See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/331396903

A Vision of 6G Wireless Systems: Applications, Trends, Technologies, and Open

Research Problems

Preprint · February 2019

CITATIONS READS

0 2,607

3 authors:

Walid Saad Mehdi Bennis

Virginia Polytechnic Institute and State University University of Oulu
507 PUBLICATIONS   11,690 CITATIONS    366 PUBLICATIONS   8,193 CITATIONS   

SEE PROFILE SEE PROFILE

Mingzhe Chen
The Chinese University of Hong Kong (Shenzhen) and Princeton University
43 PUBLICATIONS   703 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Beyond 5G Networks: Context-Awareness, mmW communications, and Caching View project

mmWave V2X communications View project

All content following this page was uploaded by Mingzhe Chen on 05 March 2019.

The user has requested enhancement of the downloaded file.

 A Vision of 6G Wireless Systems: Applications, Trends,
Technologies, and Open Research Problems
Walid Saad∗ , Mehdi Bennis† , and Mingzhe Chen‡,∗
∗ Wireless@VT, Bradley Department of Electrical and Computer Engineering, Virginia Tech, Blacksburg, VA, USA, Email: [email protected].
† CWC - Centre for Wireless Communications, University of Oulu, Finland, Email: [email protected].
‡ Beijing Key Laboratory of Network System Architecture and Convergence,
Beijing University of Posts and Telecommunications, Beijing, China 100876, Email: [email protected].

Abstract—The ongoing deployment of 5G cellular systems is being marketed will readily support basic IoE and URLLC
continuously exposing the inherent limitations of this system, services (e.g., factory automation), it is debatable whether
compared to its original premise as an enabler for Internet they can deliver the tomorrow’s smart city IoE applications.
arXiv:1902.10265v1 [cs.IT] 26 Feb 2019

of Everything applications. These 5G drawbacks are currently

spurring worldwide activities focused on defining the next- Moreover, even though 5G will eventually support fixed-
generation 6G wireless system that can truly integrate far- access at mmWave frequencies, it is more likely that early
reaching applications ranging from autonomous systems to ex- 5G roll-outs will be centered around sub-6 GHz, especially
tended reality and haptics. Despite recent 6G initiatives1 , the for supporting mobility.
fundamental architectural and performance components of the Meanwhile, an unprecedented proliferation of new IoE
system remain largely undefined. In this paper, we present a
holistic, forward-looking vision that defines the tenets of a 6G services is ongoing. Examples range from eXtended reality
system. We opine that 6G will not be a mere exploration of (XR) services (encompassing augmented, mixed, and virtual
more spectrum at high-frequency bands, but it will rather be a reality (AR/MR/VR)) to telemedicine, haptics, flying vehi-
convergence of upcoming technological trends driven by exciting, cles, brain-computer interfaces, and connected autonomous
underlying services. In this regard, we first identify the primary systems. These applications will disrupt the original 5G goal
drivers of 6G systems, in terms of applications and accompanying
technological trends. Then, we propose a new set of service of supporting short-packet, sensing-based URLLC services. To
classes and expose their target 6G performance requirements. successfully operate IoE services such as XR and connected
We then identify the enabling technologies for the introduced autonomous systems, a wireless system must simultaneously
6G services and outline a comprehensive research agenda that deliver high reliability, low latency, and high data rates, for
leverages those technologies. We conclude by providing concrete heterogeneous devices, across uplink and downlink. Emerging
recommendations for the roadmap toward 6G. Ultimately, the
intent of this article is to serve as a basis for stimulating more IoE services will also require an end-to-end co-design of
out-of-the-box research around 6G. communication, control, and computing functionalities, which
to date has been largely overlooked. To cater for this new
I. I NTRODUCTION breed of services, unique challenges must be addressed ranging
To date, the wireless network evolution was primarily driven from characterizing the fundamental rate-reliability-latency
by an incessant need for higher data rates, which mandated tradeoffs governing their performance to exploiting frequen-
a continuous 1000x increase in the network capacity. While cies beyond sub-6 GHz and transforming wireless systems
this demand for wireless capacity will continue to grow, into a self-sustaining, intelligent network fabric which flex-
the emergence of the Internet of Everything (IoE) system, ibly provisions and orchestrates communication-computing-
connecting millions of people and billions of machines, is control-localization-sensing resources tailored to the requisite
yielding a radical paradigm shift from the rate-centric en- IoE scenario.
hanced mobile broadband (eMBB) services of yesteryears To overcome these challenges and catalyze the deployment
towards ultra-reliable, low latency communications (URLLC). of new IoE services, a disruptive sixth generation (6G) wire-
less system, whose design is inherently tailored to the perfor-
Although the fifth generation (5G) cellular system [1]
mance requirements of the aforementioned IoE applications
was marketed as the key IoE enabler, through concerted 5G
and their accompanying technological trends, is needed. The
standardization efforts that led to the first 5G new radio
drivers of 6G will be a confluence of past trends (e.g., densi-
(5G NR) milestone (for non-standalone 5G) and subsequent
fication, higher rates, and massive antennas) and of emerging
3GPP releases, the initial premise of 5G – as a true carrier
trends that include new services and the recent revolution in
of IoE services – is yet to be realized. One can argue
wireless devices (e.g., smart wearables, implants, XR devices,
that the evolutionary part of 5G (i.e., supporting rate-hungry
etc.), artificial intelligence (AI), computing, sensing, and 3D
eMBB services) has gained significant momentum, however,
environmental mapping.
the promised revolutionary outlook of 5G – a system operating
The main contribution of this article is a bold, forward-
almost exclusively at millimeter wave (mmWave) frequencies
looking vision of 6G systems (see Fig. 1) that identifies
and enabling heterogeneous IoE services – has thus far re-
the applications, trends, performance metrics, and disruptive
mained a mirage. Although the 5G systems that are currently
technologies, that will drive the 6G revolution. The proposed
1 One example is the 6Genesis project in Finland (see vision will then delineate new 6G services and provide a
https://www.oulu.fi/6gflagship/). concrete research roadmap and recommendations to facilitate
 6G: Driving Applications

Multisensory Connected robotics Wireless Brain- Blockchain and

XR and Computer Distributed
Applications Autonomous Interactions Ledger
Systems Technologies

6G: Driving Trends

More Bits, From Spatial Emergence of Massive From SON Convergence of End of the
Spectrum, to Volumetric Smart Availability to Self- Communication, Smartphone
Reliability Spectral and Surfaces and of Small Sustaining Sensing, Control, Era
Energy Environments Data Networks Localization, and
Efficiency Computing

bps/Hz/
Joules/m3

6G: Enabling Technologies

Above 6 GHz Transceivers Communicati Edge AI Integrated Energy Beyond 6G
for 6G with on with Large Terrestrial, Transfer and
Integrated Intelligent Airborne, and Harvesting
Frequency Surfaces Satellite
Mobile Bands Networks
BS
mmWave band

and THz

Fig. 1. 6G Vision: Applications, Trends, and Technologies.

the leap from current 5G systems towards 6G. and limits must be factored into the engineering process (com-
puting, processing, etc.). To do so, a new concept of quality-
II. 6G D RIVING A PPLICATIONS , M ETRICS , AND N EW of-physical-experience (QoPE) measure is needed to merge
S ERVICE C LASSES physical factors from the human user itself with classical QoS
Every new cellular system generation is driven by innovative (e.g., latency and rate) and QoE (e.g., mean-opinion score)
applications. 6G is no exception: It will be borne out of inputs. Some factors that affect QoPE include brain cognition,
an unparalleled emergence of exciting new applications and body physiology, and gestures. As an example, in [2], we have
technological trends that will shape its performance targets shown that the human brain may not be able to distinguish
while radically redefining standard 5G services. In this section, between different latency measures, when operating in the
we first introduce the main applications that motivate 6G URLLC regime. Meanwhile, in [3], we showed that visual and
deployment and, then, discuss ensuing technological trends, haptic perceptions are key for maximizing wireless resource
target performance metrics, and new service requirements. utilization. Concisely, the requirements of XR services are
a blend of traditional URLLC and eMBB with incorporated
A. Driving Applications behind 6G and their Requirements perceptual factors that 6G must support.
While traditional applications, such as live multimedia 2) Connected Robotics and Autonomous Systems
streaming, will remain central to 6G, the key determinants of (CRAS): A primary driver behind 6G systems is the immi-
the system performance will be four new application domains: nent deployment of CRAS including drone-delivery systems,
1) Multisensory XR Applications: XR will yield many autonomous cars, autonomous drone swarms, vehicle platoons,
killer applications for 6G across the AR/MR/VR spectrum. and autonomous robotics. The introduction of CRAS over the
Upcoming 5G systems still fall short of providing a full immer- cellular domain is not a simple case of “yet another short
sive XR experience capturing all sensory inputs due to their packet uplink IoE service”. Instead, CRAS mandate control
inability to deliver very low latencies for data-rate intensive system-driven latency requirements as well as the potential
XR applications. A truly immersive AR/MR/VR experience need for eMBB transmissions of high definition (HD) maps.
requires a joint design integrating not only engineering (wire- The notion of QoPE applies once again for CRAS; however,
less, computing, storage) requirements but also perceptual the physical environment is now a control system, potentially
requirements stemming from human senses, cognition, and augmented with AI. CRAS are perhaps a prime use case
physiology. Minimal and maximal perceptual requirements that requires stringent requirements across the rate-reliability-
latency spectrum; a balance that is not yet available in 5G. evolution that started from 2G (bps) to 3G (bps/Hz), then
3) Wireless Brain-Computer Interactions (BCI): Be- 4G (bps/Hz/m2 ) to 5G (bps/Hz/m2 /Joules).
yond XR, tailoring wireless systems to their human user is • Trend 3 – Emergence of Smart Surfaces and Envi-
mandatory to support services with direct BCI. Traditionally, ronments: Current and past cellular systems used base
BCI applications were limited to healthcare scenarios in stations (of different sizes and forms) for transmission.
which humans can control prosthetic limbs or neighboring We are currently witnessing a revolution in electromag-
computing devices using brain implants. However, the recent netically active surfaces (e.g., using metamaterials) that
advent of wireless brain-computer interfaces and implants will include man-made structures such as walls, roads, and
revolutionize this field and introduce new use-case scenar- even entire buildings, as exemplified by the Berkeley
ios that require 6G connectivity. Such scenarios range from ewallpaper project2 . The use of such smart large intel-
enabling brain-controlled movie input to fully-fledged multi- ligent surfaces and environments for wireless communi-
brain-controlled cinema [4]. Using wireless BCI technologies, cations will drive the 6G architectural evolution.
instead of smartphones, people will interact with their envi- • Trend 4 – Massive Availability of Small Data: The
ronment and other people using discrete devices, some worn, data revolution will continue in the near future and shift
some implanted, and some embedded in the world around from centralized, big data, towards massive, distributed
them. This will allow individuals to control their environments “small” data. 6G systems must harness both big and small
through gestures and communicate with loved ones through datasets across their infrastructure to enhance network
haptic messages. Such empathic and haptic communications, functions and provide new services. This trend motivates
coupled with related ideas such as affective computing in new machine learning and data analytics techniques that
which emotion-driven devices can match their functions to go beyond classical big data.
their user’s mood, will constitute important 6G use cases. • Trend 5 – From Self-Organizing Networks (SON) to
Wireless BCI services will require fundamentally different Self-Sustaining Networks: SON has only been scarcely
performance metrics compared to what 5G delivers. Similar to integrated into 4G/5G networks due to a lack of real-
XR, wireless BCI services need high rates, ultra low latency, world need. However, CRAS and DLT technologies mo-
and high reliability. However, they are much more sensitive tivate an immediate need for intelligent SON to manage
than XR to physical perceptions and will necessitate QoPE network operations, resources, and optimization. 6G will
guarantees. require a paradigm shift from classical SON, whereby
4) Blockchain and Distributed Ledger Technologies the network merely adapts its functions to specific envi-
(DLT): Blockchains and DLT will be one of the most disrup- ronment states, into a self-sustaining network (SSN) that
tive IoE technologies. Blockchain and DLT applications can be can maintain its key performance indicators (KPIs), in
viewed as the next-generation of distributed sensing services perpetuity, under highly dynamic and complex environ-
whose need for connectivity will require a synergistic mix of ments stemming from the rich 6G application domains.
URLLC and massive machine type communications (mMTC) SSNs must be able to not only adapt their functions
to guarantee low-latency, reliable connectivity, and scalability. but to also sustain their resource usage and management
(e.g., by harvesting energy and exploiting spectrum)
B. 6G: Driving Trends and Performance Metrics to autonomously maintain high, long-term KPIs. SSN
functions must leverage the recent revolution in AI tech-
The applications of Section II-A lead to new system-wide nologies to create AI-powered 6G SSNs.
trends that will set the goals for 6G: • Trend 6 – Convergence of Communications, Comput-
• Trend 1 – More Bits, More spectrum, More Reliabil- ing, Control, Localization, and Sensing (3CLS): The
ity: Most of the driving applications of 6G require higher past five generations of cellular systems had one exclusive
bit rates than 5G. To cater for applications such as XR function: wireless communications. However, the con-
and BCI, 6G must deliver yet another 1000x increase vergence of various technologies requires 6G to disrupt
in data rates yielding a target of around 1 Terabit/second. this premise by providing multiple functions that include
This motivates a need for more spectrum resources, hence communications, computing, control, localization, and
motivating further exploration of frequencies beyond sub- sensing. We envision 6G as a multi-purpose system that
6 GHz. Meanwhile, the need for higher reliability will be can deliver multiple 3CLS services which are particularly
pervasive across most 6G applications and will be more appealing and even necessary for applications such as
challenging to meet at high frequencies. XR, CRAS, and DLT where tracking, control, localiza-
• Trend 2 – From Spatial to Volumetric Spectral and tion, and computing are an inherent feature. Moreover,
Energy Efficiency: 6G must deal with ground and aerial sensing services will enable 6G systems to provide users
users, encompassing smartphones and XR/BCI devices with a 3D mapping of the radio environment across
along with flying vehicles. This 3D nature of 6G requires different frequencies. Hence, 6G systems must tightly
an evolution towards a volumetric rather than spatial integrate and manage 3CLS functions.
bandwidth definition. We envision that 6G systems must
deliver high spectral and energy efficiency (SEE) require-
ments measured in bps/Hz/m3 /Joules. This is a natural 2 See https://bwrc.eecs.berkeley.edu/projects/5605/ewallpaper.
 • Trend 7 – End of the Smartphone Era: Smartphones 4) Multi-Purpose 3CLS and Energy Services: 6G sys-
were central to 4G and 5G. However, recent years tems must jointly deliver 3CLS services and their derivatives.
witnessed an increase in wearable devices whose func- They can also potentially offer energy to small devices via
tionalities are gradually replacing those of smartphones. wireless energy transfer. Such multi-purpose 3CLS and energy
This trend is further fueled by applications such as XR services (MPS) will be particularly important for applications
and BCI. The devices associated with those applications such as CRAS. MPS require joint uplink-downlink designs
range from smart wearables to integrated headsets and and must meet target performance for the control (e.g., stabil-
smart body implants that can take direct sensory inputs ity), computing (e.g., computing latency), energy (e.g., target
from human senses; bringing an end to smartphones and energy to transfer), localization (e.g., localization precision),
potentially driving a majority of 6G use cases. as well as sensing and mapping functions (e.g., accuracy of a
As shown in Table I, collectively, these trends impose new per- mapped radio environment).
formance targets and requirements on next-generation wireless
III. 6G: E NABLING T ECHNOLOGIES
systems that will be met in two stages: a) A major beyond 5G
evolution and b) A revolutionary step towards 6G. To enable the aforementioned services and guarantee their
performance, a cohort of new, disruptive technologies must be
C. New 6G Service Classes integrated into 6G.
1) Above 6 GHz for 6G – from Small Cells to Tiny Cells:
Beyond imposing new performance metrics, the new techno- As per Trends 1 and 2, the need for higher data rates and SEE
logical trends will redefine 5G application types by morphing anywhere, anytime in 6G motivates exploring higher frequency
classical URLLC, eMBB, and mMTC and introducing new bands beyond sub-6 GHz. As a first step, this includes further
services (summarized in Table II), as follows: developing mmWave technologies to make mobile mmWave
1) Mobile Broadband Reliable Low Latency Communi- a reality in early 6G systems. As 6G progresses, exploiting
cation: As evident from Section II-B, the distinction between frequencies beyond mmWave, at the terahertz (THz) band,
eMBB and URLLC will no longer be sustainable to support will become necessary [7]. To exploit higher mmWave and
applications such as XR, wireless BCI, or CRAS. This is THz frequencies, the size of the 6G cells must shrink from
because these applications require, not only high reliability and small cells to “tiny cells” whose radius is only a few tens
low latency but also high 5G-eMBB-level data rates. Hence, meters. This motivates new architectural designs that need
we propose a new service class called mobile broadband much denser deployments of tiny cells and new high-frequency
reliable low latency communication (MBRLLC) that allows mobility management techniques.
6G systems to deliver any required performance within the 2) Transceivers with Integrated Frequency Bands: On
rate-reliability-latency space. As seen in Fig. 2, MBRLLC their own, dense high-frequency tiny cells may not be able to
generalizes classical URLLC and eMBB services. Energy provide the seamless connectivity required for mobile 6G ser-
efficiency is central for MBRLLC, not only because of its vices. Instead, an integrated system that can leverage multiple
impact on reliability and rate, but also because 6G devices frequencies across the microwave/mmWave/THz spectra (e.g.,
will continue to shrink in size and increase in functionality. using multi-mode base stations) is needed to provide seamless
2) Massive URLLC: 5G URLLC meant meeting relia- connectivity at both wide and local area levels.
bility and latency of very specific uplink IoE applications 3) Communication with Large Intelligent Surfaces:
such as smart factories„ for which prior work [5] provided Massive MIMO will be integral to both 5G and 6G due to the
the needed fundamentals. However, 6G must scale classical need for better SEE, higher data rates, and higher frequencies
URLLC across the device dimension thereby leading to a new (Trend 1). However, for 6G systems, as per Trend 3, we en-
massive URLLC (mURLLC) service that merges 5G URLLC vision an initial leap from traditional massive MIMO towards
with legacy mMTC. mURLLC brings forth a reliability- large intelligent surfaces (LISs) and smart environments [8]
latency-scalability tradeoff which mandates a major departure that can provide massive surfaces for wireless communications
from average-based network designs (e.g., average through- and for heterogeneous devices (Trend 7). LISs enable inno-
put/delay). Instead, a principled and scalable framework which vative ways for communication such as by using holographic
accounts for delay, reliability, packet size, network architec- radio frequency (RF) and holographic MIMO. LISs will likely
ture, topology (across access, edge, and core) and decision- play a basic role in early 6G roll-outs and become more central
making under uncertainty is necessary [6]. as 6G matures.
3) Human-Centric Services: We propose a new class of 4) Edge AI: AI is witnessing an unprecedented interest
6G services, dubbed human-centric services (HCS), that pri- from the wireless community [9] driven by recent break-
marily require QoPE targets (tightly coupled with their human throughs in deep learning, the increase in available data (Trend
users, as explained in Section II-A) rather than raw rate- 4), and the rise of smart devices (Trend 7). Imminent 6G use
reliability-latency metrics. Wireless BCI are a prime example cases for AI (particularly for reinforcement learning) revolve
of HCS in which network performance is determined by the around creating SSNs (Trend 5) that can autonomously sustain
physiology of the human users and their actions. For such high KPIs and manage resources, functions, and network
services, a whole new set of QoPE metrics must be defined control. AI will also enable 6G to automatically provide MPS
and offered as function of raw QoS and QoE metrics. to its users and to send and create 3D radio environment maps
 TABLE I
R EQUIREMENTS OF 5G VS . B EYOND 5G VS . 6G.

5G Beyond 5G 6G

• eMBB. • Reliable eMBB. New applications (see Section II-C):

• URLLC. • URLLC. • MBRLLC.
Application Types • mMTC. • mMTC. • mURLLC.
• Hybrid (URLLC + eMBB). • HCS.
• MPS.
• Smartphones. • Smartphones. • Sensors and DLT devices.
• Sensors. • Sensors. • CRAS.
Device Types
• Drones. • Drones. • XR and BCI equipment.
• XR equipment. • Smart implants.

Spectral and Energy

Efficiency Gains with
10x in bps/Hz/m2 100x in bps/Hz/m2 1000x in bps/Hz/m3 (volumetric)
Respect to Today’s
Networks

Rate Requirements 1 Gbps 100 Gbps 1 Tbps

End-to-End Delay
5 ms 1 ms < 1 ms
Requirements

Radio-Only Delay
100 ns 100 ns 10 ns
Requirements

Processing Delay 100 ns 50 ns 10 ns

End-to-End Reliability
Five 9s Six 9s Seven 9s
Requirements

• Sub-6 GHz. • Sub-6 GHz. • Sub-6 GHz.

Frequency Bands • MmWave for fixed access. • MmWave for fixed access at 26 GHz • MmWave for mobile access.
and 28GHz. • Exploration of THz bands (above 140 GHz).
• Non-RF (e.g., optical, VLC, etc.).
• Dense sub-6 GHz small • Denser sub-6 GHz small cells with • Cell-free smart surfaces at high
base stations with umbrella macro base stations. frequency supported by mmWave
umbrella macro base tiny cells for mobile and fixed
stations. • < 100 m tiny and dense mmWave cells. access.
Architecture
• Temporary hotspots served by
• MmWave small cells of drone-carried base stations or
about 100 m (for fixed tethered balloons.
access).
• Trials of tiny THz cells.

(Trend 6). These short-term AI-enabled 6G functions will be low orbit satellites (LEO) and CubeSats to provide backhaul
complemented by a so-called “collective network intelligence” support and additional wide area coverage. Integrating terres-
in which network intelligence is pushed at the edge, running trial, airborne, and satellite networks [10] and [11] into a single
AI algorithms and machine learning on edge devices (Trend wireless system will be essential for 6G.
7) to provide distributed autonomy. This new edge AI leap 6) Energy Transfer and Harvesting: 6G could be the
will create a 6G system that can integrate the services of first generation of cellular systems that can provide energy,
Section II, realize 3CLS, and potentially replace classical along with 3CLS (Trend 6). As wireless energy transfer is
frame structures. maturing, it is plausible to foresee 6G base stations providing
5) Integrated Terrestrial, Airborne, and Satellite Net- basic power transfer for devices, particularly implants and sen-
works: Beyond their inevitable role as users of 6G systems, sors (Trend 7). Adjunct energy-centric ideas, such as energy
drones can be leveraged to complement ground, terrestrial harvesting (from RF or renewable sources) and backscatter
networks by providing connectivity to hotspots and to areas will also be a component of 6G.
in which infrastructure is scarce. Meanwhile, both drones and 7) Beyond 6G: A handful of technologies will mature
terrestrial base stations may require satellite connectivity with along the same time of 6G and, hence, potentially play a
 Broadband
LL
eMBB
Reliability (1-10y)
-1

-5

-7
1000
Mobile
100 -9 Broadband
1 URLLC Reliable
0 0 0.1 1 10 100
Latency (ms)
Data rate (Gbps)
Fig. 2. MBRLLC services and several special cases (including classical eMBB and URLLC) within the rate-reliability-latency space. Other involved, associated
metrics that are not shown include energy and network scale.

role towards the end of the 6G standardization and research THz environments is necessary. Another important research
process. One prominent example is quantum computing and direction is to study the co-existence of THz, mmWave, and
communications that can provide security and long-distance microwave cells across all layers, building on early works
networking. Currently, major research efforts are focused on such as [13].
the quantum realm and we expect them to intersect with 6G. 3) 3D Networking: Due to the integration of ground and
Other similar beyond 6G technologies include integration of airborne networks, as outlined in Section III, 6G must support
RF and non-RF links (including optical, neural, molecular, and communications in 3D space, including serving users in 3D
other channels). and deploying 3D base stations (e.g., tethered balloons or
temporary drones). This, in turn, requires concerted research
IV. 6G: R ESEARCH AGENDA AND O PEN P ROBLEMS on various fronts. First, measurement and (data-driven) mod-
Building on the identified trends in Section II and the eling of the 3D propagation environment is needed. Second,
enabling technologies in Section III, we now put forward a new approaches for 3D frequency and network planning (e.g.,
research agenda for 6G along with selected open problems where to deploy base stations, tethered balloons, or even
(summarized in Table III). drone-base stations) must be developed. Our work in [10]
1) 3D Rate-Reliability-Latency Fundamentals: Funda- already showed that such 3D planning is substantially differ-
mental 3D performance of 6G systems, in terms of rate- ent from conventional 2D networks due to the new altitude
reliability-latency tradeoffs and SEE is needed. Such analy- dimension and the associated degrees of freedom. Finally,
sis must quantify the spectrum, energy, and communication new network optimizations for mobility management, multiple
requirements that 6G needs to support the identified driving access, routing, and resource management in 3D are needed.
applications. Recent works in [6] and [12] provide a first step 4) Communication with LIS: As per Trend 3, 6G will
in this direction. provide wireless connectivity via smart LIS environments that
2) Exploring Integrated, Heterogeneous High- include active frequency selective surfaces, metallic passive
Frequency Bands: Exploiting mmWave and THz in reflectors, passive/active reflect arrays, as well as nonre-
6G brings forth several new open problems from hardware configurable and reconfigurable metasurfaces. Open research
to system design. For mmWave, supporting high mobility problems here range from the optimized deployment of pas-
at mmWave frequencies will be a central open problem. sive reflectors and metasurfaces to AI-powered operation of
Meanwhile, for THz, new transceiver architectures are needed reconfigurable LIS. Fundamental analysis to understand the
along with new THz propagation models [7]. High power, performance of LIS and smart surfaces, in terms of rate,
high sensitivity, and low noise figure are key transceiver latency, reliability, and coverage is needed, building on the
features needed to overcome the very high path-loss at THz early works in [8]. Another important research direction is to
frequencies. Once these physical layer aspects are well- investigate the potential of using LIS-based reflective surfaces
understood, developing new multiple access and networking to enhance the range and coverage of tiny cells and to
paradigms under the highly varying and mobile mmWave and dynamically modify the propagation environment. Using LIS
 TABLE II
S UMMARY OF 6G SERVICE CLASSES , THEIR PERFORMANCE INDICATORS , AND EXAMPLE APPLICATIONS .

Service Performance Indicators Example Applications

• Stringent rate-reliability-latency requirements. • XR/AR/VR.

• Energy efficiency. • Autonomous vehicular systems.
MBRLLC
• Rate-reliability-latency in mobile environments. • Autonomous drones.
• Legacy eMBB and URLLC.
• Ultra high reliability. • Classical Internet of Things.
• Massive connectivity. • User tracking.
mURLLC
• Massive reliability. • Blockchain and DLT.
• Scalable URLLC. • Massive sensing.
• Autonomous robotics.

• QoPE capturing raw wireless metrics as • BCI.

well as human and physical factors. • Haptics.
HCS
• Empathic communication.
• Affective communication.
• Control stability. • CRAS.
• Computing latency. • Telemedicine.
• Localization accuracy. • Environmental mapping and imaging.
MPS
• Sensing and mapping accuracy. • Some special cases of XR services.
• Latency and reliability for communications.
• Energy.

Communication &
Information
AI/Machine learning Data analytics Theory
foundations
Intelligence Communications
foundations foundations
Game Theory Electromagnetics
Autonomy Large
and sensing intelligent surfaces
foundations foundations
Control Theory 3D Stochastic
Geometry
Risk 6G System Networking
foundations foundations
Network Science
Microeconomics &
Statistics Physics Optimization
foundations foundations
Human Computing
Physics and QoPE foundations Stochastic
foundations Optimization &
Optimal Transport
Psychophysics
& Physiology

Fig. 3. Necessary foundations and associated analytical tools for 6G.

for wireless energy transfer is also an interesting direction. [9] and [14]. This joint design of ML and wireless networks
is an important area of research for 6G.
5) AI for Wireless: AI brings forward many major re-
search directions for 6G. Beyond the need for massive, small 6) QoPE Metrics: The design of QoPE metrics that in-
data analytics as well as using machine learning (ML) and AI- tegrate physical factors from human physiology (for HCS
based SSNs (realized using reinforcement learning and game services) or from a control system (for CRAS) is an important
theory), there is also a need to operate ML algorithms reliably 6G research area, especially in light of new, emerging devices
over 6G to deliver the applications of Section II. To perform (Trend 7). This requires both real-world psychophysics ex-
these critical application tasks, low-latency, high-reliability periments as well as new, rigorous mathematical expressions
and scalable AI is needed, along with a reliable infrastructure for QoPE that combine QoS, QoE, and human perceptions.
Theoretical development of QoPE can be achieved using • Recommendation 1: A first step towards 6G is to enable
techniques from other disciplines such as operations research MBRLLC and mobility management at high-frequency
(e.g., multi-attribute utility theory (see [15])) and machine mmWave bands and beyond (i.e., THz).
learning (see [2]). 6G will be the first generation to enable a • Recommendation 2: 6G requires a move from radio-
new breed of applications (wireless BCI) leveraging multiple centric system design (à-la-3GPP) towards an end-to-end
human cognitive senses. co-design 3CLS under the orchestration of an AI-driven
7) Joint Communication and Control: 6G needs to intelligence substrate.
pervasively support CRAS. The performance of CRAS is gov- • Recommendation 3: The 6G vision will not be a simple
erned by real-world control systems whose operation requires case of exploring additional, high-frequency spectrum
data input from wireless 6G links. Therefore, operating CRAS bands to provide more capacity. Instead, it will be driven
over 6G systems requires a communication and control co- by a diverse portfolio of applications, technologies, and
design, whereby the performance of the 6G wireless links is techniques (see Figs. 1 and 3).
optimized to cater for the stability of the control system and • Recommendation 4: 6G will transition from the
vice versa. Due to the traditional radio-centric focus (3GPP smartphone-base station paradigm into a new era of smart
and IEEE fora), such a co-design has been overlooked in surfaces communicating with human-embedded implants.
5G. Meanwhile, prior works on networked control abstract the • Recommendation 5: Performance analysis and opti-
specifics of the wireless network and cannot apply to cellular mization of 6G requires operating in 3D space and
communications. This makes the communication-control co- moving away from simple averaging towards fine-grained
design a key research topic in 6G. analysis that deals with tails, distributions, and QoPE.
8) 3CLS: The idea of joint communication and control R EFERENCES
must be extended to the joint design of the entire 3CLS [1] F. Boccardi, R. W. Heath, A. Lozano, T. L. Marzetta, and P. Popovski,
functions. The interdependence between computing, commu- “Five disruptive technology directions for 5G,” IEEE Communications
nication, control, localization, sensing, energy, and mapping Magazine, vol. 52, no. 2, pp. 74–80, February 2014.
[2] A. Taleb Zadeh Kasgari, W. Saad, and M. Debbah, “Human-in-the-loop
has not yet been fully explored in an end-to-end manner. Key wireless communications: Machine learning and brain-aware resource
questions range from how to jointly meet the performance of management,” arXiv preprint arXiv:1804.00209, March 2018.
[3] J. Park and M. Bennis, “URLLC-eMBB slicing to support VR multi-
all 3CLS services to multi-modal sensor fusion for reconstruct- modal perceptions over wireless cellular systems,” in Proc. of the IEEE
ing 3D images and navigating in unknown environments for Global Communications Conference (GLOBECOM), Abu-Dhabi, UAE,
Dec. 2018.
navigating robots, autonomous driving, etc. 3CLS is needed [4] P. Zioga, F. Pollick, M. Ma, P. Chapman, and K. Stefanov, “Enheduanna
for various applications including CRAS, XR, and DLT. a manifesto of falling live brain computer cinema performance: Per-
former and audience participation, cognition and emotional engagement
9) RF and non-RF Link Integration: 6G will witness a using multi brain BCI interaction,” Frontiers in neuroscience, vol. 12,
convergence of RF and non-RF links that encompass optical, pp. 191, April 2018.
[5] G. Durisi, T. Koch, and P. Popovski, “Toward massive, ultrareliable, and
visible light communication (VLC), molecular communica- low-latency wireless communication with short packets,” Proceedings
tion, and neuro-communication, among others. Design of such of the IEEE, vol. 104, no. 9, pp. 1711–1726, Sep. 2016.
[6] M. Bennis, M. Debbah, and H. V. Poor, “Ultrareliable and low-latency
joint RF/non-RF systems is an open research area. wireless communication: Tail, risk, and scale,” Proceedings of the IEEE,
10) Holographic Radio: RF holography (including holo- vol. 106, no. 10, pp. 1834–1853, Oct 2018.
[7] Y. Xing and T. S. Rappaport, “Propagation measurement system and
graphic MIMO) and spatial spectral holography can be made approach at 140 GHz-moving to 6G and above 100 GHz,” arXiv preprint
possible with 6G due to the use of LIS and similar structures. arXiv:1808.07594, 2018.
[8] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The potential
Holographic RF allows for control of the entire physical space of data transmission with large intelligent surfaces,” IEEE Transactions
and the full closed loop of the electromagnetic field through on Signal Processing, vol. 66, no. 10, pp. 2746–2758, May 2018.
[9] M. Chen, U. Challita, W. Saad, C. Yin, and M. Debbah, “Machine
spatial spectral holography and spatial wave field synthesis. learning for wireless networks with artificial intelligence: A tutorial
This greatly improves spectrum efficiency and network ca- on neural networks,” available online: arxiv.org/abs/1710.02913, Oct.
2017.
pacity, and helps the integration of imaging and wireless [10] M. Mozaffari, A. Taleb Zadeh Kasgari, W. Saad, M. Bennis, and
communication. How to realize holographic radio is a widely M. Debbah, “Beyond 5G with UAVs: Foundations of a 3D wireless
open area. cellular network,” IEEE Transactions on Wireless Communications, vol.
18, no. 1, pp. 357–372, Jan 2019.
An overview on the necessary analytical tools and funda- [11] X. Cao, S. Kim, K. Obraczka, C. Wang, D. O. Wu, and
mentals related to these open research problems is shown in H. Yanikomeroglu, “Guest editorial airborne communication networks,”
IEEE Journal on Selected Areas in Communications, vol. 36, no. 9, pp.
Fig. 3. 1903–1906, Sep. 2018.
[12] A. T. Z. Kasgari and W. Saad, “Model-free ultra reliable low latency
communication (URLLC): A deep reinforcement learning framework,”
V. C ONCLUSION AND R ECOMMENDATIONS in Proc. of the IEEE International Conference on Communications
(ICC), Shanghai, China, May 2019.
This article laid out a bold new vision for 6G systems that [13] O. Semiari, W. Saad, M. Bennis, and M. Debbah, “Integrated millimeter
wave and Sub-6 GHz wireless networks: A roadmap for joint mobile
outlines the trends, challenges and associated research. While broadband and ultra-reliable low-latency communications,” IEEE Wire-
many topics will come as a natural 5G evolution, new avenues less Communications, to appear, 2019.
[14] J. Park, S. Samarakoon, M. Bennis, and M. Debbah, “Wireless network
of research such as LIS-communication, 3CLS, holographic intelligence at the edge,” arXiv preprint arXiv:1812.02858, Dec. 2018.
radio, and others will create an exciting research agenda for [15] M. Chen, W. Saad, and C. Yin, “Virtual reality over wireless networks:
Quality-of-service model and learning-based resource management,”
the next decade. To conclude, several recommendations are in IEEE Transactions on Communications, vol. 66, no. 11, pp. 5621–5635,
order: Nov 2018.
 TABLE III
S UMMARY OF R ESEARCH A REAS

Research Area Challenges Open Problems

• Fundamental communication limits. • 3D performance analysis of rate-reliability-latency region.

3D Rate-Reliability-Latency • 3D nature of 6G systems. • Characterization of achievable rate-reliability-latency targets.
Fundamentals • 3D SEE characterization.
• Characterization of energy and spectrum needs
for rate-reliability-latency targets.

• Challenges of operation in highly mobile systems. • Effective mobility management for mmWave and THz systems.
• Susceptibility to blockage. • Cross-band physical, link, and network layer optimization.

Exploring Integrated, • Short range. • Coverage and range improvement.

Heterogeneous • Lack of propagation models. • Design of mmWave and THz tiny cells.
High-Frequency Bands
• Need for high fidelity hardware. • Design of new high fidelity hardware for THz.
• Co-existence of frequency bands. • Propagation measurements and modeling across mmWave
and THz bands.

• Presence of users and base stations in 3D. • 3D propagation modeling.

3D Networking • High mobility. • 3D performance metrics.
• 3D mobility management and network optimization.
• Complex nature of LIS surfaces. • Optimal deployment and location of LIS surfaces.
• Lack of existing performance models. • LIS reflectors vs. LIS base stations.
• Lack of propagation models. • LIS for energy transfer.
Communication with LIS • Heterogeneity of 6G devices and services. • AI-enabled LIS.
• Ability of LIS to provide different • LIS across 6G services.
functions (reflectors, base stations, etc.).
• Fundamental performance analysis of LIS transmitters and
reflectors at various frequencies.

• Design of low-complexity AI solutions. • Reinforcement learning for SON.

• Massive, small data. • Big and small data analytics.
AI for Wireless
• AI-powered network management.
• Edge AI over wireless systems.
• Incorporate raw metrics with human perceptions. • Theoretical development of QoPE metrics.
• Accurate modeling of human perceptions • Empirical QoPE characterization.
New QoPE Metrics and physiology.
• Real psychophysics experiments.
• Definition of realistic QoPE targets and measures.
• Integration of control and communication metrics. • Communication and control systems co-design.
• Handling dynamics and multiple time scales. • Control-enabled wireless metrics.
Joint Communication and Control
• Wireless-enabled control metrics.
• Joint optimization for CRAS.
• Integration of multiple functions. • Design of 3CLS metrics.
• Lack of prior models. • Joint 3CLS optimization.
3CLS
• AI-enabled 3CLS.
• Energy efficient 3CLS.
• Different physical nature of RF/non-RF interfaces. • Design of joint RF/non-RF hardware.
RF and non-RF Link Integration • System-level analysis of joint RF/non-RF systems.
• Use of RF/non-RF systems for various 6G services.
• Lack of existing models. • Design of holographic MIMO using LIS.
• Hardware and physical layer challenges. • Performance analysis of holographic RF.
Holographic Radio
• 3CLS over holographic radio.
• Network optimization with holographic radio.

View publication stats