Non-Orthogonal Multiple Access For

Cooperative Communications: Challenges,

Opportunities, And Trends
Dehuan Wan, Miaowen Wen, Fei Ji, Hua Yu, and Fangjiong Chen
arXiv:1801.04650v1 [cs.IT] 15 Jan 2018

Abstract

Non-orthogonal multiple access (NOMA) is a promising radio access technique for next-generation
wireless networks. In this article, we investigate the NOMA-based cooperative relay network. We begin
with an introduction of the existing relay-assisted NOMA systems by classifying them into three
categories: uplink, downlink, and composite architectures. Then, we discuss their principles and key
features, and provide a comprehensive comparison from the perspective of spectral efficiency, energy
efficiency, and total transmit power. A novel strategy termed hybrid power allocation is further discussed
for the composite architecture, which can reduce the computational complexity and signaling overhead
at the expense of marginal sum rate degradation. Finally, major challenges, opportunities, and future
research trends for the design of NOMA-based cooperative relay systems with other techniques are also
highlighted to provide insights for researchers in this field.

Index Terms

Non-orthogonal multiple access (NOMA), cooperative relay network, power allocation, degree of
asymmetry.

This work was supported in part by the National Natural Science Foundation of China under Grant 61431005, Grant 61771202,
Grant 61671211, Grant U1701265, and Grant 61501190, and in part by the Guangdong provincial research project under Grant
2016A030308006.
All authors are with the School of Electronic and Information Engineering, South China University of Technology, Guangzhou
510641, China (e-mail: eemwwen@scut.edu.cn).
 2

I. I NTRODUCTION

The rapid development of wireless transmission and mobile communications techniques has
lead to an explosive increase in the data traffic of wireless networks. In order to support the
tremendous demands on data traffic, the non-orthogonal multiple access (NOMA) technique,
which can be realized in the power domain, code domain, or other domains [1], [2], is widely
recognized as a key technology for the fifth generation (5G) mobile communications systems
[3], [4].
Recently, significant research efforts have been dedicated to applying NOMA techniques to
various scenarios, motivated by the following benefits [3]:
• Higher spectrum efficiency and cell-edge throughput: NOMA can serve an arbitrary number
of users in each resource block by superimposing all users’ signals, and according to the
near-far effect, more power can be allocated to the nodes with poor channel quality to
improve the system throughput.
• Good backward compatibility with other techniques: NOMA can be easily applied on top
of orthogonal frequency division multiple access (OFDMA) for downlink and single-carrier
frequency-division multiple access (SC-FDMA) for uplink, and can be easily combined with
massive multiple-input multiple-output (MIMO) and millimeter-wave (mmWave) techniques
to further support better system performance.
Among those, the application of NOMA to cooperative relaying scenarios is becoming popular.
By now, a class of dual-hop relay-aided NOMA systems has been developed, where no direct link
exists and the relay adopts either decode and forward (DF) or amplify and forward (AF) protocol.
For example, the cooperative X network [5] and the diamond network [6] considering NOMA
and the DF protocol have been constructed, where the decoding of the signals at the relay and
user must follow the same manner, or the achievable sum rate will deteriorate dramatically. On
the other hand, the study with the AF protocol has been carried out in [7]. Besides, a downlink
communication system would become a cooperative relay network if the cell-centre users can
act as relay nodes for cell-edge users via successive interference cancellation (SIC) [8]. The
superiority of NOMA over OMA in terms of achievable sum rate is highly dependent on the
system’s asymmetry. Therefore, configuring or selecting the relays to fully exploit the near-far
effect is a goal to achieve a better system performance with NOMA. However, how to realize
effective configuration and optimization for such networks is still an open issue.
 3

In this article, we develop a unified architecture for the power-domain NOMA cooperative
relay network through a sophisticated combination of three basic communication structures.
The system settings, decoding orders, system asymmetries, power allocation, and performance
are discussed. Besides, a novel hybrid power allocation strategy is proposed for the composite
architecture. The major challenges, opportunities, and future research trends are also discussed.
The rest of this article is organized as follows. Section II first illustrates the basic communica-
tion structures and then presents the categories of cooperative relay networks accordingly. Section
III briefly reviews the features of NOMA in uplink and downlink transmissions. In Section IV, we
discuss the characteristics of system settings, decoding strategies, system asymmetries, and power
allocation schemes for the cooperative relay systems. Besides, performance comparisons among
the existing cooperative relay architectures are presented. Section V proposes an effective hybrid
power allocation strategy for NOMA cooperative relay systems. In Section VI, the opportunities,
challenges, and research trends are highlighted. Finally, Section VII concludes this article.

II. S TRUCTURES OF R ELAY-A IDED W IRELESS N ETWORKS

There are three basic modes for communications between source S and destination D, namely
the one-to-one, one-to-many, and many-to-one modes, as shown in Figs. 1(a)-(c). Actually, there
exists another many-to-many mode, as shown in Fig. 1(d). However, such a structure can be
constructed by the first three basic modes. For brevity, the many-to-many mode will not be
discussed in this article.
According to the basic communication modes, the cooperative relay networks, which comprise
the sources S, relays R, and users U, can be simply divided into three categories, namely the
uplink (many-to-one structure), the downlink (one-to-many structure), and the composite (both
uplink and downlink are comprised) architectures, as shown in Figs. 2(a)-(d). Since the one-to-
one mode can be subordinated to either the uplink or downlink transmission, the cooperative
relay system constructed by either the uplink or downlink architecture always includes a one-to-
one mode. For the case that the direct links between the source and user nodes exist, this article
only focuses on the classical three-node relay-aided wireless networks, as shown in Fig. 2(e).
On the other hand, due to the orthogonality, the receiver in OMA can only decode the data by
applying the single-user detector, such that the user node cannot act as a relay node to improve
the system performance. However, for NOMA, since SIC can realize multi-user detection, the
 4

cell-centre user can act as a relay to improve the reception reliability for other users with poor
connections, as shown in Fig. 2(f).

III. F EATURES OF NOMA IN U PLINK AND D OWNLINK

For ease of demonstration, the channel between the base station and i-th user Ui is denoted
by hi ∼ CN (0, σi2 ) and we measure the channel condition for Ui by the average power of its
channel, namely σi2 . Without loss of generality, it is assumed that σi2 is arranged in ascending
order, that is σ12 ≤ · · · ≤ σN
2
, and user m can be regarded as a stronger user than user n if
2
σm ≥ σn2 .

A. Processing Procedures

1) Uplink: For uplink transmission, the optimal multiple access strategy at the transmitter is
that N users spread their signals across the entire bandwidth with different transmit powers. At
the receiver, rather than decoding every user by treating the interference from other users as
noise, the SIC technique, in which after one user is decoded, its signal is stripped away from
the aggregate received signal before the next user is decoded, is applied to achieve a higher
achievable rate. Specifically, first, the base station decodes the signal from the stronger user Ui
by treating the signals of {Un } as noise, where n ∈ {1, . . . , i − 1}. Then, the signal of Ui is
subtracted from the received signal to decode the signals of {Un }. Finally, for U1 , it is just
corrupted by the noise as the other users’ signals have been successively decoded and cancelled
out.
2) Downlink: For downlink transmission, the SIC is applied at the receiver, the superposition
coding (SC) is used at the transmitter, and more power is allocated to a weaker user, namely
Pi ≥ Pi+1 . First, each user decodes the signals from other weaker users, i.e., Ui can decode
the signals of {Un } with n < i. Then, the signals of weaker users are subtracted from the
received signal to decode the signal of user Ui , by treating the signals of {Um } with m > i as
interference. Finally, for UN , it is just corrupted by the noise as the other users’ signals have been
successively decoded and cancelled out. Since a lower transmit power is assigned to a stronger
user, the signal strength of a stronger user is not higher than that of a weaker user. Therefore,
NOMA does not contradict the basic concept of SIC, in which decoding of the strongest signal
should be performed first.
 5

3) Differences: One of the key distinctions between the uplink and downlink transmissions
is the decoding order. To be more specific, for downlink, strong users successively decode and
cancel the signals of weak ones prior to decoding their own signals; whereas for uplink, the
receiver successively decodes and cancels the signals of strong users prior to decoding the signals
of weak ones [1].

B. Capacity and Asymmetry

Reports have said that NOMA generally outperforms OMA in the rate region with two users
case [9]. In this regard, the following properties can be observed.
1) Uplink: Compared with NOMA, OMA is in general worse for uplink transmission except
at one point, where it can achieve the same capacity bound as the former. However, at this
point, the rate of the weak user is much lower than that of the strong user especially when
the difference between the channel conditions of both users is large, resulting in poor fairness.
To characterize the effect of the channel difference for uplink transmission, we introduce the
degree of asymmetry, which is defined as the ratio between the strong and weak users’ channel
2
σh
variances, namely Au = 2
2
σh
.
1
2) Downlink: For downlink transmission, the boundary of the NOMA rate region strictly
contains the OMA rate region, leaving a gap that becomes larger as the asymmetry deepens. In
those cases with severe asymmetry, however, NOMA can still provide reasonable rates for both
strong and weak users. Similarly, we define the degree of asymmetry for downlink transmission
2
σh
as Ad = 2
2
σh
.
1

IV. R ELAY-A IDED NOMA N ETWORKS

For ease of exposition, Rayleigh fading channels are considered and N = 2 in Fig. 2 is
assumed in this section.

A. System Settings

1) Relay with DF protocol: For networks with DF protocol, where the SIC is utilized at
relay and user ends, the decoding order for source to relay (S→R) and relay to user (R→U)
transmissions must be organized in the same manner. Otherwise, the achievable sum rate will
decrease dramatically because of the min function, as shown in Fig. 3(a). Note that, for the
composite architectures, such as Figs. 2(c) and (d), the decoding orders for S →R (uplink or
 6

downlink) and R→U (downlink or uplink) transmissions are different. Consequently, the channel
gains for the S→R link are configured as ascending (descending) while the one for the R→U link
as descending (ascending). In fact, such settings for the composite architectures are reasonable
since it can reach a higher achievable sum rate and balance the throughput fairness among all
multiplexed users.
2) Relay with AF protocol: For networks with AF protocol, where the SIC is only utilized
at the user end, according to the decoding order at the user, the power allocations for S→R and
R→U transmissions must be performed in the same manner. Otherwise, the system performance,
e.g., the achievable sum rate and outage probability, will deteriorate significantly. A special case
is that, for the uplink or downlink architecture, both S→R and R→U transmissions employ the
same fixed power allocation scheme [7]. On the other hand, for the composite architecture, the
channel gains for S→R and R→U links should be sorted in the same manner, namely both in
ascending or descending orders. Unfortunately, such setting is impractical because it will result
in a poor throughput fairness among all users. Therefore, an uplink or downlink architecture for
the AF relay network could be a better choice.

B. System Asymmetry
2
σh 2
σh
s s
The degree of asymmetry for the cooperative relay system can be defined as Ar = 2
σh
SR
w
· σ2 RU
w
,
h
SR RU
where r denotes the relay networks, and s and w denote the strong and weak users for S→R
and R→U links, respectively. Clearly, Ar is the product of degrees of asymmetry for uplink
and downlink transmissions. Once there exists a one-to-one mode, namely σh2sSR = σh2wSR or
2
σh 2
σh
s s
r r
σh2sRU = σh2wRU , we will have A = 2
σh
RU
w
or A = 2
σh
SR
w
. Fig. 3(b) shows the gaps of achievable
RU SR
sum rate between the NOMA [6] and OMA Max-Min schemes under different Ar , where the
OMA Max-Min scheme denotes that the relay Rk̂ is selected to transmit the message from the
 
source to the destination with k̂ = arg max min σh2SRk , σh2RkU . In general, the severer the
k∈{1,2}
asymmetry is, the greater the advantages of NOMA are. Therefore, when Ar is large enough,
e.g., Ar > 3, it is better to apply NOMA than OMA in practice. However, for the diamond
network [6], if the channel conditions satisfy k̂ 6= 2, the superiority of NOMA over OMA will
weaken or even disappear, as the black curve in Fig. 3(b) shows. This can be understood by the
fact that since the source-relay-user channel selected by OMA is better than that for the symbol
who is last decoded in NOMA, OMA achieves a larger achievable sum rate than NOMA. Hence,
for the diamond network, we would prefer NOMA to OMA when not only Ar is large enough,
 7

but also the source-relay-user channel selected by OMA is the same as that for the user who is
last decoded in NOMA. For example, the channel conditions are set as (1)-(4) in the caption of
Fig. 3.

C. Power Allocation

An effective power allocation scheme designed for cooperative relay networks depends on
the symbols’ decoding order. Generally, more power should be assigned to the earlier decoded
symbols while less power is allocated to the later decoded ones. For networks with DF protocol,
it is more complex to perform power allocation than the ones with AF protocol, especially
with global instantaneous channel state information (CSI) [6]. Therefore, finding a simple and
effective power allocation scheme for DF relay networks is demanding.

D. Performance Comparisons

1) Uplink architectures: It has been shown that NOMA can be applied in the uplink transmis-
sion. However, since different users are located in different positions and experience different
channel conditions, we have to face some ticklish problems. On one hand, it is difficult to
realize signal synchronization at both transmitting and receiving ends, and on the other hand,
realizing effective control on the transmit power among different users to avoid interference is
very challenging in practice.
2) Downlink architectures: One of the most famous downlink architectures is the classical
three-node relay network [10], where a source, a half-duplex DF relay, and a user are considered,
as shown in Fig. 2(e). However, most existing NOMA schemes only put emphasis on the Rayleigh
fading scenario [10]. For this reason, a NOMA scheme over Rician fading channel is proposed
to handle this problem [11].
Fig. 2(b) shows another downlink architecture for cooperative relay systems [7], where a
half-duplex AF or DF relay assists the communication between the source and users. Since the
signal-to-interference-and-noise ratio (SINR) with DF relaying is always higher than that with
AF relaying, the DF protocol is preferred over the AF one. As shown in Fig. 4, the outage
probability and achievable sum rate can be significantly improved when the DF protocol is
applied, especially in the low signal-to-noise ratio (SNR) region. This can be understood as
follows. If more power is allocated to the weak user, the system’s outage behavior will be
improved, while the achievable sum rate of the NOMA system will decrease since less power
 8

is allocated to the stronger user, which has a full degree of freedom in the achievable rate.
Therefore, there exists a tradeoff between the outage probability and achievable sum rate.
3) Composite architectures: Figs. 2(c) and (d) illustrate the composite architectures for DF
relay systems. In general, such systems are asymmetric, where the channel gains for uplink
transmission are sorted in ascending (descending) order while those for downlink transmission
in descending (ascending) order [5], [6]. Fig. 3(b) shows the gaps of achievable sum rate between
NOMA and OMA [6] with different degrees of asymmetry.
Since the channel conditions for uplink and downlink transmissions are different in general,
a power waste will be resulted if the transmit powers for the two links are both fixed as Pt .
Without loss of generality, assume that the channel quality for the uplink is better than that for
the downlink. In this case, we do not need to allocate full power to all symbols for the uplink,
whereas for the downlink, Pt must be fully allocated to all symbols. Therefore, if a central
processing unit that can adjust the transmit power of the symbols according to the instantaneous
CSI, the total power consumed by the system can be saved as less than 2Pt [6]. On the other
hand, provided that the requirement of quality of service has been satisfied, the saved power can
be allocated to the weak user so as to improve its data rate, outage probability, and fairness.
Assuming the knowledge of perfect CSIs, the achievable sum rate, energy efficiency, and
normalized power utilization are chosen as performance metrics for comparison. We can see
from Figs. 5(a-1) and (b-1) that NOMA shows significant superiority over OMA in terms of
achievable sum rate. However, oppositely, as shown in Figs. 5(a-2) and (b-2), NOMA performs
worse than OMA in terms of energy efficiency. This phenomenon can be explained by the fact
that NOMA can fully utilize the pre-allocated power according to the channel conditions, which
results in a boost in achievable sum rate, whereas for OMA, the pre-allocated power cannot
be used effectively since the near-far effect exists. As shown in Figs. 5(a-3) and (b-3), the real
power utilization of NOMA is much larger than that of OMA, which can be up to 30%. Although
NOMA outperforms OMA in terms of achievable sum rate, the ratio between achievable sum
rate and power consumption for NOMA is smaller than that of OMA, which means that the
energy efficiency of the former is lower than that of the latter.

V. H YBRID P OWER A LLOCATION S TRATEGY

As discussed above, NOMA can be used for capacity improvement in cooperative relay
systems. For the composite structures, however, design of a dynamic power allocation scheme for
 9

uplink transmission is very challenging. Fortunately, by borrowing the idea of “hybrid-NOMA”

[1], [8], we can solve this problem with a hybrid power allocation strategy that uses both the
knowledge of instantaneous CSI and statistical CSI.
Assume that the network has the knowledge of the statistical CSI of the S→R link and both the
statistical and instantaneous CSIs of the R→U link. The hybrid power allocation strategy can be
performed in two stages. In the first stage, the power allocation scheme in [5] can be employed to
maximize the system’s achievable sum rate with the statistical CSIs of the S→R and R→U links,
by which we can obtain fixed optimal power allocation coefficients for the S→R transmission.
In the second stage, by fixing the power allocation coefficients for the S→R transmission as
that obtained in the first stage, we can obtain new dynamic power allocation coefficients for
the R→U transmission by maximizing the system’s sum rate with the instantaneous CSI of the
R→U link.
To evaluate the performance of the hybrid power allocation strategy, we consider a specific
NOMA relay system in Fig. 2(c). Let ICSI and SCSI denote the power allocation schemes
obtained with the knowledge of instantaneous CSI and statistical CSI for the NOMA scheme,
respectively. Simulation results illustrated in Fig. 6 show that the hybrid power allocation strategy
(marked as HCSI in the figure) can achieve almost the same achievable sum rate as the one
obtained under ICSI (marked as ICSI in the figure). On the other hand, it also shows superiority
over the one obtained under SCSI (marked as SCSI in the figure), even when the near-far
effect is not severe. Simulation results on the energy efficiency ratio and normalized power
utilization are illustrated in Figs. 6(b) and (c), where FDMA is chosen as the benchmark. It is
shown that with the highest spectral efficiency and lowest power consumption, the hybrid power
allocation strategy performs the best in terms of energy efficiency. To sum up, considering its
lower computational complexity on power scheduling, effectiveness to combat the near-far effect,
and superior performance, the hybrid power allocation strategy will be a promising solution for
cooperative relay systems with NOMA.

VI. C HALLENGES , O PPORTUNITIES , AND T RENDS

A. Feedback Overhead and Computational Complexity

For the NOMA cooperative relay schemes, the spectrum efficiency and energy efficiency can be
improved significantly by applying a tailored power allocation scheme. However, such a dynamic
power allocation scheme is obtained under the condition that global instantaneous CSIs can be
 10

available, which means that the signaling overhead, used for CSI and power allocation feedbacks,
is large and would bring huge computational complexity especially when the number of users is
very large. On the other hand, since the feedback delay will cause channel estimation error, and
perfect time synchronization for the uplink is difficult to implement, obtaining perfect CSI is
challenging. To circumvent this problem, new power allocation solutions, which can reduce the
feedback overhead at the expense of marginal performance degradation, are demanding, such as
the hybrid power allocation strategy proposed in the earlier section, and the strategy that uses
limited feedback to obtain CSI [12].

B. Communication Security

Attracted by the advantages of NOMA, recently, the secrecy issue of NOMA has been
studied in [13]. However, how to realize secure communications with NOMA is still an open
issue, especially in NOMA cooperative relay networks, since relays re-transmit a copy of the
information symbols that are summed up via SC and transmitted over the same frequency band,
which means that once the carrier frequency is successfully located by the eavesdropper, all users’
messages may be intercepted. An effective solution is that the system should first gather CSIs and
then choose an appropriate relay to perform secure communications with other relays releasing
cooperative jamming. Specifically, once the appropriate relay is confirmed, the other relays should
release jamming during the two phases. However, the aforementioned discussions assume that the
relay works as a trusted transmitter. In fact, the scenarios with untrusted relays exist, especially
in the cases that the untrusted relays with DF protocol, where all users’ symbols must be decoded
first before they are re-transmitted. Therefore, how to perform security communications for the
NOMA cooperative relay networks in the presence of untrusted relays is an interesting issue.

C. Hardware Development

Although NOMA significantly outperforms the conventional OMA in the application to co-
operative relaying, the corresponding hardware implementation associated with NOMA is more
complex. For example, because of limited processing capability, it is difficult to perform multi-
user detection and interference cancellation for the mobile uesr [1]. Moreover, one of the
reasons why NOMA outperforms OMA is that it allows more users to access the network
simultaneously. Therefore, a high-performance SIC unit for the mobile receiver is a key enabler
for implementation of NOMA in the future.
 11

D. NOMA Cooperative Relay Systems with mmWave

It has been reported that the mmWave with NOMA can achieve higher spectrum and energy
efficiency [14]. On the other hand, since the wavelength of mmWave bands is very short,
effective communications generally require the transmitter and the corresponding receiver to be
located in line-of-sight (LOS) range. Due to the mobility of users and surrounding obstacles, the
blockage and non-line-of-sight (NLOS) channels are inevitable. In order to apply the mmWave
technique to 5G systems while guaranteeing effective coverage, anti-blockage mechanisms are
required through which the mmWave system is able to adaptively switch from LOS transmission
mode to NLOS transmission mode. Thus, the incorporation of cooperative relay techniques into
the mmWave communication system will bring many new possibilities and challenges for the
deployment of mmWave cellular networks [15], e.g., how to place or select the optimal relay
nodes for the network will become very important.

VII. C ONCLUSIONS

In this article, we have discussed and compared the NOMA cooperative relay schemes from
the aspects of basic principles, key features, criterion for system construction, and engineering
feasibility. Three typical structures of the cooperative relay systems have been investigated,
and simulation results have demonstrated the advantages of cooperative relaying with NOMA,
especially for the composite structures. Furthermore, a hybrid power allocation strategy, which
can provide lower computational complexity and reduce signaling overhead at the expense
of marginal sum rate degradation, has been proposed for the NOMA-based cooperative relay
networks. We have also highlighted key challenges, opportunities, and future research trends for
the design of NOMA cooperative relay systems.

R EFERENCES
[1] S. M. R. Islam, N. Avazov, O. A. Dobre, and K. S. Kwak, “Power-domain non-orthogonal multiple access (NOMA) in 5G
Systems: Potentials and challenges,” IEEE Communications Surveys & Tutorials, vol. 19, no. 2, pp. 721–742, Oct. 2017.
[2] S. Chen, B. Ren, Q. Gao, S. Kang, S. Sun, and K. Niu, “Pattern division multiple access-A novel nonorthogonal multiple
access for fifth-generation radio networks,” IEEE Trans. Veh. Technol., vol. 66, no. 4, pp. 3185–3196, Apr. 2017.
[3] L. Dai, B. Wang, Y. Yuan, S. Han, C.-L. I, and Z. Wang, “Non-orthogonal multiple access for 5G: Solutions, challenges,
opportunities, and future research trends,” IEEE Commun. Mag., vol. 53, no. 9, pp. 74–81, Sept. 2015.
[4] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, C.-L. I, and H. V. Poor, “Application of non-orthogonal multiple access in
LTE and 5G networks,” IEEE Commun. Mag., vol. 55, no. 2, pp. 185–191, Feb. 2017.
[5] M. F. Kader, M. B. Shahab, and S. Y. Shin, “Exploiting non-orthogonal multiple access in cooperative relay sharing,” IEEE
Commun. Lett., vol. 21, no. 5, pp. 1159–1162, May 2017.
[6] D. Wan, M. Wen, H. Yu, Y. Liu, F. Ji, and F. Chen, “Non-orthogonal multiple access for dual-hop decode-and-forward
relaying,” in Proc. IEEE Global Communications Conference, Washington DC, USA, Dec. 2016, pp. 1-6.
 12

[7] D. Wan, M. Wen, F. Ji, Y. Liu, and Y. Huang, “Cooperative NOMA systems with partial channel state information over
Nakagami-m fading channels,” IEEE Trans. Commun., vol. 66, no. 3, Mar. 2018.
[8] Z. Ding, P. Fan, and H. V. Poor, “Impact of user pairing on 5G non-orthogonal multiple access downlink transmissions,”
IEEE Trans. Veh. Technol., vol. 65, no. 8, pp. 1462–1465, Aug. 2016.
[9] D. Tse and P. Viswanath, Fundamentals of Wireless Communications. Cambridge University Press, 2005.
[10] J. Kim and I. Lee, “Capacity analysis of cooperative relaying systems using non-orthogonal multiple access,” IEEE
Commun. Lett., vol. 19, no. 11, pp. 1949–1952, Nov. 2015.
[11] R. Jiao, L. Dai, J. Zhang, R. MacKenzie, and M. Hao, “On the performance of NOMA-based cooperative relaying systems
over Rician fading channels,” IEEE Trans. Veh. Technol., vol. 66, no. 12, pp. 11409–11413, Dec. 2017.
[12] P. Xu, Y. Yuan, Z. Ding, X. Dai, and R. Schober, “On the outage performance of non-orthogonal multiple access with
one-bit feedback,” IEEE Trans. Wireless Commun., vol. 15, no. 10, pp. 6716–6730, Oct. 2016.
[13] Y. Liu, Z. Qin, M. Elkashlan, Y. Gao, and L. Hanzo, “Enhancing the physical layer security of non-orthogonal multiple
access in large-scale networks,” IEEE Trans. Wireless Commun., vol. 16, no. 3, pp. 1656–1672, Mar. 2017.
[14] B. Wang, L. Dai, Z. Wang, N. Ge, and S. Zhou, “Spectrum and energy efficient beamspace MIMO-NOMA for millimeter-
wave communications using lens antenna array,” IEEE J. Sel. Areas Commun., vol. 35, no. 10, pp. 2370–2382, Oct. 2017.
[15] J. Zhang, X. Ge, Q. Li, M. Guizani, and Y. Zhang, “5G millimeter-wave antenna array: Design and challenges,” IEEE
Wireless Commun., vol. 24, no. 2, pp. 106–112, Apr. 2017.

B IOGRAPHIES

Dehuan Wan (eewan e@mail.scut.edu.cn) received the B.S. degree from Henan Normal Uni-
versity, Xinxiang, China, in 2007. He is currently working toward the Ph.D. degree with the
South China University of Technology, Guangzhou, China. His recent research interests include
non-orthogonal multiple access and index modulation.

Miaowen Wen (eemwwen@scut.edu.cn) received the B.S. degree from Beijing Jiaotong Uni-
versity, Beijing, China, in 2009, and the Ph.D. degree from Peking University, Beijing, China,
in 2014. From 2012 to 2013, he was a Visiting Student Research Collaborator with Princeton
University, Princeton, NJ, USA. He is currently an Associate Professor with the South China
University of Technology, Guangzhou, China. He has authored a book and more than 80 papers
in refereed journals and conference proceedings. His research interests include index modulation,
non-orthogonal multiple access, physical layer security, and molecular communications. Dr. Wen
was the recipient of the Excellent Doctoral Dissertation Award from Peking University and Best
Paper Awards at IEEE ITST2012, ITSC2014, and ICNC2016. He was an Exemplary Reviewer for
the IEEE Communications Letters in 2017. He currently serves as an Associate Editor of the IEEE
ACCESS, and on the Editorial Board of the EURASIP Journal on Wireless Communications
and Networking, the ETRI Journal, and the Physical Communication (Elsevier).

Fei Ji (eefeiji@scut.edu.cn) received the B.S. degree in applied electronic technologies from
Northwestern Polytechnical University, Xian, China, and the M.S. degree in bioelectronics and
 13

Ph.D. degree in circuits and systems both from the South China University of Technology,
Guangzhou, China, in 1992, 1995, and 1998, respectively. She was a Visiting Scholar with the
University of Waterloo, Canada, from June 2009 to June 2010. She worked in the City University
of Hong Kong as a Research Assistant from March 2001 to July 2002 and a Senior Research
Associate from January 2005 to March 2005. She is currently a Professor with the School of
Electronic and Information Engineering, South China University of Technology. Her research
focuses on wireless communication systems and networking.

Hua Yu (yuhua@scut.edu.cn) received the B.S. degree in mathematics from Southwest Univer-
sity, Chongqing, China, in 1995 and the Ph.D. degree in communication and information system
from South China University of Technology, Guangzhou, China, in 2004. He was a Visiting
Scholar at the School of Marine Science and Policy, University of Delaware, USA, from 2012 to
2013. He is currently a Professor at the School of Electronic and Information Engineering, South
China University of Technology. He is also the Director of the National Engineering Technology
Research Center for Mobile Ultrasonic Detection, Department of Underwater Communications.
His research interests are in the physical layer technologies of wireless communications and un-
derwater acoustic communications. Dr. Yu was the Publication Chair and the Technical Program
Committee Member of the 11th IEEE International Conference on Communication Systems in
2008.

Fangjiong Chen (eefjchen@scut.edu.cn) received the B.S. degree in electronics and infor-
mation technology from Zhejiang University, Hangzhou, China, in 1997 and the Ph.D. degree
in communication and information engineering from South China University of Technology,
Guangzhou, China, in 2002. After graduation, he joined the School of Electronics and Infor-
mation Engineering, South China University of Technology, where he was a Lecturer and an
Associate Professor, from 2002 to 2005 and from 2005 to 2011, respectively. He is currently
a full-time Professor at the School of Electronics and Information Engineering, South China
University of Technology. He is also the Director of the Mobile Ultrasonic Detection National
Research Center of Engineering Technology, Department of Underwater Detection and Imaging.
His research focuses on signal detection and estimation, array signal processing, and wireless
communication. Dr. Chen received the National Science Fund for Outstanding Young Scientists
in 2013, and was elected in the New Century Excellent Talent Program of MOE, China, in 2012.
14
 15

D1 S1 S1 D1

S D S D

DN SN SM DN

(a) One-to-one (b) One-to-many (c) Many-to-one (d) Many-to-many

Fig. 1. Basic communication modes.

 16

1st phase 2nd phase

S1 U1

N N
! Pu
i 1 i
Pt N N ! i 1 i
Pd Pt
! i
Pu
1 i
Pt ! i
Pd
1 i
Pt
Si R U S R Ui
Pi u Pi d

SN UN

(a) Uplink architecture (b) Downlink architecture

(OMA or NOMA) (OMA or NOMA)

S1 R1

U1
N
! i 1 i
P u
Pt
Pi d Pi d Pi u
R Ui S Ri U
Si N N
Pi u ! i
Pd
1 i
Pt ! i
Pu
1 i
Pt
N
! i
Pd
1 i
Pt
SN
UN RN

(c) Composite architecture (d) Composite architecture

(OMA or NOMA) (OMA or NOMA)

R U1 U2

Pt N
Pjd , ! j 2 Pi d Pt
N
2
Pi d , ! i 1 Pi d Pt Pi d , ! i 1 Pi d Pt
S U S UN

(e) Downlink architecture (f) Downlink architecture

(OMA or NOMA) (NOMA)

Fig. 2. Architectures of the cooperative relay systems for OMA and NOMA, where S, R, and U denote the source
(acting as a transmitter), relay (acting as a transmitter and a receiver), and user (acting as a receiver in OMA, but
can also act as a transmitter in NOMA), respectively, Pt stands for the total power during each transmission, and
the superscript u or d denotes the uplink or downlink transmission.
 17

6 0.6

r
A =23.3
0.5
5

The Gaps of the Achievable Sum Rate [bps/Hz]

Ar=30

0.4
Ar=36
Achievable Sum Rate [bps/Hz]

Ar=45
0.3

0.2

2
0.1 Ar=7.5

1
0

The same decoding manner for SR and RU

Different decoding manner for SR and RU
0 −0.1
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SNR [dB] SNR [dB]
(a) (b)

Fig. 3. Comparison between NOMA [6] and OMA Max-min schemes: (a) achievable sum rate with DF protocol
2 2 2 2
under different decoding manners, where the channel setting is σSR 1
= 1, σSR 2
= 10, σR 1U
= 9, and σR 2U
= 2;
r 2 2
(b) the gaps of achievable sum rate under different A , where the channel settings are: (1) σSR1 = 1, σSR2 = 10,
2 2 2 2 2 2 2 2
σR 1U
= 9, and σR 2U
= 2; (2) σSR 1
= 1, σSR 2
= 10, σR 1U
= 9, and σR 2U
= 3; (3) σSR 1
= 1, σSR 2
= 10,
2 2 2 2 2 2 2 2
σR1 U = 7, and σR2 U = 3; (4) σSR1 = 1, σSR2 = 12, σR1 U = 9, and σR2 U = 3; (5) σSR1 = 4, σSR 2
= 9,
2 2
σR1 U = 10, and σR2 U = 3.
 18

0
6 10
4.05 User1−AF
User2−AF
4
5.5 User1−DF
3.95 User2−DF
User1−AF
5 3.9
User2−AF
3.85 User1−DF
4.5 User2−DF
3.8 −1 User1−TDMA
Achievable Sum Rate [bps/Hz]

10
3.75 User2−TDMA
4
3.7
Outage probability

3.65
3.5 20

0.02
3

−2
10
2.5

0.015
2

1.5 NOMA−AF
NOMA−DF 0.01
NOMA−AF
1 −3
NOMA−DF 10
TDMA
20
0.5
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SNR [dB] SNR [dB]
(a) (b)

Fig. 4. Comparisons of the: (a) achievable sum rate and (b) outage probability among the NOMA-AF, NOMA-DF
2 2 2
and TDMA schemes [7], where the channel setting is σSR = 8, σRU 1
= 2, and σRU2
= 10. The power allocation
coefficients for user 1 and user 2 in NOMA are 0.6875 and 0.3125 (solid lines), 0.8 and 0.2 (dash-dot lines),
respectively, while the resources allocated to user 1 and user 2 in TDMA are average.
 19

Achievable Sum Rate [bps/Hz]

Achievable Sum Rate [bps/Hz]

6 6
OMA OMA
NOMA [5] NOMA [6]
4 4

2 2

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SNR [dB] SNR [dB]
(a−1) (b−1)
Energy Efficiency[Kbit/joule]

Energy Efficiency[Kbit/joule]
1.5 3
OMA OMA
NOMA [5] NOMA [6]
1 2

0.5 1

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
SNR [dB] SNR [dB]
(a−2) (b−2)
Normalized Power Utilization

Normalized Power Utilization

1 1
OMA OMA
0.9 NOMA [5] 0.9 NOMA [6]

0.8 0.8

0.7 0.7

0 5 10 15 20 25 30 0 5 10 15 20 25 30
SNR [dB] SNR [dB]
(a−3) (b−3)

Fig. 5. Comparison in terms of achievable sum rate, energy efficiency, and normalized power utilization for: (a) the
NOMA [5] and OMA schemes, where the channel setting is σS2 1 R = 9, σS2 2 R = 3, σRU
2
1
2
= 2, and σRU2
= 10; (b)
2 2 2 2
the NOMA [6] and OMA schemes, where the channel setting is σSR1 = 1, σSR2 = 10, σR1 U = 9, and σR 2U
= 2.
The power allocation schemes applied for them are dynamic, namely adapted to the instantaneous CSIs.
 20

8 0.3 1
FDMA benchmark
FDMA−Benchmark
SCSI−channel1
SCSI−channel1
5.6 SCSI−channel2 0.98 SCSI−channel2
7
0.25 HSCI−channel1 HCSI−channel1
5.4 HCSI−channel2 HCSI−channel2

0.96
6 5.2
0.2
29.6 30 30.4 0.94
Achievable Sum Rate [bps/Hz]

Normalized Power Utilization

5
Energy Efficiency Ratio

0.15
0.92

0.9
0.1

3
0.88
0.05
FDMA−channel1
2 FDMA−channel2
0.86
HCSI−channel1
HCSI−channel2
ICSI−channel1 0
1
0.84
ICSI−channel2
SCSI−channel1
SCSI−channel2
0 −0.05 0.82
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
SNR [d] SNR [dB] SNR [dB]
(a) (b) (c)

Fig. 6. Comparison in terms of: (a) achievable sum rate, (b) energy efficiency ratio, and (c) normalized power
utilization among the HCSI, ICSI, SCSI, and FDMA schemes under two channel settings, namely setting 1: σS2 1 R =
9, σS2 2 R = 3, σRU
2
1
2
= 2, and σRU 2
= 10 (solid lines) and setting 2: σS2 1 R = 9, σS2 2 R = 3.8, σRU
2
1
= 5, and
2
σRU2 = 10 (dashed lines), where the degrees of asymmetry for the two settings are 15 and 4.7, respectively. The
power allocation schemes applied for SCSI and FDMA are static and dependent on the statistical CSI only.