Performance Analysis of Downlink NOMA System Relying on Energy Harvesting and Full-Duplex 1

ECTI Transactions on Computer and Information Technology

Journal homepage: https://ph01.tci-thaijo.org/index.php/ecticit/

Published by the ECTI Association, Thailand, ISSN: 2286-9131

Performance Analysis of Downlink NOMA System Relying

on Energy Harvesting and Full-Duplex
1 2
Chi-Bao Le and Dinh-Thuan Do

ABSTRACT Article information:

This paper considers the implementation of wireless power transfer scheme Keywords: Full-duplex, Power
to non-orthogonal multiple access (NOMA) network at downlink. The sin- Beacon, NOMA, Outage Proba-

gle antenna power beacon (PB) provide capability of wireless charge to two bility, Simultaneous wireless in-

NOMA users. In addition, full-duplex (FD) scheme is employed for two formation and power transfer

users. Outage probabilities related to system performance are evaluated in

the scenario of the multi-antenna base station. Improved performance can Article history:
be achieved since multi-antenna, energy harvesting capability. In particu- Received: October 11, 2020

lar, expressions of outage performance are derived in this system model to Revised: January 8, 2021

help to forward to serve far NOMA users. As the main advantage, NOMA Accepted: January 28, 2021

and full-duplex schemes are two advantages and such a system is deployed Published: February 12, 2022

to enhance the spectral eciency. Our simulation results will indicate that (Online)
main parameters such as a power allocation strategy, energy harvesting
time, self-interference factor are the main impacts on considered system.
The correct derived expressions are presented by matching Monte-Carlo
simulation curves and analytical curves in numerical result section. As our
main results, FD NOMA system corresponding the rst user and the sec-
ond user exhibits better than OMA 30% and 25% respectively when SNR
at the source is 30 dB.

DOI: 10.37936/ecti-cit.2022161.242809

1. INTRODUCTION transmit power to access the same frequency [69].

To separate the dierent signals, the successive in-
In recent work related to Full-duplex (FD) com-
terference cancellation (SIC) is implemented at the
munication, its advantage motivated many applica-
receiver [9]. In emerging network of cognitive radio,
tions since transceivers allow simultaneous downlink
the NOMA is implemented with cognitive radio net-
(DL) and uplink (UL) transmission over the same fre-
works and it will not only improve the spectrum e-
quency band [14]. However, FD in this scheme leads
ciency but also serve more secondary users. Thanks
to the expense of introducing strong self-interference
to advantages of both the cognitive radio and NOMA
(SI). A suboptimal precoding and power allocation
techniques, and it benets to the 5th generation mo-
algorithm is studied to provide maximization of the
bile networks [10].
sum throughput FD systems serving multiple users
[2]. In addition, the end-to-end throughput of an The limited performance is resulted from the lim-
FD relay system was evaluated in terms of trans- ited energy supply of small devices in wireless net-
mit beamforming design for maximization and the work. In particular, it is hard to replace the bat-
optimal joint receive [3]. The minimum signal-to- tery and/or there is no power line. To overcome this
interference-plus-noise ratio (SINR) is considered as problem, by harvesting the energy from the surround-
providing the DL and UL users in a FD system [4]. ing environments, they considered energy harvest-
The authors studied the optimal beamforming and ing technique which is so-called as Radio frequency
power allocation design [5]. In other improved spec- (RF) energy harvesting [1115]. In this model, en-
trum eciency scheme, non-orthogonal multiple ac- ergy harvesting-assisted device is able to harvest the
cess (NOMA) permits multiple users with dierent energy from the radio-frequency signals. Such energy

1 The authors is with Faculty of Electronics Technology, Industrial University of Ho Chi Minh City (IUH), Ho Chi Minh City
700000, Vietnam, E-mail: lechibao@iuh.edu.vn
2 The authors is with Wireless Communications Research Group, Faculty of Electrical & Electronics Engineering, Ton Duc
Thang University, Ho Chi Minh City 700000, Vietnam, E-mail: dodinhthuan@tdtu.edu.vn
2 ECTI TRANSACTIONS ON COMPUTER AND INFORMATION TECHNOLOGY, Vol.16, No.1, March 2022

harvesting scheme can provide exible, sustainable 2. SYSTEM MODEL

and stable energy supply for cognitive radio networks
[1620]. In [17] and [18], the primary spectrum or
transmit power in the primary network provide RF
energy for the secondary users with dynamically spec-
trum sensing technique and for battery charging. To
enhance the secondary throughput by utilizing the
harvested RF energy, an optimal channel selection
method is introduced [19]. In cognitive radio sensor
networks, the RF energy harvesting was also studied
as in [20]. In fact, the RF energy harvesting model is
ideal linear in most of recent works related to energy
harvesting. However, the non-linear behaviors need
be considered in the practical energy harvesting cir-
cuits [2123]. In addition, the non-linear RF energy
harvesting model is proposed and examined system Fig.1: System model of wireless powered NOMA
performance of the secondary performance [2123]. system relying power beacon.
In other trends of research, the integration of FD
and NOMA was advocated in thanks to the poten-
It can be seen from Fig. 1, the system include
tial benets of FD transmission [2426]. In [24],
multi-antenna base station (BS) serves two NOMA
maximization of the weighted system throughput in
users at downlink. It can be facilitated the BS with
FD MCNOMA systems can be achieved by study-
multiple antennas, NS . Link BS-user U1 , and link
ing the optimal power and subcarrier allocation al-
BS-user U2 are characterized by h1 , h2 , respectively.
gorithm design. A NOMA-based FD relaying system
Two users U1 ; U2 is facilitated with FD mode, f1 ; f2
is evaluated in terms of the outage probability and
are self-interference channel related two users U1 ; U2 ,
the ergodic sum rate [25]. In [26], outage probabil-
respectively. In this scenario, wireless charge is served
ity of NOMA-based FD relaying systems was Inves-
by the power beacon (PB) equipped with one trans-
tigated by introducing the optimal power allocation
mit antenna. Then, g1 , g2 are channels to provide
minimization.
energy harvesting ability to two users.
However, due to little amount of harvested power
at users in previous work. Such open problem mo-
tivated us to study in this paper a new model of
multi-antenna power beacon which serve as wireless
charge to far NOMA users.

Table 1: Key parameters of the system model.

Symbol Description
The power allocation coecient with
ai
i ∈ {1, 2}, a1 + a2 = 1 and a1 < a2
PS The transmit power at BS
Pp The transmit power at PB Fig.2: Energy harvesting protocol.
The AWGN noise term followed nUi ∼
ni 2


CN 0, σU I
i NS
xi The information of Ui Fig. 2 describes the key parameters related infor-
Ri The target rate at Ui mation processing and energy harvesting. In partic-
Time duration of signal frame as con- ular, there are two time slots in the time switching-
T
sideration in Fig. 2 based relaying (TSR) protocol for energy harvesting
η The energy conversion eciency
and information processing at the relay. In Fig. 2,
Time percentage to perform wireless
α information is transmitted from the source node to
power transfer
h1 The channel link between BS and U1 the destination node in αT , in which T is the block
h2 The channel link between BS and U2 time, and α is the fraction of the block time allocated
g1 The channel link between PB and U1 for energy harvesting, and its condition is 0 ≤ α ≤ 1.
g2 The channel link between PB and U2 The remaining block time, (1 − α) T is used informa-
f1 The channel self-interference at U1 tion transmission from the source to destination.
f2 The channel self-interference at U2
It is noted that proper selection of the time frac-
tion, α is important and it is used for optimal amount
of harvesting energy at the relay node, and then it
aects the achievable throughput at the destination.
Performance Analysis of Downlink NOMA System Relying on Energy Harvesting and Full-Duplex 3

The following subsections analyze performance of the (i ∈ {1, 2}) with τ ≥ 1 is integer. It is noted that the
energy harvesting and information processing at the time k satises the relationship k ≥ τ . nUi [k] the zero
relay node. mean additive white Gaussian noise (AWGN) vector.
2 2 2 2


The channel power gains kh1 k , kh2 k , |g1 | and These noise terms follow nUi ∼ CN 0, σU I
i NS
with
2 2 2
|g2 | are assumed to be exponentially distributed ran- equal variance σU IN = σ and fi ∼ CN (0, Ωfi ) is
i
dom variables (RVs) with the parameters Ωh1 , Ωh2 , self-interference channel at the Ui .
Ωg1 and Ωg2 , respectively. We call PS as trans- In NOMA, SIC is required to U1 rst detect U2
mit power of the BS. Especially, the maximum ratio who has a larger transmit power, which has less in-
transmission (MRT) is used with beamforming vector ference signal. Then the signal of U2 can be detected
to achieve optimal transmission scheme as in [29, 31], from the superposed signal.
with kwi k = 1, i = 1, 2 Therefore, the received signal-to-interference-plus-
noise ratio(SINR) need be computed and such com-
hi
wi = khi k , i ∈ {1, 2}, (1) putation performing at U1 to detect U2 's message x2

wherek.k denotes the Euclidean norm of a ma- 2

U1 a2 PS kh1 k
N ×1
trix. hi ∈ C S and gi ∈ C
1×1
are the BS − Ui and γ2→1 = 2 2
a1 PS kh1 k + PU1 |f1 | + σ 2
P B − Ui link channel vector, respectively. During en- 2
(5)
ergy harvesting time αT , the harvested energy, EUi a2 ρS kh1 k
= 2 2 2 ,
is given by [30, Eq. (2)] a1 ρS kh1 k + θ|g1 | |f1 | + 1
 2
ηPS α|g1 | T ,i ∈ 1 whereρS = PσS2 is the transmit signal-to-noise radio
EUi = 2 , (2) ηρS α
ηPS α|g2 | T ,i ∈ 2 (SNR) and θ =
(1−α) .
In NOMA, two signals x1 and x2 are superimposed
where η is the energy conversion eciency. De-
at the BS to transmit to far users. They are normal-
pending on the quality of energy harvesting electric  2 
circuitry, 0 < η < 1. ized unity power signals, i.e, E x1 = E x22 = 1
in which E {.} is the expectation operator.
From obtained EUi in (2), the transmitted power
from relay node, PUi is given by
After SIC, the received SINR at U1 can be calcu-
lated. SINR need be known to detect its own message

PUi =EUi /(1 − α) T x1 and it is given by

 .
 ηPS α|g1 |2 (1 − α) , i ∈ 1 (3) a1 ρS kh1 k
2
= . γ1U1 = 2 2 . (6)
 ηPS α|g2 |2 (1 − α) , i ∈ 2 θ|g1 | |f1 | + 1

Similarly, the received SINR at U2 need be com-

To implement NOMA scheme, PS is also as the
puted to detect x2 and it is formulated by
normalized transmission powers at the BS. x1 and x2
are the signals for U1 and U2 , respectively while a1 a2 ρS kh2 k
2
and a2 are the corresponding power allocation coef- γ2U2 = 2 2 2 . (7)
cients. To make better fairness between the relay
a1 ρS kh2 k + θ|g2 | |f2 | + 1
node, it can be assumed that a1 < a2 , and they are
constrained by a1 + a2 = 1.
3. OUTAGE PROBABILITY ANALYSIS
During the k − th time slot, U1 and U2 receives the
superposed signal and loop interference signal simul- In preliminary, we assume that all channel co-
taneously. The observation at U1 and U2 are given ecients are modeled as independent Rayleigh-
by distributed random variables (RVs).

p p  It can be denoted that fXi (x) as the probability

yU1 [k] = kh1 w1 k a1 PS x1 [k] + a2 PS x2 [k] ∆ 2
density functions (PDFs) of Xi = khi k is written as
p , [32]
+ PU1 f1 xLI−1 [k − τ ] + nU1 [k]
− Ωx
| {z } | {z }
self −interf erence AW G N xNS −1
fXi (x) = N e Xi
, i ∈ {1, 2} (8)
(4a) (NS −1)!ΩXS
i
p p 
yU2 [k] = kh2 w2 k a1 PS x1 [k] + a2 PS x2 [k] Next, FXi (x) is the cumulative distribution func-
p , tion (CDF) of Xi is given by [33]
+ PU2 f2 xLI−2 [k − τ ] + nU2 [k]
| {z } | {z }
AW G N − Ωx S −1
NP
self −interf erence xn
FXi (x) = 1 − e Xi
n!Ωn , i ∈ {1, 2} (9)
(4b) n=0 X i

where xLI−i [k − τ ] is called as loop interference Regarding self-interference channel due to FD de-
signal, τ stands for the processing delay at Ui ployment, fYi (x) is the probability density functions
4 ECTI TRANSACTIONS ON COMPUTER AND INFORMATION TECHNOLOGY, Vol.16, No.1, March 2022

∆
(PDFs) of Yi = |fi |
2
is formulated by OPUF1D is given by

− Ωx S −1 X
NX n 
1 k!ζ n θk Ωkf1

fYi (x) = e Yi
, i ∈ {1, 2}. (10) n
ΩYi OPUF1D =1 −
k n!Ωg1 Ωn−k−1
n=0 k=0 h1
We have FYi (x) is the cumulative distribution Z∞
function (CDF) of Yi is given by − Ωζ xk − Ωx
×e h1
k+1
e g
1 dx.
(Ωh1 + Ωf1 ζθx)
− Ωx 0
FYi (x) = 1 − e Yi
, i ∈ {1, 2}. (11)
(14)

 
Let t = arctan (x) − 1 ⇒ tan π(t+1)
4
π 4 = x ⇒
3. 1 Analysis on Outage performance in FD π 2 π

1
mode 4 sec 4 (t + 1) dt = dx with sec (x) = cos(x) , (14)
is expressed as

In principle, the target rates of relay node are

S −1 X
NX n 
πk!ζ n θk Ωkf1

determined since they are related quality of service n
OPUF1D =1 −
(QoS). In particular, the outage probability is con- k n!4Ωg1 Ωn−k−1
n=0 k=0 h1
sidered as an important metric and then performance
Z1
evaluation can be achieved. − Ωζ −
tan(π(t+1)/4)
×e h1
sec2 (π (t + 1)/4) e Ωg
1
We will evaluate the outage performance in two
−1
representative users in NOMA in the following.
k
Outage Probability of U1 : tan (π (t + 1)/4)
× k+1
dt.
According to NOMA protocol, the complementary (Ωh1 + Ωf1 ζθ tan (π (t + 1)/4))
events of outage at U1 can be required as: U1 is able (15)

to detect x2 rstly and then it detects its own mes-

sage x1 . Therefore, the outage probability of U1 is Unfortunately, it is dicult to derive a closed-form
formulated by expression for (15), an accurate approximation can be
  obtained using Gaussian-Chebyshev quadrature [28,
U1
OPUF1D = Pr γ2→1 < γ2F D ∪ γ1U1 < γ1F D Eq. (25.4.38)], it is given as
 
U1
=1 − Pr γ2→1 ≥ γ2F D , γ1U1 ≥ γ1F D S −1 X
NX n X
R 
n

k!π 2 ζ n θk Ωkf1
OPUF1D ≈1 −

2

2 2
 k n!4RΩg1 Ωn−k−1
=1 − Pr kh1 k ≥ ζ θ|g1 | |f1 | + 1 , n=0 k=0 r=1 h1
p
(12) 1 − ξr2 sec2 (π (ξr + 1)/4)
×
where γ1F D = 2R1 − 1 with R1 being the target rate (Ωh1 + Ωf1 ζθ tan (π (ξr + 1)/4))
k+1
at U1 to detect x1 and γ2F D = 2R2 − 1 with R2 be- ζ tan(π(ξr +1)/4)
γ1F D k − Ωh − Ωg
ing the target rate at U1 to detect x2 , ϕ1 = × tan (π (ξr + 1)/4) e 1 1 ,
a1 ρS ,
γ2F D (16)
ϕ2 = ρS (a2 −γ2DL a1 )
and ζ = max (ϕ2 , ϕ1 ) 
(2r−1)

where ξr = cos 2R π
Then, OPUF1D can be solved by


2

2 2
 Outage Probability of U2 :
OPUF1D =1 − Pr kh1 k ≥ ζ θ|g1 | |f1 | + 1
Z∞ Z∞ The outage probability of U2 is computed as
=1 − f|g1 |2 (x) f|f1 |2 (y)  
0 0 OPUF2D = Pr γ2U2 < γ2F D
h i   
× 1 − Fkh1 k2 (ζ (θxy + 1)) dxdy 2 2 2
=1 − Pr kh2 k < ϕ2 θ|g2 | |f2 | + 1
ζ (13)
S −1
NX −
ζ n e Ωh1 Z∞ Z∞
=1 − =1 − f|g2 |2 (x) f|f2 |2 (y)
n=0
n!Ωg1 Ωf1 Ωnh1
0 0
Z∞ Z∞ 
1 ζθx

− Ωx −y Ωf +Ω × F̄kh2 k2 (g (x, y)) dxdy,
× g
e 1 e 1 h1

(17)
0 0
h i
n where F̄kh2 k (g (x, y)) = 1 − Fkh2 k (ϕ2 (θxy + 1))
2 2
× (θxy + 1) dxdy.

Based on [27, Eq. (1.111)] and [27, Eq. (3.351.3)], Similarly with solving OPUF1D , it can be obtained
Performance Analysis of Downlink NOMA System Relying on Energy Harvesting and Full-Duplex 5

OPUF2D as

S −1 X
NX Q 
n X 
k!π 2 ϕn2 θk Ωkf2 OPUHD,ipSIC
n 1
OPUF2D
  
≈1 − 2 1 kh1 k2
k = 1 − Pr kh1 k ≥ Φmax , |hI | ≤ −1
n=0 k=0 q=1
n!4QΩg2 Ωn−k−1
h2  
ρS Φ1
1 x
−1
q R∞ ρS
RΦ1
1 − ξq2 sec2 (π (ξq + 1)/4) =1− fkh1 k2 (x) f|hI | (y)dxdy
× k+1 Φmax 0
(Ωh2 + Ωf2 ϕ2 θ tan (π (ξq + 1)/4)) 1
R∞ N −1 − Ωx
ϕ tan(π (ξq +1)/4)
=1− NS x S e h1
2 Γ(NS )Ωh
k − Ωh − Ωg 1 Φmax 
× tan (π (ξq + 1)/4) e 2 2 , 
− 1 x
−1

  (18) × 1 − e ρS ΩhI Φ1 dx
(2q−1)
where ξq = cos 2Q π
 Φ
− Ωmax NP S −1
Φrmax
S −1
NP
Φrmax
1 Ξr2

1
eχ
=1− e h1
r !Ω 1
r − NS NS r2 ! ,
r1 =0
1 h1 Ωh Ξ r2 =0
1
(21)
Φmax
3. 2 Consideration on half-duplex (HD) mode where Ξ = Ω1h + ρS Φ11 Ωh , χ = ρS Ω
1
hI
− Ξ and
1 I

With HD mode, the received signal at Ui , i = 1, 2 Γ (x) = (x − 1)! is the Gamma function
node in the downlink of NOMA is given by Based on (19b), the outage probabilities with per-
fect SIC for the U2 link they can be written as
p 
OPUHD =1 − Pr γ2 ≥ γ1HD
p

yU1 [k] = kh1 w1 k a1 PS x1 [k] + a2 PS x2 [k] 2
!
+ nU1 [k] , γ1HD
=Fkh2 k2

(19a) ρS a2 − a1 γ1HD
HD (22)
p p  γ1
yU2 [k] = kh2 w2 k a1 PS x1 [k] + a2 PS x2 [k] − HD )
Ωh ρS (a2 −a1 γ1
=1 − e 1

+ nU2 [k] . S −1
NX
n
γ1HD
(19b) ×
n ,
n=0
n!ρnS a2 − a1 γ1HD Ωnh2

a2 ρS kh2 k2
Based on (19a), the outage probabilities with im- where γ2 = a1 ρS kh2 k2 +1
perfect SIC for the U1 link they can be written as

OPUHD,ipSIC
1   3. 3 Asymptotic Outage Probability Analysis
= 1 − Pr γ2→1 ≥ γ2HD , γ1ipSIC ≥ γ1HD for Two User's
  2

= 1 − Pr kh1 k ≥ Φmax , |hI | ≤ ρ1S khΦ11k − 1 ,
2
We conduct the asymptotic outage probability
(20) analysis in the high SNR region. Based on analytical

2 2 result in (16) and (18), when ρS → ∞, the asymp-

S kh1 k S kh1 k
where γ2→1 = a aρ2 ρkh 2
ipSIC
, γ1 = aρ1Sρ|hI |+1
, totic outage probability of U1 andU2 for FD NOMA
1 S 1 k +1
hI ∼ CN (0, ΩhI ) with ΩhI (0 ≤ ΩhI < 1) denotes with e−x → 1 − x is given by
as the level of residual interference caused by im-
S −1 X
NX n X
R 
k!π 2 ζ n θk Ωkf1

perfect SIC. More precisely, the value of ΩhI = 0 n
and ΩhI < 1 are the perfect SIC (pSIC) and im-
OPUF1D,∞ ≈1 −
k n!4RΩg1
n=0 k=0 r=1
perfect SIC (ipSIC), respectively, γ1HD = 22R1 − 1,
k
γ2HD = 22R2 − 1, Φ1 =
γ1HD
, Φ2 =
γ2HD
and
× Ωhk+1−n
1
Υ (ξr ) tan (π (ξr + 1)/4)
a1 ρS (
ρS a2 −γ2HD a1 )   tan(π(ξ +1)/4)
ζ − r
Φmax = max (Φ1 , Φ2 ). × 1− e Ω g1
,
Ωh1
(23)
By exploiting result from (8), (10) and with the and
help of the [27, Eq. (3.351.2)], OPUHD,ipSIC is calcu-
S −1 X Q 
1 NX n X
k!π 2 ϕn2 θk Ωkf2

lated as n
OPUF2D,∞ ≈1 −
k n!4QΩg2
n=0 k=0 q=1
k
× Ωk+1−n
h2 Ψ (ξq ) tan (π (ξq + 1)/4)
  tan(π(ξq +1)/4)
ϕ2 −
× 1− Ωg
e 2 ,
Ωh2
(24)
6 ECTI TRANSACTIONS ON COMPUTER AND INFORMATION TECHNOLOGY, Vol.16, No.1, March 2022

   
(2r−1) (2q−1) In Fig. 3, the curves of outage probability for two
where ξr = cos 2R π , ξq = cos 2Q π ,
√ NOMA users versus transmit SNR ρS at the BS is
1−x2 sec2 (π(x+1)/4)
Υ (x) = k+1 and Ψ (x) = observed. It can be seen clearly that the theoreti-
√
(Ωh1 +Ωf1 ζθ tan(π(x+1)/4))
1−x2 sec2 (π(x+1)/4) cal curves achieved by mathematical analysis shows
k+1 .
4Q(Ωh2 +Ωf2 ϕ2 θ tan(π(x+1)/4)) a considerable match with the Monte Carlo simula-
The same situation can be seen for HD mode, the tions. The saturation trends can be seen it FD mode
asymptotic outage probability of U1 and U2 for HD for both users at high SNR, as ρS is greater than 40
NOMA with e−x → 1 − x is given by (dB). The reason is that self-interference exists due
to FD mode and it make system performance become
S −1
"  NX
Φmax Φrmax
1 worse. The gure clearly shows that with increased
OPUHD,ipSIC,∞ =1 − 1− SNR at the BS, outage will be improved signicantly.
1
Ωh1 r1 =0
r1 !Ωrh11
It is conrmed that outage performance of user U2
NS −1 r1
#
(1 − χ) X Φmax Ξr2 is better than that of user U1 . It can be explained
− NS , that performance gap exist due to dierent power al-
Ωh1 ΞNS r2 =0 r2 !
located a1 ; a2 . In contrast with FD mode, higher
(25)
SNR leads to improved outage at HD mode. Fur-
and
thermore, asymptotic curves tend to locate near to
the curves of exact computation.
!
γ1HD
OPUHD,∞ =1 − 1−

2
ρS a2 − a1 γ1HD

n (26)
S −1
NX
γ1HD
×
n .
n=0
n!ρnS a2 − a1 γ1HD Ωnh1

4. NUMERICAL RESULTS

To conrm accuracy of the proposed analytical ex-

pressions in term of the outage probabilities, this sec-
tion provides the simulated results which are derived
in the previous section. Furthermore, the lowest out-
age performance can be analyzed to evaluate the per-
formances of such NOMA system under dierent sim-
ulated parameters. Finally, the eect of main param-
Fig.4: Comparison on NOMA and OMA, with
eters on the performances is further determined. We
ΩhI = −40 (dB).
use a1 = 0.2 and a2 = 0.8 as power allocation fac-
tors. Regarding energy harvesting, we setη = 1 and
α = 0.2. R1 = 2 and R2 = 1. To per-
Target rates are
form approximate computation, Q = R = 500. The
channel gains are Ωh1 = Ωg1 = 0.6, Ωg2 = Ωh2 = 0.5
and Ωf1 = Ωf2 = 0.01.

Fig.5: Outage probability versus time-aware en-

ergy harvesting factor α, with ρS = 30 (dB) and
ΩhI = −30 (dB).

Next, in Fig. 4, we investigate performance dier-

Fig.3: Outage probability versus ρS at U1 and U2 , ences between OMA and NOMA. It can be observed
with ΩhI = −40 (dB) that outage behavior in NOMA with HD mode is still
Performance Analysis of Downlink NOMA System Relying on Energy Harvesting and Full-Duplex 7

better than that if OMA in entire SNR region.

Regarding energy harvesting, Fig. 5 evaluate im-
pact of level of harvested energy α on the outage
probability in FD mode. The reason is that in FD
mode, self-interference channel is resulted from har-
vested power. Increasing percentage of harvested
power leads to worse performance because less time
for information processing. While HD mode does not
depend on interference channel, and then outage per-
formance keeps stable in whole range of α.

Fig.8: Outage probabilities versus a1 , with ΩhI =

−40 (dB), R1 = 2, R2 = 0.5 and ρS = 25 (dB).

Fig. 8 shows the impact of power allocation coef-

cient a1 on the outage event of considered schemes.
It is conrmed that outage performance of U2 better
than that of U1 at very low value of a1 . Interest-
ingly, optimal outage for user U1 in HD mode can be
achieved at a1 = 0.5.

Fig.6: Impact of self-interference channel on outage

behavior, with ρS = 25 (dB).

Regarding energy harvesting, Fig. 6 evaluate im-

pact of level of interference on the outage probabil-
ity in FD mode. Furthermore, as in previous gures,
more antennas equipped at the BS lead to better out-
age performance.

Fig.9: Throughput performance of two relays node,

with ΩhI = −40 (dB), R1 = 2 and R2 = 1.

As seen from Fig. 9, since the throughput is com-

puted based on achievable outage probability, it can
be further evaluate throughput in delay-limited trans-
mission mode as τ`ι = (1 − OP`ι ) R` with ` ∈ {1, 2}
and ι ∈ {F D, HD}. The throughput of two relays
node increase signicantly since data rates are main
factor. It is seen clearly that throughput increases

Fig.7: Outage performance versus xed data rates once data rates increase.

R1 and R2 , with ρS = 25 (dB), ΩhI = −40 (dB),

a1 = 0.1 and a2 = 0.9. 5. CONCLUSION

In this paper, we developed a theoretical frame-

Fig. 7 plots the outage probability for two NOMA work to analyze NOMA downlink system perfor-
users as increasing target rates. The performance gap mance under deployment of FD mode and energy
among two users remains at low region from 0.5 to 1.5, harvesting. Analytical results are derived to show
but at high required data rate, system meets outage reasonable outage performance. For the downlink
event. NOMA system, two dierent outage performance of
8 ECTI TRANSACTIONS ON COMPUTER AND INFORMATION TECHNOLOGY, Vol.16, No.1, March 2022

two users can be shown due to dierent power alloca- systems with imperfect CSI, IEEE J. Sel. Areas
tion factors for each NOMA user. The study shows Commun., vol. 35, no. 12, pp. 2874-2885, 2017.
that increasing transmit SNR leads to signicant per- [10] F. Zhou, Y. Wu, Y.-C. Liang, Z. Li, Y. Wang
formance improvement. It is also provided guidelines and K.-K. Wong, State of the art, taxonomy,
as reasonable selection of the number of antennas at and open issues on cognitive radio networks with
the BS are performed, outage probability meets ideal NOMA, IEEE Wireless Commun., vol. 25, no. 2,
value. pp. 100-108, Apr. 2018.
[11] K. Woradit, S. Srirai, S. Kitjarunerungroj,

ACKNOWLEDGMENT T. Kodmatcha and P. Sangmahamad, Multi-

user Secrecy SWIPT for 5G OFDMA Net-
The authors would like to thank the anonymous re-
works with Particle Swarm Optimizations,
views for the helpful comments and suggestions. This
2019 IEEE International Conference on Con-
research is supported by Industrial University of Ho
sumer Electronics - Asia (ICCE-Asia), Bangkok,
Chi Minh City (IUH) under grant number 60/HD-
Thailand, 2019, pp. 1-5, doi: 10.1109/ICCE-
DHCN.
Asia46551.2019.8941603.
[12] D.-T. Do, Optimal Throughput under Time
References Power Switching based Relaying Protocol in En-

[1] D. Bharadia, E. McMilin and S. Katti, Full Du-

ergy Harvesting Cooperative Network, Wireless
plex Radios, in Proc. ACM SIGCOMM Conf., Personal Communications (Springer), vol. 87,
no. 2, pp. 551-564, 2016.
pp. 375386, 2013.
[2] D. Nguyen, L.-N. Tran, P. Pirinen and M. Latva- [13] D.-T. Do, Power Switching Protocol for Two-

aho, On the Spectral Eciency of Full-Duplex way Relaying Network under Hardware Impair-

Small Cell Wireless Systems, IEEE Trans. Wire- ments, Radioengineering, vol. 24, no. 3, pp. 765-
less Commun., vol. 13, no. 9, pp. 4896-4910, 2014. 771, 2015.

[3] M. Mohammadi, B. K. Chalise, H. A. Suraweera, [14] D.-T. Do, H.-S. Nguyen, M. Voznak and T.-S.

C. Zhong, G. Zheng and I. Krikidis, Through- Nguyen, Wireless powered relaying networks un-

put Analysis and Optimization of WirelessPow- der imperfect channel state information: system

ered Multiple Antenna Full-Duplex Relay Sys- performance and optimal policy for instantaneous

tems, IEEE Trans. Commun., vol. 64, no. 4, rate, Radioengineering, vol. 26, no. 3, pp. 869-

pp. 1769-1785, 2016. 877, 2017.

[4] T.-H. Chang, Y.-F. Liu and S.-C. Lin, Max- [15] H.-S. Nguyen, D.-T. Do, T.-S. Nguyen and M.

Min-Fairness Linear Transceiver Design for Full- Voznak, Exploiting hybrid time switching-based

Duplex Multiuser Systems, in Proc. IEEE In- and power splitting-based relaying protocol in

tern. Workshop on Signal Process. Advances in wireless powered communication networks with

Wireless Commun., pp. 1-5, 2017. outdated channel state information, Automatika,
[5] X.-X. Nguyen and Dinh-Thuan Do, Optimal vol. 58, no. 4, pp. 111-118, 2017.

power allocation and throughput performance of [16] D.-T. Do, Energy-aware two-way relaying

full-duplex DF relaying networks with wireless networks under imperfect hardware: optimal

power transfer-aware channel, EURASIP Jour- throughput design and analysis, Telecommuni-
nal on Wireless Communications and Networking, cation Systems, vol. 62, no. 2, pp. 449-459, 2016.
vol. 2017, no. 1, pp. 152, 2017. [17] H. Hu, Y.-C. Liang, H. Zhang and B.-H. Soong.
[6] X.-X. Nguyen and Dinh-Thuan Do, Maximum Cognitive radio with self-power recycling, IEEE
Harvested Energy Policy in Full-Duplex Relaying Trans. Veh. Technol., vol. 66, no. 7, pp. 6201-
Networks with SWIPT,International Journal of 6214, 2017.
Communication Systems (Wiley), vol 30, no. 17, [18] A. Celik, A. Alsharoa and A. E. Kamal, Hy-
2017. brid energy harvesting-based cooperative spec-
[7] Z. Ding et al., Application of non-orthogonal trum sensing and access in heterogeneous cogni-
multiple access in LTE and 5G networks, IEEE tive radio networks, IEEE Trans. Cogn. Com-
Commun. Mag., vol. 55, no. 2, pp. 185-191, 2017. mun. Netw., vol. 3, no. 1, pp. 37-48, 2017.
[8] Z. Ding, X. Lei, G. K. Karagiannidis, R. Schober, [19] M. Pratibha, K. H. Li and K. C. Teh, Channel
J. Yuan and V. K. Bhargava. A survey on non- selection in multichannel cognitive radio systems
orthogonal multiple access for 5G networks: Re- employing RF energy harvesting, IEEE Trans.
search challenges and future trends, IEEE J. Sel. Veh. Technol., vol. 65, no. 1, pp. 457-462, 2016.
Areas Commun., vol. 35, no. 10, pp. 2181-2195, [20] J. Ren, J. Hu, D. Zhang, H. Guo, Y. Zhang and
2017. X. Shen, RF energy harvesting and transfer in
[9] F. Fang, H. Zhang, J. Cheng, S. Roy and V. C. cognitive radio sensor networks: Opportunities
M. Leung, Joint user scheduling and power al- and challenges, IEEE Commun. Mag., vol. 56,
location optimization for energy-ecient NOMA no. 1, pp. 104-110, 2018.
Performance Analysis of Downlink NOMA System Relying on Energy Harvesting and Full-Duplex 9

[21] J.-M. Kang, I.-M. Kim and D. I. Kim, Wire- Performance Analysis, KSII Transaction Inter-
less information and power transfer: Rate- net Inf. Syst., vol. 11, no. 4, pp. 1889-1910, 2017.
energy tradeo for nonlinear energy harvesting, [32] K. Woradit, T. Q. S. Quek and Z. Z. Lei, Co-
IEEE Trans. Wireless Commun., vol. 17, no. 3, operative multicell ARQ in MIMO cellular sys-
pp. 1966-1981, 2018. tems, 2010 IEEE 11th International Workshop
[22] S. Wang, M. Xia, K. Huang and Y.-C. Wu, on Signal Processing Advances in Wireless Com-
Wirelessly powered two-way communication with munications (SPAWC), Marrakech, 2010, pp. 1-5,
nonlinear energy harvesting model: Rate regions doi: 10.1109/SPAWC.2010.5670997.
under xed and mobile relay, IEEE Trans. Wire- [33] K. Woradit, T. Q. S. Quek and Z. Lei,
less Commun., vol. 16, no. 12, pp. 8190-8204, Cooperative Multicell ARQ - Packet Er-
2017. ror Rate and Throughput Analysis, 2010
[23] R. Morsi, D. S. Michalopoulos and R. Schober, IEEE Wireless Communication and Network-
Performance analysis of near-optimal energy ing Conference, Sydney, NSW, 2010, pp. 1-6,
buer aided wireless powered communication, doi: 10.1109/WCNC.2010.5506698.
IEEE Trans. Wireless Commun., vol. 17, no. 2,
pp. 863-881, 2018.
[24] Y. Sun, D. W. K. Ng, Z. Ding and R. Schober,
Optimal Joint Power and Subcarrier Alloca-
Chi-Bao Le was born in Binh Thuan
tion for Full-Duplex Multicarrier Non-Orthogonal province, Vietnam. He has worked
Multiple Access Systems, IEEE Trans. Com- closely with Dr. Thuan at Wireless

mun., vol. 65, no. 3, pp. 1077-1091, 2017. Communications and Signal Processing
Research Group at Industrial Univer-
[25] C. Zhong and Z. Zhang, Non-Orthogonal Mul- sity of Ho Chi Minh City, Vietnam.
tiple Access with Cooperative Full-Duplex Re- He is currently pursued Master de-

laying, IEEE Commun. Lett., vol. 20, no. 12,

gree in the eld of wireless communi-
cations. His research interest includes
pp. 2478-2481, Dec 2016. electronic design, signal processing in

[26] L. Zhang, J. Liu, M. Xiao, G. Wu, Y. C. Liang wireless communications network, non-
orthogonal multiple access, physical layer security and recon-
and S. Li, Performance Analysis and Optimiza- gurable intelligent surfaces.
tion in Downlink NOMA Systems with Coopera-
tive Full-Duplex Relaying, IEEE J. Select. Areas
Commun., vol. 35, no. 10, pp. 2398-2412, 2017.
[27] I. S. Gradshteyn and I. M. Ryzhik. Table of In-
Dinh-Thuan Do received the B.S. de-
tegrals, Series and Products, 6th ed. New York, gree, M.Eng. degree, and Ph.D. de-
NY, USA: Academic Press, 2000. gree from Vietnam National University

[28] M. Abramowitz and I. A. Stegun, Handbook of (VNU-HCMC) in 2003, 2007, and 2013
respectively, all in Communications En-
Mathematical Functions with Formulas, Graphs, gineering. He was a visiting Ph.D. stu-
and Mathematical Tables. New York, NY, USA: dent with Communications Engineering
Institute, NationalTsing Hua University,
Dover, 1972.
Taiwan from 2009 to 2010. Prior to join-
[29] W. Huang, H. Chen, Y. Li, and B. Vucetic, ing Ton Duc Thang University, he was
On the performance of multi-antenna wireless- senior engineer at the VinaPhone Mobile
Network from 2003 to 2009. Dr. Thuan was recipient of golden
powered communications with energy beamform-
Globe Award from Vietnam Ministry of Science and Technol-
ing, IEEE Trans. on Vehicular Technology, ogy in 2015 (Top 10 most excellent scientist nationwide). His
vol. 65, pp. 1801-1808, 2016. name and his achievements will be reported in a special book
entitled Young talents in Vietnam 2015-2020. His research
[30] A. A. Nasir, X. Zhou, S. Durrani, and R.
interest includes signal processing in wireless communications
A. Kennedy, Relaying protocols for wireless networks, NOMA, full-duplex transmission, and energy har-
energy harvesting and information processing, vesting. His publications include over 80 SCIE/SCI-indexed

IEEE Trans. Wireless Commun., vol. 12, no. 7,

journal papers, over 45 SCOPUS-indexed journal papers and
over 50 international conference papers. He is sole author in 1
pp. 3622-3636, 2013. textbook and 1 book chapter. Heis currently serving as Editor

[31] Nguyen, X.X.; Do, D.T, An Adaptive-Harvest- of Computer Communications (Elsevier), Associate Editor of
EURASIP Journal on Wireless Communications and Network-
Then-Transmit Protocol for Wireless Powered ing (Springer), and Editor of KSII Transactions on Internet
Communications: Multiple Antennas System and and Information Systems.