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

fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI

5G Wireless Network Slicing for

eMBB, URLLC, and mMTC:
A Communication-Theoretic View
PETAR POPOVSKI1 (Fellow, IEEE), KASPER F. TRILLINGSGAARD1 (Student Member, IEEE),
OSVALDO SIMEONE2 (Fellow, IEEE), and GIUSEPPE DURISI3 (Senior Member, IEEE)
1
Department of Electronic Systems, Aalborg University, Denmark (email: {petarp,kft}@es.aau.dk)
2
King’s College London, United Kingdom (email: [email protected])
3
Department of Electrical Engineering, Chalmers University of Technology, Sweden (email: [email protected])
Corresponding author: Petar Popovski (e-mail: [email protected]).
The work of Petar Popovski and Kasper F. Trillingsgaard has been in part supported by the European Research Council (Horizon 2020
ERC Consolidator Grant Nr. 648382 WILLOW). The work of Osvaldo Simeone has received funding from the European Research Council
(ERC) under the European Union Horizon 2020 research and innovation program (grant agreement 725731). The work of Giuseppe Durisi
was partly supported by the Swedish Research Council under grant 2016-03293.

ABSTRACT The grand objective of 5G wireless technology is to support three generic services
with vastly heterogeneous requirements: enhanced mobile broadband (eMBB), massive machine-type
communications (mMTC), and ultra-reliable low-latency communications (URLLC). Service heterogeneity
can be accommodated by network slicing, through which each service is allocated resources to provide
performance guarantees and isolation from the other services. Slicing of the Radio Access Network (RAN)
is typically done by means of orthogonal resource allocation among the services. This work studies the
potential advantages of allowing for non-orthogonal sharing of RAN resources in uplink communications
from a set of eMBB, mMTC and URLLC devices to a common base station. The approach is referred to
as Heterogeneous Non-Orthogonal Multiple Access (H-NOMA), in contrast to the conventional NOMA
techniques that involve users with homogeneous requirements and hence can be investigated through a
standard multiple access channel. The study devises a communication-theoretic model that accounts for the
heterogeneous requirements and characteristics of the three services. The concept of reliability diversity is
introduced as a design principle that leverages the different reliability requirements across the services in
order to ensure performance guarantees with non-orthogonal RAN slicing. This study reveals that H-NOMA
can lead, in some regimes, to significant gains in terms of performance trade-offs among the three generic
services as compared to orthogonal slicing.

INDEX TERMS 5G wireless, network slicing, wireless communication theory, NOMA, eMBB, mMTC,
URLLC.

I. INTRODUCTION (IoT) devices, which are only sporadically active and send
URING the past few years, there has been a growing small data payloads; (c) URLLC supports low-latency trans-
D consensus that 5G wireless systems will support three
generic services, which, according ITU-R, are classified as
missions of small payloads with very high reliability from
a limited set of terminals, which are active according to
enhanced mobile broadband (eMBB), massive machine-type patterns typically specified by outside events, such as alarms.
communications (mMTC), and ultra-reliable and low-latency This paper studies the problem of enabling the coexistence
communications (URLLC) (also referred to as mission- of the three heterogeneous services within the same Radio
critical communications) [1], [2]. A succinct characteriza- Access Network (RAN) architecture. We describe below in
tion of these services can be put forward as follows: (a) more details the requirements of the three services.
eMBB supports stable connections with very high peak data eMBB traffic can be considered to be a direct extension of
rates, as well as moderate rates for cell-edge users; (b) the 4G broadband service. It is characterized by large pay-
mMTC supports a massive number of Internet of Things loads and by a device activation pattern that remains stable

VOLUME 4, 2016 1

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

over an extended time interval. This allows the network to

schedule wireless resources to the eMBB devices such that no
two eMBB devices access the same resource simultaneously.
The objective of the eMBB service is to maximize the data
rate, while guaranteeing a moderate reliability, with packet
error rate (PER) on the order of 10−3 .
In contrast, an mMTC device is active intermittently and
uses a fixed, typically low, transmission rate in the uplink.
A huge number of mMTC devices may be connected to a
given base station (BS), but at a given time only an unknown
(random) subset of them becomes active and attempt to send
their data. The large number of potentially active mMTC
devices makes it infeasible to allocate a priori resources to
individual mMTC devices. Instead, it is necessary to provide
resources that can be shared through random access. The size
eMBB URLLC mMTC
of the active subset of mMTC devices is a random variable,
whose average value measures the mMTC traffic arrival rate. FIGURE 1. The considered scenario with uplink transmissions to a common
The objective in the design of mMTC is to maximize the base station (BS) from devices using the three generic 5G services.
arrival rate that can be supported in a given radio resource.
The targeted PER of an individual mMTC transmission is
typically low, e.g., on the order of 10−1 . same type (see, e.g., [5]). This is further discussed next.
Finally, URLLC transmissions are also intermittent, but
the set of potential URLLC transmitters is much smaller than A. NETWORK SLICING OF WIRELESS RESOURCES:
for mMTC. Supporting intermittent URLLC transmissions H-OMA AND H-NOMA
requires a combination of scheduling, so as to ensure a certain Consider an uplink scenario in which a set of eMBB, mMTC
amount of predictability in the available resources and thus and URLLC devices is connected to a common BS, as shown
support high reliability; as well as random access, in order to in Fig. 1. We note that the designing uplink access is more
avoid that too many resources being idle due to the intermit- complex than the corresponding problem for the downlink
tent traffic. Due to the low latency requirements, a URLLC due to the lack of coordination among users. Orthogonal and
transmission should be localized in time. Diversity, which is non-orthogonal slicing of the RAN among the three services
critical to achieve high reliability [3], can hence be achieved are illustrated in Fig. 2(a) and (b), respectively.
only using multiple frequency or spatial resources. The rate The conventional orthogonal allocation depicted in
of a URLLC transmission is relatively low, and the main Fig. 2(a) operates in the frequency domain and allots different
requirement is ensuring a high reliability level, with a PER frequency channels to eMBB, mMTC, or URLLC devices.
typically lower than 10−5 , despite the small blocklengths. eMBB and mMTC transmissions are allowed to span mul-
In 5G, heterogeneous services are allowed to coexist tiple time resources. In contrast, in order to guarantee the
within the same network architecture by means of net- latency requirements discussed above, URLLC transmissions
work slicing [4]. Network slicing allocates the network are localized in time and are spread over multiple frequency
computing, storage, and communication resources among channels to gain diversity. Furthermore, since the URLLC
the active services with the aim of guaranteeing their iso- traffic is bursty, the resources allocated to URLLC users may
lation and given performance levels. In this paper, we be largely unused. This is because the channels reserved for
are interested in the “slicing" of RAN communication re- URLLC are idle in the absence of URLLC transmission.
sources for wireless access. The conventional approach to Importantly, orthogonal slicing does not preclude the shar-
slice the RAN is to allocate orthogonal radio resources to ing of wireless resources among devices of the same type.
eMBB, mMTC, and URLLC devices in time and/or fre- For example, multiple eMBB users may transmit on the same
quency domains, consistently with the orthogonal allocation allotted frequency channels by using NOMA [5]. Therefore,
of wired communication resources. However, wireless re- in order to distinguish orthogonality among signals origi-
sources are essentially different due to their shared nature. nating from devices of the same type, as in conventional
Using communication-theoretic analysis, this work demon- Orthogonal Multiple Access (OMA), from the orthogonality
strates that a non-orthogonal allocation that is informed by among different services, we refer to the approach in Fig. 2(a)
the heterogeneous requirements of the three services can as Heterogeneous Orthogonal Multiple Access (H-OMA).
outperform the standard orthogonal approach. Importantly, As mentioned, in this work, we investigate the potential
the considered non-orthogonal approach multiplexes het- advantages of a non-orthogonal allocation of RAN resources
erogeneous services, and is hence markedly distinct from among multiple services, which we refer to as Heterogeneous
the conventional Non-Orthogonal Multiple Access (NOMA) Non-Orthogonal Multiple Access (H-NOMA). Fig. 2(b) de-
methods that share radio resources only among devices of the picts an instance of H-NOMA. By comparison with the H-
2 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 rate of eMBB
eMBB mMTC URLLC idle user 2 NOMA rate H-NOMA

OMA H-OMA
frequency

frequency

frequencies allocated
frequencies reserved
for URLLC

for URLLC
rate of URLLC
user 1 activity
(a) (b)

URLLC mMTC
H-NOMA H-NOMA
time time reliability arrivals
(a) (b) H-OMA H-OMA

FIGURE 2. Illustration of the slicing of the wireless resources in a

time-frequency frame for supporting the three generic services with: (a)
Heterogeneous Orthogonal Multiple Access (H-OMA) (b) Heterogeneous
Non-Orthogonal Multiple Access (H-NOMA). The idle time-frequency blocks
are not used for transmission due to absence of traffic. With H-OMA, some of
the frequency channels are reserved to URLLC traffic, whereas with H-NOMA
the same channels are allocated to both URLLC and eMBB.

eMBB eMBB
rate rate
OMA solution in Fig. 2(a), under H-NOMA, the frequency (c) (d)
resources that were allocated only to mMTC or URLLC
traffic can also be granted to the eMBB users. In this way, H- FIGURE 3. Illustration of performance trade-offs for: (a) standard OMA and
NOMA within the same traffic type; (b,c,d) H-OMA and H-NOMA between
NOMA may allow for a more efficient use of radio resources heterogeneous services.
as compared to H-OMA by avoiding unused resources due
to URLLC or mMTC inactivity. This may yield a higher
spectral efficiency for the eMBB users that can benefit from devices being active, and the eMBB transmission rate or
the intermittent nature of mMTC and URLLC traffic. How- spectral efficiency. The figure illustrates the fact that the
ever, the mutual interference between eMBB and mMTC or eMBB rate is not affected by the URLLC packet arrival rate
URLLC transmissions may significantly degrade the perfor- under H-OMA, while the resulting interference impairs the
mance for all the involved services. Ensuring desired perfor- performance of H-NOMA. As an alternative performance
mance levels is hence more challenging with H-NOMA. evaluation, Fig. 3(c) shows the trade-off between the reli-
In this paper, we tackle this problem by developing a ability of URLLC transmissions and the eMBB rate. The
communication-theoretic model that aims at capturing the example highlights the fact that a non-trivial trade-off exists
essential performance trade-offs and design insights for H- for both H-OMA and H-NOMA. In fact, URLLC reliability
OMA and H-NOMA. More specifically, the main goals of can be improved by taking away frequency resources from
this work can be illustrated using Fig. 3, as discussed next. eMBB and allocating them to URLLC. As a final illustration,
To start, Fig. 3(a) depicts the type of results that are Fig. 3(d) depicts the trade-off between the arrival rate of
of interest when studying conventional OMA and NOMA mMTC devices and the eMBB rate. In a manner similar to
within a given service type, as done in a growing line of Fig. 3(c), this figure suggests that, even under H-OMA, the
work [5], [6]. These results rely on the classical analysis of spectral efficiency of the eMBB user that shares the resources
the multiple access channel, in which all users have identical with mMTC devices can be traded off for the mMTC arrival
reliability requirements and block lengths and the goal is to rate by a proper allocation of radio resources.
characterize the region of achievable rates [7] as the block
length grows large. B. FURTHER RELATED WORKS
In contrast, Fig. 3-(b,c,d) exemplify the type of results In addition to the mentioned literature on conventional
that are of interest when evaluating the performance trade- NOMA, here we briefly review works that directly tackle the
offs between heterogeneous services that are allowed by coexistence of heterogeneous services. A logical architecture
H-OMA and H-NOMA. As a first example to be further for network slicing in 5G in the presence of orthogonal
elaborated on in the paper, Fig. 3(b) shows the trade-off slicing has been presented in [4] and [8]. The downlink
between the URLLC activity, i.e., the probability of URLLC multiplexing of URLLC and eMBB is studied in [9] and
VOLUME 4, 2016 3
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

[10]. These works investigates the dynamic scheduling of adopts the classical frame error rate. Our results show
URLLC traffic over ongoing eMBB transmissions by ab- that, if reliability diversity is properly exploited, non-
stracting the operation at the physical layer. In [11], the orthogonal slicing can lead, in some regimes, to impor-
authors treat the problem of resource allocation for mMTC tant gains in terms of performance trade-offs among the
and URLLC in a new radio (NR) setting by focusing on the three generic services.
role of feedback. Orthogonal resource allocation for mMTC The paper is organized as follows. The next section
and eMBB users is studied in [12] by accounting for inter- presents the system model and provides a performance anal-
cell interference. In [13], grant-free uplink transmissions are ysis of each of the three services when considered in iso-
considered for the three services by considering concrete lation. Section III treats the slicing of resources to support
transmission/modulation/spreading methods for supporting eMBB and URLLC, while Section IV is dedicated to the
the three services. slicing of resources for eMBB and mMTC. Both sections
provide a description of the proposed theoretical framework
C. MAIN CONTRIBUTIONS
as well as numerical results illustrating the tradeoff between
The main contributions are as follows.
the services for both H-OMA and H-NOMA schemes. The
• We propose a communication-theoretic model that is
conclusions are given in Section V-A, while Section V-B
tractable and yet captures the key features and require- contains discussion on possible generalizations of the model
ments of the three services. Unlike [13], in which the considered in this paper. Two appendices, containing the
authors focus on grant-free access for all services, the technical details of some of the derivations, conclude the
proposed model takes into account the difference in paper.
arrival processes and traffic dynamics that are inherent
to each individual service. The proposed model can be
seen as an extension of the classical multiple access II. SYSTEM MODEL
channel model that underlies the analysis of conven- We are interested in understanding how the three service
tional NOMA in the sense that it accounts for the described in Section I, i.e., eMBB, URLLC, and mMTC,
coexistence of heterogeneous services. should efficiently share the same radio resources in the uplink
• We first analyze the performance of orthogonal slicing, when communicating to a common BS. We consider F radio
or H-OMA, for all three services. We focus on achiev- resources, where each resource occupies a single frequency
able transmission rates for eMBB and URLLC, under channel and a single time slot. A radio resource, which is
the respective target reliability, and on the throughput indexed by f ∈ {1, ..., F }, contains n symbols. The n
for mMTC. symbols are further divided into S minislots, where each
• We then consider the performance of H-NOMA. Al- minislot consists of nS = n/S symbols. Fig. 4 shows an
though the modeling approach allows to study an arbi- example of a time-frequency grid.
trary combination of services, in this paper we have fo- We assume that the transmission of an eMBB user oc-
cused on the analysis of two specific cases as illustrated cupies a single radio resource at a given frequency f ∈
in Fig. 2, namely: (i) slicing for URLLC and eMBB, {1, ..., F }. In contrast, due to latency constraints, a URLLC
and (ii) slicing for mMTC and eMBB. In the case user transmits within a single minislot across a subset of
of URLLC-eMBB slicing, among other schemes, we FU ≤ F frequency channels. An URLLC device may be
consider the technique of puncturing, which is currently active in an allocated minislot with probability aU . Finally,
under consideration in 3GPP [10]. It is noted that, while the set of mMTC users is allowed to access the channel only
of interest, H-NOMA between URLCC and mMTC may at a specified radio resource fM ; for the example on Fig. 4 we
be problematic due to the need to ensure reliability guar- have fM = F . The number AM of active mMTC devices in
antees for URLLC devices in the presence of the random such a resource is distributed as AM ∼ Poisson(λM ), where
interference patterns caused by mMTC transmissions. λM is the mean value, referred to as mMTC arrival rate.
• Among the main conclusions, our study demonstates Some comments regarding the modelling choices made
that non-orthogonal slicing, or H-NOMA, can achieve above are in order. First, for eMBB traffic, we focus on the
service isolation in the sense of ensuring performance standard scheduled transmission phase, hence assuming that
levels for all services by leveraging their heterogeneous radio access and competition among eMBB devices have
reliability requirements. We refer to this design princi- been resolved prior to the considered time slot. Second, we
ple as reliability diversity. As it will be discussed, the do not model collisions among URLLC devices. We assume
heterogeneity leveraged by reliability diversity is not instead that a single URLLC device is allocated a number of
only in terms of the numerical values of the reliability minislots in the given slot, over which it is active with some
levels, but also in terms of very definition of reliability probability. On the contrary, we do model the random access
across the three services. For example, the reliability phase for mMTC traffic, since this is the key transmission
metric typically considered for mMTC is the fraction of phase for this type of traffic, due to the massive population
detected devices among the massive set of active users, of devices. Extensions of our model will be discussed in
whereas for eMBB and URLLC services one typically Section V-B.
4 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 BS is assumed to have perfect CSI.
eMBB mMTC URLLC idle The error probabilities of the eMBB, URLLC, and mMTC
devices are denoted as Pr(EB ), Pr(EU ) and Pr(EM ), re-
spectively. These probabilities must satisfy the reliability
requirements Pr(EB ) ≤ B , Pr(EU ) ≤ U and Pr(EM ) ≤
7
frequency

M , where
channel

radio resource at a U  B  M . (1)

6
specific frequency channel
The large differences in reliability levels among the services,
5
as well as their different definitions, which we will introduce
4
shortly, motivate the introduction of the concept of reliability
diversity. Reliability diversity refers to system design choices
that leverage the differences among the supported services in
3
terms of reliability requirements and definitions. For exam-
2 ple, as we will see, strict per-packet reliability guarantees are
typically enforced for eMBB and URLLC devices, whereas
1 the notion of reliability for mMTC devices is less stringent
and typically involves the computation of averages over a
time
minislot large group of active devices.
time slot
A. SIGNAL MODEL
FIGURE 4. An example of H-NOMA allocation in the time-frequency grid with To summarize the main assumptions discussed so far and to
F = 7 resources and S = 6 minislots. A single resource (frequency channel) fix the notation, we assume that each eMBB user is sched-
is allocated for mMTC transmission. Each URLLC transmission is spread over
FU = 4 frequency channels.
uled on a single frequency channel within the considered F
frequency resources; each URLLC device occupies FU ≤ F
frequencies resources, numbered without loss of generality
Each radio resource f is assumed to be within the time- as f = 1, . . . , FU , in a given minislot; and a set of mMTC
and frequency-coherence interval of the wireless channel, devices is available for transmission in a channel frequency
so that the wireless channel coefficients are constant within fM ∈ {1, . . . , F }.
each radio resource. Furthermore, we assume that the channel Let Ys,f ∈ CnS denote the received vector corresponding
coefficients fade independently across the F radio resources. to the minislot s ∈ {1, 2, . . . , S} and the frequency channel
The channel coefficients of the eMBB, URLLC, and the f ∈ {1, 2, . . . F }. Based on the given assumptions, the
mMTC devices, which we denote by HB,f , HU,f , and Hm,f , received signal can be written as
m ∈ {0, . . . , AM },1 are independent and Rayleigh dis- Ys,f = HB,f XB,s,f + HU,f XU,s,f
tributed, i.e., HB,f ∼ CN (0, ΓM ), HU,f ∼ CN (0, ΓU ), and AM
Hm,f ∼ CN (0, ΓM ) for m ∈ {0, . . . , AM } across all radio X
+ H[m],f X[m],s,f + Zs,f , (2)
resources f ∈ 1, ..., F . The channel gains for the three ser- m=1
vices in a radio resource f are denoted by GB,f = |HB,f |2 ,
GU,f = |HU,f |2 , and Gm,f = |Hm,f |2 for m ∈ 1, ..., AM . where XB,s,f is the signal transmitted by an eMBB user
The average transmission power of all devices is normal- scheduled in the frequency resource f ; XU,s,f is the signal
ized to one. The differences in the actual transmission power transmitted by a URLLC device transmitting in minislot s
across various users and in the path loss are accounted for and frequency f ; X[m],s,f is the signal transmitted by one
through the average channel gains ΓB , ΓU , and ΓM . Further- of the AM active mMTC devices in frequency f ; and Zs,f
more, the power of the noise at the BS is also normalized represents the noise vector, whose entries are i.i.d. Gaussian
to one, so that the received power equals the signal-to-noise with zero mean and unit variance. The notation [m] for the
ratio (SNR) for each device. The number of symbols nS in a mMTC devices, which indicates ordering, will be formally
minislot is assumed sufficiently large to justify an asymptotic introduced in Section II-D.
information-theoretic analysis. Extensions of our analysis to We emphasize that the transmitted eMBB signal XB,s,f
capture finite-blocklength effects [14] will be considered in in (2) is zero if no eMBB user is scheduled in frequency
future works. Due to latency and protocol constraints to be channel f ; similarly, the URLLC signal XU,s,f is zero if no
detailed later, no channel-state information (CSI) is assumed URLLC device transmits in minislot s and frequency f , e.g.,
at the URLLC and at the mMTC devices. In contrast, the if f > FU ; and the mMTC signals {X[m],s,f } are similarly
eMBB devices are assumed to have perfect CSI. Finally, the all equal to zero if the channel f is not allocated to mMTC
traffic, i.e., if f 6= fM .
1 Throughout, we use the convention that the subscripts B, U , and M As discussed, with H-OMA, resources are allocated exclu-
indicate a quantity referring to eMBB, URLLC, and mMTC, respectively. sively to one of the three services, while, with H-NOMA,
VOLUME 4, 2016 5
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

resources can be shared. In the remainder of this section, we frequency channel f . This dependence is kept here in view to
study the performance of the three traffic types in an H-OMA the extension to H-NOMA in the next sections.
setting, that is, in the absence of mutual interference. We Based on the power-inversion scheme, the instantaneous
also introduce the metrics that will be used to evaluate the power PB (GB,f ) is chosen as a function of the instantaneous
performance of the three services. channel gain GB,f as
( tar
GB,f
B. EMBB if GB,f ≥ Gmin
B,f
PB (GB,f ) = GB,f (8)
Consider a radio resource f allocated exclusively to an 0 if GB,f < Gmin
B,f ,
eMBB user. As mentioned, the eMBB is aware of the CSI
where Gtar
B,f is the target SNR, which is obtained from the
GB,f and can use it in order to select its transmission power
threshold Gmin
B,f by imposing the average power constraint as
PB (GB,f ). The objective is to transmit at the largest rate
Z ∞
rB,f that is compatible with the outage probability require- 1 −x/ΓB
1 = E[PB (GB,f )] = e PB (x)dx
ment B under a long-term average power constraint. This min
GB,f ΓB
can be formulated as the optimization problem
Gtar
B,f
 Gmin 
B,f
maximize rB = γ 0, , (9)
ΓB ΓB
subject to Pr[log2 (1 + GB,f PB (GB,f )) < rB ] ≤ B with γ(·, ·) being the lower incomplete gamma function. This
E[PB (GB,f )] = 1. (3) implies that the target SNR is
The optimal solution to this problem is given by truncated ΓB
Gtar
B,f =  Gmin
. (10)
power inversion [15]. Accordingly, the eMBB device chooses γ 0, ΓB,f
B
a transmission power that is inversely proportional to the
channel gain GB,f if the latter is above a given threshold It follows from (4)–(10) that the solution to the problem
Gmin
B,f , while it refrains from transmitting otherwise. (3), which is the outage rate rB,f under outage probability
Beside being theoretically justified by the mentioned B , is given by
rate-maximization problem, the threshold-based transmis-
rB,f = log2 1 + Gtar


sion strategy discussed above also captures the fact that B,f , [bits/symbol]. (11)
eMBB devices only transmit if the current SNR is suffi- orth
We refer to the resulting rate as rB for reference. Note that
cient to satisfy minimal rate requirements. This is the case it does not depend on f under the assumptions of this section.
in most communication standards, such as LTE, in which
the transmission mode is selected from a set of allowed C. URLLC
modulation and coding schemes with given SNR constraints. The URLLC device transmits data in the allocated FU fre-
As we will discuss below, the scheme has the additional quency channels of a minislot, with activation probability aU .
analytical advantage of relating directly outage probability Hence, the number of URLLC transmissions during the time
and probability of activation for an eMBB user. We remark slot is a random variable SU ∼ Bin(S, aU ). We assume that
that the analysis could be extended to other design criteria each URLLC transmission carries a different message, and
such as the maximization of the average transmission rate. that, due to the low latency requirement, each message must
Based on the discussion above, the probability that the be decoded as soon as the relevant minislot is received. This
eMBB user transmits is given by implies that the URLLC device cannot code across multiple
−Gmin minislots.
aB = Pr GB,f ≥ Gmin B,f /ΓB .
 
B,f = e (4)
Unlike eMBB users, the URLLC device is not aware of the
Furthermore, in the absence of interference from other ser- CSI {GU,f } for the FU allocated frequency resources. This
vices, the only source of outage for an eMBB transmission is assumption is justified by the fact that CSI at the URLLC
precisely the event that an eMBB does not transmit because device would require signaling exchange before transmis-
of an insufficient SNR level. Hence, the probability of error sion, which entails extra latency as well as a potential loss
equals in terms of reliability. In fact, the high reliability constraint
Pr(EB ) = 1 − aB . (5) would enforce an even higher reliability requirement on the
auxiliary procedure of CSI signaling. As a result of the lack
Imposing the reliability condition
of CSI, no power or rate adaptation is possible for URLLC
Pr(EB ) = B (6) devices.
We choose the rate rU as the performance metric of choice.
we obtain the value of the threshold SNR
  In the absence of interference from other services, outage
min 1 occurs with probability
GB,f = ΓB ln . (7)
1 − B  
FU
Note that, in the absence of interference and under the given 1 X
Pr(EU ) = Pr log2 (1 + GU,f ) < rU  . (12)
assumptions, the threshold SNR Gmin FU
B,f does not depend on the f =1

6 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

Imposing the reliability condition P(EU ) = U allows us This quantity can be computed by means of Monte Carlo
to obtain the maximum allowed rate rU . We will refer to numerical methods.
orth
this quantity as rU (FU ) for reference. Note that increasing
FU enhances the frequency diversity and, hence, makes it III. SLICING FOR EMBB AND URLLC
possible to satisfy the reliability target U at a larger rate rU . In this section, we consider the coexistence of eMBB and
URLLC devices, while assuming that there is no mMTC
D. MMTC traffic, i.e., that the mMTC arrival rate is λM = 0. We first
The key property of the mMTC traffic is that the set of briefly recall, using the results in the previous section, how
mMTC devices that transmit in a given radio resource is the performance of the two services can be evaluated for the
random and unknown. An mMTC transmission has a fixed case of H-OMA, and then analyze the more complex scenario
rate rM and consumes one radio resource of n channel uses. of H-NOMA.
Given the rate rM and the reliability constraint M , we
focus on the maximum arrival rate λM that can be supported A. ORTHOGONAL SLICING: H-OMA FOR EMBB AND
by the system as the performance criterion of interest. As URLLC
detailed below, the probability of error measures the fraction In the case of orthogonal slicing, i.e., under H-OMA, we
of incorrectly decoded devices among the active ones. assume that FU out of the F frequency radio resources for
SIC at the BS is a useful strategy to improve the perfor- all minislots in the given radio resource are allocated to the
mance of mMTC traffic. As discussed next, a SIC decoder URLLC transmissions, while the remaining FB = F − FU
can leverage power imbalances and other mechanisms not re- radio resources are each allocated to one eMBB user. Note
viewed here (see, e.g., [16]), in order to sequentially improve that, in each minislot, the probability that the FU frequency
the reliability of simultaneous mMTC transmissions. channels allocated to URLLC traffic are unused is the com-
To characterize the performance achievable with SIC, we plement of the activation probability, i.e., 1 − aU .
let [m] denote the index of the mMTC device with the m- The performance of the system is specified in terms of the
th largest channel gain {Gm,fM } for the allocated frequency the pair (rB , rU ) of eMBB sum-rate rB and URLLC rate rU
f = fM . In the rest of this section, we drop the dependence achievable at the given reliability levels (B , U ). The eMBB
on fM for simplicity of notation. By definition, we then sum-rate is obtained as
have the inequalities G[1] ≥ G[2] ≥ . . . ≥ G[AM ] . In the
orth
absence of interference from eMBB and URLLC traffic, the rB = (F − FU )rB , [bits/symbol] (17)
SINR σ[m0 ] available when decoding the signal of the m0 −th orth
where rB is obtained as explained in Section II-B. The
mMTC device, under the additional assumption that the
URLLC rate rU is computed from (12) by imposing the
devices with indices [1], . . . , [m0 − 1] are correctly decoded,
equality Pr(EU ) = U as detailed in Section II-C.
depends only on its channel gain G[m0 ] and on the channels
gains of the other active mMTC devices as
B. NON-ORTHOGONAL SLICING: H-NOMA FOR EMBB
G[m0 ] AND URLLC
σ[m0 ] = PAM . (13)
1+ m=m0 +1 G[m] We now consider non-orthogonal slicing, or H-NOMA,
whereby all F frequency channels are used for both eMBB
The m0 −th mMTC device is correctly decoded if the in- and URLLC transmissions. Hence, FU = FB = F . With
equality log2 (1 + σ[m0 ] ) ≥ rM holds; and, if decoding non-orthogonal slicing, eMBB and URLLC transmissions
is successful, the signal from the device is subtracted from interfere, and, hence, the rate pair (rB , rU ) cannot be directly
the received signal. We let DM be the random number of obtained from the analysis in Section II. We next describe
mMTC devices in outage, i.e., DM is the largest integer different decoding architectures, and derive corresponding
in {0, . . . , AM } satisfying, for all k ∈ {1, . . . , DM }, the achievable pairs rB and rU for non-orthogonal slicing.
inequality Decoding Architectures: A key observation in the design
log2 (1 + σ[k] ) ≥ rM (14) of decoding schemes is that, due to latency constraints, the
decoding of a URLLC transmission cannot wait for, and
The error rate of the mMTC devices is then quantified as the hence depend upon, the decoding of eMBB traffic. In fact,
ratio decoding of a URLLC transmission can only rely on the
E[DM ] signal received in the given minislot. This constraint prevents
Pr(EM ) = (15)
λM SIC decoders whereby eMBB transmissions are decoded first
between the average number of users in outage, namely and canceled from the received signal prior to decoding of
E[DM ], and the average number λM of active users. The the URLLC messages. Note also that, because of the het-
maximum rate λM that can be supported under the reliability erogeneity of reliability requirements, decoding eMBB first
condition Pr(EM ) = M is defined for reference as and then URLLC in a SIC fashion would require decoding
the eMBB traffic at the same level of reliability needed for
λorth
M (rM ) = max{λM : Pr(EM ) ≤ M }. (16) the URLLC traffic. As a result of these considerations, in
VOLUME 4, 2016 7

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

H-NOMA with SIC the URLLC transmissions should be as noise. In contrast, the eMBB decoder at the BS can either
decoded while treating eMBB signals as an additional noise. leverage SIC by decoding URLLC traffic first, or rather treat
In contrast to URLLC traffic, eMBB requirements are less any minislot occupied by URLLC traffic as erased. These are
demanding in terms of latency, and hence eMBB decoding two extreme points among all possible eMBB decoders.
can wait for URLLC transmissions to be decoded first. This Rate Region: The objective of the analysis is to determine
enables a SIC mechanism whereby URLLC messages are the rate region (rB , rU ) for which the target error probabili-
decoded and then canceled from the received signal prior to ties of the two services are satisfied. To this end, we fix the
orth
decoding of the eMBB signal. Since the reliability of URLLC URLLC rate rU ∈ [0, rU (F )], and compute the maximum
is two or more orders of magnitude higher than eMBB, the attainable eMBB rate rB . We recall that the available degrees
performance of eMBB under the described SIC decoder is of freedom in the design are the target SNR Gtar B,f and the
expected to be close to the ideal orthogonal case in which minimum channel gain Gmin B,f at which an eMBB device is
no interference from URLLC traffic is present. This design active (or equivalently the activation probability (4)), as well
choice is an instance of reliability diversity. as the erasure code parameter k if a puncturing approach is
That being said, the SIC decoder may be ruled out by adopted for eMBB decoding. We also emphasize that, unlike
considerations such as complexity. In such circumstances, the orthogonal case, the target SNR Gtar B,f and the minimum
one could adopt another decoding approach that is, in a sense, SNR GminB,f are separate degrees of freedom, which are related
diametrically opposite to SIC in its treatment of URLLC by the inequality (20).
interference. Such a decoder treats any minislot that contains We start by imposing the reliability constraint for the
a URLLC transmission as erased or punctured. This option URLLC user, which yields the following condition for both
of H-NOMA with puncturing is currently being considered SIC and erasure decoder:
within the 5G community [10]. Note that this approach re-
 
FU
!
quires the decoder at the BS to be able to detect the presence 1 X G U,f
Pr(EU ) = Pr log2 1 + < rU  ≤ U .
of URLLC transmissions, e.g., via energy detection. FU 1 + δf Gtar
B,f
f =1
Encoding: If H-NOMA with SIC is used, we set the eMBB (21)
rate to Here, {δf }F U
f =1 are independent Bernoulli random variables
SIC with parameter aB given in (4). Recall that aB is a function
rB,f (Gtar tar
B,f ) = log2 (1 + GB,f ), [bits/symbol] (18)
of Gmin tar
B,f . The term δf GB,f represents the interference power
where Gtar caused by an eMBB transmission on frequency channel f
B,f represents the target SNR for eMBB transmis-
sion, which is to be determined. to the URLLC traffic. The inequality (21) imposes a joint
In contrast, in H-NOMA with puncturing and erasure constraint on both Gtar min
B,f and GB,f . Next, we impose the reli-
decoding, the eMBB device applies an outer erasure code ability constraint for eMBB traffic by considering separately
with rate 1 − k/S, which is concatenated to the codebook SIC and erasure decoders.
used in the physical-layer transmission of the eMBB encoder.
Thanks to the erasure code, the decoder is able to correct 1) SIC decoder
k ≤ S erased minislots, while, if the number of URLLC Under H-NOMA the decoding of an eMBB message is
transmissions is larger than k, the decoding process fails. generally affected by the interference from the URLLC users.
The parameter k needs to be designed so as to satisfy the However, this interference is not present if: (i) there are
target error rate B for eMBB users. The resulting data rate no URLLC transmissions, i.e., SU = 0; or (ii) if URLLC
for eMBB transmission in frequency channel f is transmissions are present, i.e., SU > 0, but the corresponding
  signals are decoded successfully and canceled by the SIC
pun k decoder. As for the latter event, since the interference from
rB,f (Gtar
B,f , k) = 1 − log2 (1 + Gtar
B,f ). (19)
S eMBB users and the fading gains are constant across the
Regarding the selection of the target SNR Gtar minislots, either all URLLC transmissions are decoded in-
B,f , we recall
that in the orthogonal case, as shown in (10), the variable correctly (event EU ) or they are all correctly decoded (event
Gtar E U ).
B,f is uniquely determined by the error probability tar-
get B via the threshold SNR Gmin Based on the discussion above, we can bound the eMBB
B,f defined in (7). In contrast,
with non-orthogonal slicing, it may be beneficial to choose a error probability by distinguishing the case in which the
smaller target SNR than the one given by (10), so as to reduce eMBB transmission is subject to interference from URLLC
the interference caused to URLLC transmissions. This yields signals, and the case in which is not, using the law of total
the inequality probability, as presented through the equations (22)-(24).
Here, equality (22) holds because the only source of outage
ΓB for eMBB in absence of URLLC interference is the instanta-
Gtar
B,f ≤  Gmin
. (20)
γ 0, ΓB,f neous SNR being below the minimum SNR, which implies
B
Pr(EB |SU = 0) = 1 − aB ; moreover, (23) follows by using
Summarizing the discussion so far, decoding of URLLC that Pr(EB |EU , SU > 0) ≤ 1 and that canceling URLLC
traffic cannot leverage SIC and treats eMBB transmissions interference results in the same performance achievable in
8 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

Pr (EB ) = Pr(SU = 0) Pr(EB |SU = 0) + Pr(SU > 0) Pr(EU |SU > 0) Pr(EB |EU , SU > 0)


+ Pr(ĒU |SU > 0) Pr(EB |ĒU , SU > 0) = (22)
S S


= (1 − aU ) (1 − aB ) + [1 − (1 − aU ) ] U Pr(EB |EU , SU > 0) + (1 − U ) Pr(EB |ĒU , SU > 0) (23)
S S
≤ (1 − aU ) (1 − aB ) + [1 − (1 − aU ) ] (U + (1 − U )(1 − aB )) . (24)

the absence of URLLC transmissions, so that we have the

equality Pr(EB |ĒU , SU > 0) = Pr(EB |SU = 0) = 1 − aB .
Imposing the reliability condition Pr(EB ) ≤ B and us-
3.0
ing (24) we obtain the inequality
1 − B 2.5
aB ≥ . (25)
1 − U (1 − (1 − aU )S ) 2.0

rU
As already pointed out, this equivalently imposes a constraint H-OMA
1.5
on Gmin
B,f through (4). H-OMA (LB)
From (25), we see that, unlike the orthogonal case, with 1.0 H-NOMA with SIC
H-NOMA with SIC (LB)
non-orthogonal slicing the eMBB activation probability aB
0.5 H-NOMA with puncturing
is larger than 1 − B . This is becasue URLLC interference H-NOMA with puncturing (LB)
may cause an eMBB decoding error even when the eMBB’s 0.0
0 2 4 6 8 10 12
SNR is above the threshold. However, the impact of URLLC rB,sum
interference is typically minimal. Indeed, the high reliability
FIGURE 5. Rate region (rB,sum , rU ) for the eMBB rate rB,sum and the
requirements for URLLC, which are reflected by the very URLLC rate rU when ΓU = 20 dB, ΓB = 10 dB,
small value of U , imply that aB is close to 1 − B . S = 5, aU = 0.1, F = 10, U = 10−5 , B = 10−3 . H-NOMA is present
with two variants, SIC and puncturing. The lower bound (LB) is derived in
To summarize, for a given feasible URLLC rate rU ∈ Appendix A.
orth
[0, rU (F )], the maximum eMBB rate is obtained by max-
imizing log2 (1 + Gtar tar
B,f ) subject on the constraints on GB
min
and GB implied by (20), (21), and (25). This maximization
1.2
requires the use of a two-dimensional numerical search. H-OMA
Note that the activation probability aB is typically very H-OMA (LB)
1.0
close to 1, and hence, when solving this problem, one can H-NOMA with SIC
H-NOMA with SIC (LB)
conservatively assume that the eMBB interference is always 0.8 H-NOMA with puncturing
present in (21), i.e., δf = 1. In contrast, the dependence H-NOMA with puncturing (LB)
of the right-hand side of (21) on Gtar
rU

B causes a non-trivial
0.6
interdependence between rU and rB .
0.4

2) Puncturing and erasure decoder

0.2
Turning now to the erasure decoder, we can write the proba-
bility of error for an eMBB user by means of the law of total 0.0
0 5 10 15 20 25 30 35 40
probability as rB,sum

Pr(EB ) = Pr (SU ≤ k) Pr(EB |SU ≤ k)+ FIGURE 6. Rate region (rB,sum , rU ) for the eMBB rate rB,sum and the
URLLC rate rU when ΓU = 10 dB, ΓB = 20 dB,
Pr (SU > k) Pr(EB |SU > k), (26) S = 5, aU = 0.1, F = 10, U = 10−5 , B = 10−3 . H-NOMA is present
with two variants, SIC and puncturing. The lower bound (LB) is derived in
where we have distinguished the case in which the erasure Appendix A.
code is able to correct the erasures caused by URLLC
transmissions, i.e., SU ≤ k, and the case in which an
error is instead declared, i.e., SU > k. When the latter Imposing equality in (27), we determine the parameter aB
event occurs, a decoding error occurs, and hence we have and, hence, Gmin
B,f via (4). Given the desired feasible URLLC
Pr(EB |SU > k) = 1. In contrast, when SU ≤ k, the
orth
rate rU ∈ [0, rU (F )], , we then obtain the target SNR Gtar
B,f
only source of outage is the instantaneous SNR being below and, hence, the eMBB rate (19) from the URLLC reliability
threshold, which results in Pr(EB |SU ≤ k) = 1 − aB . condition (21).
Overall, the resulting eMBB reliability requirement is
C. NUMERICAL ILLUSTRATION
Pr(EB ) = Pr(SU ≤ k)(1 − aB ) + Pr(SU > k) ≤ B . Here we present simulation results for the rate region
(27) (rB , rU ) for H-OMA as well as H-NOMA, with both SIC
VOLUME 4, 2016 9

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

and puncturing decoders. In addition to the results obtained in a time-sharing manner. Let α ∈ [0, 1] and 1 − α be
from the previous analysis, we also show curves obtained fraction of time in which the resources are allocated to the
from the expressions derived in Appendix A, which are easier eMBB device and the mMTC devices, respectively. We aim
to evaluate and are shown to provide a performance lower at characterizing the region of pairs (rB , λM ) of eMBB
bound (“LB"). rate rB and mMTC arrival rate λM that can be supported
In Figs. 5 and 6, we plot the rate regions for S = 5, aU = by orthogonal slicing for a given mMTC transmission rate
0.1, F = 10, U = 10−5 , B = 10−3 . Fig. 5 considers requirement rM and probability of error M .
the case ΓU > ΓB with ΓU = 20 dB and ΓB = 10 For a given time-sharing factor α, the achievable pair of
dB, while Fig. 6 focuses on the complementary set-up with eMBB rate rB and mMTC arrival rate λM can be written in
ΓB > ΓU when ΓU = 10 dB and ΓB = 20 dB. For terms of the quantities derived in Sections II-B and II-D as
both figures, H-NOMA with puncturing uses the optimal orth
rB = αrB (28)
puncturing parameter k.  
orth rM
For both set-ups considered in the figure, H-MONA with λM = λM , (29)
puncturing is outperformed by both H-OMA and H-NOMA 1−α
with SIC. Furthermore, when ΓU > ΓB , we see from respectively, where rB orth
is obtained as explained in Section
Fig. 5 that the SIC region dominates the region achievable orth
II-B and λM (·) is defined in (16). In fact, with orthogonal
by orthogonal slicing. This is thanks to the capability of the slicing, both the achievable eMBB rate rB and the achievable
BS to decode and cancel URLLC transmissions by leveraging mMTC transmission rate (specified on the right-hand-side of
reliability diversity. (14)) are scaled according to the fraction of time resources
In contrast, when ΓU < ΓB , Fig. 6 shows that orthogonal allocated to the service.
slicing can attain pairs (rB , rU ) that are not attainable by H-
NOMA with SIC. In particular, H-OMA is preferable if one B. NON-ORTHOGONAL SLICING: H-NOMA FOR EMBB
wishes to obtain large values of the URLLC rate. This is due AND MMTC
to the difficulty of ensuring high reliability in the presence In H-NOMA, the eMBB device is allowed to use the radio
of eMBB transmissions when ΓU < ΓB . We recall that this resource at the same time as the mMTC devices.
is a consequence of the impossibility to decode and cancel Decoding Architecture. As argued in Section II-D, a SIC
eMBB transmissions prior to URLLC decoding owing to the decoder may enhance the reliability of mMTC decoding.
URLLC latency constraint. In contrast, if one is interested Furthermore, when radio resources are allocated exclusively
in guaranteeing large eMBB sum-rates, H-NOMA offers sig- to mMTC devices, optimal decoding follows the order of
nificant performance gains. This is because non-orthogonal descending channel gains. The situation is more complicated
transmission allows eMBB users to operate over a larger in the presence of an interfering eMBB transmission.
number of spectral resources while not being significantly In light of the higher reliability requirements of eMBB
affected by URLLC interference. transmissions as compared to mMTC traffic, i.e., B  M ,
We see that the lower bound is able to capture the shape one may be tempted to consider decoding the eMBB traffic
of the region obtained through more accurate and time- before attempting to decode any mMTC traffic. This appears
consuming Monte-Carlo simulations. to be in line with the discussion in the previous section con-
cerning SIC for eMBB and URLLC coexistence. However,
IV. SLICING FOR EMBB AND MMTC this approach is suboptimal, since it neglects to account for
In this section, we treat the slicing of wireless resources to the different definition of reliability of mMTC traffic. In
jointly support eMBB and mMTC services, while assum- fact, the probability of error (15) measures the fraction of
ing that the URLLC traffic, if present, has been allocated incorrectly detected active users and not a per-device decod-
orthogonal resources. Analogously to the case of eMBB- ing probability. As such, some of the active mMTC devices
URLLC coexistence, we consider separately orthogonal slic- may well have very high channel gains, hence, causing large
ing (H-OMA) and non-orthogonal slicing (H-NOMA). We interference, making it beneficial to decode and cancel them
shall focus without loss of generality on the case F = 1, prior to decoding the eMBB signal. Selecting a SIC decoder
since the mMTC users are assumed to be active on a single that accounts for this important feature of mMTC traffic is
frequency channel. The extension to the case F > 1, in another example of a design choice that utilizes reliability
which the mMTC devices are allowed to randomly access all diversity.
F channels, is rather straightforward, as further elaborated Based on this discussion, we assume that, at each decoding
in Section V-B. Since a single channel is considered, in this step, the BS decodes either the eMBB device, provided that it
section we omit all frequency indices f . has not been decoded yet, or the next available mMTC device
in order of decreasing channel gains. Note that this implies
A. ORTHOGONAL SLICING: H-OMA FOR EMBB AND that the decoding step at which the eMBB device is decoded
MMTC is random, as it depends on the realization of the channel
For the case of orthogonal slicing, we assume that the eMBB gains. The process ends when no more transmissions can be
and the mMTC devices use the frequency radio resource reliably decoded.
10 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

As in non-orthogonal slicing of eMBB and URLLC, the C. NUMERICAL ILLUSTRATION

eMBB rate is set to We present numerical simulation results illustrating the trade-
offs between the eMBB rate rB and the mMTC arrival rate
rB = log2 (1 + Gtar
B) (30)
λM for orthogonal and non-orthogonal slicing. In addition
where Gtar
B is the target SNR for the eMBB transmission, to numerical results obtained by solving (34) through Monte
which is to be determined. Similar to the eMMB-URLLC Carlo methods, we also report results obtained upper and
coexistence case (see Section III-B), this quantity needs to lower bounds on λnon-orth
M (·) in (34), which are easier to
satisfy evaluate and are derived in Appendix B. Throughout this
ΓB section, we set M = 10−1 , and rM = 0.04.
Gtar
B ≤  Gmin
. (31) In Fig. 7, we plot the maximum mMTC arrival rate λM for
γ 0, ΓBB both orthogonal and non-orthogonal slicing as a function of
rB when ΓM = 5 dB, ΓB = 25 dB, and B = 10−3 . When
Again, as in Section III-B, we allow the eMBB device not
orthogonal slicing (H-OMA) is used, the supported mMTC
to use the maximal power since it may be beneficial to use a
arrival rate λM is seen to decrease in an approximately linear
value of Gtar
B lower than the right-hand side of (31) in order to
fashion with the eMBB rate. As for non-orthogonal slic-
control the impact of eMBB interference on the overall SIC
ing (H-NOMA), we observe three fundamentally different
procedure.
regimes as rB changes from zero towards is maximal value
Next, we formalize the SIC decoding procedure. When orth
rB .
the eMBB is inactive because of an insufficient SNR, i.e.,
The first regime consists of very small values of the eMMB
GB < Gmin B , the SIC decoding procedure is equivalent to
rate rB , for which the supported arrival rate λnon-orth
M is almost
the procedure described in Section II-D, namely, the mMTC
constant. At such values of rB , the eMBB device can be reli-
devices are decoded in the order of decreasing channel gains.
ably decoded before the mMTC devices. Therefore, interfer-
When the eMBB is active, the SIC procedure runs as follows.
ence from eMBB user can be cancelled, and the performance
Starting from m0 = 1, the receiver computes the SINR for
of mMTC traffic is unaffected by small increases in rB . The
the m0 -th mMTC device as
second regime spans intermediate values of rB . In this case,
G[m0 ] the eMBB signal can only be decoded after some of the
σ[m0 ] = PAM . (32)
tar strongest mMTC signals are decoded and canceled. Hence,
1 + GB + m=m 0 +1
G[m]
the mMTC performance is reduced by the interference from
If log2 (1 + σ[m0 ] ) ≥ rM , the m0 -th mMTC is decoded,
eMBB transmissions. Also, the SIC decoder tends to stop
canceled, m0 is incremented by one, and the procedure
the decoding process after detecting the eMBB user, and
starts over. Otherwise, the receiver attempts to decode the
decoding typically fails while detecting an mMTC device
eMBB user. To this end, it computes the SINR of the eMBB
because of the interference from the other, yet undecoded,
transmission as
mMTC devices. In the third regime, eMBB decoding fails
Gtar
B with a probability comparable to that of the weaker mMTC
σB = PAM (33)
1 + m=m0 G[m] devices due to the mutual interference between the two
services. As a result, in this regime, the supported mMTC
and decodes and cancels the eMBB if the condition log2 (1 +
arrival rate decays to zero as the eMBB rate rB increases.
Gtar
B ) ≥ rB is satisfied. If the eMBB is decoded successfully, The first and the third regime identified in Fig. 7 can also
the decoding procedure continues as in Section II-D. If be understood with the help of the lower and upper bounds
log2 (1 + Gtar
B ) < rB , the procedure terminates. derived in Appendix B. In particular, when rB is very low, as
Let DM ∈ {0, . . . , AM } and DB ∈ {0, 1} be the mentioned, it is almost always possible to decode the eMBB
random variables denoting the number of decoded mMTC transmission before decoding any mMTC device. This is the
and eMBB devices. With this notation, the probabilities of premise of the lower bound, which, as shown in Fig. 7, agrees
error for mMTC and eMBB users are given as Pr(EM ) = with the simulation results in the first regime. On the contrary,
1 − E[DM ] /λM and Pr(EB ) = 1 − E[DB ], respectively. the upper bound is computed by first identifying the subset of
In order to characterize the achievable pairs (rB , λM ), we mMTC devices whose channels are so weak that the additive
evaluate the maximum supported mMTC arrival rate λM as noise and the eMBB interference alone make their decoding
a function the eMBB rate rB as impossible. The upper bound is seen to agree the simulation
λnon-orth
M n (rB ) =
results in the third regime, i.e., when rB is large.
In Figs. 8 and 9, we plot the supported mMTC arrival rate
max λM ≥ 0 : ∃Gtar min
B and GB λM as a function of rB for different values of the eMBB SNR
ΓB and B = 10−3 , and for different eMBB reliability levels
o
s.t. E[DM ] /λM ≥ 1 − M and E[DB ] ≥ 1 − B .(34)
B and ΓB = 20 dB, respectively. This figures allows us to
We remark that, the probability distributions of DM and DB assess the impact of the average eMBB gain ΓB and of the
depend on the parameters λM , Gtar min
B , GB , rB , and rM . The eMBB reliability B on the operation of the system in the
computation of (34) requires Monte Carlo simulations. three regimes identified above and on relative performance
VOLUME 4, 2016 11

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

(1) (2)
rB rB

H-NOMA with SIC H-NOMA with SIC, ΓB = 10 dB

H-NOMA with SIC (UB) H-NOMA with SIC, ΓB = 20 dB
80 H-NOMA with SIC (LB) 80 H-NOMA with SIC, ΓB = 30 dB
H-OMA H-OMA, ΓB = 10 dB
H-OMA, ΓB = 20 dB
60 60 H-OMA, ΓB = 30 dB
λM

λM
40 40

20 20

0 0
0 1 2 3 4 5 0 1 2 3 4 5 6 7
rB rB

FIGURE 7. Arrival rates λorth

M and λM
non-orth
for mMTC traffic under H-OMA and FIGURE 8. Arrival rates λorth
M and λM
non-orth
for mMTC traffic under H-OMA and
H-NOMA, respectively, as a function of the eMBB rate rB . The upper bounds H-NOMA, respectively, as a function of the eMBB rate rB for
(UB) and lower bounds (LB) on the H-NOMA arrival rate λnon-orth
M (rB ) derived ΓB ∈ {10, 20, 30} dB. The parameters are ΓM = 5 dB,
in Appendix B are also shown. The parameters are ΓM = 5 dB, ΓB = 25 dB, M = 10−1 , B = 10−3 , and rM = 0.04.
−1 −3
M = 10 , B = 10 , and rM = 0.04.

of orthogonal and non-orthogonal slicing, as discussed next.

H-NOMA with SIC, B = 10−2
As it pertains to three regimes, we observe from Fig. 8 that H-NOMA with SIC, B = 10−3
the rate at which the transition from the second to the third 80 H-NOMA with SIC, B = 10−4
H-OMA, B = 10−2
regime occurs does not change as the eMBB average channel H-OMA, B = 10−3
gain ΓB is increased from 20 dB to 30 dB. On the contrary, 60 H-OMA, B = 10−4

the rB value corresponding to the transition from the first to

λM

the second regime becomes larger with this increase in ΓB . 40

This increase, in fact, allows the eMBB transmission to be

decoded earlier in the SIC process for a larger set of values of 20
rB . From Fig. 9, we observe that the eMBB error probability
constraint significantly affects the supported mMTC arrival 0
0 1 2 3 4
rate in the second and third regimes. In fact, in these case, rB
the higher eMBB transmission power required to ensure a
higher reliability impairs the decoding of mMTC users via FIGURE 9. Arrival rates λorth
M and λM
non-orth
for mMTC traffic under H-OMA and
H-NOMA, respectively, as a function of the eMBB rate rB for
interference. −2 −3 −4
B ∈ {10 , 10 , 10 }. The parameters are ΓM = 5 dB, ΓB = 20 dB,
We now elaborate on the comparison between orthogonal M = 10−1 , and rM = 0.04.

and non-orthogonal slicing. The presented figures emphasize

the fact that there are points in the rate region (rB ,λM ) that
can be attained by non-orthogonal slicing and not by or- URLLC. Albeit simple, the model accounts for the differ-
thogonal slicing, and vice versa. Specifically, non-orthogonal ences among the services in reliability, latency, and number
slicing is seen to be beneficial when rB is across the second of supported devices. Specifically, we have considered the
and the third regimes, especially for not too large reliability slicing of resources among the services in the uplink over a
levels B (see Fig. 8). For such values, the eMBB rate is large, shared multiple access resource. We have utilized the term
and yet low enough not to hamper the decoding of the mMTC “slicing” in order to emphasize the heterogeneous perfor-
users. Once again, reliability diversity is crucial to ensure the mance requirements that need to be satisfied for each service
effectiveness of non-orthogonal slicing. In contrast, for large as well as the performance isolation among services. Two
values of the rate rB , when non-orthogonal slicing is deeply slicing paradigms have been investigated, orthogonal and
in the third operating regime, orthogonal slicing is always non-orthogonal and the respective transmission schemes H-
superior. This is because in this regime the performance is OMA and H-NOMA, where the latter is inherently possible
limited by the interference caused by eMBB users. only in shared wireless channels.
We have applied the model to the study of the slicing for
V. DISCUSSION, CONCLUSIONS AND FUTURE WORK two services in two different cases: (i) eMBB and URLLC
A. DISCUSSION AND CONCLUSIONS and (ii) eMBB and mMTC. In both cases, we have shown
In this work, we have presented a communication-theoretic that, in order to be effective, the design of non-orthogonal
model that enables the investigation of the fundamental trade- slicing solutions must be guided by reliability diversity. For
offs associated with the sharing of the wireless resources the case of eMBB-URLLC coexistence, reliability diversity
among the three 5G traffic types, namely eMBB, mMTC and dictates that, in H-NOMA with SIC, the URLLC device
12 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

should be decoded first, as its decoding cannot depend relevant, as it is not feasible to reserve resources exclu-
on the decoding of eMBB, whose reliability and latency sively for replicas of packets generated by sporadically active
requirements are much looser compared to URLLC. The mMTC devices.
implications of reliability diversity are more subtle in the .
case of non-orthogonal slicing between eMBB and mMTC.
In this case, considering the fact that the number of active APPENDIX A LOWER BOUND ON THE URLLC RATE RU
mMTC devices is large with high probability, it is natural to By setting Pr(EU ) = U , we can upper bound (21) as
introduce a reliability metric that accounts for the fraction of  
FU  
correctly decoded transmissions. The analysis demonstrated X GU,f
U = Pr log2 1 + < FU rU  (35)
that there are regimes in which the decoding of eMBB should 1 + δf Gtar
B
f =1
be performed after the decoding of one or multiple mMTC  
devices in order to benefit from non-orthogonal slicing. FU  −t
Y G U,f
Our numerical results show that there are regimes in which = Pr 1+ ≥ 2−rU FU t  (36)
1 + δf Gtar
B
H-NOMA is advantageous over H-OMA and vice versa. In f =1
 −t 
the case of eMBB-URLLC, H-NOMA with SIC is always QFU 
G
beneficial when the eMBB rate is very large. In the case of E f =1 1 + 1+δfU,fGtar
B

eMBB-mMTC, non-orthogonal slicing is beneficial when the ≤ (37)

2−rU FU t
eMBB rate takes values that are small enough not to hamper  −t U
 F
G
the decoding of the mMTC devices. E 1 + 1+δ1U,1 Gtar
B
= . (38)
B. GENERALIZATIONS AND FUTURE WORK 2−rU FU t
The analysis presented in this paper is based on some sim- We obtained (35) by multiplying both terms in the inequality
plifying assumptions. However, the basic model and the by −t and then by exponentiating them; (36) follows from
methodology developed here can be extended to more gen- Markov inequality; and (37) holds because {GU,f } and {δf },
eral models and other operation regimes, as briefly discussed f ∈ {1, . . . , FU } are i.i.d., and hence we can set f = 1
here. in (38) without loss of generality. The inequality in (38) can
Starting with eMBB-URLLC coexistence, one could de- be rewritten as
" −t #
vise another H-NOMA scheme, where eMBB and URLLC 1 1 GU,1
users are allowed to access partially non-orthogonal re- rU ≥ log2 U − log2 E 1 +
tFU t 1 + δ1 Gtar
B
sources, so that only a subset of frequency channels poten- " −t #
tially occupied by URLLC traffic may be interfered by eMBB 1 1 GU,1
transmissions. Another direct generalization is to assume that > log2 U − log2 E 1 + (39)
tFU t 1 + Gtar
B
the minislots are pre-allocated to different URLLC devices.
Recall that, when all transmissions are made by the same where the strict lower bound follows by assuming that the
URLLC device, the block fading model dictates that either eMBB interference is always present, i.e., Pr(δ1 ) = 1.
all or none of the transmissions in a minislot are decoded The expectation in (39) can be calculated by a Monte Carlo
correctly. If each URLLC transmission is carried out by a simulation and the value of t ≥ 0 in (39) is chosen such as to
different device, then then error decoding events are inde- maximize the lower bound.
pendent across the minislots. As a more involved extensions
of the model, one may consider the impact of frequency APPENDIX B BOUNDS FOR NON-ORTHOGONAL
diversity also for eMBB traffic, and the performance under SLICING FOR EMBB AND MMTC
alternative decoding strategies, such as treating interference 1) A first upper bound
as noise. The idea behind the bound is as follows: if the eMBB
As for the coexistence of mMTC and eMBB services, an decoding fails, then the decoding of all mMTC devices for
rM
interesting extension is to allow multiple channels for mMTC which G[m] /(1 + GtarB) ≤ 2 − 1 must also fail. We obtain
traffic. In particular, mMTC devices may be allowed to use next a lower bound on the eMBB error probability, which will
frequency hopping, and the number of allocated frequency give us the desired upper bound on λM , by assuming that all
rM
channels may depend on the reliability requirements. An- mMTC devices that do not satisfy G[m] /(1+Gtar B) ≤ 2 −1
other aspect that deserves study is the impact of the arrival are decoded correctly and cancelled before eMBB decoding,
process, which here has been assumed to be Poisson. Namely, and that the remaining ones, which are not decoded correctly,
the higher burstiness of the arrival process can potentially cause interference to the eMBB. This yields the result de-
improve the gain that one can obtain with non-orthogonal scribed in (40), where 1{A} is the indicator function of the
slicing. Finally, following the approach in NB-IoT systems, event A.
the transmission of a single mMTC device may consist of The random variable χ in (40) is the sum of a Poisson-
replicas of the same packet in multiple time slots. This makes distributed number of truncated, exponential-distributed ran-
the non-orthogonal slicing of mMTC and eMBB even more dom variables, which allows for an efficient numerical eval-
VOLUME 4, 2016 13

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

" #
Gtar
Pr(EB ) ≥ 1 − aB + aB Pr o ≤ 2rB − 1
nB
m=1 G[m] 1
PAM G[m]
1+ ≤ 2rM − 1 1+Gtar
B
AM
" #
Gtar G[m]

G[m] 1
X
B rM
≥ 1 − aB + aB Pr rB −1≤ ≤2 −1 . (40)
2 −1 m=1
1 + Gtar
B
| {z }
=χ

low low orth

uation of the probability term in (40). Next, we set the right- for all rB ∈ [0, rB ]. When rB ∈ (rB , rB ], we obtain a
hand side of (40) equal to B , and find the values of Gmin
B and lower bound on λM non-orth
(rB ) by finding the largest value of
Gtar
B that result in the largest λ M . This value is precisely the λM for which qB (rB , λM ) ≤ M and by taking the smallest
desired upper bound on λnon-orth
M (r B ). between this value and λM,lb . We conclude by noting that
the upper bound on λnon-orth
M (rB ) on the right-hand-side of
2) A lower bound and an alternative upper bound (47), which holds for all rB , can be combined with the upper
We upper-bound the eMBB error probability by considering bound resulting from (40) to tighten it when rB is small.
the following suboptimal decoding scheme. We force the
decoder to always decode the eMBB first and subsequently REFERENCES
decode the mMTC devices. The maximal supported arrival [1] ITU-R, “ITU-R M.[IMT-2020.TECH PERF REQ] - Minimum Require-
rate with this modified decoder is clearly a lower bound on ments Related to Technical Performance for IMT- 2020 Radio Inter-
λnon-orth
M (r B ). While deriving this bound, we shall derive as face(s),” Report ITU-R M.2410-0, Nov. 2017.
non-orth [2] 3GPP, “Study on new radio (NR) access technology physical layer as-
by-product also an alternative upper bound on λM (rB ). pects,” TR 38.802, Mar. 2017.
As already mentioned, decoding the eMBB as the first de- [3] G. C. Ferrante, J. Östman, G. Durisi, and K. Kittichokechai, “Pilot-assisted
vice results in an upper bound on the eMBB error probability. short-packet transmission over multiantenna fading channels: A 5G case
study,” in Conf. Inf. Sci. Sys. (CISS), Princeton, NJ, Mar. 2018.
Mathematically, [4] H. Zhang, N. Liu, X. Chu, K. Long, A. H. Aghvami, and V. C. M. Leung,
" #
“Network slicing based 5G and future mobile networks: Mobility, resource
Gtar
B rB management, and challenges,” IEEE Commun. Mag., vol. 55, no. 8, pp.
Pr(EB ) ≤ 1 − aB + aB Pr PAM ≤ 2 − 1 (41)
1 + m=1 G[m] 138–145, Aug. 2017.
[5] Z. Wu, K. Lu, C. Jiang, and X. Shao, “Comprehensive study and compari-
AM
" #
Gtar X son on 5g noma schemes,” IEEE Access, vol. 6, pp. 18 511–18 519, 2018.
= 1 − aB + aB Pr rB B − 1 ≤ G[m] . (42) [6] L. Dai, B. Wang, Y. Yuan, S. Han, I. Chih-Lin, and Z. Wang, “Non-
2 −1 m=1 orthogonal multiple access for 5G: solutions, challenges, opportunities,
PAM and future research trends,” IEEE Communications Magazine, vol. 53,
Here, the random variable m=1 G[m] follows an Erlang dis- no. 9, pp. 74–81, Sep. 2015.
tribution. It will turn out convenient to denote by qB (rB , λM ) [7] T. M. Cover and J. A. Thomas, Elements of information theory. John
orth Wiley & Sons, 2012.
the right-hand side of (42). Let now EM (λM ) be the mMTC [8] W. Guan, X. Wen, L. Wang, Z. Lu, and Y. Shen, “A service-oriented de-
error probability as a function of mMTC arrival rate λM ployment policy of end-to-end network slicing based on complex network
in the absence of the eMBB device. By the law of total theory,” IEEE Access, vol. PP, no. 99, pp. 1–1, 2018.
[9] Z. Wu, F. Zhao, and X. Liu, “Signal space diversity aided dynamic
probability, the mMTC error probability when the eMBB is multiplexing for eMBB and URLLC traffics,” in IEEE Int. Conf. Comp.
present can be upper- and lower-bounded as Commun. (ICCC), Dec. 2017, pp. 1396–1400.
[10] A. Anand, G. de Veciana, and S. Shakkottai, “Joint scheduling of
orth E[DM ]
EM (λM ) ≤ Pr[EM ] = 1 − URLLC and eMBB traffic in 5G wireless networks,” Dec. 2017,
λM arXiv:1712.05344 [cs.IT].
orth [11] S. Y. Lien, S. C. Hung, D. J. Deng, and Y. J. Wang, “Efficient ultra-reliable
≤ EM (λM ) + B . (43)
and low latency communications and massive machine-type communica-
Set now tions in 5G new radio,” in IEEE Global Commun. Conf. (GLOBECOM),
Dec 2017, pp. 1–7.
orth

λM,lb = max λM ≥ 0 : EM (λM ) ≤ M − B (44) [12] N. H. Mahmood, M. Lauridsen, G. Berardinelli, D. Catania, and P. Mo-
gensen, “Radio resource management techniques for eMBB and mMTC
orth

λM,ub = max λM ≥ 0 : EM (λM ) ≤ M . (45) services in 5G dense small cell scenarios,” in Proc. IEEE Veh. Technol.
Conf. (VTC), Sep. 2016, pp. 1–5.
In words, these are the largest arrival rates for which the right- [13] L. Tian, C. Yan, W. Li, Z. Yuan, W. Cao, and Y. Yuan, “On uplink non-
hand side and the left-hand side of (43) are smaller than M , orthogonal multiple access for 5G: opportunities and challenges,” China
low Communications, vol. 14, no. 12, pp. 142–152, Dec. 2017.
respectively. Furthermore, let rB be given by
n o [14] G. Durisi, T. Koch, and P. Popovski, “Towards massive, ultra-reliable, and
low low-latency wireless communication with short packets,” Proc. IEEE, vol.
rB = max rB ≥ 0 : qB (rB , λM,lb ) ≤ B . (46) 104, no. 9, pp. 1711–1726, Sep. 2016.
low [15] G. Caire, G. Taricco, and E. Biglieri, “Optimum power control over fading
It follows that the pair (rB , λM,lb ) is achievable. Further- channels,” IEEE Trans. Inf. Theory, vol. 45, no. 5, pp. 1468–1489, May
more, we have that 1999.
[16] C. Bockelmann, N. Pratas, H. Nikopour, K. Au, T. Svensson, C. Ste-
λM,lb ≤ λnon-orth
M (rB ) ≤ λM,ub (47) fanovic, P. Popovski, and A. Dekorsy, “Massive machine-type communi-

14 VOLUME 4, 2016

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/ACCESS.2018.2872781, IEEE Access

cations in 5G: Physical and MAC-layer solutions,” IEEE Commun. Mag., GIUSEPPE DURISI (S’02–M’06–SM’12) re-
vol. 54, no. 9, pp. 59–65, Sep. 2016. ceived the Laurea degree summa cum laude and
the Doctor degree both from Politecnico di Torino,
Italy, in 2001 and 2006, respectively. From 2006
to 2010 he was a postdoctoral researcher at ETH
Zurich, Zurich, Switzerland. In 2010, he joined
Chalmers University of Technology, Gothenburg,
PETAR POPOVSKI (S’97–A’98–M’04–SM’10–
Sweden, where he is now professor and co-
F’16) is a Professor of Wireless Communications
director of Chalmers information and communi-
with Aalborg University. He received his Dipl.
cation technology Area of Advance. Dr. Durisi is
Ing and Magister Ing. degrees in communication
a senior member of the IEEE. He is the recipient of the 2013 IEEE ComSoc
engineering from the University of Sts. Cyril and
Best Young Researcher Award for the Europe, Middle East, and Africa
Methodius in Skopje and the Ph.D. degree from
Region, and is co-author of a paper that won a “student paper award" at
Aalborg University in 2005. He has over 300
the 2012 International Symposium on Information Theory, and of a paper
publications in journals, conference proceedings,
that won the 2013 IEEE Sweden VT-COM-IT joint chapter best student
and edited books. He holds over 30 patents and
conference paper award. In 2015, he joined the editorial board of the IEEE
patent applications. He is a Fellow of IEEE. He
Transactions on Communications as associate editor. From 2011 to 2014,
received an ERC Consolidator Grant (2015), the Danish Elite Researcher
he served as publications editor for the IEEE Transactions on Information
award (2016), IEEE Fred W. Ellersick prize (2016) and IEEE Stephen O.
Theory. His research interests are in the areas of communication theory,
Rice prize (2018). He is currently a Steering Committee Member of IEEE
information theory, and machine learning.
SmartGridComm and previously served as a Steering Committee Member
of the IEEE INTERNET OF THINGS JOURNAL. He is currently an Area
Editor of the IEEE TRANSACTIONS ON WIRELESS COMMUNICA-
TIONS. His research interests are in the area of wireless communication
and networking, and communication/information theory.

KASPER F. TRILLINGSGAARD (S’12) re-

ceived his B.Sc. degree in electrical engineering,
M.Sc. degree in wireless communications, and
Ph.D degree in wireless communications from
Aalborg University in 2011, 2013, and 2017,
respectively. He is currently a postdoctoral re-
searcher at the same institution. He was a visiting
student at New Jersey Institute of Technology in
2012 and at Chalmers University of Technology
in 2014. His research interests are in the areas of
information and communication theory.

OSVALDO SIMEONE (F’16) is a Professor

of Information Engineering with the Centre for
Telecommunications Research at the Department
of Informatics of King’s College London. He re-
ceived an M.Sc. degree (with honors) and a Ph.D.
degree in information engineering from Politec-
nico di Milano, Milan, Italy, in 2001 and 2005,
respectively. From 2006 to 2017, he was a faculty
member of the Electrical and Computer Engineer-
ing (ECE) Department at New Jersey Institute of
Technology (NJIT), where he was affiliated with the Center for Wireless
Information Processing (CWiP). His research interests include wireless
communications, information theory, optimization and machine learning. Dr
Simeone is a co-recipient of the 2017 JCN Best Paper Award, the 2015 IEEE
Communication Society Best Tutorial Paper Award and of the Best Paper
Awards of IEEE SPAWC 2007 and IEEE WRECOM 2007. He was awarded
a Consolidator grant by the European Research Council (ERC) in 2016. His
research has been supported by the U.S. NSF, the ERC, the Vienna Science
and Technology Fund, as well by a number of industrial collaborations.
He currently serves in the editorial board of the IEEE Signal Processing
Magazine, and he is a Distinguished Lecturer of the IEEE Information
Theory Society. Dr Simeone is a co-author of two monographs, an edited
book published by Cambridge University Press, and more than one hundred
research journal papers. He is a Fellow of the IET and of the IEEE and of
the IET.

VOLUME 4, 2016 15

2169-3536 (c) 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.