Aalborg Universitet
Five disruptive technology directions for 5G
Boccardi, Federico; W. Heath Jr., Robert ; Lozano, Angel ; L. Marzetta, Thomas ; Popovski,
Petar
Published in:
I E E E Communications Magazine
DOI (link to publication from Publisher):
10.1109/MCOM.2014.6736746
Publication date:
2014
Document Version
Early version, also known as pre-print
Link to publication from Aalborg University
Citation for published version (APA):
Boccardi, F., W. Heath Jr., R., Lozano, A., L. Marzetta, T., & Popovski, P. (2014). Five disruptive technology
directions for 5G. I E E E Communications Magazine, 52(2), 74 - 80 .
https://doi.org/10.1109/MCOM.2014.6736746
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 09, 2022
IEEE Communications Magazine • February 2014
74
0163-6804/14/$25.00 © 2014 IEEE
This work was carried out
when Federico Boccardi
was with Bell Labs, Alca-
tel-Lucent.
1
Most work in massive
MIMO has assumed oper-
ation at frequencies of 5
GHz or less. While the
same principles may prove
useful at millimeter wave-
lengths, a successful mar-
riage of massive MIMO
and millimeter waves may
take on a considerably
different form.
INTRODUCTION
The fifth generation (5G) cellular network is
coming. What technologies will define it? Will
5G be just an evolution of 4G, or will emerging
technologies cause a disruption requiring a
wholesale rethinking of entrenched cellular prin-
ciples? This article focuses on potential disrup-
tive technologies and their implications for 5G.
We classify the impact of new technologies,
leveraging the Henderson-Clark model [1], as
follows:
• Minor changes at both the node and archi-
tectural levels (e.g., the introduction of
codebooks and signaling support for a high-
er number of antennas). We refer to these
as evolutions in the design.
• Disruptive changes in the design of a class
of network nodes (e.g., the introduction of
a new waveform). We refer to these as com-
ponent changes.
• Disruptive changes in the system architec-
ture (e.g., the introduction of new types of
nodes or new functions in existing ones).
We refer to these as architectural changes.
• Disruptive changes that have an impact at
both the node and architecture levels. We
refer to these as radical changes.
We focus on disruptive (component, architec-
tural, or radical) technologies, driven by our
belief that the extremely higher aggregate data
rates and the much lower latencies required by
5G cannot be achieved with a mere evolution of
the status quo. We believe that the following five
potentially disruptive technologies could lead to
both architectural and component design
changes, as classified in Fig. 1.
1) Device-centric architectures: The base-sta-
tion-centric architecture of cellular systems may
change in 5G. It may be time to reconsider the
concepts of uplink and downlink, as well as con-
trol and data channels, to better route informa-
tion flows with different priorities and purposes
toward different sets of nodes within the net-
work. We present device-centric architectures.
2) Millimeter wave (mmWave): While spec-
trum has become scarce at microwave frequen-
cies, it is plentiful in the mmWave realm. Such a
spectrum “el Dorado” has led to an mmWave
“gold rush” in which researchers with diverse
backgrounds are studying different aspects of
mmWave transmission. Although far from being
fully understood, mmWave technologies have
already been standardized for short-range ser-
vices (IEEE 802.11ad) and deployed for niche
applications such as small-cell backhaul. We dis-
cuss the potential of mmWave for broader appli-
cation in 5G.
3) Massive MIMO: Massive multiple-input
multiple-output (MIMO)
1
proposes utilizing a
very high number of antennas to multiplex mes-
sages for several devices on each time-frequency
resource, focusing the radiated energy toward
the intended directions while minimizing intra-
and intercell interference. Massive MIMO may
require major architectural changes, particularly
in the design of macro base stations, and it may
also lead to new types of deployments. We dis-
cuss massive MIMO.
4) Smarter devices: 2G-3G-4G cellular net-
works were built under the design premise of
having complete control at the infrastructure
ABSTRACT
New research directions will lead to funda-
mental changes in the design of future fifth gen-
eration (5G) cellular networks. This article
describes five technologies that could lead to
both architectural and component disruptive
design changes: device-centric architectures, mil-
limeter wave, massive MIMO, smarter devices,
and native support for machine-to-machine com-
munications. The key ideas for each technology
are described, along with their potential impact
on 5G and the research challenges that remain.
5G WIRELESS COMMUNICATION SYSTEMS:
PROSPECTS AND CHALLENGES
Federico Boccardi, Vodafone
Robert W. Heath Jr., University of Texas at Austin
Angel Lozano, Universitat Pompeu Fabra
Thomas L. Marzetta, Bell Labs, Alcatel-Lucent
Petar Popovski, Aalborg University
Five Disruptive Technology
Directions for 5G
BOCCARDI_LAYOUT_Layout 1/30/14 12:57 PM Page 74
IEEE Communications Magazine • February 2014
75
side. We argue that 5G systems should drop this
design assumption and exploit intelligence at
the device side within different layers of the
protocol stack, for example, by allowing device-
to-device (D2D) connectivity or exploiting smart
caching at the mobile side. While this design
philosophy mainly requires a change at the node
level (component change), it also has implica-
tions at the architectural level. We argue for
smarter devices.
5) Native support for machine-to-machine
(M2M) communication: A native
2
inclusion of
M2M communication in 5G involves satisfying
three fundamentally different requirements asso-
ciated with different classes of low-data-rate ser-
vices: support of a massive number of low-rate
devices, sustaining a minimal data rate in virtual-
ly all circumstances, and very-low-latency data
transfer. Addressing these requirements in 5G
requires new methods and ideas at both the
component and architectural levels, and such is
the focus of a later section.
DEVICE-CENTRIC ARCHITECTURES
Cellular designs have historically relied on the
axiomatic role of “cells” as fundamental units
within the radio access network. Under such a
design postulate, a device obtains service by
establishing a downlink and an uplink connec-
tion, carrying both control and data traffic, with
the base station commanding the cell where the
device is located. Over the last few years, differ-
ent trends have been pointing to a disruption of
this cell-centric structure:
•The base station density is increasing rapid-
ly, driven by the rise of heterogeneous networks.
While heterogeneous networks were already
standardized in 4G, the architecture was not
natively designed to support them. Network den-
sification could require some major changes in
5G. The deployment of base stations with vastly
different transmit powers and coverage areas,
for instance, calls for a decoupling of downlink
and uplink in a way that allows the correspond-
ing information to flow through different sets of
nodes [5].
•The need for additional spectrum will
inevitably lead to the coexistence of frequency
bands with radically different propagation char-
acteristics within the same system. In this con-
text, [6] proposes the concept of a phantom cell
where the data and control planes are separated:
the control information is sent by high-power
nodes at microwave frequencies, whereas the
payload data is conveyed by low-power nodes at
mmWave frequencies.
•A new concept called centralized baseband
related to the concept of cloud radio access net-
works is emerging ([7]), where virtualization
leads to a decoupling between a node and the
hardware allocated to handle the processing
associated with this node. Hardware resources in
a pool, for instance, could be dynamically allo-
cated to different nodes depending on metrics
defined by the network operator.
•Emerging service classes, described later in
this article, could require a complete redefinition
of the architecture. Current works are looking at
architectural designs ranging from centralization
or partial centralization (e.g., via aggregators) to
full distribution (e.g., via compressed sensing
and/or multihop).
•Cooperative communications paradigms
such as cooperative multipoint (CoMP) or relay-
ing, which despite falling short of their initial
hype are nonetheless beneficial [8], could require
a redefinition of the functions of the different
nodes. In the context of relaying, for instance,
recent developments in wireless network coding
[9] suggest transmission principles that would
allow recovering some of the losses associated
with half-duplex relays. Moreover, recent
research points to the plausibility of full-duplex
nodes for short-range communication in the not-
so-distant future.
•The use of smarter devices could impact the
radio access network. In particular, both D2D
and smart caching call for an architectural redef-
inition where the center of gravity moves from
the network core to the periphery (devices, local
wireless proxies, relays).
Based on these trends, our vision is that the
cell-centric architecture should evolve into a
device-centric one: a given device (human or
machine) should be able to communicate by
exchanging multiple information flows through
several possible sets of heterogeneous nodes. In
other words, the set of network nodes providing
connectivity to a given device and the functions
of these nodes in a particular communication
session should be tailored to that specific device
and session. Under this vision, the concepts of
uplink/downlink and control/data channel should
be rethought (Fig. 2).
While the need for a disruptive change in
architectural design appears clear, major
research efforts are still needed to transform the
resulting vision into a coherent and realistic
proposition. Since the history of innovations [1]
indicates that architectural changes are often the
drivers of major technological discontinuities, we
believe that the trends above might have a major
influence on the development of 5G.
Figure 1. The five disruptive directions for 5G considered in this article, classi-
fied according to the Henderson-Clark model.
Component (node) disrupt.
Massive-MIMO
Smarter-devices
mmWave
Native support of
M2M services
Node-centric
networks
Architecture disrupt.
Further evolution of 4G
(out-of-scope of this
work)
2
As was learned with
MIMO, first introduced in
3G as an add-on and
then natively included in
4G, major improvements
come from native support
(i.e., from a design that is
optimized from its incep-
tion rather than amended
a posteriori).
BOCCARDI_LAYOUT_Layout 1/30/14 12:57 PM Page 75
IEEE Communications Magazine • February 2014
76
MILLIMETER WAVE
COMMUNICATION
Microwave cellular systems have precious little
spectrum: around 600 MHz are currently in use,
divided among operators [10]. There are two
ways to gain access to more microwave spectrum:
• To repurpose or refarm spectrum. This has
occurred worldwide with the repurposing of
terrestrial TV spectrum for applications such
as rural broadband access. Unfortunately,
repurposing has not freed up that much
spectrum, only about 80 MHz, and at a high
cost associated with moving the incumbents.
• To share spectrum utilizing, for instance,
cognitive radio techniques. The high hopes
initially placed on cognitive radio have been
dampened by the fact that an incumbent not
fully willing to cooperate is a major obstacle
to spectrum efficiency for secondary users.
Altogether, it appears that doubling the cur-
rent cellular bandwidth is the best case scenario
at microwave frequencies. Alternatively, there is
an enormous amount of spectrum at mmWave
frequencies ranging from 3 to 300 GHz. Many
bands therein seem promising, including most
immediately the local multipoint distribution ser-
vice at 28–30 GHz, the license-free band at 60
GHz, and the E-band at 71–76 GHz, 81–86 GHz,
and 92–95 GHz. Foreseeably, several tens of
gigahertz could become available for 5G, offering
well over an order of magnitude increase over
what is available at present. Needless to say,
work needs to be done on spectrum policy to
render these bands available for mobile cellular.
Propagation is not an insurmountable chal-
lenge. Recent measurements indicate similar
general characteristics as at microwave frequen-
cies, including distance-dependent path loss and
the possibility of non-line-of-sight communica-
tion. A main difference between microwave and
mmWave frequencies is the sensitivity to block-
ages: the results in [11], for instance, indicate a
path loss exponent of 2 for line-of-sight propaga-
tion but 4 (plus additional power loss) for non-
line-of-sight. MmWave cellular research will
need to incorporate sensitivity to blockages and
more complex channel models in the analysis,
and also study the effects of enablers such as
higher density infrastructure and relays. Another
enabler is the separation between control and
data planes, already mentioned.
Antenna arrays are a key feature in mmWave
systems. Large arrays can be used to keep the
antenna aperture constant, eliminating the fre-
quency dependence of path loss relative to omni-
directional antennas (when utilized at one side
of the link) and providing a net array gain to
counter the larger thermal noise bandwidth
(when utilized at both sides of the link). Adap-
tive arrays with narrow beams also reduce the
impact of interference, meaning that mmWave
systems could more often operate in noise-limit-
ed rather than interference-limited conditions.
Since meaningful communication might only
happen under sufficient array gain, new random
access protocols are needed that work when
transmitters can only emit in certain directions
and receivers can only receive from certain
directions. Adaptive array processing algorithms
are required that can adapt quickly when beams
are blocked by people or some device antennas
become obscured by the user’s own body.
MmWave systems also have distinct hardware
constraints. A major one comes from the high
power consumption of mixed signal components,
chiefly the analog-to-digital converters (ADCs)
and digital-to-analog converters (DACs). Thus,
the conventional microwave architecture where
every antenna is connected to a high-rate
ADC/DAC is unlikely to be applicable to
mmWave without a huge leap forward in semi-
conductor technology. One alternative is a hybrid
architecture where beamforming is performed in
analog at RF, and multiple sets of beamformers
are connected to a small number of ADCs or
DACS; in this alternative, signal processing algo-
rithms are needed to steer the analog beam-
forming weights. Another alternative is to
connect each RF chain to a 1-bit ADC/DAC,
with very low power requirements; in this case,
the beamforming would be performed digitally
but on very noisy data. There are abundant
research challenges in optimizing different
transceiver strategies, analyzing their capacity,
incorporating multiuser capabilities, and leverag-
ing channel features such as sparsity.
A data rate comparison between technologies
is provided in Fig. 3, for certain simulation set-
tings, in terms of mean and 5 percent outage
rates. MmWave operation is seen to provide very
high rates compared to two different microwave
systems. The gains exceed the 10× spectrum
increase because of the enhanced signal power
and reduced interference thanks to directional
beamforming at both the transmitter and receiver.
From the discussion above, and referring
again to the Henderson-Clark model, we con-
clude that mmWave requires radical changes in
the system, as it has a strong impact in both the
component and architecture designs. Conse-
Figure 2. Example of device-centric architecture.
Microwave
high-power BS
Mobile
device
D2D
Data
Control
mmWave BS
Centralized
baseband
Microwave
low-power BS
Uplink
Downlink
Sensors
mmWave
indoor
BOCCARDI_LAYOUT_Layout 1/30/14 12:57 PM Page 76
IEEE Communications Magazine • February 2014
77
quently, we view mmWave as a potentially dis-
ruptive technology for 5G, which, provided the
above discussed challenges can be tackled, could
lead to unrivaled data rates and a completely
different user experience.
MASSIVE MIMO
Massive MIMO (also referred to as “Large-Scale
MIMO” or “Large-Scale Antenna Systems”) is a
form of multiuser MIMO in which the number of
antennas at the base station is much larger than
the number of devices per signaling resource
[14]. Having many more base station antennas
than devices renders the channels to the different
device s quasi-orthogonal and very simple spatial
multiplexing/de-multiplexing procedures quasi-
optimal. The favorable action of the law of large
numbers smoothens out frequency dependencies
in the channel and, altogether, huge gains in
spectral efficiency can be attained (Fig. 4).
In the context of the Henderson-Clark frame-
work, we argue that massive MIMO has a dis-
ruptive potential for 5G:
• At a node level, it is a scalable technology.
This is in contrast with 4G, which, in many
respects, is not scalable: further sectoriza-
tion therein is not feasible because:
–There is limited space for bulky azimuthal-
ly directive antennas.
–There is an inevitable angle spread of the
propagation; in turn, single-user MIMO is
constrained by the limited number of anten-
nas that can fit in certain mobile devices.
In contrast, there is almost no limit on the
number of base station antennas in massive
MIMO provided that time-division duplex-
ing is employed to enable channel estima-
tion through uplink pilots.
• It enables new deployments and architec-
tures. While one can envision direct replace-
ment of macro base stations with arrays of
low-gain resonant antennas, other deploy-
ments are possible, such as conformal arrays
on the facades of skyscrapers or arrays on
the faces of water tanks in rural locations.
Moreover, the same massive MIMO princi-
ples that govern the use of collocated arrays
of antennas also apply to distributed deploy-
ments in which a college campus or an
entire city could be covered with a multi-
tude of distributed antennas that collectively
serve many users (in this framework, the
centralized baseband concept presented ear-
lier is an important architectural enabler).
While very promising, massive MIMO still
presents a number of research challenges. Chan-
nel estimation is critical and currently represents
the main source of limitations. User motion
imposes a finite coherence interval during which
channel knowledge must be acquired and uti-
lized, and consequently there is a finite number
of orthogonal pilot sequences that can be
assigned to the devices. Reuse of pilot sequences
causes pilot contamination and coherent inter-
ference, which grows with the number of anten-
nas as fast as the desired signals. The mitigation
of pilot contamination is an active research
topic. Also, there is still much to be learned
about massive MIMO propagation, although
experiments thus far support the hypothesis of
channel quasi-orthogonality. From an implemen-
tation perspective, massive MIMO can potential-
ly be realized with modular low-cost low-power
hardware with each antenna functioning semi-
autonomously, but a considerable development
effort is still required to demonstrate the cost
effectiveness of this solution. Note that at the
microwave frequencies considered in this sec-
tion, the cost and energy consumption of
ADCs/DACs are sensibly lower than at mmWave
frequencies.
From the discussion above, we conclude that
the adoption of massive MIMO for 5G could
represent a major leap with respect to today’s
state of the art in system and component design.
To justify these major changes, massive MIMO
proponents should further work on solving the
challenges emphasized above and showing realis-
tic performance improvements by means of the-
oretical studies, simulation campaigns, and
testbed experiments.
SMARTER DEVICES
Earlier generations of cellular systems were built
on the design premise of having complete con-
trol at the infrastructure side. In this section, we
discuss some of the possibilities that can be
unleashed by allowing the devices to play a more
active role and then how 5G’s design should
account for an increase in device smartness. We
focus on three different examples of technolo-
gies that could be incorporated into smarter
devices: D2D, local caching, and advanced inter-
ference rejection.
Figure 3. Cell data rate comparison between microwave systems using 50 MHz
of bandwidth (single-user single-antenna and single-user MIMO) and a
mmWave system with 500 MHz of bandwidth and a single user. Results are
given in terms of gain (%) w.r..t. the MIMO 4 × 4 baseline.
3
More details
about the comparison setup are provided in [12].
50th
percentile
200
0
Percentage
400
600
800
1000
SISO
MIMO 4 4
mmWave
5th
percentile
3
The results given in this
article have been obtained
by considering the current
state-of-the-art under-
standing of the technolo-
gies considered. However,
we emphasize that at this
point it is not yet possible
to provide a fully realistic
assessment and a compar-
ison with deployed 4G sys-
tems. Undeniably, some
research efforts are still in
the so-called hype phase,
and much work is still
required before a steady
understanding of the per-
formance and the required
enablers can be reached.
BOCCARDI_LAYOUT_Layout 1/30/14 12:57 PM Page 77