Non-Orthogonal Multiple Access and Network

Slicing: Scalable Coexistence of eMBB and URLLC

Eduardo Noboro Tominaga∗ , Hirley Alves∗ , Richard Demo Souza† , João Luiz Rebelatto‡ , Matti Latva-aho∗
2021 IEEE 93rd Vehicular Technology Conference (VTC2021-Spring) | 978-1-7281-8964-2/20/$31.00 ©2021 IEEE | DOI: 10.1109/VTC2021-Spring51267.2021.9448942

∗ 6G Flagship, Centre for Wireless Communications (CWC), University of Oulu, Finland

† Federal University of Santa Catarina (UFSC), Florianópolis, Brazil
‡ Federal University of Technology - Paraná (UTFPR), Curitiba, Brazil

{firstname.lastname}@oulu.fi, [email protected], [email protected]

Abstract—The 5G systems feature three generic services: the development of solutions for beyond-5G and 6G wireless
enhanced Mobile BroadBand (eMBB), massive Machine-Type communications systems. One of the predicted use cases for
Communications (mMTC) and Ultra-Reliable and Low-Latency beyond-5G and 6G systems is the critical wireless factory
Communications (URLLC). The diverse requirements of these
services in terms of data-rates, number of connected devices, automation that requires communication with ultra-high re-
latency and reliability can lead to a sub-optimal use of the 5G liability and ultra-low latency. Moreover, it is foreseen that
network, thus network slicing is proposed as a solution that the number of connected devices will increase substantially
creates customized slices of the network specifically designed for 6G, which also poses very stringent requirements in terms
to meet the requirements of each service. Under the network of spectral efficiency [2]. In this context, 6G will scale the
slicing, the radio resources can be shared in orthogonal and
non-orthogonal schemes. Motivated by Industrial Internet of traditional URLLC to the massive connectivity dimension,
Things (IIoT) scenarios where a large number of sensors may thus leading to a new service class defined massive URLLC
require connectivity with stringent requirements of latency and (mURLLC), which is the merge of the traditional URLLC and
reliability, we propose the use of Non-Orthogonal Multiple Access mMTC services from 5G [3].
(NOMA) to improve the number of URLLC devices that are The diverse and sometimes conflicting requirements of the
connected in the uplink to the same base station (BS), for
both orthogonal and non-orthogonal network slicing with eMBB different 5G services and applications can lead to a sub-
devices. The multiple URLLC devices transmit simultaneously optimal use of the mobile network. One efficient solution
and across multiple frequency channels. We set the reliability is the slicing of the network in multiple virtual and isolated
requirements for the two services and evaluate the pairs of logical networks running on a common physical infrastructure
achievable sum rates. We show that, even with overlapping in an efficient and economic way, thus allowing slices to be
transmissions from multiple eMBB and URLLC devices, the
use of NOMA techniques allows us to guarantee the reliability individually customized with respect to, e.g., latency, energy
requirements for both services. efficiency, mobility, massive connectivity and throughput [4].
Index Terms—5G, eMBB, IIoT, Network Slicing, NOMA, The dynamic provisioning of network slices will be one of
URLLC. the features of 6G networks. Instead of having the traditional
categorization into eMBB, mMTC and URLLC, some 6G
I. I NTRODUCTION applications will require dynamic service types according to
The Fifth Generation (5G) of wireless communication the data traffic and usage patterns of the MTC network [5].
systems is currently under standardization and deployment The previous generations of wireless communications sys-
around the world, and introduces three generic services: en- tems were mostly based on the utilization of Orthogonal
hanced Mobile Broadband (eMBB), massive Machine-Type- Multiple Access (OMA) schemes to provide connectivity to
Communications (mMTC) and Ultra-Reliable Low-Latency multiple users. In such schemes, users are allocated with
Communications (URLLC). eMBB aims to provide increased radio resources that are orthogonal in time, frequency or
data rates, with peak rates on the order of gigabits per second code domain, and ideally no interference exists among them.
to moderate rates in the order of megabits per second with high However, one drawback of OMA schemes is that the maximum
availability. The mMTC service aims at providing connectivity number of users is limited by the total amount of available
for a large number of cost- and energy-constrained devices orthogonal resources [6]. To meet the diverse requirements
(e.g. sensors) that often require low data rates. Finally, the of very high data rates, ultra-reliability, low latency, massive
challenging objective of URLLC is to provide ultra-reliable connectivity and spectral efficiency, Non-Orthogonal Multiple
connectivity while operating in short block lengths, a re- Access (NOMA) emerges as a promising technology for
quirement to achieve low latency demanded by time-critical beyond-5G and 6G. NOMA allows multiple users to share time
applications [1]. and frequency resources in the same spatial layer via power
The current 5G New Radio (NR) network is not capable domain or code domain multiplexing [6]. It is also predicted
yet to satisfy the very stringent requirements of reliability that in 6G scenarios there will be a need to new access schemes
and latency required by URLLC applications. That is one of that can dynamically change between orthogonal and non-
the reasons why the research community has already started orthogonal multiple access schemes depending on the current

978-1-7281-8964-2/21/$31.00 ©2021 IEEE

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.
state of the network [3]. overlapping transmissions from multiple URLLC devices, the
A communication-theoretic framework for network slicing use of NOMA, SIC and frequency diversity techniques allow
in 5G was presented in [7], where the same radio resources us to guarantee the reliability requirements of eMBB and
are sliced among the heterogeneous 5G services under both URLLC services in both slicing schemes.
orthogonal and non-orthogonal strategies. However, they did This paper is organized as follows. In the next section, we
not considered the use of NOMA for multiple URLLC devices, present the system model and the performance analysis of the
such that the maximum number of URLLC devices connected eMBB and URLLC when they are considered in isolation. In
to the same Base-Station (BS) was limited by the number of Section III, we show how eMBB and URLLC devices can
minislots within the timeslot. share the same radio resources for both orthogonal and non-
The coexistence of eMBB and URLLC services has also orthogonal slicing. In Section IV, we present the numerical
been studied in the other works. Joint scheduling of eMBB results illustrating the performance trade-off between the ser-
and URLLC traffic has been studied in, for example, [8], vices. Finally, the conclusions are presented in Section V.
[9] and [10]. Abreu et. al. [11] studied the multiplexing of
eMBB and URLLC traffics in the uplink using an analytical II. S YSTEM M ODEL
framework. They considered the cases where different bands We consider an uplink scenario where multiple eMBB and
are allocated for each service, and also the case where both URLLC devices transmit independent packets to a common
services share the same band. The slicing of resources for BS, as illustrated in Fig. 1. As in [7], we also consider a time-
eMBB and URLLC has been also studied in [12], where frequency grid composed of F frequency channels indexed by
the authors proposed a risk-sensitive based formulation to f ∈ {1, . . . , F } and S minislots indexed by s ∈ {1, . . . , S}.
allocate resources to URLLC devices while minimizing the The set of S minislots composes a timeslot.
risk of eMBB (i.e. protecting the eMBB devices with low The orthogonal and non-orthogonal slicing of the radio
data rate) and ensuring URLLC reliability. In [13], the authors resources for eMBB and URLLC are also based on [7] and
adopted a time/frequency resource blocks approach to address illustrated in Fig. 2, for F = 10 frequency channels, from
the sum rate maximization problem subject to latency and which FU = 5 frequency channels allocated for the URLLC
slicing isolation constraints while guaranteeing the reliability traffic, and S = 6 minislots. By employing the orthogonal
requirements with the use of adaptive modulation coding. slicing, some frequency channels are allocated exclusively
In [14], the authors analyze the coexistence of eMBB and for the eMBB traffic and some exclusively for the URLLC
URLLC in fog-radio architectures where the URLLC traffic traffic. On the other hand, with non-orthogonal slicing the
is processed at the edge while eMBB traffic is handled at the same frequency channels can be shared between the two
cloud. In [15], the authors also study the orthogonal and non- services. However, differently from [7], we consider NOMA
orthogonal slicing of radio resources for eMBB and URLLC for URLLC, as indicated by the darker blue tone in Fig. 2. In
using a max-matching diversity (MMD) algorithm to allocate other words, we allow multiple URLLC devices to transmit
the frequency channels for the eMBB devices. However, none simultaneously in the same minislot.
of the mentioned works studied the performance of the joint The transmission of an eMBB device occupies a single
combination of NOMA, SIC decoding and frequency diversity frequency channel f and extends over the entire timeslot.
for URLLC traffic. Moreover, for eMBB traffic, we model only the standard
Motivated by the mURLLC scenarios predicted for beyond- scheduled transmission phase, hence assuming that radio ac-
5G and 6G networks, and based on recent works that address cess and competition among eMBB devices have been solved
the coexistence between eMBB and URLLC (in special [7]), prior to the considered time slot. A URLLC device, in turn,
we propose a framework that allows multiple URLLC devices
to share the same radio resources with eMBB devices in
a scalable manner, for both orthogonal and non-orthogonal
slicing of radio resources in the uplink. The main difference
of our work in comparison with [7] is that we consider the eMBB URLLC
use of NOMA for multiple URLLC devices. Our contribution
consists on the joint use of NOMA, Successive Interference
Cancellation (SIC) and frequency diversity as a solution to
improve the number of URLLC devices that can be connected BS eMBB
URLLC
to the same BS. In other words, multiple URLLC devices
share the same time/frequency resource and, in order to decode
the multiple URLLC signals (and also the eMBB signals in
the case of non-orthogonal slicing), the BS performs SIC eMBB URLLC
decoding. To characterize the performance trade-offs between URLLC
eMBB and URLLC, we evaluate the pairs of achievable sum
rates under predefined reliability requirements for orthogonal Fig. 1: Uplink transmissions to a common base station (BS)
and non-orthogonal slicing scenarios. We show that, even with from multiple eMBB and URLLC devices.

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.
 channel f is
nU
X
eMBB 1 URLLC User 2 URLLC Users Idle
yf [s] = hB,f xB,f [s] + hUu ,f xUu ,f [s] + wf [s], (1)
u=1
frequency channels

frequency channels
where xB,f [s] ∈ C is the symbol transmitted by an eMBB
device scheduled in the frequency channel f , xUu ,f [s] ∈ C

frequencies shared between

frequencies allocated for
is the symbol transmitted by the u-th URLLC device in the

eMBB and URLLC

frequency channel f , and wf [s] corresponds to an Additive

URLLC
White Gaussian Noise (AWGN) sample with zero mean and
unit variance, i.e. wf [s] ∼ CN (0, 1).
To compute the pairs of maximum achievable sum rates,
we evaluate the performance of a worst case scenario where
there is always an eMBB device transmitting in each frequency
channel f and the nU URLLC devices transmitting in all
time time
minislots.
minislot minislot
time slot time slot B. Performance Analysis of eMBB
(a) (b) In this subsection, we present the performance analysis of
eMBB that was originally presented in [7]. The eMBB device
Fig. 2: Illustration of the time-frequency grid used for the
aims at transmitting at the highest rate rB that is compatible
network slicing for the eMBB and URLLC services in the (a)
with the outage probability requirement B under a long-
orthogonal and (b) non-orthogonal scenarios. The darker blue
term average power constraint1 . This can be formulated as
tone indicates the overlap of URLLC transmissions.
the optimization problem
maximize rB ,
transmits within a single minislot across a subset of FU ≤ F subject to Pr {log2 (1 + GB,f PB (GB,f )) ≤ rB } ≤ B (2)
frequency channels, as a mean to achieve frequency diversity and E {PB (GB,f )} = 1,
and meet the reliability requirements [7]. Due to the low
latency requirement, each URLLC packet must be decoded where PB (GB,f ) is the instantaneous transmit power that is a
within the duration of a minislot, not being allowed to span function of the instantaneous channel gain GB,f . The optimal
over multiple minislots. solution to this problem is given by the truncated power inver-
Each radio resource is assumed to be within the time- and sion scheme. The eMBB device chooses a transmission power
frequency-coherence interval of the wireless channel, so that that is inversely proportional to GB,f if the latter is above
the wireless channel coefficients are constant within each min- a given threshold GminB,f , while it refrains from transmitting
islot. We also assume that the coefficients fade independently otherwise.
across the radio resources. The channel envelops as seen by In the absence of interference from other services, the only
the eMBB and the URLLC traffics, which we denote by hi,f source of outage for an eMBB transmission is the event that an
with i ∈ {B, U }, are independent and complex Gaussian eMBB device does not transmit because of insufficient SNR.
distributed, i.e., hi,f ∼ CN (0, Γi ), thus the channel gains The probability that GB,f is below Gmin B,f is
Gi,f = |hi,f |2 are exponentially distributed with average Γi . Gmin
!
 min B,f
The noise power at the BS is normalized to one, such that Pr(EB ) = Pr GB,f < GB,f = 1 − exp − . (3)
the average channel gain Γi can be also interpreted as the ΓB
average received Signal-to-Noise Ratio (SNR). Moreover, no Imposing the reliability requirement Pr(EB ) = B , the
Channel-State Information (CSI) is assumed at the URLLC threshold SNR becomes
devices, whereas the eMBB devices and BS are assumed to  
1
have perfect CSI as in [7]. Gmin
B,f = Γ B ln . (4)
1 − B
The outage probabilities of the eMBB and URLLC de-
vices are denoted as Pr(EB ) and Pr(EU ), respectively, and The instantaneous power PB (GB,f ) is chosen as a function
must satisfy the reliability requirements Pr(EB ) ≤ B and of the channel gain GB,f as
Pr(EU ) ≤ U . 
tar
 GB,f if G
 min
B,f ≥ GB,f
PB (GB,f ) = GB,f (5)
A. Signal Model 
0 if GB,f < Gmin ,
B,f

Let nU ∈ {N+ } denote the maximum number of URLLC 1 Notice that full CSI of eMBB device is assumed as in [7]. Since eMBB
devices transmitting simultaneously in the same minislot. The transmissions are scheduled, devices have sufficient time to undergo through
baseband signal received at the BS in minislot s and frequency CSI acquisition procedures [7], [10].

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.
where Gtar
B,f is the target SNR, which is obtained by imposing according to their sum of mutual information across the FU
the average power constraint as frequency channels, which is defined by
Gtar
B,f

Gmin
 FU
E {PB (GB,f )} = Γ 0, ΓB,f
X
= 1. Iusum = log2 (1 + σu,f ). (10)
ΓB B

f =1
This implies that the target SNR is
Given the reliability requirement Pr(EU ) ≤ U , the objec-
ΓB
Gtar
B,f =  min  , (6) tive is to obtain the maximum rate rU that is a function of the
GB,f
Γ 0, ΓB
number of frequency channels allocated for URLLC traffic.
R∞ Then, the URLLC sum rate is given by
where Γ(a, z) = z ta−1 e−t dt is the upper incomplete sum
gamma function. Finally, the outage rate achieved by the rU = nU rU . (11)
eMBB device is Increasing FU enhances the frequency diversity and, hence,
orth
rB = log2 (1 + Gtar makes it possible to satisfy the reliability target U at a larger
B,f ). (7)
rate rU [7]. Besides, during the computation of rU , the error
C. Performance Analysis of the URLLC device probabilities for all URLLC devices are computed individually.
Since the URLLC devices are assumed to have no CSI,
III. S LICING FOR E MBB AND URLLC
and for mathematical tractability and simplicity, herein, we
consider that all of them transmit with the same data rate rU . In this section, we consider the coexistence of eMBB
We also adopt a worst case assumption that there are always and NOMA URLLC devices for both orthogonal and non-
nU ∈ {N+ } URLLC devices transmitting in all minislots, thus orthogonal slicing.
there is always NOMA interference when nU > 1. The BS
performs SIC2 decoding to decode the multiple overlapping A. Orthogonal Slicing between eMBB and URLLC
signals. Under the orthogonal slicing scenario, FU out of F fre-
Let u denote the index of the URLLC device with channel quency channels for all minislots are allocated for URLLC
gain GUu ,f in the allocated frequency channel f . Besides, let traffic, while the remaining FB = F − FU channels are each
{1, · · · , nU } denote the SIC decoding ordering. The Signal-to- allocated to one eMBB device. The performance of the system
sum sum
Interference-plus-Noise Ratio (SINR) in the frequency channel is specified in terms of the pair (rB , rU ) of eMBB sum-rate
sum sum
f when decoding the u-th active URLLC device, and assum- rB and URLLC sum-rate rU .
ing that the devices with indexes {1, · · · , u − 1} have been The eMBB sum-rate is given by [7]
correctly decoded, reads sum orth
rB = (F − FU )rB , (12)
GUu ,f
σu,f = nU
. (8) where rBorth
is given by (7). Given a number FU of frequency
P
1+ GUj ,f channels allocated for URLLC, we compute the maximum
j=u+1
URLLC rate rU that guarantees the reliability constraint
The active URLLC device is decoded successfully if Pr(EU ) ≤ U for all nU URLLC devices transmitting simul-
FU taneously, as detailed in Section II-C.
1 X
log2 (1 + σu,f ) ≥ rU . (9)
FU B. Non-Orthogonal Slicing between eMBB and URLLC
f =1
In the non-orthogonal slicing scenario, all F frequency
During the SIC decoding procedure, the BS first attempts to
channels are used for both eMBB and URLLC services.
decode the strongest user among all the active URLLC devices
Hence, FU = FB = F . Due to the latency constraints,
in a minislot. If this user is correctly decoded, its interference
the decoding of a URLLC transmission cannot wait for the
is subtracted from the received signal, then the BS attempts
decoding of eMBB traffic. The eMBB requirements are less
to decode the next user in the order of strongest users, and so
demanding in terms of latency, and hence eMBB decoding
on. The SIC decoding procedure ends when the decoding of
can wait for the URLLC transmissions to be decoded first.
one active URLLC device fails or after all the active URLLC
This enables a SIC mechanism whereby URLLC packets
devices have been correctly decoded. We assume that all SIC
are successive decoded and then canceled from the received
decoding steps can be realized within the time duration of a
signal prior to the decoding of the eMBB signal [7]. As a
minislot.
consequence, during the decoding attempts of the URLLC
For simulation purposes, and in order to emulate the be-
packets, the interference of the eMBB traffic is always present.
haviour of the BS while performing the SIC decoding of
In the orthogonal case, as shown in (6), the variable Gtar
B,f is
the URLLC devices, we define the SIC decoding ordering
uniquely determined by the error probability target B and the
2 Note that, as presented in [7], SIC outperforms other techniques of multi- threshold SNR Gmin B,f defined in (4). For the non-orthogonal
user detection, such as puncturing. slicing, it may be beneficial to choose a smaller target SNR

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.
than the one given if (6), so as to reduce the interference
caused to URLLC transmissions. This yields the inequality [7]
ΓB
Gtar
B,f ≤ . (13)
Gmin

B,f
Γ 0, ΓB

The maximum allowed SNR for the eMBB devices, which

we denote by Gtar B,max , is set by the inequality in (13).
Consequently, the maximum allowed eMBB data rate is given
max
by rB = log2 (1 + Gtar B,max ) .
The objective of the analysis is to determine the rate pair
sum sum
(rB , rU ) for which the reliability requirements of the two
services are satisfied. To this end, first we fix an eMBB data
max
rate rB ∈ [0, rB ] and then we compute the maximum
achievable rate rU . During this computation, for a given
value of rU , we search for the minimum value of the SNR
min
Gtar tar
B ∈ [GB , GB,max ] that can be used for all eMBB devices. sum sum
The error probabilities for all eMBB and URLLC devices are Fig. 3: eMBB sum rate rB versus URLLC sum rate rU
computed individually. for the the orthogonal and non-orthogonal slicing when ΓU =
Considering the SIC decoding ordering {1, · · · , nU }, the 20 dB, ΓB = 10 dB, F = 10, U = 10−5 and B = 10−3 .
SINR in the frequency channel f while decoding the u-th
active URLLC device, and assuming that the devices with
indexes {1, · · · , u − 1} have been correctly decoded, is given
by
GUu ,f
σu,f = n U
. (14)
1 + Gtar
P
B,f + GUj ,f
j=u+1

The u-th URLLC device is correctly decoded if the condi-

tion given by (9) holds, but now considering the SINR given
by (14).
The SIC decoding performed in the non-orthogonal slicing
is similar to the procedure performed in the orthogonal sce-
nario and described in Section II-C. The only difference is that
in the non-orthogonal slicing there is always interference from
the eMBB devices in all FU frequency channels during the SIC
decoding of the URLLC devices. The BS tries to decode the Fig. 4: eMBB sum rate rB sum
versus URLLC sum rate rU sum
eMBB packets only after all the URLLC devices have been for the the orthogonal and non-orthogonal slicing when ΓU =
correctly decoded. If an error occurs during the SIC decoding 10 dB, ΓB = 20 dB, F = 10, U = 10−5 and B = 10−3 .
of the URLLC devices, the packets from the eMBB devices
are lost.
Fig. 3 shows the pairs of achievable sum rates considering
IV. N UMERICAL R ESULTS
ΓU = 20 dB and ΓB = 10 dB, that is, a scenario where the
In this section we present Monte Carlo simulation results for URLLC devices have better channel conditions than the eMBB
the orthogonal and non-orthogonal slicing of radio resources devices. In this situation, due to the high levels of interference
sum
between eMBB and URLLC. For the sake of tractability, we among the URLLC devices, the highest values of rU are
consider the cases where the maximum number of URLLC achieved when only one URLLC device is transmitting in the
devices transmitting simultaneously in the same minislot is minislot, that is, nU = 1. Moreover, as we increase nU , we
sum
nU ∈ {1, 2, 3, 4}, where nU = 1 is the scenario from [7]. decrease the achievable values of rU , which renders a trade-
Besides, we consider a time-frequency grid with F = 10 off between data rate and number of connected users. Another
frequency channels and reliability requirements U = 10−5 important conclusion is that the non-orthogonal slicing outper-
and B = 10−3 [7]. Besides, we assume that the URLLC and forms its orthogonal counterpart for the whole range of rB sum
,
eMBB devices are located in a similar environment, e.g. within since the former allows us to achieve pairs of sum rates that
a sector of a factory floor, aggregated by their proximity to are not possible to achieve using the latter. Note also that,
the BS. This renders the same average received SNR to all the when adopting the non-orthogonal slicing and for nU > 1,
sum
devices belonging to the same service class. the values of rU are approximately constant for the whole

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.
 sum
range of rB . mURLLC scenarios, where a large number of devices used for
Fig. 4 shows the pairs of achievable sum rates for an the control and/or monitoring of critical processes may require
opposite scenario where ΓU = 10 dB and ΓB = 20 dB, that URLLC connectivity in the coexistence with other applications
is, when the URLLC devices have worse channel conditions that require the high data rates provided by eMBB, e.g. video
than the eMBB devices. Compared to the previous case, now surveillance.
we can achieve much higher eMBB sum rates at the cost of
ACKNOWLEDGMENT
lower URLLC sum rates. Moreover, we observe that the the
URLLC sum rates achieved with nU > 1 are higher compared This research was financially supported by 6Genesis Flag-
to the case with nU = 1 due to the lower levels of interference ship project (grant no. 318927), FIREMAN project (grant no.
among the URLLC devices. For rB sum
< 25 bits/s/Hz, the 326201) and Academy Professor project from Academy of
orthogonal slicing outperforms the non-orthogonal strategy for Finland (grant no. 307492).
all values of nU , since it allows us to achieve higher values of R EFERENCES
sum
rU . The non-orthogonal slicing becomes advantageous for
sum [1] H. Tullberg, P. Popovski, Z. Li, M. A. Uusitalo, A. Hoglund, O. Bulakci,
rB > 25 bits/s/Hz, since it allows us to achieve URLLC M. Fallgren, and J. F. Monserrat, “The METIS 5G System Concept:
sum rates that are not possible to achieve with the orthogonal Meeting the 5G Requirements,” IEEE Commun. Mag., vol. 54, no. 12,
strategy. In this higher range, increasing nU also increases pp. 132–139, December 2016.
sum [2] M. Latva-Aho and K. Leppänen, “Key Drivers and Research Challenges
rU , which represent gains in both data rates and number of for 6G Ubiquitous Wireless Intelligence,” in 6G Wireless Summit, Levi,
connected devices. Note also that, in both Figs. 3 and 4, the Finland, Mar 2019.
curves obtained for nU = 1 match those in [7]. [3] W. Saad, M. Bennis, and M. Chen, “A Vision of 6G Wireless Systems:
Applications, Trends, Technologies, and Open Research Problems,”
Considering both slicing strategies, we conclude that the IEEE Netw., vol. 34, no. 3, pp. 134–142, 2020.
best compromise between number of NOMA users and data [4] GSM Association, “An Introduction to Network Slicing,” Tech. Rep.,
rates is achieved for limited number of users, in this case 2017. [Online]. Available: https://www.gsma.com/futurenetworks/wp-
content/uploads/2017/11/GSMA-An-Introduction-to-Network-
nU ≤ 2, which corroborates with other related works, e.g. Slicing.pdf
[16]. We also conclude that the non-orthogonal slicing is the [5] N. H. Mahmood, H. Alves, O. A. López, M. Shehab, D. P. M. Osorio,
best choice in applications where URLLC devices coexist with and M. Latva-Aho, “Six Key Features of Machine Type Communication
in 6G,” in 2020 2nd 6G Wireless Summit (6G SUMMIT), 2020, pp. 1–5.
eMBB devices that require very high data rates rates. Note that [6] L. Dai, B. Wang, Y. Yuan, S. Han, C. I, and Z. Wang, “Non-Orthogonal
NOMA with SIC decoding increases the receiver complexity Multiple Access for 5G: Solutions, Challenges, Opportunities, and
and yields higher delays in processing times, which must be Future Research Trends,” IEEE Commun. Mag., vol. 53, no. 9, pp. 74–
81, Sep. 2015.
taken into consideration in practical applications. [7] P. Popovski, K. F. Trillingsgaard, O. Simeone, and G. Durisi,
“5G Wireless Network Slicing for eMBB, URLLC, and mMTC: A
V. C ONCLUSIONS Communication-Theoretic View,” IEEE Access, vol. 6, pp. 55 765–
55 779, 2018.
We proposed the joint use of NOMA, SIC decoding and [8] Z. Wu, F. Zhao, and X. Liu, “Signal Space Diversity Aided Dynamic
frequency diversity as a solution to improve the number of Multiplexing for eMBB and URLLC Traffics,” in 2017 3rd. Int. Conf.
URLLC devices that are connected in the uplink to the same Comput. Commun. (ICCC), Dec 2017, pp. 1396–1400.
[9] A. Anand, G. De Veciana, and S. Shakkottai, “Joint Scheduling of
BS and when they share the same RAN with eMBB devices. URLLC and eMBB Traffic in 5G Wireless Networks,” in 2018 IEEE
The radio resources are shared between the two services Int. Conf. Comput. Commun. (INFOCOM), April 2018, pp. 1970–1978.
by employing orthogonal or non-orthogonal network slicing [10] A. A. Esswie and K. I. Pedersen, “Opportunistic Spatial Preemptive
Scheduling for URLLC and eMBB Coexistence in Multi-User 5G
strategies. Resorting to Monte Carlo simulations, we showed Networks,” IEEE Access, vol. 6, pp. 38 451–38 463, 2018.
that the proposed method allow multiple eMBB and URLLC [11] R. Abreu, T. Jacobsen, G. Berardinelli, K. Pedersen, N. H. Mahmood,
devices to transmit overlapping signals to same BS while their I. Z. Kovacs, and P. Mogensen, “On the Multiplexing of Broadband
Traffic and Grant-Free Ultra-Reliable Communication in Uplink,” in
reliability requirements are still met. We also demonstrated 2019 IEEE 89th Veh. Technol. Conf. (VTC2019-Spring), 2019, pp. 1–6.
that when the URLLC devices have better channel conditions [12] M. Alsenwi, N. H. Tran, M. Bennis, A. Kumar Bairagi, and C. S. Hong,
than the eMBB devices, the non-orthogonal slicing is advan- “eMBB-URLLC Resource Slicing: A Risk-Sensitive Approach,” IEEE
Commun. Lett., vol. 23, no. 4, pp. 740–743, April 2019.
tageous over the orthogonal slicing for the whole range of [13] P. K. Korrai, E. Lagunas, S. K. Sharma, S. Chatzinotas, and B. Ottersten,
eMBB sum rates. However, when the eMBB devices have “Slicing Based Resource Allocation for Multiplexing of eMBB and
better channel conditions than the URLLC devices, the non- URLLC Services in 5G Wireless Networks,” in 2019 IEEE Int. Workshop
Comput. Aided Model. Des. Commun. Links Netw. (CAMAD), Sep. 2019,
orthogonal slicing outperforms the orthogonal slicing only for pp. 1–5.
very high values of eMBB sum rates. [14] R. Kassab, O. Simeone, P. Popovski, and T. Islam, “Non-Orthogonal
As stated in [2], the current 5G NR network is not yet Multiplexing of Ultra-Reliable and Broadband Services in Fog-Radio
Architectures,” IEEE Access, vol. 7, pp. 13 035–13 049, 2019.
capable of meeting the very stringent latency and reliability [15] E. J. dos Santos, R. D. Souza, J. L. Rebelatto, and H. Alves, “Network
requirements of URLLC applications. Meeting these require- Slicing for URLLC and eMBB With Max-Matching Diversity Channel
ments requires hyper-flexible networks where technologies Allocation,” IEEE Commun. Lett., vol. 24, no. 3, pp. 658–661, 2020.
[16] O. L. Alcaraz López, H. Alves, P. H. Juliano Nardelli, and M. Latva-aho,
such as Artificial Intelligence (AI) and Machine Learning “Aggregation and Resource Scheduling in Machine-Type Communica-
(ML) can be used to determine the optimal radio resource tion Networks: A Stochastic Geometry Approach,” IEEE Trans. Wireless
allocations for BSs and users. The framework developed in this Commun., vol. 17, no. 7, pp. 4750–4765, July 2018.
work can be used in the specification of such 6G networks for

Authorized licensed use limited to: Istanbul Universitesi. Downloaded on March 31,2022 at 12:00:48 UTC from IEEE Xplore. Restrictions apply.