. Commun. and Inf. Technol. (ISCIT), Ho Chi Minh City, on power-splitting (PS-) SWIPT, IEEE Trans. Veh. Technol. 66 (6) (2016)
Vietnam, Vietnam, 2019, pp. 25–27. 5445–5450.

14
Aswathi V. and Babu A.V. Physical Communication 46 (2021) 101335

[58] H. Liu, K.J. Kim, K.S. Kwak, H. Vincent Poor, Power splitting-based SWIPT Babu A.V. received the Master of Engineering degree in
with decode-and-forward full-duplex relaying, IEEE Trans. Wirel. Commun. Telecommunication from the Department of Electrical
15 (11) (2016) 7561–7577. Communication Engineering, Indian Institute of Science,
[59] I.S. Gradshteyn, I.M. Ryzhik, Table of Integrals, Series and Products, seventh Bangalore, India, in 2002 and Ph.D. degree from the
ed., Academic, New York, NY, USA, 2007. Department of Electronics and Communication Engi-
neering, National Institute of Technology, Calicut, India,
in 2008, where he is currently employed as Professor.
He has authored or coauthored more than 75 papers
Aswathi V. received B.tech degree from Calicut Uni- in reputed international journals and conferences. His
versity in electronics and Communication Engineering primary research focus is on wireless networks and the
in 2014 and M. Tech degree in Communication Engi- current research interests include resource allocation
neering and Signal Processing from Calicut University for NOMA based 5G wireless networks. He is a senior member of IEEE.
in 2016. Currently, she is pursuing Ph.D in Electronics
and Communication Department at National Institute
of Technology, Calicut. Her current research inter-
est includes Non-orthogonal multiple access based 5G
networks.

15