Received February 19, 2020, accepted March 8, 2020, date of publication March 18, 2020, date of current version

March 31, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2981745

Communications in the 6G Era

HARISH VISWANATHAN 1, (Fellow, IEEE), AND PREBEN E. MOGENSEN 2, (Member, IEEE)
1 Nokia Bell Labs, Murray Hill, NJ 07974, USA
2 Nokia Bell Labs, 9220 Aalborg, Denmark

Corresponding author: Preben E. Mogensen ([email protected])

This work was supported by Nokia.

ABSTRACT The focus of wireless research is increasingly shifting toward 6G as 5G deployments get
underway. At this juncture, it is essential to establish a vision of future communications to provide guidance
for that research. In this paper, we attempt to paint a broad picture of communication needs and technologies
in the timeframe of 6G. The future of connectivity is in the creation of digital twin worlds that are a true
representation of the physical and biological worlds at every spatial and time instant, unifying our experience
across these physical, biological and digital worlds. New themes are likely to emerge that will shape 6G
system requirements and technologies, such as: (i) new man-machine interfaces created by a collection of
multiple local devices acting in unison; (ii) ubiquitous universal computing distributed among multiple local
devices and the cloud; (iii) multi-sensory data fusion to create multi-verse maps and new mixed-reality
experiences; and (iv) precision sensing and actuation to control the physical world. With rapid advances in
artificial intelligence, it has the potential to become the foundation for the 6G air interface and network,
making data, compute and energy the new resources to be exploited for achieving superior performance.
In addition, in this paper we discuss the other major technology transformations that are likely to define 6G:
(i) cognitive spectrum sharing methods and new spectrum bands; (ii) the integration of localization and
sensing capabilities into the system definition, (iii) the achievement of extreme performance requirements
on latency and reliability; (iv) new network architecture paradigms involving sub-networks and RAN-Core
convergence; and (v) new security and privacy schemes.

INDEX TERMS 6G, AI/ML driven air interface, network localization and sensing, cognitive spectrum
sharing, sub-terahertz, RAN-Core convergence, subnetworks, security, privacy, network as a platform.

I. INTRODUCTION On the other hand, modifications that are a fundamental shift

With the deployment of 5G systems in full swing, the research and are incompatible with the existing 5G framework or can
focus toward 6G mobile cellular systems has begun [1]–[5]. only be incorporated with high cost to the network or devices
Keeping up with the tradition of a new generation of cellular will be part of the next generation.
system once every ten years or so, there is an expectation that Besides enhanced mobile broadband for consumers, 5G is
a 6G system will be standardized with deployments starting widely expected to enable the Fourth Industrial Revolution,
before 2030. Since it often takes more than ten years for a or Industry 4.0, through the digitalization and connectivity
novel technology to see the commercial daylight, it is time to of all things big and small. Digital twins of various objects
begin research on novel technology components for 6G. created in edge clouds will form the essential foundation of
It is essential to establish a vision of future communica- the future digital world. Digital twin worlds of both physi-
tions to provide guidance for research, which is the purpose cal and biological entities will be an essential platform for
of this paper. We attempt to paint a broad picture of commu- the new digital services of the future. The realization of
nication needs and technologies in the timeframe of 6G. It is a comprehensive digital world that is a complete and true
possible that some of these requirements can already be met representation of the physical world at every spatial and time
by incorporating new technologies within the 5G framework. instant will require an enormous amount of capacity at low
In general, we expect to see as part of the 5G evolution the latency. Digitalization will also pave the way for the creation
introduction of any modifications that can be introduced in of new virtual worlds with digital representations of imagi-
a backward-compatible fashion at a reasonable cost within nary objects that can be blended with the digital twin world
the 5G framework to meet new performance requirements. to various degrees to create a mixed-reality, super-physical
world. As smart watches and heart rate monitors transform
The associate editor coordinating the review of this manuscript and into skin patchables, ingestables, body implants, body armor
approving it for publication was Adnan Shahid . skeleton and brain activity detectors, the biology of humans
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
VOLUME 8, 2020 57063
 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

FIGURE 1. 6G for the inter-connection of physical, biological, and digital worlds.

will be mapped accurately every instant and integrated into mobile networking as it seeks to address increasingly spe-
the digital and virtual worlds, enabling new super-human cialized requirements, and the consequent implications on the
capabilities. Augmented reality user interfaces will enable evolution of standardization. We conclude Section VII with a
efficient and intuitive human control of all these worlds, brief summary.
whether physical, virtual or biological.
The connectivity of the future is therefore about enabling II. WHAT WILL 6G BE USED FOR?
the seamless integration of these different worlds, illustrated What will life and our digital society on the other side of
in Figure 1, to create a unified experience for humans, the 2030s look like? We begin with devices that humans
or should we say create an internet of cyborgs. When consid- may use to connect to the network. While the smartphone
ering such a future, the following major new themes emerge and the tablet will still be around, we are likely to see
in addition to the new communication needs: (i) end devices new man-machine interfaces that will make it substantially
extending from being single entities to a collection of multiple more convenient for us to consume and control information.
local entities acting in unison to create the new man-machine We expect that:
interface; (ii) ubiquitous universal computing distributed • Wearable devices, such as earbuds and devices embed-
among multiple local devices and the cloud; (iii) knowledge ded in our clothing, will become common, and skin
systems that store, process and convert data into actionable patches and bio-implants may not be so uncommon.
knowledge; and (iv) precision sensing and actuation to con- We might even become reliant on new brain sensors
trol the physical world. Recently, several publications have to actuate machines. We will have multiple wearables
espoused their views on 6G, see e.g., [4], [6]–[10]. We take that we carry with us and they will work seamlessly
a unique and broader perspective, focusing not only on the with each other, providing natural, intuitive interfaces.
technologies but also the human transformation we expect in Figure 2 shows the potential evolution in devices.
the 6G era, which helps to provide a view of the performance • Touch-screen typing will likely become outdated.
requirements and design principles for 6G. Our view on the Gesturing and talking to whatever devices we use to get
technology transformations starts from where the current 5G things done will become the norm.
systems are, moving to how they are evolving, and then to • The devices we use will be fully context-aware, and
what may become fundamentally different beyond that. We the network will become increasingly sophisticated at
also address transformations likely to happen in the nature of predicting our needs. This context awareness combined
standardization needed in a world with open platforms. with new human-machine interfaces will make our inter-
The paper is organized as follows. In Section II we dive action with the physical and digital world much more
deeper into what the world might become in the 2030s, and intuitive and efficient.
from that draw relevant new use cases for 6G. This leads to The computing needed for these devices will likely not all
a description in Section III of the potential requirements and reside in the devices themselves because of form factor and
performance indicators that will distinguish 6G. In Section IV battery power considerations. Rather, they may have to rely
we discuss some new fundamental dimensions to consider on locally available computing resources to complete tasks,
in the design of 6G radio interfaces, in addition to the tradi- beyond the edge cloud. Networks will thus play a significant
tional dimensions of space, spectrum and spectral efficiency. role in the man-machine interface of tomorrow.
In Section V we elaborate on the new technologies that may As consumers, we can expect that:
form the basis for a new generation of cellular networks. • The self-driving concept cars of today will be available
In Section VI we discuss the open platform approach to to the masses by the 2030s. They will be self-driving

57064 VOLUME 8, 2020

H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

FIGURE 2. Likely evolution in devices.

most of the time but will still likely need at least a • Health care will be substantially transformed, with
remote driver or the passenger to take control under 24/7 monitoring of vital parameters for both the healthy
certain conditions. This will substantially increase the and the sick through numerous wearable devices. Health
time available for us to consume data from the internet monitoring will also include in-body devices that com-
in the form of more entertainment, rich communications municate with wearables outside, which in turn can
or education. The cars themselves will also consume transport the data to the internet.
significantly more data: vehicle sensor data will be The transformation to Industry 4.0 and the first wave of
uploaded in real-time to the network, high-resolution wireless-enabled automation will already have happened
maps will be downloaded and cars will link directly to before the 2030s. 5G networks providing ultra-reliable
one another. low-latency communications (URLLC) will have facilitated
• There will be a massive deployment of wireless cameras real-time processing in the cloud. However, industrial use
as sensors. With advances in AI and machine vision and cases relying on much more extreme requirements for wire-
their capacity to recognize people and objects (or more less communication will require 6G.
generally, automatically gather information from images • Holographic telepresence will become the norm for both
and videos), the camera will become a universal sensor work and social interaction. It will be possible to make it
that can be used everywhere. Privacy concerns will be appear as though one is in a certain location while really
addressed by limiting access to data and anonymizing being in a different location, for example, appearing to
information. Also, radio and other sensing modalities be in the office while actually being in the car. We will
like acoustics will be used to gather information on the have systems that combine current facial expressions
environment. with a virtual self within the digital representation of any
• Advanced techniques will be used in security-screening physical world.
procedures to eliminate security lines. A combination of • We will see massive use of mobile robot swarms and
various sensing modalities will be used to screen people drones in various verticals such as hospitality, hospitals,
as they move through crowded areas rather than only at warehouses and package delivery.
entrances. Radio sensing will be an essential component • Dynamic digital twins in the digital world with increas-
of achieving this, supported by the communication sys- ingly accurate, synchronous updates of the physical
tems of the future. world will be an essential platform for augmenting
• Digital cash and keys may become the norm, with trans- human intelligence.
actions in both the physical and digital worlds being con- Based on the above vision of the future, we can extrapolate
ducted through the plethora of devices that we will have. the following key use cases. These are a combination of what
The network of the future should provide the security 5G will enable but with adoption at scale in the timeframe
and privacy that is fundamental to such a transformation. of 6G using new technologies, plus a set of new use cases
• Numerous domestic service robots will complement the enabled by the new 6G technologies. See Table 1.
vacuum cleaners and lawn mowers we know today.
These may take the form of a swarm of smaller robots III. SIX REQUIREMENTS AND KEY PERFORMANCE
that work together to accomplish tasks. The robots will MEASURES FOR 6G
be equipped with video cameras streaming to a local The array of new use cases we expect to see by 2030 and
compute server for real-time processing. Thus, we will beyond will drive the new requirements that need to be
see an increase in the number of devices and higher achieved by 6G. The 5G key performance indicators (KPIs)
capacity requirements within our home networks. of data rate/throughput/capacity, latency, reliability, scale and

VOLUME 8, 2020 57065

 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

TABLE 1. Use cases for 6G.

flexibility will continue to be important measures for 6G 3) Finally, we expect a major revolution in the end device
performance. Several new characteristics will also become in the timeframe of 6G. Hence, we introduce a few
important for 6G given the potential use cases described in characteristics under a device category to point out
the previous section. In Figure 3, we group the requirements the major transitions that we expect. First, we believe
for 6G into six categories - three categories with KPIs similar that the end device will evolve in many scenarios to
to 5G and three new categories: be a network of devices or a sub-network. As exam-
1) Localization and sensing using the communication net- ples, we can imagine a machine-area network or a
work will be an important feature of 6G. We identify robot-area network involving connecting multiple parts
precision and accuracy as the corresponding perfor- of a machine such as a controller and its drives. Another
mance measures for localization and sensing, respec- hallmark of the device in the timeframe of 6G will
tively. We expect that centimeter-level precision will be be that we will have much more intuitive interfaces,
achieved. Object sensing accuracy can be measured in with access through gesturing rather than typing, for
terms of missed detection (MD) and false alarm (FA) example. Finally, another possibility for a certain class
probabilities and parameter estimation errors. of device is one that will be extremely low-power
2) The network will be engineered with distributed AI/ML and potentially battery-less, relying on the network to
techniques embedded in various nodes, and how power the device.
quickly they adapt to new conditions in the network
is an important measure. Network automation will be IV. SIX FUNDAMENTAL DIMENSIONS TO DESIGN 6G
the norm, and thus how close a network is to com- In every generation until 5G, the three fundamental dimen-
plete automation with zero manual intervention will be sions of spectrum, spectral efficiency and spatial reuse have
another criterion. dictated how we can grow capacity, as represented by the light
blue triangle in Figure 4. It will continue to be the case for
6G. RF technology will advance to power and cost-effectively
use spectrum in even higher bands. There is the opportunity
of at least a tenfold increase in the amount of spectrum
by going to terahertz-frequency bands. Spectral efficiency
will improve through the use of massive multi-user MIMO
not only in centimeter wave (cmWave) but also millime-
ter wave (mmWave) bands as we transition from analog to
hybrid/digital beamforming in these lower mmWave bands.
As the cost of massive MIMO falls, even larger arrays may
be deployed to further increase spectral efficiency. Network
densification will undoubtedly continue to increase - not
only for capacity reasons but simply to provide increased
coverage at higher-frequency bands, at higher data rates and
with higher reliability. Furthermore, more pervasive spec-
FIGURE 3. Key requirements and characteristics of 6G. Object sensing is trum access will come into play; the sharing between oper-
characterized though missed detection (MD) and false alarm (FA). ators of even licensed spectrum powered by software-defined

57066 VOLUME 8, 2020

H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

change solutions will become a major focus everywhere in

the world by the 2030s, and the growing energy consumption
of networks and devices will be highly scrutinized. Thus,
energy becomes another important dimension for the design
of 6G. 6G will therefore have six fundamental dimensions
that research will need to explore to achieve flexible perfor-
mance targets, as illustrated in Figure 4.

V. SIX KEY TECHNOLOGIES FOR 6G

A new generation is ultimately characterized by the num-
ber of novel, essential technologies that shape the commu-
nication system. Truly fundamental new technologies typi-
cally take a decade or more to become realized in practice.
In view of this, the truly novel technologies forming 6G
FIGURE 4. Three new fundamental dimensions for system design.
must be research concepts today. Keeping with the theme
of ‘‘six’’ for 6G, we have identified six new potential tech-
nology transformation that we expect to be part of shaping
radio (SDR) and AI/ML will allow much higher reuse of the 6G system: (i) AI/ML-driven air interface design and
spectrum. Efficient spectrum reuse is especially important in optimization; (ii) expansion into new spectrum bands and
the lower bands as they have good non-line-of-sight (NLOS) new cognitive spectrum sharing methods; (iii) the integration
propagation properties and spectrum resources in those bands of localization and sensing capabilities into system defini-
are scarce. tion; (iv) the achievement of extreme performance require-
We suggest that 6G will fundamentally differ from the ments on latency and reliability; (v) new network architecture
previous generations in that three new fundamental dimen- paradigms involving sub-networks and RAN-Core conver-
sions will come into play in addition to the above three tradi- gence; and (vi) new security and privacy schemes. Each of
tional dimensions, as shown by the reverse dark-blue triangle these is described in the following subsections.
in Figure 4. These dimensions represent the fundamental In our choice of the six key technologies for 6G, we are
resources of data, compute and energy. As is well known, aware that certain emerging technologies, not yet part of the
AI/ML techniques are data-driven, and whoever has access current 5G specification are omitted; still we see them as
to large volumes of domain-specific data will be success- part of the 5G era. Examples are, full duplex communica-
ful in applying these techniques. Application of AI/ML to tion, cell free communication [11], Unmanned Aerial Vehicle
the design of 6G systems will be fundamental, and simi- (UAV) [12], High Altitude Platform Systems (HAPS) [13]
lar to various other domains, network and sensor data will and Low Earth Orbit (LEO) satellite based communica-
become fundamental resources to be exploited to improve tion [14]. At the other end of the time axis for the 6G era,
the performance of the system. While computing power has we have also excluded more long term emerging technologies
always been an important resource for cellular systems, two like Quantum communication [15], molecular communica-
major trends point in the direction of that becoming a limited tion [16] and Visible Light Communication [17] to become
resource, and hence how that is exploited in 6G will become key technologies for 6G. We rather see these technologies
significant. The first trend we observe is the emerging sat- either for special solutions in near term or more generalized
uration in the number of transistors that can be packed into in the time context of beyond 6G networks.
a unit volume, which limits the computing power of devices.
The second trend is that we will adopt multiple end devices to A. THE EXPANDING ROLE OF ARTIFICIAL INTELLIGENCE
augment human sensing capabilities, such as glasses, earbuds AND MACHINE LEARNING
and other wearables, which all have very small form factors AI and ML techniques, especially deep learning, have rapidly
and hence will have limited computing capability. The current advanced over the last decade and are now central to several
approach of computing offloaded to the edge cloud is unlikely domains involving image classification and computer vision,
to be sufficient to meet the synchronous computing needs ranging from social networks to security. They are applied in
across the different devices. Leveraging computing that is problem areas where significant amounts of data are readily
available in the local area but separate from the devices will available for training. Reinforcement learning is beginning
be a new theme in the 6G timeframe. In this sense, we treat to be applied in a variety of robotic control applications
compute as another essential dimension driving the design of following various demonstrations of its prowess in gaming
the new communication system. Finally, available energy at environments, such as AlphaGo.
every element of the network will determine the achievable Recently there has been much exploration of the applica-
performance. This ranges from near-to-zero energy at some tion of deep-learning techniques to wireless systems. Over the
types of devices, to power supply limits at radio base stations, next few years, we expect AL/ML to be applied to 5G systems
and to power constraints in data centers. In addition, climate in at least three different ways. First, they have the potential

VOLUME 8, 2020 57067

 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

FIGURE 5. End-to-end learning systems adapting to hardware and channels.

to replace some of the model-based Layer 1 and Layer receiver learn to pick the best design for these parameters,
2 algorithms such as channel estimation, preamble detection, as illustrated in Figure 5. While such an end-to-end learn-
equalization and user scheduling, either because they perform ing approach may be unfeasible for complex, dynamically
better or are less complex. Second, they are likely to be changing multi-user environments, the 6G communication
applied extensively in deployment optimization, for example framework will be designed in such a way as to allow learning
for configuring an optimal subset of beams with which to in the field to make some design choices. This will enable
illuminate the coverage area, taking cell traffic patterns into optimization of the air-interface characteristics based on the
account. Given the complexity of 5G systems in terms of the choice of spectrum, environment, hardware deployed and
sheer number of parameters to be configured at the time of target requirements. One important shift will be to include
deployment, AI/ML techniques will play an important role in the capabilities of the hardware in the optimization of the
the vision of zero human touch network optimization. Finally, communication framework. In the current approach, the air-
we can expect some other use cases such as localization interface is designed taking into account some practical limits
of end devices using 5G technology to exploit learning on implementation. But after the design phase, it is expected
techniques for improved accuracy. In [18], a comprehensive that all implementations will have the hardware required for
review of the possible applications of AI/ML to 5G systems the chosen air-interface design. In the future, we can expect
and beyond is provided. Similarly, [19] provides a great the air-interface to adapt to the capabilities of the hardware.
summary of applications classified by the type of learning For example, a certain implementation may have a limited
technique, in addition to an overview of Quantum computing number of analog-to-digital conversion (ADC) or digital-to-
and communications. analog conversion (DAC) resolutions, which can be taken
In addition to the use of AI/ML in the RAN, AI/ML will into account by the learning systems to determine the optimal
become essential for the 5G end-to-end network automation signaling choice.
dealing with the complexity of orchestration across multiple
network domains and layers. This will allow for dynamic
adaptation of network and cloud resources according to 2) COGNITIVE SPECTRUM USE
changing demands, rapid deployment of new services and Low-frequency spectrum will continue to be of paramount
fast mitigation of failures, while significantly reducing oper- importance for wide-area coverage due to the superior prop-
ational expenditures. agation properties in NLOS compared to higher-frequency
We envision 6G systems to employ AI/ML in a more bands. Over the next decade, substantial amounts of new
fundamental way than the above 5G approach. We expect spectrum will be allocated to 5G and its evolutions, and that is
to go from AI as an enhancement to AI as a foundation likely to lead to near-exhaustion of spectrum in bands below
for air interface design and optimization - self-optimizing 6 GHz. Thus, in the timeframe of 6G, new spectrum-use
transmitters and receivers, cognitive spectrum use and context methods will be required even within the licensed spectrum
awareness. regime to allow better local access to the spectrum and
coexistence with other users. Operators may need to share
spectrum among themselves and with other private dedicated
1) SELF-OPTIMIZING TRANSMITTERS AND RECEIVERS networks. And even within a single operator, multiple gener-
Ongoing research has demonstrated that deep-learning sys- ations of technologies will coexist and share spectrum. With
tems can learn to communicate over quasi-static links more advances in radio technology enabling multi-band operation
efficiently than model-based system designs [20], [21]. and learning techniques such as deep reinforcement learning,
No explicit design of waveform, constellation, or reference efficient autonomous sharing of spectrum can alleviate major
signals is required. Through extensive training, a single spectrum sharing hurdles [22], [23]. With increasing use of
deep-learning network at the transmitter and one at the advanced beamforming techniques and densification, use of

57068 VOLUME 8, 2020

H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

spectrum becomes highly local, facilitating more reuse spec- B. EXPLOITING NEW SPECTRUM BANDS
trum and hence allowing various forms of coexistence among The need for higher peak rates and capacity has con-
cognitive sharing systems that will be highly beneficial. tinuously driven mobile communication systems to utilize
higher-spectrum bands. New spectrum bands between 3GHz
and 6GHz and from 24GHz to 50GHz have been allocated
3) CONTEXT AWARENESS for 5G in various regions. 5G evolution will surely include
Another major development we can expect in the timeframe new bands between 52.6GHz and 114GHz [26]. Entirely
of 6G is seamless integration of awareness of the environ- new physical layer designs with transition to single-carrier
ment, traffic patterns, mobility patterns and location into waveforms may be possible within the highly flexible 5G
the optimization of communication schemes aided by new framework. The major challenge to using these high bands
AI/ML techniques [24]. For example, in environments such has been the realization of high-output power devices at
as factory floors, video cameras will be able to capture the reasonable cost. Massive antenna arrays are employed to
presence and movement of various machines and devices that form narrow-beam-width transmitters to increase the effec-
can be processed in real time through deep-learning networks tive isotropic radiated power (EIRP) and range [27]. Signal
to predict changes to the propagation environment, which propagation in these high bands also poses a challenge, since
in turn can be used to optimize communication. Essentially, signals are easily blocked owing to their small wavelength.
new data acquisition and processing techniques integrated Diffraction around objects is limited and signal absorption by
into the communication system can reduce the randomness water is significant. On the other hand, in dense urban and
in the communication links. Long-term mobility patterns indoor environments, reflections from buildings and walls
can be derived in indoor and outdoor settings that can then allow NLOS coverage along the same street or corridor as the
be used to optimize the service experience by establish- access point [28]. Despite these challenges and idiosyncrasies
ing connectivity to the right technology at the right time. of high-band propagation, a significant amount of research
Another important element of future systems may be the use and development is leading to viable deployments.
of digitally controlled passive elements such as large-scale The trend of using ever-higher band spectrum will con-
meta-surfaces [25]. These are likely to be distributed oppor- tinue. In the time of 6G systems, we expect sub-terahertz
tunistically, especially in indoor environments, and new bands from 114GHz to 300GHz (see Figure 6) to become
methods are required to exploit this for improving com- available and practical for use in cellular systems in specific
munications. Determining optimal control of these elements scenarios. An obvious use case for sub-terahertz spectrum
using model-based optimization methods may be intractable. is for backhaul in integrated access and backhaul networks
It will be challenging to exactly model signal propagation of the future. Narrow beam point-to-point communication
incorporating their collective effects, which in turn depend in these bands can free up spectrum for access in mmWave
on how they are controlled. AI/ML techniques will likely be bands. Other potential use cases include short-range commu-
used to solve such complex problems in the 6G era. Higher- nications across display and compute devices and rack-to-
level semantic knowledge of how the communication is being rack communications in rapidly deployable edge data centers.
used, for example whether it is for robot control or augmented
reality in a factory or for gaming, can be learned from traffic
patterns and device characteristics, and appropriate services
can be provided automatically. Accurate service personal-
ization down to the lower layers of the communication can
be achieved through learning techniques. Moving from AI
for 5G to AI native 6G, we expect that various forms of
learning will be employed to realize the above applications.
Transfer learning and federated learning will play critical
roles. Systems will have to be trained offline in simulation
environments to a sufficient extent first so that basic commu-
nications can be established, and then be subsequently trained
in the field to optimize performance. So there will be transfer
of learning from the simulation to the field environment. FIGURE 6. Spectrum options for 6G.
Devices and network infrastructure have to co-learn to incor-
porate end-to-end operations, and here, federated learning A significant amount of ongoing research to improve
will play a role. Rather than sharing large data sets between mmWave systems will also be beneficial for the sub-terahertz
various devices and the network, models will be shared. systems. Initial product design in those bands will follow
At the higher layers, deep reinforcement learning will be nec- the approach established in mmWave systems today. New
essary for optimization of resource allocation and control of radio frequency integrated circuits (RFICs) with on-chip
various parameters. Hierarchical and multi-agent reinforce- or on-board antenna arrays and with phase shifters capa-
ment learning will need to be used across different nodes. ble of forming narrow beams will be implemented [29].

VOLUME 8, 2020 57069

 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

New component technologies such as antenna-on-glass could capabilities to improve localization accuracy and can become
emerge that help reduce the cost of devices [30]. Hybrid the single system for both URLLC and localization in indus-
beamforming will be needed to achieve massive capaci- trial automation environments.
ties using single-user or multiple-user MIMO. New receiver As we move toward 6G, we expect the network to perform
architectures with pre-combing before the low-noise ampli- various sensing tasks in addition to high-precision local-
fier could be introduced [31]. To reduce power consumption, ization. Localization solutions will be enhanced to achieve
new waveforms designed using AI/ML techniques suitable centimeter-level accuracy indoors over a larger area where
for single-bit converters are being explored [32]. The propa- line-of-sight visibility to a large number of access points
gation delay across the antenna array becomes comparable to is limited. New channel charting methods based on AI/ML
the symbol duration with large bandwidth signals, and hence techniques applied to large antenna array systems [34] as well
new signal processing schemes to handle this beam squint as data fusion across RF, camera and other sensors on robots
will be required. New channel measurements and models for will improve accuracy of sensing even with a limited number
using these spectrum bands for access will also be required. of visible access points.
In addition to exploiting sub-terahertz spectrum bands, 6G systems will be used for imaging of passive objects.
lower-cost massive MIMO techniques will enable much bet- System design will not only be optimized for communica-
ter use of spectrum in mmWave and cmWave bands. Early tion but will also incorporate special capabilities for sens-
mmWave systems relied on analog beamforming and thus ing. For example, waveforms suitable for sensing such as
are restricted in terms of number of users that can be served chirp signals can be multiplexed with waveforms optimized
simultaneously from a single panel. As network density for communications [35]. Large antenna arrays deployed for
increases and massive MIMO technologies are cost-reduced, massive MIMO communications can be leveraged for form-
multi-user MIMO will be widely applied in mmWave bands ing narrow beams that can be periodically swept for sensing.
to enable massive-scale, multi-user massive MIMO to exploit Multiple transmitters and receivers can coordinate to enhance
the available spectrum. What are considered high bands the sensing capabilities of the network. The evolution to
today will essentially become mid bands in the time of 6G. sub-terahertz and terahertz bands, with the associated large
In the lowest-frequency bands for 6G, namely the lower signaling bandwidth, increases the opportunity for precision
cmWave band, the applicability of massive MIMO becomes sensing. Millimeter-precision imaging using terahertz-band
gradually restricted by the large size of the antenna ele- infrastructure will enable a significant number of new use
ments. The basic path loss and material penetration properties cases in industry automation and health care, such as fault
are much better towards the sub-gigahertz frequencies, and detection in extrusion manufacturing processes or detection
the lower-frequency bands will remain being essential for of cancerous tissue and tooth cavities [36]. There are plenty of
wide-area coverage at the time of 6G. Spectrum availability is applications that will benefit from radio points being turned
scarce and research on improved spectrum utilization for the into sensors, such as food quality control in supermarkets,
lower-frequency bands is important. Spectrum assignment or invisible metal detectors in airport or event infrastructures
will move from a static split between operators and services to replace the current security gate systems.
toward much more dynamic AI-based spectrum access in Combining the multi-modal sensing capabilities with the
time, frequency and space (see Figure 6). Visible light com- cognitive technologies enabled by the 6G platform will allow
munication is likely to be used in limited scenarios but is for analyzing behavioral patterns and people’s preferences
unlikely to become a mainstream 6G technology, since radio and even emotions, hence creating a sixth sense that antici-
communications will be cheaper to achieve at the same data pates user needs. It will allow for interactions with the phys-
rates. ical world in a much more intuitive way.

C. THE NETWORK WITH THE SIXTH SENSE D. EXTREME NETWORKING

One of the critical requirements for industrial automation The new Industrial Internet of Things (IIoT) use cases tar-
is high-accuracy localization. While real-time kinematics geted by 5G rest on achieving ultra-low latency of 1 ms over
global navigation satellite system (RTK GNSS) can provide the air with five-nines reliability. The main techniques intro-
highly accurate localization under the conditions of good duced to achieve these performance targets are mini-slots and
satellite visibility, many of the automation use cases are grant-free faster channel access for low latency and multiply
indoors, where that is not possible. The current approach connected links involving multiple access points, carriers and
to solving the localization problem is to rely on specialized packet duplication for reliability.
systems based on ultra-wideband (UWB) or Bluetooth Low Even 1ms latency over the air is insufficient for many use
Energy (BLW) and requires additional access points and cases, for example in replacing traditional industrial wire-
devices to be installed [33]. A separate system for localization line connectivity solutions such as Sercos or EtherCAT. For
in addition to the communication system incurs additional such use cases, substantially lower radio latencies of the
infrastructure expense and ongoing maintenance costs that order of 100 µs at gigabit-per-second data rates are required.
can be avoided if the communication system is also able Furthermore, the actual requirement on reliability is based on
to perform accurate localization. As a result, 5G includes the actual downtime of the equipment, the onset of which is

57070 VOLUME 8, 2020

H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

triggered by multiple consecutive packet losses. The target for

such a reliability measure can be of the order of nine-nines for
some industrial automation use cases.
6G will be designed to achieve these extreme requirements
cost-effectively. Recent measurements show that, contrary to
conventional wisdom, the blocking of signals in mmWave
is not that severe in factory floors. By relying on wider
bandwidth available in mmWave spectrum, it will possible
to achieve such extreme low latencies at the desired high
data rates. Reliability can be enhanced through simultane-
ous transmission through multiple paths involving multiple
wireless hops. Cooperative relaying through device-to-device
connections can be used to create separate paths from FIGURE 7. 6G architecture with sub-networks, multi-path connectivity
the network to the device. Predictive beam management and distributed function placement extending to the device.
using AI/ML prediction techniques [37] can also substan-
tially reduce uncertainties in link quality. In 4G, low-power uninterrupted despite poor or no connectivity to the wider
wireless access was introduced through Narrowband IoT network. Multiple path connectivity employing infrastruc-
(NB-IoT) technology. Another extreme form of networking ture and opportunistic device-to-device connections will be
that we can expect in 6G is going from low-power to zero- required for the ultra-reliability, potentially leading to truly
energy devices for IoT. Wireless zero-energy devices are cell-less architectures. Integration of these sub-networks to
well known as passive radio frequency identification (RFID) 6G as one holistic architecture has some advantages:
tags, with low-cost active RFIDs supporting sensors typically • The 6G sub-network will ensure high data rates, extreme
having a battery life of three to five years, but these are low latency, reliability and resilience
generally restricted to very short ranges. There are several • 6G security and resilience features are enforced to the
use cases where sustainable sensor devices and extremely lowest level of devices in the sub-network
long operation times are desired in a wider-area context. For • 6G service execution can dynamically be split between
construction inspection of bridges or tunnels, as an example, execution in the edge cloud or in the device that is part
it is desirable to have wireless sensor devices that can be of the sub-network.
fully embedded in the construction and that can operate on Time-sensitive communications (TSC), through integration
the order of 100 years without human intervention. Potential of the 5G network and TSN, with the 5G network acting
solutions space will likely include a mixture of low-power as a TSN bridge, will evolve to 6G to provide native TSN,
communication, extreme low-idle current, energy harvesting including over wider areas with mobility.
(potentially from the communication network), and reachable
energy storage. 2) HYPER-SPECIALIZED SLICING
Beyond the extension of the traditional connectivity architec-
ture into a variety of sub-network and multiply connected
E. 6G NETWORK ARCHITECTURE CONCEPTS scenarios, we expect further advances in slicing and virtu-
1) SUB-NETWORKS alization. Slices can become highly specialized, potentially
The cellular network architecture in previous generations has with separate software stacks in each slice for different func-
been designed primarily for extending the voice and data tional treatment of the flows, as illustrated in Figure 8. The
internet to individual, mobile end points. 5G is the first current trend in virtualization of the higher layers of the RAN
system designed to make inroads into the industrial envi- will lead to further disaggregation of RAN functions into
ronment, meeting the challenging requirements through new modular micro-services that can be flexibly composed into
architectural evolutions such as supporting time-sensitive net- slice-specific RAN implementations. For example, one can
working (TSN) bridge functionality. To build on the road envision slice specializations for a video service slice incor-
paved by 5G, and become truly entrenched in the indus- porating specific video optimization micro-services included
trial environment and replace wired connectivity everywhere, in that RAN slice but not necessary in other slices. Similarly,
6G should provide deterministic wire-grade reliability for low-throughput IoT slices can incorporate functions allowing
a variety of connectivity scenarios (Figure 7), from static, connectionless access [38] while other slices are based on
isolated devices, to inter-related locally interacting devices, a traditional access approach. In addition, we can expect
to rapidly moving swarms of robots and drones that need flexible slice-specific function placement in gateway devices,
to inter-connect but also connect directly to the network relays, cell sites, far edges, edges and regional clouds across a
when separated from the swarm. To ensure both high time variety of different hardware platforms according to the needs
and spatial domain reliability and determinism, we see the of the slice. New innovations in service management and
need for semi-autonomous 6G sub-networks, where at least orchestration are needed to create and manage such highly
the most critical services in the sub-network will continue specialized slices.

VOLUME 8, 2020 57071

 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

between sub-networks but also between the network and

sub-networks. It will be crucial for the sub-networks to act
as an independent network, empowered as an authorization
authority and responsible for sub-network asset management.
Due to the dynamic behavior of devices joining/leaving the
sub-network, maintaining sub-network privacy and potential
anonymity will be a challenge to be solved in the 6G network
architecture. When the physical and biological worlds are
mirrored in the digital world with high precision and when
new mixed-reality worlds combining digital representations
FIGURE 8. Slicing with slice-specific software optimization.
of real and virtual objects are created, privacy solutions of
today are unlikely to be sufficient. Although multi-modal
3) RAN-CORE CONVERGENCE sensing will capture nearly everything in their surroundings,
In 5G, the base station has been compartmentalized into users will want to restrict what others are allowed to expe-
the distributed unit (DU) and centralized unit (CU). The rience in the content that they share. Users must be able to
DU includes the lower layers of the user and control plane set a rich set of preferences in a simple fashion on what
protocol stack, namely the physical Layer 1 and real-time they wish to share, and data processing should automatically
Layer 2, while the CU includes the non-real-time Layer 2 and ensure that. A variety of new signal processing techniques are
Layer 3 functions. The CU is further split into the control emerging [39] to address security in the mixed-reality world
plane and user plane, with a well-defined interface between of tomorrow that will become an integral part of the 6G net-
the two. The CU is typically implemented as a virtualized work. Physical layer security mechanisms [40], [41] typically
function in the edge or metro cloud and can serve multiple depend on the uniqueness of the wireless channel to establish
DUs. The 5G core functions, on the other hand, are becoming authentication, confidentiality and key exchange, and may
more decentralized as the amount traffic through the core become more mature in the timeframe of 6G, addressing new
increases substantially. The various core functions are also issues such as jamming. Trust in the network is critical for the
being virtualized and implemented in regional or metro success of 6G.
clouds, or even edge clouds for low-latency services. With
increasing centralization of the higher-layer RAN functions VI. 6G AS AN OPEN PLATFORM FOR THE CREATION OF
and the distribution of the core functions, simplification HIGHLY SPECIALIZED SOLUTIONS
can be achieved by combining some of the RAN and core One important factor for the huge success of mobile com-
functions into single entities. Thus, in the timeframe of 6G, munication has been global open standards. Until recently,
we expect a reduced set of functional blocks implementing mobile communication services have been dominated by
a combination of 5G RAN and core, resulting in a coreless human-centric services like voice, short messaging service
RAN, especially on the user plane. (SMS) and best-effort broadband data. Tremendous volumes
have been achieved, with over a billion smartphones in
F. NEW SECURITY, PRIVACY AND TRUST PARADIGMS use and tens of millions of mobile network base stations
Wireline-grade reliability also implies that the network deployed. During the last five years, mobile communication
must be designed with new security and privacy measures. services have expanded to now include IoT. First, low-power
Jamming in industrial networks is a new threat that the wide-area (LPWA) connectivity solutions such as NB-IoT
networks will have to be protected against. Attackers could and Cat-M were standardized. Next, Cellular V2X connec-
attempt to jam networks from outside the industrial facil- tivity for cars followed. UAV connectivity for drones has
ity, and so physical security will be insufficient. In the been specified. Most recently, in 5G, URLLC and TSC for
future, jamming may also take the form of simply delaying real-time control are being standardized. LPWA and C-V2X
packet delivery by creating interference only sporadically. connectivity devices have the potential to achieve volumes
This can seriously impact industrial operations relying on in the order of billions. As the applicability of the cellu-
time-sensitive networks. 6G networks will be designed to pro- lar network expands into new IIoT and future home and
tect against such new threats. The definition of sub-networks enterprise environments, we foresee the need for much more
in networks requires a change to the authorization strategy. specialized connectivity solutions, optimized for the specific
It is no longer authorization by the network, but by the requirements. The natural consequences are that there will be
sub-network. When analyzing a body area network (BAN), significantly smaller volumes for each specialized connectiv-
the assets in the sub-network belong to the sub-network, and ity solution. To efficiently support this long-tailed distribution
therefore the authorization and asset management must be of new wireless connectivity solutions, we foresee the need to
handled in that trust boundary. The network will connect sub- specify and introduce mobile network specifications as a plat-
networks, and a second level of authorization might be needed form with a few core capabilities at the lower layers. This will
at network level. Different sub-networks might belong to be needed for the different scenarios, which will then be uti-
mutual untrusted entities, which calls for clear separation lized by several different interest groups to specify their own

57072 VOLUME 8, 2020

H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

higher-layer specifications to achieve a complete connectivity coupled with the use of wider bandwidth signals at high-band
solution for their own use case family. The latter may in some spectrum, especially indoors, new opportunities to utilize this
cases be realized through joint software development within for localization and sensing will encourage a 6G design that is
the interest group, or open source software. This is illustrated not only optimized for communication but also for perception
in Figure 9. A major advantage of this platform approach is and understanding of the physical world and people’s needs,
the open interfaces from the platform to the specialized layers thus augmenting human existence in the most intuitive way.
of connectivity. We believe that the lower-layer communica- In this paper, we identified the following key technology
tion protocols typically implemented in silicon still need to be transformations as having the highest potential to be defining
specified by a global standardization body to ensure economy for the 6G system: (i) AI/ML-driven air interface design and
of scale and coexistence among connectivity services. optimization; (ii) expansion into new spectrum bands and new
cognitive spectrum sharing methods; (iii) the integration of
localization and sensing capabilities into system definition;
(iv) the achievement of extreme performance requirements
on latency and reliability; (v) new network architecture
paradigms involving sub-networks and RAN-Core conver-
gence; and (vi) new security and privacy schemes. Finally,
the expansion into many varied use cases calls for a shift to a
platform approach to the network, with decoupling of the air
FIGURE 9. Illustration of 6G as a solution platform model.
interface from networking.

The current trend toward open RANs with multiple ven- ACKNOWLEDGMENT
dors providing different pieces of the network will be further The authors would like to thank a number of colleagues at
facilitated by the above network as a platform approach in 6G. Nokia Bell Labs with whom they had numerous discussions
Apart from the radio and some of the processing-intensive about the future of wireless. In particular, they would also like
functions that are best realized in custom hardware, the rest to thank H. Holma, P. Merz, M. Uusitalo, P. Vetter, T. Wild,
will be software functions running on any commercial com- and V. Ziegler for their comments on an early draft of this
pute hardware, and the interface specifications will also article. They would also like to thank A. Hu and R. Adeogun
become open. The intelligent control layers of the mobile for helping with the preparation of the manuscript.
network can thus be optimized to suit the connectivity needs
of the specific use case family and can come from any vendor. REFERENCES
Note that this is also in line with the notion of the next [1] B. Aazhang et al, ‘‘Key drivers and research challenges for 6G ubiqui-
tous intelligence (white paper),’’ 6G Flagship, Univ. Oulu, Oulu, Finland,
generation of slicing that was described earlier, where each White Paper, Sep. 2019.
slice can have its own specialized function. [2] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, ‘‘From a human-centric
Mobile networks and terminals as a platform as such is perspective: What might 6G be?’’ 2019, arXiv:1906.00741. [Online].
Available: http://arxiv.org/abs/1906.00741
not a 6G technology, but rather a new way of efficiently and [3] M. N. Patwary, S. J. Nawaz, M. A. Rahman, S. K. Sharma, M. M. Rashid,
rapidly specifying solutions for the long-tailed distribution of and S. J. Barnes, ‘‘The potential short- and long-term disruptions and
specialized industrial and other indoor use cases. transformative impacts of 5G and beyond wireless networks: Lessons
learnt from the development of a 5G testbed environment,’’ IEEE Access,
vol. 8, pp. 11352–11379, 2020.
VII. SUMMARY [4] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y.-J.-A. Zhang, ‘‘The roadmap
to 6G: AI empowered wireless networks,’’ IEEE Commun. Mag., vol. 57,
The tradition of deploying a new cellular generation approx- no. 8, pp. 84–90, Aug. 2019.
imately every decade will continue into the future, with [5] G. Berardinelli, N. H. Mahmood, I. Rodriguez, and P. Mogensen, ‘‘Beyond
6G becoming a reality in the 2030s. 6G will be optimized 5G wireless IRT for industry 4.0: Design principles and spectrum aspects,’’
in Proc. IEEE Globecom Workshops (GC Wkshps), Dec. 2018, pp. 1–6.
and cost-reduced for the new use cases introduced in 5G, [6] Z. Zhang, Y. Xiao, Z. Ma, M. Xiao, Z. Ding, X. Lei, G. K. Karagiannidis,
driving their adoption at scale. At the same time, it will and P. Fan, ‘‘6G wireless networks: Vision, requirements, architecture, and
enable new use cases that we cannot yet imagine or describe key technologies,’’ IEEE Veh. Technol. Mag., vol. 14, no. 3, pp. 28–41,
Sep. 2019.
in detail. The expansion of mobile cellular to verticals that [7] P. Yang, Y. Xiao, M. Xiao, and S. Li, ‘‘6G wireless communications: Vision
began with the introduction of low-cost IoT technologies and potential techniques,’’ IEEE Netw., vol. 33, no. 4, pp. 70–75, Jul. 2019.
in 4G and ultra-reliable low latency IIoT in 5G will con- [8] T. S. Rappaport, Y. Xing, O. Kanhere, S. Ju, A. Madanayake, S. Mandal,
A. Alkhateeb, and G. C. Trichopoulos, ‘‘Wireless communications and
tinue, becoming both broader and deeper in 6G. The rapid applications above 100 GHz: Opportunities and challenges for 6G and
advance of AI/ML technology and its effectiveness in solving beyond,’’ IEEE Access, vol. 7, pp. 78729–78757, 2019.
problems in several domains points toward a 6G system [9] K. David and H. Berndt, ‘‘6G vision and requirements: Is there any need
for beyond 5G?’’ IEEE Veh. Technol. Mag., vol. 13, no. 3, pp. 72–80,
that will fundamentally exploit these new capabilities to Sep. 2018.
improve performance by better adapting to the operational [10] E. C. Strinati, S. Barbarossa, J. L. Gonzalez-Jimenez, D. Ktenas,
environment. The inexorable demand for higher capacity N. Cassiau, L. Maret, and C. Dehos, ‘‘6G: The next frontier: From holo-
graphic messaging to artificial intelligence using subterahertz and visible
and peak rates points toward technologies that will exploit light communication,’’ IEEE Veh. Technol. Mag., vol. 14, no. 3, pp. 42–50,
ever-higher bands. As the density of infrastructure increases, Sep. 2019.

VOLUME 8, 2020 57073

 H. Viswanathan, P. E. Mogensen: Communications in the 6G Era

[11] D. Wang, M. Wang, P. Zhu, J. Li, J. Wang, and X. You, ‘‘Performance of [32] S. Bender, P. Seiler, B. Klein, M. Dorpinghaus, D. Plettemeier, and
network-assisted full-duplex for cell-free massive MIMO,’’ IEEE Trans. G. Fettweis, ‘‘Pathways towards Tb/s wireless,’’ in Proc. IEEE 17th Int.
Commun., to be published. Conf. Ubiquitous Wireless Broadband (ICUWB), Sep. 2017, pp. 1–4.
[12] A. Fotouhi, H. Qiang, M. Ding, M. Hassan, L. G. Giordano, [33] J. Rantala, ‘‘High accuracy indoor positioning—Technology solution and
A. Garcia-Rodriguez, and J. Yuan, ‘‘Survey on UAV cellular com- business implications,’’ in Proc. 3rd Invitational Workshop Opportunistic
munications: Practical aspects, standardization advancements, regulation, RF Localization Next Gener. Wireless Devices, May 2012, pp. 1–10.
and security challenges,’’ IEEE Commun. Surveys Tuts., vol. 21, no. 4, [34] M. Arnold, J. Hoydis, and S. Ten Brink, ‘‘Novel massive MIMO channel
pp. 3417–3442, 4th Quart., 2019. sounding data applied to deep learning-based indoor positioning,’’ in Proc.
[13] A. Mohammed, A. Mehmood, F.-N. Pavlidou, and M. Mohorcic, ‘‘The role Int. ITG Conf. Syst., Commun. Coding (SSC), Feb. 2019, pp. 1–6.
of high-altitude platforms (HAPs) in the global wireless connectivity,’’ [35] C. Sturm and W. Wiesbeck, ‘‘Waveform design and signal processing
Proc. IEEE, vol. 99, no. 11, pp. 1939–1953, Nov. 2011. aspects for fusion of wireless communications and radar sensing,’’ Proc.
[14] B. Di, H. Zhang, L. Song, Y. Li, and G. Y. Li, ‘‘Ultra-dense LEO: Integrat- IEEE, vol. 99, no. 7, pp. 1236–1259, Jul. 2011.
ing terrestrial-satellite networks into 5G and beyond for data offloading,’’ [36] P. H. Siegel, ‘‘Terahertz technology in biology and medicine,’’ in IEEE
IEEE Trans. Wireless Commun., vol. 18, no. 1, pp. 47–62, Jan. 2019. MTT-S Int. Microw. Symp. Dig., vol. 3, Jun. 2004, pp. 1575–1578.
[15] S.-T. Cheng, C.-Y. Wang, and M.-H. Tao, ‘‘Quantum communication for [37] A. Kaya and H. Viswanathan, ‘‘Machine learning based predictive beam
wireless wide-area networks,’’ IEEE J. Sel. Areas Commun., vol. 23, no. 7, management for 5G mmWave systems,’’ IEEE Trans. Wireless Commun.,
pp. 1424–1432, Jul. 2005. submitted for publication.
[16] B.-H. Koo, C. Lee, H. B. Yilmaz, N. Farsad, A. Eckford, and C.-B. Chae, [38] C. Kahn and H. Viswanathan, ‘‘Connectionless access for mobile cellular
‘‘Molecular MIMO: From theory to prototype,’’ IEEE J. Sel. Areas Com- networks,’’ IEEE Commun. Mag., vol. 53, no. 9, pp. 26–31, Sep. 2015.
mun., vol. 34, no. 3, pp. 600–614, Mar. 2016. [39] J. A. de Guzman, K. Thilakarathna, and A. Seneviratne, ‘‘A first look
[17] A. K. Gupta and J. G. Andrews, ‘‘Comments on coverage analysis of into privacy leakage in 3D mixed reality data,’’ in Proc. 24th Eur. Symp.
multiuser visible light communication networks,’’ IEEE Trans. Wireless Res. Comput. Secur. Comput. Secur. (ESORICS), Luxembourg, Europe,
Commun., vol. 18, no. 9, pp. 4605–4606, Sep. 2019. Sep. 2019, pp. 149–169.
[18] J. Wang, C. Jiang, H. Zhang, Y. Ren, K.-C. Chen, and L. Hanzo, [40] A. Mukherjee, S. A. A. Fakoorian, J. Huang, and A. L. Swindlehurst,
‘‘Thirty years of machine learning: The road to Pareto-optimal next- ‘‘Principles of physical layer security in multiuser wireless networks:
generation wireless networks,’’ IEEE Commun. Surveys Tuts., early access, A survey,’’ IEEE Commun. Surveys Tuts., vol. 16, no. 3, pp. 1550–1573,
Jan. 2020. 3rd Quart., 2014.
[19] S. J. Nawaz, S. K. Sharma, S. Wyne, M. N. Patwary, and M. Asaduzzaman, [41] W. Trappe, ‘‘The challenges facing physical layer security,’’ IEEE Com-
‘‘Quantum machine learning for 6G communication networks: State-of- mun. Mag., vol. 53, no. 6, pp. 16–20, Jun. 2015.
the-art and vision for the future,’’ IEEE Access, vol. 7, pp. 46317–46350,
2019.
[20] S. Dorner, S. Cammerer, J. Hoydis, and S. T. Brink, ‘‘Deep learning based
communication over the air,’’ IEEE J. Sel. Topics Signal Process., vol. 12,
no. 1, pp. 132–143, Feb. 2018. HARISH VISWANATHAN (Fellow, IEEE)
[21] T. J. Oshea, T. Roy, N. West, and B. C. Hilburn, ‘‘Physical layer commu- received the B.Tech. degree from the Depart-
nications system design over-the-air using adversarial networks,’’ in Proc. ment of Electrical Engineering, Indian Institute of
26th Eur. Signal Process. Conf. (EUSIPCO), Sep. 2018, pp. 529–532.
Technology Madras, Chennai, India, and the M.S.
[22] F. Zhou, G. Lu, M. Wen, Y.-C. Liang, Z. Chu, and Y. Wang, ‘‘Dynamic
and Ph.D. degrees from the School of Electrical
spectrum management via machine learning: State of the art, taxon-
omy, challenges, and open research issues,’’ IEEE Netw., vol. 33, no. 4, Engineering, Cornell University, Ithaca, NY, USA.
pp. 54–62, Jul. 2019. Since joining Bell Labs in October 1997, he has
[23] P. Tilghman, ‘‘Will rule the airwaves: A DARPA grand challenge seeks worked extensively on wireless research, ranging
autonomous radios to manage the wireless spectrum,’’ IEEE Spectr., from physical layer to network architecture and
vol. 56, no. 6, pp. 28–33, Jun. 2019. protocols, including multiple antenna technology
[24] D. Korpi, P. Yli-Opas, M. R. Jaramillo, and M. Uusitalo, ‘‘Visual detection- for cellular wireless networks, multihop relays, network optimization, net-
based blockage prediction for beyond 5G wireless systems,’’ in Proc. 6G work architecture, and the IoT communications. From 2007 to 2015, he was
Wireless Summit, Levi, Finland, Mar. 2020. in the Corporation CTO organization, where as a CTO Partner, he advised
[25] M. D. Renzo, M. Debbah, D.-T. Phan-Huy, A. Zappone, M.-S. Alouini, the Corporate CTO on technology strategy through in-depth analysis of
C. Yuen, V. Sciancalepore, G. C. Alexandropoulos, J. Hoydis, H. Gacanin, emerging technology and market needs. He is currently the Head of Radio
J. D. Rosny, A. Bounceur, G. Lerosey, and M. Fink, ‘‘Smart radio envi- Systems Research Group, Nokia Bell Labs. He has published extensively
ronments empowered by reconfigurable AI meta-surfaces: An idea whose
with over 100 publications. He is also a Bell Labs Fellow.
time has come,’’ EURASIP J. Wireless Commun. Netw., vol. 2019, no. 1,
p. 129, Dec. 2019.
[26] A. Ghosh, A. Maeder, M. Baker, and D. Chandramouli, ‘‘5G evolution:
A view on 5G cellular technology beyond 3GPP release 15,’’ IEEE Access,
vol. 7, pp. 127639–127651, 2019.
[27] M. Cudak, T. Kovarik, T. A. Thomas, A. Ghosh, Y. Kishiyama, and
T. Nakamura, ‘‘Experimental mm wave 5G cellular system,’’ in Proc. IEEE PREBEN E. MOGENSEN (Member, IEEE)
Globecom Workshops (GC Wkshps), Dec. 2014, pp. 377–381. received the M.Sc. and Ph.D. degrees from
[28] J. Du, D. Chizhik, R. A. Valenzuela, R. Feick, G. Castro, M. Rodriguez, Aalborg University, Denmark, in 1988 and 1996,
T. Chen, M. Kohli, and G. Zussman, ‘‘Directional measurements respectively. He has been with Aalborg University,
in urban street canyons from macro rooftop sites at 28 GHz for since graduating in 1988. Since 1995, he has
90% outdoor coverage,’’ 2019, arXiv:1908.00512. [Online]. Available: also been a part-time Associate with Nokia in
https://arxiv.org/abs/1908.00512 various research positions and has made contribu-
[29] S. Shahramian, Y. Baeyens, N. Kaneda, and Y.-K. Chen, ‘‘A 70–100 GHz tions ranging from 2G to 5G cellular technologies.
direct-conversion transmitter and receiver phased array chipset demon- In 2000, he became a Full Professor at Aalborg
strating 10 Gb/s wireless link,’’ IEEE J. Solid-State Circuits, vol. 48, no. 5,
University, where he is currently leading the Wire-
pp. 1113–1125, May 2013.
[30] N. Ranjkesh, A. Taeb, S. Gigoyan, M. Basha, and S. Safavi-Naeini,
less Communication Networks Section, Department of Electronic Systems.
‘‘Millimeter-wave silicon-on-glass integrated tapered antenna,’’ IEEE He is also a Principal Scientist with the Standardization Research Lab, Nokia
Antennas Wireless Propag. Lett., vol. 13, pp. 1425–1428, 2014. Bell Labs. He has coauthored over 400 articles in various domains of wireless
[31] L. Iotti, G. LaCaille, and A. M. Niknejad, ‘‘A 12 mW 70-to-100 GHz communication, and his Google Scholar H-index is 63. His current research
mixer-first receiver front-end for mm-wave massive-MIMO arrays in interest includes industrial use cases for 5G, 5G evolution, and 6G. He is also
28 nm CMOS,’’ in Proc. IEEE Int. Solid-State Circuits Conf. (ISSCC), a Bell Labs Fellow.
Feb. 2018, pp. 414–416.

57074 VOLUME 8, 2020