Aalborg Universitet
Interference Coordination for 5G New Radio
Alvarez, Beatriz Soret; De Domenico, Antonio; Bazzi, Samer; Mahmood, Nurul Huda;
Pedersen, Klaus I.
Published in:
I E E E Wireless Communications Magazine
DOI (link to publication from Publisher):
10.1109/MWC.2017.1600441
Publication date:
2018
Document Version
Accepted author manuscript, peer reviewed version
Link to publication from Aalborg University
Citation for published version (APA):
Alvarez, B. S., De Domenico, A., Bazzi, S., Mahmood, N. H., & Pedersen, K. I. (2018). Interference Coordination
for 5G New Radio. I E E E Wireless Communications Magazine, 25(3), 131-137.
https://doi.org/10.1109/MWC.2017.1600441
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
- Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal -
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.
Downloaded from vbn.aau.dk on: August 10, 2022
1536-1284/17/$25.00 © 2017 IEEE
IEEE Wireless Communications • Accepted for Publication
2
AbstrAct
The arrival of the 5G NR provides a unique
opportunity for introducing new inter-cell interfer-
ence coordination (ICIC) mechanisms. The objec-
tive is twofold: to better exploit the benefits of
ICIC in coherence with the rest of radio resource
management (RRM) principles in 5G, and to
support new services and deployment scenarios.
We propose several enhanced techniques. In the
uplink, inter-cell coordination of the pilot sequence
configuration mitigates the inter-cell interference
problem of such pilots, which is especially severe
for cell-edge users. In the downlink, coordinated
small cell DTX aims at network interference control
and energy consumption reduction, whereas inter-
cell rank coordination can unleash the potential of
advanced receivers with minimal control overhead.
Besides, on-demand power boosting and coor-
dinated muting can be tailored to meet URLLC
requirements. The simulation results quantify the
performance benefits of the different techniques
under heterogeneous key performance indicators
(KPIs). We also discuss the standardization effort
required for having each of these techniques
included in the 5G NR specifications.
OverAll Inter-cell Interference cOOrdInAtIOn
desIgn PrIncIPles fOr 5g new rAdIO
Co-channel inter-cell interference is known to be
one of the limiting factors of cellular systems, and it
has triggered numerous academic research studies
and industrial standardization and implementation
efforts in LTE/LTE-A. Standardized network-based
inter-cell interference coordination (ICIC) schemes
for LTE range from basic coordination to further
enhanced ICIC, and more elaborate coordinated
multi-point (CoMP) communication techniques
[1–3]. The solutions for LTE were mainly designed
to offer spectral efficiency benefits for data channel
transmissions by applying various forms of inter-
cell coordinated muting (or power adjustments)
or interference randomization, while offering only
limited benefits for control channel performance.
Fifth generation new radio (5G NR) [4, 5] is
expected to experience a proliferation in the num-
ber of emerging use cases, categorized into three
broad service groups [6]. Enhanced mobile broad-
band (eMBB), an evolution of today’s broadband
traffic, will still be a key driver, with a main key
performance indicator (KPI) in the form of a tar-
get peak data rate of 20 Gb/s. Also, 5G opens the
door to new use cases with heterogeneous require-
ments, like ultra-reliable low-latency communica-
tions (URLLC), where messages must be correctly
decoded with very high probability (10
–5
outage
probability) and in a very short time (1 ms); and
massive machine type communications (mMTC),
catering to a large number (1 million devices/km2)
of low-data rate, low-cost services. The first phase
of the standardization process will primarily focus
on the first two categories, namely eMBB and
URLLC [4].
The set of radio features to support eMBB
and URLLC is broad, and it can be categorized
as follows: spectrum enhancements, with the use
of licensed, lightly licensed, and unlicensed bands
spanning microwave and millimeter wave frequen-
cies; deployment enhancements, for instance in the
form of ultra-dense networks with self-backhauling;
and capacity enhancements, like non-orthogonal
access, device-to-device and massive MIMO [7].
One important design principle in NR is to have
a flexible and efficient use of radio resources and
available spectrum [6]. As per the architecture, the
latency requirements also pose new challenges for
the backhaul, both in classical distributed cases and
in emerging centralized RAN (C-RAN) [4], where
a shared pool of centralized baseband resources
serves a large number of remote radio heads. In
this context, the ICIC framework evolution must go
hand in hand with the new radio access.
In this article, we present a set of interference
management advances for 5G NR. To fulfill the
promise of a comprehensive and integrated net-
work, 5G should move from a network-oriented
to a service-oriented paradigm, where differenti-
ated services with diverse KPIs can coexist on the
same infrastructure. Moreover, ICIC design prin-
ciples must exploit the new degrees of freedom
that come with 5G NR, especially taking advan-
tage of flexible physical-layer and medium access
(MAC) design [5], as well as the richer architecture
options [4]. In this light, we describe solutions to
address major interference challenges. First, we
consider the uplink (UL) inter-cell pilot (also known
as the reference symbol) interference problem.
Inter-cell pilot sequence coordination techniques
are proposed, which improve the link performance
because of enhanced channel estimation and
coherent demodulation [8]. Another advantage
of such techniques is their ability to support more
users than the current LTE solutions. In the down-
link (DL), we propose a scheme for joint interfer-
ence control and energy efficiency in dense small
cell scenarios [9] by means of enhanced methods
for discontinuous transmissions at the cell level
(cell discontinuous transmission DTX) based on
fuzzy Q-learning [10]. We elaborate on the new
Beatriz Soret, Antonio De Domenico, Samer Bazzi, Nurul H. Mahmood, and Klaus I. Pedersen
I C 5G N R
ACCEPTED FROM OPEN CALL
Beatriz Soret is with GomSpace; Antonio De Domenico is with the CEA-LETI MINATEC; Samer Bazzi is with Huawei Technologies
Duesseldorf GmbH; Nurul Huda Mahmood is with Aalborg University; Klaus I. Pedersen is with Nokia Bell Labs.
Digital Object Identifier:
10.1109/MWC.2017.1600441
This article has been accepted for inclusion in a future issue of this magazine. Content is final as presented, with the exception of pagination.
IEEE Wireless Communications • Accepted for Publication
3
opportunities that come from assuming multi-user
multiple-input-multiple-output (MU-MIMO) and
advanced interference-aware receivers as the base-
line for 5G [11]. Building on the earlier work in
[12], novel solutions for coordination of the maxi-
mum transmission rank between neighboring cells
is also presented. Another proposal is to support
the challenging reliability and delay requirements
of URLLC through highly agile and fast coordina-
tion techniques, offering benefits for both control
and the data channel performance [13].
It is worth highlighting that the proposed
schemes are complementary, addressing differ-
ent interference challenges but sharing the 5G
NR philosophy of more dynamic coordination
for a multi-service air interface. The details of the
proposed mechanisms are presented in the next
sections. The delay over the backhaul in the sig-
naling exchange between base station (BS) nodes
(through the so called Xn interface [4]) is a limit-
ing factor in inter-cell coordination. In all cases, we
strive for a generic design that is applicable both in
distributed architectures with Xn interface as well
as in C-RAN scenarios with a centralized control-
ler. For the sake of conciseness, each downlink
solution is tailored for a given service, although all
UL and DL proposals are applicable to both eMBB
and URLLC.
UPlInk Inter-cell PIlOt cOOrdInAtIOn
In UL, inter-cell pilot interference arises when the
assigned uplink pilot sequences across multiple
cells, which are non-orthogonal, are scheduled on
the same time-frequency resources. The received
pilots from a target user suffer from pilot interfer-
ence coming from neighboring cells, resulting in
poor channel estimation. This problem is especial-
ly severe for cell-edge users, as the power of the
interfering pilots is comparable to that of desired
pilots. It leads to errors in uplink coherent demod-
ulation, and it is very detrimental in uplink multi-us-
er MIMO scenarios that heavily rely on accurate
channel knowledge to perform receive filtering.
Additionally, in a calibrated time-division-duplex
(TDD) system where channel reciprocity holds, the
BS can acquire the channel knowledge necessary
for downlink multi-user MIMO precoding via the
uplink pilots sent by the users. In this case, pilot
interference leads to erroneous channel knowl-
edge, which affects the precoding quality and the
downlink throughput.
In LTE-A, users across cells are assigned non-or-
thogonal yet distinguishable sequences. These
sequences are cyclic extended Zadoff-Chu (ZC)
sequences, which are spread over the subcarriers
of interest. Cyclic-extension is necessary to maxi-
mize the number of distinguishable sequences. The
available sequences in each cell are constructed by
phase rotation of a root sequence identified by a
root index, and are mutually orthogonal. The root
sequences (and the corresponding root indices)
across cells are different. Different root sequenc-
es or phase rotations thereof are not orthogonal,
though they are distinguishable via their root indi-
ces.
Few solutions exist to mitigate inter-cell pilot
interference via a distributed or centralized
sequence assignment over the cells. A related work
is [14], where the authors propose an assignment
of ZC sequences in an Orthogonal Frequency Divi-
sion Multiplexed (OFDM) system, such that the
worst-case channel estimation mean square error
(MSE) is minimized. However, a key assumption
of [14] is that user pilots occupy all available sub-
carriers, which is not the case in a practical system,
rendering the performed analysis inapplicable. Fur-
thermore, the BSs treat pilot interference as noise,
which is suboptimal at high uplink signal-to-noise
ratios (SNRs) occurring in, for example, small cell
scenarios. In such scenarios, a better approach
would be the suppression of pilot interference at
the BS to recover the desired pilots with as little
interference as possible.
LTE-A can allow for pilot orthogonality among
multiple cells: a BS assigns, from its pool of avail-
able orthogonal sequences, pilot sequences for
users in neighboring cells. Such a solution is not
scalable for many 5G applications, as the number
of users a BS can serve within its cell decreases.
One possibility to suppress the pilot interfer-
ence in 5G NR and leave the number of served
users within a cell unchanged can be realized by
exchanging ZC root indices among BSs through
the backhaul Xn interface. An alternative imple-
mentation is a centralized approach with a central
controller sending the indices of all concerned BSs
to each BS. Both implementations allow a given
BS to construct the sequences used in neighboring
cells and perform channel estimation, including not
only the channel of the desired user but also that
of users in neighboring cells [8]. The channel of the
former is then estimated with some residual inter-
ference (due to the non-orthogonality of sequenc-
es across the cells), while the estimated channels
of the latter can be dropped or used according
to the desired application (e.g., CoMP beamform-
ing or joint transmission rely on the knowledge of
channels of users in neighboring cells). The chan-
nel estimation is performed in the time domain and
exploits the fact that, in practical OFDM systems,
the number of taps is (much) smaller than the
number of subcarriers, which results in a reduced
number of variables in the time domain (i.e., taps)
that can be efficiently estimated. Going one step
further, [8] proposes to optimize the choice of the
used sequences such that the channel estimation
MSE is further reduced. The gains of optimized
sequence selection are mainly seen in the medium
to high SNR regime where the non-orthogonality
of used sequences becomes the limiting factor. Fig-
ure 1 shows the signaling steps necessary both for
a centralized and a decentralized implementation.
The first step consists of the signaling/exchange of
sequence indices, while the second one involves
informing the users within each cell of the chosen
sequence within the respective cell.
Summing up, this procedure generalizes the
idea of uplink CoMP data reception to pilot
sequence reception. It can be implemented for
ZC as well as other types of sequences (e.g.,
pseudo-noise sequences). It improves the channel
estimation quality for non-CoMP applications and
allows efficient CoMP operation without reducing
the number of users that can be simultaneously
assigned pilot sequences. As observed in [8], a
careful choice of sequences can allow the achiev-
able MSE to closely follow the interference-free
MSE. In contrast to LTE solutions, keeping the num-
ber of users that can be served unchanged is espe-
cially important for eMBB and URLLC 5G services.
LTE-A can allow for
pilot orthogonality
among multiple cells:
a BS assigns, from
its pool of available
orthogonal sequences,
pilot sequences for
users in neighboring
cells. Such a solution is
not scalable for many
5G applications, as the
number of users a BS
can serve within its
cell decreases.
This article has been accepted for inclusion in a future issue of this magazine. Content is final as presented, with the exception of pagination.
IEEE Wireless Communications • Accepted for Publication
4
dOwnlInk Inter-cell
Interference cOOrdInAtIOn
In the DL, the trend is toward more dynamic ICIC
solutions, as already agreed in 3GPP for 5G NR
[5], as well as addressing various network deploy-
ments (small cell and macro scenarios), key 5G
technologies (dense small cell networks and
MU-MIMO), and KPI requirements (spectral effi-
ciency, energy, and reliability).
dOwnlInk Inter-cell cOOrdInAted smAll cell dtX
Cell discontinuous transmission is an energy sav-
ing technology that adapts the cell activity to its
instantaneous load. Within each frame, the cell
DTX will instantaneously activate (deactivate) the
cell components and the associated functional-
ities when the user data is present (absent) in the
cell queue. Furthermore, it is possible to increase
the period in which a cell switches off or mutes by
maximizing the usage of the available frequency
resources at each active TTI, that is, trading off
latency for energy efficiency.
In dense small cell deployments, this approach
comes with the challenge of orchestrating the net-
work activity in order to limit simultaneous activa-
tion of nearby cells. First, the optimal selection of
the subset of small cells to activate at each frame
is a combinatorial problem, which is complex to
solve. Second, dormant cells cannot exchange
information and implement baseline ICIC solu-
tions or CoMP schemes. Finally, a reliable solution
needs to take into account the stochastic nature
of both the traffic and the radio channel. Existing
ICIC mechanisms are not designed to deal with
multi-objective optimization problems, e.g., jointly
reducing interference and energy consumption
while satisfying traffic latency constraints.
Reinforcement learning solutions provide an
efficient framework to learn an optimal activation
strategy by interacting with stochastic environ-
ments [9]. We design a fuzzy Q-learning based
cell DTX controller that uses its decisions in the
previous time slot to estimate the interference level
experienced by the active small cells. In addition,
the controller observes the queued data pending
for transmission per cell, the expected capacity,
and the requirements of the active services to
decide whether to activate a small cell.
The sketch of the architecture and the detailed
signaling exchange required by the scheme are
shown in Figs. 2a and 2b, respectively. Notice that
the pictured solutions here utilize the enhanced
support for different architectures and functional
splits that comes with 5G NR [4]. In those exam-
ples, the aggregation node buffers the data relat-
ed to nearby small cells while the orchestration
functions are deployed at the network controller.
Additionally, the measurements related to the
radio access network capacity can be forwarded
to the controller node during the small cell acti-
vation. Radio resource management (RRM) and
lower-layer functions are implemented locally at
the transmission points; thus, the controller and the
small cells do not need to continuously exchange
messages through the backhaul. On the contrary, a
fully distributed architecture requires coordination
across nearby small cells, which in turn increases
the small cell (and the backhaul) energy consump-
tion. In the same way, implementing centralized
scheduling or coordinated beamforming schemes
at the network controller 1) increases network
complexity; 2) needs regular transmission of channel
quality indicators (CQIs) over the backhaul link; and
3) is affected by the backhaul latency and capacity
constraints. In any case, the proposed solution with
reinforcement learning manages the small cell activ-
ity to limit network energy consumption without
reducing the system quality of service (QoS).
dOwnlInk Inter-cell rAnk AdAPtAtIOn
In a MIMO setting, the downlink serving rank (or
number of transmission streams) plays a major role
in the interference suppression levels of interfer-
ence rejection combining (IRC) receivers. This is
because significant interference suppression is only
possible when the number of desired data streams
and dominant interference streams are collectively
fewer than the receiver dimension, i.e., the number
of receive antennas. Traditionally, rank selection
at each user is essentially performed in a selfish
manner independently per link, without taking into
account the interference caused by such selec-
tions. For 5G NR, an inter-cell rank coordination
mechanism can improve the network and the cell-
edge user throughput by coordinating the gener-
ated inter-cell interference from the aggressor cell.
Consider a MU-MIMO TDD system. The
FIGURE 1. Signaling steps in a centralized or decentralized implementation of pilot sequence allocation.
i(s
1
),
i(s
2
)
Cell 1 uses
sequence s
1
Network controller selects
sequences s
1
and s
2
Cell 2 uses
sequence s
2
i(s
1
),
i(s
2
)
i(s
l
): index of sequence s
l
, l=1,2
Signaling step 1
i(s
1
)
i(s
2
)
Signaling step 2
Signaling step 1
Signaling step 2
i(s
1
)
Cell 1 selects
sequence s
1
Cell 2 selects
sequence s
2
i(s
2
)
i(s
l
): index of sequence s
l
, l=1,2
i(s
1
)
i(s
2
)
This article has been accepted for inclusion in a future issue of this magazine. Content is final as presented, with the exception of pagination.
IEEE Wireless Communications • Accepted for Publication
5
available resources are divided into time-frequen-
cy slots, with the smallest unit being a physical
resource block (PRB), corresponding to the dura-
tion of a single time transmission interval (TTI)
over a single frequency channel. The transmission
toward a desired UE from its serving BS generates
interference toward out-of-cell interfered receivers.
A cell-edge user scheduled on a given set of PRBs
in a neighboring cell is most likely to be affected
by the transmission on the same PRBs, and hence
requires interference coordination. Studies have
shown that coordinating the transmission rank can
help improve the performance of interference sup-
pressing receivers, such as the IRC [11].
The proposed inter-cell rank coordination aims
at limiting the maximum rank of an aggressor,
thus providing a guarantee on the experienced
interference. To further limit the complexity of
the scheme, only the strongest interferer, known
as the dominant interferer (DI), of the victim UE
is considered in the coordination. The victim
UE reports the DI physical cell id together with
a measure of the dominant to interference ratio
(DIR), defined as the ratio between the DI power
to the rest of interference and noise power in the
network. As happens with the interference, the
DIR can change very fast in fractional load sce-
narios [15], and therefore the LTE-A measures of
received signal power are not sufficient.
The proposed coordination mechanism involves
the following steps, as illustrated in Fig. 3:
1) The UEs report the DIR and the CQI to the
serving BS. The serving BS determines whether the
DIR is above a certain pre-specified threshold, and
rank coordination is only invoked for those UEs with
a strong DI. The UEs selected for rank coordination
are then grouped according to the DI, to avoid con-
flicting coordination requests from the same BS.
2) The serving BS decides what will be the
maximum transmission rank for each of the UEs
in each group, along with the interference rank it
would like to have. The signal to interference and
noise (SINR) ratio is used for the decision. The
ranks are chosen based on the estimated post IRC
SINR.
1
The proposed rank coordination mecha-
nism is not bound to any specific rank adaptation
algorithm, though interference-aware rank adap-
tation algorithms such as those presented in [12]
are best suited for such applications.
3) The serving BS sends the desired rank mes-
sage to the respective interfering BS. The desired
rank message is indicated as the maximum allow-
able transmission rank for a given set of PRBs.
These messages can be per single PRB, or several
PRBs can be grouped into a single desired rank
and priority level. The granularity provides a trade-
off among performance, overhead and complexity.
4) The serving BS updates its transmission
parameters according to the feedback mes-
sage from the interfering BS. Such updates can
include re-scheduling the users, re-adjusting
the transmission parameters, or re-adapting the
transmission rank with respect to the feedback
message. The 5G TTI is expected to be shorter
than the current 1 ms of LTE [5]. The rank coor-
dination could occur over a longer time basis
(in the range of 5-10 ms), therefore suitable for
heavy payload traffic spanning over multiple
TTIs. For random intermittent traffic with small
payload, the interference rank can be pre-co-
ordinated to cater to such bursty but critical
payloads.
The algorithm in Fig. 3 is applicable to both dis-
tributed and C-RAN architectures, taking advan-
tage of the flexible architecture options that come
with 5G NR [4]. Naturally, when having a central-
ized unit, the scheme simplifies since there is no
FIGURE 2. Downlink inter-cell coordinated small cell DTX: a) sketch of the architecture; b) signaling exchange.
Aggregation
node
User plane buffers
Network
controller
1
2
3
Power saving
mode
Backhaul
Low buffer
(a)
(b)
High buffer
Interference level
1
2
3
Low
interference
High
interference
SC set
Aggregation
node
Backhaul
node
Network
controller
1. RAN information
(Capacity estimate,
dominant interferer)
2. Buffer information
(cell load, packet TTL)
4. Fuzzy controller
5. BH node mode control
8. Data Forwarding
6. SC mode control
B) if SC switched on
A) if SC switched off
7. Buffering
3. Interference estimate
FIGURE 3. Flowchart of Message Flow with multiple UEs.
Dominant interferer (DIR, AP
ID) info + CQI measurements
DIR >
DIR ?
t
Group UEs
according to
dominant
interfering AP
No rank
coordination
Joint rank
coordination
for each
group
Coordination
message
Accept?
Reason, and additional
response (e.g., alternative rank)
UE domain
Serving eNB
domain
DL transmission
Update parameters
(if required)
Interfering
eNB domain
NACK
+ info
ACK
1
The post IRC SINR is the
criterion for eMBB services.
For URLLC, the probability of
satisfying a target SINR can
be used instead.
This article has been accepted for inclusion in a future issue of this magazine. Content is final as presented, with the exception of pagination.