Journal of Engineering Science and Technology

Vol. 14, No. 3 (2019) 1173 - 1186

© School of Engineering, Taylor’s University

RELAY SELECTION ON DEVICE-TO-DEVICE-ASSISTED

FULL-DUPLEX NON-ORTHOGONAL MULTIPLE ACCESS:
SYSTEM MODEL AND PERFORMANCE ANALYSIS

HUU-PHUC DANG1,2, MINH-SANG V. NGUYEN3,

DINH-THUAN DO3,*, HONG-LIEN PHAM1
1Faculty of Electrical and Electronics Technology, Ho Chi Minh City University of Technology
and Education, Vietnam. 1 Vo Van Ngan, Thu Duc Dist., Ho Chi Minh City, Vietnam
2Tra Vinh University, 126 Nguyen Thien Thanh, Ward 5,

Tra Vinh City, Tra Vinh Province, Vietnam

3 Faculty of Electronics Technology, Industrial University of Ho Chi Minh City (IUH),

12 Nguyen Van Bao, Go Vap Dist., Ho Chi Minh City, Vietnam

*Corresponding Author: dodinhthuan@iuh.edu.vn

Abstract
In recent year, bandwidth efficiency and performance of Non-Orthogonal
Multiple Access (NOMA) system can be enhanced by using a full-duplex
transmission model. However, previous papers have not investigated
satisfactorily in the full-duplex scenario in which, near device (i.e., so-called as
relay) is selected for Device-To-Device (D2D) transmission employing NOMA
scheme. The outage performance of Full-Duplex D2D NOMA and relay selection
structure for D2D is studied in this paper. By considering the outage and
throughput performance of a considered system with its vital result, general
wireless networks with various selected devices to help D2D communication in
the cellular network. Simulation results show that the device selection scheme
can achieve significant performance improvement.
Keywords: D2D, NOMA, Outage probability, Relay selection.

1173
1174 H.-P. Dang et al.

1. Introduction
In recent years, NOMA technology has attracted a lot of attention. It is considered
as one of the modern technologies applied in 5th Generation (5G) mobile networks
in order to considerably enhance the system spectral efficiency for the next
generation communication networks [1-4]. The information of multiple users is
incorporated into a mixed signal in the power domain, whereas the users are served
in the same domain such as time, frequency or code domain. NOMA procedure
mainly allocates greater transmit power dedicated for users suffering from weak
channel conditions. In this scenario, by treating others signal as noise, these weak
users can detect and decode its symbols precisely due to greater power level
distributed for the far user. On the other hand, based on the successive interference
cancellation (SIC) technique, users, which have strong channel conditions, are
capable to detect its own signal. In addition, as compared to the Orthogonal
Multiple Access (OMA) system, there is a significant improvement in NOMA
throughput performance [5].
Chen et al. [6] and Wang et al. [7], focused on the multiple user channel capacity
and the spectrum efficiency as well as the number of subscribers under bad channel
condition can be enhanced by employing NOMA scheme into the system model.
Simultaneously, Gupta and Singer [8] utilized SIC, the received message, in which,
is decoded at the receiver while the transmitter treating interference signal applies
the NOMA technique. NOMA is known as combination scheme as relaying scheme
[9, 10] can be integrated into such a system. Moreover, based on studies by Luan
and Do [10], the scenario of impacts on co-channel interference is applied in order
to increase the wirelesses power transfer performance and the mathematical
formula in the closed form of probability has been found. Nguyen et al. [11]
explained that full-duplex are evaluated as architecture to enhance bandwidth usage
efficiency. As a result, such full-duplex is conveyed in several NOMA schemes.
For the purpose of improving the outage performance of all devices in the relay
selection (RS) system, Ding et al. [12] have proposed a two-stage RS in cooperative
NOMA model. Furthermore, similar to studies by Ding et al. [12], in order to
improve system performance, a full-duplex NOMA (FD-NOMA) model has been
proposed by Zhang et al. [13]. According to Men and Ge [14], Zhong and Zhang
[15], NOMA design with a selected cooperative relay node as well as outage
probability is examined. Asadi et al. [16] proposed that the Device-To-Device
(D2D) communication has shown the possibility of being widely applicable, which
based on spectrum sharing with cellular networks. In principle, the equipment in
such D2D design can connect directly with another equipment through the control
of the base station (BS) and it leads to the result that the end-to-end latency can be
descended. It can be proved that the area spectral effectiveness can be enhanced in
the D2D scenario and the cellular network is able to accommodate more devices
[17]. Motivated by advantages of relay selection [18, 19] and D2D model, this
paper aims to combine cooperative NOMA with D2D scheme in full-duplex
scenario into considered D2D NOMA system. In particular, outage and throughput
performance in such D2D NOMA are investigated. In addition, the D2D
transmission with the utilization of full-duplex scenario and relay selection to
perform NOMA scheme has been studied by evaluating related parameters.
As the main contribution of this paper, we derive the closed-form expression in
term of outage event to evaluate system performance of transmission from the first

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1175

device in each pair user (so-called near NOMA users), which is treated as relay
node to assist forwarding signal to the far NOMA devices.

2. System Model
We consider a downlink system model is shown as Fig. 1, where a base station (BS)
would like to send a message to the N near users (NU), which will act as the relay
to broadcast BS's data signal to a pair of far users (FU) that set as device 1 ( U1 )
and device 2 ( U 2 ).
This paper develops a system model presented in recent work [13].
Transmission procedure happens in two phases, i.e., at a time t , t  1 . More
specifically, Fig. 1 illustrates the full-duplex D2D devices applying NOMA, which
can serve D2D transmission to far devices, for example transferring signalling
signal as D2D is underlying with a traditional cellular network. In this model, we
denote, Dn , n 1, 2,..., N as D2D devices so-called as NUs. The BS is a single
antenna transmission source while the NUs are designed with two separated
antennas to serve full-duplex transmission. As a result, self-interference (SI) due to
simultaneous operation of two antennas at near devices. We assume that there does
not exist a direct link from the BS to FU, hn and g n ,i ( i 1,2 ) are denoted the
Rayleigh fading channel coefficients of the link BS-NU and NU-FU, respectively.
2 2
The random variables hn and g n,i follow the exponential distribution with
parameters Φhn and Φgn ,i , respectively. Following the principle of normal NOMA,
users are distributed in order based on the channel conditions. For example, we
assumed that U1 and U 2 are used for different data transmission, in which, U1 is
used for low-speed applications, while U 2 serves high-speed data rate.

(P1 x1  P2 x2 )
Dn U1 Decoding x1
BS Dn
(P1 x1  P2 x2 )
Dn U2 SIC-Decoding x 2
(P1 x1  P2 x2 )

t t +1

hn gn,1
USER 1
Relay D1

hSI
Source
BS
gn,2

Relay Dn USER 2

Communication Link Self Interference Link

Fig. 1. System model of relay selection

D2D NOMA with full-duplex transmission.

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1176 H.-P. Dang et al.

The transmission between the BS  NU and Di -th  Ui -th, (i 1, 2) are

divided into two phases. In the early phase, the BS sends the superposition signal,
P1 x1  P2 x2 , where x1 and x2 denotes the message for BS sent forward to first
devices and then these signal intend to transmit to U1 and U 2 in D2D link using
the N device considered as cooperative relays, respectively.

In this full-duplex (FD) scheme, xSI denotes self-interference signal. In such

case, the residual loop self-interference (SI) is set as a Rayleigh fading feedback
2
channel with the coefficient hSI , and random variables hSI also follow the
exponential distribution with mean value ΦSI , Pi , (i 1, 2) is the power allocation
coefficients, Ps and PR (assuming that Ps  PR ) are the transmission power at the
BS and NU user, respectively. Because it must be satisfied Quality-of-Service
(QoS) requirements, we denote P1  P2  1 with assumption as P1  P2 .
2 2

Therefore, the received signal at the NU user, i.e., relay n-th, 1  n  N can be
given as

ynr  hn PS ( P1 x1  P2 x2 )  hSI PR xSI   nr (1)

where  n stands for the additive Gaussian noise (AWGN) at NU-th user with zero
r

mean and variance N 0 .

To successfully decode the signal, which is containing x1 and x2 , the

conditions for relay n-th are written as:
 
 
2 2
hn P12 h P2
log 2  1    R1 , log 2 ( 1  n 2 )  R2 (2)
1 1
 hn P22  hSI   hSI 
2 2 2

  

where  and Ri , i  (1,2) are the signal-to-noise ratio (SNR) and the target data
Ps P
rate for user Ui-th, respectively,    R .
N0 N0
During the following time slot, consider that relay n-th can successfully decode
the signal and is selected to forward the message P1 x1  P2 x2 . In this case, the
received signal at far user Ui-th are given as the following equation:
d
yn,i  gn,i PR ( P1 x1  P2 x2 )   n,id ,i 1,2 (3)

where  nd,i is the AWGN at FU-th user with zero mean and variance N 0 . It is
similar to the first time slot, the condition for User U1 decodes its message, i.e., x1

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1177

2
g n ,1 P12
with the SINR1  and the other user decodes its own message with
2 1
g n ,1 P22 

 2 
 g n ,2 P12 
SINR2   P gn ,2
2
, provided that log 2  1   R1
2
2
1
 gn ,2 P22  
2

 
Device Selection Scheme
To enhance system performance in such a network, full duplex NOMA cooperates
with the device selection model. In this section, we propose a device selection
scheme as follow.
For a chosen relay n-th, only if the instantaneous rates for U1 and U 2 are
greater or equal to their target data rates R1 and R2 , respectively, the near user can
successfully decode the BS’s messages x1 and x2 , i.e., this condition must be
satisfied.

 
 P12 hn
2
  hn P22 
2

log 2  1    R1  log 2  1   R2
1
(4)
 1 /   P2 h 2  h 2

   hSI  
2
2 n SI
 
This equivalent with the following equation:

 ( 2 R1  1 )( 1   h 2 ) ( 2 R2  1 )( 1   h 2 ) 
SI SI
max  , 
hn
2

 1
 P 2
 P 2
2
 2 R1
 1  P 2
2

(5)

2 2R1  1
It worth noting that if g n,1  then U 1 needs to be treated first, i.e., the

target rate conditions of U1 need to be satisfied.

The set of selected devices, which can meet the Quality of Service (QoS)
requirements of, U1 can be expressed as follows

 2  2  
DSn   hn  , gn ,1  1  (6)
  


in which,   1   hSI 2
 max  P 1
2
1  
, 2  ,1  2 R1  1,2  2 R2  1 . In the
 P221 P22 
other hand, this condition can be happened as the following probability.
  1  m 
  1 Φ
 2  2  Φ e hn
Pr  hn  , g n ,1  1   e hn  (7)
   ΦSI m  1

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1178 H.-P. Dang et al.

Noted that it can be chosen Φhn  Φgn ,i  1

   
where m  max  2 1 2 , 22  .
 P1  P2 1 P2 
In addition, the further criterion for this concerned system for device selection
adapting U 2 is:

n*  arg max gn ,2
n 1,2 ,...,N
 2
 (8)

3. System Performance Analysis

In this section, in order to determine whether or not the outage ability occur on the

whole system, we denote Ψ 1 is the outage ability that relay n is unable to decode
x1 , or both of the FUs and selected NU cannot decode x1 successfully, Ψ 2 denotes

the outage ability that x2 cannot be decoded successfully by relay n and by U 2 .
Hence, the outage ability of the entire system can be expressed as follows:
Ψ  Ψ1 Ψ 2 (9)

Consequently, the overall outage probability can be given as:

Pr(Ψ )  Pr(Ψ1 )  Pr(Ψ 2 ) (10)

in which, Pr(Ψ1 ) can be calculated as follows:

 
 
  
N
Pr Ψ 1    1   Pr hn  hSI   1 
2 2

n 1
 
  A1


 2
 
2 
 Pr g n ,1  1  Pr g n ,2  1 

 (11)
A2 A3 

 2 R1  1 
1  2 R2  1  . It is worth
where 1  ,   max  , 
  P12  1 P2 2  

 P2
2
 P 1
2
 
2 R2
 1 P 1
2
 

Φhn   / Φhn
noting that using exponent distribution we obtain A1  e , and
Φhn   ΦSI
1 1
 
Φgn ,1 Φgn ,2
A2  e , A3  e .

The second term Pr(Ψ 2 ) can be derived as follows:

Pr Ψ 2   Pr Q1 , DSn  n   Pr Q2 ,Q1 , DSn  n  (12)

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1179


where Q1 is the case that relay n is unable to decode x2 , Q1 stands for the
complementary circumstance of Q1 , and Q2 denotes when x2 cannot be decoded
at U 2 . The previous equation can be denoted as follows:

Pr Ψ 2   Pr Q1 , DSn  n   Pr Q2 ,Q1 , DSn  n  (13)

A B

Next, we consider the first term of Eq. (13) as below:

A  Pr Q1 , DSn  n 
   
  h P2 
2

 Pr  log 2  1  n* 2   R2 , DSn  n 
 1 
 hSI  
2

   
    h  2  
2

 1  Pr  hn*  2 SI 2    Pr  DSn  n  (14)

2

   P2 
 

It is worth noting that the probability for the n relay devices, which it is
selected in DSn , can be expressed as follows [18]
n N n
  1  m      1  m 

 e    e  
Pr( DSn  n )   
N
n  Φ m 1 1  Φ m  1  (15)
 SI   SI 
   
From Eq. (14) and Eq. (15), the expected result for Eq. (14) can be rewritten as
follow:
 N
N Φhn  P22  n2  
A   1      1
n 1
exp   2 

 n 1  n  ΦSI n2  Φhn  P2  
 Φhn  P2  
2

N n
  1 m     1  m 
n
 
N  e   e  
 n   1 
 Φ m 1
(16)
Φ
 SI m  1   SI 
   
Similarly, it can be expressed in the second term of Eq. (13) as below
B  Pr Q2 ,Q1 , DSn  n 
  
  h P2  
2
2
 P r log 2 ( 1   g n*,2 P22 )  R2 ,log 2  1  n* 2   R2 , DSn  n 
1
 hSI   
2

  
 N
N  n2  
  1      1 exp  
n 1

 n 1  n 
 Φg  P22  
  n ,2 
 N N Φhn  P22  n2  
      1
n 1
exp   2 

 n 1  n  Φ n    Φ  P 2  Φ 
hn  P2  
 SI 2 hn 2 
 1m  
n
  1  m 

N n
(17)
  nN  
e  1  e  
  Φ m 1
 ΦSI m  1   SI 
   

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1180 H.-P. Dang et al.

After some manipulations, we obtain:

Pr(Ψ 2 )  A  B
 N
N Φhn  P22  n2  
  1      1
n 1
exp   2 

 n 1  n  ΦSI n2  Φhn  P2  
 Φhn  P2  
2

N n
  1 m     1  m 
n
 
e 
 N
n   Φe m  1  
1 
 Φ m 1

 SI   SI 
   
 N
N  n2  
  1      1 exp  
n 1

 n 1  n   Φg  P22  
  n ,2 
 N N Φhn  P22  n2  
      1
n 1
exp   2 

 n 1  n  Φ n   Φ  P 2  Φ 
hn  P2  
 SI 2 hn 2 
N n
  1 m     1  m 
n
 
N  e   e  
 n 
(18)
 1  Φ m  1 
 ΦSI m  1   SI 
   

From Eqs. (11) to (18), the required probability in Eq. (10) can be calculated in
the closed-form as:
 
N  Φhn  1  1 
Pr Ψ    1 
 / Φ Φ Φ
e hn e gn ,1 e gn ,2 
n 1  Φhn   ΦSI 

 N
N Φhn  P22  n2  
  1      1
n 1
exp   
 n 1  n  ΦSI n2  Φhn  P22  Φh  P22  
  n 
N n
  1 m     1  m 
n
 
e 
 N
n   Φe m  1  
1 
 Φ m 1

 SI   SI 
   
 N
N  n2  
  1      1 exp  
n 1

 n 1  n   Φg  P22  
  n ,2 
 N N Φhn  P22  n2  
     1
n 1
exp   
 n 1  n  ΦSI n2  Φhn  P22  Φh  P22  
  n 
N n
  1 m     1  m 
n
 
e  (19)
 N
n   Φe m  1  
1 
 Φ m 1

 SI   SI 
   

Remark 1: Here, it is assumed that channel gain of each link in the first hop and
self-interference channel are similar value and equals to Φhn , ΦSI , respectively.
This is an important contribution of this paper to evaluate device selection scheme
in D2D NOMA.
In this paper, it can calculate the throughput from the obtainable sum rate of the
system. Particularly, the BS sends the message at a fixed target rate R0 , which is
subject to the impact of outage probability because of random wireless fading
channels. The overall FD NOMA sum throughput can be given as
Γ 1 Pr(Ψ ) R0 (20)

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1181

4. Simulation Result
In this section, we present numerical results to evaluate analytical expressio ns
calculated in the previous part. Without loss of generality, it can be pretended
that the distance from BS to near users-th, which is standardised to the unit,
and average channel power gains are assumed that ΦSI Φhn Φgn ,1 Φgn ,2 1
2
. The power allocation coefficients of NOMA are P1 0.75 and P2 2 0.25 for
U 1 and U 2 , respectively. Note that the simulation results are done with the help of
Matlab.
As the observation, Figs. 2 and 3 plot the outage probability for proposed
NOMA with two scenarios described as in figures, where the relay selection
scheme is applied. Observing Fig. 2, one can conclude that more relay brings
better outage performance. Besides, in studies by Ding et al. [19], we also
compared our system's outage performance with the max-min relay selection
criterion, in which, can be seen in Fig. 2 that at SNR from 0 dB to 20 dB, our
system's performance is better than max-min relay selection criterion.
However, at the high SNR, i.e., SNR from 25 dB to 40 dB, the performance
is better than our system model.
This could be explained because of our system is FD, so when the SNR is
high, the self-interference is also stronger and it crucial affects the system's
performance. In addition, both Figs. 2 and 3 manifest that D2D NOMA can
remarkably increase the outage performance as if a reasonable selection of
number of device is given.
Moreover, when the number of the NUs, i.e., N is increased, the performance
gap regarding the outage probabilities achieved by D2D NOMA can be seen clearly
at high transmit SNR.

Fig. 2. Outage probability with different number of NU nodes and

comparison with the max-min relay selection criterion [19].
Key parameters R1 0.5 , R2 2 .

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1182 H.-P. Dang et al.

Fig. 3. Outage probability with number of NUs,

N  2 and different target rates are R1  0.1;0.3;0.5 and R2  2 .

In Fig. 4, it can be observed that the outage probability varies according to the
different values of SI. In this figure, we set different self-interference ratios are
ΦSI    0.6;0.8;1 . The exact outage probability curves of proposed
E hSI
2

D2D NOMA with a higher level of SI will result in worse outage performance. It
is observed that the superiority of full duplex function in D2D NOMA is no longer
   1 dB ). Therefore, it is
appear with the very large values of SI (i.e., E hSI
2

essential to consider the influence of SI when designing practical full duplex

antenna in such NOMA systems. Besides, the perfect combination of analytical
lines and simulation lines has confirmed the accuracy of our analytical expressions.

Fig. 4. Outage probability with number of NUs,

N  3 , 1 0.5 , R2  2 , Φhn Φgn ,1 Φgn ,2 1 and ΦSI 0.6; 0.8;1 .
R 

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1183

In Fig. 5, we compare the throughput of four different cases considering the

number of relay. Our setting are R1  1 (b / s / Hz) , R2  2 (b / s / Hz) ,
R0  2(b / s / Hz) and P22 0.25 . It can be observed from Fig. 5 that for the
proposed scheme, the analytical throughput remains at the maximal level at high
transmit SNR, i.e. transmit SNR is greater than approximately 20 (dB). In addition,
Fig. 5 also shows that the proposed scheme can obtain clear throughput
performance gap as SNR between 5dB and 20 dB. For the case that the value of
SNR is low, the throughput in this NOMA scheme becomes very small and the
proposed scheme almost works in low quality.

Fig. 5. System throughput with different

numbers of relay. R0  2 , R1  1 , R2  2 .

5. Conclusion
In this study, we suggested a device selection scheme integrated with NOMA
scheme to evaluate system outage performance for a D2D transmission with
multiple devices. The considered NOMA scheme is assessed and compared with
the different scenarios in terms of outage behaviour and throughput. By achieving
outage with appropriate selection of the number of device, it can be observed from
the simulation result that the proposed NOMA can be applied in real NOMA
design, then such NOMA scheme can achieve outage and throughput criteria for
each user in pair of NOMA users. For further research topics, with more than two
users we may consider a generalization of relay selection policies and performance
investigation for randomly distributed NOMA users.

Nomenclatures

D1...Dn NUs user

DSn Set of selected devices for U1
E 
. Expectation operation, dB
Φgn ,1 2
Mean of g n,1 , dB

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1184 H.-P. Dang et al.

Φgn ,2 Mean of g n ,2 , dB
2

Φhn 2
Mean of hn , dB
ΦSI Mean of hSI , dB
2

Γ System throughput, b/s/Hz

g n ,1 Rayleigh fading channel coefficient of the link relay- U1
g n ,2 Rayleigh fading channel coefficient of the link relay- U 2

g n,1
2 Channel power gain of the link relay- U1 , dB

g n,2
2 Channel power gain of the link relay- U 2 , dB

hn Rayleigh fading channel coefficient of the link BS-relay

hSI Rayleigh fading channel self-interference coefficient

hn
2 Channel power gain of the link BS-relay, dB

hSI
2 Channel power gain of the self-interference, dB
m Temporary coefficient in calculation (7)
N0 Noise power of AWGN  nr , dB
n; N Number of relay
n* Selected relay in the set of relays
P1 Power allocation coefficient of U1 , dB
P2 Power allocation coefficient of U 2 , dB
Pr(.) Probability operator
PS Power of BS, dB
PR Power of Relay, dB

Q1 Outage ability that relay n is unable to decode x2
Q1 Complementary circumstance of Q1
Q2 Outage ability when x2 cannot be decoded at U 2
R0 Fixed target rate of BS, b/s/Hz
R1 Target data rate for U1 , b/s/Hz
R2 Target data rate for U 2 , b/s/Hz
SINR1 Condition of signal-to-noise ratio for user U1
decodes its message
SINR2 Condition of signal-to-noise ratio for user U 2
decodes its message
t Time, s
U1 User device 1
U2 User device 2
x1 Source message of U1

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

 Relay Selection on Device-to-Device-Assisted Full-Duplex . . . . 1185

x2 Source message of U 2
xSI Self-interference signal of relay n-th

ynr Received signal at the relay n-th

Greek Symbols
 Temporary coefficient in calculation (11)
 r
n
Additive white Gaussian noise, dB
 Signal-to-noise ratio ( Ps / N0  PR / N0 ), dB
1 SNR threshold of U1 ( 2R1  1 ), dB
2 SNR threshold of U 2 ( 2R2  1 ), dB
 Temporary coefficient in calculation (6)
Ψ Outage ability of the entire system
Ψ1 Outage ability to decode x1
Ψ2 Outage ability to decode x2
1 Temporary coefficient in calculation (11)

Abbreviations
AWGN Additive White Gaussian Noise
BS Base Station
D2D Device-to-Device
FD Full-Duplex
FU Far Users
NOMA Non-Orthogonal Multiple Access
NU Near Users
OMA Orthogonal Multiple Access
QoS Quality-of-Service
RS Relay Selection
SI Self-Interference
SIC Successive Interference Cancellation
SNR Signal-to-Noise

References
1. Yu, Y.; Chen, H.; Li, Y.; Ding, Z.; and Vucetic, B. (2017). Antenna selection
for MIMO-NOMA networks. Proceedings of the IEEE International
Conference on Communications (ICC). Paris, France, 1-6.
2. Saito, Y.; Kishiyama, Y.; Benjebbour, A.; Nakamura, T.; Li, A.; and Higuchi,
K. (2013). Non-orthogonal multiple access (NOMA) for cellular future radio
access. Proceedings of the IEEE 77th Vehicular Technology Conference (VTC
Spring). Dresden, Germany, 1-5.
3. Dai, L.; Wang, B.; Yuan, Y.; Han, S.; I., C.-L.; Lin, C.; and Wang, Z. (2015).
Non-orthogonal multiple access for 5G: solutions, challenges, opportunities,
and future research trends. IEEE Communications Magazine, 53(9), 74-81.

Journal of Engineering Science and Technology June 2019, Vol. 14(3)

1186 H.-P. Dang et al.

4. Wei, Z.; Yuan, J.; Ng, D.W.K.; Elkashlan, M.; and Ding, Z. (2016). A survey
of downlink non-orthogonal multiple access for 5G wireless communication
networks. Retrieved May 5, 2018, from https://arxiv.org/abs/1609.01856.
5. Ding, Z.; Yang, Z.; Fan, P.; and Poor, H.V. (2014). On the performance of non-
orthogonal multiple access in 5G systems with randomly deployed users. IEEE
Signal Processing Letters, 21(12), 1501-1505.
6. Chen, Z.; Ding, Z.; Dai, X.; and Karagiannidis, G.K. (2016). On the application
of quasi-degradation to MISO-NOMA downlink. IEEE Transactions Signal
Processing, 64(23), 6174-6189.
7. Wang, X.; Labeau, F.; and Mei, L. (2017). Closed-form BER expressions of
QPSK constellation for Uplink Non-Orthogonal Multiple Access (NOMA).
IEEE Communications Letter, 21(10), 2242-2245.
8. Gupta, A.S.; and Singer, A. (2007). Successive interference cancellation
using constellation structure. IEEE Transactions Signal Processing, 55(12),
5716-5730.
9. Nguyen, H.-S.; Bui, A.-H.; Do, D.-T.; and Voznak, M. (2016). Imperfect
channel state information of AF and DF energy harvesting cooperative
networks. China Communications, 13(10), 11-19.
10. Luan, N.T.; and Do, D.-T. (2017). A new look at AF two-way relaying
networks: energy harvesting architecture and impact of co-channel
interference. Annals of Telecommunications, 72(11-12), 669-678.
11. Nguyen, T.N.; Do, D.-T.; Tran, P.-T.; and Voznak, M. (2016). Time switching
for wireless communications with full-duplex relaying in imperfect CSI
condition. KSII Transactions on Internet and Information Systems, 10(9),
4223-4239.
12. Ding, Z.; Dai, H.; and Poor, H.V. (2016). Relay selection for cooperative
NOMA. IEEE Wireless Communications Letters, 5(4), 416-419.
13. Zhang, Z.; Ma, Z.; Xiao, M.; Ding, Z.; and Fan, P. (2017). Full-duplex device-
to-device aided cooperative non-orthogonal multiple access. IEEE
Transactions on Vehicular Technology, 66(5), 4467-4471.
14. Men, J.; and Ge, J. (2015). Non-orthogonal multiple access for multiple-
antenna relaying networks. IEEE Communications Letters, 19(10), 1686-1689.
15. Zhong, C.; and Zhang, Z. (2016). Non-orthogonal multiple access with
cooperative full-duplex relaying. IEEE Communications Letters, 20(12),
2478-2481.
16. Asadi, A.; Wang, Q.; and Mancuso, V. (2013). A survey on device-to-device
communication in cellular networks. IEEE Communication Survey and
Tutorials, 16(4), 1801-1819.
17. Kim, J.-B.; Lee, I.-H.; and Lee, J. (2017). Capacity scaling for D2D aided
cooperative relaying system using NOMA. IEEE Communications Letters,
7(1), 42-45.
18. Yang, Z.; Ding, Z.; Wu, Y.; and Fan, P. (2017). Novel relay selection strategies
for cooperative NOMA. IEEE Transactions on Vehicular Technology, 66(11),
10114-10123.
19. Ding, Z.; Dai, H.; and Poor, H.V. (2016). Relay selection for cooperative
NOMA. IEEE Wireless Communications Letters, 5(4), 416-419.

Journal of Engineering Science and Technology June 2019, Vol. 14(3)