Non-Orthogonal Multiple Access for 5G and

Beyond
Proceedings of the IEEE, Dec. 2017

Yuanwei Liu, Zhijin Qin, Maged Elkashlan, Zhiguo Ding,

Arumugam Nallanathan and Lajos Hanzo

Dec. 2017

1 / 92
Outline

1 Overview and Motivation

2 Power-Domain NOMA Basics

3 Sustainability of NOMA Networks

4 Compatibility of NOMA in 5G Networks

5 Security Issues in NOMA Networks

6 Other Research Contributions on NOMA

7 Research Opportunities and Challenges for NOMA

2 / 92
Brief History of Wireless Standardization

MFAA
MIMO MFAA St.
LS-MIMO
BF Close Sq. Terrace

4G
Sq.
5G
OMA/ Place
NOMA
OVSF-CDMA St.

Telepr. Ave.
Sq.

HetNets
CR

MPEG St.
SDN
Turbo St. Sq.
FEC
LDPC St. Sq.
UL/DL decoupling St.
BICM-ID St.

3 / 92
Future 5G network architecture.

IoT Health ... Safety VR

Applications
Telco API

Virtualization Software defined networking controller

Forwarding

• Ultra Wideband
(cmWave, mmWave)
Cloud RAN
Macro Massive
IoT
...
cell MIMO

f Fronthaul
• NOMA
M2M
Power
…… D2D

Small cells V2V

f Radio access unit

[1] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo, “Non-Orthogonal Multiple Access for

5G”, Proceedings of the IEEE ; Dec 2017. 4 / 92

From OMA to NOMA

1 Question: What is multiple access?

2 Orthogonal multiple access (OMA): e.g., FDMA, TDMA,
CDMA, OFDMA.
3 New requirements in 5G
High spectrum efficiency.
Massive connectivity.
4 Non-orthogonal multiple access (NOMA): to break
orthogonality.
5 Standard and industry developments on NOMA
Whitepapers for 5G: DOCOMO, METIS, NGMN, ZTE, SK
Telecom, etc.
LTE Release 13: a two-user downlink special case of NOMA.
Next generation digital TV standard ATSC 3.0: a variation
of NOMA, termed Layer Division Multiplexing (LDM).

5 / 92
Introduction to NOMA Systems

The non-orthogonal nature of a multiple access system may

manifest itself in the time-, frequency-, code- or
spatial-domains as well as in their arbitrary combinations;
Even if originally an OMA scheme is used, the deleterious
effects of the wireless channel may erode the orthogonality.
For example, the channel-induced dispersion may ’smear’ the
originally orthogonal time-slots of a TDMA system into each
other, because the transmitted signal is convolved with the
dispersive channel’s impulse response (CIR).
Similarly, the Orthogonal Variable Spreading Factor (OVSF)
codes of the 3G systems rely on orthogonal Walsh-Hadamard
codes, but upon transmission over the dispersive channel their
orthogonality is destroyed.

6 / 92
Introduction to NOMA Systems

This realization has then led to the concept of NOMA based

on the Spatial Division Multiple Access (SDMA) philosophy,
where the unique, user-specific non-orthogonal channel
impulse responses are used for distinguishing the uplink
transmissions of the users - provided that their CIR is
estimated sufficiently accurately.
In simple tangible terms this implies that a NOMA system is
capable of supporting more users than the number of distinct
time-, frequency-, code-domain resources, provided that their
channels can be sufficiently accurately estimated even under
these challenging interference-contaminated conditions.
Naturally, this challenging channel estimation and
user-separation process typically imposes an increased signal
processing complexity.
Many of these NOMA-user-separation techniques are surveyed
in this paper, with a special emphasis on the power-domain 7 / 92
Power-Domain NOMA Basics

Time User m Subtract user User n

detection m’s signal detection
User n
Power

SIC
BS
Superimposed signal of
User m User m and n User m
0detection
User m
User n
Frequency

1 Supports multiple access within a given resource block

(time/frequecy/code), using different power levels for
distinguishing/separating them [1].
2 Apply successive interference cancellation (SIC) at the
receiver for separating the NOMA users [2].
3 If their power is similar, PIC is a better alternative.
[1] Y. Liu, Z. Qin, M. Elkashlan, Z. Ding, A. Nallanathan, and L. Hanzo, “Non-Orthogonal Multiple Access for
5G”, Proceedings of the IEEE ; Dec 2017.

[2] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, Chih-Lin I, and H. V. Poor (2017), “Application of

Non-orthogonal Multiple Access in LTE and 5G Networks”, IEEE Communication Magazine.

8 / 92
NOMA Basics

1 Question: Why NOMA is a popular proposition for 5G?

2 Consider the following two scenarios.
If a user has poor channel conditions
The bandwidth allocated to this user via OMA cannot be used
at a high rate.
NOMA - improves the bandwidth-efficiency.
If a user only needs a low data rate, e.g. IoT networks.
The use of OMA gives the IoT node more capacity than it
needs.
NOMA - heterogeneous QoS and massive connectivity.

[1] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, Chih-Lin I, and H. V. Poor (2017), “Application of

Non-orthogonal Multiple Access in LTE and 5G Networks”, IEEE Communication Magazine.

9 / 92
NOMA Basics

1 Question: Why NOMA is a popular proposition for 5G?

2 Consider the following two scenarios.
If a user has poor channel conditions
The bandwidth allocated to this user via OMA cannot be used
at a high rate.
NOMA - improves the bandwidth-efficiency.
If a user only needs a low data rate, e.g. IoT networks.
The use of OMA gives the IoT node more capacity than it
needs.
NOMA - heterogeneous QoS and massive connectivity.

[1] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, Chih-Lin I, and H. V. Poor (2017), “Application of

Non-orthogonal Multiple Access in LTE and 5G Networks”, IEEE Communication Magazine.

9 / 92
NOMA Basics

1 Question: Why NOMA is a popular proposition for 5G?

2 Consider the following two scenarios.
If a user has poor channel conditions
The bandwidth allocated to this user via OMA cannot be used
at a high rate.
NOMA - improves the bandwidth-efficiency.
If a user only needs a low data rate, e.g. IoT networks.
The use of OMA gives the IoT node more capacity than it
needs.
NOMA - heterogeneous QoS and massive connectivity.

[1] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, Chih-Lin I, and H. V. Poor (2017), “Application of

Non-orthogonal Multiple Access in LTE and 5G Networks”, IEEE Communication Magazine.

9 / 92
NOMA Basics

1 Question: Why NOMA is a popular proposition for 5G?

2 Consider the following two scenarios.
If a user has poor channel conditions
The bandwidth allocated to this user via OMA cannot be used
at a high rate.
NOMA - improves the bandwidth-efficiency.
If a user only needs a low data rate, e.g. IoT networks.
The use of OMA gives the IoT node more capacity than it
needs.
NOMA - heterogeneous QoS and massive connectivity.

[1] Z. Ding, Y. Liu, J. Choi, Q. Sun, M. Elkashlan, Chih-Lin I, and H. V. Poor (2017), “Application of

Non-orthogonal Multiple Access in LTE and 5G Networks”, IEEE Communication Magazine.

9 / 92
Research Contributions in NOMA

Compatibility

NOMA for 5G

Security Sustainability

10 / 92
From NOMA to Cooperative NOMA

NOMA can pair a user having better channel conditions with

another user having worse channel conditions and then detect
them using SIC. For example, consider a downlink scenario in
which there are two groups of users:
Cell-centre users: close to the base station (BS) and have
better channel conditions.
Cell-edge users: close to the edge of the cell controlled by
the BS and therefore have worse channel conditions.
While the bandwidth efficiency of NOMA is superior to OMA, the
fact that the near users co-exist with the far users causes
performance degradation to the far users. This motivates us to
invoke cooperative NOMA.
But again, the cell-edge user suffers from some performance
erosion in NOMA
The cell-centre user may infer the information sent to the
cell-edge user. 11 / 92
What is Cooperative NOMA?

Solution – Cooperative NOMA

3 time slots are needed for cooperative OMA, while
cooperative NOMA only needs 2.
Cooperative NOMA: cell-centre users act as relays to help
the cell-edge users having poor channel conditions.

SIC of User A User B signal

signal detection
Base Station
User B

User A signal
Non-cooperative NOMA detection
Cooperative NOMA User A

Advantages: SIC is used and hence the information of the

cell-edge users is known by the cell-centre users, which may
act as DF relays.
12 / 92
A Simple Example (1/3)

Consider a NOMA downlink with two users.

Time slot I: BS sends the superimposed messages to both
users
Time slot II: The user with strong channel conditions is to
help its partner by acting as a relay
Simulation parameters are set as follows:
The BS is located at (0, 0).
User 2 is located at (5m, 0).
The x -y plane denotes the location of User 1.
A bounded path loss model is used to ensure all distances are
greater than one. The path loss exponent is 3.
The transmit signal-to-noise ratio (SNR) is 30 dB.
The power allocation coefficient for User 2 and User 1 are
(aA , aB ) = 45 , 15 .

The targeted data rate is 0.5 bits per channel use (BPCU).

13 / 92
A Simple Example (2/3 – Overall Outage)

10 0
Outage Probability

10 -1

10 -2

OMA
non-cooperative NOMA
cooperative NOMA
10 -3
0 10 20 30 40
SNR
14 / 92
A Simple Example (3/3 – Overall Outage)
Outage probability of the poor user

-1
10

10-2

10-3
2 Non-cooperative NOMA
Cooperative NOMA 2
0 1
0
y -1
-2 -2 x

15 / 92
SWIPT—Background (1/2)

Wireless energy Transfer (WET)

Key Idea: Energy is transmitted from a power source to a
destination over the wireless medium.
Motivation: 1) Ambient radio frequency signals are
everywhere; 2) WET could be the only means of increasing
useful lifetime of energy constrained networks.
Tesla had already provided a successful demonstration of
lighting an electric bulb wirelessly in 1891, but WET has been
forgotten owing to its low energy efficiency.
What has changed then?
We have numerous low-power devices.
Advanced energy-beamformers have become available.
[1] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor (2016), “Cooperative Non-orthogonal Multiple Access with

Simultaneous Wireless Information and Power Transfer”, IEEE Journal on Selected Areas in Communications

(JSAC).
16 / 92
SWIPT—Background (2/2)

i T
Energy Energy
Harvesting Harvesting
! ! i "T
Tx
Tx
j T
Information
Information Decoding
Decoding ! ! "T j

(a) Separated Receiver (b) Time Switching Receiver

Power Energy
Harvesting Energy
Splitting
Harvesting
! i

Tx Tx
j
Information
Power
Splitting
Information Decoding
Decoding
! j

(c) Power Splitting Receiver (d) Antenna Switching Receiver

[1]Z. Ding, C. Zhong, D. W. Ng, M. Peng, H. A. Suraweera, R. Schober and H. V. Poor, Application of Smart
Antenna Technologies in Simultaneous Wireless Information and Power Transfer, IEEE Commun. Magazine, 2015.
17 / 92
Sustainability of NOMA Networks

1 Transmission reliability - cooperative NOMA.

2 Energy consumption - radio signal energy harvesting.
SIC Procedure

User B

Base Station Energy flow

Direct Information flow
Cooperative information flow

User A

3 Propose a wireless powered cooperative NOMA protocol [1].

4 The first contribution on wirelessly powered NOMA networks.
[1] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor (2016), “Cooperative Non-orthogonal Multiple Access with
Simultaneous Wireless Information and Power Transfer”, IEEE Journal on Selected Areas in Communications
(JSAC).

18 / 92
Sustainability of NOMA Networks

1 Transmission reliability - cooperative NOMA.

2 Energy consumption - radio signal energy harvesting.
SIC Procedure

User B

Base Station Energy flow

Direct Information flow
Cooperative information flow

User A

3 Propose a wireless powered cooperative NOMA protocol [1].

4 The first contribution on wirelessly powered NOMA networks.
[1] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor (2016), “Cooperative Non-orthogonal Multiple Access with
Simultaneous Wireless Information and Power Transfer”, IEEE Journal on Selected Areas in Communications
(JSAC).

18 / 92
Sustainability of NOMA Networks

1 Transmission reliability - cooperative NOMA.

2 Energy consumption - radio signal energy harvesting.
SIC Procedure

User B

Base Station Energy flow

Direct Information flow
Cooperative information flow

User A

3 Propose a wireless powered cooperative NOMA protocol [1].

4 The first contribution on wirelessly powered NOMA networks.
[1] Y. Liu, Z. Ding, M. Elkashlan, and H. V. Poor (2016), “Cooperative Non-orthogonal Multiple Access with
Simultaneous Wireless Information and Power Transfer”, IEEE Journal on Selected Areas in Communications
(JSAC).

18 / 92
Motivation for SWIPT + Cooperative NOMA

To improve the reliability of the distant NOMA users without

draining the near users’ batteries, we consider the application
of SWIPT to NOMA, where SWIPT is performed at the near
NOMA users.
Therefore, the aforementioned pair of communication
concepts, namely cooperative NOMA and SWIPT, can be
naturally linked together.
Cooperative SWIPT NOMA – a new spectral-efficient and
energy-efficient wireless multiple access protocol.

19 / 92
Network Model

Illustration of a downlink
SWIPT NOMA system
RDA RDC ≫ RDB
A1 with a base station S (blue
A6
circle). The spatial
RDC B5
B4
RDB B1
distributions of the near
A3 B6 ... S ... B3 users (yellow circles) and
A2
B2 hB
the far users (green circles)
i

Bi
hA i
obey a homogeneous
gi Poisson Point Process
A5
A4 (PPP).
... ...
Ai

Direct Transmission Phase with SWIPT

Cooperative Tansmission Phase

20 / 92
Network Model

The locations of the near and far users are modeled as

homogeneous PPPs Φκ (κ ∈ {A, B}) with densities λΦκ .
The near users are uniformly distributed within the disc and
the far users are uniformly distributed within the ring.
The users in {Bi } are energy harvesting relays that harvest
energy from the BS and forward the information to {Ai } using
the harvested energy as their transmit powers.
The DF strategy is applied at {Bi } and the cooperative
NOMA system consists of two phases.
It is assumed that the two phases have the same transmission
periods.

21 / 92
Phase 1: Direct Transmission

During the first phase, the BS sends two messages

pi1 xi1 + pi2 xi2 to two selected users Ai and Bi based on
NOMA, where pi1 and pi2 are the power allocation
coefficients and xi1 and xi2 are the messages of Ai and Bi ,
respectively. The observation at Ai is given by
p X hAi
yAi ,1 = PS pik xik q + nAi ,1 . (1)
k∈{1,2} 1 + dAαi

Without loss of generality, we assume that |pi1 |2 > |pi2 |2 with

|pi1 |2 + |pi2 |2 = 1. The received signal to interference plus
noise ratio (SINR) at Ai to detect xi1 is given by
xi1 ρ|hAi |2 |pi1 |2
γS,A = , (2)
i
ρ|pi2 |2 |hAi |2 + 1 + dAαi
PS
where ρ = σ2
is the transmit signal to noise ratio (SNR).
22 / 92
Phase 1: Direct Transmission

We assume that the near users have rechargeable batteries

and that power splitting is applied to perform SWIPT. Thus,
the observation at Bi is given by
√
p X 1 − βi hBi
yBi ,1 = PS pik xik q + nBi ,1 , (3)
k∈{1,2} 1 + dBαi
where βi is the power splitting coefficient.
The receiver’s SINR at Bi used for detecting xi1 of Ai is
xi1 ρ|hBi |2 |pi1 |2 (1 − βi )
γS,B = . (4)
i
ρ|hBi |2 |pi2 |2 (1 − βi ) + 1 + dBαi
The receiver’s SNR at Bi used for detecting xi2 of Bi is
xi2 ρ|hBi |2 |pi2 |2 (1 − βi )
γS,B = . (5)
i
1 + dBαi
23 / 92
Phase 1: Direct Transmission

Based on (4), the data rate supported by the channel from

the BS to Bi for decoding xi1 is given by
ρ|hBi |2 |pi1 |2 (1 − βi )
!
1
Rxi1 = log 1 + . (6)
2 ρ|hBi |2 |pi2 |2 (1 − βi ) + 1 + dBαi
In order to ensure that Bi can successfully decode the
information of Ai , we have a rate, i.e., R1 = Rxi1 . Therefore,
the power splitting coefficient is set as follows:
   
 τ1 1 + dBαi 
βi = max 0, 1 −   , (7)
 ρ |pi1 |2 − τ1 |pi2 |2 |hBi |2 

where τ1 = 22R1 − 1. Here βi = 0 means that all the energy is

used for information decoding and no energy remains for
energy harvesting.
24 / 92
Phase 1: Direct Transmission

Based on (3), the energy harvested at Bi is given by

T ηPS βi |hBi |2
EBi =   , (8)
2 1 + dBαi

where T is the time period for the entire transmission

including the direct transmission phase and the cooperative
transmission phase, and η is the energy harvesting coefficient.
We assume that the two phases have the same transmission
period, and therefore, the transmit power at Bi can be
expressed as follows:

ηPS βi |hBi |2
Pt = . (9)
1 + dBαi

25 / 92
Phase 2: Cooperative Transmission

During this phase, Bi forwards xi1 to Ai by using the

harvested energy during the direct transmission phase. In this
case, Ai observes
√
Pt xi1 gi
yAi ,2 = q + nAi ,2 . (10)
1 + dCαi

Based on (9) and (10), the received SNR for Ai to detect xi1
forwarded from Bi is given by

xi1 Pt |gi |2 ηρβi |hBi |2 |gi |2

γA i ,Bi
=  =  . (11)
1 + dCαi σ 2 1 + dCαi 1 + dBαi

26 / 92
Phase 2: Cooperative Transmission

At the end of this phase, Ai combines the signals from the BS

and Bi using maximal-ratio combining (MRC).
Combining the SNR of the direct transmission phase (2) and
the SINR of the cooperative transmission phase (11), we
obtain the received SINR at Ai as follows:

xi1 ρ|hAi |2 |pi1 |2 ηρβi |hBi |2 |gi |2

γA i ,MRC
= + .
ρ|hAi |2 |pi2 |2 + 1 + dAαi
 
1 + dBαi 1 + dCαi
(12)

27 / 92
Non-Orthogonal Multiple Access with User Selection

A natural question arises: which specific near NOMA user

should help which particular far NOMA user?
To investigate the performance of a specific pair of selected
NOMA users, three opportunistic user selection schemes may
be considered, based on the particular locations of users to
perform NOMA as follows:
random near user and random far user (RNRF) selection,
where both the near and far users are randomly selected from
the two groups.
nearest near user and nearest far user (NNNF) selection, where
a near user and a far user closest to the BS are selected from
the two groups.
nearest near user and farthest far user (NNFF) selection, where
a near user which is closest to the BS is selected and a far user
which is farthest from the BS is selected.
28 / 92
RNRF Selection Scheme—Outline

This selection scheme provides a fair opportunity for each user

to access the source with the aid of the NOMA protocol.
Advantage: it does not require the knowledge of
instantaneous channel state information (CSI).
1 Outage Probability of the Near Users of RNRF
2 Outage Probability of the Far Users of RNRF
3 Diversity Analysis of RNRF
4 System Throughput in Delay-Sensitive Transmission Mode of
RNRF

29 / 92
Outage Probability of the Near Users of RNRF

An outage of Bi can occur for two reasons.

1 Bi cannot detect xi1 .
2 Bi can detect xi1 but cannot detect xi2 .
Based on this, the outage probability of Bi can be expressed
as follows:

ρ|hBi |2 |pi1 |2
!
PBi = Pr < τ1
ρ|hBi |2 |pi2 |2 + 1 + dBαi
ρ|hBi |2 |pi1 |2
!
xi2
+ Pr > τ1 , γS,B < τ2 . (13)
ρ|hBi |2 |pi2 |2 + 1 + dBαi i

30 / 92
Outage Probability of the Far Users of RNRF

Outage experienced by Ai can occur in two situations.

1 Bi can detect xi1 but the overall received SNR at Ai cannot
support the targeted rate.
2 Neither Ai nor Bi can detect xi1 .
Based on this, the outage probability can be expressed as follows:
 
xi1 xi1
PAi = Pr γA i ,MRC
< τ1 , γS,B i β =0
> τ1
i
 
xi1 xi1
+ Pr γS,A i
< τ1 , γS,B i β =0
< τ1 . (14)
i

31 / 92
Diversity Analysis of RNRF—Near Users

The diversity gain is defined as follows:

log P (ρ)
d = − lim . (15)
ρ→∞ log ρ
Near users: When ε → 0, a high SNR approximation with
1 − e −x ≈ x is given by
N
1X
q
FYi (ε) ≈ ωN 1 − φn 2 cn εAi (φn + 1). (16)
2 n=1

Substituting (16) into (15), we obtain that the diversity gain

for the near users is one, which means that using NOMA with
energy harvesting will not decrease the diversity gain.

32 / 92
Diversity Analysis of RNRF—Far Users

Far users: For the far users, substituting (??) into (15), we obtain
 
log − ρ12 log ρ1
d = − lim
ρ→∞ log ρ
log log ρ − log ρ2
= − lim = 2. (17)
ρ→∞ log ρ
Remarks:
This result indicates that using NOMA with an energy
harvesting relay will not affect the diversity gain.
At high SNRs, the dominant factor for the outage probability
is ρ12 ln ρ.
The outage probability of using NOMA with SWIPT decays at
a rate of lnSNR
SNR
2 . However, for a conventional cooperative
system without energy harvesting, a faster decreasing rate of
1
SNR 2
can be achieved.
33 / 92
System Throughput in Delay-Sensitive Transmission Mode
of RNRF

In this mode, the transmitter sends information at a fixed rate

but the ’goodput’ becomes lower, as determined by the
outage probability.
As a result, the system throughput of RNRF in the
delay-sensitive transmission mode is given by

RτRNRF = (1 − PAi ) R1 + (1 − PBi ) R2 . (18)

34 / 92
NNNF Selection Scheme and NNFF Selection Scheme

Advantage of NNNF: it can minimize the outage probability

of both the near and far users.
Advantage of NNFF: NOMA can offer a larger performance
gain over conventional MA when user channel conditions are
more distinct.
Following a procedure similar to that of RNRF, we can obtain the
outage probability, diversity gain, and the throughput of NNNF
and NNFF.

35 / 92
Numerical Results

10 0 Lower outage probability is

Outage probability of the near users

R1 = 0.5, R2 = 1
(BPCU)
achieved than with RNRF.
10−1 NNN(F)F RNRF All curves have the same
R1 = R2 = 1
Simulation (BPCU) slopes, which indicates the
10−2 Incorrect choice of rate
RNRF analytical (α = 2) same diversity gains.
RNRF analytical-appro (α = 3)
10−3 RNRF analytical-appro (α = 4) The incorrect choice of
NNN(F)F analytical (α = 2)
NNN(F)F analytical-appro (α = 3) rate results in an outage
NNN(F)F analytical-appro (α = 4)
10−4
10 15 20 25 30 35 40 45 50
probability for the near
SNR (dB) users, which is always one.

36 / 92
Numerical Results

The outage of the near

Outage probability of the near users

users occurs more

10 0
frequently as the rate of
NNN(F)F
10−1
RNRF the far user, R1 , increases.
For the choice of R1 , it
10−2 should satisfy the condition
1.0 (|pi1 |2 − |pi2 |2 τ1 > 0).
10−3 0.5
1.0 0.8 0.6 0.4 0.2 0 0 R2 (BPCU) For the choice of R2 , it
R1 (BPCU)
should satisfy the
condition that the split
energy for detecting xi1 is
also sufficient to detect xi2
(εAi ≥ εBi ).
37 / 92
Numerical Results

10 0 NNNF achieves the lowest

Outage probability of the far users

10−1
α=3
outage probability.
NNFF achieves lower
10−2
outage than RNRF, which
10−3
RNRF simulation indicates that the distance
NNNF simulation
10−4
NNFF simulation of the near users has more
RNRF analytical-appro
10−5 NNNF analytical-appro α=2 impact than that of the far
NNFF analytical-appro
users.
10−6
10 15 20 25 30 35 40 45 50 All of the curves have the
SNR (dB)
same slopes, which
indicates that the diversity
gains of the far users are
the same.
38 / 92
Numerical Results

10 0 Cooperative NOMA has a

Outage probability of the far users

steeper slope than that of

10−1
non-cooperative NOMA.
10−2 NNNF achieves the lowest
RNRF Cooperative NOMA outage probability.
10−3 NNNF Cooperative NOMA
NNFF Cooperative NOMA
RNRF Non-cooperative NOMA
NNFF has higher outage
10−4 NNNF Non-cooperative NOMA probability than RNRF in
NNFF Non-cooperative NOMA

10−5
non-cooperative NOMA,
10 15 20 25 30 35 40 45 50 however, it achieves lower
SNR (dB)
outage probability than
RNRF in cooperative
NOMA.

39 / 92
Numerical Results

2.0 NNNF achieves the highest

System Throughput (BPCU)

1.8 throughput since it has the

1.6 R1 =1, R2 =1
1.4 (BPCU)
lowest outage probability.
R1 =1, R2 =0.5
1.2 (BPCU) The existence of the
1.0 throughput ceilings in the
0.8 R1 =1, R2 =2
0.6
(BPCU) high SNR region.
0.4 RNRF Increasing R2 from
0.2 NNNF

0
NNFF R2 = 0.5 BPCU to R2 = 1
10 15 20 25 30 35
SNR (dB)
40 45 50 BPCU can improve the
throughput; however, for
the case R2 = 2 BPCU,
the throughput is lowered.

40 / 92
NOMA in 5G Networks—HetNets

1 Heterogenous networks (HetNets): meet the requirements

of high data traffic in 5G.
Question: How to support massive connectivity in HetNets?
Question: How to further improve the spectral efficiency of
HetNets?

Pico BS
Femto BS
Marco BS

OMA

2 New framework: NOMA-enabled HetNets.

3 Challenge: Complex co-channel interference environment.

41 / 92
NOMA in 5G Networks—HetNets

1 Heterogenous networks (HetNets): meet the requirements

of high data traffic in 5G.
Question: How to support massive connectivity in HetNets?
Question: How to further improve the spectral efficiency of
HetNets?

Pico BS
Femto BS
Marco BS

OMA

2 New framework: NOMA-enabled HetNets.

3 Challenge: Complex co-channel interference environment.

41 / 92
NOMA in 5G Networks—HetNets

1 Heterogenous networks (HetNets): meet the requirements

of high data traffic in 5G.
Question: How to support massive connectivity in HetNets?
Question: How to further improve the spectral efficiency of
HetNets?

Pico BS
Femto BS
Marco BS

NOMA

2 New framework: NOMA-enabled HetNets.

3 Challenge: Complex co-channel interference environment.

41 / 92
NOMA in 5G Networks—HetNets

1 Heterogenous networks (HetNets): meet the requirements

of high data traffic in 5G.
Question: How to support massive connectivity in HetNets?
Question: How to further improve the spectral efficiency of
HetNets?

Pico BS
Femto BS
Marco BS

NOMA

2 New framework: NOMA-enabled HetNets.

3 Challenge: Complex co-channel interference environment.

41 / 92
NOMA in HetNets I — Resource Allocation

Fig.: System model.

K-tier HetNets: One macro base station (MBS), B small base stations
(SBSs)
M macro cell users (MCUs), M RBs, K small cell users (SCUs) served by
each SBS
Each SBS serves K SCUs simultaneously on the same RB via NOMA
[1] J. Zhao, Y. Liu, K. K. Chai, A. Nallanathan, Y. Chen and Z. Han (2017),“Spectrum Allocation and Power

Control for Non-Orthogonal Multiple Access in HetNets”, IEEE Transactions on Wireless Communications 42 / 92
Channel Model

Received signal at the k-th SCU, i.e., k ∈ {1, ..., K }, served by

the b-th SBS, i.e., b ∈ {1, ..., B}, on the m-th RB is given by
m √ √
XK
n m m m m
yb,k = fb,k pb ab,k xb,k + fb,k 0
pb ab,k 0 xb,k 0 + ζb,k
| {z } k =k |{z}
| {z }
desired signal interference from NOMA users noise
XM √ X
m √ m
+ λm,b hm,b,k pm xm + λb∗,b gb∗,b,k pb ∗ xb∗ .
| m=1 {z } |
b∗6=b
{z }
cross-tier interference co-tier interference
(19)
Received SINR:
2
m p am
fb,k b b,k
m
γb,k,k = , (20)
INk,k + Ico
k + I k + σ2
cr

where INk,k = |fb,k

m |2 p PK m
b i=k+1 ab,i
43 / 92
Problem Formulation

Maximize the sum-rate:

B X
X K X
M
m
max Rb,k (λ), (21a)
λ
b=1 k=1 m=1

s.t. λm,b ∈ {0, 1} , ∀m, b, (21b)

X
λm,b ≤ 1, ∀b, (21c)
m
X
λm,b ≤ qmax , ∀m, (21d)
b

Im ≤ Ithr , ∀m. (21e)

Solution:
NP-hard =⇒ High complexity
Solution: Many-to-one matching theory
44 / 92
Matching Model

Two-sided matching between SBSs and RBs

: “Preference” based on players’ utility
SBSs’ utility: sum-rate of all the serving SCUs minus its cost for
occupying RB m
K
X m
Ub = Rb,k − βpb |gb,m |2 , (22)
k=1

RBs’ utility: sum-rate of the occupying SCUs

B K
!
X X m 2
Um = λm,b Rb,k + βpb |gb,m | , (23)
b=1 k=1

45 / 92
Matching Model (cont’)

Peer effects among players’ preferences=⇒ Swap operations

Swap matching:

Φba = {Φ \ {(a, Φ(a)), (b, Φ(b))}} ∪

{(a, Φ(b)), (b, Φ(a))} . (24)

Φ: matching state
Swap-blocking pair (a, b) ⇔

1) ∀s ∈ {a, b, Φ(a), Φ(b)} , Us (Φba ) ≥ Us (Φ) and;

2) ∃s ∈ {a, b, Φ(a), Φ(b)}, such that Us (Φba ) > Us (Φ)

46 / 92
Matching Algorithm

Step 1: Initialization: GS algorithm to obtain initial matching state

Step 2: Swap operations: keep finding swap-blocking pairs —- until no

swap-blocking pair exists;

Flag SRa,b to record the time that SBS a and b swap their allocated
RBs=⇒ prevent flip flop

Step 3: Final matching result

47 / 92
Numerical Results

22
Centralized
SOEMA
20 IA
Sum rate of SCUs (bits/(s*Hz))

18
B=10, M=5
16

14

12
B=7, M=5
10

6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Number of iterations

Fig.: Convergence of the proposed algorithms for different number of RBs and SBSs.

48 / 92
Numerical Results (cont’)

18
SOEMA
IA
16 SOEMA−OMA
Sum rate of SCUs (bits/(s*Hz))

IA −OMA

14

12

10

6
10 12 14 16 18
Number of SBS (B)

Fig.: Sum-rate of the SCUs for different number of small cells, with M = 10.

49 / 92
Numerical Results (cont’)

−92
Average cross−tier interference at each MCU (dBm)
β=2×1013
−93
β=4×1013
−94 β=6×1013
β=8×1013
−95

−96

−97

−98

−99

−100

−101

−102
6 7 8 9 10
Number of RBs (M)

Fig.: Average received cross-tier interference at each MCU, with B = 12.

50 / 92
Summary

NOMA-enabled HetNets

Novel resource allocation algorithm based on matching

theory
Complexity: O(B 2 )
Performance: near-optimal performance

NOMA-enabled HetNets outperform OMA-based one

51 / 92
NOMA in HetNets II — Large-Scale Analysis

User n signal
Massive MIMO detection

User m signal SIC of User

detection m signal

User 1
Pico BS
User m User n
Marco BS
NOMA
……
User 2
User N

Fig.: System model.

High spectrum efficiency

Low complexity: The complex precoding/cluster design for
MIMO-NOMA systems can be avoided.
Fairness/throughput tradeoff: allocating more power to weak users.

[1] Y. Liu, Z. Qin, M. Elkashlan, A. Nallanathan, JA McCann (2017),“Non-orthogonal Multiple Access in

Large-Scale Heterogeneous Networks”, IEEE Journal on Selected Areas in Communications (JSAC).

52 / 92
Network Model

K-tier HetNets model: the first tier represents the macro

cells and the other tiers represent the small cells such as pico
cells and femto cells.
Stochastic Geometry: the positions of macro BSs and all
the k-th tier BSs are modeled as homogeneous poisson point
processes (HPPPs).
Hybrid access: massive MIMO transmissions in macro cells
and NOMA transmissions in small cells.
Flexible User association: based on the maximum average
received power.

53 / 92
Information Signal Model

The signal-to-interference-plus-noise ratio (SINR) that a

typical user experiences at a macro BS is

P1 /Nho,1 L (do,1 )
. (25)
IM,1 + IS,1 + σ 2

The SINR that user n experiences at the k-th tier small cell is

an,k Pk go,k L (do,kn )

γk n = . (26)
IM,k + IS,k + σ 2

The SINR experienced by user m in the k-th tier small cell is

am,k Pk go,k L (Rk )

γkm∗ = . (27)
Ik,n + IM,k + IS,k + σ 2

54 / 92
User Association Probability

The user association probability of a typical user connecting

to the NOMA-enhanced small cell BSs in the k-th tier and to
the macro BSs can be calculated as:
λk
Ãk = K  δ  δ , (28)
P P̃1k GM
λi P̃ik B̃ik + λ1 Nan,k Bk
i=2
and
λ1
Ã1 = δ , (29)
K

P an,i P̃i1 Bi N
λi GM + λ1
i=2

Remark 4.1
By increasing the number of antennas at the macro cell BSs, the
user association probability of the macro cells increases and the
user association probability of the small cells decreases.
55 / 92
Coverage Probability

A typical user can successfully transmit at a target data rate of Rt .

1 Near User Case: successful decoding when the following
conditions hold.
The typical user can decode the message of the connected user
served by the same BS.
After the SIC process, the typical user can decode its own
message.

Pcov ,k (τc , τt , x0 )|x0 ≤rk = Pr {γkn→m∗ > τc , γkn > τt } , (30)

2 Far User Case: successful decoding when the following

condition holds

εf x α i I k + σ 2
(
)
Pcov ,k (τt , x0 )|x0 >rk = Pr go,km > t 0 . (31)
Pk η

56 / 92
Spectrum Efficiency

The spectral efficiency of the proposed hybrid Hetnet is

XK
τSE,L = A1 Nτ1,L + Ak τk , (32)
k=2

where Nτ1 and τk are the lower bound spectrum efficiency of

macro cells and the exact spectral efficiency of the k-th tier
small cells.

57 / 92
Energy Efficiency

The energy efficiency is defined as

Total data rate
ΘEE = . (33)
Total energy consumption
The energy efficiency of the proposed hybrid Hetnets is as
follows:
XK
ΘHetnets
EE = A1 Θ1EE + Ak ΘkEE , (34)
k=2

Here, Ak and A1 are the user association probability of the

k-th tier small cells and macro cell, respectively.
τk Nτ1,L
ΘkEE = Pk,total and Θ1EE = P1,total are the energy efficiency of
k-th tier small cells and macro cell, respectively.

58 / 92
Numerical Results—User Association Probability

0.7
Marco cells
As the number of antennas
0.6
Pico cells
Femto cells
at each macro BS
User association probability

Simulation

B2 =10
increases, more users are
0.5
likely to associate to macro
0.4 cells — larger array gain.
0.3 Increasing the bias factor
can encourage more users
0.2 B2 =20
to connect to the small
50 100 150 200 250 300 350 400 450 500
M cells — an efficient way to
Fig.: User association probability versus extend the coverage of
antenna number with different bias small cells or control the
factor.
load balance among each
tier of HetNets.
59 / 92
Numerical Results — Coverage Probability

Observe the cross-over of

am =0.9, a n =0.1
am =0.6, a n =0.4
these two surfaces —
1 optimal power sharing for
the target-rate considered.
Coverage probability

0.8

0.6
For inappropriate power
0.4

0.2 6
and target-rate selection,
0
4 the coverage probability is
5
4
3
2
1
2
Rc (BPCU) always zero.
0 0
Rt (BPCU)

Fig.: Successful probability of typical

user versus targeted rates of Rt and Rc .

60 / 92
Numerical Results — Spectrum Efficiency

3.5 NOMA-based small cells

3 outperform the
Spectrum efficiency (bit/s/Hz)

NOMA
2.5 conventional OMA based
2 small cells.
1.5 The spectral efficiency of
Analytical NOMA, P 2 = 20 dBm
1 Analytical NOMA, P 2 =30 dBm small cells is reduced as
Simulation
0.5 OMA,P2 =30 dBm
OMA
the bias factor is increased
OMA,P2 =20 dBm
0
5 10 15 20 25 30
— larger bias factor results
B2 in associating more macro
Fig.: Spectrum efficiency comparison of users having a low SINR to
NOMA and OMA based small cells. small cells.

61 / 92
Numerical Results — Energy Efficiency

3 The energy efficiency of

the macro cells is reduced
Energy efficiency (bits/Hz/Joule)

2.5

2 Macro cells M=200

as the number of antennas
NOMA small cells M=200
HetNets M=200
is increased owing to the
1.5 NOMA small cells
Macro cells M=50
HetNets

OMA small cells

NOMA small cells M=50 power consumption of the
HetNets M=50
1
Macro cells OMA small cells M=200
OMA small cells M=50
baseband signal processing
0.5 of massive MIMO.
0
5 10 15 20 25 30
NOMA-assisted small cells
B2 may achieve higher energy
Fig.: Energy efficiency of the proposed efficiency than the massive
framework. MIMO aided macro cells
as a benefit of densely
deploying the BSs in
NOMA-aided small cells.
62 / 92
NOMA-based D2D Communications

D2D communications underlaying cellular networks

Non-Orthogonal Multiple Access (NOMA) protocol:
facilitates the access of multiple users in the power domain
New framework: NOMA-enhanced D2D, to further improve
the spectral efficiency
Challenge: Complex co-channel interference environment
⇓
Intelligent resource allocation design is needed

63 / 92
System Description

DT1
... DR1
DRLn
...
DRk
...
DRLn D2D Group D1
DTn BS Reuse
DR1 Subchannel
Reuse
Subchannel
D2D Group Dn

Cellular User

Fig.: System model.

Single-cell uplink scenario

Set of traditional cellular users: C = {C1 , ..., CM }
Set of D2D groups: D = {D1 , . . . , Dn , . . . , DN }
[1] J. Zhao, Y. Liu, K. K. Chai, Y. Chen, and M. Elkashlan (2017),“Joint Subchannel and Power Allocation for

NOMA Enhanced D2D Communications”, IEEE Transactions on Communications (TCOM), 2017.

64 / 92
Channel Model

The signal received by the BS corresponding to subchannel

SCm :
p X p
ym = Pc hm,b xm + ηn,m Pd gn,b tn + ζm , (35)
| {z } | n {z } |{z}
desired signal interference from D2D links noise

The signal at the k-th receiver in the n-th D2D group:

q XLn q
zn,k = fn,k an,k Pd sn,k + fn,k an,k 0 Pd sn,k 0 + ζn,k
k 0 =k+1 |{z}
| {z } | {z }
desired signal interference from NOMA users noise
X p p
+ ηn∗,n Pd gn∗,n,k tn∗ + Pc hm,n,k xm ,
n∗6=n | {z }
| {z }
interference from other D2D groups interference from CU

(36)

65 / 92
Problem Formulation

Maximize the sum-rate:

maxηn,m Rsum , (37a)

k thr
s.t. γn,k ≥ γn,k , ∀n, k, (37b)
thr
γm ≥ γm , ∀m, (37c)
ηn,m ∈ {0, 1} , ∀n, m, (37d)
X
ηn,m ≤ 1, ∀n. (37e)
m
Solution:
NP-hard =⇒ High complexity
Solution: Many-to-one matching theory

66 / 92
Matching Model

: “Prefer” PL = P (D1 ) , . . . , P (DN ) , P† (RB1 ) , . . . , P† (RBM )


0
RBm Dn RBm0 ⇔ Rnm > Rnm
0
S RBm S 0 ⇔ RmS + Dn ∈S Rnm > RmS + Dn ∈S 0 Rnm
P P

D2D Group D1

...

DR1,1 DT1 DR1,k

...
RB1
D2D Group D2

... RB2

DR2,1 DT2 DR2,k RB3

...

...
D2D Group D3

...

DR3,1 DT3 DR3,k

...
...

67 / 92
Matching Algorithm

Step 1: Initialization: PL
propose to
Step 2: Matching Phase: D2D groups −−−−−→ RBs;
acceps/reject
RBs −−−−−−−→ D2D groups
Step 3: Final matching result
D2D Group D1

...

DR1,1 DT1 DR1,k

...
RB1
D2D Group D2

... RB2

DR2,1 DT2 DR2,k RB3

...

...
D2D Group D3

...

DR3,1 DT3 DR3,k

...
...

68 / 92
Numerical Results

6
Optimal
MTBSA
Number of accessed D2D groups 5 One−to−one matching

0
1 3 5 7 9 11
Number of D2D groups (N)

Fig.: Number of accessed D2D groups versus the number of D2D groups in the
network, with K=3.

69 / 92
Numerical Results (cont’)

35
Optimal
MTBSA
One−to−one matching
30 Optimal (OMA)
Total sum rate (bits/(s*Hz))

Many−to−one matching (OMA)

One−to−one matching (OMA)

25

20

15

10
1 3 5 7 9 11
Number of D2D groups (N)

Fig.: Total sum-rate versus the number of D2D groups in the network, with K=3.

70 / 92
Numerical Results (cont’)

16
Optimal
MTBSA
14 One−to−one matching
Optimal (OMA)
Number of accessed receivers

Many−to−one matching (OMA)

12
One−to−one matching (OMA)

10

0
1 3 5 7 9 11
Number of D2D groups (N)

Fig.: Number of accessed receivers versus the number of D2D groups in the network,
with K=3.

71 / 92
Numerical Results (cont’)

18
Optimal
MTBSA
One−to−one matching
17.5
Total sum rate (bits/(s*Hz))

17

16.5

16

15.5
2 3 4
Number of receivers in each D2D group (K)

Fig.: Total sum-rate versus the number of receivers in each D2D group, with N=3.

72 / 92
Conclusions

NOMA-enhanced D2D framework

Novel resource allocation algorithm based on matching
theory
Complexity: O(NM 2 )
Performance: near-optimal performance
NOMA-enhanced D2D framework outperforms OMA-based
D2D framework
sum-rate
number of users supported

73 / 92
Security in NOMA Networks

1 Question: Is NOMA still secure when there are eavesdroppers

in the networks?

Main Channel
Bob n

Alice Wiretap Channel

for Bob m & Bob n
Bob m

Eve

2 The use of insecure wireless communication. channels

3 Strong detection ability at the eavesdropper side.
[1] Y. Liu, Z. Qin, M. Elkashlan, Y. Gao, and L. Hanzo(2017), “Enhancing the Physical Layer Security of
Non-orthogonal Multiple Access in Large-scale Networks”, IEEE Transactions on Wireless Communications
(TWC).

74 / 92
Network Model

Network model for the

NOMA transmission
protocol under malicious
attempt of eavesdroppers in
rp large-scale networks, where
RD rp , RD , and ∞ are the
radius of the protected zone,
∞ NOMA user zone, and an
infinite two dimensional
plane for eavesdroppers,
respectively.
Base station User Eavesdropper

75 / 92
Network Model—SINR for NOMA users

Based on the aforementioned assumptions, the instantaneous

signal-to-interference-plus-noise ratio (SINR) for the m-th user and
signal-to-plus-noise ratio (SNR) for the n-th user can be given by

am |hm |2
γBm = , (38)
an |hm |2 + ρ1b

and

γBn = ρb an |hn |2 , (39)

PA
respectively. We denote ρb = σb2
as the transmit SNR, where PA is
the transmit power at Alice and σb2 is the variance of additive
white Gaussian noise (AWGN) at Bobs.

76 / 92
Network Model—SNR for the Eavesdroppers

The instantaneous SNR for detecting the information of the m-th

user and the n-th user at the most detrimental Eve can be
expressed as follows:
n o
γEκ = ρe aκ max |ge |2 L (de ) . (40)
e∈Φe ,de ≥rp

It is assumed that κ ∈ {m, n}, ρe = PσA2 is the transmit SNR with

e
σe2 is the variance of AWGN at Eves.
In this paper, we assume that Eves can be detected if they are
close enough to Alice. Therefore, a protect zone with radius
rp is introduced to keep Eves away from Alice.

77 / 92
Secrecy Outage Probability

The secrecy rate of the m-th user and the n-th user can be
expressed as

Im = [log2 (1 + γBm ) − log2 (1 + γEm )]+ , (41)

and

In = [log2 (1 + γBn ) − log2 (1 + γEn )]+ , (42)

respectively, where [x ]+ = max{x , 0}.

78 / 92
Exact Secrecy Outage Probability

Given the expected secrecy rate Rm and Rn for the m-th and n-th
users, a secrecy outage is declared when the instantaneous secrecy
rate drops below Rm and Rn , respectively. Based on (41), the
secrecy outage probability for the m-th and n-th user is given by

Pm (Rm ) = Pr {Im < Rm }

Z ∞  
= fγEm (x ) FγBm 2Rm (1 + x ) − 1 dx . (43)
0

and

Pn (Rn ) = Pr {In < Rn }

Z ∞  
= fγEn (x ) FγBn 2Rn (1 + x ) − 1 dx , (44)
0

respectively.
79 / 92
Secrecy Diversity Analysis

The secrecy diversity order can be given by

∞ + P ∞ − P ∞P ∞)
log (Pm n m n
ds = − lim = m, (45)
ρb →∞ log ρb
The asymptotic secrecy outage probability for the user pair can be
expressed as
∞
Pmn ∞
=Pm + Pn∞ − Pm
∞ ∞ ∞
Pn ≈ Pm Gm (ρb )−Dm . (46)

Remarks: It indicates that the secrecy diversity order and the

asymptotic secrecy outage probability for the user pair are
determined by the m-th user.

80 / 92
Numerical Results

100 The red curves and the

α=4
black curves have the same
Secrecy outage probability

α=3
10-1
slopes. While the blue
asymptotic, m=1, n=3
exact, m=1 n=3 curves can achieve a larger
10-2 simulation, m=1, n=3
asymptotic, m=1, n=2 secrecy outage slope.
exact, m=1, n=2
10
-3 simulation, m=1, n=2 It is due to the fact that the
asymptotic, m=2, n=3
exact, m=2, n=3 secrecy diversity order of the
simulation, m=2, n=3
10
-4
0 10 20 30 40 50
user pair is determined by
ρ (dB)
b the poor one m.
This phenomenon also
consists with the obtained
insights in Remark 1.

81 / 92
Numerical Results

100 λ e= 10-3 The secrecy outage

λ e= 10-4
probability decreases as the
Secrecy outage probability

10-1
radius of the protected zone
increases, which
RD = 5 m, λ e= 10-3 demonstrates the benefits of
10-2
R = 10 m, λ = 10-3
D e
the protected zone.
R = 5 m, λ = 10-4
D e
R =10 m, λ = 10-4
Smaller density λe of Eves
D e
10
-3
1 2 4 6 8 10 12 14 16 18 20
can achieve better secrecy
r (m)
p performance, because
smaller λe leads to less
number of Eves, which lower
the multiuser diversity gain
when the most detrimental
Eve is selected.
82 / 92
Multi-antenna Aided Security Provisioning for NOMA

Bob n
Bob n
Alice
Alice

Bob m

Bob m
& Eve

Eve
Main Channel

Wiretap Channel
Wiretap Channel for Bob n
for Bob m & Bob n

(a) PLS of NOMA with External Eves (b) PLS of NOMA with Internal Eves

1 Artificial Noise for enhancing the security [1].

2 Multi-antenna to create channel differences [2].
[1] Y. Liu, Z. Qin, M. Elkashlan, Y. Gao, and L. Hanzo(2017), “Enhancing the Physical Layer Security of
Non-orthogonal Multiple Access in Large-scale Networks”, IEEE Transactions on Wireless Communications
(TWC).
[2] Z. Ding, Z. Zhao, M. Peng, and H. V. Poor (2017), “On the Spectral Efficiency and Security Enhancements of
NOMA Assisted Multicast-Unicast Streaming”, IEEE Transactions on Communications (TCOM).
83 / 92
Other Research Contributions on NOMA

1 Interplay between NOMA and cognitive radio networks.

2 MIMO-NOMA design.
3 NOMA in mmWave Networks.
4 Cross layer design for NOMA — a QoE perspective.
5 Full-duplex design for NOMA.
6 Relay-selection for NOMA.

84 / 92
Interplay between NOMA and cognitive radio networks

PT PR BS SR (User n)
PT (user m)+ST
(user n)

ST SR PR (User m)

Transmission link Interfernce link

(a) Conventional CR (b) CR Inspired NOMA

1 Cognitive radio inspired NOMA [1].

2 NOMA in cognitive radio networks [2].
[1] Z. Ding, P. Fan, and H. V. Poor (2016), “Impact of User Pairing on 5G Nonorthogonal Multiple-Access
Downlink Transmissions”, IEEE Trans. Veh. Technol. (TVT).
[2] Y. Liu, Z. Ding, M. Elkashlan, and J. Yuan, “Non-orthogonal Multiple Access in Large-Scale Underlay Cognitive
Radio Networks”, IEEE Trans. Veh. Technol. IEEE Trans. Veh. Technol. (TVT).

85 / 92
MIMO-NOMA Design - Beamformer Based Structure

1 Centralized Beamforming.
2 Coordinated Beamforming.

User m detection Subtract user m’s User n detection

with Rn→m signal with Rn →n

wn User n
SIC

wm
User m detection
BS
with Rm→m
User m

[1] Y. Liu, H. Xing, C. Pan, A. Nallanathan, M. Elkashlan, and L. Hanzo, “Multiple Antenna Assisted
Non-Orthogonal Multiple Access”, IEEE Wireless Communications.

86 / 92
MIMO-NOMA Design - Beamformer Based Structure

1 Centralized Beamforming.
2 Coordinated Beamforming.

Unserved
Near User BS User

Unserved User BS
Centric Near User
Far Cell Edge User

BS
Near User
Unserved User

Coordinated Data link for

beamforming link Interference link
near user

[1] Y. Liu, H. Xing, C. Pan, A. Nallanathan, M. Elkashlan, and L. Hanzo, “Multiple Antenna Assisted
Non-Orthogonal Multiple Access”, IEEE Wireless Communications.

87 / 92
MIMO-NOMA Design - Cluster Based Structure

1 Inter-Cluster Interference Free Design.

2 Inter-Cluster Interference Contaminated Design.

……
User L1
User 2 ……
User 1
……
User 1 User 2 User Lm

BS ……
User 1
User 2 ……
User LM

[1] Y. Liu, H. Xing, C. Pan, A. Nallanathan, M. Elkashlan, and L. Hanzo, “Multiple Antenna Assisted
Non-Orthogonal Multiple Access”, IEEE Wireless Communications.

88 / 92
NOMA in MmWave Networks

1 User Scheduling — Matching Theory.

2 Power Allocation — Branch-and-bound.
M superposed data streams

Baseband processing

...
1 5)

...
...
...

M beams
K users

Partial channel information feedback

[2] J. Cui, Y. Liu, Z. Ding, P. Fan, and A. Nallanathan, “Optimal User Scheduling and Power Allocation for
Millimeter Wave NOMA Systems”, IEEE Transactions on Wireless Communications (TWC) accept to appear.

89 / 92
Cross layer design for NOMA — a QoE perspective

1 QoE-Aware NOMA Framework [1].

2 Multi-cell Multi-carrier QoE aware resource allocation [2].

Content
Context

Codec, bitrate Application

display

Clustering, Packet queue

1 2 1 Buffering
scheduling 2 1 2

Superposition coding/non-
Power\code

UserN Subtract
orthogonal multi-carrier design ... 1 2 1 1
2
User2 Decoding
User1 1
Frequency

Transmitter Receiver

[1] W. Wang, Y. Liu, L. Zhiqing, T. Jiang, Q. Zhang and A. Nallanathan, “Toward Cross-Layer Design for
Non-Orthogonal Multiple Access: A Quality-of-Experience Perspective”, IEEE Wireless Communications (Under
revision).

[2] J. Cui, Y. Liu, Z. Ding, P. Fan, and A. Nallanathan, “QoE-based Resource Allocation for Multi-cell NOMA

Networks”, IEEE Transactions on Wireless Communications (TWC) (Under Review).

90 / 92
Research Opportunities and challenges for NOMA

1 Error Propagation in SIC.

2 Imperfect SIC and limited channel feedback.
3 Synchronization/asynchronization design for NOMA.
4 Different variants of NOMA.
5 Novel coding and modulation for NOMA.
6 Hybrid multiple access
7 Efficient resource management for NOMA
8 Security issues of NOMA
9 Different variants of NOMA

91 / 92
Thank you!

Thank you!

92 / 92