You are on page 1of 30

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

A Survey of Non-Orthogonal Multiple Access for


5G
Linglong Dai, Senior Member, IEEE, Bichai Wang, Zhiguo Ding, Member, IEEE, Zhaocheng Wang, Senior
Member, IEEE, Sheng Chen, Fellow, IEEE, and Lajos Hanzo, Fellow, IEEE

Abstract—In the 5th generation (5G) of wireless communica- 4G Fourth generation mobile system
tion systems, hitherto unprecedented requirements are expected 5G Fifth generation mobile system
to be satisfied. As one of the promising techniques of addressing AWGN Additive white gaussian noise
these challenges, non-orthogonal multiple access (NOMA) has
been actively investigated in recent years. In contrast to the family BDM Bit division multiplexing
of conventional orthogonal multiple access (OMA) schemes, the BER Bit error rate
key distinguishing feature of NOMA is to support a higher BLER Block error ratio
number of users than the number of orthogonal resource slots BPSK Binary phase shift keying
with the aid of non-orthogonal resource allocation. This may be BS Base station
realized by the sophisticated inter-user interference cancellation
at the cost of an increased receiver complexity. In this article, CDMA Code division multiple access
we provide a comprehensive survey of the original birth, the CIR Channel impulse response
most recent development, and the future research directions CoMP Coordinated multi-point
of NOMA. Specifically, the basic principle of NOMA will be CS Compressive sensing
introduced at first, with the comparison between NOMA and CSI Channel state information
OMA especially from the perspective of information theory. Then,
the prominent NOMA schemes are discussed by dividing them CTU Contention transmission unit
into two categories, namely, power-domain and code-domain DEA Differential evolution algorithm
NOMA. Their design principles and key features will be discussed DF Decision feedback
in detail, and a systematic comparison of these NOMA schemes EA Evolutionary algorithm
will be summarized in terms of their spectral efficiency, system ESE Elementary signal estimator
performance, receiver complexity, etc. Finally, we will highlight
a range of challenging open problems that should be solved EXIT Extrinsic information transfer-function
for NOMA, along with corresponding opportunities and future FDMA Frequency division multiple access
research trends to address these challenges. GA Genetic algorithm
Index Terms—5G, non-orthogonal multiple access (NOMA), i.i.d independent and identically distributed
multi-user detection (MUD), spectral efficiency, massive connec- IDMA Interleave-division multiple access
tivity, overloading, low latency. IoT Internet of things
IrCC Irregular convolutional code
N OMENCLATURE ISI Inter-symbol-interference
JMPA Joint message passing algorithm
1G First generation mobile system KPI Key performance indicators
2G Second generation mobile system LDPC Low-density-parity-check
3G Third generation mobile system LDS Low-density spreading
L. Dai, B. Wang, and Z. Wang are with Tsinghua National Lab- LDS-CDMA Low-density spreading code division multiple
oratory for Information Science and Technology (TNList), Departmen- access
t of Electronic Engineering, Tsinghua University, Beijing 100084, P. LDS-OFDM Low-density spreading aided orthogonal fre-
R. China (E-mail: daill@tsinghua.edu.cn, wbc15@mails.tsinghua.edu.cn,
zcwang@tsinghua.edu.cn). quency division multiplexing
Z. Ding is with the School of Electrical and Electronic Engineer- LMS Least mean square
ing, The University of Manchester, Manchester, M13 9PL, UK. (E-mail: Log-MPA Logarithmic-domain message passing algorithm
zhiguo.ding@manchester.ac.uk).
S. Chen is with Electronics and Computer Science, University of Southamp- LTE Long term evolution
ton, Southampton SO17 1BJ, UK (E-mail: sqc@ecs.soton.ac.uk), and also MAP Maximum a posteriori probability
with King Abdulaziz University, Jeddah 21589, Saudi Arabia. MBER Minimum bit error rate
L. Hanzo is with Electronics and Computer Science, University of
Southampton, Southampton SO17 1BJ, U.K. (E-mail: lh@ecs.soton.ac.uk). MC-CDMA Multi-carrier code division multiple access
This work was supported by the National Natural Science Foundation of MIMO Multiple input multiple output
China for Outstanding Young Scholars (Grant No. 61722109), the National ML Maximum likelihood
Natural Science Foundation of China (Grant No. 61571270), and the Royal
Academy of Engineering under the UK-China Industry Academia Partnership MMSE Minimum mean square error
Programme Scheme (Grant No. UK-CIAPP\49). The work of Z. Ding was MPA Message passing algorithm
supported by the UK EPSRC under grant number EP/ L025272/1 and by MRC Maximum ratio combining
H2020-MSCA-RISE-2015 under grant number 690750. L. Hanzo would also
like to acknowledge the financial support of the European Research Council’s MU-SCMA Multi-user sparse code multiple access
Advanced Fellow Grant QuantCom. MUD Multi-user detection

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

FDMA
(1G-1980’s)

TDMA
Orthogonal (2G-1990’s)
Multiple Access
(OMA) CDMA
(3G-2000’s)

OFDMA
Multiple (4G-2010’s)
Access
Technology
IDMA
(2003)

LDS-CDMA LDS-OFDM SCMA


(2006) (2010) (2013)
Non-Orthogonal
Multiple Access Power-Domain
(NOMA) NOMA SAMA
(2012) (2014)

BDM
(2013) PDMA
(2014)
MUSA
(2014)

Fig. 1. The milestones of multiple access technologies.

MUSA Multi-user shared access traffic increase by the year 2020 [1]. Specifically, the key
NOMA Non-orthogonal multiple access performance indicators (KPI) advocated for 5G solutions can
OFDMA Orthogonal frequency division multiple access be summarized as follows [2]: 1) The spectral efficiency is
OMA Orthogonal multiple access expected to increase by a factors of 5 to 15 compared to 4G;
PAPR Peak-to-average-power ratio 2) To satisfy the demands of massive connectivity for IoT,
PDMA Pattern division multiple access the connectivity density target is ten times higher than that of
PIC Parallel interference cancelation 4G, i.e. at least 106 /km2 ; 3) 5G is also expected to satisfy the
PSO Particle swarm optimization requirements of a low latency (radio latency ≤ 1 ms), low cost
QAM Quadrature amplitude modulation (≥ 100 times the cost efficiency of 4G), and the support of
QPSK Quaternary phase shift keying diverse compelling services. In order to satisfy these stringent
RWBS Repeated weighted boosting search requirements, advanced solutions have to be conceived.
SAMA Successive interference cancellation amenable multi-
Over the past few decades, wireless communication systems
ple access
have witnessed a “revolution” in terms of their multiple access
SC-FDMA Single carrier-frequency division multiple access
techniques. Specifically, for 1G, 2G, 3G, and 4G wireless
SCMA Sparse code multiple access
communication systems, frequency division multiple access
SD Sphere detector
(FDMA), time division multiple access (TDMA), code di-
SDMA Spatial division multiple access
vision multiple access (CDMA), and orthogonal frequency
SE Spectral efficiency
division multiple access (OFDMA) have been used as the
SIC Successive interference cancellation
corresponding key multiple access technologies, respective-
SINR Signal-to-interference-and-noise ratio
ly [3] [4]. From the perspective of their design principles, these
SNR Signal-to-noise ratio
multiple access schemes belong to the category of orthogonal
ST-DFE Space-time decision feedback equalization
multiple access (OMA), where the wireless resources are or-
STE Space time equalizer
thogonally allocated to multiple users in the time-, frequency-,
TDMA Time division multiple access
code-domain or according in fact based on their combinations.
UE User
We might collectively refer to these domains as “resources”. In
URC Unity rate code
this way the users’ information-bearing signals can be readily
UWB Ultra-wideband
separated at a low complexity by employing relatively cost-
efficient receivers. However, the number of supported users
I. I NTRODUCTION is limited by the number of available orthogonal resources in
HE rapid development of the mobile Internet and the OMA. Another problem is that, despite the use of orthogonal
T Internet of things (IoT) leads to challenging requirements
for the 5th generation (5G) of wireless communication sys-
time-, frequency- or code-domain resources, the channel-
induced impairments almost invariably destroy their orthogo-
tems, which is fuelled by the prediction of 1000-fold data nality. More specifically, considering that two signals s1 (t) and

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

s2 (t), which are orthogonal either in time-, frequency- or code- I Introduction


II Basic principles and advantages of NOMA
domain, are transmitted over dispersive channels h1 (t) and
II-A Channel capacity comparison of OMA and NOMA
h2 (t), separately, the received signals x1 (t) = s1 (t) ⊗ h1 (t)
II-B Advantages of NOMA compared with OMA
and x2 (t) = s2 (t) ⊗ h2 (t) are usually non-orthogonal due to III Dominant NOMA solutions
time or frequency expansion. Hence, some high-complexity III-A Power-Domain NOMA
“orthogonality restoring measures”, such as multi-user equal- III-A1 Basic NOMA relying on a SIC receiver
izers, have to be invoked. Consequently, it remains a challenge III-A2 NOMA in MIMO systems
for OMA to satisfy the radical spectral efficiency and massive III-A3 Cooperative NOMA
connectivity requirements of 5G. III-A4 NOMA in CoMP
The innovative concept of non-orthogonal multiple access III-A5 Application of power-domain NOMA
III-B Code-Domain NOMA
(NOMA) has been proposed in order to support more users
III-B1 LDS-CDMA
than the number of available orthogonal time-, frequency-,
III-B2 LDS-OFDM
or code-domain resources. The basic idea of NOMA is to III-B3 SCMA
support non-orthogonal resource allocation among the users III-B4 MUSA
at the ultimate cost of increased receiver complexity, which is III-B5 SAMA
required for separating the non-orthogonal signals. Recently, III-C Other NOMA schemes
several NOMA solutions have been actively investigated [5]– III-D User grouping and resource allocation
[13], which can be basically divided into two main cate- III-E Comparison of NOMA solutions
gories, namely power-domain NOMA [14]–[65] and code- III-F Performance evaluations and transmission experiments
domain NOMA [66]–[99], including multiple access solutions IV Challenges, opportunities, and future research trends
relying on low-density spreading (LDS) [66]–[75], sparse code V Conclusions
multiple access (SCMA) [76]–[97], multi-user shared access
(MUSA) [98], successive interference cancellation amenable Fig. 2. The structure of this article.
multiple access (SAMA) [99], etc. Some other closely-related
multiple access schemes, such as spatial division multiple opportunities and future research trends will be highlighted in
access (SDMA) [100]–[113], pattern division multiple access order to provide some insights into potential future research
(PDMA) [114] [115] and bit division multiplexing (BD- in this promising field. Finally, our conclusions are offered in
M) [116] have also been proposed. The milestones of NOMA Section V. The structure of this article is shown in Fig. 2 at a
techniques are summarized in Fig. 1. Note that most the glance.
existing survey papers on NOMA [5]–[10] only focus on the
power-domain NOMA. Some code-domain NOMA schemes
are briefly introduced in [11] and [12]. On the contrary, in this II. BASIC P RINCIPLES AND A DVANTAGES OF NOMA
paper, both power-domain NOMA and code-domain NOMA, In this section, we will firstly compare the basic principles
as well as other NOMA schemes proposed as part of the Rel- of OMA and NOMA. Then, the pros and cons of NOMA are
14 3GPP NR Study Item shown in Table I are introduced contrasted to those of OMA in detail.
to provide a more comprehensive review of the most recent In conventional OMA schemes, such as FDMA, TDMA,
progress on NOMA. Furthermore, the comparison of 15 NO- CDMA and OFDMA used for 1G, 2G, 3G, and 4G, re-
MA schemes proposed for Rel-14 3GPP NR Study Item are spectively, multiple users are allocated to orthogonal radio
also provided to shed some insights on the unified design resources in the time-, frequency-, code-domain or to their
and implementation of NOMA. Moreover, in addition to the combinations. More specifically, each user transmits a unique,
basic principle and theoretical analysis, prototype evaluations user-specific signal over its unique frequency resource in
and field test results are also presented to verify the actual FDMA, hence the receiver can readily detect all users’ data in
performance gain of NOMA in this paper, which are not their corresponding frequency bands, respectively. Similarly,
discussed in the existing papers. in TDMA, an exclusive time slot is allocated to each user,
In this article, we will discuss the basic principles as well hence it is easy to distinguish the different users’ signals at
as pros and cons of NOMA in Section II. In Section III, the receivers in the time domain. In CDMA, multiple users can
the design principles and key features of these dominant share the same time-frequency resources, while the transmitted
NOMA solutions as well as the user grouping and resource symbols of different users may be mapped to orthogonal
allocation will be discussed, and a systematic comparison of spreading sequences, such as Walsh-Hadamard codes. Hence,
these NOMA schemes will be provided in terms of their a low-complexity decorrelation receiver can be used for multi-
spectral efficiency, system performance, receiver complexity, user detection (MUD). OFDMA can be regarded as a smart
etc. Performance evaluations and transmission experiments of integration of FDMA and TDMA, where the radio resources
NOMA are also introduced to verify the analytical results. are orthogonally partitioned in the time-frequency grid. Theo-
In Section IV, we will highlight a range of challenging open retically, as a benefit of orthogonal resource allocation, there
problems that should be solved for supporting NOMA, such is no interference among users in OMA systems, hence low-
as the theoretical analysis, the sophisticated transmitter design, complexity detectors with linear complexity can be used
and the tradeoff between the attainable system performance to separate the different users’ signals [117]. However, the
versus receiver complexity. Accordingly, the corresponding maximum number of supportable users is rigidly restricted by

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

TABLE I
NOMA SCHEMES PROPOSED FOR THE R EL -14 3GPP NR S TUDY I TEM .
Uplink (UL)
NOMA schemes Company Full Name
Downlink (DL)
Power-domain Power-domain
1 DCM UL/DL
NOMA [14]–[65] non-orthogonal multiple access
2 SCMA [76]–[97] Huawei Sparse code multiple access UL/DL
3 MUSA [98] ZTE Multi-user shared access UL/DL
4 PDMA [114] [115] CATT Pattern division multiple access UL/DL
Low code rate and signature based
5 LSSA [134] ETRI UL
shared access
6 RSMA [135]–[137] Qualcomm Resource spread multiple access UL
7 IGMA [138] Samsung Interleave-grid multiple access UL/DL
8 IDMA [139] Nokia Interleaver division multiple access UL
9 NCMA [140] LGE non-orthogonal coded multiple access UL
10 NOCA [142] Nokia Non-orthogonal coded access UL
11 GOCA [143] MTK Group orthogonal coded access UL
Low density spreading-signature
12 LDS-SVE [151] Fujitsu UL/DL
vector extension
13 FDS [152] Intel Frequency domain spreading UL
14 LCRS [152] Intel Low code rate spreading UL
15 RDMA [143] MTK Repetition division multiple access UL

R2 (b/s)
N 0 = 10 dB
2
the number of orthogonal resources available in conventional p1 h1
A
OMA schemes, which becomes a hard limit when massive 3.46 p2 h2
2
N 0 = 10 dB
connectivity is required for 5G. Additionally, it has been
2.20
theoretically shown that OMA cannot always achieve the C NOMA
maximum attainable sum-rate of multi-user wireless systems, 0.93 B OMA
while NOMA is capable of achieving the multi-user capacity
with the aid of time-sharing or rate-splitting if necessary [118], R1 (b/s)
0 0.93 2.20 3.46
which will be explained in detail in the following subsection.
(a)
In order to circumvent the above limitation of OMA, R2 (b/s)
N 0 = 20 dB
2
p1 h1
NOMA has been recently investigated as a design alternative. A
p2 h2
2
N 0 = 0 dB
1.00
The key distinguishing feature of NOMA is to support a 0.80
higher number of users than the number of orthogonal resource NOMA
slots, with the aid of non-orthogonal resource allocation. This OMA
0.067
may be achieved by the sophisticated inter-user interference C B
cancellation at the cost of an increased receiver complexity, 0 R1 (b/s)
3.70 5.67 6.61 6.66
e.g., computational complexity with polynomial or exponential
order. Furthermore, it is worth bearing in mind that even (b)
if we opt for OMA schemes, the time-domain signals are
smeared by their convolution with the dispersive channel Fig. 3. Channel capacity comparison of OMA and NOMA in the uplink
impulse response (CIR). In addition, a historic concept of AWGN channel: (a) Symmetric channel; (b) Asymmetric channel [5] ⃝IEEE.
c
NOMA is constituted by the family of CDMA systems relying
on more non-orthogonal sequences than the number of chips A. Channel capacity comparison of OMA and NOMA
Nc in a sequence to support more than Nc users at the
cost of imposing inter-user interference. Such interference can From the perspective of information theory, for the capacity
only be efficiently mitigated with the aid of powerful multi- of multiple access channels in both additive white Gaussian
user detectors. Hence, the NOMA concept is appealing. The noise (AWGN) and fading scenarios, we have the following
family of NOMA schemes can be basically divided into two results for OMA and NOMA (applicable to both power-
categories: power-domain NOMA and code-domain NOMA. domain NOMA and code-domain NOMA):
In power-domain NOMA, different users are assigned different • AWGN channel: In the uplink of an AWGN channel

power levels according to their channel quality, while the same supporting K users (K can be larger than 2), the capacity
time-frequency-code resources are shared among multiple of the multiple access channel can be formulated as [119]
users. At the receiver side, power-domain NOMA exploits the  ∑K 
users’ power-difference, in order to distinguish different users ∑ P
K
 i

based on successive interference cancellation (SIC). Code- Ri ≤ W log  i=1
1 + N0 W  ,
 (1)
domain NOMA is similar to CDMA or multi-carrier CDMA i=1
(MC-CDMA), except for its preference for using low-density
sequences or non-orthogonal sequences having a low cross- where W is the bandwidth, Pi is the transmitted pow-
correlation. er, and N0 is the power spectral density of Gaussian

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

R2 (b/s)
N 0 = 10 dB
2
p1 h1 pairs of NOMA is in general beyond that of OMA in
asymmetric channels.
N 0 = 10 dB
2
3.46 p2 h2
• Fading channels: In fading channels, the sum capacity
in the uplink - provided that the channel state information
NOMA
(CSI) is only known at the receiver - can be represented
OMA as
{ ( ∑K )}
k=1 |hk | Pave
R1 (b/s) 2
3.46 Csum = E log 1 + , (3)
N0
(a)
where we assume that each user has the same average
R2 (b/s) power Pave . In this way, OMA is strictly suboptimal
N 0 = 20 dB
2
p1 h1 in the uplink, while NOMA relying on MUD is opti-
1
p2 h2
2
N 0 = 0 dB
mal [118].
0.8
• MIMO-NOMA and MIMO-OMA: NOMA can be also
0.6 extended to MIMO scenarios, where the BS has M
NOMA
0.4 antennas and each user is equipped with N antennas.
OMA
0.2 Additionally, multiple users can be randomly grouped
R1 (b/s) into M clusters with two users in each cluster. It has
0 1 2 3 4 5 6 7
been shown in [121] that MIMO-NOMA is strictly better
(b) than MIMO-OMA in terms of the sum channel capacity
(except for transmission to a single user in MIMO sys-
Fig. 4. Channel capacity comparison of OMA and NOMA in the downlink tems), i.e., for any rate pair achieved by MIMO-OMA
AWGN channel: (a) Symmetric channel; (b) Asymmetric channel [5] ⃝IEEE.
c schemes, there is a power split for which MIMO-NOMA
is capable of achieving rate pairs that are strictly higher.
noise. More particularly, according to the capacity anal-
ysis found in the pioneering contribution of Tse and B. Advantages of NOMA compared to OMA
Viswanath [118], Fig. 3 and Fig. 4 from [5] provided We can see from the capacity analysis that it is feasible
the channel capacity comparison of OMA and NOMA, for NOMA to achieve a higher transmission rate than OMA.
where a pair of users communicating with a base station Specifically, the main advantages of NOMA compared to the
(BS) over an AWGN channel is considered as an example classical OMA can be summarized as follows:
without loss of generality. Fig. 3 showed that the uplink
• Improved spectral efficiency and cell-edge through-
of NOMA is capable of achieving the capacity region,
put: The time-frequency resources are shared non-
while OMA is suboptimal in general, except at one
orthogonally among users both in the power-domain
point. However, at this optimal point, rate-fairness is
NOMA and in the code-domain of NOMA. As described
not maintained, since the rate of the low-power user is
above, in the uplink of AWGN channels, although both
much lower than that of the higher-power user when the
OMA and NOMA are capable of achieving the maxi-
difference of the received powers of the two users is high.
mum attainable sum capacity, NOMA supports a more
Note that the results for the simple two-user case can be
equitable user fairness. Additionally, the capacity bound
extended to the general case of an arbitrary number of
of NOMA is higher than that of OMA in the downlink
users [118]. Explicitly, it is shown in [118] that there
of AWGN channels. In multi-path fading channels sub-
are exactly K! corner points when the K-user scenario
jected to inter-symbol-interference (ISI), although OMA
is considered and the K-user NOMA system can achieve
is indeed capable of achieving the maximum attainable
the same optimal sum rate at all of these K! corner points.
sum capacity in the downlink, NOMA relying on MUD
In the downlink, the boundary of the capacity region is
is optimal, while OMA remains suboptimal, if the CSI is
given by the rate tuples [120]:
only known at the downlink receiver.
 
• Massive connectivity: Non-orthogonal resource alloca-
  tion in NOMA indicates that the number of supportable
 P |h |
2 
Rk = W log 
1 + (k k )  , (2)

users/devices is not strictly limited by the number of or-
 ∑K
2 thogonal resources available. Therefore, NOMA is capa-
N0 W + Pj |hk |
j=k+1 ble of significantly increasing the number of simultaneous
connections in rank-deficient scenarios, hence it has the

K potential to support massive connectivity. Of course, it
which is valid for all possible splits P = Pk of total should be noted that some practical implementation issues
k=1
power at the BS. The optimal points can be achieved in NOMA systems, such as its hardware imperfections
by NOMA with the aid of superposition coding at the and computational complexity, may hinder the realization
transmitter and SIC at each of the receivers [118]. More of massive connectivity, which will be detailed in Section
particularly, Fig. 4 showed that the boundary of the rate IV.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

•Low transmission latency and signaling cost: In con- and key features of power-domain NOMA have been described
ventional OMA relying on access-grant requests, a user in detail. In contrast to the multiple access schemes relying on
first has to send a scheduling request to the base station the time-, frequency-, code-domain or on their combinations,
(BS). Then, upon receiving this request, the BS schedules NOMA can be realized in a recently emerged new domain,
the user’s uplink transmission by responding with a clear- namely in the power domain. At the transmitter, different
to-send signal in the downlink channel. Thus, a high signals generated by different users are directly superimposed
transmission latency and a high signaling overhead will on each other after classic channel coding and modulation.
be imposed, which becomes unacceptable in the case of Multiple users share the same time-frequency resources, and
massive 5G-style connectivity. Specifically, the access- then are detected at the receivers by MUD algorithms such
grant procedure in LTE takes about 15.5 ms before the as SIC. In this way, the spectral efficiency can be enhanced
data is transmitted [122]. In this way, the radical require- at the cost of an increased receiver complexity compared to
ment of maintaining a user delay below 1 ms cannot be conventional OMA. Additionally, it is widely recognized based
readily satisfied [114]. By contrast, dynamic scheduling on information theory that non-orthogonal multiplexing using
is not required in some of the uplink NOMA schemes. superposition coding at the transmitter and SIC at the receiver
To elaborate a little further, in the uplink of a SCMA not only outperforms classic orthogonal multiplexing, but it
system, grant-free multiple access can be realized for is also optimal from the perspective of achieving the capacity
users associated with pre-configured resources defined in region of the downlink broadcast channels [118].
the time- and frequency-domain, such as the codebooks, Some practical considerations for power-domain NOMA,
as well as the pilots. By contrast, at the receiver blind such as multi-user power allocation, signalling overhead, SIC
detection and compressive sensing (CS) techniques can be error propagation and user mobility, were discussed in [14].
used for performing joint activity and data detection [93]. To achieve a further enhancement of its spectral efficiency,
Hence again, beneficial grant-free uplink transmission can the authors of [14]–[25] discussed the combination of NOMA
be realized in NOMA, which is capable of significantly with MIMO techniques. Particularly, the capacity comparison
reducing both the transmission latency and the signaling between MIMO-NOMA and MIMO-OMA has been inves-
overhead. Note that in some NOMA schemes using SIC tigated in [18] [19], and the superiority of MIMO-NOMA
receivers, the SIC process may impose extra latency. over MIMO-OMA in terms of both sum channel capacity and
Therefore, the number of users relying on SIC should ergodic sum capacity was proved analytically. Furthermore,
not be excessive, and advanced MIMO techniques can be in [20] [21], the potential gains of MIMO-NOMA were
invoked for serving more users, as discussed in Section shown based on both link-level as well as on system-level
III. simulations and using the NOMA test-bed developed, where
• Relaxed channel feedback: The requirement of channel a hardware SIC receiver was used taking into account the
feedback will be relaxed in power-domain NOMA, be- realistic hardware impairments quantified in terms of the error
cause the CSI feedback is only used for power allocation. vector magnitude (EVM), the number of quantization bits in
Hence there is no need for accurate instantaneous CSI the analog/digital (A/D) converter, etc. The simulation results
knowledge. Therefore, regardless whether fixed or mobile and the measurements obtained showed that under multiple
users are supported, having a limited-accuracy outdated configurations, the cell throughput achieved by NOMA is
channel feedback associated with a certain maximum in- about 30% higher compared to OFDMA. Furthermore, some
accuracy and delay will not severely impair the attainable open implementation issues were also discussed in [20] [21],
system performance, as long as the channel does not including the granularity of the multi-user power allocation
change rapidly. both in time and frequency, as well as the signaling overhead,
Given the above prominent advantages, NOMA has been feedback enhancements and receiver design. Additionally, the
actively investigated, with a views for employment in 5G as receiver design was discussed in [26]–[28]. A novel NOMA
a promising solution. In the next section, we will discuss and transmitter and receiver design was proposed in [26], where
compare the dominant NOMA solutions. the signals of multiple users are jointly modulated at the
transmitter side and detected at the receiver side. In this
III. D OMINANT NOMA S OLUTIONS scheme, the desired signal of the cell center user can be
In this section, we will discuss the families of promi- directly detected without detecting the signal of the cell edge
nant NOMA schemes by dividing them into two categories, user, i.e., without SIC processing. Thus, a low complexity is
namely power-domain and code-domain NOMA. Their design achieved. Furthermore, the associated simulation results have
principles and key features will be highlighted, respectively. shown that compared to the ideal SIC, the downlink NOMA
We will also provide their comparison in terms of their link-level performance depends both on the actual receiver
spectral efficiency, system performance, receiver complexity, design and on the difference in the power ratio split between
etc. At the end of this section, performance evaluations and the cell edge user and cell center user. Besides, the design
transmission experiments of NOMA will be discussed. and performance of the SIC receiver for downlink NOMA
combined with 2-by-2 open-loop SU-MIMO based on LTE
A. Power-Domain NOMA TM3 (Transmission mode 3) were investigated in [27], where
In this subsection, we will discuss the first category of NO- different receiver weight generation schemes were introduced
MA, namely, power-domain NOMA. In [14]–[17], the concept both before SIC and after SIC according to the transmission

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

rank combination between the users. The link-level simulation SIC of User 2 User 1 signal

Power
signal detection
results showed that the codeword level SIC achieves higher User 2 User 1
performance than the symbol level SIC and in fact approaches User 1 User 2 signal
the performance of ideal SIC. The impact of applying the SIC Frequency detection
receiver for cell-edge users in downlink NOMA using SU- User 2
(a)
MIMO was investigated in [28]. The simulation results showed
that there is an improvement of the NOMA gains over OMA
in conjunction with SIC processing for the cell-edge users. User 1 Beam 1
Furthermore, in order to increase the attainable performance
of the SIC receiver, cooperative NOMA transmission has been User 2
User 5
proposed in [29]–[31]. A range of investigations related to User 4
Beam 2
multi-cell NOMA schemes were carried out in [32]–[34]. User 3
Moreover, since having an increased number of cell-edge users
(b)
typically degrades the efficiency of coordinated multi-point
(CoMP) transmissions, this limitation was circumvented by BS 1 BS 2
User 1
promising NOMA solutions proposed for a CoMP system User 3
Cell 2
Cell 1
in [35] [36]. Additionally, the performance of NOMA tech- User 2 User 4
niques supporting randomly distributed users was evaluated NOMA user pair NOMA user pair
in [37]. These simulation results demonstrated that the outage (c)
performance of NOMA substantially depended both on the
users’ target data rates and on their allocated power. In [38]– Fig. 5. Power-domain NOMA: (a) Basic NOMA relying on a SIC receiver;
[47], the system-level performance of power-domain NOMA (b)
. NOMA in MIMO systems; (c) NOMA in CoMP
was evaluated and the associated simulation results showed
that both the overall cell throughput and the cell-edge user
by different power-scaling coefficients and then they are
throughput, as well as the degree of proportional rate-fairness
superimposed at the BS according to:
of NOMA were superior to those of OMA. Furthermore,
the impact of the residual interference imposed by realistic ∑
K

imperfect channel estimation on the achievable throughput x= pi xi , (4)
performance was investigated in [48]–[50]. On the one hand, i=1

the channel estimation error results in residual interference where E[|xi |2 ] = 1 (i = 1, 2, · · · , K) denotes the normalized
in the SIC process, which hence reduces the achievable user ∑
K
power of the user signals, and we have P = pi . The
throughput. On the other hand, the channel estimation error i=1
causes error in the transmission rate control for the respective received signal yi of the ith user is
users, which may result in decoding errors not only at the
yi = hi x + vi , (5)
destination user terminal but also at other user terminals
owing to the error propagation imposed by the SIC process. where hi denotes the channel gain between the BS and the ith
A simple transmission rate back-off algorithm was considered user, while vi associated with the power density Ni represents
in [48] [49], and the impact of the channel estimation error was the Gaussian noise plus the inter-cell interference.
effectively mitigated. Simulation results showed that NOMA At the receiver, SIC is used for MUD. The optimal order of
achieves beneficial user throughput gains over OMA in a SIC detection relies on detecting the strongest to the weakest
2
scenario subject to channel estimation errors, which is similar user (determined by |hi | /Ni , i = 1, 2, · · · , K). Based on this
to the case associated with perfect channel estimation. optimal SIC-detection order, any user can detect its informa-
Let us now elaborate on the power-domain NOMA tech- tion without substantial interference-contamination imposed
niques in this subsection. Firstly, the basic principle of power- the other users whose normalized channel gain is smaller
domain NOMA relying on a SIC receiver will be discussed. than that of this user. In this way, the user with the strongest
Then, a promising extension relying on integrating NOMA normalized channel gain can cancel interference from all the
with MIMOs will be discussed for the sake of increasing its other users, and thus is least contaminated. It is intuitive
attainable spectral efficiency. Another compelling extension that the users associated with small normalized channel gains
to a cooperative NOMA transmission scheme will also be should be allocated higher power levels in order to improve
presented. Finally, the networking aspects of NOMA solutions their received signal-to-interference and noise ratio (SINR),
will be discussed. so that a high detection reliability can be guaranteed. More
1) Basic NOMA relying on a SIC receiver: Firstly, we particularly, it has been validated in [123] that, to maximize
consider the family of single antenna systems relying on a the sum rate, it is optimal for each user to decode the signals
single BS and K users. of users with poorer normalized channel gains first. Although
In the downlink, the total power allocated to all K users the users having larger normalized channel gains require less
is limited to P , and the BS transmits the signal xi to the power, they are capable of correctly detecting their data with a
ith user subjected to the power-scaling coefficient pi . In other high probability, as a benefit of SIC. Without loss of generality,
2 2 2
words, the signals destined for different users are weighted we assume that |h1 | /N1 ≥ |h2 | /N2 ≥ · · · ≥ |hK | /NK ,

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

and a descending-order-based power allocation p1 ≤ p2 ≤


· · · ≤ pK can be considered. Assuming perfectly error-free

Power
decoding of the interfering signals, the achievable rate of user
i (i = 1, 2, · · · , K) can be written as
  R1 = 3.33bps/HZ
R2 = 0.50bps/HZ
  2 2
| h1 | | h2 |
= 20dB = 0dB
 pi |hi |
2 
Ri = W log  . N1 N2
1 + ( )  (6)
 ∑
i−1
2

Power
Ni W + pj |hi | p1 =
1
P
j=i 5
4
p2 = P
In the case of two users as shown in Fig. 5 (a), we assume 5
that the normalized channel gain of the second user is lower R1 = 4.39bps/HZ
2 2
than that of the first one, i.e., |h1 | /N1 > |h2 | /N2 , and thus R2 = 0.74bps/HZ
p1 < p2 . The second user detects its signal by regarding the
signal of the first user as interference. The first user firstly
Fig. 6. Comparison of NOMA and OMA [14] ⃝IEEE.
c
detects the signal of the second user, and then subtracts its
remodulated version from the received signal, so that the first
user can detect its own signal without interference from the As illustrated in Section II-A, this NOMA scheme is capable
second user. of achieving the maximum attainable multi-user capacity in
Assuming that the transmission bandwidth is normalized to AWGN channels both in the uplink and downlink. Further-
1Hz, the data rates of the first user and the second user can more, this NOMA scheme has the potential of striking a more
be represented as attractive tradeoff between the spectral efficiency and user-
( ) fairness.
2
p1 |h1 |
R1 = log2 1 + , (7) When the number of users is sufficiently high, the SIC-
N1 induced error propagation may have a severe effect on the error
( ) probability in the absence of preventative measures. However,
2
p2 |h2 |
R2 = log2 1 + 2 , (8) some advanced user pairing and power allocation methods,
p1 |h2 | + N2 as well as powerful channel coding schemes can be used for
respectively. Thus, by tuning power allocation coefficients, the reducing the error probability. Indeed, it has been shown that
BS can adjust the data rate of each user. More particularly, it error propagation only has a modest impact on the NOMA
has been shown in Fig. 6 [14] that this NOMA scheme is performance even under the worst-case scenario [14] [39].
capable of achieving higher rates than OFDMA. On the other 2) NOMA in MIMO systems: Although the same time-
hand, this NOMA scheme makes a full use of the natural frequency resources can be shared by multiple users in the
difference of channel gains among the users, which implies basic NOMA employing SIC, the improvement of spectral
that the near-far effect is effectively harnessed to achieve efficiency still remains limited, hence may not satisfy the ex-
higher spectral efficiency. As a result, both the attainable sum pected spectral efficiency improvements of 5G. An appealing
capacity and the cell-edge user data rate can be improved [15]. solution is the extension of the basic NOMA using SIC by
In the uplink, the signal received at the BS is given by amalgamating it with advanced MIMO techniques [16] [17].

K As illustrated in Fig. 5 (b), in downlink NOMA of MIMO

y= hi pi xi + v, (9) systems, MBS BS antennas are used for generating B different
i=1 beams in the spatial domain with the aid of beamforming.
where pi and xi are the transmit power and signal transmitted Within each beam, the signals of multiple users may be trans-
by the i-th user, respectively. Furthermore, v associated with mitted by superimposing them, hence leading to the concept
the power density N0 represents the Gaussian noise plus the of intra-beam superposition modulation, which is similar to
inter-cell interference at the BS. SIC is used for reliable signal the basic NOMA using SIC, as discussed above. The bth
detection at the BS. Without loss of generality, we assume (1 ≤ b ≤ B) transmitter beamforming vector is denoted as
2 2 2 mb . Let us assume that the number of users in the bth beam
that p1 |h1 | ≥ p2 |h2 | ≥ · · · ≥ pK |hK | , and accordingly the
optimal decoding order for SIC is x1 , x2 , · · · , xK . Before the is kb , the transmitted symbol of the ith user in the bth beam is
BS detects the i-th user’s signal, it decodes the j-th (j < i) xb,i , and the corresponding power-scaling coefficient is pb,i .
user’s signal first and then removes (i − 1) users’ signals from Then, by accumulating all signals of the B different beams,
the observation y. The remaining (K − i) signals are regarded the MBS -dimensional transmitted downlink signal vector at
as interference. As a result, the achievable data rate of the i-th the BS can be formulated as
user becomes ∑ ∑
  B kb

x0 = mb pb,i xb,i . (11)
 pi |hi |
2 
  b=1 i=1
Ri = W log 1 + . (10)
 ∑
K
2  Assuming that each user has Nr receiver antennas, the Nr -
N0 W + |hj | pj
j=i+1 dimensional received signal vector of the ith user encapsulated

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

in the bth beam can be represented as 50

yb,i = Hb,i x0 + vb,i , (12)


40
where Hb,i denotes the channel matrix of size Nr × MBS

Total user rate (Mb/s)


between the BS and the ith user in the bth beam, and vb,i
30
denotes the Gaussian noise plus the inter-cell interference.
At the receiver, a pair of interference cancellation ap-
20
proaches are used for removing the inter-beam interference
and the intra-beam interference, respectively. The inter-beam NOMA−2*2 MIMO with random beamforming
interference can be suppressed by spatial filtering, which is 10 OFDMA−2*2 MIMO with random beamforming
NOMA−1*2 SIMO
similar to the signal detection algorithm of spatial division OFDMA−1*2 SIMO
multiple access (SDMA) systems. Assuming that the spatial 0
0 10 20 30 40 50
filtering vector of the ith user in the bth beam is fb,i , the signal Number of users per cell
zb,i after spatial filtering can be represented as
H Fig. 7. System-level performance for NOMA applying opportunistic random
zb,i =fb,i yb,i
beamforming in downlink [15] ⃝IEEE.
c

kb

H
=fb,i Hb,i mb pb,j xb,j
j=1 antennas, the number of orthogonal downlink reference signals

B ∑
kb (13) is not increased beyond the number of transmit antennas.
H √
+ fb,i Hb,i mb′ pb′ ,j xb′ ,j Obviously, we can see from (14) and (15) that the design
b′ =1 j=1 of beamforming vectors {mb } and spatial filtering vectors
b′ ̸=b
{fb,i } is crucial for interference cancellation, which should be
H
+ fb,i vb,i . carefully considered in MIMO NOMA systems. Particularly,
By normalizing the aggregated power of the inter-beam inter- the beamforming and spatial filtering optimization problem
ference and the receiver noise plus inter-cell interference to is usually formulated as the maximization of the sum rate
unity, we can rewrite (13) as problem for perfect CSI scenarios and is formulated as the
minimization of the maximum outage probability (MMOP)
√ ∑√ kb
problem for imperfect CSI scenarios [124]. For the perfect CSI
zb,i = ab,i pb,j xb,j + qb,i , (14)
scenarios, an iterative minorization-maximization algorithm
j=1
(MMA) was proposed for beamforming design [125], where
where qb,i is the normalized term representing the sum of a second-order cone program (SOCP) convex problem should
the inter-beam interference and receiver noise plus inter-cell be solved in each iteration. In addition, the duality scheme
interference, while ab,i is formulated as has been introduced in [126], which can be regarded as the
quasi-degraded solution [127] for the sum rate maximization
|fb,i Hb,i mb |2
ab,i =  B k . (15) problem, and a quadratic constrained quadratic programs (QC-

 ∑ ∑b 

 pb′ ,j |fb,i Hb,i mb′ |2  QP) convex problem should be solved. Furthermore, to real-
b′ =1 j=1 ize low-complexity beamforming, a multiple-user CSI-based

 b′ ̸=b 

 +fH H  singular value decomposition (MU-CSI-SVD) algorithm was
b,i E[vb,i vb,i ]fb,i proposed [124] to solve the sum rate maximization problem
After spatial filtering, the system model (14) becomes in perfect CSI scenarios. What’s more, this algorithm can
similar to that of the basic NOMA combined with SIC, as be easily used to simplify the outage probability expressions
described above. Therefore, the inter-beam interference can be in imperfect CSI scenarios. Therefore, for imperfect CSI
suppressed, and then intra-beam SIC is invoked for removing scenarios, the MU-CSI-SVD algorithm can be also used with
the inter-user interference imposed by superposition coding low complexity.
within a beam. In the uplink systems associated with K users, where the
Naturally, more users can also be simultaneously supported, BS is equipped with MBS antennas and each user has a
because more than two users can share a single beamforming single transmitter antenna, the signal received at the BS can
vector. To elaborate a little further, observe in Fig. 7 [15] that be represented as
both the basic NOMA combined with SIC and the extended
NOMA relying on MIMO are capable of achieving a higher ∑
K

y= hi pi xi + v, (16)
sum-rate than OFDMA. Furthermore, the sum-rate of NOMA
i=1
in the context of MIMO systems is higher than that of the basic
NOMA using a single antenna at the BS. Additionally, in this where hi is the MBS -dimensional channel vector between user
NOMA scheme, the number of reference signals required is i and the BS. Furthermore, pi and xi are the transmit power
equal to the number of transmitter antennas, regardless of the and the signal transmitted by the i-th user, respectively, while
number of non-orthogonal user-signals. In this way, when the v represents the Gaussian noise plus the inter-cell interference
number of users is increased beyond the number of transmitter vector.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

10

At the receiver, MMSE-SIC receiver can be used to realize To elaborate a little further, in practice the participation of
signal detection [5]. It has been shown in [5] that for any all users in the cooperative NOMA cannot be readily realized
decoding order, the sum throughput for all K users is equal due to the extra requirement of short-range communication
to the maximum of the total user throughput given the received resources as well as owing to the complex signal processing as-
signal vector in (16). In practice, the decoding order can be sociated with a high signaling overhead. A promising solution
adjusted according to the actual requirements, such as user to this problem is to reduce the number of cooperating users.
fairness. Without loss of generality, we consider the appealingly simple
3) Cooperative NOMA: Recently, a cooperative NOMA case of having only two cooperative users as an example. Let
transmission scheme was proposed in [29]. Similar to the us assume that the users are sorted in the order of improving
basic NOMA, cooperative NOMA also uses a SIC receiver for channel qualities and that the mth and nth users are paired
detecting the multi-user signal. Therefore, the users associated together, where we have m < n. It has been shown in [29]
with better channel conditions can be relied upon as relays in that the worst choice of m and n is n = m + 1, while the
order to improve the reception reliability of the users suffering optimal choice is to group two users experiencing significantly
from poor channel conditions. For cooperative transmission, different channel qualities.
for example short-range communication techniques - such as Again, the cooperative NOMA further exploits the specific
Bluetooth and ultra-wideband (UWB) schemes - can be used feature that the users having better channel qualities have the
for delivering signals from the users benefitting from better knowledge of other users’ signals, whose channel qualities are
channel conditions to the users with poor channel conditions, poor. In this way, the maximum diversity gain can be achieved
which is the key difference with respect to the basic NOMA for all users by transmitting the signals of those specific users
associated with SIC. who have better channel qualities to other users.
Without loss of generality, let us now consider a downlink 4) NOMA in CoMP: NOMA also exhibits its own benefits
cooperative NOMA system relying on a single BS for sup- in multi-cell applications, leading to the concept of NOMA
porting K users, where the K users are ordered based on in CoMP. However, by directly applying a single-cell NOMA
their channel qualities, with the first user having the worst design to multi-cell scenarios, NOMA in CoMP may result
channel condition, while the Kth user having the best channel in severe inter-cell interference. As an example, a downlink
condition. Cooperative NOMA relies on the following two cellular system having two cells and four users is depicted in
phases [29]. Fig. 5 (c), where a two-user NOMA scheme is considered,
The first phase, also termed the broadcast phase, represents with user 1 and user 2 being served by BS 1, while user 3
the direct transmission. In this phase, the BS sends downlink and user 4 being served by BS 2. However, at user 1, strong
messages to all the K users based on the principle of basic interference may be imposed by the signals transmitted from
NOMA relying on SIC, where the superimposed information BS 2 (similarly, user 3 also suffers from severe interference
of the K users obeys the total power constraint. The SIC caused by the signals transmitted from BS 1), which poten-
process is implemented at the user side. As a result, the tially leads to a significant performance degradation of the
users having better channel conditions have the knowledge NOMA in CoMP scheme.
of the signals intended to the users having poor channel To mitigate the inter-cell interference in the downlink,
conditions [128]. joint transmit-precoding of all NOMA users’ signals can be
The second phase represents the cooperative transmission. utilized. However, all users’ data and channel information
During this phase, the cooperating users transmit their signals should be available at the BSs involved, and finding the
via their short range communication channels, such as Blue- optimal transmit-precoder is not trivial. Moreover, multi-user
tooth or UWB. Particularly, the second phase includes (K −1) transmit precoding of single-cell NOMA may not be efficient
time slots. In the first time slot, the Kth user broadcasts the in a NOMA in CoMP setting, since a beam generated via
superposition of the (K −1) signals destined for the remaining geographically separated BS antennas may not be capable of
users. Then the SIC process is invoked again at these (K − 1) covering more than one angularly separated user for intra-
users. The (K −1)th user combines the signals received during beam NOMA. By exploiting that the CIRs of different users
both phases by using maximum ratio combining (MRC), and are likely to be rather different in the multi-cell scenario, a
it detects its own information at a higher SNR than that of reduced-complexity transmit precoding scheme was proposed
the traditional SIC. Similarly, in the kth time slot, where for NOMA in CoMP [32], where the precoder is applied only
1 ≤ k ≤ K − 1, the (K − k + 1)st user also broadcasts to the signals of the cell edge users, such as users 1 and 3
the (K − k) superposed signals for the remaining (K − k) of Fig. 5 (c). Additionally, a multi-cell uplink NOMA system
users, whose channel conditions are worse than that of this has been considered in [33], and the rate coverage probability
user. Then the (K − k + 1)st user combines the observations (the probability that a given user’s achievable rate remains
gleaned from both phases and it detects its own information above the target data rate) of a user who is at rank m (in
at a higher SNR than the traditional SIC. Therefore the terms of the distance from its serving BS) among all users in
emplyment of cooperation is indeed capable of enhancing a cell and the mean rate coverage probability of all users in
the reception reliability. Note that it can only be invoked at a cell were analyzed using the theory of order statistics and
low user loads, because the above-mentioned regime requires poisson cluster process. It has been shown that the average
potentially excessive resources for cooperation. This might rate coverage of a NOMA cluster is better than that of its
reduce the extra gain of NOMA. counterpart OMA cluster for higher number of users per cell

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

11

SNRm = 0 dB
B. Code-Domain NOMA
LDM
0.8 The NOMA schemes discussed in the previous subsection
Capacity of moble service (b/s/HZ)

TDM/FDM
0.7 realize multiplexing in the power domain. By contrast, in
this subsection, we introduce the other main category of
0.6
NOMA schemes, which achieves multiplexing in the code
0.5 domain. The concept of code-domain NOMA is inspired by
0.4 the classic CDMA systems, in which multiple users share the
same time-frequency resources, but adopt unique user-specific
0.3
spreading sequences. However, the key difference compared to
0.2 SNRf = 0 dB CDMA is that the spreading sequences are restricted to sparse
SNRf = 20 dB SNRf = 30 dB
0.1 SNRf = 10 dB sequences or non-orthogonal low cross-correlation sequences
in NOMA. In this subsection, we first present the initial form
0
0 2 4 6 8 10 of NOMA based on sparse spreading sequences, i.e., LDS-
Capacity of fixed service (b/s/HZ)
CDMA [66]–[69]. Then, the family of LDS aided multi-carrier
OFDM systems (LDS-OFDM) [70]–[75] will be discussed,
Fig. 8. Channel capacity advantage of LDM vs. TDM/FDM [129] ⃝IEEE.
c which retains all the benefits of OFDM-based multi-carrier
transmissions in terms of its ISI avoidance, together with
MUD-assisted LDS-CDMA operating at a lower complexity
and for higher target rate requirements. Furthermore, an up-to- than that of the optimal maximum a posteriori probability
date literature review of interference management techniques (MAP) detector. Another important extension of LDS-CDMA
that apply NOMA in multi-cell networks has been provided in is SCMA [76]–[97], which still enjoys the benefit of low-
[34], including both NOMA using joint processing and NOMA complexity reception, but has a better performance than LDS-
relying on coordinated scheduling/beamforming. The major CDMA. A suite of other improved schemes and special forms
practical issues and challenges that arise in the implementation of CDMA, such as MUSA [98] and SAMA [99] will also be
of multi-cell NOMA have also been highlighted in [34], such discussed in this subsection.
as the SIC implementation issues, imperfect CSI, as well as
1) Low-density spreading CDMA (LDS-CDMA): Develope-
multi-user power allocation and clustering.
d from the classic concept of CDMA, LDS-CDMA is designed
5) Application of power-domain NOMA: Recently, the for limiting the amount of interference imposed on each chip
concept of power-domain has been successfully applied to of conventional CDMA systems by using LDS instead of
ATSC 3.0 [129], which is a new next-generation broadcasting conventional spreading sequences. The basic principle of LDS-
standard in US, and this physical-layer non-orthogonal mul- CDMA has been discussed in [66]–[68]. Additionally, [66]
tiplexing technology is named layered-division-multiplexing and [67] also discussed the iterative MUD based on the mes-
(LDM). sage passing algorithm (MPA) imposing a lower complexity
Specifically, a two-layer LDM structure consisting of the than that of the optimal MAP detector. Specifically, in [66], the
upper layer (UL) and the lower layer (LL) is accepted by performance of LDS-CDMA communicating over memoryless
ATSC 3.0 to improve spectral efficiency and provide more Gaussian channels using BPSK modulation was analyzed.
versatile broadcasting services. The UL with higher power The simulation results showed that the performance of LDS-
allocation is used to deliver mobile services to indoor, portable CDMA is capable of approaching the single-user performance
and handheld receivers, while the LL is designed to deliver for a normalized user-load as high as 200%. However, the
high data rate services, such as UHDTV or multiple HDTV performance of LDS-CDMA operating in multipath fading
services to fixed reception terminals, where the operational channels is still under investigation at the time of writing. The
SNR is usually high due to the large and possibly directional challenge is that the multipath fading channels will destroy
receive antennas [129]. At the transmitter, the data of each the original LDS structure. On the other hand, a structured
layer is firstly processed by its own physical-layer signal approach of designing LDS codes for LDS-CDMA has been
processing modules, including channel encoding, interleaving, proposed in [68], where the basic idea is to map the signature
modulation, etc., and then the signals from all layers are constellation elements to the spreading matrix hosting the
superimposed over the same time-frequency resources. At the spreading sequences. Furthermore, the capacity region of LDS-
receivers, to decode the UL signal, the lower-power LL service CDMA was calculated using information theoretic analysis
is treated as an additional interference. To decode the LL in [69], and the accompanying simulation results showed how
signal, the receiver firstly needs to cancel the UL signal, which the attainable capacity depended on the spreading sequence
is referred as SIC procedure in power-domain NOMA. density factor as well as on the maximum number of users as-
As shown in Fig. 8 from [129], the channel capacity for sociated with each chip, which provided insightful theoretical
the mobile and fixed services of the LDM and TDM/FDM guidelines for practical LDS system designs.
systems is compared. It is observed that LDM offers better Let us now consider a classic synchronous CDMA system
performance than TDM/FDM in all scenarios, and the higher operating in the uplink and supporting K users with the aid
the SNR threshold of the fixed service, the larger the advantage of Nc chips (Nc equals to the number of observations at
of the LDM systems. the receiver). The transmitted symbol xk of user k is firstly

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

12

generated by mapping a sequence of independent information


bits to a constellation alphabet, i.e., xk is taken from a
complex-valued constellation set X. Then, the transmitted x1 sn,k ≠ 0
symbol xk is mapped to a spreading sequence sk , such as the y1
x2
set of widely used PN sequences, which is unique for each
user. The signal received during chip n can be represented by y2
x3


K
x4 y3
yn = gn,k sn,k xk + wn , (17)
k=1 x5
y4
where sn,k is the nth component of the spreading sequence x6
sk , gn,k is the channel gain of user k on chip n, and
wn is a complex-valued Gaussian noise sample with a zero
mean and a variance of σ 2 . When we combine the signals Fig. 9. Illustration of LDS-CDMA: 6 users only employ 4 chips for
transmission, which implies that a normalized user-load of 150% can be
received during all the Nc chips, the received signal vector achieved.
T
y = [y1 , y2 , · · · , yNc ] is formulated as
rewritten as
y = Hx + w, (18) ∑ ∑
yn = gn,k sn,k xk + wn = hn,k xk + wn , (19)
T
where x = [x1 , x2 , · · · , xK ] , H is the channel matrix k∈N (n) k∈N (n)
of size (Nc × K), and the element hn,k in the nth row
where N (n) denotes the set of users whose sparse spreading
and the kth column of H is denoted by ( gn,k)sn,k . Finally,
T sequences have a non-zero element at chip n, namely, N (n) =
w = [w1 , w2 , · · · , wNc ] , and w ∼ CN 0, σ 2 I .
{k|sn,k ̸= 0}.
In classic CDMA systems, the elements of the spreading At the receiver, MUD based on message passing algo-
sequences sk (k = 1, · · · , K) are usually non-zero, i.e., the rithm may be performed. Given the joint probability func-
spreading sequences are not sparse. Consequently, the signals tion p (x1, x2 , . . . , xE ) for random variables x1, x2 , . . . , xE ,
received from all the active users are overlaid on top of each the message-passing algorithm is capable of simplifying the
other at each chip, and every user will be subjected to inter- calculation of the marginal probability distribution for each
user interference imposed by all the other users. If the spread- variable as follows:
ing sequences are orthogonal, it is straightforward to eliminate ∑
the interferences, hence the information of all users can be p (xe ) = p (x1, x2 , . . . , xE ) , (20)
accurately detected by a low-complexity correlation receiver. ∼{xe }
However, the classical orthogonal spreading sequences can
where ∼ {xe } represents all variables except for xe . We
only support as many users as the number of chips. By con-
assume that the joint probability function can be decomposed
trast, the above-mentioned PN-sequence family has many more
into the product of some positive functions, namely,
codes than the number of chips in a sequence, but since the
1 ∏
codes are non-orthogonal, they impose interference even in the D

absence of non-dispersive channels. Hence they require more p (x1, x2 , . . . , xE ) = fd (Xd ) , (21)
Z
complex MUDs. Another natural idea, which leads to LDS- d=1

CDMA, is to use sparse spreading sequences instead of the where Z is a normalized constant, Xd is a subset of
classic “fully-populated” spreading sequences to support more {x1, x2 , . . . , xE }, and f1 (X1 ) , f2 (X2 ) , . . . , fD (XD ) are
users, where the number of non-zero elements in the spreading positive functions which are not necessarily the probability
sequence is much lower than Nc for the sake of reducing the functions. Then, we can translate this form into the factor
interference imposed on each chip. Therefore, LDS-CDMA graph, which is a bipartite graph, as shown in Fig. 10 [130],
is potentially capable of improving the attainable system where the circles represent variable nodes corresponding to
performance by using low-density spreading sequences [66], x1, x2 , . . . , xE , while the squares indicate observation nodes
which is the key distinguishing feature between conventional corresponding to f1 (X1 ) , f2 (X2 ) , . . . , fD (XD ). An edge is
CDMA and LDS-CDMA. present between a variable node xe and an observation node
In LDS-CDMA, all transmitted symbols are modulated onto fd (Xd ) if and only if xe ∈ Xd .
sparse spreading sequences. In this way, each user will only In general, the message passing algorithm relies on the
spread its data over a small number of chips, as shown in factor graph representation of the problem as its input and re-
Fig. 9. As a result, the number of the superimposed signals at turns the marginal distribution of all variable nodes. Messages
each chip will be less than the number of active users, which can be passed between the variable node and the observation
means that the interference imposed on each chip will be effi- node through the edge between them, and the message can be
ciently reduced, hence mitigating the multi-user intereference interpreted as the soft-value that represents the reliability of the
by carefully designing the spreading sequences. Therefore, variable associated with each edge. The marginal distribution
the received signal at chip n in LDS-CDMA systems can be of a variable node can be interpreted as a function of the

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

13

x1 x2 x3 x4 x5 x6
Variable
nodes

Observation
nodes
f1 ( X 1 ) f2 ( X 2 ) f3 ( X 3 ) f4 ( X 4 )

Fig. 10. Factor graph representation of MPA.

messages received by that variable node. The iterative form of


the message passing algorithm can be represented as
(t)
∑ ∏ (t−1)
md→e (xe ) ∝ fd (Xd ) mi→d (xi ) , Fig. 11. Comparison of LDS-CDMA and DS-CDMA [66] ⃝IEEE.
c
{xi |i∈N (d)\e} i∈N (d)\e
∏ (22)
(t) (t−1) where
me→d (xe ) ∝ mi→e (xe ) , (23)  
i∈N (e)\d
( ) 1  1 ∑ 
p yn |x[n] = √ exp − 2 (yn − hn,k xk )2 .
(t)
where md→e (xe ) denotes the message transmitted from the 2πσ  2σ 
k∈N (n)
observation node fd (Xd ) to the variable node xe at the tth (27)
(t)
iteration. Similarly, me→d (xe ) presents the message trans- Observe that (26) represents a marginal process of
mitted from the variable node xe to the observation node ∏c
N
p(yn |x[n] ), which is similar to the form of the decom-
fd (Xd ). If the maximum number of iterations is T , the n=1
marginal probability distribution for each variable can be posable joint probability function of the message passing
finally calculated as algorithm, apart from a normalization constant Z. To elaborate
∏ (T )
a little further, we can regard each term p(yn |x[n] ) as a positive
p (xe ) ∝ md→e (xe ) . (24) function fd (Xd ) in the message passing algorithm. Then, the
d∈N (e) factor graph shown in Fig. 9 can be constructed just like that of
It has been theoretically shown that the marginal distribu- Fig. 10. Therefore, we can rewrite the iterative Equations (22)
tion can be accurately estimated with the aid of a limited and (23) as follows:
number of iterations, provided that the factor graph does (t)
∑ 1 1
not have loops [130]. However, in many practical situations mn→k (xk ) ∝ √ exp{− 2 (yn − hn,k xk
2πσ 2σ
the presence of loops cannot be avoided. Fortunately, the {xi |i∈N (n)\k}
∑ ∏ (t−1)
message passing algorithm is quite accurate for “locally tree − hn,i xi )2 } mi→n (xi ) ,
like” graphs, which implies that the length of the shortest i∈N (n)\k i∈N (n)\k
loop is restricted to O (log (E)). Therefore, in most practical ∏
(28)
(t) (t−1)
applications associated with a sparse structure, we can obtain mk→n (xk ) ∝ mi→k (xk ) . (29)
an accurate merginal distribution estimate by an appropriate i∈N (k)\n
design of the factor graph.
In LDS-CDMA, the optimum MAP detection of x in (18) Finally, the (approximate) marginal probability distribution
can be formulated as: of each variable after T iterations can be calculated by (24).
∑ In the case of LDS, the number of edges in the factor graph
x̂k = arg max p(x|y). (25) is relatively low, hence less and longer loops can be expected
a∈X
∼ {xk } based on a meritorious design of the factor graph based on
xk = a a beneficial sparse spreading sequence design. Additionally,
assuming that the maximum number of users superimposed at
Without loss of generality, we assume that the transmitted the same chip is w, the receiver
( ) complexity is on the order
symbols and noise are identically and independently dis- w
of O (|X| ) instead of O |X|
K
(K > w) for conventional
tributed (i.i.d), and the transmitted symbols obey the uniform
CDMA.
distribution. Then according to Bayes’ rule, (25) can be
The performance of LDS-CDMA and direct sequence-
reformulated as
CDMA (DS-CDMA), which is adopted by the 3G WCDMA
∑ ∏
Nc
systems, have been compared in [66]. More specifically, as
x̂k = arg max p(yn |x[n] ), (26) shown in Fig. 11 [66], LDS-CDMA outperforms DS-CDMA
a∈X
∼ {xk } n=1
using the best-found spreading sequences, where the MMSE-
xk = a based partial parallel interference cancellation (PPIC) receiver

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

14

Codebook 1 Codebook 2 Codebook 3 Codebook 4 Codebook 5 Codebook 6


has been adopted by both schemes. Furthermore, when LDS- Bit streams
are mapped
CDMA relies on an MPA receivers, its performance approach- to sparse
es the single user bound within a small margin of 1.17 dB at codewords
a BER of 10−4 . (0,0) (1,0) (0,1) (1,1) (1,1) (0,0)

2) Low-density spreading aided OFDM (LDS-OFDM):


OFDM and MC-CDMA are close relatives, especially when 6 sparse codewords MUD
considering frequency-domain-spreading, which spreads each are transmitted over 4 based on
user’s symbols across all the OFDM subcarriers, provided that orthogonal resources MPA
the number of spreading-code chips is identical to the number
of subcarriers. Then multiple users may be supported by Fig. 12. SCMA encoding and multiplexing.
overlaying the unique, user-specific spreading sequences of all
users on top of each other across all subcarriers. As always, the
spreading sequences may be chosen to be orthogonal Walsh- Furthermore, in order to improve the achievable performance
Hadamard codes or non-orthogonal m-sequences, as well as of LDS-OFDM, a joint subcarrier and power allocation method
LDSs, for example. was proposed in [74], with the objective of maximizing the
Hence LDS-OFDM can be interpreted as an integrated weighted sum-rate using an efficient greedy algorithm. As a
version of LDS-CDMA and OFDM, where for example each further result, a pair of PAPR reduction techniques have been
user’s symbol is spread across a carefully selected number of proposed for LDS-OFDM in [75].
subcarriers and overlaid on top of each other in the frequency- 3) Sparse code multiple access (SCMA): The recently pro-
domain. To elaborate a little further, in the conventional OFD- posed SCMA technique constitutes another important NOMA
MA system, only a single symbol is mapped to a subcarrier, scheme, which relies on code domain multiplexing developed
and different symbols are transmitted on different subcarriers, from the basic LDS-CDMA scheme. In [76], SCMA was
which are orthogonal and hence do not interfere with each extensively discussed in terms of its transmission and multi-
other. Therefore, the total number of transmitted symbols plexing aspects, as well as in terms of its factor graph represen-
is restricted by the number of orthogonal subcarriers. By tation and receiver architecture relying on the message passing
contrast, in the LDS-OFDM system, the transmitted symbols algorithm. In contrast to the basic LDS-CDMA, as illustrated
are firstly multiplied with LDS sequences, whose length is in Fig. 12, the bit-to-constellation mapping and spreading
equal to the number of subcarriers and the resultant chips operations in SCMA are intrinsically amalgamated, hence the
are transmitted on different subcarriers. When using LDS original bit streams are directly mapped to different sparse
spreading sequences, each original symbol is only spread to a codewords, where each user has its own codebook. Without
specific fraction of the subcarriers. As a result, each subcarrier loss of generality, we assume that there are J codebooks,
carries chips related to a fraction of the original symbols. where each codebook contains M codewords of length Kl , and
Suffice to say that apart from the already accentuated bene- the number of non-zero elements in every codeword is Nnz .
fits, frequency-domain spreading is particularly advantageous For example, in Fig. 12 we have J = 6, M = 4, Kl = 4, and
in strongly frequency-selective channels, which would often Nnz = 2. We consider the rank-deficient scenario of Kl < J,
obliterate some of the subcarriers and their information, whilst which is capable of supporting the massive connectivity ex-
in the presence of frequency-domain spreading they would pected in 5G. All codewords in the same codebook contain
only affect some of the chips conveying the original symbols. zeros in the same (Kl − Nnz ) dimensions, and the positions
This is likely to allow us to still recover the original symbols. of zeros in the different codebooks are unique and distinct for
We note in closing that the family of MUDs designed using the sake of facilitating collision avoidance for any pair of users.
the message passing algorithm for LDS-CDMA can also be Therefore, the maximum number of codebooks ( ) by
is restricted
used for LDS-OFDM in order to separate the overlaid symbols Kl
the selection of Kl and Nnz , which is equal to . For
at the receiver. Nnz
At the time of writing a number of insightful LDS-OFDM each user, log2 M bits are mapped to a complex codeword. The
investigations have already been disseminated in the literature. codewords of all users are then multiplexed onto Kl shared
For example, the system model and properties of LDS-OFDM, orthogonal resources, such as the OFDM subcarriers. Due to
including its frequency diversity order, receiver complexity, the sparsity of codewords, the signal received on subcarrier k
and its ability to operate under rank-deficient conditions in can be represented by
the presence of more users than chips have been presented ∑
yk = hkj xkj + wk , (30)
in [70]. An upper limit was imposed on the number of users
j∈N (k)
per subcarrier, in order to control the receiver complexity
in [71]. Additionally, in [72] [73], the performance of LDS- where xkj is the kth component of the codeword xj for user
OFDM was compared to both SC-FDMA and OFDMA in j, hkj is the channel gain of user j at the kth subcarrier, and
terms of the peak-to-average-power ratio (PAPR), as well as wk denotes the complex-valued Gaussian noise with zero mean
the link-level and system-level performance. The associated and variance σ 2 . Similar to LDS-CDMA, the message passing
simulation results showed that LDS-OFDM is capable of algorithm can also be used for MUD at the SCMA receiver.
significantly improving the attainable system performance at However, the receiver complexity may become excessive. To
a given transmission power, spectral efficiency and fairness. circumvent this problem, improved variants of the message

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

15

0
10 f

−1
10
BLER

−2
10

LDS−QPSK, code rate 3/4,


t
6 signatures
SCMA−4 point constellation,
code rate 3/4, 6 layers
−3
10 CB1 CB2 ... CBJ
2 4 6 8 10 12 14 16 18 20
SNR(dB) P1 PL+1 ... PL( J −1)+1
. . . .
. . . .
. . . .
Fig. 13. Performance comparison between SCMA and LDS [76] ⃝IEEE.
c PL P2 L ... PLJ

passing algorithm have been proposed in [77]–[85]. Specifi-


cally, a low-complexity logarithmic-domain message passing Fig. 14. Definition of a contention transmission unit (CTU) [92] ⃝IEEE.
c
algorithm (Log-MPA) was proposed in [86]. The associated
simulation results showed that the performance degradation
of Log-MPA over the full-complexity MPA was negligible As a further solution, a multi-stage design approach has
in practical applications, despite the fact that the Log-MPA been proposed for finding a meritorious sub-optimal solution
achieved over 50% complexity reduction. For Log-MPA, the in [76]. More details concerning the codebook design can be
conditional channel probability calculation imposes up 60% found in [89].
of the total computational complexity in the whole decoding Specifically, in order to simplify the optimization prob-
procedure. In [77], a dynamic search algorithm based on lem of the multi-dimensional constellation design, a mother-
classic signal uncertainty theory was proposed for eliminating constellation can be generated first by minimizing the average
any unnecessary conditional channel probability calculation alphabet energy for a given minimum Euclidian distance
without degrading the decoding performance. On the other between any two constellation points. More particularly, an op-
hand, in order to improve the BER performance of SCMA, the timized design of the mother-constellation based on the classic
powerful turbo-principle has been invoked in [87] for exchang- star-QAM signaling constellation has been proposed in [90].
ing extrinsic information between the SCMA detector and the The resultant simulation results showed that the star-QAM
channel decoder. By contrast, in [88], a low-complexity turbo- based codebooks are capable of significantly enhancing the
like combination of iterative detection and iterative decoding BER performance of the square-QAM based codebooks. Once
was conceived for striking a compelling performance versus the mother-constellation has been obtained, the codebook-
complexity balance. specific operation can be applied to the mother-constellation in
Quantitatively, the performance of SCMA relying on a order to obtain specific constellations for each codebook. More
multi-dimensional constellation having four points was com- specifically, the codebook-specific operations, such as phase
pared to that of LDS using QPSK modulation in Fig. 13 [76]. rotation, complex conjugation and dimensional permutation,
These simulation results show that SCMA outperforms LDS can be optimized for introducing correlation among the non-
in terms of its block error ratio (BLER). The key difference zero elements of the codewords, which is beneficial in terms
between LDS and SCMA is that SCMA relies on multi- of recovering the codewords contaminated by the interference
dimensional constellations for generating its codebooks, which imposed by other tones. Additionally, different power can
results in the so-called “constellation shaping gain” [76]. This be assigned to the symbols superimposed over the same
gain is unavailable for other NOMA schemes. More explicitly, time-frequency index for ensuring that the message passing
the “shaping gain” terminology represents the average symbol algorithm can operate more efficiently by mitigating the inter-
energy gain, when we change the shape of the modulation con- ferences between the paired layers. Furthermore, inspired by
stellation. In general, the shaping gain is higher when the shape the family of irregular low density parity check (LDPC) codes,
of the constellation becomes similar to a sphere. However, the an irregular SCMA structure has been proposed in [91], where
SCMA codebook design is complex [76], since the different the number of non-zero elements of the codewords can be
layers are multiplexed with the aid of different codebooks. different for different users. In this way, users having different
However, the best design criterion to be used for solving the QoS requirements can be simultaneously served.
multi-dimensional constellation problem is unknown at the Again, in the uplink of a SCMA system, grant-free multiple
time of writing. Having said this, it is anticipated that using the access can be realized by carefully assigning the codebooks
powerful semi-analytical tool of extrinsic information transfer- and the pilots to the users based on [92]. As mentioned
function (EXIT) charts for jointly designing the channel code in Section II, a user does not have to send a scheduling
and the constellation would lead to near-capacity performance. request to the BS in the grant-free transmission scheme, thus

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

16

Each user’s symbols are


a significant latency- and signaling overhead-reduction can be spread by a specially
User 1
expected. As shown in Fig. 14, the pre-configured resource to designed sequence
be assigned to the users may be referred to as a contention Spreading
transmission units (CTU). There are J codebooks defined User 1
over a time-frequency resource, and L pilot sequences are User 2

associated with each codebook. The grant-free multiple access Spreading User 2
regime allows contentions to occur, when multiple users are . SIC
.
. .
assigned to the same CTU. The network detects the uplink .
.
.
. .
User n User n
packets by attempting their reception using all possible access
codes assigned to the predefined contention region. Then a Spreading Spread symbols
random back-off procedure can be invoked, when collisions are transmitted on
same orthogonal
occur. At the receiver, blind detection and compressive sensing resources
(CS) techniques can be used for performing joint activity and
data detection, e,g., with the aid of the joint message passing
Fig. 15. Uplink MUSA system.
algorithm (JMPA) of [93].
SCMA can also be used in the downlink in order to improve
the system throughput, leading to the multi-user SCMA (MU-
SCMA) concept [94]. Based on a limited knowledge of the 0
10
Length=4
channel conditions of different users, the BS simply pairs −1
Length=8
10 Length=16
the users, where the transmit power is appropriately shared
among multiple users. This regime is hence reminiscent of the −2

Average BLER
10

NOMA scheme relying on the previously mentioned power- −3


10
domain multiplexing. Compared to MU-MIMO, MU-SCMA
−4
is more robust to channel-quality variations, and indeed, the 10

provision of near-instantaneous CSI feedback is unnecessary −5


10
for this open-loop multiple access scheme [94]. In [95], the
−6
concept of single-cell downlink MU-SCMA is extended to 10
100 150 200 250 300 350 400
Overloading (%)
an open-loop downlink CoMP solution, which was termed
as MU-SCMA-CoMP. In this scheme, the SCMA layers and
transmits power are shared among multiple users within a Fig. 16. The performance of MUSA with different-length spreading se-
quences at different normalized user-loads [98] ⃝IEEE.
c
CoMP cluster. The analysis and simulation results in [95]
demonstrated its robustness both to user mobility and channel
aging. Furthermore, the capacity of downlink Massive MIMO
order to separate the different users’ data according to their
MU-SCMA was analyzed in [96] based on random matrix
channel conditions.
theory and it has been shown that compared to Massive MIMO
MU-OFDMA systems, Massive MIMO MU-SCMA is capable In the downlink of the MUSA system the users are separated
of achieving a higher sum rate. into G groups. In each group, the different users’ symbols
In a nutshell, the efficiency of SCMA has been verified both are weighted by different power-scaling coefficients and then
by simulations and real-time prototyping in [97]. Both the lab they are superimposed. Orthogonal sequences of length G
tests and the field tests demonstrated that SCMA is capable can be used as spreading sequences in order to spread the
of supporting upto three times more users than the number of superimposed symbols from G groups. More specifically,
resource-slots, whilst still maintaining a link-integrity close to the users from the same group employ the same spreading
that of orthogonal transmissions. sequence, while the spreading sequences are orthogonal across
4) Multi-user shared access (MUSA): MUSA is another the different groups. In this way, the inter-group interferences
NOMA scheme relying on code-domain multiplexing, which can be removed at the receiver. Then, SIC can be used for
can be regarded as an improved CDMA-style scheme. carrying out intra-group interference cancellation by exploiting
In the uplink of the MUSA system of Fig. 15, all transmitted the associated power difference.
symbols of a specific user are multiplied with the same In MUSA, the spreading sequences should have low cross-
spreading sequence (Note that different spreading sequences correlation in order to facilitate near-perfect interference can-
can also be used for different symbols of the same user, which cellation at the receiver. The MUSA technique is capable of
results in beneficial interference averaging). Then, all symbols improving the downlink capacity, which is an explicit benefit
after spreading are transmitted over the same time-frequency of the associated SINR difference and SIC. As a further
resources, such as OFDM subcarriers. Without any loss of compelling benefit, MUSA is capable of guaranteeing fairness
generality, we assume that each user transmits a single symbol amongst the multiplexed users without any capacity loss. In
every time, and that there are K users as well as N subcarriers. a nutshell, with the advent of advanced spreading sequences
Rank-deficient scearios can also be supported by MUSA, i.e., and powerful state-of-the-art SIC techniques, substantial gains
K > N , which will impose interference amongst the users. can be obtained by MUSA, even for a normalized user-load
At the receiver, linear processing and SIC are performed in as high as say 300%, which is shown in Fig. 16 [98].

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

17

5) Successive interference cancellation aided multiple ac- destroyed by their convolution with the CIR even in the
cess (SAMA): Let us consider an uplink SAMA system absence of co-channel interference. Hence we end up with
supporting K users with the aid of N orthogonal OFDM a potentially infinite variety of received sequences. This leads
subcarriers, where we have K > N , i.e. when the system to the appealing concept of simply using the unique, user-
is rank-deficient. The system model of SAMA is similar to specific CIRs for distinguishing the users, instead of unique,
that of MUSA, but in SAMA, the non-zero elements of any user-specific spreading sequences. Naturally, when the users
spreading sequence bk for user k are equal to one, and the transmitting in the uplink are close to each other, their CIRs
spreading matrix B = (b1 , b2 , · · · , bK ) is designed based on become quite similar, which aggravates the task of the MUD
the following principles [99]: in separating their signals. The beneficial properties of this
• The number of groups with different number of 1’s in the family of solutions have attracted substantial research efforts,
spreading sequence should be maximized. as detailed in [100]–[113].
• The number of the overlapped spreading sequences which To elaborate a little further, given the potentially infinite va-
have the same number of 1’s should be minimized. riety of CIRs, these sophisticated SDMA systems are capable
Then the maximum number of user supported with the aid of of operating under highly rank-deficient conditions, namely
N orthogonal subcarriers can be calculated as when the number of mobile users transmitting in the uplink is
( ) ( ) ( ) much higher than the number of BS uplink-receiver antennas.
N N N This would avoid the hard-limited user-load of the Walsh-
+ + ··· + = 2N − 1. (31)
1 2 N Hadamard code based CDMA systems, since the system-
For example, spreading matrices for N = 2, K = 3, N = performance would only gracefully decay upon increasing
3, K = 7, and N = 4, K = 15 can be designed as follows: the number of users. The resultant SDMA systems tend to
( ) exhibit a similar performance to their rank-deficient CDMA
1 1 0
B2,3 = , (32) counterparts relying on m-sequences for example.
1 0 1
Since these SDMA systems rely on the CIR for distinguish-
 
1 1 0 1 1 0 0 ing the users, they require accurate CIR estimation, which
B3,7 =  1 1 1 0 0 1 0  , (33) becomes extremely challenging, when the number of users
1 0 1 1 0 0 1 is much higher than the number of BS receiver antennas.
This logically leads to the concept of joint iterative channel
B4,15
  and data estimation, which attracted substantial research in-
1 1 0 1 1 1 1 1 0 0
0 0 1 0 0 terests [104] [110]. These high-end solutions often rely on
 1 1 1 0 1 1 0 0 1 0 
1 0 0 1 0 .powerful non-linear bio-inspired MUDs exchanging their soft-
=
 1

1 1 1 0 0 1 0 1 0 
0 1 0 0 1 information with the channel estimator.
1 0 1 1 1 0 0 1 0 1
1 1 0 0 0 Below we will elaborate further on a variety of powerful
(34)
solutions in a little more detail. In [100] the family of
At the receiver, the message passing algorithm is invoked for
minimum bit error rate (MBER) MUDs was shown to be
separating the signals of the different users. The design objec-
capable of outperforming the classic minimum mean-squared-
tive used for determining the spreading matrices in SAMA is
error (MMSE) MUD in term of the achievable BER owing to
to facilitate convenient interference cancellation [99]. Consider
directly minimizing the BER cost function. In this paper, ge-
B4,15 for example. The spreading sequence of the first user
netic algorithms (GAs) were invoked for finding the optimum
has 4 non-zero elements, hence the resultant diversity order
weight vectors of the MBER MUD in the context of multiple
is 4. Thus the first user’s symbol is the most reliable one.
antenna aided multi-user OFDM. It was shown that the MBER
Therefore, the first user’s symbol can be readily determined
MUD is capable of supporting more users than the number of
in a few iterations, which is beneficial for the convergence
receiver antennas available in highly rank-deficient scenarios.
of the symbol detection process of all the other users having
A novel parallel interference cancelation (PIC) based turbo
lower diversity orders.
space time equalizer (STE) structure was designed in [101] for
multiple antenna assisted uplink receivers. The proposed re-
C. Other NOMA schemes ceiver structure allowed the employment of non-linear type of
Apart from the prominent power-domain NOMA and code- detectors such as the Bayesian decision feedback (DF) assisted
domain NOMA solutions discussed in the previous two sub- turbo STE or the MAP STE, while operating at a moderate
sections, recently a range of alternative NOMA schemes computational complexity. The powerful receivers based on
have also been investigated, which will be discussed in this the proposed structure tend to outperform the linear turbo
subsection. detector benchmarker based on the classic MMSE criterion,
1) Spatial division multiple access (SDMA): SDMA is even if the latter aims for jointly detecting all transmitters’
one of the powerful NOMA schemes, and the philosophy signals. Additionally, the PIC based receiver is also capa-
of SDMA may be deemed to be related to that of classic ble of equalizing non-linear binary pre-coded channels. The
CDMA, based on the following philosophy. Even if orthog- performance difference between the presented algorithms was
onal Walsh-Hadamard spreading sequences are employed for discussed using the powerful semi-analytical tool of extrinsic
distinguishing the users in CDMA systems, when they are information transfer-function (EXIT) charts.
transmitted over dispersive channels, their orthogonality is Wang et al. demonstrated [102] that the iterative exchange

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

18

of extrinsic information between the K-best sphere detector a discrete search space. Hence the capability of the GA, RWB-
(SD) and the channel decoder is appealing, since it is capable S, PSO, and DEA to achieve optimal solutions at an affordable
of achieving a near MAP performance at a moderate complexi- complexity was investigated in this challenging application by
ty. However, the computational complexity imposed by the K- Zhang et al. [104]. Their study demonstrated that the EA-
best SD significantly increases when using a large value of K assisted joint channel estimation and turbo MUD/decoder are
for the sake of maintaining a near-MAP performance in a high- capable of approaching both the Cramer-Rao lower bound
throughput uplink SDMA/OFDM system supporting a large of the optimal channel estimation and the BER performance
number of users and/or a high number of bits/symbols. This of the idealized optimal Maximum likelihood (ML) turbo
problem is further aggravated when the number of users/MSs MUD/decoder associated with perfect channel estimation, re-
exceeds that of the receive antennas at the BS, namely, in spectively, despite imposing only a fraction of the idealized
the challenging scenario of rank-deficient systems. It was turbo ML-MUD/decoder’s complexity.
demonstrated that the iterative decoding convergence of this From the discussions above, we can see that the concept of
two-stage system may be improved by incorporating a unity NOMA has already existed in various systems, such as SDMA,
rate code (URC) having an infinite impulse response, which where users are distinguished using the unique, user-specific
improves the efficiency of the extrinsic information exchange. CIRs. Actually, these systems require accurate CIR estimation
Although this results in a slightly more complex three-stage to successfully realize MUD, which becomes extremely chal-
system architecture, it allows us to use a low-complexity SD lenging when the number of users is much higher than that of
having a significantly reduced detection candidate list size. receiver antennas. Solving this CIR estimation problem logi-
Alternatively, a reduced SNR is required. For example, given cally leads to the concept of joint channel and data estimation,
a target BER of 10−5 and a candidate list size of 32 for and these high-end solutions often rely on powerful non-linear
the SD, the three-stage receiver is capable of achieving a MUDs. In fact, most of the studies focus on MUD design,
performance gain of 2.5 dB over its two-stage counterpart in a and a series of non-linear MUD algorithms such as parallel
rank-deficient SDMA/OFDM 4-QAM system supporting eight interference cancellation (PIC) [101], and space-time decision
co-channel users and employing for receive antennas at the BS, feedback equalization (ST-DFE) [103] have been proposed.
namely, in an (8×4) rank-deficient system having a normalized Further, with the development of evolutionary algorithms,
user-load of two. For the sake of further enhancing the three- algorithms like genetic algorithm (GA), and particle swarm
stage concatenated receiver, the proposed iterative center- optimization (PSO) may be explored to acquire accurate CIR
shifting SD scheme and the so-called irregular convolutional estimation [104]. In contrast, power-domain NOMA transmits
codes (IrCCs) were intrinsically amalgamated, which led to an the superposition of multi-user signals with different power-
additional performance gain of 2 dB. allocation coefficients, and usually SIC is used at the receiver
In [103] Chen et al. proposed a space-time decision feed- to detect multi-user signals. As the channel gain difference
back equalization (ST-DFE) assisted MUD scheme for multi- among users is translated into different multiplexing gains [5],
ple receiver antenna aided SDMA systems. Again, a sophisti- both user grouping and resource allocation have substantial
cated MBER MUD design was invoked, which was shown to effects on the achievable throughput. As a result, most of
be capable of improving the achievable BER performance and the studies concerning power-domain NOMA focus on user
enhancing the attainable system capacity over that of the stan- grouping, resource (power) allocation, and performance anal-
dard MMSE design. An appealing adaptive implementation of ysis. Recently, with the development of mmWave communi-
the MBER ST-DFE assisted MUD was also proposed using a cation and massive MIMO, combining power-domain NOMA
stochastic gradient-based least bit error rate algorithm, which with mmWave and massive MIMO has become a promising
was demonstrated to consistently outperform the classical technique [131]–[133].
least mean square (LMS) algorithm, while imposing a lower 2) Pattern division multiple access (PDMA): Apart from
computational complexity than the LMS algorithm for the the SDMA scheme mentioned above, the family of PDMA
binary signalling scheme considered. It was demonstrated that schemes [114] [115] constitutes another promising NOMA
the MBER ST-DFE assisted MUD is more robust to channel class that can be implemented in multiple domains. At the
estimation errors as well as to potential error propagation transmitter, PDMA employs non-orthogonal patterns, which
imposed by decision feedback errors, than the MMSE ST-DFE are designed by maximizing the diversity and minimizing
assisted MUD. the correlations among the users. Then, multiplexing can be
The development of evolutionary algorithms (EAs) [104], realized in the code-, power- or spatial-domains, or in fact
such as GAs, repeated weighted boosting search (RWBS), in their combinations. Multiplexing in the code domain is
particle swarm optimization (PSO), and differential evolution reminiscent of SAMA [99]. Multiplexing in the power domain
algorithms (DEAs) stimulated wide interests in the com- has a system model similar to multiplexing in the code domain,
munication research community. However, the quantitative but power-scaling has to be considered under the constraint
performance-versus-complexity comparison of GA, RWBS, of a given total power. Multiplexing in the spatial domain
PSO, and DEA techniques applied to the joint channel estima- leads to the concept of spatial PDMA, which relies on multi-
tion and turbo MUD/decoding in the context of SDMA/OFDM antenna aided techniques. In contrast to MU-MIMO, spatial
systems is a challenging problem, which has to consider both PDMA does not require joint precoding for realizing spatial
the channel estimation problem formulated over a continuous orthogonality, which significantly reduces the system’s design
search space and the MUD optimization problem defined over complexity. Additionally, multiple domains can be combined

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

19

C1
j
optional
Bit/symbol
User 0 FEC QAM symbol level signature IFFT
data encoding mapping multiplexing add CP 1
channel
User 1 Bit/symbol
FEC QAM symbol IFFT
data
encoding mapping level signature 1
add CP SF
multiplexing
channel
Date Serial/Parallel
. receiver
IFFT
Stream converter

CS
. P

jF
channel
.
User n-1 FEC QAM symbol
Bit/symbol IFFT PxSF
level signature add CP
data encoding mapping
multiplexing

Fig. 19. NOCA transmitter structure [142] ⃝IEEE.


c
Fig. 17. The LSSA transmitter structure [134] ⃝IEEE.
c
which can distinguish the different users based on their differ-
ent bit-level interleavers, different grid mapping patterns, or
User Channel Bit-level Carrier Baseband
data coding Interleaving
Modulation Grid Mapping
Modulation to RF the combinations of these two techniques [138]. The typical
FEC Repetition Zero padding
Symbol-level
transmitter structure of IGMA is shown in Fig. 18. Specifically,
interleaving
the channel coding process can be either simple repetition
Fig. 18. The schematic of IGMA transmitter [138] ⃝IEEE.
c coding (spreading) of moderate coding rate for classic for-
ward error correction (FEC) or low-rate FEC. By contrast,
in PDMA to make full use of the various wireless resources the grid mapping process of Fig. 18 may vary from sparse
available. The simulation results of [115] demonstrated that mapping based on zero padding to symbol-level interleaving,
compared to LTE, PDMA is potentially capable of achieving which could provide another dimension for user multiplexing.
a 200% normalized throughput gain in the uplink, and more Whilst we need well-designed FEC codes and spreading code
than 50% throughput gain may be attained in the downlink. sequences, the design of bit-level interleavers and/or grid
3) Signature-based NOMA: Signature-based NOMA mapping patterns in Fig. 18 is somewhat more related. They
schemes are also proposed as promising candidates for 5G. provide scalability to support different connection densities,
Low code rate and signature based shared access (LSSA) whilst striking a trade off between the channel coding gain and
is one of them, and the transmitter structure conceived for the benefit gleaned from sparse resource mapping. Moreover,
uplink massive machine-type communication (mMTC) is the symbol-level interleaving of Fig. 18 randomizes the symbol
depicted in Fig. 17 [134]. LSSA [134] multiplexes each user’s sequence order, which may bring about further benefits in
data either at bit or symbol level with the aid of a specific terms of combating the deleterious effects of frequency se-
signature pattern, which consists of a reference signal (RS), lective fading and inter-cell interference. Besides, a relatively
complex/binary sequence, and permutation pattern of a short low-complexity multi-user detector can be applied and the
length vector. All the users’ signatures share the same short employment of a sparse grid mapping pattern could further
vector length, which can be chosen randomly by the mobile reduce the detection complexity imposed.
terminal or assigned to the user by the network. Moreover, Another interleaver-based multiple access scheme, namely
LSSA can be optionally modified to have a multi-carrier interleave-division multiple access (IDMA), has also been pro-
variant in order to exploit the frequency diversity provided posed. Explicitly, IDMA interleaves the chips after the sym-
by a wider bandwidth, and to reduce the latency. It can also bols have been multiplied by the spreading sequences. Hence,
support asynchronous uplink transmissions, because the BS is IDMA is effectively chip-interleaved CDMA. As shown in
capable of distinguishing/detecting the overlaid user signals [139], compared to CDMA, IDMA is capable of achieving
by correlating them with the signature patterns, even if the about 1 dB Eb /N0 gain at a BER of 10−3 in highly loaded
transmission timing is different from each other. systems having a normalized user-load of 200%. The gain
Similar to LSSA, resource spread multiple access (RSMA) is mostly attributable to the fact that chip-interleaving results
[135] [136] also assigns unique signatures to separate the in an increased diversity-gain compared to conventional bit-
different users and spreads their signals over all the available interleaving.
time and frequency resources. The unique signatures may be 5) Spreading-based NOMA: There are also many other
constituted by the power, spreading/scrambling codes with NOMA schemes based on spreading codes, which are con-
good correlation properties, interleavers, or their combinations, sistent with the concept of the aforementioned code-domain
and interference-cancellation type receivers can be utilized. NOMA, and non-orthogonal coded multiple access (NCMA)
Depending on the specific application scenarios, RSMA may is one of them [140]. NCMA is based on resource spreading
include [137]: by using non-orthogonal spreading codes having a low corre-
• single carrier RSMA: It is optimized for battery power lation. These codes can be obtained by finding the solutions
consumption and link budget extension by utilizing single of a Grassmannian line packing problem [141]. By imposing
carrier waveforms and very low peak to average power additional layers using superposition coding, it can provide
ratio (PAPR) modulations. It allows grant-free transmis- an increased throughput and improve connectivity at a low
sion and potentially allows asynchronous access. block error ratio (BLER). Furthermore, since the receiver of
• Multi-carrier RSMA: It is optimized for low-latency the NCMA system adopts parallel interference cancellation
access and allows for grant-free transmissions. (PIC), it has a scaleable performance vs complexity. Conse-
4) Interleaver-based NOMA: Interleave-grid multiple ac- quently, NCMA is eminently suitable for a larger number of
cess (IGMA) is an interleaver-based multiple access scheme, connections exchanging small packets in massive machine-

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

20

type communication (mMTC), or for reducing the collision D. User grouping and resource allocation
probability in contention based multiple access.
In power-domain NOMA, the channel gain difference a-
Non-orthogonal coded access (NOCA) is also a spreading mong users is translated into different multiplexing gains [5].
based multiple access scheme [142]. Similar to other spreading Therefore, both the user grouping and resource allocation
based schemes, the basic idea of NOCA is that the data techniques have substantial effects on the achievable through-
symbols are spread using non-orthogonal sequences before put [153]. In order to optimize the user grouping and resource
transmission, which can be applied both in the frequency allocation, a reasonable optimization criterion has to be found
domain and/or in time domain based configurations. The basic first. Then, a compelling performance versus complexity trade-
transmitter structure is shown in Fig. 19, where SF denotes off has to be struck. Specifically, the classic proportional
the spreading factors and Cj is the spreading sequence of the fair (PF) scheduler is known to strike an attractive tradeoff
jth user. Specifically, the original modulated data sequence is between the capacity and fairness attained. Hence it has been
first converted into P parallel sequences, then each of the P used both in the orthogonal 3G and 4G multiple access
sequences is spread across SF number of subcarriers. In order systems. Therefore, it has also been widely considered as
to meet the requirements of different scenarios, the spreading a beneficial optimization criterion [5], [54]–[57] in NOMA
factors can be adaptively varied. systems. Specifically, both multi-user scheduling and power
allocation per frequency block can be realized by maximizing
6) Bit division multiplexing (BDM): The basic concept
the product of the average user throughput of all the users
of BDM [116] relies on hierarchical modulation, but the
within a cell [5] [54]. In the uplink, the transmission power is
resources of multiplexed users are partitioned at the bit level
usually independently determined for all users, thus scheduling
instead of the symbol level. Strictly speaking, the resource
a user set can be configured at a given power level. In the
allocation of BDM is orthogonal in the bit domain, but multi-
downlink, under the constraint of a fixed total power, user
user signals can share the same constellation, which implies
scheduling and power allocation should be jointly optimized.
that they are superimposed in the symbol domain.
More particularly, for a given scheduling user set, the classic
7) Compressive sensing (CS)-based NOMA: CS may be iterative water-filling based power allocation algorithm [5] [54]
readily combined with NOMA schemes for exploiting either can be used, which achieves the maximum weighted sum of
the user activity sparsity or data sparsity [145] [146]. A range the user throughput, when exploiting the uplink-downlink du-
of CS-based random access schemes have been conceived ality. Given the optimal power allocation for each scheduling
recently, such as the family of asynchronous random access user set, the scheduling user set is selected by maximizing the
protocols [147] and compressive random access arrangement optimization criterion.
of [148]. Additionally, in [149], random multiple access re- More particularly, matching theory has been shown an
lying on CS was invoked for maximizing the system’s total effective tool for user grouping in NOMA. Matching theory
throughput. Furthermore, the attainable throughput associated can be used to efficiently solve the combinatorial problem of
with different amount of channel knowledge was discussed, matching players from two distinct sets according to players’
which provided useful insights into the quantitative benefits individual information and their preferences [154] [155]. Due
of CS in the context of throughput maximization in random to the low complexity, matching theory has been widely
multiple access schemes. Furthermore, the joint detection of used to address the resource optimization problems in NO-
both node activity and of the received data was proposed MA wireless networks [156]–[159]. The resource allocation
in [150] for machine-type communication, which exploited the in NOMA systems, such as user grouping and subchannel
sporadic nature of the expected communication. All in all, CS allocation, can be modeled as classical matching structures
is expected to play an important role in NOMA schemes. (one-to-one [156], many-to-one [157] [158], and many-to-
many [159]). The simplest matching model is one-to-one
8) Miscellaneous NOMA Schemes: Apart from the NO- matching, where each user from one set can match with at
MA schemes introduced above, there are four other schemes most one user from the opposite set. In a downlink cognitive
proposed by different companies as part of the Rel-14 3GPP radio inspired NOMA networks, a one-to-one matching prob-
NR Study Item shown in Table I. Among these four NOMA lem was formulated to optimize the user pairing to improve
schemes, three are spreading-based NOMA arrangements, the system throughput [156]. Two users in the matched user
namely the low density spreading-signature vector extension pair can share the same spectrum to improve each user data
(LDS-SVE) proposed by Fujitsu [151], as well as the fre- rate and the entire system sum rate. Regarding NOMA het-
quency domain spreading (FDS) [152] and the low code rate erogeneous networks, a many-to-one matching was modeled
spreading (LCRS) [152] schemes proposed by Intel. They to address the spectrum allocation problem [157], where a
spread the user symbols to multiple RBs in order to attain an swap-operation enabled matching algorithm was proposed to
increased diversity gain. Repetition division multiple access match the small base stations with the suitable resource block
(RDMA) proposed by MTK [143] belongs to the family of aiming at maximizing small cell users throughout. In many-to-
interleaver-based NOMA schemes, which can readily separate many matching, at least one player in one set can be matched
the different users’ signals, whilst exploiting both time- and with multiple players in the opposite set. In [159], in order
frequency-diversity with the aid of the cyclic-shift repetition to maximize the system sum rate, subchannel allocation was
of the modulated symbols. formulated as a two-sided many-to-many matching process

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

21

TABLE II
C OMPARISON OF NOMA SCHEMES PROPOSED FOR THE R EL -14 3GPP NR S TUDY I TEM [144].
Candidate
NOMA Schemes Signature Collision Pattern Target Scenarios
Receivers
Low/medium SE per UE;
1 Power-domain NOMA Power domain superposition Full collision SIC
High connection efficiency
Low/medium SE per UE;
2 SCMA Symbol level spreading Partial collision MPA
High connection efficiency
Low/medium SE per UE;
3 MUSA Symbol level spreading Partial collision MMSE-SIC
High connection efficiency
MPA;
Low/medium SE per UE;
4 PDMA Symbol level spreading Partial collision SIC-MPA;
High connection efficiency
SIC
Low/medium SE per UE;
5 LSSA Symbol level spreading Full collision MMSE-SIC
High connection efficiency
MMSE-SIC; Very low SE;
6 RSMA Symbol level scrambling Full collision
ESE-PIC Large coverage extension
MPA;
Low/medium SE per UE;
7 IGMA Symbol level interleaving Partial collision SIC-MPA;
High connection efficiency
SIC
Low/medium SE per UE;
8 IDMA Bit level interleaving Full collision ESE-PIC
High connection efficiency
MMSE-SIC; Low/medium SE per UE;
9 NCMA Symbol level spreading Full collision
ESE-PIC High connection efficiency
MMSE-SIC; Low/medium SE per UE;
10 NOCA Symbol level spreading Full collision
ESE-PIC High connection efficiency
Symbol level spreading and Low/medium SE per UE;
11 GOCA Full collision MMSE-SIC
scrambling High connection efficiency
MPA;
Low/medium SE per UE;
12 LDS-SVE Symbol level spreading Partial collision SIC-MPA;
High connection efficiency
SIC
Low/medium SE per UE;
13 FDS Symbol level spreading Full collision MMSE-SIC
High connection efficiency
Low/medium SE per UE;
14 LCRS Bit level spreading Full collision ESE-PIC
High connection efficiency
Low/medium SE per UE;
15 RDMA Symbol level interleaving Full collision MMSE-SIC
High connection efficiency

for a downlink NOMA network. Sequentially, a two-to-one NOMA systems [59], where a dynamic user allocation and
matching was utilized to improve the energy efficiency of a power optimization problem was investigated. Specifically, a
downlink NOMA network [158]. In this research work, a low sub-optimal two-step method has been proposed. In the first
complexity algorithm was proposed to allocate multiple users step, the power allocation is optimized by fixing a specific
on subchannels to maximize the system energy efficiency. The combination of user allocation according to the max-min
subchannel allocation was considered as dynamic matching fairness, while the second step considered all the user alloca-
process between user set and subchannel set. According to tion combinations. Furthermore, the joint power and channel
the predefined preference lists, each user can send match- allocation optimization has been shown to be NP-hard in [60],
ing request to its most preferred subchannel. However, the and an algorithm combining Lagrangian duality and dynam-
subchannel can accept or reject the user depending on the ic programming was proposed for delivering a competitive
energy efficiency the user can provide on this subchannel. The suboptimal solution. Furthermore, in [61], power allocation
matching process will terminate when there is no user left has been conceived for the NOMA downlink supporting two
to match. Besides, for uplink NOMA, user grouping based users when practical modulation schemes are employed. To
on CSI was investigated considering some predefined power elaborate a little further, the mutual information metric rather
allocation schemes in [160], where the optimization problem than the Shannon-theoretic throughput metric has been used
of user grouping to achieve the maximum sum rate was ana- for deriving a more accurate result, and it has been shown that
lyzed in the large system limit for various scenarios, and some the power allocation problem formulated for maximizing the
optimum and sub-optimum algorithms with a polynomial-time total mutual information depends on the modulation schemes
complexity were proposed. employed.
Additionally, the max-min fairness criterion associated with In addition to spectral efficiency, energy efficiency is also
instantaneous CSI knowledge, i.e. maximizing the minimum a KPI for 5G. In [161], with the assumption of the known
user rate, and the min-max fairness criterion relying on the of perfect CSI at BS, the energy efficiency maximization
average-CSI knowledge, i.e. minimizing the maximum outage problem was investigated in a downlink NOMA system by
probability, have been considered in [58] for deriving the jointly optimizing the subchannel assignment and power allo-
power allocation. Low-complexity algorithms have also been cation. Due to the fact that the perfect CSI is challenging to
developed for solving the associated non-convex problems. obtain, [162] considered the energy efficiency maximization
The max-min fairness criterion has also been used in MIMO problem in a downlink NOMA system with imperfect CSI.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

22

[163] studied the energy efficiency optimization for the fading increase the signalling cost, especially when the receiver does
MIMO NOMA systems with statistical CSI at the transmitter, not know which users are active. On the other hand, in LDS-
where the energy efficiency is defined by ergodic capacity CDMA, LDS-OFDM, SCMA, and SAMA, the complexity of
under unit power consumption. [164] investigated the energy the MPA-based receiver is proportional to O(|X|w ), where |X|
efficiency optimization problem in a cognitive radio inspired denotes the cardinality of the constellation set X, and w is
multiuser downlink NOMA system subject to an individual the maximum number of non-zero signals superimposed on
quality of service constraint for each primary user, and an each chip or subcarrier. As a result, the complexity becomes
efficient algorithm based on the sequential convex approxi- high in typical scenarios of massive connectivity. The BER
mation method was presented to solve the formulated non- performance of three typical code-domain NOMA schemes is
convex fractional programming problem. From another hand, also compared in [166].
the wireless power transfer can also integrated into NOMA
system to further improve the energy efficiency of the NOMA
F. Performance Evaluations and Transmission Experiments of
systems, e.g., [165] considered the application of simultaneous
NOMA
wireless information and power transfer into a cooperative
NOMA system to improve the energy efficiency of the system. We have provided some theoretical analysis of NOMA in
Considering the inter-cell interference, fractional frequency the previous sections, which shows that NOMA yields a better
reuse (FFR), which allows the users under different channel performance than traditional OMA schemes, and this makes
conditions to rely on different frequency reuse factors, has it a promising candidate for 5G wireless communication. In
been employed in [54], [60], [62] for further enhancing the this part, we intend to present some performance evaluations
performance of the cell-edge users. FFR-based power allo- and transmission experiments of NOMA, so as to verify the
cation strikes a tradeoff between the frequency bandwidth analytical results.
utilization per cell and the impact of inter-cell interference. To assess the efficiency of NOMA, NTT DOCOMO per-
formed performance evaluations and transmission experiments
using prototype equipment [20], [21], [167]–[170]. Specifi-
E. Comparison of NOMA Solutions cally, in their experiments, the radio frame configuration was
Based on the discussion above, Table II summarizes the designed based on LTE Release 8, and the targets of these eval-
comparison of existing dominant NOMA schemes. uations were Transmission Mode 3 (TM3) and Transmission
From a theoretical perspective, code-domain NOMA is Mode 4 (TM4), operating without and with feeding back the
capable of achieving a beneficial “spreading gain” with the aid user’s precoding matrix index to the base station, respectively
of using spreading sequences, which may also be termed as [171].
codewords. However, this benefit cannot be readily reaped by Researchers also performed experiments in an indoor radio-
the above-mentioned power-domain NOMA regime. Achiev- wave environment using the prototype equipment in [171]. In
ing a “spreading gain” is an innate benefit of classic CD- this experiment, both UE1 and UE2 are stationary, and the
MA, which may also be viewed as a low-rate repetition- former was near the base station, while the latter was at a
style channel coding scheme, where the code-rate is given point about 50 meters from the base station. It has been shown
by the spreading factor. In simple plausible terms, when for that NOMA was capable of obtaining a throughput gain of
example one of the chips is corrupted by a high noise- or approximately 80% over OFDMA when a 2-by-2 SU-MIMO
interference-sample, the specific spreading code may still be is adopted.
recovered by a matched filter or correlator based receiver. In Apart from NTT DOCOMO, Huawei Technologies have
contrast to CDMA, SCMA is capable of achieving an extra also developed a SCMA-based multi-user uplink prototype
“shaping gain” due to the optimization of the associated multi- to verify the advantages of the SCMA technology in real
dimensional constellation [89]. communication systems [172]. The Huawei demo system
We can also compare the dominant NOMA schemes in consists of one base station using two antennas for diversity
terms of their signalling techniques and complexity. In power- combined reception, and 12 single-antenna aided users for
domain NOMA, SIC constitutes one of the popular inter- uplink access and data transmission.
ference cancellation techniques. The complexity of the SIC- The basic system configurations of the demo are aligned
MMSE is O(K 3 ), where K is the number of users supported. with the current LTE TDD system. In particular, the re-
Therefore, the complexity of SIC is significantly lower than searchers use the LTE TDD Configuration 1 and use OFD-
that of the optimal full-search-based MUD, especially when MA as the baseline for their performance comparison. The
K is high, which is expected to be the case in practical specifications of the prototype system are shown in Table III.
5G systems. Note that the implementation of NOMA in The in-lab prototype system relies on a software-defined
CoMP imposes a relatively high complexity, and a promising baseband, which means that all the baseband processing is
congenial solution to this problem has been discussed in [32]. realized by a CPU instead of FPGA/DSP. At the base station
However, side-information has to be transmitted in order to side, a single server (Huawei Tecal RH2288) is responsible
signal the associated power assignment, which imposes a for all the baseband processing, which is then connected to
signalling overhead. Additionally, in code-domain NOMA, the standard commercial radio frequency components (Huawei
specific spreading sequences or codebooks have to be known product RRU3232). At the user side, the CPU of a lap-
at the receiver in order to support the MUD, which will top (MacBook Pro ME294CH/A) is used for modeling the

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

23

TABLE III
S PECIFICATION FOR SCMA PROTOTYPE OF H UAWEI [172] ⃝IEEE.
c 800
700
Mode Sparse code multiple access

Throughput (kbps)
Number of active UEs 12 out of 14 600
23 dBm (max) with open-loop 500
UE transmit power SCMA
power control
Basic waveform OFDM / F-OFDM 400
LTE
MIMO mode 1-by-2 SIMO 300
Center frequency/bandwidth 2.6GHz/20MHz 200
Scheduled resource 48RBs/4 RBs
Code rate 0.3-0.92 100
SCMA codebook 24-by-8, 4 points 0
TDD configuration 1, 4 Subframes Fixed case Mobility case
Frame structure
for PUSCH
Fig. 20. SCMA throughput gain over OFDM in field testing [172] ⃝IEEE.
c
baseband processing for two users, which is connected to
two mobile RF modules for testing. A user interface (UI) limited to 4 RBSs in each subframe. The comparative test
is developed to show the real-time throughput of each UE, results of OFDM and SCMA are shown in Fig. 20, which
supporting the real-time change of both the user status and of indicates that SCMA achieves a nearly 300% throughput gain
the system operational modes as well. over OFDM.
The prototype can run in either OFDMA or SCMA mode,
and supports real-time switching from one to the other. To IV. C HALLENGES , O PPORTUNITIES , AND F UTURE
ensure a fair comparison, same data rate is maintained for R ESEARCH T RENDS
each user to guarantee the same quality of service. It is shown Existing NOMA schemes relying on either power-domain
by the test that compared to the orthogonal multiple access or code-domain multiplexing are capable of improving the
baseline of 4G LTE, SCMA technology attains up to 300% spectral efficiency with the aid of non-orthogonal resource
throughput gain. For instance, 150% throughput gain can be sharing. What’s more, NOMA techniques are capable of oper-
observed by considering the fact that, if each user requires 12 ating in rank-deficient scenarios, which facilitates the support
physical resource blocks (RBs), a system having a total of 48 of massive connectivity. Therefore, NOMA solutions are con-
RBs can serve at most 4 users using orthogonal LTE OFDMA. sidered as potentially promising 5G candidates. However, there
However, with SCMA, the codebook design supports 6 users are still numerous challenging problems to be solved. Hence
with the same throughput to simultaneously share 48 RBs, thus below some of the key challenges of NOMA designs will
the equivalent delivered amount of data is actually 12 × 6 = be highlighted, along with opportunities and future research
72 RBs rather than 12 × 4 = 48 RBs, which results in the trends addressing these challenges.
throughput gain of about 72/48 = 150%. The 300% gain can
be calculated in a similar way, but needs a different codebook A. Theoretical analysis
associated with a larger spreading factor and larger number To elaborate a little further, in-depth theoretical analysis is
of data layers. In the prototype, a 24-by-8 SCMA codebook required to provide additional insights to guide and inform the
is used to support 12 users (each having 2 data streams) to associated system design. The attainable capacity of multiple
transmit simultaneously. By contrast, for LTE OFDMA, only access schemes constitutes one of the most essential system
4 users out of 12 can transmit data. performance criteria. Specifically, the capacity bounds of code-
Apart from the fading simulator-based test in the Lab, a domain NOMA relying sophisticated spreading sequences has
SCMA prototype has also been deployed in field trials to to be investigated. Similar methods and tools can also be
evaluate the performance. Specifically, four different test cases conceived for MC-CDMA. On the other hand, the maximum
are designed, and UEs are deployed at different locations to normalized user-load that may be supported is limited both by
evaluate the performance of SCMA under different conditions. the achievable interference cancellation capability and by the
The four test cases are as follows: affordable receiver complexity, which is related to the specific
• Case 1: 12 UEs closely located in an area without design of both the spreading sequences and the receiver.
mobility.
• Case 2: 12 UEs located in an area with distant separation B. Design of spreading sequences or codebooks
but no mobility In LDS systems, due to the non-orthogonal resource allo-
• Case 3: 12 UEs moving along a road about 120 meters cation, there is mutual interference amongst the users. The
away from the BS (open-loop power control maintains maximum number of superimposed symbols at each orthogo-
mediocre transmit power at UE)) nal resource “index” is determined by the particular spreading
• Case 4: 12 UEs are moving along a road about 180 sequences or codewords of the users, which has a direct impact
meters away from the BS (open-loop power control on the interference cancellation capability achieved at the
provides a comparatively high power for the UE)) receiver. Therefore, the factor graph of the message passing
In all field trial tests, typical small packets of 20 bytes algorithm should be optimized to strike a compelling tradeoff
(METIS definition) are used as payload for both LTE and between the normalized user-load supported and the receiver
SCMA, and the scheduling resources in the whole system are complexity imposed.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

24

In addition, it has been shown that the message passing NOMA was investigated in [173] [174], and some optimization
algorithm is capable of determining the exact marginal distri- algorithms have been proposed for reducing the channel esti-
bution in case of an idealized cycle-free factor graph, and an mation error. Nevertheless, with the increase of the number of
accurate solution can be obtained with the aid of “locally tree users in future 5G systems, more grave inter-user interference
like” factor graphs, which implies that the cycle girth should be will be caused, which in turn may result in severe channel
sufficiently high. Graph theory can be used to design cycle- estimation error. Therefore, more advanced channel estimation
free or “locally tree like” factor graphs for NOMA without algorithms are required to achieve accurate channel estimation
any loss of spectral efficiency. On the other hand, realistic in NOMA systems.
factor graphs exhibiting cycles can be decomposed into cycle-
free graphs in some practical applications. In this way, the E. Grant-free NOMA
message passing based receiver is capable of attaining the
optimal performance at the cost of a moderately increased As illustrated in Section II, a high transmission latency
receiver complexity. Additionally, the classic matrix design and a high signaling overhead are encountered by an access-
principle and low-density-parity-check (LDPC) code based grant based transmission scheme due to the uplink scheduling
design methods can be invoked for constructing the factor requests and downlink resource assignments required. It is
graph of NOMA solutions. expected that NOMA is capable of operating without grant-
Apart from the challenge of factor graph design, we should free transmissions at a low transmission latency, at a small
also consider how to choose the non-zero values for each signaling overhead, whilst supporting massive connectivity,
sequence. The non-zero values superimposed at the same especially in case transmitting short packets, as expected
resource indices should be distinct. A promising technique is in 5G. Hence contention-based NOMA schemes constitute
to select different values from a complex-valued constellation a promising solution, in which one or more pre-configured
for these non-zero elements in order to maintain the maximum resources are assigned to the contending users. On the oth-
possible Euclidean distance. er hand, integrated protocols - including random back-off
schemes - can be considered as a technique of resolving
non-orthogonal collisions, whilst reducing the packet dropping
C. Receiver design
rates. Additionally, without relying on any access-grant proce-
The complexity of an MPA-based receiver may still become dure, the BS cannot obtain any information on the associated
excessive for massive connectivity in 5G. Therefore, some user activity, which however can be fortunately detected by
approximate solutions of the MPA can be used for reducing CS-aided recovery algorithms due to the sparsity of user
receiver complexity, such as a Gaussian approximation of activity.
the interference, which models the interference-plus-noise as
a Gaussian distribution. This approximation becomes more
F. Resource allocation
accurate, when the number of connections becomes high,
as expected in 5G. Additionally, the MPA can be used to In power-domain NOMA, the interference cancellation ca-
jointly detect and channel-decode the received symbols, where pability of receivers is closely related to the accuracy of the
the constructed graph consists of variable nodes, observation power allocation scheme. On the other hand, the accuracy
nodes and check nodes corresponding to the check equations of allocating the power of each user directly affects the
of the LDPC code. In this way, extrinsic information can be throughput of both power-domain NOMA and of code-domain
more efficiently exchanged between the decoder and demod- NOMA. By carefully adjusting the power allocation under
ulator used at the receiver for improving the signal detection a specific total power constraint, the BS becomes capable
performance. of flexibly controlling the overall throughput, the cell-edge
For an SIC-based receiver, the associated error propagation throughput and the rate-fairness of the users. The optimal
may degrade the performance of some users. Therefore, at resource allocation scheme has to search through the entire
each stage of SIC, a high-performance non-linear detection search space of legitimate solutions, and thus the complexity
algorithm can be invoked for alleviating the influence of error may become excessive. Both dynamic programming algo-
propagation. rithms and greedy algorithms may be considered for realizing
a near-optimal power allocation operating at a low complexity.
D. Channel estimation Additionally, in order to support various applications, dynamic
power allocation constitutes a promising research topic for
In most NOMA contributions, perfect CSI is assumed for future work.
resource allocation or multi-user detection. However, it is not
practical to obtain perfect CSI in practical systems, hence
channel estimation errors exist in NOMA. The impact of the G. Extension to MIMO
residual interference imposed by realistic imperfect channel It is desirable to extend the existing NOMA schemes to their
estimation on the achievable throughput of NOMA systems MIMO-aided counterparts, especially to large-scale MIMO
has been investigated in [48] [49], and a low-complexity trans- systems, in order to further improve the attainable spectral
mission rate back-off algorithm was conceived for mitigating efficiency by exploiting the spatial diversity gain and/or the
the impact of the channel estimation errors. Furthermore, multiplexing gain of MIMO systems. However, the design
the design of the practical channel estimators conceived for of MIMO-aided NOMA techniques is by no means trivial.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

25

Consider the power-domain NOMA as an example. Recall employment of a single scheme for all applications, regardless
that the key idea of power-domain NOMA is to allocate of their diverse requirements. Accordingly, various system
the transmission power to users inversely proportionally to design factors have to be considered in order to accommodate
their channel conditions. For scenarios associated with single- the worst-case condition, which leads to inefficient multiple
antenna nodes, it is possible to compare the users’ chan- access design in many applications. Therefore, the software
nel conditions, since channel gains/attenuations are scalars. defined multiple access technology is expected to support the
However, in MIMO scenarios the channels are represented flexible configuration of multiple access schemes, and thus
by a matrix. Hence it becomes difficult to decide, which different services as well as applications can be supported in
user’s channel is better. This dilemma leads to implementional 5G. It is expected that NOMA solutions will achieve further
difficulties for NOMA-solutions. This is still a promising open performance improvements by addressing these challenges.
area at the time of writing, with very few solutions proposed
in the open literature. A possible solution is to request the BS V. C ONCLUSIONS
to form multiple beams, where NOMA techniques are invoked
for supporting the users covered by the same directional In this article, we have discussed the key concept and
beam and MIMO precoding/detection is used to cancel the advantages of NOMA techniques, which constitute one of the
inter-beam interference [175]. Another possible solution is to promising technologies for future 5G systems. The dominant
assign different beams to different users individually, where NOMA schemes have been introduced together with their
the NOMA power allocation constraint has to be taken into comparison in terms of their operating principles, key features,
consideration for the design of beamforming [125]. receiver complexity, pros and cons, etc. We also highlighted
a range of key challenges, opportunities and future research
trends related to the design of NOMA, including the theoreti-
H. Cognitive radio inspired NOMA cal analysis, the design of spreading sequences or codebooks,
The advantage of NOMA techniques can be simply illus- the receiver design, the design issues of access-grant-free
trated by exploiting the concept of cognitive radio networks. NOMA, resource allocation schemes, extensions to massive
Specifically, the user associated with poorer channel conditions MIMO systems and so on. It is expected that NOMA will
in a NOMA system can be viewed as a primary user in the play an important role in future 5G wireless communication
context of cognitive radio networks. If conventional OMA is systems supporting massive connectivity and low latency.
used, the bandwidth resources assigned to this primary user,
such as the time slots or frequency slots are solely occupied R EFERENCES
by this user, and no other users can access these bandwidth
resources, even if the primary user has a poor connection [1] A. Osseiran, F. Boccardi, V. Braun, K. Kusume, P. Marsch, M. Maternia,
O. Queseth, M. Schellmann, H. Schotten, H. Taoka, H. Tullberg, M. A.
to the BS. The benefit of using NOMA is analogous to Uusitalo, B. Timus, and M. Fallgren, “Scenarios for 5G mobile and
that of cognitive radio networks, where additional secondary wireless communications: The vision of the METIS project,” IEEE
users can be admitted into the specific beam occupied by Commun. Mag., vol. 52, no. 5, pp. 26–35, May 2014.
[2] F. Boccardi, R. W. Heath Jr, A. Lozano, T. L. Marzetta, and P. Popovski,
the primary user. While these secondary users may impose “Five disruptive technology directions for 5G,” IEEE Commun. Mag.,
performance degradations on the primary user, the overall vol. 52, no. 2, pp. 74–80, Feb. 2014.
system throughput can be significantly improved, particularly [3] A. W. Scott and R. Frobenius, “Multiple access techniques: FDMA,
TDMA, and CDMA,” RF Measurements for Cellular Phones and Wireless
if the secondary users have better connections to the BS. By Data Systems, pp. 413–429, Jan. 2008.
exploiting the appealing concept of cognitive radio networks [4] H. Li, G. Ru, Y. Kim, and H. Liu, “OFDMA capacity analysis in MIMO
we can readily illustrate the performance gain of NOMA over channels,” IEEE Trans. Inf. Theory, vol. 56, no. 9, pp. 4438–4446, Sep.
2010.
conventional OMA, which significantly simplifies the design [5] K. Higuchi and A. Benjebbour, “Non-orthogonal multiple access (NO-
of NOMA systems [175]. For example, for the scenarios MA) with successive interference cancellation for future radio access,”
associated with MIMO schemes or in the presence of co- IEICE Trans. Commun., vol. E98-B, no. 3, pp. 403–414, Mar. 2015.
[6] S. M. R. Islam, N. Avazov, O. A. Dobre, and K.-S. Kwak, “Power-domain
channel interference, the design of optimal power allocation is non-orthogonal multiple access (NOMA) in 5G systems: Potentials and
difficult, since it is challenging to decide the quality-order of challenges,” IEEE Commun. Surveys & Tutorials, vol. 19, no. 2, pp. 721–
the users’ channel conditions. The exploitation of cognitive 742, May 2017.
[7] Z. Wei, J. Yuan, D. W. K. Ng, M. Elkashlan, and Z. Ding, “A survey of
radio networks may impose new constraints on the power downlink non-orthogonal multiple access for 5G wireless communication
allocation, which has to strike a similar throughput and fairness networks,” ZTE Communications, vol. 14, no. 4, 2016.
tradeoff as conventional NOMA. [8] M. S. Ali, H. Tabassum, and E. Hossain, “Dynamic user clustering and
power allocation in non-orthogonal multiple access (NOMA) systems,”
IEEE Access, vol. 4, pp. 6325-6343, Aug. 2016.
I. Further challenges [9] M. S. Ali, E. Hossain, and D. I. Kim, “Non-orthogonal multiple access
(NOMA) for downlink multiuser MIMO systems: User clustering, beam-
Some further challenges should also be resolved in the forming, and power allocation,” IEEE Access, vol. 5, pp. 565-577, Dec.
context of NOMA systems, including the associated refer- 2016.
[10] S. M. R. Islam, M. Zeng, and O. A. Dobre, “NOMA in 5G systems:
ence signal design and channel estimation, the reduction of Exciting possibilities for enhancing spectral efficiency,” IEEE 5G Tech
the PAPR in multi-carrier NOMA systems, such as LDS- Focus, vol. 1, no. 2, pp. 1-6, Jun. 2017.
OFDM, maintaining system scalability, the issues of channel- [11] Z. Ding, X. Lei, G. K. Karagiannidis, R. Schober, J. Yuan, and V. K.
Bhargava, “A survey on non-orthogonal multiple access for 5G networks:
quality feedback design, BS cooperation, etc. Additionally, Research challenges and future trends,” IEEE J. Sel. Areas Commun., vol.
the existing multiple access design routinely assumes the 35, no. 10, pp. 2181-2195, Oct. 2017.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

26

[12] Y. Cai, Z. Qin, F. Cui, G. Y. Li, and J. A. McCann, “Modulation and [32] S. Han, C.-L. I, Z. Xu, and Q. Sun, “Energy efficiency and spectrum
multiple access for 5G networks,” IEEE Commun. Surveys & Tutorials, efficiency co-design: From NOMA to network NOMA,” IEEE MMTC
vol. 20, no. 1, pp. 629-646, Oct. 2017. E-Letter, vol. 9, no. 5, pp. 21–24, Sep. 2014.
[13] L. Dai, B. Wang, Y. Yuan, S. Han, C.-L. I, and Z. Wang, “Non- [33] H. Tabassum, E. Hossain, and M. J. Hossain, “Modeling and analysis
orthogonal multiple access for 5G: Solutions, challenges, opportunities, of uplink non-orthogonal multiple access (NOMA) in large-scale cellular
and future research trends,” IEEE Commun. Mag., vol. 53, no. 9, pp. networks using poisson cluster processes,” IEEE Trans. Commun., vol.
74–81, Sep. 2015. 65, no. 8, pp. 3555-3570, Aug. 2017.
[14] A. Benjebbour, Y. Saito, Y. Kishiyama, A. Li, A. Harada, and T. Naka- [34] W. Shin, M. Vaezi, B. Lee, D. J. Love, J. Lee, and H. V. Poor, “Non-
mura, “Concept and practical considerations of non-orthogonal multiple orthogonal multiple access in multi-cell networks: Theory, performance,
access (NOMA) for future radio access,” in Proc. IEEE Intelligent Signal and practical challenges,” IEEE Commun. Mag., vol. 55, no. 10, pp. 176-
Processing and Communications Systems (IEEE ISPACS’13), Nov. 2013, 183, Oct. 2017.
pp. 770–774. [35] J. Choi, “Non-orthogonal multiple access in downlink coordinated two-
[15] Y. Saito, Y. Kishiyama, A. Benjebbour, T. Nakamura, A. Li, and point systems,” IEEE Commun. Lett., vol. 18, no. 2, pp. 313–316, Jan.
K. Higuchi, “Non-orthogonal multiple access (NOMA) for future radio 2014.
access,” in Proc. IEEE Vehicular Technology Conference (IEEE VTC’13 [36] M. S. Ali, E. Hossain, A. A.-Dweik, and D. I. Kim, “Downlink power
Spring), Jun. 2013, pp. 1–5. allocation for CoMP-NOMA in multi-cell networks,” to appear in IEEE
[16] K. Higuchi and Y. Kishiyama, “Non-orthogonal access with random Trans. Commun., 2018.
beamforming and intra-beam SIC for cellular MIMO downlink,” in Proc. [37] Z. Ding, Z. Yang, P. Fan, and H. V. Poor, “On the performance of non-
IEEE Vehicular Technology Conference (IEEE VTC’13 Fall), Sep. 2013, orthogonal multiple access in 5G systems with randomly deployed users,”
pp. 1–5. IEEE Signal Process. Lett., vol. 21, no. 12, pp. 1501–1505, Jul. 2014.
[38] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, “System-level
[17] N. Nonaka, Y. Kishiyama, and K. Higuchi, “Non-orthogonal multiple
performance evaluation of downlink non-orthogonal multiple access (NO-
access using intra-beam superposition coding and SIC in base station co-
MA),” in Proc. IEEE Personal Indoor and Mobile Radio Communications
operative MIMO cellular downlink,” in Proc. IEEE Vehicular Technology
(IEEE PIMRC’13), Sep. 2013, pp. 611–615.
Conference (IEEE VTC’14 Fall), Sep. 2014, pp. 1–5.
[39] A. Benjebbovu, A. Li, Y. Saito, Y. Kishiyama, A. Harada, and T. Naka-
[18] M. Zeng, A. Yadav, O. A. Dobre, G. I. Tsiropoulos, and H. Vincent mura, “System-level performance of downlink NOMA for future LTE
Poor, “Capacity comparison between MIMO-NOMA and MIMO-OMA enhancements,” in Proc. IEEE Global Communications Conference Work-
with multiple users in a cluster,” IEEE J. Sel. Areas Commun., vol. 35, shops (IEEE Globecom Workshops’13), Dec. 2013, pp. 66–70.
no. 10, pp. 2413-2424, Oct. 2017. [40] Y. Saito, A. Benjebbour, A. Li, K. Takeda, Y. Kishiyama, and T.
[19] M. Zeng, A. Yadav, O. A. Dobre, G. I. Tsiropoulos, and H. Vincent Nakamura, “System-level evaluation of downlink non-orthogonal mul-
Poor, “On the sum rate of MIMO-NOMA and MIMO-OMA systems,” tiple access (NOMA) for non-full buffer traffic model,” in Proc. IEEE
IEEE Wireless Commun. Lett., vol. 6, no. 4, pp. 534-537, Aug. 2017. Conference on Standards for Communications and Networking (IEEE
[20] A. Benjebbour, K. Saito, A. Li, Y. Kishiyama, and T. Nakamura, “Non- CSCN’15), Oct. 2015, pp. 94-99.
orthogonal multiple access (NOMA): Concept, performance evaluation [41] A. Benjebbour, A. Li, Y. Kishiyama, H. Jiang, and T. Nakamura,
and experimental trials,” in Proc. IEEE International Conference on “System-level performance of downlink NOMA combined with SU-
Wireless Networks and Mobile Communications (IEEE WINCOM’15), MIMO for future LTE enhancements,” in Proc. IEEE Global Commu-
Oct. 2015, pp. 1-6. nications Conference Workshops (IEEE Globecom Workshops’14), Dec.
[21] A. Benjebbour, A. Li, K. Saito, Y. Saito, Y. Kishiyama, and T. Nakamu- 2014, pp. 706-710.
ra, “NOMA: From concept to standardization,” in Proc. IEEE Conference [42] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, “System-level
on Standards for Communications and Networking (IEEE CSCN’15), Oct. performance of downlink non-orthogonal multiple access (NOMA) under
2015, pp. 18-23. various environments,” in Proc. IEEE Vehicular Technology Conference
[22] B. Kim, W. Chung, S. Lim, S. Suh, J. Kwun, S. Choi, and D. Hong, (IEEE VTC’15 Spring), May 2015, pp. 1-5.
“Uplink NOMA with multi-antenna,” in Proc. IEEE Vehicular Technology [43] M. Kimura and K. Higuchi, “System-level throughput of NOMA with
Conference (IEEE VTC’15 Spring), May 2015, pp. 1-5. SIC in cellular downlink under FTP traffic model,” in Proc. IEEE
[23] Z. Ding, F. Adachi, and H. V. Poor, “The application of MIMO to non- International Symposium on Wireless Communication Systems (IEEE
orthogonal multiple access,” IEEE Trans. Wireless Commun., vol. 15, no. ISWCS’15), Aug. 2015, pp. 1-5.
1, pp. 537-552, Jan. 2016. [44] Y. Endo, Y. Kishiyama, and K. Higuchi, “Uplink non-orthogonal access
[24] Z. Ding, L. Dai, and H. V. Poor, “MIMO-NOMA design for small packet with MMSE-SIC in the presence of inter-cell interference,” in Proc. IEEE
transmission in the internet of things,” IEEE Access, vol. 4, pp. 1393- International Symposium on Wireless Communication Systems (IEEE
1405, Apr. 2016. ISWCS’12), Aug. 2012, pp. 261-265.
[25] Y. Lan, A. Benjebboiu, X. Chen, A. Li, and H. Jiang, “Considerations [45] P. Sedtheetorn and T. Chulajata, “Spectral efficiency evaluation for
on downlink non-orthogonal multiple access (NOMA) combined with non-orthogonal multiple access in Rayleigh fading,” in Proc. IEEE
closed-loop SU-MIMO,” in Proc. IEEE Signal Processing and Commu- International Conference on Advanced Communication Technology (IEEE
nication Systems (IEEE ICSPCS’14), Dec. 2014, pp. 1-5. ICACT’16), Jan. 2016, pp. 747-750.
[46] Z. Ding, F. Adachi, and H. V. Poor, “Performance of MIMO-NOMA
[26] C. Yan, A. Harada, A. Benjebbour, Y. Lan, A. Li, and H. Jiang, “Receiver
downlink transmissions,” in Proc. IEEE Global Communications Confer-
design for downlink non-orthogonal multiple access (NOMA),” in Proc.
ence (IEEE GLOBECOM’15), Dec. 2015, pp. 1-6.
IEEE Vehicular Technology Conference (IEEE VTC’15 Spring), May
[47] A. Li, A. Benjebbour, and A. Harada, “Performance evaluation of non-
2015, pp. 1-5.
orthogonal multiple access combined with opportunistic beamforming,”
[27] K. Saito, A. Benjebbour, Y. Kishiyama, Y. Okumura, and T. Nakamura, in Proc. IEEE Vehicular Technology Conference (IEEE VTC’14 Spring),
“Performance and design of SIC receiver for downlink NOMA with open- May 2014, pp. 1-5.
loop SU-MIMO,” in Proc. IEEE International Conference on Communi- [48] K. Yamamoto, Y. Saito, and K. Higuchi, “System-level throughput
cation Workshop (IEEE ICCW’15), Jun. 2015, pp. 1161-1165. of non-orthogonal access with SIC in cellular downlink when channel
[28] X. Chen, A. Beiijebbour, A. Li, H. Jiang, and H. Kayama, “Considera- estimation error exists,” in Proc. IEEE Vehicular Technology Conference
tion on successive interference canceller (SIC) receiver at cell-edge users (IEEE VTC’14 Spring), May 2014, pp. 1-5.
for non-orthogonal multiple access (NOMA) with SU-MIMO,” in Proc. [49] N. Nonaka, A. Benjebbour, and K. Higuchi, “System-level throughput of
IEEE Annual International Symposium on Personal, Indoor, and Mobile NOMA using intra-beam superposition coding and SIC in MIMO down-
Radio Communications (IEEE PIMRC’15), Aug. 2015, pp. 522-526. link when channel estimation error exists,” in Proc. IEEE International
[29] Z. Ding, M. Peng, and H. V. Poor, “Cooperative non-orthogonal multiple Conference On Communication Systems (IEEE ICCS’14), Nov. 2014, pp.
access in 5G systems,” IEEE Commun. Lett., vol. 19, no. 8, pp. 1462– 202-206.
1465, Jun. 2015. [50] Z. Yang, Z. Ding, P. Fan, and G. K. Karagiannidis, “On the perfor-
[30] Z. Ding, H. Dai, and H. V. Poor, “Relay selection for cooperative mance of non-orthogonal multiple access systems with partial channel
NOMA,” IEEE Wireless Commun. Lett., vol. 5, no. 4, pp. 416-419, Jun. information,” IEEE Trans. Commun., vol. 64, no. 2, pp. 654-667, Feb.
2016. 2016.
[31] D. Wan, M. Wen, F. Ji, H. Yu, and F. Chen, “Non-orthogonal multiple [51] Q. Sun, S. Han, C.-L. I, and Z. Pan, “On the ergodic capacity of MIMO
access for cooperative communications: Challenges, opportunities, and NOMA systems,” IEEE Wireless Commun. Lett., vol. 4, no. 4, pp. 405–
trends,” IEEE Wireless Commun., vol. 25, no. 2, pp. 109-117, Apr. 2018. 408, Apr. 2015.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

27

[52] S. Timotheou and I. Krikidis, “Fairness for non-orthogonal multiple [73] M. Al-Imari, M. A. Imran, and R. Tafazolli, “Low density spreading
access in 5G systems,” IEEE Signal Process. Lett., vol. 22, no. 10, pp. for next generation multicarrier cellular systems,” in Proc. IEEE Future
1647–1651, Mar. 2015. Communication Networks (IEEE ICFCN’12), Apr. 2012, pp. 52–57.
[53] K. Yakou, and K. Higuchi, “Downlink NOMA with SIC using unified [74] M. Al-Imari, M. A. Imran, R. Tafazolli, and D. Chen, “Subcarrier
user grouping for non-orthogonal user multiplexing and decoding order,” and power allocation for LDS-OFDM system,” in Proc. IEEE Vehicular
in Proc. IEEE International Symposium on Intelligent Signal Processing Technology Conference (IEEE VTC’11 Spring), May 2011, pp. 1–5.
and Communication Systems (IEEE ISPACS’15), Nov. 2015, pp. 508-513. [75] M. Al-Imari and R. Hoshyar, “Reducing the peak to average power ratio
[54] J. Umehara, Y. Kishiyama, and K. Higuchi, “Enhancing user fairness in of LDS-OFDM signals,” in Proc. IEEE Wireless Communication Systems
non-orthogonal access with successive interference cancellation for cellu- (IEEE ISWCS’10), Sep. 2010, pp. 922–926.
lar downlink,” in Proc. IEEE International Conference on Communication [76] H. Nikopour and H. Baligh, “Sparse code multiple access,” in Proc.
Systems (IEEE ICCS’12), Nov. 2012, pp. 324–328. IEEE 24th International Symposium on Personal Indoor and Mobile
[55] N. Otao, Y. Kishiyama, and K. Higuchi, “Performance of non-orthogonal Radio Communications (IEEE PIMRC’13), Sep. 2013, pp. 332–336.
access with SIC in cellular downlink using proportional fairbased resource [77] Y. Zhou, H. Luo, R. Li, and J. Wang, “A dynamic states reduction
allocation,” in Proc. IEEE International Symposium on Wireless Commu- message passing algorithm for sparse code multiple access,” in Proc.
nication Systems (IEEE ISWCS’12), Aug. 2012, pp.476–480. IEEE Wireless Telecommunications Symposium (IEEE WTS’16), Apr.
[56] X. Chen, A. Benjebbour, A. Li, and A. Harada, “Multi-user proportional 2016, pp. 1-5.
fair scheduling for uplink non-orthogonal multiple access (NOMA),” in [78] Y. Du, B. Dong, Z. Chen, J. Fang, and X. Wang, “A fast convergence
Proc. IEEE Vehicular Technology Conference (IEEE VTC’14 Spring), multiuser detection scheme for uplink SCMA systems,” IEEE Commun.
May 2014, pp. 1–5. Lett., vol. 5, no. 4, pp. 388-391, May 2016.
[57] F. Liu, P. Mähönen, and M. Petrova, “Proportional fairness-based user [79] H. Mu, Z. Ma, M. Alhaji, P. Fan, and D. Chen, “A fixed low complex-
pairing and power allocation for non-orthogonal multiple access,” in Proc. ity message pass algorithm detector for uplink SCMA system,” IEEE
IEEE 26th Annual International Symposium on Personal, Indoor, and Wireless Commun. Lett., vol. 4, no. 6, pp. 585-588, Aug. 2015.
Mobile Radio Communications (IEEE PIMRC’15), Aug. 2015, pp. 1127– [80] Z. Jia, Z. Hui, and L. Xing, “A low-complexity tree search based quasi-
1131. ML receiver for SCMA system,” in Proc. IEEE International Conference
[58] S. Timotheou and I. Krikidis, “Fairness for non-orthogonal multiple on Computer and Communications (IEEE ICCC’15), Oct. 2015, pp. 319-
access in 5G systems,” IEEE Signal Process. Lett., vol. 22, no. 10, pp. 323.
1647–1651, Oct. 2015. [81] Y. Liu, J. Zhong, P. Xiao, and M. Zhao, “A novel evidence theory based
[59] Y. Liu, M. Elkashlan, Z. Ding, and G. K. Karagiannidis, “Fairness of row message passing algorithm for LDS systems,” in Proc. IEEE Inter-
user clustering in MIMO non-orthogonal multiple access systems,” IEEE national Conference on Wireless Communications & Signal Processing
Commun. Lett., vol. 20, no. 7, pp. 1465–1468, Apr. 2016. (IEEE WCSP’15), Oct. 2015, pp. 1-5.
[60] L. Lei, D. Yuan, C. K. Ho, and S. Sun, “Joint optimization of power and [82] D. Wei, Y. Han, S. Zhang, and L. Liu, “Weighted message passing al-
channel allocation with non-orthogonal multiple access for 5G cellular gorithm for SCMA,” in Proc. IEEE International Conference on Wireless
systems,” in Proc. IEEE Global Communications Conference (IEEE Communications & Signal Processing (IEEE WCSP’15), Oct. 2015, pp.
GLOBECOM’15), Dec. 2015, pp. 1–6. 1-5.
[61] J. Choi, “On the power allocation for a practical multiuser superposition [83] K. Xiao, B. Xiao, S. Zhang, Z. Chen, and B. Xia, “Simplified multiuser
scheme in NOMA systems,” IEEE Commun. Lett., vol. 20, no. 3, pp. detection for SCMA with sum-product algorithm,” in Proc. IEEE Inter-
483–441, Jan. 2016. national Conference on Wireless Communications & Signal Processing
[62] Y. Hayashi, Y. Kishiyama, and K. Higuchi, “Investigations on power (IEEE WCSP’15), Oct. 2015, pp. 1-5.
allocation among beams in non-orthogonal access with random beam- [84] Y. Du, B. Dong, Z. Chen, J. Fang, and L. Yang, “Shuffled multiuser
forming and intra-beam SIC for cellular MIMO downlink,” in Proc. IEEE detection schemes for uplink sparse code multiple access systems,” IEEE
Vehicular Technology Conference (IEEE VTC’13 Fall), Sep. 2013, pp. 1– Commun. Lett., vol. 20, no. 6, pp. 1231-1234, Jun. 2016.
5. [85] A. Bayesteh, H. Nikopour, M. Taherzadeh, H. Baligh, and J. Ma, “Low
[63] K. Higuchi and Y. Kishiyama, “Non-orthogonal access with successive complexity techniques for SCMA detection,” in Proc. IEEE Global Com-
interference cancellation for future radio access,” IEEE Vehicular Tech- munications Conference Workshops (IEEE Globecom Workshops’15),
nology Society, Asia Pacific Wireless communications symposium (IEEE Dec. 2015, pp. 1-6.
APWCS’12), pp. 1–5, Aug. 2012. [86] S. Zhang, X. Xu, L. Lu, Y. Wu, G. He, and Y. Chen, “Sparse code multi-
[64] N. Otao, Y. Kishiyama, and K. Higuchi, “Performance of non-orthogonal ple access: An energy efficient uplink approach for 5G wireless systems,”
access with SIC in cellular downlink using proportional fair-based re- in Proc. IEEE Global Communications Conference(IEEE Globecom’14),
source allocation,” in Proc. IEEE Wireless Communication Systems (IEEE Dec. 2014, pp. 4782–4787.
ISWCS’12), Aug. 2012, pp. 476–480. [87] Y. Wu, S. Zhang, and Y. Chen, “Iterative multiuser receiver in sparse
[65] T. Takeda and K. Higuchi, “Enhanced user fairness using non-orthogonal code multiple access systems,” in Proc. IEEE International Conference
access with SIC in cellular uplink,” in Proc. IEEE Vehicular Technology on Communications (IEEE ICC’15), Jun. 2015, pp. 2918-2923.
Conference (IEEE VTC’11 Fall), Sep. 2011, pp. 1–5. [88] B. Xiao, K. Xiao, S. Zhang, Z. Chen, B. Xia, and H. Liu, “Iterative
[66] R. Hoshyar, F. P. Wathan, and R. Tafazolli, “Novel low-density signature detection and decoding for SCMA systems with LDPC codes,” in Proc.
for synchronous CDMA systems over AWGN channel,” IEEE Trans. IEEE International Conference on Wireless Communications & Signal
Signal Process., vol. 56, no. 4, pp. 1616–1626, Apr. 2008. Processing (IEEE WCSP’15), Oct. 2015, pp. 1-5.
[67] D. Guo and C.-C. Wang, “Multiuser detection of sparsely spread [89] M. Taherzadeh, H. Nikopour, A. Bayesteh, and H. Baligh, “SCMA
CDMA,” IEEE J. Sel. Areas Commun., vol. 26, no. 3, pp. 421–431, Apr. codebook design,” in Proc. IEEE Vehicular Technology Conference (IEEE
2008. VTC’14 Fall), Sep. 2014, pp. 1–5.
[68] J. Van De Beek and B. M. Popovic, “Multiple access with low-density [90] L. Yu, X. Lei, P. Fan, and D. Chen, “An optimized design of SCMA
signatures,” in Proc. IEEE Global Communications Conference (IEEE codebook based on star-QAM signaling constellations,” in Proc. IEEE In-
Globecom’09), Dec. 2009, pp. 1–6. ternational Conference on Wireless Communications & Signal Processing
[69] R. Razavi, R. Hoshyar, M. A. Imran, and Y. Wang, “Information (IEEE WCSP’15), Oct. 2015, pp. 1-5.
theoretic analysis of LDS scheme,” IEEE Commun. Lett., vol. 15, no. 8, [91] S. Zhang, B. Xiao, K. Xiao, Z. Chen, and B. Xia, “Design and analysis
pp. 798–800, Jun. 2011. of irregular sparse code multiple access,” in Proc. IEEE International
[70] R. Hoshyar, R. Razavi, and M. Al-Imari, “LDS-OFDM an efficient mul- Conference on Wireless Communications & Signal Processing (IEEE
tiple access technique,” in Proc. IEEE Vehicular Technology Conference WCSP’15), Oct. 2015, pp. 1-5.
(IEEE VTC’10 Spring), May 2010, pp. 1–5. [92] K. Au, L. Zhang, H. Nikopour, E. Yi, A. Bayesteh, U. Vilaipornsawai,
[71] M. Al-Imari, P. Xiao, M. A. Imran, and R. Tafazolli, “Uplink non- J. Ma, and P. Zhu, “Uplink contention based SCMA for 5G radio access,”
orthogonal multiple access for 5G wireless networks,” in Proc. 11th in Proc. IEEE Global Communications Conference (IEEE Globecom’14),
International Symposium on Wireless Communications Systems (IEEE Dec. 2014, pp. 1–5.
ISWCS’14), Aug. 2014, pp. 781–785. [93] A. Bayesteh, E. Yi, H. Nikopour, and H. Baligh, “Blind detection of
[72] M. Al-Imari, M. A. Imran, R. Tafazolli, and D. Chen, “Performance SCMA for uplink grant-free multiple-access,” in Proc. IEEE Wireless
evaluation of low density spreading multiple access,” in Proc. IEEE Communications Systems (IEEE ISWCS’14), Aug. 2014, pp. 853–857.
Wireless Communications and Mobile Computing Conference (IEEE [94] H. Nikopour, E. Yi, A. Bayesteh, K. Au, M. Hawryluck, H. Baligh, and
IWCMC’12), Aug. 2012, pp. 383–388. J. Ma, “SCMA for downlink multiple access of 5G wireless networks,”

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

28

in Proc. IEEE Global Communications Conference (IEEE Globecom’14), [117] R. Zhang and L. Hanzo, “A unified treatment of superposition coding
Dec. 2014, pp. 1–5. aided communications: Theory and practice,” IEEE Commun. Surveys &
[95] U. Vilaipornsawai, H. Nikopour, A. Bayesteh, and J. Ma, “SCMA for Tutorials, vol. 13, no. 3, pp. 503-520, Jul. 2011.
open-loop joint transmission CoMP,” in Proc. IEEE Vehicular Technology [118] D. Tse and P. Viswanath, Fundamentals of Wireless Communication.
Conference (IEEE VTC’15 Fall), Sep. 2015, pp. 1-5. Cambridge: Cambridge University Press, 2005.
[96] T. Liu, X. Li, and L. Qiu, “Capacity for downlink massive MIMO [119] T. M. Cover and J. A. Thomas, Elements of information theory. John
MU-SCMA system,” in Proc. IEEE International Conference on Wireless Wiley & Sons, 2006.
Communications & Signal Processing (IEEE WCSP’15), Oct. 2015, pp. [120] P. P. Bergmans, “A simple converse for broadcast channels with additive
1-5. white Gaussian noise,” IEEE Trans. Inf. Theory, vol. 20, no. 2, pp. 279–
[97] L. Lu, Y. Chen, W. Guo, H. Yang, Y. Wu, and S. Xing, “Prototype 280, Mar. 1974.
for 5G new air interface technology SCMA and performance evaluation,” [121] Y. Liu, G. Pan, H. Zhang, and M. Song, “On the capacity comparison
China Communications, Supplement No. 1, pp. 38-48, Dec. 2015. between MIMO-NOMA and MIMO-OMA,” IEEE Access, vol. 4, pp.
[98] Z. Yuan, G. Yu, and W. Li, “Multi-user shared access for 5G,” Telecom- 2123–2129, May 2016.
munications Network Technology, vol. 5, no. 5, pp. 28–30, May 2015. [122] L. Wang, X. Xu, Y. Wu, S. Xing, and Y. Chen, “Sparse code multiple
[99] X. Dai, S. Chen, S. Sun, S. Kang, Y. Wang, Z. Shen, and J. Xu, “Suc- access-Towards massive connectivity and low latency 5G communica-
cessive interference cancelation amenable multiple access (SAMA) for tions,” Telecommunications Network Technology, vol. 5, no. 5, pp. 6–15,
future wireless communications,” in Proc. IEEE International Conference May 2015.
on Communication Systems (IEEE ICCS’14), Nov. 2014, pp. 1–5. [123] Z. Yang, W. Xu, C. Pan, Y. Pan, and M. Chen, “On the optimality of
[100] M. Y. Alias, S. Chen, and L. Hanzo, “Multiple-antenna-aided OFDM power allocation for NOMA downlinks with individual QoS constraints,”
employing genetic-algorithm-assisted minimum bit error rate multiuser IEEE Commun. Lett., vol. 21, no. 7, pp. 1649-1652, Jul. 2017.
detection,” IEEE Trans. Veh. Technol., vol. 54, no. 5, pp. 1713–1721, [124] C. Chen, W. Cai, X. Cheng, L. Yang, and Y. Jin, “Low complexity
Sep. 2005. beamforming and user selection schemes for 5G MIMO-NOMA systems,”
[101] A. Wolfgang, S. Chen, and L. Hanzo, “Parallel interference cancellation IEEE J. Sel. Areas Commun., vol. 35, no. 12, pp. 2708-2722, Dec. 2017.
based turbo space-time equalization in the SDMA uplink,” IEEE Trans. [125] M. F. Hanif, Z. Ding, T. Ratnarajah, and G. K. Karagiannidis, “A
Wireless Commun., vol. 6, no. 2, pp. 609–616, Feb. 2007. minorization-maximization method for optimizing sum rate in non-
[102] L. Wang, L. Xu, S. Chen, and L. Hanzo, “Three-stage irregular orthogonal multiple access systems,” IEEE Trans. Signal Process., vol. 64,
convolutional coded iterative center-shifting K-best sphere detection for no. 1, pp. 76–88, Sep. 2015.
soft-decision SDMA-OFDM,” IEEE Trans. Veh. Technol., vol. 58, no. 4, [126] Q. Sun, S. Han, Z. Xu, S. Wang, I. Chih-Lin, and Z. Pan, “Sum
pp. 2103–2109, May 2009. rate optimization for MIMO non-orthogonal multiple access systems,” in
[103] S. Chen, L. Hanzo, and A. Livingstone, “MBER space-time decision Proc. IEEE Wireless Communications and Networking Conference (IEEE
feedback equalization assisted multiuser detection for multiple antenna WCNC’15), Mar. 2015, pp. 747-752.
aided SDMA systems,” IEEE Trans. Signal Process., vol. 54, no. 8, pp. [127] Z. Chen, Z. Ding, X. Dai, and G. K. Karagiannidis, “On the application
3090–3098, Aug. 2006. of quasi-degradation to MISO-NOMA downlink,” IEEE Trans. Signal
Process., vol. 64, no. 23, pp. 6174-6189, Dec. 2016.
[104] L. Hanzo, S. Chen, J. Zhang, and X. Mu, “Evolutionary algorithm
[128] L. Li, L. Wang, and L. Hanzo, “Differential interference suppression
assisted joint channel estimation and turbo multi-user detection/decoding
aided three-stage concatenated successive relaying,” IEEE Trans. Com-
for OFDM/SDMA,” IEEE Trans. Veh. Technol., vol. 63, no. 3, pp. 1204–
mun., vol. 60, no. 8, pp. 2146–2155, May 2012.
1222, Mar. 2014.
[129] L. Zhang, W. Li, Y. Wu, X. Wang, S-I. Park, H. M. Kim, J-Y. Lee,
[105] S. Chen, A. Wolfgang, C. J. Harris, and L. Hanzo, “Symmetric
P. Angueira, and J. Montalban, “Layered-division-multiplexing: Theory
RBF classifier for nonlinear detection in multiple-antenna-aided systems,”
and practice,” IEEE Trans. Broadcast., vol. 62, no. 1, pp. 216–232, Mar.
IEEE Trans. Neural Networks, vol. 19, no. 5, pp. 737–745, May 2008.
2016.
[106] A. Wolfgang, J. Akhtman, S. Chen, and L. Hanzo, “Reduced-
[130] F. R. Kschischang, B. J. Frey, and H.-A. Loeliger, “Factor graphs and
complexity near-maximum-likelihood detection for decision feedback
the sum-product algorithm,” IEEE Trans. Inf. Theory, vol. 47, no. 2, pp.
assisted space-time equalization,” IEEE Trans. Wireless Commun., vol. 6,
498–519, Feb. 2001.
no. 7, pp. 2407–2411, Jul. 2007.
[131] Z. Ding, P. Fan, and H. V. Poor, “Random beamforming in millimeter-
[107] J. Akhtman, A. Wolfgang, S. Chen, and L. Hanzo, “An optimized- wave NOMA networks,” IEEE Access, vol. 5, pp. 7667–7681, Feb. 2017.
hierarchy-aided approximate Log-MAP detector for MIMO systems,” [132] B. Wang, L. Dai, Z. Wang, N. Ge, and S. Zhou, “Spectrum and energy
IEEE Trans. Wireless Commun., vol. 6, no. 5, pp. 1900–1909, May 2007. efficient beamspace MIMO-NOMA for Millimeter-Wave communications
[108] S. Chen, A. Livingstone, H. Q. Du, and L. Hanzo, “Adaptive min- using lens antenna array,” IEEE J. Sel. Areas Commun., vol. 35, no. 10,
imum symbol error rate beamforming assisted detection for quadrature pp. 2370–2382, Oct. 2017.
amplitude modulation,” IEEE Trans. Wireless Commun., vol. 7, no. 4, pp. [133] A. Marcano, and H. L. Christiansen, “Performance of Non-Orthogonal
1140–1145, Apr. 2008. Multiple Access (NOMA) in mmWave wireless communications for
[109] J. Zhang, S. Chen, X. Mu, and L. Hanzo, “Turbo multi-user detection 5G networks,” in Proc. IEEE International Conference on Computing,
for OFDM/SDMA systems relying on differential evolution aided iterative Networking and Communications (IEEE ICNC’ 17), Jan. 2017, pp. 26-
channel estimation,” IEEE Trans. Commun., vol. 60, no. 6, pp. 1621– 29.
1633, Jun. 2012. [134] 3GPP, “Low code rate and signature based multiple access scheme for
[110] J. Zhang, S. Chen, X. Mu, and L. Hanzo, “Joint channel estimation and New Radio,” TSG RAN1 #85, Nanjing, China, 23rd-27th, May. 2016.
multi-user detection for SDMA/OFDM based on dual repeated weighted [135] 3GPP, “Discussion on multiple access for new radio interface,” TSG
boosting search,” IEEE Trans. Veh. Technol., vol. 60, no. 7, pp. 3265– RAN WG1 #84bis, Busan, Korea, 11th-15th, Apr. 2016.
3275, Jun. 2011. [136] 3GPP, “Initial views and evaluation results on non-orthogonal multiple
[111] C.-Y. Wei, J. Akhtman, S.-X. Ng, and L. Hanzo, “Iterative near- access for NR uplink,” TSG RAN WG1 #84bis, Busan, Korea, 11th-15th,
maximum-likelihood detection in rank-deficient downlink SDMA system- Apr. 2016.
s,” IEEE Trans. Veh. Technol., vol. 57, no. 1, pp. 653–657, Jan. 2008. [137] 3GPP, “Candidate NR multiple access schemes,” TSG RAN WG1
[112] A. Wolfgang, J. Akhtman, S. Chen, and L. Hanzo, “Iterative MIMO #84b, Busan, Korea, 11th-15th, Apr. 2016.
detection for rank-deficient systems,” IEEE Signal Process. Lett., vol. 13, [138] 3GPP, “Non-orthogonal multiple access candidate for NR,” TSG RAN
no. 11, pp. 699–702, Nov. 2006. WG1 #85, Nanjing, China, 23rd-27th, May. 2016.
[113] L. Xu, S. Chen, and L. Hanzo, “EXIT chart analysis aided turbo MUD [139] K. Kusume, G. Bauch, and W. Utschick, “IDMA vs. CDMA: Analysis
designs for the rank-deficient multiple antenna assisted OFDM uplink,” and comparison of two multiple access schemes,” IEEE Trans. Wireless
IEEE Trans. Wireless Commun., vol. 7, no. 6, pp. 2039–2044, Jun. 2008. Commun., vol. 11, no. 1, pp. 78–87, Jan. 2012.
[114] “5G: Rethink mobile communications for 2020+,” FuTURE Mobile [140] 3GPP, “Considerations on DL/UL multiple access for NR,” TSG RAN
Communication Forum, Nov. 2014. WG1 #84bis, Busan, Korea, 11th-15th, Apr. 2016.
[115] S. Kang, X. Dai, and B. Ren, “Pattern division multiple access for 5G,” [141] A. Medra and T. N. Davidson, “Flexible codebook design for limited
Telecommunications Network Technology, vol. 5, no. 5, pp. 43–47, May feedback systems via sequential smooth optimization on the grassmannian
2015. manifold,” IEEE Trans. Signal Process., vol. 62, no. 5, pp. 1305–1318,
[116] J. Huang, K. Peng, C. Pan, F. Yang, and H. Jin, “Scalable video Mar. 2014.
broadcasting using bit division multiplexing,” IEEE Trans. Broadcast., [142] 3GPP, “Non-orthogonal multiple access for New Radio,” TSG RAN
vol. 60, no. 4, pp. 701–706, Dec. 2014. WG1 #85, Nanjing, China, 23rd-27th, May. 2016.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

29

[143] 3GPP, “New uplink non-orthogonal multiple access schemes for NR,” tional Symposium on Broadband Multimedia Systems and Broadcasting
TSG RAN WG1 #86, Gothenburg, Sweden, 22nd-26th, Aug. 2016. (IEEE BMSB’15), Jun. 2015, pp. 1–5.
[144] 3GPP, “Categorization and analysis of MA schemes,” TSG RAN WG1 [167] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, “System-
Meeting #86bis, Lisbon, Portugal, 10th-14th, Oct. 2016. level performance evaluation of downlink non-orthogonal multiple access
[145] B. Wang, L. Dai, Y. Yuan, and Z. Wang, “Compressive sensing (NOMA),” in Proc. IEEE Annu. Symp. PIMRC, London, U.K., Sep. 2013,
based multi-user detection for uplink grant-free non-orthogonal multiple pp. 611–615.
access,” in Proc. IEEE Vehicular Technology Conference (IEEE VTC’15 [168] Y. Saito, A. Benjebbour, Y. Kishiyama, and T. Nakamura, “System-
Fall), Sep. 2015, pp. 1–5. level performance evaluation of downlink non-orthogonal multiple access
[146] B. Wang, L. Dai, T. Mir, and Z. Wang, “Joint user activity and data (NOMA) under various environments,” in Proc. IEEE Vehicular Technol-
detection based on structured compressive sensing for NOMA,” IEEE ogy Conference (IEEE VTC-Spring’ 15), May. 2015, pp. 1–5.
Commun. Lett., vol. 20, no. 7, pp. 1473-1476, Jul. 2016. [169] K. Saito, A. Benjebbour, A. Harada,Y. Kishiyama, and T. Nakamu-
[147] V. Shah-Mansouri, S. Duan, L.-H. Chang, V. W. Wong, and J.-Y. Wu, ra, “Link-level performance evaluation of downlink NOMA with SIC
“Compressive sensing based asynchronous random access for wireless receiver considering error vector magnitude,” in Proc. IEEE Vehicular
networks,” in Proc. IEEE Wireless Communications and Networking Technology Conference (IEEE VTC-Spring’ 15), May. 2015, pp. 1–5.
Conference (IEEE WCNC’13), Apr. 2013, pp. 884–888. [170] K. Saito, A. Benjebbour, Y. Kishiyama, Y. Okumura, and T. Nakamura,
[148] G. Wunder, P. Jung, and C. Wang, “Compressive random access for “Performance and design of SIC receiver for downlink NOMA with
post-LTE systems,” in Proc. IEEE International Conference on Commu- open-Loop SU- MIMO,” in Proc. IEEE International Conference on
nications Workshops (IEEE ICC’14), Jun. 2014, pp. 539–544. Communication Workshop (IEEE ICCW’ 15), June. 2015, pp. 1161–1165.
[149] J.-P. Hong, W. Choi, and B. D. Rao, “Sparsity controlled random mul- [171] A. Benjebbour, K. Saito, Y. Saito, and Y. Kishiyama, “5G radio access
tiple access with compressed sensing,” IEEE Trans. Wireless Commun., technology,” NTT DOCOMO Technical Journal, vol. 17, no. 4, pp. 16–28,
vol. 14, no. 2, pp. 998–1010, Feb. 2015. 2015.
[150] “Components of a new air interface - building blocks and performance,” [172] L. Lu, Y. Chen, W. Guo, H. Yang, Y. Wu, and S. Xing, “Prototype for
Mobile and wireless communications Enablers for the Twenty-twenty 5G new air interface technology SCMA and performance evaluation”,
Information Society METIS, Mar. 2014. China Communications, vol. 12, no. supplement, pp. 38–48, Dec. 2015.
[151] 3GPP, “Initial LLS results for UL non-orthogonal multiple access,” [173] Y. Tan, J. Zhou, and J. Qin, “Novel channel estimation for non-
TSG RAN WG1 #85, Nanjing, China, 23rd-27th, May. 2016. orthogonal multiple access systems,” IEEE Signal Process. Lett., vol. 23,
[152] 3GPP, “Multiple access schemes for new radio interface,” TSG RAN no. 12, pp. 1781-1785, Dec. 2016.
WG1 #84bis, Busan, South Korea, 11th-15th, Apr. 2016. [174] K. Struminsky, S. Kruglik, D. Vetrov, and I. Oseledets, “A new ap-
[153] M. A. Sedaghat and R. R. Mĺźller, “On user pairing in NOMA uplink”, proach for sparse Bayesian channel estimation in SCMA uplink systems,”
https://arxiv.org/abs/1707.01846, Jul. 2017. in Proc. IEEE International Conference on Wireless Communications &
Signal Processing (IEEE WCSP’16), Oct. 2016, pp. 1-5.
[154] A. Roth and M. Sotomanyor, Two Sided Matching: A Study in Game-
[175] Z. Ding, F. Adachi, and H. V. Poor, “The application of MIMO to
Theoretic Modeling and Analysis. 1st ed. Cambridge, U.K.: Cambridge
non-orthogonal multiple access,” IEEE Trans. Wireless Commun., vol. 15,
University. Press, 1989.
no. 1, pp. 537–552, Sep. 2015.
[155] Y. Gu, W. Saad, M. Bennis, M. Debbah, and Z. Han, “Matching
theory for future wireless networks: Fundamentals and applications,”
IEEE Commun. Mag., vol. 53, no. 5, pp. 52-59, May 2015.
[156] W. Liang, Z. Ding, Y. Li, and L. SongH, “User pairing for downlink
non-orthogonal multiple access networks using matching algorithm,”
IEEE Trans. Commun., vol. 65, no. 12, pp. 5319-5332, Dec. 2017.
[157] J. Zhao, Y. Liu, K. K. Chai, A. Nallanathan, Y. Chen, and Z. Han,
“Spectrum allocation and power control for nonorthogonal multiple access
in HetNets,” IEEE Trans. Wireless Commun., vol. 16, no. 9, pp. 5825-
5837, Sep. 2017. Linglong Dai (M’11-SM’14) received the B.S. de-
[158] F. Fang, H. Zhang, J. Cheng, and V. C. M. Leung, “Energy-efficient gree from Zhejiang University in 2003, the M.S.
resource allocation for downlink non-orthogonal multiple access (NOMA) degree (with the highest honor) from the China A-
network,” IEEE Trans. Commun., vol. 64, no. 9, pp. 3722-3732, Jul. 2016. cademy of Telecommunications Technology in 2006,
[159] B. Di, S. Bayat, L. Song, and Y. Li, “Radio resource allocation and the Ph.D. degree (with the highest honor) from
for downlink non-orthogonal multiple access (NOMA) networks using Tsinghua University, Beijing, China, in 2011. From
matching theory,” in Proc. IEEE Global Communications Conference 2011 to 2013, he was a Post-Doctoral Research
(IEEE Globecom’15), Dec. 2015, pp. 1-6. Fellow with the Department of Electronic Engineer-
[160] H. Tabassum, M. S. Ali, E. Hossain, M.. J. Hossain, and Dong ing, Tsinghua University, where he was an Assistant
In Kim, “Non-orthogonal multiple access (NOMA) in cellular u- Professor from 2013 to 2016 and has been an
plink and downlink: Challenges and enabling techniques”, http- Associate Professor since 2016. He coauthored the
s://arxiv.org/abs/1608.05783, Aug. 2016. book “mmWave Massive MIMO: A Paradigm for 5G” (Academic Press,
[161] F. Fang, H. Zhang, J. Cheng, and V. C. M. Leung, “Energy-efficient re- Elsevier, 2016). He has published over 60 IEEE journal papers and over
source allocation for downlink non-orthogonal multiple access network,” 40 IEEE conference papers. He also holds 15 granted patents. His current
IEEE Trans. Commun., vol. 64, no. 9, pp. 3722-3732, Sep. 2016. research interests include massive MIMO, millimeter-wave communications,
[162] F. Fang, H. Zhang, J. Cheng, S. Roy, and V. C. M. Leung, “Joint user NOMA, sparse signal processing, and machine learning. He has received four
scheduling and power allocation optimization for energy-efficient NOMA conference Best Paper Awards at the IEEE ICC 2013, the IEEE ICC 2014,
systems with imperfect CSI,” IEEE J. Sel. Areas Commun., vol. 35, no. the IEEE ICC 2017, and the IEEE VTC 2017-Fall. He has also received the
12, pp. 2874-2885, Dec. 2017. Tsinghua University Outstanding Ph.D. Graduate Award in 2011, the Beijing
[163] Q. Sun, S. Han, C. L. I, and Z. Pan, “Energy efficiency optimization for Excellent Doctoral Dissertation Award in 2012, the China National Excellent
fading MIMO non-orthogonal multiple access systems,” in Proc. IEEE Doctoral Dissertation Nomination Award in 2013, the URSI Young Scientist
International Conference on Communications (IEEE ICC’15), Jun. 2015, Award in 2014, the IEEE Transactions on Broadcasting Best Paper Award in
pp. 2668-2673. 2015, the Second Prize of Science and Technology Award of China Institute
[164] Y. Zhang, Q. Yang, T. X. Zheng, H. M. Wang, Y. Ju, and Y. Meng, “En- of Communications in 2016, the Electronics Letters Best Paper Award in
ergy efficiency optimization in cognitive radio inspired non-orthogonal 2016, the IEEE Communications Letters Exemplary Editor Award in 2017, the
multiple access,” in Proc. IEEE Annual International Symposium on National Natural Science Foundation of China for Outstanding Young Scholars
Personal, Indoor, and Mobile Radio Communications (IEEE PIMRC’16), in 2017, and the IEEE ComSoc Asia-Pacific Outstanding Young Researcher
Sep. 2016, pp. 1-6. Award in 2017. He currently serves as an Editor of the IEEE Transactions
[165] Y. Xu, C. Shen, Z. Ding, X. Sun, S. Yan, G. Zhu, and Z. Zhong, on Communications, the IEEE Transactions on Vehicular Technology, and the
“Joint beamforming and power splitting control in downlink cooperative IEEE Communications Letters.
swipt noma systems,” IEEE Trans. Signal Process., vol. 15, no. 18, pp.
4874-4886, Sep. 2017.
[166] B. Wang, K. Wang, Z. Lu, T. Xie, and J. Quan, “Comparison study of
non-orthogonal multiple access schemes for 5G,” in Proc. IEEE Interna-

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/COMST.2018.2835558, IEEE
Communications Surveys & Tutorials

30

Bichai Wang (S’15) received her B.S. degree in Sheng Chen (M’90-SM’97-F’08) received his BEng
Electronic Engineering from Tsinghua University, degree from the East China Petroleum Institute,
Beijing, China, in 2015. She is currently working Dongying, China, in 1982, and his PhD degree from
towards the Ph.D. degree in the Department of Elec- the City University, London, in 1986, both in control
tronic Engineering, Tsinghua University, Beijing, engineering. In 2005, he was awarded the higher
China. Her research interests are in wireless com- doctoral degree, Doctor of Sciences (DSc), from
munications, with the emphasis on non-orthogonal the University of Southampton, Southampton, UK.
multiple access, mmWave massive MIMO, and deep From 1986 to 1999, He held research and academic
learning-based wireless communications. She has appointments at the Universities of Sheffield, Edin-
received the Freshman Scholarship of Tsinghua U- burgh and Portsmouth, all in UK. Since 1999, he has
niversity in 2011, the Academic Merit Scholarships been with the School of Electronics and Computer
of Tsinghua University in 2012, 2013, and 2014, respectively, the Excellent Science, the University of Southampton, UK, where he holds the post of
Thesis Award of Tsinghua University in 2015, the National Scholarship in Professor in Intelligent Systems and Signal Processing. Dr Chen’s research
2016, the IEEE VTC’17 Fall Best Student Paper Award in 2017, and the interests include adaptive signal processing, wireless communications, mod-
IEEE Transactions on Communications Exemplary Reviewer Award in 2017. elling and identification of nonlinear systems, neural network and machine
learning, intelligent control system design, evolutionary computation methods
and optimisation. He has published over 600 research papers. Dr. Chen is a
Fellow of the United Kingdom Royal Academy of Engineering, a Fellow of
IET, a Distinguished Adjunct Professor at King Abdulaziz University, Jeddah,
Saudi Arabia, and an ISI highly cited researcher in engineering (March 2004).
Professor Chen has 11,400 Web of Science citations and 24,000+ Google
Scholar citations.
Zhiguo Ding (S’03-M’05) received his B.Eng in
Electrical Engineering from the Beijing University
of Posts and Telecommunications in 2000, and from
Jul. 2005 to Apr. 2018, he was working in Queen’s
University Belfast, Imperial College, Newcastle U-
niversity and Lancaster University. Since Apr. 2018,
he has been with the University of Manchester as
a Professor in Communications. From Oct. 2012 to
Sep. 2018, he has also been an academic visitor in
Princeton University. Dr Ding’ research interests are
5G networks, game theory, cooperative and energy
harvesting networks and statistical signal processing. He is serving as an
Editor for IEEE Transactions on Communications, IEEE Transactions on
Vehicular Technology, and Journal of Wireless Communications and Mobile
Computing, and was an Editor for IEEE Wireless Communication Letters,
IEEE Communication Letters from 2013 to 2016. He received the best paper
award in IET ICWMC-2009 and IEEE WCSP-2014, the EU Marie Curie Lajos Hanzo (M’91-SM’92-F’04) FREng, FIEEE,
Fellowship 2012-2014, the Top IEEE TVT Editor 2017 and IEEE Heinrich FIET, Fellow of EURASIP, DSc received his degree
Hertz Award 2018. in electronics in 1976 and his doctorate in 1983. In
2009 he was awarded an honorary doctorate by the
Technical University of Budapest and in 2015 by the
University of Edinburgh. In 2016 he was admitted
to the Hungarian Academy of Science. During his
40-year career in telecommunications he has held
various research and academic posts in Hungary,
Zhaocheng Wang (M’09-SM’11) received his B.S., Germany and the UK. Since 1986 he has been with
M.S. and Ph.D. degrees from Tsinghua University the School of Electronics and Computer Science,
in 1991, 1993 and 1996, respectively. From 1996 to University of Southampton, UK, where he holds the chair in telecommu-
1997, he was a Post-Doctoral Fellow with Nanyang nications. He has successfully supervised 111 PhD students, co-authored 18
Technological University, Singapore. From 1997 to John Wiley/IEEE Press books on mobile radio communications totalling in
1999, he was a Research Engineer/Senior Engineer excess of 10 000 pages, published 1741 research contributions at IEEE Xplore,
with OKI Techno Centre Pte. Ltd., Singapore. From acted both as TPC and General Chair of IEEE conferences, presented keynote
1999 to 2009, he was a Senior Engineer/Principal lectures and has been awarded a number of distinctions. Currently he is
Engineer with Sony Deutschland GmbH, Germany. directing a 60-strong academic research team, working on a range of research
Since 2009, he has been a Professor with De- projects in the field of wireless multimedia communications sponsored by
partment of Electronic Engineering, Tsinghua Uni- industry, the Engineering and Physical Sciences Research Council (EPSRC)
versity, where he is currently the Director of Broadband Communication UK, the European Research Council’s Advanced Fellow Grant and the Royal
Key Laboratory, Beijing National Research Center for Information Science Society’s Wolfson Research Merit Award. He is an enthusiastic supporter of
and Technology (BNRist). His research interests include millimeter wave industrial and academic liaison and he offers a range of industrial courses.
communications, optical wireless communications and digital broadcasting. He is also a Governor of the IEEE ComSoc and VTS. During 2008-2012 he
Prof. Wang hold 34 US/EU granted patents (23 of them as the first inventor) was the Editor-in-Chief of the IEEE Press and a Chaired Professor also at
and published more than 135 peer-reviewed international journal papers. He Tsinghua University, Beijing. For further information on research in progress
authored or co-authored two books, which have been selected by IEEE Series and associated publications please refer to http://www-mobile.ecs.soton.ac.uk.
on Digital & Mobile Communication and published by Wiley-IEEE Press.
Prof. Wang is a Fellow of the Institution of Engineering and Technology. He
received ICC2013 Best Paper Award, OECC2015 Best Student Paper Award,
2016 IEEE Scott Helt Memorial Award, 2016 National Award for Science and
Technology Progress (First Prize) and ICC2017 Best Paper Award. He was
an Associate Editor of IEEE Transactions on Wireless Communications from
2011 to 2015 and an Associate Editor of IEEE Communications Letters from
2013 to 2016. He was also technical program co-chairs of many international
conferences including ICC and GlobeSIP.

1553-877X (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.