Content uploaded by Annapurna Pradhan
Author content
All content in this area was uploaded by Annapurna Pradhan on Oct 06, 2020
Content may be subject to copyright.
Joint Preference Metric for Efficient Resource
Allocation in Co-existence of eMBB and URLLC
Annapurna Pradhan
Department of Electrical Engineering
National Institute of Technology Rourkela
Rourkela, India
pradhan.annapurna37@gmail.com
Susmita Das
Department of Electrical Engineering
National Institute of Technology Rourkela
Rourkela, India
sdas@nitrkl.ac.in
Abstract—The co-existence of enhanced Mobile Broadband
(eMBB) and Ultra-Reliable Low Latency Communication
(URLLC) services in emerging 5G technology are essential to
support many diversified cutting edge applications with the
requirement of high data rate and low latency. In this multi-
service co-existence, optimizing resource allocation with efficient
scheduling is a challenging task. The hard latency bound URLLC
traffic selectively overlapped by puncturing the ongoing eMBB
transmission. This results in a potential loss to the data rate of
eMBB traffic. Therefore, we propose resource allocation based
on a joint metric of the minimum achievable rate of eMBB and
optimal URLLC placement demand for minimizing loss. The
resource allocation to each URLLC users in any mini-slot is
done based on the proposed metric.
Index Terms—eMBB, URLLC, coexistence, scheduling, punc-
turing, resource allocation.
I. INTRODUCTION
With a paradigm shift in technology, a number of emerging
applications and support for various services has imbibed the
LTE service and moved on towards the 5G mobile network.
Applications like augmented reality (AR), virtual reality (VR),
smart cities, unmanned aerial vehicles (UAVs), autonomous
cars, smart industries, etc. demands for high data rate (in terms
of gigabits), low latency (less than 1ms), reliability (99.999%),
energy efficiency, etc. To handle these diversified requirements
three service categories have been proposed by International
Telecommunication Union (ITU): massive machine type com-
munication (mMTC), enhanced Mobile Broadband (eMBB)
and Ultra-Reliable Low Latency Communication (URLLC)
[1]. Billions of interconnected devices in 5G will generate
huge amount of data traffic. Handling these data traffics
(i.e.,video,audio,text) with current wireless network model
and service classes is challenging. Therefore, to evade some
of these shortcomings co-existence mechanisms and resource
sharing methodologies can be adopted in order to provide an
uninterrupted data rate to users using services like eMBB and
URLLC.
In the co-existence scenario, the intermittent URLLC traffic
is per-emptively scheduled with high priority and instant re-
source allocation to satisfy its latency constraint. The resources
are allocated by superposition/puncturing the ongoing eMBB
allocation. This leads to a performance loss for eMBB users.
The total transmission time is divided into smaller time slots
in which eMBB traffic is scheduled. These time slots are
eMBB transmission time intervals (TTI). Each eMBB slot is
further sub divided into some mini-slots of smaller time period
called as URLLC transmission time intervals. URLLC traffic
is dynamically scheduled on these mini-slots as shown in Fig.
1.
Recently, research activities on coexistence of eMBB and
URLLC are gaining a lot of attention. The study in [2]
proposed the utilization of unlicensed spectrum resources for
serving punctured eMBB traffic. An optimal joint schedul-
ing for URLLC and eMBB traffic is proposed in [3],with
linear, convex/threshold rate loss models in case of superpo-
sition/puncturing of eMBB allocation due to URLLC traffic.
Scheduling of entire URLLC payload avoiding segmentation is
considered in [4] for developing packet scheduling algorithm
in mixed traffic type.The proposed algorithm uses throughput
to average metric for scheduling URLLC payload and pro-
portional fair metric for eMBB payload scheduling. The type
of eMBB traffic(low and high data rate requirement) is not
addressed during scheduling. Authors in [5], proposed a risk
sensitive approach for allocating resources for URLLC traffic
protecting low data rate eMBB users.A recent work in [6],
demonstrated a null space based spatial preemptive scheduling
for eMBB and URLLC traffic in dense multi user 5G network.
However for recovering data rate of punctured eMBB service,
we propose a joint preference metric based resource allocation
policy for efficient co-existence of eMBB and URLLC users
on sharable resources.
The rest of the paper is organized as follows. Section II
elaborates the system model and problem formulation for
coexistence of eMBB and URLLC traffic. We present the solu-
tion to the resource allocation problem in section III for multi
service coexistence. Performance evaluation of the proposed
solution is presented in section IV. Finally, conclusion and
future work are given in section V.
II. SY ST EM MO DE L AN D PROB LE M FO RM UL ATIO N
Deployment scenario consists of downlink transmission of
a Base Station (BS) with EeMBB users and UURLLC users.
The time domain is divided into slots with 1ms duration with
a further division into Mmini slots to accommodate the low
2020 12th International Conference on Communication Systems & Networks (COMSNETS)
978-1-7281-3187-0/20/$31.00 ©2020 IEEE 897
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on October 06,2020 at 06:19:59 UTC from IEEE Xplore. Restrictions apply.
Fig. 1. Superposition/ Puncturing mechanism for coexistence of eMBB and
URLLC
latency requirement of URLLC users. The BS uses uniform
system bandwidth Band SResource Blocks (RBs).
Base station allocates resource blocks at the beginning of
each slot to eMBB users. In the meantime if intermittent
URLLC traffic arrives, it is scheduled at the immediate next
mini slot of the ongoing eMBB transmission. This scheduling
of URLLC is done by puncturing the ongoing eMBB transmis-
sion because of low latency requirement. This leads to a loss
in data rate of eMBB users. Zero transmit power is allocated to
eMBB user at the time of puncturing resulting in a linear data
loss [5]. Initially, the maximum achievable rate of a particular
eMBB user ewithout puncturing at time slot t, as per [7] is
denoted as:
Rt
e=∑
s∈S
ϕe,s Belog2(1 + γe), for e ∈E(1)
Where ϕe,s is the resource allocation vector of eMBB user
ewith signal to noise ratio (SNR) γe. Let sis the number
of RBs allocated to eMBB users e.Beis the bandwidth part
associated to eMBB user e.ϕ∈[0,1], highest value associated
required RB allocation, otherwise zero. The achievable rate of
immediately scheduled URLLC in mini slot mwith resource
allocation vector θu,s ∈[0,1] is denoted by:
Rm,t
llc =∑
s∈S
θu,s Ru(2)
Where Ruis the data rate of a finite block length URLLC
transmission. Rllc is the total achievable rate for URLLC pay-
load.The scheduling of URLLC transmission should consider
the target outage probability εof total incoming URLLC traffic
Uand maximum latency constraint Tmax.
Puncturing causes performance degradation of eMBB data
transmission. Hence, the achievable rate of eMBB user in
punctured scheduling (PS) can be expressed in terms of a rate
loss function Πps as:
Rt
e=∑
e∈E∑
u∈U
Ke
SKu
S(1 −Πps)(3)
Where Ke
Sand Ku
Sare fraction of resources allocated to
eMBB user eand URLLC user ufrom assigned eMBB full
load RBs. The rate loss function Πps=Ke
S
Ku
S
,where Π∈[0,1]
shows the affected RBs of eMBB user edue to URLLC traffic.
In case of joint scheduling of eMBB and URLLC users
maximization of minimum achievable eMBB data rate while
scheduling all incoming URLLC traffic is a complex optimiza-
tion problem and can be formulated as:
Rmax,e= max
ϕ,θ (min
e∈E(E[∑
t∈T
Rt
e])) (4)
s.t. P (αm,t
u< U )≤ε , ∀u∈U(4a)
∑
u∈U
αm,t
uδ≤Tmax (4b)
Here 4(a) and 4(b) denotes the URLLC reliability and
latency constraint. αis an indicator variable for URLLC
service allocation and α∈[0,1]. It is 1 for serving user u
of URLLC traffic in mini-slot mor 0 otherwise. δis the time
requirement for one URLLC transport block transmission.
III. SOL UT IO N OF T HE PRO BL EM
We propose a scheduling method for eMBB and URLLC
traffic with pending HARQ retransmission following a
scheduling policy π. The retransmission of dropped URLLC
data packets are scheduled with highest priority to preserve
reliability of URLLC transmission and to avoid unnecessary
queuing delay. HARQ data will adaptively reuse the allotted
resources of initial data transmission. The URLLC scheduler
pre-emptively schedules [8] the URLLC traffic at the eMBB
mini-slots because of its low latency deadline. The sporadic
URLLC traffic is scheduled with high priority by puncturing
ongoing eMBB transmission. The eMBB users are scheduled
at the boundary of each time slot with a Proportional Fair
(PF) resource allocation to each user. In case of punctured
scheduling the data rate loss of corresponding eMBB user is
linear and directly proportional to the punctured resources.
Hence, maximization of minimum achievable eMBB data rate
can be preserved by optimally allocating resources to URLLC
users following the proposed scheduling policy π. This policy
is a function of two preference matrices P Meand P Mu. The
first Preference Metric P Meis proposed for an eMBB user e
at time slot t, depending on the expected achieved eMBB rate
till the previous time slot t−1and can be represented as:
P Me=E(t−1
∑
t=1
Rt
max,e), for all e ∈E, t ∈T(5)
Another Preference Metric P Muis proposed on the basis
of an optimal placement policy of URLLC traffic in mini-
slots of eMBB user. P Muis dependent on the placement
policies of URLLC traffic prior to current a mini slot m. Op-
timal placement decision can be taken depending on previous
URLLC placement demand Dm, where the random variable
D={D(1), D(2), ..D(m−1)}. The proposed method decides
the allocation of the resources to URLLC users depending
on the joint Preference Metric of expected achievable rate
2020 12th International Conference on Communication Systems & Networks (COMSNETS)
898
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on October 06,2020 at 06:19:59 UTC from IEEE Xplore. Restrictions apply.
and URLLC demand per slot basis.The scheduling policy π
choose a pair (P Me,P Mu) at the beginning of each slot t.
This policy specifies a history dependent resource allocation
to the URLLC traffic while maximizing eMBB data rate with
expected mean URLLC demand E[Dm] = ρper mini-slot.
IV. PERFORMANCE EVALUATION
The performance of the proposed solution is evaluated
using MATLAB based simulation parameters given in Table
I. A single cell scenario is considered using 5G gNB base
station and multiple UEs. The base station follows a Poission
downlink data flow model.Total number of RBs allotted to
each eMBB slot is 100. Each slot contains 8 mini slots.
Maximum 30 users are supported by the BS, out of which 10
active eMBB users and 20 URLLC users (maximum value) is
assumed. Active URLLC users per mini slot vary randomly.
P Muis based on binomial distribution of demand.
TABLE I
SIM ULATI ON PAR AM ETE RS
Parameter Value
Environment 3GPP Urban Macro (UMa) network
with 500 meter inter-site distance
Carrier Configuration 4GHz carrier frequency, 20MHz bandwidth
Sub carrier spacing 15KHz
Number of RB 100 resource blocks, 2 OFDM symbols
per TTI (0.143ms) 12 sub carriers per PRB
UE distribution Uniformly distributed outdoor, speed 3km/hr.
HARQ Asynchronous HARQ with chase combining,
4 TTI round trip delay, Maximum 6 HARQ retransmission
Fig. 2. Comparison of achievable rate of eMBB users
The performance of proposed scheduling and resource al-
location scheme is compared with a baseline of Random
Scheduler (RS). The RS selects resource blocks randomly
from allocates resources of eMBB users to serve URLLC
users without following any resource allocation and scheduling
policy. The minimum achievable rate of eMBB users for both
the methods are represented with the help of corresponding
empirical cumulative distribution function (ECDF). The sim-
ulation result after 1500 iterations is presented above in Fig.
2. In 60%cases the data rate of proposed method is more
than 15 Mbps where as in 30%cases random approach shows
same result.The result in Fig. 3, shows the reliability of eMBB
users for proposed method and baseline RS method. It can be
clearly observed that the reliability of eMBB decreases with
Fig. 3. Reliability of eMBB users Vs. URLLC load
the increase in URLLC load per cell, due to more puncturing.
At low URLLC load condition, the reliability of eMBB with
proposed method is 2.06%higher than RS method, whereas
17.3%higher for high URLLC load (U=20) condition. The
result of the proposed method shows an improved data rate
as well as reliability of eMBB users compared to the random
scheduling approach.
V. CONCLUSION
In this paper, we propose an efficient resource allocation
strategy based on a joint Preference metric for the co-existence
of eMBB and URLLC. The initial simulation result shows a
significant improvement of the proposed method over random
RB allocation for URLLC traffic. In the future, we aim to
use potential machine learning methods to serve co-existing
services over limited radio resources. Using learning based
approach to make decisions for flexible resource allocation
and scheduling will provide solutions for avoiding complex
coexistence problems in 5G and beyond wireless systems.
REFERENCES
[1] R. Abreu, T. Jacobsen, K. Pedersen, G. Berardinelli, and P. Mogensen,
“System level analysis of embb and grant-free urllc multiplexing in
uplink,” in 2019 IEEE 89th Vehicular Technology Conference (VTC2019-
Spring). IEEE, 2019, pp. 1–5.
[2] A. K. Bairagi, M. S. Munir, S. F. Abedin, and C. S. Hong, “Utilization
of unlicensed spectrum in effective coexistence of embb and urllc,” , pp.
1058–1060, 2018.
[3] A. Anand, G. De Veciana, and S. Shakkottai, “Joint scheduling of urllc
and embb traffic in 5g wireless networks,” in IEEE INFOCOM 2018-IEEE
Conference on Computer Communications. IEEE, 2018, pp. 1970–1978.
[4] A. Karimi, K. I. Pedersen, N. H. Mahmood, G. Pocovi, and P. Mogensen,
“Efficient low complexity packet scheduling algorithm for mixed urllc and
embb traffic in 5g,” in 2019 IEEE 89th Vehicular Technology Conference
(VTC2019-Spring). IEEE, 2019, pp. 1–6.
[5] M. Alsenwi, N. H. Tran, M. Bennis, A. K. Bairagi, and C. S. Hong,
“embb-urllc resource slicing: A risk-sensitive approach,” IEEE Commu-
nications Letters, vol. 23, no. 4, pp. 740–743, 2019.
[6] A. A. Esswie and K. I. Pedersen, “Opportunistic spatial preemptive
scheduling for urllc and embb coexistence in multi-user 5g networks,”
Ieee Access, vol. 6, pp. 38 451–38 463, 2018.
[7] A. K. Bairagi, M. S. Munir, S. F. Abedin, and C. S. Hong, “Coexistence
of embb and urllc in 5g wireless networks,” , pp. 1377–1379, 2018.
[8] N. B. Khalifa, M. Assaad, and M. Debbah, “Risk-sensitive reinforcement
learning for urllc traffic in wireless networks,” in 2019 IEEE Wireless
Communications and Networking Conference (WCNC). IEEE, 2019,
pp. 1–7.
2020 12th International Conference on Communication Systems & Networks (COMSNETS)
899
Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY ROURKELA. Downloaded on October 06,2020 at 06:19:59 UTC from IEEE Xplore. Restrictions apply.