Vous êtes sur la page 1sur 6

Cooperative NOMA-Based D2D Communications:

A Survey in the 5G/IoT Context


Rajaa Elouafadi and Mustapha Benjillali
INPT
Rabat, Morocco
Emails: elouafadi@inpt.ac.ma, benjillali@ieee.org

Abstract—Non-orthogonal multiple access (NOMA) and On the other hand, multiple access techniques are also
device-to-device (D2D) communications have been recognized being discussed and, by many considerations, redefined to
as promising techniques for the fifth generation (5G) networks meet the high spectral efficiency targets of 5G. While the
due to their potential role in increasing the spectral efficiency.
Cooperative NOMA presents a more attractive option since the previous generations of mobile networks have been relying
two techniques are combined to provide better benefits and on user orthogonality in the time/frequency/code domains,
address some of 5G deployment purposes, and Internet of things and suffering from the inherently reduced resource allocation
(IoT) constraints and use cases. In this paper, considering recent efficiency; the introduction of non-orthogonal multiple access
and popular cooperative NOMA schemes, we first provide a (NOMA) is a key enabler to alleviate the increasing resource
compact taxonomical classification of the various approaches
in the literature, and we present a comparative performance scarcity issues, and enhancing different performance metrics
analysis based on outage probability and average throughput. (e.g., multiplexing gains, spectral efficiency, scheduling com-
Then, we illustrate the applicability of cooperative NOMA by a plexity, and latency). We here focus on the power domain
few use cases in industrial IoT environments. Finally, we discuss NOMA, where the signals from multiple users are multiplexed
open research challenges that still need to be investigated in this (superposed) on the transmitter side, and multi-user signal
area.
separation is performed on the receiver side based on a
I. I NTRODUCTION successive interference cancellation (SIC) process [5].
In this paper, we discuss the different approaches to the
Based on the targets set by IMT2020 in terms of spectrum implementation of NOMA-based D2D schemes in the context
efficiency, connection density, mobility, energy efficiency, and of emerging 5G and IoT eco-systems. First, as a reminder,
user experience, the 5th generation (5G) of mobile systems we revisit the performance of basic NOMA (in contrast with
is under standardization, and the first products are highly the conventional orthogonal multiple access (OMA)), then
expected to be on the market by the end of next year. we extend the discussion to cooperative NOMA schemes.
Compared to the previous generations, 5G networks should For instance, we propose a taxonomical classification of the
be supporting unprecedented numbers of connected devices different system models in the literature, and we discuss
as well as considerably larger volumes of data traffic. their effectiveness through a comparative study showing the
Many surveys in the literature discussed the key technolo- advantages and limitations of each scheme. In addition, in
gies to enable the emergence of 5G systems and networks; order to illustrate the discussion with practical examples, we
both from evolutionary points of view and disruptive revolu- present D2D use cases in industrial IoT applications. Finally,
tionary visions. For example, in [1], the authors have back we discuss open research challenges, and propose a few
then advocated five technologies that could lead to disruptive directions of interest that may be investigated by the research
approaches to the design of 5G cellular networks, while community in the near future.
the authors of [2] present an extensive discussion of the II. C ONCEPT AND P ERFORMANCE C OMPARISON
architectural changes to the radio access network, and the role
other key technologies can play in the new 5G eco-system. We consider a common, basic, system model like the one
in [5] consisting of one base station source BS and two users
Device-to-device (D2D) communications are one of the
UE-1 and UE-2. BS sends two superposed signals x1 and x2
promising options that have been highlighted by the research
to UE-1 and UE-2, respectively, with transmit powers P1 and
community in the past years, and have been adopted by
P2 (without any loss of generality, we consider that P2 >
several standardization groups. The aim is to boost the per-
P1 ). The SIC process is implemented at the UE receiver in
formance of conventional cellular networks, e.g. in terms
the order of decreasing channel gain normalized by noise and
of power consumption, spectrum efficiency, and throughput,
inter-cell interference power, thus in the case of 2 users: UE-
by exploiting the possibility of direct interactions between
1 first decodes x2 and subtracts its component from received
devices in proximity [3]. In the particular context of Internet of
signal, then, it decodes x1 . UE-2 treats x1 as noise and decodes
things (IoT) applications and use cases, D2D relaying schemes
x2 . In NOMA, x1 and x2 are superposed as follows:
present several advantages, and can help improving reliability, √ √
fault tolerance, and network access scalability [4]. 𝑥 = 𝑃1 𝑥 1 + 𝑃2 𝑥 2

978-1-5386-3738-8/18/$31.00 ©2018 IEEE 132


Thus, the received signal at UE-i is represented as : III. C OOPERATIVE NOMA
Among the current research works on the cooperative re-
𝑦 𝑖 = ℎ𝑖 𝑥 + 𝑤 𝑖 laying systems using NOMA we selected the articles [6-14].
Fig. 3 summarizes the use cases of the considered system
Where hi is the complex channel coefficient between UE-i models.
and the BS. Term wi denotes additive white Gaussian noise Kim et al. in [6] proposed a cooperative relaying sys-
(AWGN) including inter-cell interference. The power spectral tem (CRS) using NOMA which can achieve more spectral
2 2
density of wi is Ni . We assume that ∣ℎ𝑁11∣ > ∣ℎ𝑁22∣ . Assuming efficiency than the conventional CRS; the destination can
successful decoding and no error propagation, the throughput receive two symbols during two time slots unlike the case of
of UE-i, Ri , is represented as : conventional CRS where it can only receive one symbol. The
( ) ( ) system model consists of a source, a half-duplex (HD) relay,
𝑃1 ∣ℎ1 ∣2 𝑃2 ∣ℎ2 ∣2
𝑅1 = log 1 + , 𝑅2 = log 1 + and a destination, all the links are assumed to be available.
𝑁0,1 𝑃1 ∣ℎ1 ∣2 + 𝑁0,2 In [7] a cooperative NOMA scheme was introduced where
one base station sends a broadcast message to K users in the
For OMA, the bandwidth of 𝛼 (0 < 𝛼 < 1) Hz is assigned to direct transmission phase and each of the K users resends it
UE-1 and the remaining bandwidth, (1 − 𝛼) Hz, is assigned to the users with worst channel conditions in the cooperative
to UE-2. The throughput of UE-i, Ri , is represented as : phase during K time slots. [8] proposed a system model where
( ) a D2D device acts as a HD relay to serve the user with poor
𝑃1 ∣ℎ1 ∣2
𝑅1 = 𝛼 log 1 + channel conditions. In phase 1, the source performs NOMA
𝛼𝑁0,1
to serve UE-l and UE-2. In phase 2, Dl performs NOMA to
serve D2 and to boost UE-2. Lastly, UE-2 performs maximal
( )
𝑃2 ∣ℎ2 ∣2 ratio combining (MRC) using received signal at phase 1 and
𝑅2 = (1 − 𝛼) log 1 + phase 2. This system improves the system reliability and the
(1 − 𝛼)𝑁0,2
spectral efficiency.
We simulate the ergodic capacity of the system for [9] and [10] have considered the case of a full duplex (FD)
both NOMA and OMA schemes. For NOMA case, in relaying. In [9], Zhong et al. considered a two user NOMA
the ergodic capacity simulation we take different values of system where UE-1 directly communicates with the BS, while
𝑎1 = 𝑃1 /(𝑃1 + 𝑃2 ). For the case of OMA, we take 𝛼 = 1/2. UE-2 requires the assistance of a FD relay employing decode-
We present the results in Fig. 1. and-forward (DF). The system performance was compared
with the case of HD relay and showed that it achieves
lower outage probability for both users, and attains higher
10 1
ergodic sum capacity in the low to moderate SNR range.
[10] considered the case of D2D relaying. One source sends
a superimposed signal to two users using NOMA, and the
𝑎1 = 0.3 user with strongest channel conditions acts as a FD relay to
curves converge for high SINR help the weakest user. The authors have also proposed an
𝐸𝑟𝑔𝑜𝑑𝑖𝑐𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦 [bps/Hz]

adaptive multiple access scheme (AMA) to improve outage


performance, which dynamically switched to a proper mul-
𝑎1 = 0.1 tiple access scheme (OMA, NOMA or Cooperative NOMA)
based on the level of residual self-interference and channel
conditions. In [11], the authors propose a system with a
𝑎1 = 0.05
10 0
buffer aided relay and an adaptive transmission scheme in
which the system adaptively chooses its working mode in
Theoretical NOMA each time slot. In [12] two source-destination pairs share
Simulation NOMA a common half-duplex relay employing decode and forward
Theoretical OMA
Simulation OMA strategy. NOMA is applied at the relay level, both in uplink
-10 -5 0 5 10 15 20 25 30
and downlink. Different from [6-12], [13] and [14] have
𝑆𝑁 𝑅 [dB] considered cooperative NOMA system with multiple relays.
In [13] a two-stage relay-selection scheme was proposed. In
Fig. 1. Ergodic capacity of NOMA and OMA systems
[14], a hybrid decode-forward and amplify-forward relaying
scheme was proposed for a multiple relay cooperative NOMA
Simulation shows that NOMA presents better performance system.
than OMA as regards of system ergodic capacity. This rate
gain depends on the power allocation coefficient chosen. IV. TAXONOMY
Works such as [6] derived a sub-optimal power allocation In this section we categorize the cooperative NOMA models
scheme which maximizes the gain. based on the use of NOMA (uplink or downlink) and based

133
on the relaying type used in the cooperation. Compared to CRS scheme, the proposed system in [6] can
achieve more spectral efficiency when SNR is high and source
A. First categorization to relay link is better than those of source to destination and
A first categorization is based on the use of NOMA, either relay to destination links. Whereas it has worse performance
it is used in uplink at the relay level or in the downlink at the in low SNR regime. To solve this issue a suboptimal power
source level. Fig. IV-A gives a schematic representation. allocation scheme is proposed. Simulations show that, then,we
have more rate gain over the conventional CRS. However,
this introduction of suboptimal power allocation scheme adds
NOMA SIC at relay &
Cooperative

Multi−users complexity to the system, especially in case of multiple users


NOMA

On Downlink user levels


and in a mobility context where the calculations become
NOMA SIC at relay
heavy and the base station needs to dynamically change the
Multi−source power allocation coefficients (suboptimal power coefficient
On Uplink level
is dependent on the average channel power of the links).
Fig. 2. NOMA use in cooperative NOMA systems
[11] considers a transmission between a base station and
two users via a buffer relay capable of adaptively choosing
the optimal transmission scheme for fixed users rates. The
proposal improves the system throughput but this is at the
Cooperative NOMA on downlink
expense of system complexity. Indeed, it is generating addi-
The source sends a superimposed signal to the users on the tional signalization between the nodes to provide feedback on
power domain. The relay performs SIC to extract its own data the working mode chosen.
and forward the other messages to the related users in one In [12], the authors propose a model where two sources
of the cooperation modes. This is mainly used in the case share the same relay to communicate with their respective
of serving far users or with bad channel conditions from one destinations. The authors assumed that the sources send over
single base station or source. Papers [6] to [10] considers the same frequency band, cooperate to adapt their power
the cooperative models with one source. transmission coefficient and perfectly synchronize their mes-
sages. This is on the purpose to get a superimposed signal
Cooperative NOMA on uplink
at the relay level which decodes the messages using SIC and
NOMA is used on the uplink channel, one or more relays resend them using NOMA to the destinations. Capacity gain
receive a data from multiple sources simultaneously, perform of this model over OMA depends on the level of residual
SIC to extract the data and help the destinations to get their interference. It outperforms OMA under perfect SIC, whereas,
message. The relay can perform on its turn NOMA to serve under imperfect SIC scenario, the performance is worse in high
the destinations [12] or use another protocol based on the and moderate SNR regimes.
channel conditions and the quality of received data to serve Diffrently from the previous mentionned works which all
the users [14]. Use cases of this model is boosting the spectral considered the case of decode and forward relay (DF) (to
efficiency in serving users linked to two base stations, and/or adapt the data to be sent depending on the users locations
are at the edge of the BS’s cells. and channel conditions), [15] has investigated the case of an
amplify and forward (AF) relay to serve a far user.
B. Second categorization
FD : Zhong et al. in [9] justify the choice of using
A second categorization of the cooperative NOMA models a FD relay cooperative NOMA by the fact that with HD,
is proposed based on the relaying type used in the cooperation additional time resource are used for HD cooperation mech-
phase. We distinguish three types: Cooperative NOMA with anism which effectively offsets the spectral efficiency gain
dedicated relay, a D2D communications capable user which promised by NOMA systems. Simulations are showing that,
acts as relay for other users and a multi-relay scheme, where in perfect interference cancellation scenario, the proposed FD
the base station uses multiple relays to serve the destination cooperative NOMA system attains better performance than the
through one or many of them. HD cooperative NOMA system. However, under imperfect
We provide in Fig. 4 a schematic presentation of the interference cancellation, FD cooperative NOMA system is
proposed taxonomy. better than HD one only in low to moderate SNR regime .
Dedicated relay D2D relay
We distinguish between half-duplex (HD) and full duplex In this category as well we distinguish between the half-
(FD) relays. duplex users which are D2D communications capable and the
HD : [6], [11], [12] and [15] have considered cooperative full duplex ones.
NOMA system models with a dedicated half-duplex relay. HD : [7] proposes a NOMA multi-user cooperative model
Sum rate was used as metric to validate the system per- where a base station sends, using NOMA, a superimposed
formance over conventional cooperation relay system (CRS) message to K users in the direct transmission phase, then, the
schemes. K users cooperate during K-1 time slots in the cooperation

134
Fig. 3. Use cases for NOMA cooperative relaying systems

Cooperative to a proper multiple access scheme based on the level of


NOMA residual self-interference and the quality of channels. This
AMA scheme outperforms the cooperative NOMA, conven-
tional NOMA and OMA, but, like in [11], the gain will be
Dedicated Relay D2D user Multi−relay at the expense of the system complexity.

Mutli-relay
HD HD Relay selection
[13] is proposing two-stage scheme to select the optimal
relay which can achieve not only the optimal diversity gain,
FD FD Hybrid DF−AF but also the minimal outage probability. In this scheme, first,
the user with the high quality of service target is served with
its target data rate then the second one is served with a rate as
Fig. 4. Proposed classification of NOMA cooperative relaying systems
large as possible. [14] is considering a system model where
two-source communicate with one user through N relays. The
authors propose a hybrid decode and forward (DF)- amplify
phase. A diversity order of K is achievable by all users and forward (AF) scheme where a group from the total relays,
unlike the conventional NOMA where the nth user achieves, which can decode the first message, use a DF protocol to
at maximum, a diversity order of n. This configuration is not send data from source 1 and the others use an AF protocol to
very realistic in the case of many users due to the system send the superimposed message. This model is compared with
complexity for coordinating user cooperation. However, user DF-NOMA Best Relay, the DF-NOMA Multi-Relay, the DF-
pairing is a promising solution to reduce system complexity. TDMA Best Relay and the DF-TDMA Multi-Relay schemes.
The authors in [8] considered a different pattern, where a Simulations shows that it can achieve larger sum capacity for
source communicates with two users and a near D2D device high transmit power. The outage performance is, on the other
overhears the far users message, decodes it and resends it hand, much worse compared to other schemes.
to the destination while it is in communication, in parallel, TABLE I provides a comparative summary of the considered
with another D2D device. This proposal comes with spectral system models following the proposed taxonomy.
efficiency benefit; only half of time slot or resources are used
because of D2D NOMA relay compared to traditional D2D V. I OT INDUSTRIAL USE CASES
relay. The concept of smart industry is not new; it stands mainly
FD :Differently from [7], the authors in [8] considered a for accelerating the digitalization of industrial processes and
D2D- NOMA cooperative system at a smaller scale. A source the supply chain from the material supply to the end con-
communicates with a predefined NOMA user pair which has sumer. IoT technologies and especially D2D communica-
the capability of FD D2D communications. Compared to [9], tions technique is a great enabler to this vision. Indeed,
an AMA scheme is proposed which dynamically switches D2D communications enables proximity applications where

135
TABLE I
S UMMARY OF THE CONSIDERED MODELS PROPOSING COOPERATIVE NOMA

#Symboles/
Mode Ref. #Users Metric Advantage Complexity
Time slot

[6] 1 2 System-models outperform Scalability: Suboptimal power


Dedicated HD [11] 2 1 Sum Rate OMA in low SNR regimes. calculation can be heavy in the
Whereas, performance is case of many users and in mo-
Relay [12] 3 1 worse in high SNR regime. bility context [6]. Higher sys-
Suboptimal power coefficient tem complexity to adaptively
calculated to maximize the switch between different trans-
gain. mission modes for instance
signalization is heavy [11].

Sum Rate FD scheme outperforms HD FD cooperative NOMA sys-


FD [9] 2 1 & scheme in the low and mod- tem saturates in the high SNR
erate regime. regime where the relay be-
Outage comes interference-limited.

Sum Rate Reception reliability: Diversity When 𝐾 is large, the relia-


[8] 2 1 order of 𝐾 is achievable by bility will be at the expense
D2D HD &
[7] 𝐾 1/𝐾 all users [7]. Only half of the complexity. User grouping is
User Outage time slot is used because of one of the ideas to simplify the
D2D NOMA relay compared system.
to conventional D2D relay.

FD [10] 2 1 Outage Compared to [9], an AMA Like in [11]. The gain will be
scheme is proposed which dy- at the expense of complexity
namically switches to a proper with AMA, which will gener-
multiple access scheme based ate heavy signalization.
on the level of residual self-
interference and the qual-
ity of channels. This AMA
scheme outperforms conven-
tional ones.

Relay The scheme can achieve diver-


Multi- [13] 2 1/2 Outage sity gain and a minimal outage
Selection
Relay probability.

Sum Rate Hybrid DF-AF schemes can Outage performance is worse.


Hybrid achieve larger sum capacity for
[14] 1 2/2 &
DF-AF the transmission compared to
Outage other schemes.

devices can trigger different services, such as local information elucidate the missing parts and go for standardization, but
exchange, network coverage extension, local advertisement, still, many areas need to be explored and further studied. The
etc. [3]. This comes with many benefits for the industrial field following are some of the areas that need further investigations
such as increasing productivity, improving product quality, [3], :
lowering manufacturing costs and effective advertising. Other ∙ Multi user power allocation: as seen in the previous
benefits of the D2D communications are listed in [3], we sections, power allocation is key in achieving the targeted
can cite for instance, higher throughput, lower delay and quality of service. Finding the optimum value is not
energy consumption which are typical characteristics of short- always feasible and easy to compute in the case of many
range communications when compared to long-range commu- users. A trade-off between system complexity and achiev-
nications. Coverage extension is another key benefit of D2D ing targeted data rate need to be further investigated;
communications where devices can act as relays to serve non- ∙ Signaling overhead: extra signalization is needed for D2D
covered areas by the cellular network. users to find and explore others in proximity and to
Fig. 5 illustrates some of the applications of D2D and IoT get channel measurement. This might become intolerable
in the industry. overhead to the system if it needs instantaneous CSI
feedback. The trade-off between accuracy of CSI and its
VI. O PEN RESEARCH CHALLENGES
resulting overhead needs to be found;
NOMA and D2D communications are still new emergent ∙ Interference management: this comes as result of resource
technologies. Intensive researches have been conducted to allocation problematic, especially in the case of inband

136
IoT use cases in the industry and shared the open research
challenges on this field.
R EFERENCES
[1] F. Boccardi, R. W. Heath, A. Lozano, T. L. Marzetta, and P. Popovski,
“Five disruptive technology directions for 5G,” IEEE Communications
Magazine, vol. 52, no. 2, pp. 74–80, Feb. 2014.
[2] M. Agiwal, A. Roy, and N. Saxena, “Next generation 5G wireless
networks: A comprehensive survey,” IEEE Communications Surveys &
Tutorials, vol. 18, no. 3, pp. 1617–1655, Feb. 2016.
[3] L. Militano, G. Araniti, M. Condoluci, I. Farris, and A. Iera, “Device-
to-device communications for 5G internet of things,” IOT, EAI, vol. 1,
no. 1, pp. 1–15, Oct. 2015.
[4] M. R. Palattella, M. Dohler, A. Grieco, G. Rizzo, J. Torsner, T. Engel,
and L. Ladid, “Internet of things in the 5G era: Enablers, architecture,
and business models,” IEEE Journal on Selected Areas in Communica-
tions, vol. 34, no. 3, pp. 510–527, Mar. 2016.
[5] A. Benjebbour, Y. Saito, Y. Kishiyama, A. Li, A. Harada, and T. Naka-
mura, “Concept and practical considerations of non-orthogonal multiple
access (NOMA) for future radio access,” in International Symposium on
Fig. 5. Some of the industrial applications in IoT context Intelligent Signal Processing and Communications Systems (ISPACS).
Naha, Okinawa, Japan: IEEE, Nov. 2013, pp. 770–774.
[6] J.-B. Kim and I.-H. Lee, “Capacity analysis of cooperative relaying
systems using non-orthogonal multiple access,” IEEE Communications
D2D, where D2D pairs use the same resources as cellular Letters, vol. 19, no. 11, pp. 1949–1952, Nov. 2015.
users. Interference need to be monitored to not affect the [7] Z. Ding, M. Peng, and H. V. Poor, “Cooperative non-orthogonal multiple
cellular users; access in 5G systems,” IEEE Communications Letters, vol. 19, no. 8,
pp. 1462–1465, Aug. 2015.
∙ Energy consumption: although D2D communications is [8] Y. B. Song, H. S. Kang, and D. K. Kim, “5G cellular systems with D2D
supposed to improve the energy efficiency of the UE, this assisted NOMA relay,” in URSI Asia-Pacific Radio Science Conference
might be affected if the user equipment needs to wake (URSI AP-RASC). Seoul, South Korea: IEEE, Aug. 2016, pp. 1–3.
[9] C. Zhong and Z. Zhang, “Non-orthogonal multiple access with coopera-
up very often to listen to transmit discovery messages tive full-duplex relaying,” IEEE Communications Letters, vol. 20, no. 12,
frequently. A proper discovery protocol needs to be pp. 2478–2481, Dec. 2016.
further investigated; [10] Z. Zhang, Z. Ma, M. Xiao, Z. Ding, and P. Fan, “Full-duplex device-to-
device-aided cooperative nonorthogonal multiple access,” IEEE Trans-
∙ Privacy and security issues, combination with MIMO and actions on Vehicular Technology, vol. 66, no. 5, pp. 4467–4471, May
receiver design are other areas that need to be sealed. 2017.
[11] S. Luo and K. C. Teh, “Adaptive transmission for cooperative NOMA
VII. C ONCLUSION system with buffer-aided relaying,” IEEE Communications Letters,
vol. 21, no. 4, pp. 937–940, Apr. 2017.
In this work, we presented a comparison study of some [12] M. F. Kader, M. B. Shahab, and S. Y. Shin, “Exploiting non-orthogonal
of the use cases of future radio access networks in the IoT multiple access in cooperative relay sharing,” IEEE Communications
Letters, vol. 21, no. 5, pp. 1159–1162, May 2017.
context. First we gave a reminder on the NOMA concept and [13] Z. Ding, H. Dai, and H. V. Poor, “Relay selection for cooperative
check its performance over OMA scheme, then we presented NOMA,” IEEE Wireless Communications Letters, vol. 5, no. 4, pp. 416–
the considered system models on cooperative NOMA, pro- 419, Aug. 2016.
[14] Y. Liu, G. Pan, H. Zhang, and M. Song, “Hybrid decode-forward &
posed a taxonomy based on the use of NOMA whether it is amplify-forward relaying with non-orthogonal multiple access,” IEEE
on uplink or in downlink and another one based on the relay Access, vol. 4, pp. 4912–4921, 2016.
type used in the cooperation. We discussed the advantage and [15] X. Liang, Y. Wu, D. W. K. Ng, Y. Zuo, S. Jin, and H. Zhu, “Outage
performance for cooperative NOMA transmission with an AF relay,”
complexity of each system model. Lastly we presented some IEEE Communications Letters, Mar. 2017.

137

Vous aimerez peut-être aussi