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

net/publication/348155658

The Road Towards 6G: A Comprehensive Survey

Preprint · January 2021

DOI: 10.36227/techrxiv.13382765.v1

CITATIONS READS

0 270

4 authors:

Wei Jiang Bin Han

Deutsches Forschungszentrum für Künstliche Intelligenz Technische Universität Kaiserslautern
74 PUBLICATIONS   703 CITATIONS    57 PUBLICATIONS   929 CITATIONS   

SEE PROFILE SEE PROFILE

Mohammad Asif Habibi Hans D. Schotten

Technische Universität Kaiserslautern Deutsches Forschungszentrum für Künstliche Intelligenz
24 PUBLICATIONS   218 CITATIONS    431 PUBLICATIONS   5,260 CITATIONS   

SEE PROFILE SEE PROFILE

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

National Reference Project for IT Security in Industry 4.0 (IUNO) View project

AI for network security View project

All content following this page was uploaded by Bin Han on 11 May 2021.

The user has requested enhancement of the downloaded file.

 Received 25 January 2021; accepted 31 January 2021. Date of publication 8 February 2021; date of current version 24 February 2021.
Digital Object Identifier 10.1109/OJCOMS.2021.3057679

The Road Towards 6G: A Comprehensive Survey

WEI JIANG 1 (Senior Member, IEEE), BIN HAN 2 (Member, IEEE), MOHAMMAD ASIF HABIBI 2,

AND HANS DIETER SCHOTTEN1,2 (Member, IEEE)

1 Intelligent Networking Research Group, German Research Center for Artificial Intelligence (DFKI), 67663 Kaiserslautern, Germany
2 Division of Wireless Communications and Radio Navigation, Department of Electrical and Computer Engineering,

University of Kaiserslautern, 67663 Kaiserslautern, Germany

CORRESPONDING AUTHOR: W. JIANG (e-mail: wei.jiang@dfki.de)
This work was supported in part by the German Federal Ministry of Education and Research (BMBF) through KICK under Grant 16KIS1105; in part by the
TACNET4.0 project under Grant KIS15GTI007; in part by the European Commission’s H2020 program through 5G-PPP SELFNET project
under Grant H2020-ICT-2014-2/671672; in part by 6G flagship research project Hexa-X under Grant H2020-ICT-2020-2/101015956;
and in part by 6G research project AI@EDGE under Grant H2020-ICT-2020-2/101015922.

ABSTRACT As of today, the fifth generation (5G) mobile communication system has been rolled out
in many countries and the number of 5G subscribers already reaches a very large scale. It is time for
academia and industry to shift their attention towards the next generation. At this crossroad, an overview
of the current state of the art and a vision of future communications are definitely of interest. This article
thus aims to provide a comprehensive survey to draw a picture of the sixth generation (6G) system in
terms of drivers, use cases, usage scenarios, requirements, key performance indicators (KPIs), architecture,
and enabling technologies. First, we attempt to answer the question of “Is there any need for 6G?” by
shedding light on its key driving factors, in which we predict the explosive growth of mobile traffic until
2030, and envision potential use cases and usage scenarios. Second, the technical requirements of 6G
are discussed and compared with those of 5G with respect to a set of KPIs in a quantitative manner.
Third, the state-of-the-art 6G research efforts and activities from representative institutions and countries
are summarized, and a tentative roadmap of definition, specification, standardization, and regulation is
projected. Then, we identify a dozen of potential technologies and introduce their principles, advantages,
challenges, and open research issues. Finally, the conclusions are drawn to paint a picture of “What 6G
may look like?.” This survey is intended to serve as an enlightening guideline to spur interests and further
investigations for subsequent research and development of 6G communications systems.

INDEX TERMS 5G, 6G, artificial intelligence, blockchain, cell-free MIMO, digital twin, edge computing,
holographic-type communications, Internet of Everything, Internet of Things, machine learning, mobile
networks, non-terrestrial networks, optical wireless communications, O-RAN, tactile Internet, terahertz,
visible light communications, wireless communications.

GLOSSARY AoI age of information

1G first generation AR augmented reality
2D two-dimensional AoS age of synchronization
3D three-dimensional AES advanced encryption standard
3G third generation ATIS Alliance for Telecommunications Industry
3GPP Third Generation Partnership Project Solutions
4G fourth generation CAPEX capital expenditure
5G fifth generation CDMA code division multiple access
5G-PPP Fifth Generation Private Public Partnership CoMP coordinated multi-point
6G sixth generation CR cognitive radio
AI artificial intelligence CPU central processing unit
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/

334 VOLUME 2, 2021

 CU centralized unit NFV network function virtualization
DARPA Defense Advanced Research Projects NG-RAN next-generation radio access network
Agency NGMN Next Generation Mobile Networks
DSA digital signature algorithm NIST National Institute of Standards and
DLT distributed ledge technology Technology
DSM dynamic spectrum management NOMA non-orthogonal multiple access
DU distributed unit NR new radio
E2E end-to-end NTN non-terrestrial network
ECC elliptic curve cryptosystem OFDM orthogonal frequency-division multipling
EI edge intelligence OMA orthogonal multiple access
eMBB enhanced mobile broad-band OPEX operational expenditure
ENI experiential network intelligence O-RAN open radio access network
ER extended reality PD power domain
ETSI European Telecommunication Standard PDMA pattern-division multiple access
Institute PNF physical network function
FCC Federal Communications Commission PoP point of presence
FSO free-space optical PPP Public-Private Partnership
GEO geostationary Earth orbit PRB physical resource block
GDPR General Data Protection Regulation QoS quality of service
gNB next-generation NodeB RAN radio access network
GSM Global System for Mobile communications RAT radio access technology
HAP high-altitude platform RF radio frequency
HTC holographic-type communication RIS reconfigurable intelligent surfaces
IAB integrated access and backhaul RSA Rivest-Shamir-Adleman
ICT information and communication technology RU radio unit
IDMA interleave division multiple access SCMA sparse-code multiple access
IMT International Mobile Telecommunications SDN software-defined networking
IoE internet of everything SLAM simultaneous localization and mapping
IoIT internet of intelligent things SLM spatial light modulator
IoT internet of things SRE smart radio environment
IR infrared SWIPT simultaneous wireless information and
IRS intelligent reflecting surfaces power transfer
ISG industry specification group TCMA trellis-coded multiple access
ISTN integrated space and terrestrial network TD-SCDMA time division synchronous code division
ITU-R International Telecommunciation Union - multiple access
Radiocommunication THz terahertz
ITU-T International Telecommunciation Union - UAV unmanned aerial vehicle
Telecommunication ULBC ultra-reliable low-latency broadband com-
KPI key performance indicator munication
LBT listen-before-talk uMBB ubiquitous mobile broadband
LD Laser diode URLLC ultra-reliable low-latency communications
LED light-emitting diodes VLC visible light communication
LEO low Earth orbit VNF virtual network function
LOS line of sight VR virtual reality
LTE long term evolution WCDMA wideband code division multiple access
M2M machine to machine WSN wireless sensor network
MANO management and orchestration WRC World Radiocommunication Conference.
MBB mobile broadband
MIMO multi-input multi-output I. INTRODUCTION
ML
mMTC
mmWave
machine learning
massive machine-type communications
millimeter wave
T HE MOBILE telecommunication industry stems from the
first generation (1G) analog cellular systems represented
by Advanced Mobile Phone System in the United States and
mULC massive ultra-reliable low-latency Nordic Mobile Telephone in Europe, which firstly offered
communication mobile voice-calling service around the year 1980. Since
MUSA multi-user shared access then, a new generation of mobile communications was intro-
NFMF network function management function duced to market nearly every ten years. The 1G analog

VOLUME 2, 2021 335

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

systems were replaced by the second generation digital cel- where Windows 10 is the ultimate version, and even there is an
lular networks in around 1990. Despite of several competing opposition to talking about 6G [5], several pioneering works on
systems, the Global System for Mobile Communications the next-generation wireless networks have been initiated. A
known as GSM [1] achieved a great commercial success and focus group called Technologies for Network 2030 within the
allowed more than one billion of the world’s population to International Telecommunication Union Telecommunication
enjoy the convenience brought by mobile voice, short texting, (ITU-T) standardization sector was established in July 2018.
and low-rate data services. Exploiting a revolutionary tech- The group intends to study the capabilities of networks for
nology named Code-Division Multiple Access (CDMA), the 2030 and beyond [6], when it is expected to support novel
third generation (3G) systems [2] represented by WCDMA, forward-looking scenarios, such as holographic-type commu-
CDMA2000, and TD-SCDMA, were developed and firstly nications, ubiquitous intelligence, Tactile Internet, multi-sense
deployed in 2001 to support high-speed data access with a experience, and digital twin. The European Commission ini-
rate of several megabits per second. In December 2009, the tiated to sponsor beyond 5G research activities, as its recent
commercial Long Term Evolution (LTE) networks [3] were Horizon 2020 calls - ICT-20 5G Long Term Evolution and
launched in the Scandinavian capitals Stockholm and Oslo, ICT-52 Smart Connectivity beyond 5G – where a batch of
providing the world’s first fourth generation (4G) mobile pioneer research projects for key 6G technologies were kicked
broadband service. The 4G system that is empowered by off at the early beginning of 2020. The European Commission
a genius combination of multi-input multi-output (MIMO) has also announced its strategy to accelerate investments in
and orthogonal frequency-division multiplexing (OFDM) Europe’s “Gigabit Connectivity” including 5G and 6G to
spurs the proliferation of smart phones, fostering the mobile shape Europe’s digital future [7]. In October 2020, the Next
Internet industry that is worth trillions of dollars a year. Generation Mobile Networks (NGMN) has launched its new
In April 2019, when South Korea’s three mobile opera- “6G Vision and Drivers” project, intending to provide early
tors and U.S. Verizon were arguing with each other about and timely direction for global 6G activities. At its meet-
who is the world’s first provider of the fifth generation ing in February 2020, the International Telecommunication
(5G) communication services, we stepped into the era of Union Radiocommunication sector (ITU-R) decided to start
5G. In the past two years, the term of 5G has been the study on future technology trends for the future evolution
remaining one of the hottest buzzwords in the media, of International Mobile Telecommunications (IMT) [8]. In
attracting unprecedented attention from the whole society. Finland, the University of Oulu began ground-breaking 6G
It even went beyond the sphere of technology and economy, research as part of Academy of Finland’s flagship program [9]
becoming the focal point of geopolitical tension. Unlike called 6G-Enabled Wireless Smart Society and Ecosystem
the previous generations that focused merely on improv- (6Genesis), which focuses on several challenging research
ing network capacities, 5G expands mobile communication areas including reliable near-instant unlimited wireless con-
services from human to things, and also from consumers nectivity, distributed computing and intelligence, as well as
to vertical industries. The potential scale of mobile sub- materials and antennas to be utilized in future for circuits
scription is substantially enlarged from merely billions of and devices. Besides, other traditional main players in mobile
the world’s population to almost countless inter-connectivity communications, such as the United States, China, Germany,
among humans, machines, and things. It enables a wide Japan, and South Korea, already initiated their 6G research
variety of services from traditional mobile broadband to officially or at least announced their ambitions and tentative
Industry 4.0, virtual reality (VR), Internet of Things (IoT), roadmaps. At this crossroad, an overview of the current state
and automatic driving [4]. In 2020, the outbreak of the of the art and a vision of future communications to provide an
COVID-19 pandemic leads to a dramatic loss of human life enlightening guideline for subsequent research and develop-
worldwide and imposes unprecedented challenges on soci- ment works is of interest. Recently, the articles focusing on 6G
etal and economic activities. But this public health crisis topics, e.g., use cases, application scenarios, requirements, and
highlights the unique role of networks and digital infras- promising technological pillars, are emerging in the literature,
tructure in keeping society running and families connected, as summarized in the following subsection.
especially the values of 5G services and applications, such
as remote surgeon, online education, remote working, driver-
less vehicles, unmanned delivery, robots, smart healthcare, A. STATE-OF-THE-ART RELATED WORKS
and autonomous manufacturing. The earliest article [10] that discusses the topic of 6G was
Currently, 5G is still on its way being deployed across the published in September 2018, where David and Berndt tried
world, but it is already the time for academia and industry to to address the question of “Is there any need for beyond
shift their attention to beyond 5G or the sixth generation (6G) 5G?” by reviewing the key services and innovations from
systems, in order to satisfy the future demands for information the 1G analog system to the virtualized and software-defined
and communications technology (ICT) in 2030. Even though 5G infrastructure. Nawaz et al. [11] checked the state-of-
discussions are ongoing within the wireless community as to the-art advances in the fields of machine learning (ML)
whether there is any need for 6G or whether counting the genera- and quantum computing, and then envisaged their syner-
tions should be stopped at 5, adopting the Microsoft’s approach gies with communication networks to be considered in the

336 VOLUME 2, 2021

6G system. In [12], Rappaport et al. described the chal- the future 6G network from the perspectives of communi-
lenges and potentials of terahertz (THz) communications in cation, networking, and computing. Guo outlined the core
the development and implementation of 6G networks. Later, concepts of explainable AI for 6G in [28], including pub-
the authors of [13] provided a brief description of vision and lic and legal motivation, definition, the trade-off between
potential techniques. In [14], Letaief et al. discussed poten- explainability and performance, explainable methods, and
tial technologies to enable ubiquitous Artificial Intelligence an explainable AI framework for future wireless systems. A
(AI) service in 6G networks (i.e., networking for AI) and AI- survey paper [29] provides a comprehensive view of 6G in
enabled methodologies for the design and optimization of 6G terms of applications, requirements, challenges, and research
(i.e., AI for networking). Zong et al. proposed two candidate directions. Some key technologies such as AI, terahertz,
system architectures for 6G in [15] and identified several 6G blockchain, three-dimensional networking, and wireless opti-
technologies including photonics-defined radio, holography, cal communications are briefly introduced. In [30], the
and AI. In the context, the authors of [16] aimed to shed light authors aimed to extend the vision of 5G to more ambitious
on requirements, network architecture, key technologies, scenarios in a more distant future and speculate on the vision-
and new applications in the upcoming 6G system. In [17], ary technologies that can provide the step changes needed
Strinati et al. envisioned five technology enablers for 6G, for enabling 6G. Liu et al. identified the vision of the soci-
i.e., pervasive AI at network edge, three-dimensional (3D) ety development towards 2030 and derived key performance
coverage consisting of terrestrial networks, aerial platforms, requirements from new applications and services. Taken into
and satellite constellation, a new physical layer incorpo- account the convergence of information and communica-
rating sub-THz and Visible Light Communications (VLC), tion technologies, a logical mobile network architecture is
distributed security mechanisms, and a new architecture. proposed to resolve the lessons from 5G network design [31].
Huang et al. surveyed the new architectural changes associ- Gui et al. gave a brief presentation on various 6G issues
ated with green 6G networks in [18], which also briefly in [32], including core services, use cases, requirements,
introduces several potential technologies such as ubiqui- enabling technologies, architectures, typical use scenarios,
tous 3D coverage, pervasive AI, THz, VLC, and blockchain. challenges, and research directions. The authors of [33]
In [19], Jiang and Luo provided a comprehensive and highly provided an overview of holographic MIMO surface com-
coherent treatment on all the technology aspects related to munications as a promising technological enabler for 6G
ML for wireless communications and networks. wireless communications.
Since the beginning of 2020, the number of publications Recently, Bariah et al. gave a comprehensive 6G vision
related to 6G grows a bit faster than that of the past two in [34], identifying seven disruptive technologies, associ-
years. In [20], Dang et al. argued that 6G should be human ated requirements, challenges, and open research issues.
centric, and therefore security, secrecy, and privacy are key From the viewpoints of radio frequency (RF) hardware and
features. To support this vision, a systematic framework, antenna, [35] analyzes link budget of THz communication
required technologies, and anticipated challenges were out- links with justified estimates of calculus terms, such as
lined. Then, a survey on various ML technologies applied for the achievable or required noise figure, transmit power, and
communication, networking, and security aspects of vehicu- antenna gain. It also evaluates communication distances for
lar networks, and a vision of the ways toward an intelligent links implemented with different technologies and complex-
6G vehicular network were provided in [21]. Giordani et al. ity at 300 GHz looking towards anticipated 6G use scenarios.
foresaw several possible 6G use cases and present a num- In [36], Polese et al. provided an overview of the issues
ber of technologies, which are considered by the authors that need to be overcome to introduce the terahertz spec-
as the enablers for these use cases [22]. Viswanathan and trum in mobile networks, from the perspectives of medium
Mogensen attempted to paint a broad picture of commu- access control, network, and transport layer, with consider-
nication needs and technologies in the era of 6G [23], ation on the performance of end-to-end (E2E) data flows
where new themes that are likely to shape 6G require- on terahertz connections. Wang et al. [37] provided a sur-
ments are presented. Zhang et al. argued in their article [24] vey of 6G wireless channel measurements and models for
that 1000 times price reduction from the customer’s view full frequency bands covering mmWave, THz, and optical
point is the key to success for the 6G system. In [25], wireless communication (OWC) channels, full coverage such
Chen et al. contributed a comprehensive discussion that cov- as satellite, maritime, and underwater acoustic communica-
ers visions, requirements, technology trends, and challenges, tion channels, and full application scenarios, e.g., high-speed
aiming at clarifying the ways to tackle the challenges of train, vehicle-to-vehicle, and industry IoT communication
coverage, capacity, data rate, and mobility of 6G mobile channels. Reference [38] presents the potential of integrat-
communication systems. The authors of [26] shared their ing VLC in 6G, and discusses its technological advances
viewpoints in terms of applications, technological trends, including new materials and devices, modulation, under-
service classes, and requirements, and then give their identifi- water transmission, and ML-based signal processing. The
cation on enabling technologies and open research problems. authors of [39] shed light on the advantages of unmanned
Kato et al. [27] recognized possible challenges and poten- aerial vehicle (UAV) to improve the coverage and capac-
tial research directions of advancing ML technologies into ity of 6G, and proposed a network setup utilizing tethered

VOLUME 2, 2021 337

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

UAVs. In [40], the authors outlined the concept of trust- 5) It summarizes the ambitions, efforts, and research
worthy autonomy for 6G, clarified how explainable AI activities on 6G across the world, while a tentative
can generate the qualitative and quantitative modalities of roadmap of definition, specification, standardization,
trust, and provided associated key performance indicators and regulation is envisaged. To the best knowledge of
(KPIs) for measuring trust. In [41], Du et al. summa- the authors, that is the first time in the literature to
rized some intelligent approaches of applying AI and ML provide such an investigation from this perspective.
tools to optimize 6G networks, including THz communica- 6) An architecture of 3D coverage integrating non-
tions, energy management, security, mobility management, terrestrial and terrestrial networks is envisioned and
and resource allocation. To facilitate a clearer illustration, illustrated within envisioned 6G deployment scenarios.
the aforementioned works with major contributions and 7) Unlike all previous works that simply list tech-
categorized topics are listed chronologically in Table 1. nological candidates in a line, we categorize 6G
enabling technologies into the following groups: New
Spectrum, New Networking, New Air Interface, New
B. CONTRIBUTIONS Architecture, and New Paradigm. This methodology is
It is noticed that most of the aforementioned previous works the first time being used in 6G publications.
focus merely on one specific aspect of 6G, such as THz [12], 8) It gives a complete view of potential 6G technolo-
AI [14], green networks [18], use cases [22], ML [27], and gies, which identifies the largest set of enablers by
VLC [38]. There are a few surveys attempting to provide far and the number of identified enablers is far more
a complete view, but a comprehensive survey is still miss- than any existing survey. The principle, advantages,
ing until now. To fill this gap, this article comprehensively challenges, and open issues for each enabler are
surveys the latest advances of 6G research and provides elaborated. Some of the technologies are introduced
a broad vision in terms of drivers, requirements, efforts, in detail for the first time from the perspective of
and enablers. Upon a thorough state-of-the-art analysis of 6G, e.g., large-scale satellite constellation and post-
related works, this article is started from envisioning the quantum security. It includes: new spectrum consist-
driving forces, potential use cases, and usage scenarios so ing of mmWave, THz communications, VLC, OWC,
as to address the concern on the necessity of developing and dynamic spectrum management (DSM), new
6G. Then, the technical requirements needed to support 6G networking that covers softwarization and virtualiza-
applications and services are clarified in terms of a set of tion, radio access network (RAN) slicing, open-RAN
KPIs, and promising technologies are identified and elab- (O-RAN), and post-quantum security, new air interface
orated. The up-to-date research activities across the world including massive MIMO, intelligent reflecting sur-
are summarized, and the roadmap for research, specification, faces (IRS), coordinated multi-point (CoMP), cell-free
standardization, and development toward 2030 is projected. massive MIMO, and new modulation techniques, new
Finally, the conclusions are drawn to paint a picture of “What architecture providing 3D coverage by means of inte-
6G may look like?.” grating large-scale satellite constellation, high-altitude
Compared with existing 6G articles, the main contributions platform (HAP), and UAV with traditional terrestrial
of this article can be listed as follows: networks, and new paradigm empowered by the con-
1) A thorough state-of-the-art analysis, which provides vergence of computing-communication resources and
the most complete summary of related works with the integration of mobile networks, AI, blockchain,
latest advances, is made. and digital twin.
2) It attempts to answer the question of “Do we really 9) It concludes this article by painting a picture of “What
need 6G?” by shedding light on its key drivers, includ- 6G may look like?.” The authors envision that 6G
ing the explosive growth of mobile traffic and mobile would be a radio-optical system, a connected intel-
subscriptions until 2030, and disruptive use cases. It ligent platform, an integrated space-aerial-terrestrial
goes beyond the state of the art by identifying the con- network, and a smart compute-connect entity to trans-
sensus of previous works on use cases and proposes form the whole Earth into a huge brain, which
novel use cases that have never been reported, i.e., fully supports the informationized and intelligentized
Global Ubiquitous Connectability, Enhanced On-Board society in 2030 and beyond.
Communications, and Pervasive Intelligence.
3) Using a holistic methodology, three novel usage sce- C. ORGANIZATION OF THE ARTICLE
narios for 6G are proposed, i.e., ubiquitous mobile The rest of this article is organized as follows: Section II
boardband (uMBB), ultra-reliable low-latency broad- clarifies the key driving forces for the necessity of devel-
band communication (ULBC), and massive ultra- oping 6G, including the explosive growth of mobile traf-
reliable low-latency communication (mULC). fic and mobile subscriptions, disruptive use cases, and
4) It discusses the technical requirements of 6G in terms advanced usage scenarios. Section III analyzes the tech-
of a set of KPIs, which are compared with the KPIs nical requirements for the 6G system in terms of a
of 5G quantitatively, if applicable. number of KPIs. The ambitions and efforts from the

338 VOLUME 2, 2021

TABLE 1. Summary of state-of-the-art contributions related to 6G communication systems.

main players in the mobile communication industry are II. DRIVERS

summarized and a development roadmap is estimated in Since the middle of 2019, commercial 5G mobile networks
Section IV. Section V provides a complete view of a have been rolled out across the world and already reached a
dozen of key technologies for 6G. Finally, Section VI very large scale in some countries. For example, the number
concludes this article by painting a picture about what of deployed 5G base stations in China exceeds 500 000 at the
6G is. end of 2020, serving more than 100 million 5G subscribers.

VOLUME 2, 2021 339

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

Following the tradition that a new generation appears every

one decade, it is time for both academia and industry to
initiate the exploration of the successor of 5G. On the road
towards 6G, however, the first problem we encounter is that
there are many concerns like “Do we really need 6G?” or
“Is 5G already enough?.” To address such questions, we first
need to clarify the key driving forces for 6G.
The development of a next-generation system is driven by
not only the exponential growth of mobile traffic and mobile
subscriptions but also new disruptive services and applica-
tions on the horizon. In addition, it is also driven by the
intrinsic need of mobile communication society to continu-
ously improve network efficiencies namely cost efficiency,
energy efficiency, spectrum efficiency, and operational effi-
ciency. With the advent of advanced technologies such as
AI, THz, and large-scale satellite constellation, the commu-
FIGURE 1. Estimated global mobile subscriptions and mobile traffic from 2020 to
nication network is able to evolve towards a more powerful 2030. Source: ITU-R Report M.2370-0 [42].
and more efficient system to better fulfil the requirements
of current services and open the possibility for offering dis-
ruptive services that have hitherto never been seen. In this Netflix, and more recently Tik-Tok, as well as the stable
section, we intend to shed light on three drivers: i) the explo- improvement of screen resolution on mobile devices. The
sive growth of mobile traffic, ii) disruptive use cases, and traffic coming from mobile video services already account
iii) novel usage scenarios. Its technological drivers will be for two thirds of all mobile traffic nowadays [43] and
discussed in detail in Section V. is estimated to be more dominant in the future. In some
developed countries, a strong traffic growth before 2025 is
A. EXPLOSIVELY GROWING MOBILE TRAFFIC
driven by rich-video services and a long-term growth wave
We are in an unprecedented era where a large number of will continue due to the penetration of augmented reality
smart products, interactive services, and intelligent applications (AR) and VR applications. The average data consumption
emerge and evolve in a prompt manner, imposing a huge for every mobile user per month, as illustrated in Fig. 1,
demand on mobile communications. It can be foreseen that will increase from around 5 GB in 2020 to over 250 GB
the 5G system is hard to accommodate the tremendous volume in 2030. In addition to human-centric communications, the
of mobile traffic in 2030 and beyond. Due to the proliferation of scale of M2M terminals will increase more rapidly and will
rich-video applications, enhanced screen resolution, machine- become saturated no earlier than 2030. It is predicted that
to-machine (M2M) communications, mobile cloud services, the number of M2M subscriptions will reach 97 billion,
etc., the global mobile traffic will continuously increase in around 14 times over that of 2020 [42]. This serves as
an explosive manner, up to 5016 EB1 per month in the year another driving force for the explosive growth of mobile
2030 compared with 62 EB per month in 2020, according to the traffic.
estimation by ITU-R [42] in 2015. A report from Ericsson [43]
reveals that the global mobile traffic has reached 33 EB per B. POTENTIAL USE CASES
month at the end of 2019, which justifies the correctness of With the advent of new technologies and continuous evo-
ITU-R’s estimation. lution of existing technologies, e.g., holography, robotics,
In the last decade, the number of smartphones and tablets microelectronics, photo-electronics, AI, and space technol-
has experienced an exponential growth due to the prolifera- ogy, many unprecedented applications can be fostered in
tion of mobile broadband (MBB). This trend will continue in mobile networks. To explicitly highlight the unique char-
the 2020s since the penetration of smartphones and tablets is acteristics and define the technical requirements of 6G, we
not saturated especially in developing countries. Meanwhile, foresee several representative use cases as follows:
new-style user terminals, such as wearable electronics and Holographic − Type Communication (HTC): Compared
VR glasses, emerge in the market quickly and are adopted to traditional 3D videos using binocular parallax, true holo-
by consumers in a fast pace. It is expected that the total num- grams can satisfy all visual cues of observing 3D objects
ber of MBB subscribers worldwide will reach 17.1 billion by the naked eye as natural as possible. With a significant
by 2030, as shown in Fig. 1. On the other hand, the traffic advance of holographic display technology in recent years
demand per MBB user continuously raises in the company such as Microsoft’s HoloLens [44], it is envisioned that
of the rising number of MBB users. That is mainly because its application will become a reality in the next decade.
of the popularity of mobile video services such as Youtube, Remote rendering high-definition holograms through a
mobile network will bring truly immersive experience.
1. 1 exabyte (EB)=1 000 000 terabytes (TB), 1 TB=1000 gigabytes (GB) For example, holographic telepresence will allow remote

340 VOLUME 2, 2021

participants to be projected as holograms into a meeting range of applications such as remote surgery, remote control-
room or allow the attendee of online training or education ling, and immersive gaming. This use case brings a stringent
to interact with ultra-realistic objects. However, HTC requirement on low latency.
leads to huge bandwidth demands on the order of terabits Digital Twin: is used to create a comprehensive and
per second even with image compression. In addition detailed virtual copy of a physical (a.k.a. real) object.
to consider the frame rate, resolution, and color depth The softwarized copy is equipped with a wide range of
in two-dimensional (2D) video, the quality of hologram characteristics, information, and properties related to the
also involves the volumetric data such as tilt, angle, and original object. Such a twin is then used to manufacture
position. If representing an object with images every 0.3◦ , multiple copies of an object with full automation and intel-
an image-based hologram with 30◦ field of view and a tilt ligence. The early rollouts of digital twin have attracted
of 10◦ needs a 2D array of 3300 separate images [45]. significant attention of a number vertical industries and man-
HTC also requires ultra-low latency for true immersiveness ufacturers. However, its full deployment is expected to be
and high-precision synchronization across massive bundles realized with the development of 6G networks.
of interrelated streams for reconstructing holograms. Pervasive Intelligence: With the proliferation of mobile
Extended Reality (ER): Combining augmented, virtual, smart devices and the emergence of new-style connected
and mixed realities, ER starts stepping into practical applica- equipment such as robots, smart cars, drones, and VR
tions in the age of 5G, but it is still in its infancy analogue glasses, over-the-air intelligent services are envisioned to
to the video service at the beginning of mobile Internet. boom. These intelligent tasks mainly rely on traditional
To achieve the same level of image quality, ER devices computation-intensive AI technologies: computer vision,
with 360◦ field of view need much higher data through- simultaneous localization and mapping (SLAM), face and
put in comparison to 2D video streaming. For an ideal speech recognition, natural language processing, motion con-
immersion experience, the quality of video with higher reso- trol, to name a few. To overcome stringent computation,
lution, higher frame rate, more color depth, and high dynamic storage, power, and privacy constrains on mobile devices,
range are required, leading to a bandwidth demand of over 6G networks will offer pervasive intelligence in an AI-as-
1.6 Gbps per device [46]. Similar to video traffic that satu- a-Service manner [14] by utilizing distributed computing
rates the 4G networks, the proliferation of ER devices will resources across the cloud, mobile edge, and end-devices,
be blocked by the limited capacity of 5G with the peak and cultivating communication-efficient ML training and
rate of 20 Gbps, especially at the cell edge. In addition to interference mechanisms. For example, a humanoid robot
bandwidth, interactive ER applications such as immersive such as Atlas from Boston Dynamics [49] is possible to
gaming, remote surgery, and remote industrial control, low off-load its computational load for SLAM towards edge
latency and high reliability are mandatory. computing resources, in order to improve motion accuracy,
Tactile Internet: It provides extremely low E2E latency to prolong battery life, and become more lightweight by remov-
satisfy the 1-millisecond (ms) or less reaction time reaching ing some embedded computing components. In addition to
the limit of human sense [47]. In combination with high reli- computation-intensive tasks, pervasive intelligence also facil-
ability, high availability, high security, and sometimes high itates time-sensitive AI tasks to avoid the latency constraint
throughput, a wide range of disruptive real-time applications of cloud computing when fast decisions or responses to
are enabled. It will play a critical role in the field of real-time conditions are required.
monitoring and remote industrial management for Industry Intelligent Transport and Logistics: In 2030 and
4.0 and Smart Grid. For example, with immersive audio- beyond, millions of autonomous vehicles and drones
visual feeds provided by ER or HTC streaming, together provide a safe, efficient, and green movement of people
with haptic sensing data, a human operator can remotely and goods. Connected autonomous vehicles have stringent
control the machinery in a place surrounded by biologi- requirements on reliability and latency to guarantee the
cal or chemical hazards, as well as remote robotic surgery safety of passengers and pedestrians. Unmanned aerial
carried out by doctors from hundreds of miles away [48]. vehicles, especially swarm of drones, open the possibility
The typical closed-loop controlling, especially for devices for a wide variety of unprecedented applications, while
or machinery moving rapidly, is very time-sensitive, where bringing also disruptive requirements for mobile networks.
an E2E latency below 1 ms is expected. Enhanced On − Board Communications: With the
Multi − Sense Experience: Human has five senses (sight, development of economy, the activity sphere of human and
hearing, touch, smell, and taste) to perceive external envi- the frequency of their movement will rapidly increase in
ronment, whereas current communications focus only on the next decade. The number of passengers travelling by
optical (text, image, and video) and acoustic (audio, voice, commercial planes, helicopters, high-speed trains, cruise
and music) media. The involvement of the senses of taste ships, and other vehicles, will be very huge, bringing
and smell can create fully-immersive experience, which may skyrocketing demands on high-quality communication
bring some new services for example in food and texture services on board. Despite the efforts in the previous
industries [6]. Furthermore, the application of haptic com- generations until 5G, it is undeniable that on-board
munication will play a more important role and raise a wide connectivity is far from satisfactory in most cases due

VOLUME 2, 2021 341

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

to high mobility, frequent handover, sparse coverage of

terrestrial networks, and limited bandwidth and high cost
of satellite communications. Relying on reusable space
launching technologies and massive production of satellites,
the deployment of large-scale satellite constellation such
as SpaceX’s Starlink [50] becomes a reality, enabling
cost-efficient and high-throughput global coverage. Keep
this in mind, 6G is expected to be an integrated system
of terrestrial networks, satellite constellation, and other
aerial platforms to provide seamless 3D coverage, which
offers high-quality, low-cost, and global-roaming on-broad
communication services.
Global Ubiquitous Connectability: The previous gen-
erations of mobile communications focused mainly on
the dense metropolitan areas, especially indoor scenarios.
However, a large population in remote, sparse, and rural
areas even have no access to basic ICT services, digging a
big digital divide among humans around the world. Besides,
more than 70 % of the Earth’s surface is covered by water,
FIGURE 2. In addition to typical 5G usage scenarios (eMBB, ULRRC, and mMTC),
where the growth of maritime applications require network three enhanced scenarios named uMBB, ULBC, and mULC are proposed by the
coverage for both water surface and underwater. However, authors of this article for the 6G system in order to support disruptive use cases and
applications.
ubiquitous coverage across the whole planet with sufficient
capacity, acceptable quality of service (QoS), and affordable
cost is far from a reality. On the one hand, it is techni-
cally impossible for terrestrial networks to cover remote smart devices (smartphones, tablets, and wearable elec-
areas and extreme topographies such as ocean, desert, and tronics). It emphasizes wide-area coverage to provide
high mountain areas, while it is too costly to offer terres- seamless access and high capacity in hot spots.
trial communication services for sparsely-populated areas. • ultra-reliable low-latency communications (URLLC) is
On the other hand, geostationary Earth orbit (GEO) satel- a disruptive promotion over the previous generation
lites are expensive to deploy and their capacity is limited to systems that focus on human users. It opens the
several Gbps per satellite [51], which is dedicated only for possibility for mission-critical connectivity for new
high-end users such as maritime and aeronautic industries. applications such as automatic vehicles, Smart Grid,
As mentioned above, the deployment of large-scale low Earth and Industry 4.0, which have stringent requirements on
orbit (LEO) satellite constellation will enable low-cost and reliability, latency, and availability.
high-throughput global communication services [52]. The 6G • massive machine-type communications (mMTC) sup-
system is envisioned to make use of the synergy of terrestrial ports dense connectivity with a very large number of
networks, satellite constellation, and other aerial platforms connected devices typically deployed in IoT scenarios.
to realize ubiquitous connectability for global MBB users The devices such as sensors are low-cost, low-power
and wide-area IoT applications. consumption but typically transmitting a low volume
The connection of the aforementioned use cases with usage of delay-tolerate data.
scenarios proposed in the next subsection is depicted in Fig. 2, It can be seen that these 5G usage scenarios cannot sat-
while some representative use cases are demonstrated within isfy the technical requirements of the aforementioned 6G
envisaged 6G deployment scenarios, as shown in Fig. 3. use cases. For instance, an user that wears a lightweight
VR glass to play interactively-immersive games requires not
only ultra-high bandwidth but also low latency. Autonomous
C. USAGE SCENARIOS vehicles on the road or flying drones need ubiquitous connec-
The 5G system is designed to meet more diverse QoS tivity with high throughput, high reliability, and low latency.
requirements arising from a wide variety of vertical applica- Although a few related works discussed potential usage
tions and services, which have never encountered by mobile scenarios for 6G, their proposed scenarios are simply an
subscribers in the previous generations. To define 5G, three enhancement or extension of 5G scenarios. The definition of
usage scenarios were firstly recommended by ITU-R M.2083 scenarios are arbitrary and piece-wise, where the interwork-
in 2015 [53]: ing among 6G scenarios and their relations with 5G scenarios
• enhanced mobile broad-band (eMBB) addresses the are not correctly clarified. In this article, we propose a holis-
human-centric applications for a high-data-rate access tic and more reasonable methodology to define 6G scenarios
to mobile services, multi-media content, and data. This through extending the scope of current usage scenarios, as
scenario fosters new services and applications over shown in Fig. 2. Three new scenarios are proposed to meet

342 VOLUME 2, 2021

FIGURE 3. Envision of the deployment scenarios and architecture for 6G, demonstrating representative use cases and some of key technological enablers.

the requirement of aforementioned use cases, which cover the wearing lightweight VR glasses gathering in a small room
overlapping areas of 5G scenarios so as to form a complete where a data rate of several Gbps per user is needed. The
set. To support high-quality on-board communications and uMBB scenario will be the foundation of digital twin, per-
global ubiquitous connectability, the MBB service should be vasive intelligence, enhanced on-board communications, and
available across the whole surface of the Earth in the era of global ubiquitous connectability, as the mapping relationship
6G, called ubiquitous MBB or uMBB. In addition to its ubiq- shown in Fig. 2. In addition to KPIs that are applied to evalu-
uitousness, another enhancement of uMBB is a remarkable ate eMBB (such as peak data rate and user-experienced data
boost of network capacity and transmission rate for hot spots rate), other KPIs become as same critical as the others in
so as to support disruptive services, e.g., a group of users uMBB, i.e., mobility, coverage, and positioning, as indicated

VOLUME 2, 2021 343

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

TABLE 2. Performance requirements (KPIs) to support the implementation of usage scenarios.

in Table 2. Ultra-reliable low-latency broadband communi- of times that of 5G, which has the peak rate of 20 Gbps
cation also called ULBC supports the applications requiring for downlink and 10 Gbps for uplink.
not only URLLC but also extreme high throughput, e.g., • User − experienced data rate is defined as the 5th
HTC-based immersive gaming. It is expected that the use percentile point (5%) of the cumulative distribution
cases of HTC, ER, Tactile Internet, multi-sense experience, function of user throughput. In other words, a user can
and pervasive intelligence will benefit from this scenario. get at least this data rate at any time or location with
Furthermore, the third scenario called massive ultra-reliable a possibility of 95%. It is more meaningful to mea-
low-latency communication or mULC combines the charac- sure the perceived performance, especially at the cell
teristics of both mMTC and URLLC, which will facilitate edge, and reflect the quality of network design such as
the deployment of massive sensors and actuators in verti- site density, architecture, inter-cell optimization, etc. In
cal industries. Together with eMBB, URLLC, and mMTC, the 5G deployment scenario of dense urban, the target
three new scenarios fill the gaps in-between and then a com- of user-perceived rate is 100 Mbps for downlink and
plete set of usage scenarios is formed to support all kinds 50 Mbps for uplink. It is expected that 6G can offer
of use cases and applications in 6G, as shown in Fig. 2. 1 Gbps even higher, which is 10 times that of 5G.
Performance requirements (KPIs) required during the design • Latency can be differentiated into user plane and con-
and implementation of such usage scenarios are listed in trol plane latency. The former is the time delay induced
Table 2. in a radio network from a packet being sending out at
the source until the destination receives it, assuming a
mobile station is in the active state. In 5G, the min-
III. REQUIREMENTS imum requirement for user plane latency is 4 ms for
To well support disruptive use cases and applications in 2030 eMBB and 1 ms for URLLC. This value is envisioned
and beyond, the 6G system will provide extreme capacity, to be further reduced to 100 µs or even 10 µs. Control
reliability, efficiency, etc. Like the minimal requirements for plane latency refers to the transition time from a most
IMT-2020 specified in [54], a number of quantitative or qual- “battery efficient” state (e.g., the idle state) to the start
itative KPIs are utilized to indicate the technical requirements of continuous data transfer (e.g., the active state). The
for 6G. Most of the KPIs that are applied for evaluating minimum latency for control plane should be 10 ms in
5G are still valid for 6G while some new KPIs would be 5G and is expected to be also remarkably improved in
introduced for the assessment of new technological features. 6G. In addition to over-the-air delay, round-trip or E2E
The first eight KPIs in the following part are considered as delay is more meaningful but also complicated due to
key requirements in the definition of 5G, which are briefly the large number of network entities involved. In 6G,
introduced as follows: the E2E latency may be considered as a whole.
• Peak data rate is the highest data rate under ideal con- • Mobility means the highest moving speed of a mobile
ditions, in which all available radio resources are totally station supported by a network with the provisioning
assigned to a single mobile station. Traditionally, it is of acceptable Quality of Experience (QoE). To support
the most symbolic parameter to differentiate different the deployment scenario of high-speed trains, the high-
generations of mobile systems. Driven by both user est mobility supported by 5G is 500 km/h. In 6G, the
demand and technological advances such as THz com- maximal speed of 1000 km/h is targeted if commercial
munications, it is expected to reach up to 1 Tbps, tens airline systems are considered.

344 VOLUME 2, 2021

 • Connection density is the KPI applied for the purpose strong demand in many vertical and industrial appli-
of evaluation in the usage scenario of mMTC. Given a cations, especially in indoor environment that cannot
limited number of radio resources, the minimal number be covered by satellite-base positioning systems. With
of devices with a relaxed QoS per square kilometer the application of THz radio station, which has a strong
(km2 ) is 106 in 5G, which is envisioned to be further potential in high-accuracy positioning, the accuracy sup-
improved 10 times to 107 per km2 . ported by 6G networks is expected to reach centimeter
• Energy efficiency is important to realize cost-efficient (cm) level.
mobile networks and reduce the total carbon emission • Coverage in the definition of 5G requirement mainly
for green ICT, playing a critical role from the societal- focuses on the received quality of radio signal within a
economic respective. After the early deployment of 5G single base station. The coupling loss, which is defined
networks, there is already some complaints about its as the total long-term channel loss over the link between
high energy consumption although the energy efficiency a terminal and a base station and includes antenna gains,
per bit has been substantially improved in comparison path loss, and shadowing, is utilized to measure the
with the previous generations. In 6G networks, this KPI area served by a base station. In 6G networks, the con-
would be 10 − 100 times better over that of 5G so as notation of coverage should be substantially extended
to improve the energy efficiency per bit while reducing considering that the coverage will be globally ubiqui-
the overall power consumption of the mobile industry. tous and will be shifted from only 2D in terrestrial
• Peak spectral efficiency is an important KPI to mea- networks to 3D in a terrestrial-satellite-aerial integrated
sure the advance of radio transmission technologies. system.
The minimum requirement in 5G for peak spectral effi- • Timeliness is an emerging time-domain performance
ciencies are 30 bps/Hz in the downlink and 15 bps/Hz in requirement to future communication systems. Typical
the uplink. Following the empirical data, it is expected metrics of timeliness include the well-known age-
that advanced 6G radio technologies can achieve three of-information (AoI) [55], and its recently proposed
times higher spectral efficiency over the 5G system. variants such like age-of-task (AoT) [56] and age-of-
• Area traffic capacity is a measurement of the total synchronization (AoS) [57]. Differing from the classical
mobile traffic that a network can accommodate per memoryless metric of latency, which focuses on the
unit area, relating to the available bandwidth, spec- overall delay experienced by all data packets or service
trum efficiency, and network densification. The minimal sessions throughout their delivery process, the concept
requirement for 5G is 10 Mbps per square meter of timeliness emphasizes the freshness of the latest data
(m2 ), which is expected to reach 1 Gbps/m2 in some and service that are successfully delivered to the end
deployment scenarios such as indoor hot spots. user. This brings to the system an endogenous birth-time
discrimination against outdated data/service, as well as
In addition to the aforementioned key capabilities, there a memory to its historical state(s), and therewith raises
are several extended or novel KPIs that may be also required both the impact and the complexity of task scheduling
so as to properly evaluate the requirements of 6G. in system optimization.
• Security and privacy are necessary for assessing
• Reliability relates to the capability of transmitting a whether the operation of a network is secure enough
given amount of traffic within a predetermined time to protect infrastructure, devices, data, and assets. The
duration with high success probability. This requirement main security tasks for mobile networks are confiden-
is defined for the purpose of evaluation in the usage tiality that prevents sensitive information from being
scenario of URLLC. In 5G networks, the minimum exposing to unauthorized entities, integrity guarantee-
requirement for the reliability is measured by a success ing that information is not modified illegally, and
probability of 1 − 10−5 when transmitting a data packet authentication ensuring that the communicating par-
of 32 bytes within 1 ms given the channel quality of ties are who they say they are. On the other hand,
coverage edge for the deployment scenario of urban privacy becomes a high priority to address growing con-
macro environment. It is expected to improve at least cern and privacy legislation such as the General Data
two orders of magnitude, i.e., 1 − 10−7 or 99.999 99 % Protection Regulation (GDPR) in Europe. Some KPIs
in the next-generation system. can be applied to quantitatively measure security and
• Signal bandwidth is the maximum aggregated system privacy, e.g., percentage of security threats that are iden-
bandwidth. The bandwidth may be supported by single tified by threat identification algorithms, with which the
or multiple RF carriers. The requirement for bandwidth effectiveness of anomaly detection can be evaluated.
in 5G is at least 100 MHz, and 6G will support up to • Capital and operational expenditure is a critical fac-
1 GHz for operation in higher frequency bands or even tor to measure the affordability of mobile services,
higher in THz communications or OWC. influencing substantially the commercial success of a
• Positioning accuracy of the 5G positioning service is mobile system. The expenditure of a mobile oper-
better than 10 m. Higher accuracy of positioning has a ator can be divided into two main aspects: capital

VOLUME 2, 2021 345

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

focus group “Technologies for Network 2030” was estab-

lished under ITU-T [6]. This group intends to study and
review existing technologies, platforms, and standards for
identifying the gaps and challenges towards the capabili-
ties of networks for the year 2030 and beyond, with the
emergence of novel forward-looking scenarios such as holo-
graphic applications, Tactile Internet, multi-sense networks,
and digital twin. Although it mainly focuses on fixed data
communication networks, the vision, requirements, architec-
ture, and novel use cases identified in this group also has
reference values for the definition of the 6G mobile system.
According to the empirical timeline, ITU-R section will initi-
ate the study of 6G vision and will publish the requirements
for IMT for 2030 (as the requirements of IMT-2020 [54]
published in 2017) in the middle of the 2020s, and then
will step into the evaluation phase afterwards. Considering
the successful accomplishments by ITU for the evolution
of IMT-2000, IMT-Advanced and IMT-2020, similar actions
FIGURE 4. Quantitative comparison of the technical requirements between 5G and
are proposed for the evolution of IMT towards 2030 and
6G w.r.t eight representative KPIs. The vertices of the inner polygon stands for the beyond. At its meeting in February 2020, ITU-R working
KPIs for 5G, while the vertices of the outer octagon represent that of 6G. Different
dotted circles indicate quantities in an exponential manner instead of proportional
party 5D decided to start the study on future technology
scale, where the value in a bigger circle stands for one order of magnitude “better" trends for the future evolution of IMT [8]. It is planned to
than that of the neighboring smaller circle. For example, the minimal latency of 5G is
defined as 1 ms in comparison with 0.1 ms expected in 6G, amounting to 10 times
complete this study at the meeting in June 2022. A pre-
better, while the peak rate of 6G is envisioned to be 1 Tbps that is 50 times over 5G. liminary draft report will be developed and will consider
related information from various external organizations and
country/regional research programs. ITU-R is also responsi-
expenditure (CAPEX) that is the cost spent to build ble for organizing the world radiocommunication conference
communication infrastructure and operational expen- (WRC) that governs the frequency assignment, being hold
diture (OPEX) used for maintenance and operation. every three to four years. In WRC-19 held in 2019, the spec-
Due to the network densification, mobile operators trum allocation issue for the 5G system was determined.
suffer from a pressure of high CAPEX. Meanwhile, It is expected that the WRC probably scheduled in 2023
mobile networks’ troubleshooting (systems failures, (WRC-23) will discuss the spectrum issues for 6G and the
cyber-attacks, and performance degradations, etc.) still spectrum allocation for 6G communications may be formally
cannot avoid manual operations. A mobile operator has decided in 2017 (WRC-27).
to keep an operational group with a large number of In the early of 2019, the third generation partnership
network administrators with high expertise, leading to project (3GPP) has frozen the Release 15 (Rel. 15 or R15)
a costly OPEX that is currently three times that of specifications, which are the first phase of 5G standards.
CAPEX and keeps rising [58]. During the design of Rel. 15 mainly focuses on eMBB and provides the basis for
6G, the expenditure will be a key factor to consider. URLLC, especially in respect of the support of low latency.
To provide a quantitative performance comparison In July 2020, the subsequent release (i.e., Rel. 16) has been
between 5G and 6G, eight representative KPIs are visu- completed as the second phase of 5G standards [59]. In
alized, as shown in Fig. 4. The KPIs required to assess 5G addition to enhancements for the existing Rel. 15 features,
and 6G usage scenarios and the enabling technologies that new features such as non-public network, new radio (NR)
can support the implementation of each KPI are summarized unlicensed, NR positioning, NR-Light, and integrated access
in Table 2 and Table 5, respectively. and backhaul (IAB) have been introduced so as to fully sup-
port URLLC and industrial IoT. Currently, a more advanced
IV. ROADMAP AND EFFORTS version (Rel. 17) is being standardized by 3GPP and is
Even though discussions are ongoing within the wireless expected to be completed in the early of 2021, in spite of a
community as to whether there is any need for 6G and delay due to the COVID-19 pandemic. Driven by a multitude
whether counting the generations should be stopped at 5, of key stakeholders from the traditional mobile industry, a
a few pioneering works on the next-generation wireless wide range of verticals, and the non-terrestrial industry, it is
networks have been initiated. In this section, the up-to-date envisioned as the most versatile release in 3GPP history in
advances on 6G research from representative institutions terms of features content, including NR over non-terrestrial
and countries are summarized, while a tentative roadmap networks, NR beyond 52.6 MHz, NR sidelink enhancement,
of definition, specification, standardization, and regulation network automation, etc. 3GPP is expected to standardize
is projected, as demonstrated in Fig. 5. In July 2018, a several subsequent releases to further evolve the 5G system,

346 VOLUME 2, 2021

FIGURE 5. Predicted roadmap of research, definition, specification, spectrum regulation, development, and deployment for 6G mobile systems.

which can be called 5G+ or 5G Evolution. According to Smart Society and Ecosystem (6Genesis), which focuses on
the experiences got in previous generations, 6G will be a several challenging research areas including reliable near-
disruptive system that cannot be developed following such instant unlimited wireless connectivity, distributed computing
a backward-compatible manner. In parallel, therefore, 3GPP and intelligence, as well as materials and antennas to be
is expected to initiate the study item for 6G around the year utilized in future for circuits and devices. As early as
2025, followed by the phase of specification, to guarantee 2016, the U.S. Defense Advanced Research Projects Agency
the first commercial deployment roll-out of 6G by 2030. (DARPA), along with companies from the semiconductor
In October 2018, the European Commission has initi- and defense industries, has initiated the joint university
ated to sponsor beyond 5G research activities by opening microelectronic project (JUMP), among which the center
the ICT-20-2019 call “5G Long Term Evolution” under for converged THz communications and sensing seeks to
the eighth Framework Programmes for Research and develop technologies for a future cellular infrastructure. In
Technological Development (FP8) being named Horizon March 2019, the U.S. spectrum regulator - the Federal
2020. Eight projects such as 5G-COMPLETE [60] and Communications Commission (FCC) announced to open up
5G-CLARITY [61] have been selected from a total of 66 experimental licences for the use of frequencies between
proposals and kicked off in early 2020. In its recent call 95 GHz and 3 THz for 6G and beyond. In October 2020, the
ICT-52-2020 “Smart Connectivity beyond 5G,” the accepted Alliance for Telecommunications Industry Solutions (ATIS)
projects selected from a high-competitive evaluation process announced the launch of the “Next G Alliance” [62], an
explicitly shown that their ambition is to provide the early industry initiative that aims to advance North American
research efforts on 6G. The details of ICT-20 and ICT-52 mobile technology leadership in 6G over the next decade.
research projects are summarized in Table 3. In its upcoming Its ambition it to encompass the full lifecycle of 6G
research and innovation framework program called Horizon research and development, manufacturing, standardization,
Europe or FP9, a large number of efforts and funding will and market readiness. The founding members include AT&T,
concentrate on the research and development of 6G and will T-Mobile, Verizon, Qualcomm, Ericsson, Nokia, Apple,
be organized in the framework of Public Private Partnership Samsung, Google, Facebook, Microsoft, etc. In November
(PPP) “Smart Network & Services,” following the successful 2019, China has officially kicked off the 6G technology
strategy of the 5G Infrastructure Public Private Partnership research and development works coordinated by the Ministry
(5G-PPP) under Horizon 2020. Furthermore, the European of Science and Technology, together with five other min-
Commission has announced in February 2020 its strategy to istries or national institutions. A promotion working group
accelerate investments in Europe’s “Gigabit Connectivity” from government that is in charge of management and coor-
including 5G and 6G to shape Europe’s digital future [7]. dination, and an overall expert group that is composed of
Besides, many countries have announced and are imple- 37 experts from universities, research institutes, and indus-
menting ambitious plans to launch 6G research and devel- try were established at this event. Later, it was announced
opment initiatives. In Finland, the University of Oulu that China aims to form 6G overall development ideas by
began ground-breaking 6G research as part of Academy of the end of 2020. In November 2020, China launched what it
Finland’s flagship program [9] called 6G-Enabled Wireless

VOLUME 2, 2021 347

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

TABLE 3. Summary of EU beyond-5G and 6G research projects.

claimed is the first 6G experimental satellite to test communi- communications, VLC, OWC, and DSM, new networking
cations from space using high-frequency terahertz spectrum. that covers softwarization and virtualization, RAN slicing,
In early 2020, the Japanese government set up a dedicated O-RAN, and post-quantum security, new air interface includ-
panel including representatives from the private sector and ing massive MIMO, IRS, CoMP, cell-free massive MIMO,
academia to discuss technological development, potential and new modulation techniques, new architecture providing
use cases, and policy. Japan reportedly intends to dedicate 3D coverage by means of integrating large-scale satellite
around $2 billion to encourage private-sector research and constellation, HAP, and UAV with traditional terrestrial
development for 6G technology. In late 2020, the govern- networks, and new paradigm empowered by the conver-
ment of South Korea has confirmed a plan to carry out a gence of communication, computing, and storage resources,
6G trial in 2026 and is expected to spend approximately as well as the integration of AI, blockchain, digital twin,
$169 million over the course of five years to develop 6G and mobile networks. The principle, advantages, challenges,
technology. The trial aims to achieve 1 Tbps in data trans- and open research issues for each identified technology are
mission speeds and latency reduction to one-tenth of current introduced.
5G services. The government will initially push for tasks in
six key areas (hyper-performance, hyper-bandwidth, hyper-
precision, hyper-space, hyper-intelligence, and hyper-trust) A. NEW SPECTRUM
to preemptively secure next-generation technology. Next generation cellular networks provide a good capability
of heterogeneous radio access technology (RAT), where the
V. TECHNOLOGICAL ENABLERS legacy RAT with low radio frequencies and the line-of-sight
To well support disruptive use cases and applications, (LOS)-dependent RATs (THz, VLC, and OWC) can co-exist
advanced technologies on transmission, networking, and well. THz, VLC, and OWC may construct a new layer in
computing would be developed and then applied in the 6G the hierarchical RAN architecture (e.g., picocells), where
system. This section provides a complete view of poten- heterogeneous cells with different RAT are overlaying on
tial 6G technological enablers, which are categorized into each other. The ideology is similar to the introduction of
several groups: new spectrum consisting of mmWave, THz mmWave in the 5G networks.

348 VOLUME 2, 2021

TABLE 4. Categorized key technology enablers with advantages and challenges.

TABLE 5. Categorized key technology enablers with impact on KPIs.

1) MILLIMETER WAVE
The mmWave technology has been introduced by the 5G new
radio, and believed to remain as an essential component in

VOLUME 2, 2021 349

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

future 6G networks. Compared to legacy RF technologies bands, are expected to play an important role in the next
working below 6 GHz, it significantly broadens the avail- generation RAN, providing extremely high bandwidth.
able bandwidth with new carrier frequencies up to 300 GHz. Similar to mmWave, THz waves also suffer from high
Such a huge new bandwidth, as Shannon’s theorem reveals, path loss and therefore highly rely on directive antennas
will inflate the radio channel capacity and quench the immi- and LOS channels while providing a very limited cover-
nent thirsty for higher data rate. Meanwhile, the shorter age. However, when a satisfactory LOS link is available, the
wavelength also leads to smaller antenna size. This not only high carrier frequency brings a bandwidth that is significantly
improves the portability and integration level of device, but higher than any legacy technology, which makes it possible
also allows to increase the dimension of antenna arrays and to simultaneously provide ultra-high performance in aspects
therewith narrow the beams [63], which is beneficial to spe- of throughput, latency, and reliability. Moreover, compared
cific applications such like detection radars and physical to both mmWave systems working at lower frequencies
layer security. Furthermore, the atmospheric and molecular and wireless optical systems working in higher frequency
absorption exhibit highly variant characteristics at different bands, THz communication systems are insensitive to atmo-
frequencies across the mmWave band, providing potential spheric effects, which eases the tasks of beamforming and
for diverse use cases. On the one hand, low attenuation can beam tracking. This shapes THz communication into a good
be observed at some special bands such as 35 GHz, 94 GHz, supplementary solution in addition to the mainstream RF
140 GHz, and 220 GHz, making long-distance peer-to-peer technologies for specific use cases, such as indoor commu-
communications possible at these frequencies; on the other nications and wireless backhaul; and a competitive option
hand, severe propagation loss is experienced at some “atten- for future cyber-physical applications with extreme QoS
uation peaks” such as 60 GHz, 120 GHz, and 180 GHz, requirements, such like real-time VR/AR [70].
which can be exploited by short-range covert networks with Furthermore, the high carrier frequency also allows
stringent safety requirements [63], [64]. Currently, the stan- smaller antenna size for higher integration level. It is
dardization efforts in the mmWave field are mainly focusing expected [16] that over 10 000 antennas can be embedded
on the 60 GHz band for indoor use, e.g., the ECMA-387 [65], into a single THz BS to provide hundreds of super-narrow
the IEEE 802.15.3c [66], and the IEEE 802.11ad [67]. beams simultaneously, so as to overcome the high propaga-
Accompanying to the benefits of mmWave technologies, tion loss, and to simultaneously achieve extremely high traf-
there also come new challenges. First of all, the broad band- fic capacity together with massive connectivity, which assem-
width in mmWave band and high transmission power can ble to unlock its applications in ultra-massive machine-type
lead to severe non-linear signal distortions, which proposes communications such as Internet-of-Everything (IoE) [71].
higher technical requirements for the integrated circuits than Nevertheless, while THz outperforms mmWave in many
those for RF devices. Meanwhile, since the effective trans- ways, it also faces stronger technical challenges, especially
mission range of mmWave, especially in the 60 GHz band, is from the aspect of implementing essential hardware circuits,
severely limited by the atmospheric and molecular absorp- including antennas [72], amplifiers [73], and modulators.
tion, mmWave channels are commonly dominated by the Especially, it has been since long the most critical challenge
LOS path. This becomes a major drawback that is further for practical deployment of THz technologies, to efficiently
magnified by the poor diffraction at this short wavelength, modulate baseband signals onto such high frequency car-
which causes a strong blockage loss in scenarios with dense riers with integrated circuits. To address this issue, a great
presence of small-scale obstacles such as vehicles, pedes- effort has been made over the past decade, leading to a pros-
trians, or even the human body of user itself [68]. The perous set of developments, most of which are solid state
high propagation loss and LOS-dependency also signifi- THz systems that rely on frequency mixing, such as the
cantly raises the channel state sensitivity to the mobility, one reported in [74]. Recently, it has also been discussed
i.e., the impact of fading is much stronger than that in the to apply spatial direct modulation in THz systems, so as to
RF bands. The demand for an outstanding mobility manage- directly modulate baseband signals to THz band without any
ment becomes therefore unprecedentedly high. Furthermore, intermediate frequency stage.
in scenarios with dense links coexisting, especially for indoor
environments, the interference among different access points 3) VISIBLE LIGHT COMMUNICATIONS
will be significant, interference management approaches are VLC works in the frequency range of 400 THz to 800 THz.
therefore called for [69]. Differing from the RF technologies in lower THz range
that use antennas, VLC relies on illumination sources
– especially light-emitting diodes (LEDs) – and image-
2) TERAHERTZ COMMUNICATIONS
sensor or photodiode arrays to implement the transceivers.
Despite of its current abundance in spectral redundancy, With these transceivers, a high bandwidth can be easily
mmWave is hardly adequate to tackle down the increas- achieved with low power consumption (100 mW for 10 Mbps
ing cravenness on bandwidth for another decade. Looking to 100 Mbps) without generating electromagnetic or radio
forward to the 6G era, wireless technologies operating at interference [75]. The good power efficiency, the long life-
even higher frequencies, such as THz or optical frequency time (up to 10 years) and low cost of mainstream LEDs, in

350 VOLUME 2, 2021

addition to the unlicensed access to spectrum, makes VLC availability of optical emitters and detectors. Since the IR
an attractive solution for use cases sensitive to battery life and UV waves have a similar behavior as the visible light, the
and access cost, such like massive IoT and wireless sen- security risks and interference can be significantly confined
sor network (WSN). Moreover, VLC also exhibits better and the concern about the radio radiation to human health
propagation performance than RF technologies do in some can be eliminated. It is expected to have obvious advantages
non-terrestrial scenarios, such as aerospace and underwater, in deployment scenarios such as vehicular communications
which can be important part of the future 6G ecosystem, as in smart transportation systems, airplane passenger lighting,
we will introduce later in Section V-D. and medical machines that are sensitive to electromagnetic
Compared to RF, the MIMO gain in VLC, especially interference. Despite its advantage, OWC suffers from the
in indoor scenarios, is very frail. This roots in the high impairments such as ambient light noise, atmospheric loss,
coherence among the propagation paths, i.e., the low spatial nonlinearity of LEDs, multi-path dispersion, and pointing
diversity. Though this coherence can be somehow reduced errors [86].
by using spaced LED arrays [76], MIMO-VLC is also In OWC, LED or laser diode (LD) is applied to convert
challenged by the design and implementation of receivers: an electrical signal to an optical signal at the transmitter
non-imaging receivers are extremely sensitive to their spa- and the receiver uses a photodiode (PD) to convert the
tial alignments with the transmitters, while imaging receivers optical signal into electrical current. The information is
are not applicable in cost-critical use cases for their high conveyed by modulating the intensity of optical pulse sim-
prices [77]. Therefore, there has been so far no MIMO ply through widely-used scheme such as on-off keying or
method standardized into the mainstream VLC physical layer pulse-position modulation, as well as advanced multi-carrier
of IEEE 802.15.7, despite of the persistent efforts made in schemes [87] such as OFDM to get higher transmission rate.
academia since a decade [78], [79], [80]. Therefore, the To support multiple users in a single optical access point,
beamforming in VLC, differing from the MIMO-based RF OWC can apply not only typical electrical multiplexing
beamforming, is implemented by a special optical device technologies such as time-division, frequency-division, and
known as spatial light modulator (SLM) [81]. code-division multi-access, but also optical multiplexing such
Similar to mmWave and THz technologies, VLC also as wavelength-division multi-access. Optical MIMO technol-
relies on LOS channels, since it has neither ability to ogy [78] is also implemented in OWC, where multiple LEDs
penetrate, nor sufficient diffraction to bypass common obsta- and multiple PDs are applied, in contrast to the multiple
cles. Meanwhile, due to the concerns about adjacent cell antennas in a typical MIMO system operating in the RF
interference and almost ubiquitous environment light noise, band.
VLC systems generally require directive antennas with nar- The optical system applying image sensors to detect the
row beams. These facts make VLC systems highly sensitive optical pulse also called optical camera system. The imagine
to the position and mobility of users, leading to a high sensor can convert the optical signal into the electrical sig-
requirement of beam tracking. On the other hand, this feature nal, which has the advantage of easier implementation due
can be also exploited for advantages in certain use scenar- to the wide spread of camera-embedded smart phones. On
ios, such like better accuracy in indoor positioning [82], and the other hand, terrestrial point-to-point OWC also known as
lower interference in vehicular communications [83]. free-space optical (FSO) communications [88] take place at
Another key technical challenge for VLC roots in the the near IR band. Using high-power high-concentrated laser
open and unregulated (to be more specific, unregulatable) beam at the transmitter, the FSO system can achieve high
access to visible light spectrum, which implies a higher data rate, i.e., 10 Gbps per wavelength, over long-distance (up
security risk and calls for a more stringent security require- to 10 000 km). It offers a cost-effective solution for the back-
ment in VLC systems, when compared to legacy cellular haul bottleneck in terrestrial networks, enables crosslinks
systems in licensed RF bands. Regarding this, physical layer among space, air, and groud platforms, and facilitates high-
security has been extensively investigated as a promising capacity inter-satellite links for the emerging LEO satellite
solution [84]. constellation. Furthermore, there has also been a growing
interest on ultraviolet communication [89] as a result of
4) OPTICAL WIRELESS COMMUNICATIONS recent progress in solid state optical transmitter and detector
OWC points to wireless communications [85] that use for non-LOS UV communications that offer broad coverage
infrared (IR), visible light, or ultraviolet (UV) as transmis- and high security.
sion medium. It is a promising complementary technology
for traditional wireless communications operating over RF 5) DYNAMIC SPECTRUM MANAGEMENT
bands. OWC systems operating in the visible band are com- Alongside the continuous mining of unused spectrum at ever-
monly referred to as VLC, which attracts much attention higher frequencies, there is a second approach towards our
recently and is discussed separately in Section V-A3. The vision of bandwidth prosperity in 6G: to improve the radio
optical band can provide almost unlimited bandwidth with- resource utilization rate by DSM.
out the permission from the regulators worldwide. It can be The idea of DSM dates back to the well-known listen-
applied to realize high-speed access at low cost thanks to the before-talk (LBT) protocol applied in IEEE 802.11, which

VOLUME 2, 2021 351

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

treats all users equally in a contention-based control of B. NEW NETWORKING

access to the spectrum, and one can exploit a band only To deploy the aforementioned use cases utilizing the dis-
when it is not occupied by the others. In the unlicensed cussed key enabling technologies, the network infrastruc-
industrial, scientific and medical (ISM) band, LBT has ture is expected to be flexible, intelligent, and open to
demonstrated a great success in collision and interference multi-vendor equipment and multi-tenancy. To that aim, soft-
control. Meanwhile, in the licensed spectrum, as reported by warization and virtualization of 6G network is the first and
the FCC of the U.S., it is the under-utilization of spectrum foremost step. Based on such architecture, the network slic-
caused by regulated access “a more significant problem than ing and resource isolation in both core and RAN domains
the physical scarcity of spectrum” [90]. This fact has raised are considered as novel mechanisms to decrease complex-
intense research interest on the topic of LBT-like dynamic ity in the management and orchestration operations. These
spectrum sharing among various systems with heterogeneous aspects, along with their privacy and security concerns, of
RATs and different priorities to access licensed/unlicensed 6G mobile networks are discussed in this subsection.
bands. Driven by the successful development of software-
defined radio technologies, these research efforts gave birth 1) SOFTWARIZATION AND VIRTUALIZATION
to the technology of cognitive radio (CR) [91], which became Most of the network functions of the 5G core network
rapidly mature in the first decade of this century. Since (5GC) and next-generation RAN (NG-RAN) architecture are
the LTE era, it has become a distinguished topic in the virtualized using network function virtualization (NFV) tech-
field of wireless networks to study the DSM in coexis- nology, which provides high flexibility in order to adapt to
tence of licensed cellular systems and unlicensed ISM-band various scenarios, requirements, and use cases of next gen-
technologies [92]. eration communication systems [95]. In practice, however,
Regarding the future 6G systems, the demand for DSM NFV faces some critical challenges that are required to be
is becoming unprecedentedly imperative. On the one hand, solved, such as the increased amount of virtual network
radio access to ISM bands (especially the IEEE 802.11 functions (VNFs), the various requirements of the different
bands) is becoming nowadays almost a standard functional- tenants, the resources orchestration of the VNFs in a shared
ity of mainstream cellular terminals, making it a universal infrastructure, the complexity in management and orches-
solution to provide extra network capacity in scenarios with tration operations, and others [96]. Therefore, the allocation
dense users. On the other hand, since it is impossible to of virtual network resources, the management and orches-
reserve the broad bands of 6G new spectrum for licensed tration of the VNFs in a multi-tenant environment require
use only (saliently the visible light spectrum), 6G system is cutting-edge tools and promising solutions to be addressed,
foreseen to suffer in the unlicensed part of its spectrum from such as the AI techniques [58] and ML algorithms [97].
the ubiquitous existence of interference by other systems In this context, the European Telecommunication Standards
and environmental noises, which are highly dynamic and Institute (ETSI) Industry Specification Group (ISG) intro-
environment-dependent. 6G systems therefore must be able duced the Experiential Network Intelligence (ENI) working
to dynamically and cognitively select the most appropriate group in order to improve the experience of network oper-
operating band with respect to the instantaneous situation. ators and add value to the teleco provided services [98].
There are numerous technical challenges that 6G DSM The main objectives of the ENI is to exploit AI and ML
is facing. In perspective of the hardware implementation, it techniques in order to adjust the VNFs of the networked
is mainly troubled by the broadness of 6G new spectrum services based on dynamic changes in the requirements of
that leads to a difficulty in designing transceivers capable of the end users, the conditions of the environments, and the
dynamic all-spectrum sensing [93]. The 6G front-end must goals of the business. The current specification of the ENI
be capable to carry out a rapid and power-efficient spectrum is in its initial phase. Further work is needed to explore the
sensing across the ultra-wide 6G bands, so as to enable deployment of the AI and ML in the NFV of the beyond
an online radio environment cognition and a timely adap- 5G and 6G mobile networks.
tation to spectrum access. On the system level, for better The software-defined networking (SDN) is considered as
efficiency and safety of the DSM, the physical layer CR one of the most critical key enablers of 5G mobile networks,
based on spectrum sensing shall be further completed by integrated with NFV capabilities, they are able to offer high
an awareness of cyber-physical level context information, in flexibility in network management and achieve high effec-
order to obtain a deeper understanding in the communication tiveness in service modularity [95]. Based on its integral role
environment, including terrain scenario, traffic pattern, local and performance in the 5G mobile network, the SDN is going
regulation, etc. This leads to challenges in various aspects to be indispensable for the evolution of beyond 5G and 6G
of context-awareness, ranging from data provisioning to data mobile networks as well and will continue to play a crucial
ownership [94]. role within their management, orchestration, architectures,
etc., [14]. Despite its numerous advantages in 5G, the SDN
technology has a number of key research problems that are
needed to be addressed in order to thwart full exploitation
of its potentials in the future generation of communication

352 VOLUME 2, 2021

networks. These key challenges that continue to confront the NG-RAN performance, and maintain a significant level
SDN in the wake of 6G include, but not limited to: effec- of isolation and customization among RAN slices of dif-
tively resolving the optimal placement of the SDN controller ferent types while considering the metrics of service level
in the future networks [99], [100], efficiently maintaining agreements between the mobile operator and tenants [113].
the end-to-end and timely global views of the dynamic The NG-RAN is expected to support a massive amount of
network topology and its links states [101], exploiting AI/ML RAN slice subnets. Each type of RAN slice subnet fulfills
techniques for automated network management [102] and the service requirements of a singe type of use case of a sin-
orchestration [103], and traffic engineering with guaranteed gle tenant. Nevertheless, there is a large number of vertical
stringent QoS requirements of heterogeneous services [104]. industries – such as automobile, manufacturing, power grid,
and others – which are consisting of multiple use cases [111].
2) RAN SLICING Providing RAN slice subnets for multiple use cases consist-
Slicing the RAN architecture, using SDN and NFV tech- ing of vertical industries is a crucial research problem that
nologies, is an emerging research direction towards cloudifi- is needed to be addressed in the next generation of mobile
cation, virtualization, and centralization of RAN resources in networks. One of the crucial parts of this research problem
beyond 5G and 6G mobile networks [105]. The slicing-aware is to design an extensive architectural framework for the 6G
NG-RAN architecture helps mobile operators to efficiently RAN domain in order to support the RAN slice subnets of
slice the entire infrastructure (or a part of it) based on the multi-use-case verticals. This management and orchestration
requirements of end users and vertical industries [106]. Such framework shall effectively manage the CU, DU, and RU of
a classification is also applicable in terms of slicing the per-vertical per-use-case RAN slices in 6G RAN using AI
NG-RAN into eMBB, URLLC, and mMTC subnets. The and ML.
resources required by these three types of RAN slices are
divided into physical and virtual resources. The physical 3) OPEN-RAN
resources are managed by the 3GPP network slicing manage- The key concepts of O-RAN, including its vision, architec-
ment system [107] whereas the virtual resources are managed ture, interfaces, technologies, objectives, and other important
by the ETSI network function virtualization-management and aspects, were introduced for the first time by the O-RAN
orchestration (NFV-MANO) [108]. alliance in a white paper [114]. The O-RAN alliance has then
Towards an efficient RAN slicing, currently, some of the further studied the use cases leveraging the O-RAN archi-
radio processing functionalities of a next-generation NodeB tecture to demonstrate its capability in real time behavior
(gNB) in the NG-RAN are accommodated as VNFs [109], in [115]. The main objective of the openness and intelli-
namely the centralized unit (CU) and distributed unit (DU), gence in RAN architecture is to build a radio network that
and some others are distributed as physical network function is resource efficient, cost effective, software-driven, virtual-
(PNF), namely the radio unit (RU). The VNFs run on points ized, slicing-aware, centralized, open source, open hardware,
of presence (PoPs) [110], while the PNFs are implemented intelligent, and therefore more flexible and dynamic than
on dedicated hardware in the cellular network sites. In the any previous generation of mobile networks. To do so, the
next generation of mobile networks, the full virtualization of research community has introduced the utilization of AI and
CU and DU, and the partial/full virtualization of RU, can lead ML techniques on every single layer of the RAN architecture
to a number of advantages, such as increased performance to fulfill the requirements of dense network edge in beyond
of the RAN architecture, deployment of RAN slice subnets, 5G and 6G mobile communication systems.
decreased network expenditure, simplified operations and By opening up the RAN from a singe vendor environ-
management of the network, etc., [111]. We, thus, firmly ment to a standardized, open, multi-vendor, and ML and
believe that further studies are needed to address the vir- AI powered hierarchical controller structure, it allows third
tualization of RU functionalities towards a virtualized and parties and mobile operators to dynamically deploy inno-
slicing-aware RAN of the 6G mobile networks [112]. vative applications and emerging services that cannot be
Focusing on the function split in the NG-RAN, the current deployed or supported in legacy RAN architectures. In addi-
distribution of the radio processing functions over gNB com- tion, the O-RAN is built upon the NFV-MANO reference
ponents is executed statically without taking into considera- architecture proposed by ETSI, which deploys commercial
tion the type of RAN slices. When supporting a large amount off-the-shelf hardware components, virtualization techniques,
RAN slices in types of eMBB, URLLC, and mMTC, the and software pieces. The virtual machines abstracted (or vir-
one-size-fits-all function splitting architecture in NG-RAN tualized) from the underlying physical resources are easily
is not efficient in terms of resource allocation and network created, deployed, configured, and decommissioned. Such a
performance [109]. We strongly believe that a customized virtualized environment and virtual resources, therefore, do
and dynamic distribution of the gNB functionalities is needed not only bring flexibility to the O-RAN architecture, but
in order to satisfy the service requirements of the three afore- also decrease the CAPEX/OPEX and energy consumption
mentioned types of RAN slices in future networks. Such a towards 6G communication networks.
dynamic distribution of gNB functionalities shall improve Despite the flexibility and interoperability that O-RAN
the utilization of physical and virtual resources, enhance offers to mobile operators, it also has a number of key

VOLUME 2, 2021 353

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

problems that require further research efforts for its full real- search for all possible keys can be substantially accelerated
ization in future mobile networks, such as the convergence by Grover’s algorithm, making such cryptosystems insecure
of different vendors’ technologies and various operators on once quantum computers come. In an asymmetric cryp-
the same platform, the harmonization of different manage- tosystem, a public key is applied to encrypt messages by
ment and orchestration frameworks to deliver enhanced QoE, anybody while only the owner of the corresponding private
lacked standards for the validation and troubleshooting of key can decrypt cipher-texts. The public key schemes are
performance issues related to the network, and others. To also employed for the implementation of digital signature,
overcome these challenges, researchers from industry and where a signature is generated from the private key while
academia are also expected to take part in theoretical analy- everyone can utilize the public key to validate this signature.
sis and practical roll-outs of this technology towards an open Current asymmetric cryptography schemes such as Rivest-
and intelligent RAN for 6G mobile networks. Shamir-Adleman (RSA), elliptic curve cryptosystem (ECC),
and digital signature algorithm (DSA) are based on the hard-
4) POST-QUANTUM SECURITY ness of solving some number-theoretic problems such as
In 5G communication systems, security and privacy are integer factorization and discrete logarithms. However, math-
considered as the most critical components for business ematician Peter Shor [117] revealed that quantum computers
continuity. This issue has even been raised to the stage can solve such problems efficiently, which makes all these
of international politics, for example, some countries are public-key schemes completely broken.
proposing sanctions to ban 5G hardware and software from It is still unclear when practical quantum computers will
certain vendors, claiming “to protect their networks and be available, but recent advances in quantum technologies
citizens”. Focusing on the technical aspect of the 5G secu- demonstrate the urgency of exploring post-quantum secu-
rity, together with user equipment, the rise of new business rity for communication networks. Since 6G networks will
and innovations such as vertical industries, mobile virtual be deployed around 2030 and will last for several decades,
network operators, and multi-tenancy, have put a burden on long-term threats arising from quantum computing could be
the commercial deployment of 5G networks. For instance, seriously taken into account during the design and imple-
mMTC types of communication services require lightweight mentation of 6G systems. Consequently, the research and
security components while eMBB and URLLC types of com- development of quantum-resistant cryptographic algorithms
munication services demand high efficient security schemes. and technologies, also called post-quantum cryptography,
Another example is multi-tenant aspect, which implies the which are secure against both quantum and classical com-
lack of a central authentication server so that the subscribers’ puters play a vital role for the success of 6G. According to
identities must be confirmed in either a decentralized or the initial recommendation of NIST [118], the cryptographic
a collaborative fashion. It is envisioned that 6G systems schemes based on lattices, codes, hash, and multivariate
will encounter more challenging security problems over cur- polynomials could be usable in the quantum era.
rent 5G systems. Many cutting-edge mechanisms are under Moreover, another quantum technology called quantum
exploration to meet the requirement of high security and pri- communication [119] can significantly enhance the security
vacy in the next-generation networks, such as E2E security, level of data transmission, which might have application
distributed authentication, and deep learning-based intrusion potential in 6G networks. In terms of the laws of quantum
detection. To ensure the E2E security for 6G, the deploy- physics, if an adversary eavesdropper measures the state
ment of the AI techniques can play a significant role in the of superposition from particles, typically photons of light
design, implementation, and optimization of security proto- for transmitting data along optical fibers, their super-fragile
cols in order to protect network, user equipment, and vertical quantum state “collapses” to either 1 or 0. Since a single
industries from unauthorized access and threats. To allow for particle carrying qubits is inseparable, the eavesdropper can-
trustworthy among communication participants, distributed not replicate it. It means that the eavesdropper can’t tamper
authentication relevant to users authenticating against gNB, with the qubits without leaving behind a telltale sign of the
and among network functions (RU, DU, and CU), will be activity. In theory, quantum communication could provide
designed by leveraging blockchain technology. absolute security and offer new solutions to a high level of
In addition to conventional threats, the rise of quantum security that traditional communication systems are unable
computing imposes a big challenge on network security. The to implement [120].
cryptosystems used today can be divided into two categories:
symmetric and asymmetric [116]. The former shares a com- C. NEW AIR INTERFACE
mon secret key between two communicating parties, and The union of OFDM and small-scale active MIMO antenna
a message is encrypted at the sender and is decrypted at arrays has ruled the world of cellular radio access network
the receiver using this key. An example of symmetric cryp- over the entire era of 4G, and is still continuously show-
tography that is widely used today is advanced encryption ing its dominance in recent progresses of 5G. However, the
standard (AES), which has been standardized in 2001 by grandiose march towards ever-higher carrier frequency, as
National Institute of Standards and Technology (NIST) for we have seen in Section V-A, is squeezing the last drop
confidentiality and integrity. The process of an exhaustive of technical potential from this air interface. Facing new

354 VOLUME 2, 2021

challenges such as high propagation loss and low NLOS also introduces new challenges, such as higher propagation
path diversity, various emerging technologies are expected loss, lower diffraction and more blockage. In the frequency
to play their key roles in the evolution towards the next gen- range of mmWave, massive MIMO has been proven effec-
eration air interface, which shall be capable to fully exploit tive in realizing active beamforming to provide high antenna
the advantages of 6G new spectrum and support future use gain for overcoming the channel loss. Nevertheless, its capa-
cases with extreme performance requirements. This evolu- bility can be insufficient for the future 6G new spectrum,
tion is primarily expected to make several significant shifts as we have discussed earlier in Section V-A3. Among all
and extensions by means of MIMO: from small-scale to potential candidate solutions to enhance current beamform-
massive, from active antenna to passive reflective surface, ing approaches, the technology of IRS has been widely
from physical layer to network layer. New modulation and considered promising for 6G mobile networks.
multiplexing schemes, as complements, are also arising on The so-called IRS, a.k.a. reconfigurable intelligent sur-
the horizon of the next decade. faces (RIS) [122], is assembled by a category of pro-
grammable and reconfigurable material sheets that are
1) MASSIVE MIMO capable of adaptively modifying their radio reflecting char-
In legacy cellular networks the MIMO is divided into two acteristics. When attached to environmental surfaces, e.g.,
types: the point-to-point MIMO and the multi-user MIMO. walls, glass, ceilings, etc., IRS enables to convert parts of
In the former, multiple antennas are installed in both user the wireless environment into smart reconfigurable reflec-
equipment and base station, however a single user equipment tors, known as smart radio environment (SRE) [123], and
is served at a single time. In the latter, an antenna array is therewith to exploit them for a passive beamforming that can
installed in the base station which provides connection to significantly improve the channel gain, at low costs of imple-
many user equipment under its respective coverage area. In mentation and power consumption in comparison to active
order to further enhance user experience, increase through- massive MIMO antenna arrays. Moreover, unlike antenna
put, and scale up the statistical multiplexing gain, the concept arrays that must be compact enough for integration, SREs
of massive MIMO was introduced in order to address the are implemented on large-size surfaces apart from the UEs,
shortcomings of conventional multi-user MIMO [95]. Since making it easier for them to realize accurate beamform-
then, the massive MIMO has been considering one of the ing with ultra-narrow beams, which are essential for some
key enablers of the legacy wireless communication systems. applications such like physical layer security. Furthermore,
In addition, the massive MIMO is also expected to provide a unlike active mMIMO antenna arrays that must be specif-
significant increase in the system capacity, higher statistical ically implemented for every individual RAT, the passive
multiplexing gain, spectral efficiency, lower CAPEX/OPEX, reflection mechanism that IRS is relying on works almost
decreased energy consumption, and many other advantages universal for all RF and optical frequencies, which is espe-
in beyond 5G and 6G cellular networks. cially cost beneficial for the 6G systems that work in an
To commercially deploy massive MIMO, many operators ultra broad spectrum.
have configured base stations with 64 fully digital transceiver Though IRS is showing a great technical competitiveness
chains which have proved its realization in the 5G mobile in context of the 6G new spectrum, it still lacks mature tech-
networks [121]. These rollouts proved that limitations due niques for accurate modeling and estimation of the channels
to pilot contamination have addressed and for a better spec- and the surface themselves, especially in the near-field range.
tral efficiency the relevant signal processing methods have Moreover, a commercial deployment is only possible after
been developed and deployed. To pave the way towards addressing the business concern, that IRS relies on exter-
the realization of massive MIMO in beyond 5G and 6G nal assessments such like buildings that do not belong to
mobile networks, the authors in [121] and [33] have outlined the MNOs. Therefore, it calls for a thoughtful design and
a number of research challenges including the (i) deploy- standardization of framework providing essential interfaces,
ment of extremely large aperture arrays; (ii) limitation in agreements, and signaling protocols, so that 6G operators
channel prediction; (iii) implementation of intelligent envi- become capable to widely access and exploit IRS-equipped
ronment aware adaptation; (iv) fundamental limits of wireless objects in public and private domains.
communication with holographic Massive MIMO; (v) six-
dimensional positioning; and (vi) large-scale MIMO radar. 3) COORDINATED MULTI-POINT AND CELL-FREE
These research problems give a chance to researchers work- CoMP refers to a class of technologies that allow multiple
ing in academia, industry, and standardization organizations access points to jointly serve multiple mobile stations, so
to concentrate their attention towards further improvement that a network layer MIMO can be realized to increase the
in the realization of massive MIMO in next generation of spatial diversity on the top of classical physical layer MIMO
communication networks. approaches. Therefore, it is also known as network MIMO
or cooperative MIMO. CoMP was initially introduced by
2) INTELLIGENT REFLECTING SURFACES 3GPP in its Release 11 [124] for LTE Advanced systems.
While releasing a significant bandwidth to support high With recent evidences of its potentials in mitigating down-
throughput, the use of high frequency bands over 10 GHz link inter-cell interference and joint user detection in uplink,

VOLUME 2, 2021 355

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

CoMP is expected to play an important role in 5G [125]. 6G systems, which relies on AI techniques to efficiently
In the upcoming 6G era, regarding the new spectrum over solve the problem of online MIMO precoding and resource
10 GHz, the CoMP technologies that make use of base sta- mapping [16].
tion level diversity will become an important complement Alongside with a further evolution in OFDMA technolo-
to the traditional antenna-level spatial diversity, as the lat- gies, non-orthogonal multiple access (NOMA) technologies
ter can be minimized by the dense blockage phenomena in are also widely considered as an answer to the new chal-
high-frequency bands. lenges in the next generation of mobile communication
Furthermore, since CoMP is generally suggesting every networks. In contrast to OMA, NOMA allow multiple users
UE to simultaneously hold multiple links to different access to reuse the same PRB, which can be achieved by complex
points (even when they are of the same RAT), it reveals the inter-user interference cancellation. NOMA approaches can
feasibility of a novel “cell-free” RAN architecture, where be generally divided into two categories, namely the power-
numerous single-antenna access points distributed over the domain (PD) NOMA and the code-domain NOMA. While
coverage area are connected to a central processing unit, PD-NOMA has been recently proposed and is attracting
and jointly serving all UEs by coherent transmission in a a lot of research interest in context of 5G [131], code-
CoMP fashion [126]. Recent study has shown that such domain NOMA has a longer history in legacy systems (e.g.,
cell-free massive MIMO are able to outperform traditional CDMA in 3G), and provides an alternative to PD-NOMA
cellular massive MIMO while also reducing the fronthaul with numerous variations, such as trellis-coded multiple
signaling [127]. access (TCMA) [132], interleave division multiple access
As a trade-off for the performance gain originated from its (IDMA) [133], multi-user shared access (MUSA) [134],
nature of cooperative decoding, CoMP also has to face some pattern-division multiple access (PDMA) [135], and sparse-
key technical challenges caused by the same reason. First of code multiple access (SCMA) [136].
all, the performance of CoMP highly relies on the cluster- Since the beyond 5G and the 6G networks are expected to
ing of cooperating base stations, so an appropriate clustering simultaneously manage a massive amount of links, e.g., in
scheme must be found, which has been a focus of research the mMTC scenario and its future extensions, NOMA solu-
over the past years [128]. Second, the synchronization among tions appear promising since they provide higher bandwidth
cooperating base stations have to be accomplished without efficiency than OMA approaches. Recent studies have also
inter-carrier and inter-symbol interference [129]. The chan- demonstrated that NOMA can be effectively exploited in new
nel estimation and equalization also must be carried out spectrum, including mmWave, THz, and optical frequencies.
in a inter-BS-coherent manner, which greatly increases the Additionally, when deployed together with CoMP, NOMA
computation complexity. has been proven as capable to outperform CoMP-OMA in
both power efficiency and spectral efficiency.
4) NEW MODULATION Being completely based on successive interference can-
The 3GPP LTE-Advanced (LTE-A) networks are imple- cellation, NOMA has a significantly higher complexity in
mented based on orthogonal frequency division multiple its receiver design than OMA, which increases in polyno-
access (OFDMA) [130], which is a typical instance of mial or even exponential order along with the number of
orthogonal multiple access (OMA) technologies prohibiting users. Especially, in some scenarios that require cooperative
physical resource block (PRB) sharing by multiple users. In decoding across different UEs, specific D2D interfaces must
comparison to CDMA, which is deployed in 3G systems, be reserved for this functionality, and security/trust concerns
OFDMA shows a conspicuous superiority in combating shall be taken into account, to enable the deployment of
multi-path fading by simple and robust carrier-based chan- NOMA in 6G.
nel equalization. Furthermore, when combined with MIMO,
OFDMA is capable to overwhelmingly outperform CDMA
in spectral efficiency. Nevertheless, the full performance of D. NEW ARCHITECTURE
MIMO-OFDM highly relies on the MIMO precoding and So far, all legacy and existing cellular systems have been
resource mapping, which have to be precisely adapted to the designed to substantially rely on terrestrial base stations. For
channel condition to achieve the optimum. As the dimension marine, oceanic, as well as wild terrestrial areas, which are
of MIMO increase, from up to 8 × 4 in LTE-A gradually impossible or economically challenging to be covered by ter-
to over 256 × 32 massive MIMO, and eventually to the restrial cellular networks, satellites have been since long the
future ultra-massive MIMO (e.g., 1024 × 64), the complex- most common communication solution. Aiming at a better
ity of MIMO-OFDM adaptation is dramatically increasing. coverage rate, deployment of non-terrestrial infrastructures
Meanwhile, in response to the demand of supporting higher as part of the 6G network is being treated as an emerging
mobility – which implies higher fading dynamics – the topic, known as the integrated space and terrestrial network
computation latency constraint to this online adaptation pro- (ISTN). An ISTN is expected to consist of three layers: the
cedure is also becoming more and more strict. To cope with ground-based layer constructed by terrestrial base stations,
these emerging challenges, a new architecture of AI-driven the airborne layer empowered by HAP and UAV, and the
MIMO-OFDM transceivers has been proposed towards future spaceborne layer implemented by satellites. An envision of

356 VOLUME 2, 2021

the architecture for the 6G systems, demonstrating represen- services [50]. The U.S. FCC has approved its first-stage
tative use cases and key technological enablers, is illustrated plan to launch 12 000 satellites and another application for
in Fig. 3. deploying 30 000 additional satellites is under considera-
tion. With the advancement of electronic technology, the
1) LARGE-SCALE SATELLITE CONSTELLATION weight of each satellite reduces to approximately 260 kg and
Until now, the coverage of terrestrial networks has only a compact, flat-panel design minimizes the volume, allow-
reached a small portion of the whole surface of the globe. ing for a dense launch of 60 Starlink satellites to make full
First, it is technically impossible to install terrestrial base use of the capability of Falcon 9. Each satellite becomes
stations for offering large-scale coverage in ocean and cheap due to the massive production, in combination with
desert [51]. Second, it is difficult to cover extreme topogra- reusable rockets, building a large-scale constellation becomes
phies, e.g., high mountain area, valley, and cliff, while it feasible from both commercial and technological perspec-
is not cost-efficient to use a terrestrial network to provide tives. Since its first launch in May 2019, more than 1000
services for sparsely-populated areas. Additionally, terres- satellites have been successfully deployed through 17 times
trial networks are vulnerable to natural disasters such as launches and the speed of deployment is planned to reach
earthquake, flood, hurricane, and tsunami, where there is 120 satellites per month. In October 2020, SpaceX started
a vital demand of communications but the infrastructure is to invite some early users to join public testing and it was
destroyed or in outage of service. With the expansion of reported that data rate vary from 50 Mbit/s to 150 Mbit/s and
human activity, e.g., the passengers in commercial planes and latency from 20 ms to 40 ms can be expected. Considering
cruise ships, the demand of MBB services in uncovered areas its announced monthly fee of 99 USD for active services and
increasingly grows. Also, the connectivity demand of IoT its steadily improvement of capacity and performance with
deployment scenarios like wild environmental monitoring, the increasing number of deployed satellites, its impact on
Offshore wind farm, and smart grid require wide-area ubiq- terrestrial networks should be seriously taken into account.
uitous coverage. Satellite communications have been since Looking forward to a future global ubiquitous converge that
long the most common solution for wide coverage but cur- is available anywhere and anytime, it is strongly suggested
rently the mobile communication service offered by GEO to integrate satellite networks into the 6G network as part
satellites is costly, low data rate, and high latency due to of it.
the expensive cost for launching and its wide area coverage
(1/3 surface of Earth per GEO satellite). 2) HIGH ALTITUDE PLATFORM
The satellites in LEO [52] have some advantages over In general, terrestrial networks and satellite communications
GEO satellite for providing communication services. A LEO are two technologies that dominate mobile communications
satellite operates in an orbit generally lower than 1000km, for long years. HAP, a quasi-stationary aerial platform oper-
which can substantially lower the latency due to the signal ating in the stratosphere at an altitude between 17 km to
propagation compared to the GEO satellite in the orbit of 22 km above the Earth’s surface represents a new alternative
36000km. Meanwhile, the propagation loss of LEO is much to provide a multitude of telecommunication services in a
smaller, facilitating the direction connectivity to mobile and cost-efficient way [139]. In comparison with terrestrial base
IoT devices that are strictly constraint by the battery supply. stations, HAP can cover a larger area, offer unobstructed
Moreover, a stationary ground terminal like an IoT device connectivity with high signal arrival angle, and provide the
mounted in a monitoring position may suffer from an obsta- flexibility of quick deployment with less temporal and spatial
cle in the line of sight from GEO. There is an early attempt to constraints [140]. Compared to satellite systems, HAP has
implement a global satellite mobile communication system, the following advantages: much lower cost of implementa-
i.e., the Iridium constellation [137] that became commer- tion and deployment due to the avoidance of space launching,
cially available in November 1998. It consists of 66 LEO the possibility of upgrading, repairing, and redeployment,
satellites at an altitude of approximately 781 kilometers and and much shorter propagation distance that corresponds to
provides mobile phone and data services over the entire Earth higher signal strength and lower latency [141].
surface. Even though it fails due to expensive costs and lack An aerial platform can carry on-board base stations of
of demands at that time, it is a great technological break- both terrestrial and satellite segments, providing connectiv-
through. It still operates today and the second generation ity to the terminals of end users transparently. It can keep
Iridium system has been successfully deployed last year. quasi-stationary, be redeployed, or be moved from one site to
In recent years, the high-tech company SpaceX gains another on demand. This is therefore an efficient solution to
a lot of attention due to its revolutionary development of improve the coverage of terrestrial and satellite systems. The
space launching technologies. Its reusable rocket namely aerial platform is not only a host of communication services.
Falcon 9 dramatically lower the cost of space launching, It well adapts to different applications such as high-definition
opening the possibility for deploying large space infrastruc- multimedia broadcasting, remote sensing, surveillance, intel-
tures. In January 2015, SpaceX announced its ambition called ligent transportation, and environmental monitoring. It is also
Starlink [138], utilizing a very large-scale constellation with applied for navigation and positioning, working standalone
thousands of LEO satellites to provide global Internet access or as a local enhancement component of Global Navigation

VOLUME 2, 2021 357

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

Satellite System, for a higher accuracy and better availabil- 6G. With its well developed integration with multimedia
ity [142]. Last but not least, it has unique value in local devices such as video camera and microphone, UAV has
temporary events, emergence communications, and natural since long been widely used in offline photographing and
disaster (e.g., earthquake, flood, and tsunami) relief, where videographing. By nature it is promised to play a larger role
the terrestrial infrastructure is destroyed or the power grid in future online video streaming, and can be enabling remote
is in outage of services. sensing and multi-sense experience when equipped with vari-
In 6G systems, the synergy among terrestrial, satellite, and ant sensors. The high flexibility, mobility and continuously
HAP is worth exploring to provide ubiquitous, robust, and decreasing load cost of UAVs also lead to their foreseeable
resilient network infrastructure and communication connec- deployment in future intelligent logistic applications.
tivity. For example, using HAP to provide high-throughput
backhaul links for the small cells deployed in the sites that E. NEW PARADIGM
are hard or expensive to provide wired links, or applying The recent revival of AI technology spurs the discussions
HAP to enhance or relay the satellite signals. From the per- of whether 6G will be an integrated system of AI and
spective of the integrated terrestrial-HAP-satellite systems, mobile networks. The 6G system is expected to support
there are several technical challenges to be solved such the upsurge of diversified mobile AI applications, and in
as the power supply of the aerial platform, the stabil- turn AI will play a critical role in designing and opti-
ity of antenna array, channel models, seamless handover, mizing the wireless architecture [150]. The similar trends
admission control, interference management, etc. To this are happening in other fields like blockchain and digital
end, extensive research on the topics such as 3D channel twin, which are recognized as strong drivers to shape the
modeling [143], advanced multi-antenna technologies [144], next generation mobile system. It is foreseen that 6G will
spectrum-awareness, dynamic spectrum management, and transform into a huge computer, which converges distributed
FSO [145] has been carried. communication, computing, storage, sensing, and controlling
resources for provisioning services of pervasive computing,
3) UNMANNED AERIAL VEHICLE AI, blockchain, digital twin, etc.
Alongside the space satellite and HAP, UAV also play an
indispensable role, as the last piece of the puzzle that fills 1) ARTIFICIAL INTELLIGENCE
the gap of near-Earth altitude, in the foreseen ISTN. Over On the list of 6G enabling technologies, AI is recognized
the past years, it has been widely discussed to use UAVs as the most potential one. As mobile networks are increas-
in cellular networks as flying BSs or relays [146], [147]. ingly complex and heterogeneous, many optimization tasks
Generally, it makes a flexible mobile supplement to the become intractable, offering an opportunity for advanced ML
fixed terrestrial gNBs and space satellites, offering a possi- techniques. Categorized typically into supervised, unsuper-
bility to dynamically re-plan the RAN by flexibly deploying vised, and reinforcement learning, ML is being considered
UAVs to different locations. Compared to HAPs, UAVs are as a promising data-driven tool to provide computational
deployed closer to the ground, which not only makes them radio and network intelligence from the physical layer [151]
much cheaper, but also grants them better channel gain with to network management [152]. As a sub-branch of ML,
lower path loss. Especially, upon emergencies (e.g., in disas- deep learning [153] can mimic biological nervous systems
ter reliefs or search & rescue), UAV also provides a low-cost and automatically extract features, extending across all three
solution of temporary wireless service delivery to inaccessi- mentioned learning paradigms. It has a wide variety of
ble areas, such like caves, tunnels, and dilapidated buildings applications to against the big challenges in wireless com-
in earthquakes/fires, which cannot be covered by satellites munications and networking, being applied to form more
and HAPs. adaptive transmission (power, precoder, coding rate, and
Furthermore, empowered by the latest achievements of modulation constellation) in massive MIMO [154], to enable
wireless power transmission, UAVs can also be exploited as more accurate estimation and prediction of fading chan-
mobile and automated wireless chargers [148]. Especially, nels [155], [156], to provide a more efficient RF design
with the simultaneous wireless information and power trans- (pre-distortion for power amplifier compensation, beam-
fer (SWIPT) technologies, the missions of battery charging forming, and crest-factor reduction), to deliver a better
and information transmission can be accomplished in a solution for intelligent network management [58], and to
seamless joint [149]. This significantly raises the feasi- offer more efficient orchestration for mobile edge computing,
bility of massive deployment of battery-life-critical UEs, networking slicing, and virtual resources management [97].
which is essential in some emerging use scenarios such as In addition to deep learning, a few cutting-edge ML
dense WSN. With such rich potentials, UAVs are nowadays techniques represented by federated learning and transfer
widely considered as an essential component of future 6G learning start showing strong potential in wireless commu-
infrastructure. nications. Data-driven methods always have to take into
In addition to its irreplaceable position in the 6G network account the issue of data privacy, which limits the manner
infrastructure, UAV is also expected to contribute to the of processing collected data. In some scenarios, distributing
prosperity of new use cases and emerging applications in data is strictly prohibited and only local processing on the

358 VOLUME 2, 2021

device where the data was collected is allowed. Federated blockchain is unchangeable once it is recorded since each
learning is a method achieving the fulfillment of this require- block is linked with other blocks via the hash value. The
ment by processing the raw data locally and distributing the possibility of breaking the whole chain and modifying the
processed data in a masked form. The mask is designed content of all blocks is very limited. Decentralization: it
such that each of the individual data processing expose no applies the consensus mechanisms to manage and maintain
information, whereas their cooperation allows for meaning- the distributed ledge without the need of a centralized entity
ful parameter adjustments towards a universal model. While or third party. The blocks are replicated and shared over
federated learning gives a method of training ML models an entire blockchain network, thereby avoiding the risk of
from a large number of data sources without ever exposing single point of failure, enhancing data persistency and secu-
sensitive data, it creates only one shared model for uni- rity, and providing flexibility. Transparency: all blockchain
versal applicability. When individual adjustments of models participants have equal right and can access all transac-
are required for their deployment to be successful, transfer tion information of blockchain. Security and privacy: the
learning can be used as a tool enabling these adjustments adoption of asymmetric cryptography, the inherit feature of
and doing so in a manner requiring a much lower amount of data immutability, consensus mechanism, and anonymous
data. By reusing the major part of pre-trained models in a addressing ensure the security, trustworthiness, and privacy
different environment and only adjusting some of the param- of the blockchain. Despite these promising merits, scalability
eters, transfer learning is able to provide quick adaptations is a key barrier when the blockchain technology is widely
using only a low amount of local data. applied from the perspectives of throughput, storage and
In addition to using AI to assist with the operation of the networking. Enabling technologies [159] related to the num-
networks (i.e., AI for Networking), it is also important to ber of transactions in each block, block interval time, data
use the ubiquitous computing, connectivity, storage resources transmission, and data storage to realize scalable blockchain
to provide mobile AI services to end users in an AI-as-a- systems are recently studied.
Service paradigm (i.e., Networking for AI) [14]. Principally, Recently, the potential of blockchain in 5G and beyond
this provides deep edge resources to enable AI-based com- systems has been initially investigated in the literature [160].
putation for new-style terminals such as robots, smart cars, It is applied to enhance the technologies such as edge
drones, and VR glasses, which demand large amount of computing, NFV, network slicing, and device-to-device com-
computing resources but limited by embedded computing munications, to implement important services, e.g., the
components and power supply. Such AI tasks mainly means sharing of spectrum and radio resource, data storage and
the traditional computation-intensive AI tasks, e.g., com- sharing, network virtualization, security and privacy, in the
puter vision, SLAM, speech and facial recognition, natural use case domains of smart city, smart transportation, smart
language processing, and motion control. grid, smart healthcare, and UAV. On the other hand, the
deployment of 5G networks can boost the application of
2) BLOCKCHAIN blockchain systems. The ubiquitous connectivity, comput-
With the great success of a kind of cryptocurrency known as ing, and storage resources provided by mobile networks can
Bitcoin, the blockchain technology has received enormous be employed to provide local computing power for mobile
attention in both industry and academia [157]. A blockchain blockchain systems [161] so as to support solving PoW
is essentially a distributed public ledger spreading across all puzzles, hashing, encryption, and consensus execution. It
participants deployed typically in a peer-to-peer network. A is envisioned that the blockchain will be converged into the
chain of blocks originates from the first block called the gen- upcoming 6G system for more flexible, secure, and efficient
esis block. A new block is appended to the chain via a hash information infrastructure.
value that is generated according to the information of its
parent block. Each block typically consists of two parts: the 3) DIGITAL TWIN
block header and transaction data. In particular, the header Digital twin is an emerging technology and one significant
mainly contains the following information: block version use case of 6G communication system. It refers to the logical
indicating the validation rule, the hash of its parent block, copy (a.k.a., virtual object or softwarized copy) of a physical
timestamp, the number of transactions, and MerkelRoot that object [162]. The virtual representation shall reflect all the
concatenates the hash values of all the transactions in this dynamics, characteristics, critical components, and impor-
block. A chain continuously grows as blockchain users per- tant properties of an original physical object that operates
form transactions. A miner records and packs a batch of and lives throughout its life cycle [163]. The digital twin
transactions into a block by solving a computationally diffi- is following the life cycle of a physical twin. Therefore,
cult problem called Proof of Work (PoW). The newly mined its monitoring, controlling, maintenance, prediction, and
block is then broadcasted to the whole blockchain network optimization processes are started and ended in parallel with
and all the nodes join the consensus process to validate its its respective physical twin. Each digital twin is linked to its
trustfulness and update the new block into the chain. respective physical twin through a unique key. The unique
The blockchain has the following technological advan- key is used to identify the physical twin and allows to estab-
tages [158]: Immutability: the transaction data in the lish a bijective (one-to-one) relationship between the digital

VOLUME 2, 2021 359

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

twin and its real twin. Recently, digital twin has become demand for the EI in the edge of 6G mobile networks to
the center of attention and has attracted significant attention automate its respective tasks [112].
from the Industry 4.0, research and development commu- One of the major use cases of EI can be the automation
nity, manufacturing, and others due to its importance in the of the management and orchestration tasks of the virtual
improvement of the quality of products, services, processes, resources in NG-RAN architecture. In this use case, the EI
devices, etc. in a specific context using AI techniques. is extended to the NG-RAN in order to automate all tasks
There is a large number of industries such as manu- related to the RAN network slice subnet management func-
facturing and aviation, which have been developing and tion (NSSMF) and network function management functions
commercializing the digital twin in order to optimize their (NFMFs) in order to lower the management and orchestra-
processes since last years [164]. In addition, the digital twin tion complexity. To that aim, the ETSI has launched the
is also in its initial stage in healthcare and medicine fields ENI ISG to investigate and provides recommendations to
and is expected to be fully commercialized with the devel- operators [169]. In each of the use cases, including the afore-
opment and deployment of 6G communication networks in mentioned, the EI is constituted of a set of connected devices
the near future [165]. Despite, it is also being extensively which are used to collect, normalize, process, and analyze
applied in the IoT and Industry 4.0 domains. In both sce- the data. Subsequently, the processed data is sent back to
narios, the AI techniques are used to collect, analyze, and the assisted systems in the form of recommendations and/or
test (in different conditions) the data from a physical object orders to be executed in order to automate the target tasks
to build its softwarized copy [166]. The more information or functionalities [169].
about the physical object is provided to the AI analyzer, the The applications of EI in the context of vertical indus-
more accurate and better the performance and prediction of tries have also attracted the attention of both academia and
the virtual object will be during its life cycle [167]. industrial organizations. For example, the EI plays signif-
Towards full utilization of digital twin in 6G communica- icant role in addressing mission-critical applications and
tion system and specifically in the context of Industry 4.0 and massive and critical mMTC types of services in centralized,
IoT, there are still a number of very critical research prob- semi-centralized, and localized resource allocation scenar-
lems, which are needed to be addressed from both academia ios. Despite, the EI also provides energy-efficient solution
and industry. They are, including, but not limited to: dynam- which are expected to reduce the energy consumption of
ically scaling up the platform of digital twin to millions and the communication networks. Therefore, it is considered a
billions of IoT devices, the deployment of the zero-touch and novel opportunity for both operators and vertical industries
self-management approaches to the devices and processes, in order to digitize its application and businesses – using
lack of existing models and methodologies in the area of personal computing, fog computing, urban computing, and
AI to virtualize the physical object, security and privacy, other mechanisms – in the 6G mobile networks.
and many others that are thoroughly addressed in [162], Despite the aforementioned key advantages, there are still
[163], and [167]. These challenges shall be, first and fore- a number of unsolved research challenges in realizing EI
most, explored aiming to develop more suitable digital twin in beyond 5G mobile networks. It is, therefore, crucial to
solutions for a wide range of IoT deployment scenarios and identify and analyze such open research problems and seek
industrial adaptations in the coming decade. for their theoretical and technical solutions. Among others,
the most prominent challenges are: data consistency on every
4) INTELLIGENT EDGE COMPUTING device at edge, data scarcity at edge, bad adaptability of
Edge computing plays a significant role in increasing the statically trained model, and data privacy and security. We
performance of network services, efficient utilization of expect that more research efforts are needed to completely
network resources (both physical and virtual), decreasing realize EI in beyond 5G and 6G networks upon addressing
CAPEX/OPEX of a mobile operator, and lowering network such challenges.
complexity (both in control plane and user plane) [168].
However, the existence of a large number of end users, each 5) COMMUNICATION-COMPUTING-CONTROL
with a diverse set of business and technical requirements, CONVERGENCE
challenges the network operator to think on different alter- Mobile edge networks provide computing and caching capa-
natives in order to address the existing limitations related bilities at the network edge, which makes low-latency, high-
to edge computing using cutting-edge AI tools and modern bandwidth, location-aware pervasive computing services a
ML methods. To that objective, the edge intelligence (EI) is reality. With the proliferation of IoT and Tactile Internet,
introduced aiming to integrate the AI and ML techniques at a huge number of sensing devices and actuators are con-
the edge of the mobile networks in order to bring automation nected to mobile networks. The next-generation system is
and intelligence. The EI is envisioned to be one of the key envisioned to become a huge computer that would converge
enabling technologies for beyond 5G and 6G communication ubiquitous communication, computation, storage, sensing,
networks. Thanks to the increasing number of smart portable and controlling as a whole to provide disruptive applications.
devices, user equipment, Internet of Intelligent Things (IoIT), Due to their superiority in integration and mobility over
and the proliferation of intelligent services; there is a strong wired connections, wireless links are gradually relied by

360 VOLUME 2, 2021

modern and future controlling systems to close the sig- terminate at 5G and the first 6G network is expected to be
nal loop, giving birth to the use scenario of URLLC. The deployed in 2030 or even earlier taking into account great
spirit of system design behind this concept follows the clas- passions of developing 6G from both academia and indus-
sic methodology of control & communication independent try. 6G will accommodate the use cases and applications
design: it starts from designing the controlling component introduced in 5G such as IoT, Industry 4.0, virtual reality,
without concerning the characteristics of communication and automatic driving with better quality of experience in
system, where a set of communication requirements will be a more cost-efficient, energy-efficient, and resource-efficient
generated regarding the expected controlling performance; manner. Meanwhile, it will enable unprecedented use cases
then there follows the designing of wireless system, aim- that cannot be supported by 5G, e.g., holographic-type com-
ing at achieving the target performance proposed by the last munications, pervasive intelligence, and global ubiquitous
stage. The KPI requirements of URLLC, such as 99.999% connectability, as well as other disruptive applications that
reliability and 10 ms, were formulated for a generic con- we are unable to yet imagine. The trend of mobile com-
trolling scenario in this fashion. Nevertheless, recent studies munication services expanded from only human centric to
have revealed a necessity of in-loop co-designing of com- connecting also machines and things, started when MTC
munication and controlling systems tightly coupled to each and IoT were introduced in the age of 5G will continue, and
other. For instance, the close-loop reliability of controlling Internet-of-Everything will be realized when 6G comes. The
system has been proven exponentially decreasing along with 6G system has to meet extremely stringent requirements on
the AoI over the feedback channel [170], [171]. Meanwhile, latency, reliability, mobility, and security, as well as provi-
the AoI over control/feedback channels, as a communica- sioning a substantial boost of coverage, peak data rate, user
tion metric, is convex about the arrival rates of controlling experienced rate, system capacity, and connectivity density,
command and feedback information, which correspond to gaining KPIs generally 10 to 100 times better in comparison
the sampling rate of sensors and decision rate of controller with 5G.
in the control system [172]. The performance of communi- It is envisioned that 6G will take unprecedented transfor-
cation system, is therefore limited by the design of sensing mations that will make it dramatically distinguishing with
and controlling systems. the previous generations. That is:
Similar issues are also to be addressed in cloud comput-
ing. While it occasionally happens that some computing task
occupies the cloud server for long time, blocking all other
pending tasks in the waiting queue and causing a severe • It will be shifted from a radio communication network
congestion, its source may have re-issued the same task based on electronic technologies to a radio-optical
with a more up-to-date status, making the previous task out- system taking advantage of both electronic and pho-
dated and lack of utility. A preemption of the server, i.e., tonic technologies so as to exploit the abundant spectral
terminating the ongoing task in advance to its completion, resources in terahertz and visible light bands, especially
will in this case help reduce the age of task and improve in indoor optical wireless coverage, to meet the inex-
the quality of cloud computing service. Furthermore, it is orable demand on higher system capacity and peak data
also critical for reducing the AoI, to schedule the order of rates.
computing tasks from multiple applications to be offloaded • It will become a connected intelligent platform that
to the cloud [56]. Optimal decisions of such preemption maximizes the synergy between AI and mobile
and scheduling, however, cannot be solely solved by the networks. On the one hand, 6G will facilitate the
computing server, nor by the communication system, but provisioning of over-the-air AI applications, where AI-
only achievable in a joint collaboration among the terminal as-a-Service is offered to end users through pervasive
devices, the network controller, and the cloud computing intelligence. On the other hand, AI-driven air interface,
server. algorithms, protocols, and approaches are applied to
With the dramatically increasing demand for wireless- implement highly-efficient transmission, optimization,
connected industrial automation, and the irreversible trend control, and management of resources and networks.
that not only most mobile applications but also the mobile • With the disruptive advances in large-scale LEO satellite
networking service itself are gradually cloudified, new constellation, the 6G system will go beyond terrestrial
approaches are therefore required in 6G to jointly design networks and provide ubiquitous 3D coverage of the
and optimize this communication-computing-control sym- whole surface of our planet through an integrated space-
biosis [173]. aerial-terrestrial network.
• It will be a smart compute-connect entity by means of
VI. CONCLUSION converging distributed communication, computing, stor-
This article provided a comprehensive survey on the drivers, age, and big data resources, integrating sensing, local-
requirements, efforts, and enablers for the next-generation ization, and controlling capabilities, and inter-working
mobile system beyond 5G. It can be concluded that the tra- with emerging paradigms such as AI, blockchain, digital
ditional evolution of a new generation every decade will not twin, quantum computing and communications.

VOLUME 2, 2021 361

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

•
It should be an intelligent, green, sustainable, and [22] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi,
secure system to fully support the informationized and “Toward 6G networks: Use cases and technologies,” IEEE Commun.
Mag., vol. 58, no. 3, pp. 55–61, Mar. 2020.
intelligentized society in 2030 and beyond. [23] H. Viswanathan and P. E. Mogensen, “Communications in the 6G
In 1926, engineer and inventor Nikola Tesla stated that era,” IEEE Access, vol. 8, pp. 57063–57074, 2020.
“When wireless is perfectly applied the whole Earth will be [24] S. Zhang, C. Xiang, and S. Xu, “6G: Connecting everything by
1000 times price reduction,” IEEE Open J. Veh. Technol., vol. 1,
converted into a huge brain”. This prophecy will transform pp. 107–115, Mar. 2020.
into a reality when 6G comes. [25] S. Chen, Y.-C. Liang, S. Sun, S. Kang, W. Cheng, and M. Peng,
“Vision, requirements, and technology trend of 6G: How to tackle the
challenges of system coverage, capacity, user data-rate and movement
REFERENCES speed,” IEEE Wireless Commun. Mag., vol. 27, no. 2, pp. 218–228,
[1] J. D. Vriendt, P. Laine, C. Lerouge, and X. Xu, “Mobile network Apr. 2020.
evolution: A revolution on the move,” IEEE Commun. Mag., vol. 40, [26] W. Saad, M. Bennis, and M. Chen, “A vision of 6G wireless systems:
no. 4, pp. 104–111, Apr. 2002. Applications, trends, technologies, and open research problems,”
[2] E. Dahlman, B. Gudmundson, M. Nilsson, and A. Skold, IEEE Netw., vol. 34, no. 3, pp. 134–142, May/Jun. 2020.
“UMTS/IMT-2000 based on wideband CDMA,” IEEE Commun. [27] N. Kato, B. Mao, F. Tang, Y. Kawamoto, and J. Liu, “Ten chal-
Mag., vol. 36, no. 9, pp. 70–80, Sep. 1998. lenges in advancing machine learning technologies toward 6G,” IEEE
[3] D. Astely, E. Dahlman, A. Furuskär, Y. Jading, M. Lindström, Wireless Commun., vol. 27, no. 3, pp. 96–103, Jun. 2020.
and S. Parkvall, “LTE: The evolution of mobile broadband,” IEEE [28] W. Guo, “Explainable artificial intelligence for 6G: Improving trust
Commun. Mag., vol. 47, no. 4, pp. 44–51, Apr. 2009. between human and machine,” IEEE Commun. Mag., vol. 58, no. 6,
[4] J. G. Andrews et al., “What will 5G be?” IEEE J. Sel. Areas pp. 39–45, Jun. 2020.
Commun., vol. 32, no. 6, pp. 1065–1082, Jun. 2014. [29] M. Z. Chowdhury, M. Shahjalal, S. Ahmed, and Y. M. Jang,
[5] F. H. P. Fitzek and P. Seeling, “Why we should not talk about 6G,” “6G wireless communication systems: Applications, requirements,
Mar. 2020. [Online]. Available: arXiv:2003.02079. technologies, challenges, and research directions,” IEEE Open J.
[6] “A blueprint of technology, applications and market drivers towards Commun. Soc., vol. 1, pp. 957–975, 2020.
the year 2030 and beyond,” document ITU-T FG-NET-2030, ITU, [30] F. Tariq, M. R. A. Khandaker, K.-K. Wong, M. A. Imran, M. Bennis,
Geneva, Switzerland, May 2019. and M. Debbah, “A speculative study on 6G,” IEEE Wireless
[7] Shaping Europe’s Digital Future, Eur. Commission, Brussels, Commun. Mag., vol. 27, no. 4, pp. 118–125, Aug. 2020.
Belgium, Feb. 2020. [31] G. Liu et al., “Vision, requirements and network architecture of
[8] “Future technology trends for the evolution of IMT towards 2030 6G mobile network beyond 2030,” China Commun., vol. 17, no. 9,
and beyond,” in Liaison Statement, ITU-R Working Party 5D, ITU, pp. 92–104, Sep. 2020.
Geneva, Switzerland, Oct. 2020.
[32] G. Gui, M. Liu, F. Tang, N. Kato, and F. Adachi, “6G: Opening new
[9] “Key drivers and research challenges for 6G ubiquitous wireless
horizons for integration of comfort, security, and intelligence,” IEEE
intelligence,” 6G Flagship and Univ. Oulu, Oulu, Finland, White
Wireless Commun. Mag., vol. 27, no. 5, pp. 126–132, Oct. 2020.
Paper, Sep. 2019.
[10] K. David and H. Berndt, “6G vision and requirements: Is there any [33] C. Huang et al., “Holographic MIMO surfaces for 6G wireless
need for beyond 5G?” IEEE Veh. Technol. Mag., vol. 13, no. 3, networks: Opportunities, challenges, and trends,” IEEE Wireless
pp. 72–80, Sep. 2018. Commun. Mag., vol. 27, no. 5, pp. 118–125, Oct. 2020.
[11] S. J. Nawaz, S. K. Sharma, S. Wyne, M. N. Patwary, and [34] L. Bariah et al., “A prospective look: Key enabling technologies,
M. Asaduzzaman, “Quantum machine learning for 6G communi- applications and open research topics in 6G networks,” IEEE Access,
cation networks: State-of-the-art and vision for the future,” IEEE vol. 8, pp. 174792–174820, 2020.
Access, vol. 7, pp. 46317–46350, 2019. [35] K. Rikkinen, P. Kyosti, M. E. Leinonen, M. Berg, and A. Parssinen,
[12] T. S. Rappaport et al., “Wireless communications and applications “THz radio communication: Link budget analysis toward 6G,” IEEE
above 100 GHz: Opportunities and challenges for 6G and beyond,” Commun. Mag., vol. 58, no. 11, pp. 22–27, Nov. 2020.
IEEE Access, vol. 7, pp. 78729–78757, 2019. [36] M. Polese, J. M. Jornet, T. Melodia, and M. Zorzi, “Toward end-to-
[13] P. Yang, Y. Xiao, M. Xiao, and S. Li, “6G wireless communica- end, full-stack 6G terahertz networks,” IEEE Commun. Mag., vol. 58,
tions: Vision and potential techniques,” IEEE Netw., vol. 33, no. 4, no. 11, pp. 48–54, Nov. 2020.
pp. 70–75, Jul./Aug. 2019. [37] C.-X. Wang, J. Huang, H. Wang, X. Gao, X. You, and Y. Hao, “6G
[14] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y.-J. A. Zhang, “The wireless channel measurements and models: Trends and challenges,”
roadmap to 6G: AI empowered wireless networks,” IEEE Commun. IEEE Veh. Technol. Mag., vol. 15, no. 4, pp. 22–32, Dec. 2020.
Mag., vol. 57, no. 8, pp. 84–90, Aug. 2019. [38] N. Chi, Y. Zhou, Y. Wei, and F. Hu, “Visible light communication in
[15] B. Zong, C. Fan, X. Wang, X. Duan, B. Wang, and J. Wang, “6G 6G: Advances, challenges, and prospects,” IEEE Veh. Technol. Mag.,
technologies: Key drivers, core requirements, system architectures, vol. 15, no. 4, pp. 93–102, Dec. 2020.
and enabling technologies,” IEEE Veh. Technol. Mag., vol. 14, no. 3, [39] M. Kishk, A. Bader, and M.-S. Alouini, “Aerial base station deploy-
pp. 18–27, Sep. 2019. ment in 6G cellular networks using tethered drones: The mobility
[16] Z. Zhang et al., “6G wireless networks: Vision, requirements, archi- and endurance tradeoff,” IEEE Veh. Technol. Mag., vol. 15, no. 4,
tecture, and key technologies,” IEEE Veh. Technol. Mag., vol. 14, pp. 103–111, Dec. 2020.
no. 3, pp. 28–41, Sep. 2019.
[40] C. Li, W. Guo, S. C. Sun, S. Al-Rubaye, and A. Tsourdos,
[17] E. C. Strinati et al., “6G: The next frontier: From holographic mes-
“Trustworthy deep learning in 6G-enabled mass autonomy: From
saging to artificial intelligence using subterahertz and visible light
concept to quality-of-trust key performance indicators,” IEEE Veh.
communication,” IEEE Veh. Technol. Mag., vol. 14, no. 3, pp. 42–50,
Technol. Mag., vol. 15, no. 4, pp. 112–121, Dec. 2020.
Sep. 2019.
[18] T. Huang, W. Yang, J. Wu, “A survey on green 6G [41] J. Du, C. Jiang, J. Wang, Y. Ren, and M. Debbah, “Machine learning
network: Architecture and technologies,” IEEE Access, vol. 7, for 6G wireless networks: Carrying forward enhanced bandwidth,
pp. 175758–175768, 2019. massive access, and ultrareliable/low-latency service,” IEEE Veh.
[19] W. Jiang and F.-L. Luo, “Computational radio intelligence: One key Technol. Mag., vol. 15, no. 4, pp. 122–134, Dec. 2020.
for 6G wireless,” ZTE Commun., vol. 17, no. 4, pp. 1–3, Dec. 2019. [42] IMT Traffic Estimates for the Years 2020 to 2030, ITU-R Standard
[20] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should M.2370-0, Jul. 2015.
6G be?” Nat. Electron., vol. 3, pp. 20–29, Jan. 2020. [43] “Mobile data traffic outlook,” Ericsson, Kista, Stockholm, Rep.,
[21] F. Tang, Y. Kawamoto, N. Kato, and J. Liu, “Future intelligent and Jun. 2020.
secure vehicular network toward 6G: Machine-learning approaches,” [44] Microsoft HoloLens. Accessed: Jan. 25, 2021. [Online]. Available:
Proc. IEEE, vol. 108, no. 2, pp. 292–307, Feb. 2020. https://www.microsoft.com/en-us/hololens/

362 VOLUME 2, 2021

[45] A. Clemm, M. T. Vega, H. K. Ravuri, T. Wauters, and [71] B. Han, M. A. Habibi, and H. D. Schotten, “Optimal resource
F. D. Turck, “Toward truly immersive holographic-type communica- dedication in grouped random access for massive machine-type com-
tion: Challenges and solutions,” IEEE Commun. Mag., vol. 58, no. 1, munications,” in Proc. IEEE Conf. Stand. Commun. Netw. (CSCN),
pp. 93–99, Jan. 2020. Helsinki, Finland, 2017, pp. 72–77.
[46] “Cloud VR network solution white paper,” HUAWEI, Shenzhen, [72] H. Vettikalladi, W. W. T. Sethi, A. F. B. Abas, W. Ko, M. A. Alkanhal,
China, White Paper, 2018. and M. Himdi, “Sub-THz antenna for high-speed wireless commu-
[47] G. Fettweis et al., “The tactile Internet,” ITU-T, Geneva, Switzerland, nication systems,” Int. J. Antennas Propag., vol. 2019, Mar. 2019,
Technology Watch Rep., Aug. 2014. Art. no. 9573647.
[48] G. P. Fettweis, “The tactile Internet: Applications and challenges,” [73] J. C. Tucek, M. A. Basten, D. A. Gallagher, and K. E. Kreischer,
IEEE Veh. Technol. Mag., vol. 9, no. 1, pp. 64–70, Mar. 2014. “Operation of a compact 1.03 THz power amplifier,” in Proc. IEEE
[49] Boston Dynamics. Accessed: Jan. 25, 2021. [Online]. Available: Int. Vac. Electron. Conf. (IVEC), Monterey, CA, USA, 2016, pp. 1–2.
https://www.bostondynamics.com/atlas [74] S. Koenig et al., “Wireless sub-THz communication system with
[50] Spacex Starlink. Accessed: Jan. 25, 2021. [Online]. Available: high data rate,” Nat. Photon., vol. 7, no. 12, pp. 977–981, 2013.
https://www.starlink.com/ [75] A. Sevincer, A. Bhattarai, M. Bilgi, M. Yuksel, and N. Pala,
[51] Z. Qu, G. Zhang, H. Cao, and J. Xie, “LEO satellite constellation for “LIGHTNETs: Smart LIGHTing and mobile optical wireless
Internet of Things,” IEEE Access, vol. 5, pp. 18391–18401, 2017. NETworks—A survey,” IEEE Commun. Surveys Tuts., vol. 15, no. 4,
pp. 1620–1641, 4th Quart., 2013.
[52] Y. Hu and V. O. K. Li, “Satellite-based Internet: A tutorial,” IEEE
Commun. Mag., vol. 39, no. 3, pp. 154–162, Mar. 2001. [76] A. Al-Kinani, C.-X. Wang, L. Zhou, and W. Zhang, “Optical wireless
communication channel measurements and models,” IEEE Commun.
[53] IMT Vision-Framework and Overall Objectives of the
Surveys Tuts., vol. 20, no. 3, pp. 1939–1962, 3rd Quart., 2018.
Future Development of IMT for 2020 and Beyond, ITU-R
Standard M.2083-0, Sep. 2015. [77] P. H. Pathak, X. Feng, P. Hu, and P. Mohapatra, “Visible light com-
munication, networking, and sensing: A survey, potential and chal-
[54] Minimum Requirements Related to Technical Performance for IMT-
lenges,” IEEE Commun. Surveys Tuts., vol. 17, no. 4, pp. 2047–2077,
2020 Radio Interface(s), ITU-R Standard M.2410-0, Nov. 2017.
4th Quart., 2015.
[55] A. Kosta, Age of Information: A New Concept, Metric, and Tool.
[78] L. Zeng et al., “High data rate multiple input multiple output (MIMO)
Hanover, MA, USA: Now Publ., 2017. optical wireless communications using white LED lighting,” IEEE
[56] X. Song, X. Qin, Y. Tao, B. Liu, and P. Zhang, “Age based task J. Sel. Areas Commun., vol. 27, no. 9, pp. 1654–1662, Dec. 2009.
scheduling and computation offloading in mobile-edge computing [79] A. Nuwanpriya, S.-W. Ho, and C. S. Chen, “Indoor MIMO visible
systems,” in Proc. IEEE Wireless Commun. Netw. Conf. (WCNCW), light communications: Novel angle diversity receivers for mobile
Marrakech, Morocco, 2019, pp. 1–6. users,” IEEE J. Sel. Areas Commun., vol. 33, no. 9, pp. 1780–1792,
[57] H. Tang, J. Wang, Z. Tang, and J. Song, “Scheduling to mini- Sep. 2015.
mize age of synchronization in wireless broadcast networks with [80] N. Huang, X. Wang, and M. Chen, “Transceiver design for MIMO
random updates,” IEEE Trans. Wireless Commun., vol. 19, no. 6, VLC systems with integer-forcing receivers,” IEEE J. Sel. Areas
pp. 4023–4037, Jun. 2020. Commun., vol. 36, no. 1, pp. 66–77, Jan. 2018.
[58] W. Jiang, M. Strufe, and H. D. Schotten, “Experimental results for [81] L. Wu, Z. Zhang, and H. Liu, “Transmit beamforming for MIMO
artificial intelligence-based self-organized 5G networks,” in Proc. optical wireless communication systems,” Wireless Pers. Commun.,
IEEE Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC), vol. 78, no. 1, pp. 615–628, 2014.
Montreal, QC, Canada, Oct. 2017, pp. 1–6.
[82] Y. Zhuang et al., “A survey of positioning systems using visible LED
[59] A. Ghosh, A. Mäder, M. Baker, and D. Chandramouli, “5G evolu- lights,” IEEE Commun. Surveys Tuts., vol. 20, no. 3, pp. 1963–1988,
tion: A view on 5G cellular technology beyond 3GPP release 15,” 3rd Quart., 2018.
IEEE Access, vol. 7, pp. 127639–127651, 2019.
[83] A. Memedi and F. Dressler, “Vehicular visible light communications:
[60] 5G-Complete. Accessed: Jan. 25, 2021. [Online]. Available: A survey,” IEEE Commun. Surveys Tuts., early access, Oct. 27, 2020,
https://5gcomplete.eu/ doi: 10.1109/COMST.2020.3034224.
[61] 5G-Clarity. Accessed: Jan. 25, 2021. [Online]. Available: [84] M. A. Arfaoui et al., “Physical layer security for visible light commu-
https://www.5gclarity.com/ nication systems: A survey,” IEEE Commun. Surveys Tuts., vol. 22,
[62] Netx G Alliance. Accessed: Jan. 25, 2021. [Online]. Available: no. 3, pp. 1887–1908, 3rd Quart., 2020.
https://nextgalliance.org/ [85] H. Elgala, R. Mesleh, and H. Haas, “Indoor optical wireless com-
[63] X. Wang et al., “Millimeter wave communication: A comprehensive munication: Potential and state-of-the-art,” IEEE Commun. Mag.,
survey,” IEEE Commun. Surveys Tuts., vol. 20, no. 3, pp. 1616–1653, vol. 49, no. 9, pp. 56–62, Sep. 2011.
3rd Quart., 2018. [86] J. M. Kahn and J. R. Barry, “Wireless infrared communications,”
[64] L. J. Ippolito, “Radio propagation for space communications Proc. IEEE, vol. 85, no. 2, pp. 265–298, Feb. 1997.
systems,” Proc. IEEE, vol. 69, no. 6, pp. 697–727, Jun. 1981. [87] T. Ohtsuki, “Multiple-subcarrier modulation in optical wireless com-
[65] H. Ajorloo and M. T. Manzuri-Shalmani, “Modeling beacon period munications,” IEEE Commun. Mag., vol. 41, no. 3, pp. 74–79,
length of the UWB and 60-GHz mmwave WPANs based on ECMA- Mar. 2003.
368 and ECMA-387 standards,” IEEE Trans. Mobile Comput., [88] J. C. Juarez, A. Dwivedi, A. R. Hammons, S. D. Jones;
vol. 12, no. 6, pp. 1201–1213, Jun. 2013. V. Weerackody, and R. A. Nichols, “Free-space optical communica-
[66] (2017). IEEE 802.15 WPAN Millimeter Wave Alternative tions for next-generation military networks,” IEEE Commun. Mag.,
PHY Task Group 3c (TG3c). [Online]. Available: vol. 44, no. 11, pp. 46–51, Nov. 2006.
http://www.ieee802.org/15/pub/TG3c.html [89] Z. Xu and B. M. Sadler, “Ultraviolet communications: Potential and
[67] T. Nitsche, C. Cordeiro, A. B. Flores, E. W. Knightly, E. Perahia, and state-of-the-art,” IEEE Commun. Mag., vol. 46, no. 5, pp. 67–73,
J. C. Widmer, “IEEE 802.11ad: Directional 60 GHz communication May 2008.
for multi-gigabit-per-second Wi-Fi [invited paper],” IEEE Commun. [90] M. Marcus, J. Burtle, B. Franca, A. Lahjouji, and N. McNeil,
Mag., vol. 52, no. 12, pp. 132–141, Dec. 2014. “Federal communications commission spectrum policy task force,”
[68] B. Han, L. Wang, and H. D. Schotten, “A 3D human body blockage Rep. Unlicensed Devices and Experimental Licenses Working Group,
model for outdoor millimeter-wave cellular communication,” Phys. FCC, Washington, DC, USA, 2002.
Commun., vol. 25, pp. 502–510, Dec. 2017. [91] S. Haykin, “Cognitive radio: Brain-empowered wireless communi-
[69] A. M. Al-samman, M. H. Azmi, and T. A. Rahman, “A survey cations,” IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 201–220,
of millimeter wave (mm-Wave) communications for 5G: Channel Feb. 2005.
measurement below and above 6 GHz,” in Proc. Int. Conf. Rel. Inf. [92] A. M. Voicu, L. Simić, and M. Petrova, “Inter-technology coexistence
Commun. Technol., 2018, pp. 451–463. in a spectrum commons: A case study of Wi-Fi and LTE in the 5-
[70] Z. Chen et al., “A survey on terahertz communications,” China GHz unlicensed band,” IEEE J. Sel. Areas Commun., vol. 34, no. 11,
Commun., vol. 16, no. 2, pp. 1–35, Feb. 2019. pp. 3062–3077, Nov. 2016.

VOLUME 2, 2021 363

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

[93] I. F. Akyildiz, A. Kak, and S. Nie, “6G and beyond: The [114] O-RAN: Towards an Open and Smart RAN, O-RAN Alliance, Alfter,
future of wireless communications systems,” IEEE Access, vol. 8, Germany, Oct. 2018. Accessed: Sep. 26, 2020.
pp. 133995–134030, 2020. [115] O-RAN Use Cases and Deployment Scenarios: Towards Open and
[94] A. Kliks et al., “Beyond 5G: Big data processing for better spectrum Smart RAN, O-RAN Alliance, Alfter, Germany, Feb. 2020. Accessed:
utilization,” IEEE Veh. Technol. Mag., vol. 15, no. 3, pp. 40–50, Sep. 26, 2020.
Sep. 2020. [116] C. Cheng, R. Lu, A. Petzoldt, and T. Takagi, “Securing the Internet
[95] M. A. Habibi, M. Nasimi, B. Han, and H. D. Schotten, “A of Things in a quantum world,” IEEE Commun. Mag., vol. 55, no. 2,
comprehensive survey of RAN architectures toward 5G mobile pp. 116–120, Feb. 2017.
communication system,” IEEE Access, vol. 7, pp. 70371–70421, [117] T. Monz et al., “Realization of a scalable Shor algorithm,” Science,
2019. vol. 351, no. 6277, pp. 1068–1070, Mar. 2016.
[96] A. Laghrissi and T. Taleb, “A survey on the placement of virtual [118] “Report on post-quantum cryptography,” NIST, Gaithersburg, MD,
resources and virtual network functions,” IEEE Commun. Surveys USA, Rep. IR8105 Draft, Oct. 2016.
Tuts., vol. 21, no. 2, pp. 1409–1434, 2nd Quart., 2019. [119] J.-W. Pan, S. Gasparoni, R. Ursin, G. Weihs, and A. Zeilinger,
[97] W. Jiang, M. Strufe, and H. D. Schotten, “Intelligent network man- “Experimental entanglement purification of arbitrary unknown
agement for 5G systems: The SELFNET approach,” in Proc. IEEE states,” Nature, vol. 423, pp. 417–422, May 2003.
Eur. Conf. Netw. Commun. (EUCNC), Oulu, Finland, Jun. 2017, [120] S.-K. Liao et al., “Long-distance free-space quantum key distribu-
pp. 109–113. tion in daylight towards inter-satellite communication,” Nat. Photon.,
[98] Y. Wang et al., “Network management and orchestration using arti- vol. 11, pp. 509–513, Jul. 2017.
ficial intelligence: Overview of ETSI ENI,” IEEE Commun. Stand. [121] E. Björnson, L. Sanguinetti, H. Wymeersch, J. Hoydis, and
Mag., vol. 2, no. 4, pp. 58–65, Dec. 2018. T. L. Marzetta, “Massive MIMO is a reality—What is next?: Five
[99] A. Ksentini, M. Bagaa, and T. Taleb, “On using SDN in 5G: The promising research directions for antenna arrays,” Digit. Signal
controller placement problem,” in Proc. IEEE Global Commun. Conf. Process., vol. 94, pp. 3–20, Nov. 2019.
(GLOBECOM), Washington, DC, USA, 2016, pp. 1–6. [122] M. A. El Mossallamy, H. Zhang, L. Song, K. G. Seddik, Z. Han,
[100] V.-M. Alevizaki, M. Anastasopoulos, A. Tzanakaki, and and G. Y. Li, “Reconfigurable intelligent surfaces for wireless com-
D. Simeonidou, “Joint fronthaul optimization and SDN con- munications: Principles, challenges, and opportunities,” IEEE Trans.
troller placement in dynamic 5G networks,” in Proc. IFIP Intl. Cogn. Commun. Netw., vol. 6, no. 3, pp. 990–1002, Sep. 2020.
Conf. Optical Network Design and Modeling, Athens, Greece, May [123] M. D. Renzo et al., “Smart radio environments empowered by recon-
2019, pp. 181–192. figurable intelligent surfaces: How it works, state of research, and
[101] S. Khan, A. Gani, A. W. A. Wahab, M. Guizani, and M. K. Khan, the road ahead,” IEEE J. Sel. Areas Commun., vol. 38, no. 11,
“Topology discovery in software defined networks: Threats, taxon- pp. 2450–2525, Nov. 2020.
omy, and state-of-the-art,” IEEE Commun. Surveys Tuts., vol. 19,
[124] “Coordinated multi-point operation for LTE physical layer aspects,”
no. 1, pp. 303–324, 1st Quart., 2017.
3GPP, Sophia Antipolis, France, Rep. 3GPP TR 36.819, 2013.
[102] W. Jiang, M. Strufe, and H. D. Schotten, “Autonomic network man-
[125] Q. C. Li, H. Niu, A. T. Papathanassiou, and G. Wu, “5G network
agement for software-defined and virtualized 5G systems,” in Proc.
capacity: Key elements and technologies,” IEEE Veh. Technol. Mag.,
Eur. Wireless, Dresden, Germany, May 2017, pp. 1–6.
vol. 9, no. 1, pp. 71–78, Mar. 2014.
[103] P. Chemouil et al., “Special issue on artificial intelligence and
machine learning for networking and communications,” IEEE J. Sel. [126] H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, and
Areas Commun., vol. 37, no. 6, pp. 1185–1191, Jun. 2019. T. L. Marzetta, “Cell-free massive MIMO versus small cells,” IEEE
Trans. Wireless Commun., vol. 16, no. 3, pp. 1834–1850, Mar. 2017.
[104] S. Tomovic and I. Radusinovic, “Toward a scalable, robust, and
QoS-aware virtual-link provisioning in SDN-based ISP networks,” [127] E. Björnson and L. Sanguinetti, “Cell-free versus cellular massive
IEEE Trans. Netw. Service Manag., vol. 16, no. 3, pp. 1032–1045, MIMO: What processing is needed for cell-free to win?” in Proc.
Sep. 2019. IEEE 20th Int. Workshop Signal Process. Adv. Wireless Commun.
[105] S. E. Elayoubi, S. B. Jemaa, Z. Altman, and A. Galindo-Serrano, “5G (SPAWC), Cannes, France, 2019, pp. 1–5.
RAN slicing for verticals: Enablers and challenges,” IEEE Commun. [128] S. Bassoy, H. Farooq, M. A. Imran, and A. Imran, “Coordinated
Mag., vol. 57, no. 1, pp. 28–34, Jan. 2019. multi-point clustering schemes: A survey,” IEEE Commun. Surveys
[106] M. A. Habibi, B. Han, and H. D. Schotten, “Network slicing in Tuts., vol. 19, no. 2, pp. 743–764, 2nd Quart., 2017.
5G mobile communication: Architecture, profit modeling, and chal- [129] V. Kotzsch and G. Fettweis, “Interference analysis in time and
lenges,” in Proc. 14th Int. Symp. Wireless Commun. Syst., Bologna, frequency asynchronous network MIMO OFDM systems,” in Proc.
Italy, Sep. 2017. pp. 1–6. IEEE Wireless Commun. Netw. Conf., Sydney, NSW, Australia, 2010,
[107] Technical Specification Group Services and System Aspects; pp. 1–6.
Telecommunication management; Study on Management and [130] W. Jiang and T. Kaiser, “From OFDM to FBMC: Principles
Orchestration of Network Slicing for Next Generation Network and comparisons,” in Signal Processing for 5G: Algorithms and
(Release 15). V15 1.0, Standard 3GPP TS 28.801, Jan. 2018. Implementations, F.-L. Luo and C. Zhang, Eds. Chichester, U.K.:
[108] Network Functions Virtualisation (NFV) Release 3; Evolution and United Kindom: Wiley, 2016, ch. 3.
Ecosystems; Report on Network Slicing Support With ETSI NFV [131] S. M. R. Islam, N. Avazov, O. A. Dobre, and K.-S. Kwak, “Power-
Architecture Framework, Standard ETSI GR NFV-EVE 012 V3.1.1, domain non-orthogonal multiple access (NOMA) in 5G systems:
Dec. 2017. Potentials and challenges,” IEEE Commun. Surveys Tuts., vol. 19,
[109] H. Hirayama, Y. Tsukamoto, S. Nanba, and K. Nishimura, “RAN no. 2, pp. 721–742, 2nd Quart., 2017.
slicing in multi-CU/DU architecture for 5G services,” in Proc. IEEE [132] T. Aulin and R. Espineira, “Trellis coded multiple access (TCMA),”
Veh. Tech. Conf. (VTC), 2019, pp. 1–5. in Proc. IEEE Int. Conf. Commun. (ICC), vol. 2. Vancouver, BC,
[110] Network Functions Virtualisation (NFV); Infrastructure Overview, Canada, 1999, pp. 1177–1181.
Standard ETSI GS NFV-INF 001 V1.1.1, Jan. 2015. [133] L. Ping, L. Liu, K. Wu, and W. K. Leung, “Interleave division
[111] M. A. Habibi, B. Han, F. Za Yousaf, and H. D. Schotten, “How multiple-access,” IEEE Trans. Wireless Commun., vol. 5, no. 4,
should network slice instances be provided to multiple use cases pp. 938–947, Apr. 2006.
of a single vertical industry?” IEEE Commun. Stand. Mag., vol. 4, [134] Z. Yuan, G. Yu, W. Li, Y. Yuan, X. Wang, and J. Xu, “Multi-user
no. 3, pp. 53–61, Sep. 2020. shared access for Internet of Things,” in Proc. IEEE Veh. Tech. Conf.
[112] M. A. Habibi, B. Han, M. Nasimi, N. P. Kuruvatti, A. Fellan, and (VTC), Nanjing, China, 2016, pp. 1–5.
H. D. Schotten, “Towards a fully virtualized, cloudified, and slicing- [135] S. Chen, B. Ren, Q. Gao, S. Kang, S. Sun, and K. Niu, “Pattern
aware RAN for 6G mobile networks”, in 6G Mobile wireless netw., division multiple access—A novel nonorthogonal multiple access for
Y. Wu et al., Eds. London, U.K.: Springer, 2021. Ch. 9. fifth-generation radio networks,” IEEE Trans. Veh. Technol., vol. 66,
[113] M. A. Habibi, B. Han, M. Nasimi, and H. D. Schotten, “The structure no. 4, pp. 3185–3196, Apr. 2017.
of service level agreement of slice-based 5G network,” in Proc. IEEE [136] H. Nikopour and H. Baligh, “Sparse code multiple access,” in Proc.
Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC), Bologna, IEEE Int. Symp. Pers. Indoor Mobile Radio Commun. (PIMRC),
Italy, Sep. 2018. pp. 1–6. London, U.K., 2013, pp. 332–336.

364 VOLUME 2, 2021

[137] J. L. Grubb, “The traveler’s dream come true (satellite personal [160] D. C. Nguyen, P. N. Pathirana, M. Ding, and A. Seneviratne,
communication),” IEEE Commun. Mag., vol. 29, no. 11, pp. 48–51, “Blockchain for 5G and beyond networks: A state of the art survey,”
Nov. 1991. J. Netw. Comput. Appl., vol. 166, pp. 1–38, Sep. 2020.
[138] J. Foust, “SpaceX’s space-Internet woes: Despite technical glitches, [161] Z. Xiong, Y. Zhang, D. Niyato, P. Wang, and Z. Han, “When mobile
the company plans to launch the first of nearly 12,000 satellites in Blockchain meets edge computing,” IEEE Commun. Mag., vol. 56,
2019,” IEEE Spectr., vol. 56, no. 1, pp. 50–51, Jan. 2019. no. 8, pp. 33–39, Aug. 2018.
[139] S. Karapantazis and F.-N. Pavlidou, “The role of high altitude plat- [162] R. Minerva, G. M. Lee, and N. Crespi, “Digital twin in the IoT
forms in beyond 3G networks,” IEEE Commun. Mag., vol. 12, no. 6, context: A survey on technical features, scenarios, and architectural
pp. 33–41, Dec. 2005. models,” Proc. IEEE, vol. 108, no. 10, pp. 1785–1824, Oct. 2020.
[140] W. Jiang, H. Cao, M. Wiemeler, and T. Kaiser, “Achieving high relia- [163] F. Pires, A. Cachada, J. Barbosa, A. P. Moreira, and P. Leitão,
bility in aerial-terrestrial networks: Opportunistic space-time coding,” “Digital twin in Industry 4.0: Technologies, applications and chal-
in Proc. IEEE Eur. Conf. Netw. Commun. (EUCNC), Bologne, Italy, lenges,” in Proc. IEEE Int. Conf. Ind. Informat. (INDIN), vol. 1,
Jun. 2014, pp. 1–5. 2019, pp. 721–726.
[141] A. Mohammed, A. Mehmood, F.-N. Pavlidou, and M. Mohorcic, [164] B. Schleich, N. Anwer, L. Mathieu, sna S. Wartzack, “Shaping the
“The role of high-altitude platforms (HAPs) in the global wireless digital twin for design and production engineering,” CIRP Ann.,
connectivity,” Proc. IEEE, vol. 99, no. 11, pp. 1939–1953, Nov. 2011. vol. 66, no. 1, pp. 141–144, 2017.
[142] W. Jiang, H. Cao, L. Goratti, M. Wiemeler, and T. Kaiser, [165] A. J. Zakrajsek and S. Mal, “The development and use of a digital
“Opportunistic relaying over aerial-to-terrestrial and device-to-device twin model for tire touchdown health monitoring,” in Proc. Amer.
radio channels,” in Proc. IEEE Intl. Conf. Commun. (ICC), Sydney, Inst. Aero. and Astro. (AIAA), Jan. 2017, p. 0863.
NSW, Australia, Jul. 2014, pp. 206–211. [166] S. Luscinski, “Digital twinning for smart industry,” in Proc. 3rd EAI
[143] F. Dovis, R. Fantini, M. Mondin, and P. Savi, “Small-scale fading Int. Conf. Manag. Manuf. Syst., Dec. 2018, pp. 1–9.
for high-altitude platform (HAP) propagation channels,” IEEE J. Sel. [167] B. R. Barricelli, E. Casiraghi, and D. Fogli, “A survey on digital twin:
Areas Commun., vol. 20, no. 3, pp. 641–647, Apr. 2002. Definitions, characteristics, applications, and design implications,”
[144] E. T. Michailidis and A. G. Kanatas, “Three-dimensional HAP- IEEE Access, vol. 7, pp. 167653–167671, 2019.
MIMO channels: Modeling and analysis of space-time correlation,” [168] M. Nasimi, M. A. Habibi, B. Han, and H. D. Schotten, “Edge-assisted
IEEE Trans. Veh. Technol., vol. 59, no. 5, pp. 2232–2242, Jun. 2010. congestion control mechanism for 5G network using software-defined
networking,” in Proc. IEEE Int. Symp. Wireless Commun. Syst.
[145] F. Fidler, M. Knapek, J. Horwath, and W. R. Leeb, “Optical commu-
(ISWCS), 2018, pp. 1–5.
nications for high-altitude platforms,” IEEE J. Sel. Topics Quantum
[169] Experiential Networked Intelligence (ENI); System Architecture,
Electron., vol. 16, no. 5, pp. 1058–1070, Sep./Oct. 2010.
Standard ETSI GS ENI 005 V1.1.1, Sep. 2019.
[146] A. Fotouhi et al., “Survey on UAV cellular communications: Practical
[170] O. Ayan, M. Vilgelm, and W. Kellerer, “Optimal scheduling for
aspects, standardization advancements, regulation, and security chal-
discounted age penalty minimization in multi-loop networked con-
lenges,” IEEE Commun. Surveys Tuts., vol. 21, no. 4, pp. 3417–3442,
trol,” in Proc. IEEE Consum. Commun. Netw. Conf. (CCNC), 2020,
4th Quart., 2019.
pp. 1–7.
[147] H. Wang, H. Zhao, W. Wu, J. Xiong, D. Ma, and J. Wei, “Deployment [171] B. Han, S. Yuan, Z. Jiang, Y. Zhu, and H. D. Schotten, “Robustness
algorithms of flying base stations: 5G and beyond with UAVs,” IEEE analysis of networked control systems with aging status,” in Proc.
Internet Things J., vol. 6, no. 6, pp. 10009–10027, Dec. 2019. IEEE Conf. Comput. Commun. Workshops, 2020, pp. 1360–1361.
[148] C. Su, F. Ye, L.-C. Wang, L. Wang, Y. Tian, and Z. Han, [172] Z. Jiang, S. Fu, S. Zhou, Z. Niu, S. Zhang, and S. Xu, “AI-
“UAV-assisted wireless charging for energy-constrained IoT devices assisted low information latency wireless networking,” IEEE Wireless
using dynamic matching,” IEEE Internet Things J., vol. 7, no. 6, Commun. Mag., vol. 27, no. 1, pp. 108–115, Feb. 2020.
pp. 4789–4800, Jun. 2020. [173] G. Zhao, M. A. Imran, Z. Pang, Z. Chen, and L. Li, “Toward real-
[149] W. Feng et al., “UAV-enabled SWIPT in IoT networks for emergency time control in future wireless networks: Communication-control
communications,” IEEE Trans. Wireless Commun., vol. 27, no. 5, co-design,” IEEE Commun. Mag., vol. 57, no. 2, pp. 138–144,
pp. 140–147, Oct. 2020. Feb. 2019.
[150] W. Jiang and H. D. Schotten, “A deep learning method to predict
fading channel in multi-antenna systems,” in Proc. IEEE Veh. Tech.
Conf. (VTC), Antwerp, Belgium, May 2020, pp. 1–5.
[151] W. Jiang and H. D. Schotten, “Multi-antenna fading channel WEI JIANG (Senior Member, IEEE) received the Ph.D. degree in computer
prediction empowered by artificial intelligence,” in Proc. IEEE Veh. science from the Beijing University of Posts and Telecommunications in
Technol. Conf. (VTC), Chicago, IL, USA, Aug. 2018, pp. 1–6. 2008. From 2008 to 2012, he was with the 2012 Laboratory, HUAWEI
[152] W. Jiang, M. Strufe, and H. D. Schotten, “A SON decision-making Technologies. From 2012 to 2015, he was with the Institute of Digital
framework for intelligent management in 5G mobile networks,” in Signal Processing, University of Duisburg–Essen, Germany. Since 2015, he
Proc. IEEE Int. Conf. Comput. Commun. (ICCC), Chengdu, China, has been a Senior Researcher with German Research Center for Artificial
Dec. 2017, pp. 1158–1162. Intelligence (DFKI), which is the biggest European AI research institution
[153] W. Jiang and H. D. Schotten, “Deep learning for fading channel and is the birthplace of “Industry 4.0” strategy. He is a Senior Lecturer
prediction,” IEEE Open J. Commun. Soc., vol. 1, pp. 320–332, 2020. with the University of Kaiserslautern, Germany. He has authored three book
[154] H. Huang, Y. Song, J. Yang, G. Gui, and F. Adachi, “Deep-learning- chapters and over 60 conference and journal papers, holds around 30 granted
based millimeter-wave massive MIMO for hybrid precoding,” IEEE patents, and participated in a number of EU and German research projects:
Trans. Veh. Technol., vol. 68, no. 3, pp. 3027–3032, Mar. 2019. ABSOLUTE, 5G COHERENT, 5G SELFNET, AI@EDGE, TACNET4.0, and
[155] W. Jiang and H. D. Schotten, “Neural network-based fading chan- KICK. He was the Guest Editor for the Special Issue on “Computational
nel prediction: A comprehensive overview,” IEEE Access, vol. 7, Radio Intelligence: A Key for 6G Wireless” in ZTE Communications in
pp. 118112–118124, 2019. December 2019. He currently serves as an Associate Editor for IEEE
[156] W. Jiang and H. D. Schotten, “Recurrent neural networks with long ACCESS and is a Moderator for IEEE TechRxiv.
short-term memory for fading channel prediction,” in Proc. IEEE
Veh. Technol. Conf. (VTC), Antwerp, Belgium, May 2020, pp. 1–5.
[157] F. Tschorsch and B. Scheuermann, “Bitcoin and beyond: A technical BIN HAN (Member, IEEE) received the B.E. degree from Shanghai Jiao
survey on decentralized digital currencies,” IEEE Commun. Surveys Tong University, in 2009, the M.Sc. degree from the Technical University
Tuts., vol. 18, no. 3, pp. 2084–2123, 3rd Quart., 2016. of Darmstadt, in 2012, and the Ph.D. degree from the Karlsruhe Institute
[158] H.-N. Dai, Z. Zheng, and Y. Zhang, “Blockchain for Internet of Technology in 2016. Since July 2016, he has been with the University
of Things: A survey,” IEEE Internet Things J., vol. 6, no. 5, of Kaiserslautern as a Senior Lecturer, researching in the broad area of
pp. 8076–8094, Oct. 2019. wireless communication and networking, with recent special focus on
[159] J. Xie, F. R. Yu, T. Huang, R. Xie, J. Liu, and Y. Liu, “A survey on B5G/6G networks, network slicing, finite blocklength information theory,
the scalability of blockchain systems,” IEEE Netw., vol. 33, no. 5, and information freshness. He has authored over 30 conference and journal
pp. 166–173, Sep./Oct. 2019. papers, and participated in multiple EU collaborative research projects.

VOLUME 2, 2021 365

JIANG et al.: ROAD TOWARDS 6G: A COMPREHENSIVE SURVEY

MOHAMMAD ASIF HABIBI received the B.Sc. degree in telecommu- HANS DIETER SCHOTTEN (Member, IEEE) received the Ph.D. degree
nication engineering from Kabul University, Afghanistan, in 2011, and from the RWTH Aachen University of Technology, Germany, in 1997.
the M.Sc. degree in systems engineering and informatics from the Czech From 1999 to 2003, he worked with Ericsson. From 2003 to 2007, he
University of Life Sciences, Czech Republic, in 2016. He is currently pursu- worked with Qualcomm. He became the Manager of a R&D Group, a
ing the Ph.D. degree with Technische Universität Kaiserslautern, Germany, Research Coordinator for Qualcomm Europe, and the Director for Technical
where has been working as a Research Fellow since January 2017. From Standards. In 2007, he accepted the offer to become the Full Professor with
2011 to 2014, he was working as a Radio Access Network Engineer with the University of Kaiserslautern. In 2012, he became a Scientific Director
HUAWEI. His main research interests include network slicing, network of the German Research Center for Artificial Intelligence (DFKI) and the
function virtualization, resource allocation, machine learning, and radio Head of the Department for Intelligent Networks. He served as the Dean of
access network. the Department of Electrical Engineering, University of Kaiserslautern from
2013 until 2017. He has authored more than 200 papers and participated over
40 European and national collaborative research projects. Since 2018, he
has been the Chairman of the German Society for Information Technology
and a Member of the Supervisory Board of the VDE.

366 VOLUME 2, 2021

View publication stats