RAN Moderation in 5G Dynamic Radio Topology
Ömer Bulakci, Alexandros Kaloxylos, Josef Eichinger, Chan Zhou
Huawei Technologies GRC, Munich, Germany
{oemer.bulakci, alexandros.kaloxylos, josef.eichinger, chan.zhou}@huawei.com
Abstract—The standardization for the fifth generation (5G) of
mobile and wireless networks is at its early phase and has
recently completed the first study item in Release 14.
Nevertheless, there is a consensus that 5G will address the diverse
service requirements of high-variety use cases. The network shall
cope with such variation effectively and cost-efficiently even
though the requirements can change over space and time. The
design of the radio topology for the peak service demand is, thus,
not desirable for network operators. As a consequence, the trend
is towards more flexible network deployment. In this context,
dynamic radio topology through vehicular nomadic nodes
(VNNs) is an emerging concept towards 5G to efficiently address
non-uniformly distributed traffic. VNNs are aimed to overcome
the lack of flexibility induced by small cells that are deployed at
fixed locations via network planning in current wireless
networks. A VNN is a low-power access node with wireless self-
backhaul, which can be activated temporarily to provide
additional system capacity and/or coverage on demand. VNNs
can be integrated into vehicles, e.g., in car-sharing fleets. In this
paper, we evaluate the performance of radio access network
(RAN) moderation of VNNs in a multi-cell environment
considering composite fading/shadowing environments with co-
channel interference, where active VNNs are selected from a set
of available candidate VNNs based on the signal-to-interference-
plus-noise ratio (SINR) on the wireless backhaul link. The results
show that RAN moderation can significantly improve the end-to-
end rate and SINR performances along with clear amount-of-
fading (AoF) reduction.
Keywords—5G; Dynamic Radio Topology; HetNet; METIS-II;
Moving Networks; RAN Moderation; Vehicular Nomadic Node
I. INTRODUCTION
In current mobile and wireless networks, one approach for
addressing increased coverage and/or capacity demands is to
deploy fixed small cells. Small cells are typically deployed by
network operators at certain locations, where the locations can
be determined by network planning. However, the full
operation of such dense fixed small cell deployment is not
always needed due to the inhomogeneous distribution of traffic
over time and space. Hence, fixed network deployment has the
disadvantage of increased capital expenditure (CAPEX), e.g.,
due to deployment of additional wireless access nodes, and
operational expenditure (OPEX), e.g., due to the incurred site
leasing costs, although small cells need to be operated only
partially. In addition, the need for proper site locations, e.g.,
due to power supply facility, can further limit the achievable
network topology.
Towards the Fifth Generation (5G) system, the concept of
dynamic radio topology has emerged [1]-[9]. Within the
framework of dynamic radio topology, the network shall react
quickly and dynamically to fulfill the increased service
requirements in a certain time period and at a target service
region. On this basis, one component to enable dynamic radio
topology is vehicular nomadic node (VNN) operation, which
provides a complementary approach to fixed small cells. A
VNN is a movable access node with wireless backhaul link,
which can provide coverage extension and/or capacity
improvement on demand. VNNs can be integrated into vehicles
as shown in Fig. 1, e.g., within a car-sharing fleet. VNNs are
assumed to be stationary during their operation; however, their
availability changes with respect to time and space according to
their battery state or driver needs, and, hence, the term
“nomadic” is applied. Furthermore, as VNNs are integrated
into vehicles, due to low height of 1.5 m like the one of user
equipments (UEs), severe fading characteristics can be
expected on the wireless backhaul link as opposed to well-
elevated small cells (e.g., 5-10 m for fixed relay nodes).
Accordingly, to ensure the expected benefits of VNN
operation, active VNNs shall be properly selected such that the
wireless backhaul link quality is optimized.
In this work, RAN moderation of VNNs is presented in a
cellular wireless network to determine the active VNNs, where
the radio channels are modeled by composite
fading/shadowing. Performance of the proposed RAN
moderation strategies are shown in terms of the resultant
signal-to-interference-plus-noise ratio (SINR) on the wireless
backhaul link, end-to-end rates, and severity of fading via
amount-of-fading (AoF) metric.
The remainder of the paper is organized as follows.
Section II sets the scene for RAN moderation strategies in
dynamic radio topology along with the channel and system
models. In Section III, performance results are provided.
Finally, Section IV concludes the paper.
II. RAN MODERATION SCHEMES
A. System and Channel Models
The considered system model is depicted in Fig. 1, where
the network is represented by a regular hexagonal layout with
seven macrocells served by base stations (BSs) deployed in the
centers of hexagons. In the exemplary illustration, there are
three candidate VNNs available in the target service region,
i.e., three vehicles are parked in the region of interest and can
be activated to serve the UEs in the proximity of the active
VNN on the access link (VNN-UE link). It is worth noting that
the target service regions can be determined by the operators,
e.g., based on UE density due to an event. In order to take into
account the uncertainty for the availability of VNNs, we utilize
the parking lot model given in [1], [4], [10], which is based on
continuous-time Markov chains. We consider a parking lot
Fig. 1. The considered system model with UEs, VNNs and macro BSs. The
active VNN to serve the target service region and the associated serving BS
are selected based on the backhaul link SINR.
Fig. 2. SINR gains with different RAN moderation strategies considering M
max
= 5 (left) and M
max
=15 (right).
where a maximum of
}15,5{
max
M
places are available on a
line road. The distance between two nearby VNNs is taken as
6 m. Moreover, we set the parking lot model parameters, i.e.,
departure and arrival rates, such that a regular day time (09:00-
17:00) is simulated [1]. The flexible wireless backhaul (BS-
VNN link) is realized by in-band half-duplex relaying in this
work, while different relaying options can also be considered
for the VNN operation, such as, full-duplex and out-band
relaying.
The channel models pertain to a two-hop decode-and-
forward relaying operation through VNNs, where end-to-end
performance is degraded also by interference on the backhaul
link. The direct link (BS-UE) and backhaul link are modelled
by Rayleigh-lognormal (a.k.a. Suzuki) composite distribution,
and the access link is modeled by Rician-lognormal composite
distribution, which are the two common models in the literature
[11]. Interfering signals on the backhaul link are assumed to be
subject to Rayleigh-lognormal composite fading/shadowing. It
is worth noting that severe fading characteristics are also
assumed on the backhaul link due to low height of VNNs like
the UEs. Further, a single UE is connected to a single VNN on
the access link and is communicating via this VNN with a BS.
The shadowing standard deviation is set to 8 dB on direct and
backhaul links, unless otherwise stated. The rest of the system
parameters are in line with [4]. Besides, the simulations are
conducted using MATLAB R2011a as the computational
environment.
B. Problem Formulation and Selection Schemes
RAN moderation strategies take into account the backhaul
link qualities at different candidate VNNs towards the available
K BSs. At a given time instant, there are M available VNN
candidates in cell k out of which we select the VNN m* (VNN
selection) and associate it with the BS k* (serving BS
selection) such that downlink SINR γ on the backhaul link is
maximized as
max
,
,
opt
subject to
},...,2,1&,...,2,1:{max
*km*,
MM
KkMm
km
km
(1)
Accordingly, the serving BS may not necessarily be the closest
BS to the candidate VNNs. For instance, the closest BS may be
shadowed due to a large obstacle and, thus, a neighbor BS may
provide the best backhaul link conditions. Two VNN selection
schemes are considered within the RAN moderation
framework. Namely, in case of optimal VNN selection, both
shadowing and multi-path fading are considered in SINR
measurement in (1). On the other hand, in case of coarse VNN
selection only shadowing is factored in the SINR measurement
in (1). That is, optimal VNN selection takes into account short-
term changes in radio conditions, whereas coarse VNN
selection is focusing on the long-term radio conditions.
Consequently, the optimal selection requires more frequent
channel quality indications to be sent.
In [1] and [4], the analyses on were limited to VNN
selection schemes, where the selected VNN was assumed to be
served by the midmost BS only (k*=1). Herein, the analyses
are conducted such that VNN and serving BS selections are
jointly optimized, and the impact of serving BS selection is
highlighted. On this basis, in the following, single BS refers to
the case where the selected VNN is served by the midmost BS
only, and the multi BS refers to the case where the serving cell
of the VNN can be any of the available K=7 BSs based on the
selection criterion given in (1).
III. PERFORMANCE EVALUATION
In this section, the performance assessment of different
selection schemes is performed in terms of downlink SINR,
end-to-end rates, and AoF. Random VNN selection is taken as
reference, where no particular channel knowledge is utilized.
Fig. 3. End-to-end rate gains with different RAN moderation strategies considering M
max
= 5 (left) and M
max
=15 (right). The Hull Curve indicates the upper
bound of achievable gains when the cell selection for the UE (BS or VNN) is optimized.
A. SINR Distributions
The cumulative distribution function (CDF) plots of SINR
values on the backhaul link and on the direct link are illustrated
in Fig. 2 for M
max
=5 and 15. It is first observed that the
performances of random VNN selection and direct link are
comparable. This is due to the assumption of the severe fading
characteristics on the backhaul link similar to the direct link.
This outcome motivates the need for proper VNN selection
schemes to attain the promised benefits of the dynamic radio
topology. In this regard, it can be seen that VNN selection
schemes can clearly improve the performance especially at the
low SINR regime. For instance, in case of M
max
= 5, single-BS
coarse selection can increase the SINR at 5%-ile CDF level by
9 dB relative to the direct link, while the joint optimization
considering the serving BS selection (i.e., multi BS) can
provide an additional 4-dB SINR gain. Optimal VNN selection
can further improve the SINR performance at the cost of
increased measurement overhead. In particular, the optimal
VNN selection results mark the upper bound of achievable
SINR gains in a given scenario.
Additionally, when there are more available VNNs in the
parking lot (see, M
max
= 15), the achievable SINR gains can be
clearly improved, e.g., the SINR gain at the 5%-ile SINR CDF
is then increased to 14 dB in case of single-BS coarse VNN
selection. Nevertheless, the extra gains obtained via multi-BS
optimization are decreased when the available number of
VNNs increases, as the probability of having the best backhaul
link toward the midmost BS increases.
B. End-to-end Rate Distributions
The CDF plots of end-to-end rates are illustrated by Fig. 3
for M
max
=5 and 15. Herein, equal-time resource operating point
is assumed between backhaul and access links, e.g., in a long-
term evolution (LTE) system, this would correspond to the case
where five subframes of a ten-subframe radio frame are
allocated to each of the backhaul and access links. The results
further highlight that VNN selection is vital because without
VNN selection (see random selection), the VNN performance
becomes worse than that of the direct link due to half-duplex
constraint, i.e., the total time resources are shared between
backhaul and access links.
The previously observed SINR gains translate into end-to-
end rate gains. The VNN selection schemes can significantly
improve the end-to-end rate performance particularly at the low
and mid throughput regimes. When serving BS is jointly
determined with the VNN selection, clear gains can be
observed, where these gains are higher in case of optimal VNN
selection (compare single BS and multi BS). In case of multi
BS (Optimal) and M
max
=5, the rate performance can be
improved by 40 times at 5%-ile rate level compared to the
direct link. Yet, VNN performance is worse than the direct link
as of 80%-ile because of the half-duplex limitation; yet, by
optimal cell selection the shown hull curve performance can be
approached.
Moreover, in case of a larger number of available VNNs
(see, M
max
=15), the overall performance can be further
improved. In particular, the rate performance is improved by
73 times at 5%-ile rate level compared to the direct link.
Nevertheless, the extra gain through multi-BS optimization
decreases, as also observed for the SINR gains in Section III-A.
C. AoF
The AoF, which reflects the severity of fading, can be
calculated from the first and the second moments of the SINR
as [11]
2
)]([
)var(
E
AoF
, (2)
where var(∙) denotes variance. Accordingly, AoF provides
further insights into fading mitigation via the proposed RAN
moderation strategies.
The AoF values on the backhaul link are depicted in Fig. 4
as a function of the shadowing standard deviation for M
max
=5
and 15. The case of zero shadowing standard deviation
indicates the absence of shadowing, where the channel is
impaired by multi-path fading only. It is seen that AoF on the
backhaul link decreases clearly when selection schemes are
applied. In particular, when the shadowing standard deviation
Fig. 4. AoF on the backhaul link as a function of the shadowing standard deviation (σ
dB
) on the backhaul link with different RAN moderation strategies
considering M
max
= 5 (left) and M
max
=15 (right).
is large, i.e., heavy shadowing, the selection schemes can
effectively reduce AoF. Coarse VNN selection performs well
in mitigating the deleterious impact of shadowing on the
backhaul link, while optimal VNN selection can yield further
reduction. That is, the lower bound for AoF is reached when
optimal NN selection is utilized. With the increasing number of
available VNNs, more AoF reduction can be observed (see,
M
max
=15). In addition, multi-BS optimization can assist in
reducing the AoF, where its impact is more pronounced in case
of smaller number of available VNNs (see, M
max
=5 and
compare single BS and multi BS) compared to a larger number
of available VNNs (see, M
max
=15 and compare single BS and
multi BS).
IV. CONCLUSION
In this paper, RAN moderation is demonstrated in dynamic
radio topology consisting of VNNs. Various RAN moderation
strategies are analyzed, and their implications on the SINR and
end-to-end rate distributions as well as AoF are shown. First of
all, the vital role of the considered selection schemes on the
VNN operation is highlighted taking into account different
performance metrics. The results indicate that coarse VNN
selection is a promising practical scheme that can substantially
improve the overall performance with less-frequent link quality
measurements. Optimal VNN selection would, however,
require frequent measurements to be able to follow the changes
in the channel conditions; thus, it can be considered illustrative
for showing the achievable gains via VNN selection schemes.
Additionally, serving BS selection together with the VNN
selection schemes can clearly improve the performance,
particularly when a small number of VNNs are available in the
target service region. The achieved gains, even in case of a
small number of available VNNs, motivate the utilization of
VNNs as a promising enhancement to heterogeneous networks
(HetNets) by enabling demand-driven dynamic radio topology
in 5G mobile and wireless communication networks.
ACKNOWLEDGMENT
Part of this work has been performed in the framework of
the H2020 project METIS-II co-funded by the EU. Authors
would like to acknowledge the contributions of their colleagues
from METIS-II although the views expressed are those of the
authors and do not necessarily represent the views of the
METIS-II project.
REFERENCES
[1] Ö. Bulakci, et al., "Dynamic Nomadic Node Selection for Performance
Enhancement in Composite Fading/Shadowing Environments," 2014
IEEE VTC Spring, Seoul, 2014, pp. 1-5.
[2] Z. Ren, S. Stanczak and P. Fertl, "Activation of nomadic relay nodes in
dynamic interference environment for energy saving," 2014 IEEE
Global Communications Conference, Austin, TX, 2014, pp. 4466-4471.
[3] METIS Deliverable D6.6, "Final report on the METIS 5G system
concept and technology roadmap," April 2015.
[4] Ö. Bulakci et al., "Towards flexible network deployment in 5G:
Nomadic node enhancement to heterogeneous networks," 2015 IEEE
ICC, London, 2015, pp. 2572-2577.
[5] NGMN Alliance 5G Initiative Team, "NGMN 5G White Paper," NGMN
LTD., Frankfurt, 2015.
[6] METIS-II Deliverable D5.2, “Final Considerations on Synchronous
Control Functions and Agile Resource Management Framework for
5G,” March 2017.
[7] G. Tsoulos, et al., “Performance of Vehicular Nomadic Node Operation
in Realistic Multicellular Environments,” 2nd International Workshop
on Research Advancements in Future Networking Technologies
(RAFNET 2017) at 2017 IEEE VTC Spring, Sydney, Australia, 2017.
[8] G. Tsoulos, et al.,, “Dynamic Wireless Network Shaping via Moving
Cells: The Nomadic Nodes Case,” Wiley Transactions on Emerging
Telecommunications Technologies (ETT), January 2017.
[9] T. Sahin, et al. “Nomadic Nodes for Dynamic Radio Topology in 5G
Wireless Networks: System-level Performance Evaluation,” 2017 IEEE
VTC Spring, Sydney, Australia, 2017.
[10] M. Caliskan, et al., “Predicting parking lot occupancy in vehicular ad
hoc networks,” in IEEE VTC2007-Spring, Dublin, Ireland, April 2007.
[11] M. Simon and M. Alouini, Digital communication over fading channels,
2nd ed. New York: John Wiley & Sons Inc., 2005.
