SCIENCE CHINA

Information Sciences
. REVIEW . May 2022, Vol. 65 151301:1–151301:27
https://doi.org/10.1007/s11432-021-3351-5

Vision, application scenarios, and key technology

trends for 6G mobile communications
Zhiqin WANG1* , Ying DU1 , Kejun WEI1 , Kaifeng HAN1 , Xiaoyan XU1 ,
Guiming WEI1* , Wen TONG2 , Peiying ZHU2 , Jianglei MA2 , Jun WANG3 ,
Guangjian WANG4 , Xueqiang YAN5 , Jiying XIANG6 , He HUANG6 , Ruyue LI6 ,
Xinhui WANG6 , Yingmin WANG7 , Shaohui SUN7 , Shiqiang SUO7 ,
Qiubin GAO7 & Xin SU7
1
China Academy of Information and Communications Technology, Beijing 100191, China;
2
Huawei Technologies Canada Co., Ltd., Ottawa K2K 3J1, Canada;
3
Huawei Technologies Co., Ltd. (Hangzhou), Hangzhou 310028, China;
4
Huawei Technologies Co., Ltd. (Chengdu), Chengdu 611730, China;
5
Huawei Technologies Co., Ltd. (Shanghai), Shanghai 200040, China;
6
ZTE Corporation, Shenzhen 518057, China;
7
Datang Mobile Communications Equipment Co., Ltd., Beijing 100083, China

Received 29 May 2021/Revised 5 August 2021/Accepted 11 October 2021/Published online 23 March 2022

Abstract With the global commercialization of the fifth-generation (5G) network, many countries, includ-
ing China, USA, European countries, Japan, and Korea, have started exploring 6G mobile communication
network, following the tradition of “planning the next while commercializing one generation”. Currently,
studies on 6G networks are at the infancy stage. Research on the vision and requirements for 6G is still
ongoing, and the industry is yet to clarify the key enabling technologies for 6G. However, 6G will certainly
build on the success of 5G. Therefore, developing high-quality 5G networks and seamlessly integrating 5G
with verticals are the priorities before 2030, when 6G is projected to be commercialized. Also, global 5G
standards will keep evolving to better support vertical applications. As a milestone, the Third-Generation
Partnership Project (3GPP) published Release 16 in July 2020, which continuously enhanced the capabilities
of mobile broadband service based on Release 15 and realized the support for low-delay and high-reliability
applications, such as Internet of Vehicles and industrial Internet. Currently, 3GPP is working on Releases
17 and 18, focusing on meeting the demands of medium- and high-data-rate machine communication with
low-cost and high-precision positioning, which will be published in June 2022. Thus, 6G networks will further
expand the application fields and scope of the Internet of Things to accommodate those services and appli-
cations that are beyond the capabilities of 5G networks. Herein, we present our vision, application scenarios,
and key technological trends for 6G networks. Furthermore, we propose several future research opportunities
in 6G networks with regard to industrialization and standardization.
Keywords sixth-generation (6G) mobile communication, 6G vision, application scenarios of 6G, tera-
hertz communication, integrated sensing and communication, integrated intelligence and communication,
ultra-massive MIMO, reconfigurable intelligent surface, co-frequency co-time full-duplex, holographic radio
technology

Citation Wang Z Q, Du Y, Wei K J, et al. Vision, application scenarios, and key technology trends for 6G mobile
communications. Sci China Inf Sci, 2022, 65(5): 151301, https://doi.org/10.1007/s11432-021-3351-5

1 Introduction
In the past decade, we witnessed and experienced the development of the fifth-generation (5G) technolo-
gies, standards, trails, and commercialization efforts. With the commercialization of the 5G network,
both academia and industry have shifted their focus toward the development of 6G technologies. From
2030, society will enter an era of intelligence. Balanced, high-quality social services, scientific, precise
* Corresponding author (email: [email protected], [email protected])

c Science China Press and Springer-Verlag GmbH Germany, part of Springer Nature 2022 info.scichina.com link.springer.com
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:2

social governance, green, energy-saving social development will become the trend of society in the future.
From mobile Internet to the Internet of Everything, and then to the intelligent connection of all things,
6G will realize the transition from serving people, people and things to supporting the efficient connec-
tion of intelligent agents. Through the intelligent interconnection of people, machines, and things and
collaborative integration, it will meet the needs of high-quality economic and social development, serve
smart production and life, and promote the construction of an inclusive and intelligent human society.
Based on breakthroughs in mathematics, physics, materials, biotechnology, and other fields, 6G network
will integrate information technologies, such as advanced computing, big data, artificial intelligence (AI),
and blockchain. This integration will further accelerate the fusion of communications with sensing,
computing, and control, making the 6G network the foundation of our daily lives, industrial production,
and green development. 6G network will also maximize the potential of low-, medium-, and high-band
spectra to achieve a seamless global coverage, meeting the requirements for unlimited secure and reliable
connections among people, machines, and things anytime, anywhere.
6G network will provide fully immersive interaction scenarios and support precise spatial interactions to
meet the requirements for multiple senses, feeling, and mind communications. Communication for sensing
and inclusive intelligence will not only improve communications but also supercharge the digitalization
and intelligence of physical objects, greatly enhancing the quality of information and communication
services. In summary, the 6G network will realize the digitization and intelligence of physical entities in
the real environment and penetrate all aspects of production, life, and social governance.
Research on the 6G network is already ongoing, and both academia and industry have started to define
what 6G networks should be. The IMT-2030 (6G) Promotion Group officially released the white paper
on 6G networks, named “6G Vision and Candidate Technologies” [1], to illustrate the vision, driving
forces, candidate use cases, and technologies of 6G, highlighting several thoughts on 6G development.
The latest book of 6G [2] provides a comprehensive view of the 6G wireless network through a technical
lens, including the vision, use cases, and target key performance indicator (KPI) of 6G, theoretical
foundations for 6G, enabling technologies for 6G air interface design, and new features for 6G network
architecture. Specifically, the book proposes the key driving force of 6G: shifting from connected people
and things to connected intelligence. A comprehensive survey of recent advances and future trends of
6G wireless networks has been provided [3]. It is envisioned that 6G networks will have the following
four new paradigm shifts: developing space-air-ground-sea-integrated networks to provide a complete
global coverage; leveraging all spectra (sub-6 GHz, millimeter wave (mmWave), terahertz (THz), optical
frequency bands, etc.) to deliver a higher data rate and network capacity; machine learning (ML),
big data, and artificial intelligence techniques are explored to support full applications; strong network
security for both physical and network layers would be guaranteed. Authors in [4] illustrated several
potential physical layer technologies for 6G, including four revolutionary technologies (i.e., holographic
radio, THz communication, large intelligent surface, and orbital angular momentum) of the exploratory
nature and three technologies with more maturity (i.e., advanced channel coding/modulation, visible
light communication, and advanced duplex) for possible usage in the near future.
Unlike previous studies, herein, we present the vision, application scenarios, and key technological
trends of 6G from the perspective of the industry. The rest of this article is organized as follows: the
vision and three types of application scenarios of 6G are introduced in Section 2; key enabling technologies
of 6G, including THz communication, integrated sensing and communication, integrated intelligence and
communication, ultra-massive multiple-input multiple-output (MIMO), reconfigurable intelligent surface,
co-frequency co-time full-duplex, holographic radio technology, and ultralow-power Internet of Thing
(IoT) communication, are described in Section 3; the conclusion is presented in Section 4.

2 Vision and application scenarios of 6G

6G would provide terabit-level transmission rates, submillisecond-level latency, centimeter-level position-
ing accuracy, trillion device connectivity capabilities, a thousand times increase in network capacity, and
a further reduction in overall network energy consumption. 6G would also build up a commonly inter-
connected intelligent network, realizing the vision of “intelligent connection of everything, digital twin”
society.
Considering the services of 6G, several new application scenarios would appear with the features of
immersive, intelligent, and ubiquitous. In this section, three application scenarios (immersive, intelligent,
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:3

ubiquitous applications) of 6G, including eight candidate use cases, are introduced, which present higher
performance requirements for 6G and penetrate all aspects of production, life, and governance. For each
application scenario, we envisage the corresponding technical trends to illustrate the motivation and
guiding principle for 6G technology development.

2.1 Immersive application

The immersive application of 6G means that the 6G network will enable deep connections and interactions
for everything, offering the ultimate immersive experience for humans. This includes four typical use
cases: immersive cloud extended reality (XR), holographic communications, sensory interconnection,
and intelligent interaction.

2.1.1 Immersive cloud XR: a broad virtual space

XR is a term encompassing virtual reality (VR), augmented reality (AR), and mixed reality (MR). In
cloud-based XR, contents are stored, and all rendering and computing operations are implemented on
the cloud. This greatly reduces the computing load and energy consumption of XR devices while freeing
users from the constraints of cables. Thus, lightweight XR devices would become the mainstream to
guarantee more immersive and intelligent experiences, facilitating commercialization.
From 2030, immense improvement of networks and XR terminals will drive XR technology to an era of
full immersion. Cloud-based XR will be integrated with networks, cloud computing, big data, and AI to
enable digital transformation in various fields, such as business and trade, industrial production, culture
and entertainment, education and training, and healthcare.
Cloud-based XR will enable users to interact with environments using audio and motion, including
eye, head, and hand movements. This premium experience can be delivered only when ultralow latency,
ultrareliable, and ultrahigh bandwidth are provided in deterministic environments, which demand novel
designs, ranging from physical layers, through link layers to network layers [5].

2.1.2 Holographic communications: extremely immersive experience

With the continuous development of wireless networks, high-resolution rendering, and terminal displays,
holographic technology will enable three-dimensional (3D) dynamic interactions among people, things,
and surrounding environments through natural and lifelike visual restoration, greatly empowering future-
oriented communications.
Holographic communications will be widely used in culture and entertainment, healthcare, education,
and production to bring a fully immersive experience to users without the restrictions of time, space,
and the boundary between the real and virtual worlds. However, this will not be possible if future
communication networks cannot meet the requirements for real-time 3D display and fast transfer of
holographic images. To send 1920 × 1080 × 50 3D target data with a 24-bit RGB and refresh rates of
60 frames per second (FPS), networks must support a peak throughput of approximately 149.3 Gbps or
an average throughput of 1.5 Gbps with a compression ratio of 100:1. Since immersive multidimensional
interactions will involve thousands of concurrent data streams, the networks must be able to provide
Tbps-level perceived throughput. For holographic targeted therapy and remote microsurgical operations,
loss of information results in a retransmission, which, in turn, fails to meet the requirements of reliability
and latency. This further raises the standards for networks in transmission security and reliability.

2.1.3 Sensory interconnection: fusion of all senses

Sight, hearing, touch, smell, and taste are essential for us to understand our world. From 2030, signal
transmission of not only the hearing and sight but also other main senses, including touch, smell, and
taste, will become a major part of communications and be used in various fields, including healthcare,
education, entertainment, traffic control, production, and social interactions. In the future, people will
be able to feel the warmth of a hug from a family member through their mobile devices when living far
apart. Even in their own homes, users may be able to enjoy experiences, such as beautiful scenery and
even a walk on the sandy beach while feeling the sea breeze in the Maldives.
Fused interaction of all major senses requires coordinated and synchronized transmission of data related
to different senses. To maintain a high-quality experience, millisecond-level latency is required. The
feedback of the touch sense is strongly related to body movement and location, raising the requirements
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:4

for high-precision positioning. The in-step transfer of all-sense information will not be possible if the
maximum throughput of networks is not increased. With more senses interlinked during communication,
robust data security must be assured to protect the privacy and avoid infringement. In addition, each
sense should have a unique digital representation, demanding new joint and independent encoding and
decoding modes to facilitate data transmission.

2.1.4 Intelligent interaction: interactions of feelings and thoughts

6G mobile communications will provide opportunities to achieve breakthroughs in many areas of research,
such as emotional and brain-computer interactions. Intelligent agents will be able to perceive, recognize,
and think, resulting in a complete alternative to traditional intelligent devices. The user-tool relationship
between people and intelligent agents will evolve to human-like interactions with emotions and mutual
understanding. Such intelligent agents will be able to sense the psychological and emotional states of
users through dialogues and facial expressions to help users mitigate health risks. Supporting lossless
transmission of brain information, mind-controllable machines will be available to help the disabled
overcome physiological difficulties in their daily lives and work while quickly accumulating knowledge
and skills.
Considering the technical trends for immersive applications, we conclude the following points.
• 6G will leverage all spectra up to THz or even visible light to deliver ultrahigh data
rates. To support ultimate immersive applications, such as immersive XR and holographic communica-
tion, extremely high data rates (up to tens of Tbps) are required, hence the need for additional spectrum
resources. Therefore, in addition to sub-6 GHz and mmWave, new spectra of up to THz and optical
bands would be fully explored and used to enable remarkable data-hungry and delay-sensitive services
and applications that require high-resolution sensing. To this end, the THz communication techniques,
illustrated in Subsection 3.1, will be developed in 6G.
• 6G will further upgrade the massive MIMO technique for higher network throughput
and link reliability. To satisfy the ever-increasing demand for network throughput, user density, link
reliability, and massive MIMO techniques with even higher orders (ultra-massive MIMO) are investigated
and developed along with the evolution of wireless communication. We envision that the success of ultra-
massive MIMO in 6G would hinge on not only the enhancement of existing 5G methods but also new
solutions, including new antenna architectures, materials, and radio frequency (RF) techniques [2], which
are illustrated in Subsection 3.4.
• 6G will explore multiple physical dimensions to revolutionize data transmission, achiev-
ing higher spectrum efficiency and capacity. Since the spectrum resources, particularly the sub-
6 GHz bands, are becoming scarce, it is necessary to fully explore multiple physical dimensions, especially
the spatial and spectral dimensions, to support higher spectrum efficiency and deliver ultrahigh data rate
and capacity for better user experience. To this end, novel techniques, including reconfigurable intelligent
surfaces (RIS), full-duplex design, and holographic radio, have been developed for potential use in 6G. Us-
ing programmable meta-materials, RIS can dynamically control electromagnetic waves (adjust their am-
plitudes, phases, etc.) through digital encoding to form controllable signals, enhancing network coverage
and improving the cell-edge transmission rate, which is illustrated in Subsection 3.5. The co-frequency co-
time full-duplex technique has been proposed to support concurrent transmission and reception in a single
time and frequency resource to deliver ultrahigh spectral efficiency and throughput and low latency, which
are discussed in Subsection 3.6. The holographic radio technique employs RF spatial spectrum holog-
raphy and holographic spatial wave field synthesis technologies to achieve ultrahigh-resolution (UHR)
spatial multiplexing with ultrahigh capacity and spectrum efficiency, enabling immersive human-centric
communication, such as XR, holographic display, and multisensory communication (Subsection 3.7).

2.2 Intelligent application

As society moves toward “intelligent of everything”, the future 6G is the key for proliferating AI and
delivering several use cases with high intelligence to humans and everything, such as the proliferation of
intelligence, communication for sensing, and digital twins.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:5

2.2.1 Proliferation of intelligence: ubiquitous smart core

By 2030, there will be more smart terminals, including smart personal and household devices, sensors in
cities, unmanned vehicles, and smart robots. Unlike current smart phones, these new terminals will not
only transmit data at high speeds but also work with and learn from various smart devices. The number
of devices connected through the 6G network would reach trillions in the future. Through continuous
learning, communication, cooperation, and competition, these devices can efficiently simulate and predict
physical world scenarios and provide optimal decisions.
Applying AI is essential to explore and continuously learn from big data with enhanced computing
power. In the era of 6G, network self-learning, -operation, and -maintenance would be developed based
on AI and ML. By then, networks will be robust enough to adapt to various real-time changes. Through
self-learning and collaboration between devices, 6G networks will empower society to achieve ubiquitous
learning and continuous updates. AI services and applications will be brought to end-users, making
real-time and reliable AI intelligence accessible to every individual, home, and industry. Through this,
real inclusive intelligence can be realized.

2.2.2 Communication for sensing: extending the functions of converged communications

6G networks will utilize communication signals to sense, detect, locate, identify, and image targets. This
will help wireless communication systems obtain information about the environment to further improve
resource allocation and user experience. Leveraging higher-band spectra, such as mmWave or THz, will
facilitate the acquisition of environmental information, which will, in turn, enhance the performance of
wireless systems. It will also facilitate the digitalization of physical entities in the environment and pave
the way for more applications.
With real-time wireless sensing, advanced signal processing algorithms, and the exploration of edge
computing and AI technologies, 6G will help sense environmental information and reconstruct the target
environment with ultra high definition (UHD) RF images and enable centimeter-level positioning. This
significantly accelerates the realization of virtual and smart cities. Sensor networks based on wireless
signals can supplement or even replace laser radars and cameras, which are easily affected by light,
to improve sensing resolution and detection precision in all weather conditions. With these networks,
surrounding objects, such as pedestrians, bicycles, and baby strollers, can be categorized through sensing.
In addition, to implement applications, such as collaboration between robots, contactless motion controls,
and action recognition, a millimeter-level resolution will be required to provide high-precision real-time
sensing services to users. Besides, higher frequency bands, such as THz, environmental pollution source
detection, and air composition monitoring, such as PM2.5 analysis, can also be implemented with sense.

2.2.3 Digital twins: digital mirror of the physical world

With advances in sensing, communication, and AI technologies, objects or processes are digitally repli-
cated. Interactions, including people-to-people, people-to-thing, and thing-to-thing, will be intelligently
mapped into the digital world. Leveraging data mining and advanced algorithm models, the digital world
can utilize abundant historical and real-time data to simulate, verify, predict, and control physical objects
or processes, delivering the optimal solution for issues in the physical world.
The 6G era will usher in a world of digital twins. In healthcare, medical systems can use digital twin
information to diagnose and provide optimal treatment. In the industrial field, product design can be
digitally optimized to reduce costs and improve efficiency. In agriculture, the production process can
be simulated and deduced to predict adverse factors and improve the production and use of land. In
network O&M, networks can quickly adapt to complex and dynamic environments through physical and
digital interactions, cognitive intelligence, and automatic O&M, resulting in autonomy throughout the
O&M lifecycle, from planning through construction, monitoring, optimization, to self-healing.
To realize use cases in intelligent application scenarios, the following technique trends are considered.
• 6G will support native AI and ML to intelligently connect intelligent things and devices.
One of the key features and objectives of 6G is to enable ubiquitous intelligence by fully integrating both
AI and ML into communication systems natively. This demands a new design principle in 6G to combine
wireless and AI functionalities at the beginning instead of designing the wireless system first and applying
AI next. The technique named integrating intelligence and communications would create opportunities
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:6

for post-Shannon communication theory breakthrough and invocations in learning theories, as illustrated
in Subsection 3.3.
• 6G will seamlessly integrate sensing capability, enabling the fusion of digital and phys-
ical worlds. We envision that diversified new killer applications and services, such as autonomous
driving, intelligent industrial, and intelligent healthcare, will be achieved and supported by 6G networks.
These trends inspire novel technical design principles that seamlessly integrate two originally decoupled
functionalities, i.e., wireless communication and sensing, into one system symbiotically, which is dubbed
integrated sensing and communications (ISAC), to endow the 6G wireless network with the ability to
“see” and “talk” to the physical world simultaneously [6]. Specifically, ISAC can utilize electromag-
netic wave signals to wirelessly detect, localize, track, and image objects, recognize different activities
or statuses, or even reconstruct the environments, and the sensing results can be used to enhance the
performance of radio access and resource management in wireless communications. The ISAC technique
is discussed in Subsection 3.2.

2.3 Ubiquitous application

6G network will be able to provide broadband access anytime and anywhere, particularly for remote
areas, airplanes, unmanned aerial vehicles (UAVs), vehicles, and ships, called ubiquitous applications
with globally seamless coverage capability.
Currently, more than 3 billion people around the world do not have basic Internet access, most of whom
live in rural and remote areas. Due to the high cost of constructing terrestrial communication networks, it
is difficult for telephone communication companies to afford them. Moreover, terrestrial networks cannot
provide high-speed communication required in uninhabited or oceanic areas for Antarctic expeditions and
ocean freighters. In addition, there is an increase in the demand for connections for aerial devices, such as
UAVs and airplanes. As services converge and deployment scenarios increase, terrestrial cellular networks
would be integrated with nonterrestrial networks, including high-, medium-, and low-orbit satellites, high-
altitude platforms, and UAVs, to build a 3D integrated network with global coverage, providing users
with ubiquitous broadband mobile communication services.
Global seamless coverage can enable broadband access anytime and anywhere for everything, which
implies providing ubiquitous connections for remote areas, airplanes, UAVs, vehicles, and ships. It
can also enable wide-area IoT access in areas not covered by terrestrial networks, ensuring emergency
communications, crop monitoring, endangered animal monitoring in uninhabited areas, and the collection
of information about marine buoys and ocean containers. In addition, centimeter-level positioning would
enable high-precision navigation and precise agriculture. With high-precision surface imaging of the
Earth, services, such as emergency rescue and traffic dispatching, can be implemented.
Consider the technical trend for enabling ubiquitous applications. 6G will support seamless mobile
services everywhere, realizing ubiquitous connectivity and global coverage for all users. To achieve this
goal, 6G networks will integrate the terrestrial network, satellites at various latitudes (high-, medium-,
and low-orbit), and aircraft operating within different airspace to form a new mobile information network
that can provide mobile services anytime and anywhere. Specifically, terrestrial networks can implement
standard coverage for urban hotspots and most rural scenarios, whereas space- and air-based networks
can achieve on-demand coverage in remote areas, at sea, and in the air, resulting in flexible networking,
high resilience, and ultra-reliability [1, 3].

3 Potential key technologies of 6G

To achieve the above-mentioned vision of 6G, traditional air-interface technologies will be further devel-
oped, and new technological breakthroughs must be achieved. Based on the current research progress
and the requirements of the proposed application scenarios of 6G, key enabling technologies include
THz communication, integrated sensing and communication, integrated intelligence and communication,
ultra-massive MIMO, reconfigurable intelligent surface, co-frequency co-time full-duplex, holographic ra-
dio, and ultralow-power IoT communication. These technologies will work together to further improve
the technical abilities of future 6G networks and provide users with even more diversified applications
and better service experiences.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:7

Electronics Photonics

Wavelength 10 cm 1 cm 1 mm 100 µm 10 µm

Traditional communication Advanced communication bands

bands (complementary for various applications)
THz
Sub-6 GHz
mmWave
Light
(infrared/visible
Cellular bands /X-ray)

GHz

Figure 1 (Color online) THz frequency bands.

3.1 Terahertz communications

With the wide applications of big data analytics and ever-increasing popularization of immersive expe-
rience in the era of intelligent Internet of Everything, the expected data rate of 6G may reach the Tbps
level, which is much higher than that of existing 5G systems. To achieve this goal, higher-frequency
bands will be explored to obtain more spectrum resources.
THz waves are electromagnetic waves in the frequency range of 100 GHz to 10 THz (see Figure 1),
which have the characteristics of microwave and light waves. Leveraging rich frequency resources, THz
communication can support ultrahigh communication rates and is considered a useful supplement to
existing air-interface transmission solutions. THz waves would be mainly employed in scenarios with
ultrahigh rate transmission requirements, such as holographic communication, short- and medium-range
wireless access, and data backhaul/fronthaul, as well as new communication scenarios, such as micro-
size communication. In addition, with the help of ultralarge bandwidths and the ability to penetrate
nonmetallic or nonpolarized materials with extremely low losses, THz communication can also enable
high-accuracy positioning and high-resolution sensing and imaging.
To fully utilize the benefits of THz communications for 6G, further research is required in the following
directions.

3.1.1 Ultra-high-speed baseband signal processing

THz communication systems based on the 6G network should process Tbps-level transmission rates in
real-time, and existing high-speed signal processing capabilities and (analog-to-digital/digital-to-analog)
AD/DA sampling capabilities will become important technical challenges. A potential solution is to
utilize low-quantization precision signal processing technologies, including joint optimization design of
bit quantization and signal algorithm, joint adaptive quantization threshold single-bit demodulation
optimization, and low-complexity-hardware-integrated circuit designs based on probability calculations.
By reducing the number of quantization bits per unit symbol and maintaining system performance within
an acceptable range, a tradeoff between performance and complexity can be achieved. Furthermore, new
signal processing system architectures, air-interface architectures, and compensation algorithms can be
explored for low-performance components to tackle the challenges in THz communication.

3.1.2 Ultra-large-scale array antenna

Since THz wavelengths are very short, ultra-massive MIMO structures can be integrated into a smaller
size, thereby realizing high gain of THz beams and flexibility of spatial signal preprocessing. However,
current high-gain THz antennas mainly use reflector antennas. Developing miniaturized and arrayed
ultra-massive THz MIMO is a key research issue to be addressed in the future 6G wireless network.
For example, considering the miniaturization of array-associated circuits, it is challenging to arrange the
array elements in a small package due to the small wavelength. As the physical space margin reserved
for associated circuits is extremely limited, arranging the associated circuits of devices, such as power
amplifiers and phase shifters, to be within several millimeters is unrealistic. A solution is a subarray
design based on analog-digital hybrids, which transform associated circuits based on array elements into
physically larger subarray-associated circuits. Another solution is the use of new antennas, such as liquid-
crystal or reconfigurable intelligence surface antennas, instead of phased array antennas, to address the
issues of miniaturization and power loss of associated circuits.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:8

cm-wave
300 3000

2500 Rough surface

Distance (m)

200 2000
mmWave
1500

100 1000
Rough surface
50 500
10 THz Wave
0.1 1 2 3
Frequency (THz)
Rough surface

Figure 2 (Color online) Frequency and distance selective fad- Figure 3 Reflection differences of electromagnetic waves at
ing in THz transmission. different frequencies.

3.1.3 Channel measurement and modeling

Considering air-interface algorithm and system designs, establishing an accurate channel model to char-
acterize the propagation characteristics of THz waves is one of the key tasks to be conducted. Although
THz wireless channel models have been rapidly developed in the past decade and have a certain research
foundation, there is still a lack of a standard channel model for 6G. Therefore, the following research
should be conducted.

• Modeling THz-range propagation losses. Polar molecules in the atmosphere absorb the energy
of THz waves and convert it to molecular kinetic energy, i.e., the molecular absorption effect. Under
different frequencies, the degree of molecular absorption varies, and strong absorption peaks are generated
at certain specific frequencies, thereby dividing the entire continuous THz frequency band into discrete
spectral windows of different sizes (Figure 2). Therefore, THz channels have dual selective fading of
frequency and distance. Thus, building a propagation loss model in the full THz frequency domain is
the prerequisite to evaluating the performance of 6G systems and designing related key technologies.
To facilitate compatibility with existing 5G new radio (NR) evaluation models, path losses should be
considered as a characteristic function of the frequency and distance based on the measured data in
typical frequency bands and scenarios, and characteristic factors are added to improve existing models.
• Modeling and characterizing THz physical propagation. Since the interaction mechanism
between THz wave and propagation environment differs from that of millimeter waves and optical wave-
bands, existing propagation models cannot be applied. For example, surfaces that can be considered
smooth in the low-frequency band appear relatively rough in the THz band (Figure 3), and the diffuse
reflection and scattering effects are further enhanced. Figure 4 shows the test results of the angular
spectrum of the reflected beam after illuminating the wall with directional beams at different operating
frequencies. There are frequency-related boundary conditions to validate the specular reflection. There-
fore, it is necessary to conduct theoretical research on the mechanism of THz propagation and construct
models for characterizing the reflection, scattering, and transmission features of THz waves on different
materials and rough surfaces to better characterize small-scale propagation characteristics. This is vital
for air-interface designs.
• Methodology of THz-channel modeling. Existing channel modeling methods include determin-
istic, stochastic, and hybrid channel modeling. Due to the large path loss in the THz band, its spatial
multi-path propagation distribution model differs from the existing centimeter wave channel model. Tra-
ditional statistical channel modeling methods cannot fully characterize channel propagation features of
the THz band in all scenarios. Therefore, there is a need to explore novel channel modeling methods to
accurately describe the propagation characteristics of THz bands in the space-time domain and wider
architecture models in typical scenarios with reasonable complexity. Considering factors, such as the
universality and complexity of channels, deterministic channel modeling is unsuitable as an independent
modeling method as it requires propagation environment and high complexity information. A more fea-
sible method is deterministic channel modeling for some features based on stochastic channel modeling,
i.e., hybrid channel modeling.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:9

2.7 m

Figure 4 (Color online) Impact of material reflection on directional beams at different frequencies.

40
2020

30 2005

20
THz Gap
10
Pout (dBm)

0 Amplifier Laser

−10

−20 Multiplier

Osillator
−30
10 100 1000 10000 100000
Frequency (GHz)

Figure 5 (Color online) Current technical capabilities of key circuits/devices in the THz system.

3.1.4 Key technologies for THz circuits and devices

THz circuits and devices are key components of THz systems; their performance directly affects the
performance of the systems (Figure 5). THz circuits and devices are key components to realizing a THz
system, and their performance directly affects the overall performance of that system. However, due to
the constraints of materials and techniques, the current working frequency and output power of THz
devices are limited, and it is difficult to meet commercial requirements, such as low power consumption,
high efficiency, and long battery life. Current mature frequency bands for THz are mainly low bands
(100–300 GHz). Thus, subsequent research on THz circuits and device technologies should focus on the
following: higher frequency bands (> 300 GHz), lower loss, and higher efficiency. This includes new types
III-V compound devices, which provide better electron mobility and higher saturation voltage than silicon
devices; system-on-a-chip (SOC), system-in-a-package (SIP), and heterogeneous integration technologies
to reduce the insertion loss caused by the discretization of existing THz systems or circuits and devices;
microsystem integration and packaging technology to reduce the system volume and create conditions for
large-scale integrated production.
In summary, as an important supplement for existing 5G systems, THz communications would enable
various data-hungry and delay-sensitive applications, including holographic communications, short-range
communications (e.g., inter-chip/nano communications), and ultrahigh capacity data backhaul. More-
over, THz-enabled high-accuracy positioning and high-resolution sensing are promising use cases for 6G.
However, several technical breakthroughs should be achieved in THz, including novel transceiver ar-
chitecture designs with high compact and low cost, novel antenna designs, development for advanced
semiconductor materials, baseband signal processing algorithms with low complexity, channel measure-
ment, and modeling. The research roadmap for THz communications should be further clarified step by
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:10

Integrated sensing and communication

Base station n
Environment Base station Reflectio

Inform the
Active routine
sensing

UE
Environment
Environment Multi-site
sensing Base station
Active sensing center Reflection
Environment

Figure 6 (Color online) Enhanced positioning by combing Figure 7 (Color online) High-resolution imaging.
active positioning and passive sensing.

step, and its industrialization should be developed.

3.2 Integrated sensing and communications

In the era of the intelligent Internet of Everything, as digital transformation continues to develop, services,
such as interactive immersive experiences, machine collaboration in unstaffed factories, real-time sensing
smart healthcare, and advanced autonomous driving, are promoted, and they have been applied on a
large scale within this decade. These services require communication networks to support millimeter-
level sensing accuracy.
RF sensing obtains information from the environment through RF signals and enables positioning,
motion detection, imaging, and other functions. RF sensing has long been a separate system developing
in parallel with mobile communication systems, and positioning is the only sensing service that existing
5G systems can provide.
Future communication systems would leverage higher frequencies (tens of GHz or even hundreds of GHz
and THz), larger bandwidth, and larger antenna apertures, making it possible to integrate RF sensing
into communication systems. Therefore, in high-frequency and large-bandwidth application scenarios, we
can use radio-wave transmission, echo, reflection, and scattering analysis to explore the material world.
Wireless sensing abilities are natively integrated into communication systems to provide high-accuracy
positioning, imaging, environment reconstruction, etc. This improves communication performance, ex-
pands the service scenarios, and provides data portals for building a smart digital world. Using AI, the
wireless sensing ability will be further integrated into people’s life.
Integrated sensing and communication are based on the implementation of two independent functions
(RF sensing and wireless communications) in a single system to achieve mutual benefits. On the one
hand, communication systems can be used to complete different types of sensing services, using the same
spectrum or even hardware or signal processing modules. On the other hand, sensing results can be used to
enhance communication access or management and improve service quality and communication efficiency.
This concept has the following prospective application scenarios and possible research directions.
• Enhanced positioning. 6G will further improve positioning accuracy to the centimeter level for
both indoor and outdoor scenarios. Based on the integration of active positioning and passive sensing, ob-
jects (scatterers) in the environment can be used as anchor points to provide more confidence information
for positioning, thereby improving positioning accuracy (Figure 6).
• High-resolution imaging. Sensing services should provide secure, high-precision, and low-power
sensing and imaging capabilities. Based on the principle of integrated sensing and communications,
portable terminals or base stations can sense their surrounding environments, making it easier to collect
and integrate the information needed to maximally translate information about the physical world into
the cyber world (Figure 7).
• Simultaneous localization and mapping (SLAM). In an unknown environment, mobile sensing
devices can be used to recognize surrounding objects (or landmarks), and then a 2D/3D map of the
environment can be reconstructed to further improve positioning accuracy. Due to the complexity of
the actual environment, non-line-of-sight (NLoS) coverage occupies a large part of target service areas.
In this case, reconstructing the virtual environment based on 6G can provide the latest environmental
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:11

Mono-static sensing channel

^
S2
Est H2

Communication channel
P(Sn)
W H2 Est
^
W
environment

?
^
S3
H3 Est

Bi-static sensing channel

Figure 8 (Color online) Environment-associated system of integrated sensing and communication.

information. For example, in NLoS scenarios, 6G-based SLAM can achieve accurate centimeter-level
positioning when the distance between the sensing device and the object is within 10 m.
• Gesture and activity recognition. Sensing services include typical gesture recognition, heartbeat
detection, fall detection, breathing detection, sneeze detection, and intrusion detection. The recognition
of these gestures and activities is associated with the detection of Doppler shift information.
Integrated sensing and communication include multilevel integration from the spectrum, RF hardware,
baseband algorithm, protocol layer, and high-level applications. Future research directions are listed
below.

3.2.1 Fundamental theories

Due to the various purposes of communication and sensing, there are still many technical challenges in
integrated designs of communication and sensing. Achieving an appropriate balance between sensing
resolution and communication ability, i.e., gaining sensing ability without reducing communication effi-
ciency, should be considered. There would be a basic tradeoff because the more pilot resources used, the
better channel estimation performance can be achieved but at the expense of communication spectrum
efficiency.
Therefore, a proper cost function should be designed as a KPI to describe this tradeoff and facilitate
the design of integrated sensing and communications. Assuming the environment function is P (Sn )
(Figure 8), the error function can be used as a cost function for assessing both communication and
sensing. For example, the cost function for communication is given by

lim P (Ŵ 6= W ) = 0. (1)

n→∞

And the cost function for sensing is given by

n
" #
1 X
lim sup E d(Si , Ŝi ) 6 D. (2)
n→∞ N i=1

An alternative is to let the sensed information be the messages for radar transmission, which can be
characterized by the radar transmission rate. Then, the rates of communication and radar transmission
form paired parameters for joint optimization and design.
Resource scheduling (time, frequency, and space) and interference management for integrated sensing
and communication based on these basic theories are important research topics.

3.2.2 Integrated hardware design

Baseband and RF hardware sharing are important research topics for designing integrated sensing and
communication. Hardware integration solutions can reduce the overall power consumption, system size,
and internal exchange delay and enable sensing and communication systems to benefit from each other
in distortion correction and compensation. Since they use different KPIs and algorithms to evaluate
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:12

Distance estimation error

103
FMCW CBLR
FMCW Sim RMSE
102 OFDM CRLB
OFDM Sim RMSE
101
RMSE [Distance] (m)

H2
100
H1
10−1 H3

10−2

10−3

10−4
−30 −25 −20 −15 −10 −5 0
SNR (dB)

Figure 9 (Color online) Distance estimation performance of Figure 10 (Color online) Communication and sensing chan-
two different waveforms. nels in vehicle-to-everything (V2X) cases.

system performance, sensing and communication systems have varying hardware requirements. Given
both the cost and scale of communication and radar systems, hardware designs of integrated sensing
and communication systems are similar to traditional communication architectures. Thus, it is essential
to assess the impact of architecture on sensing performance. For example, full-duplex is the optimal
solution to implement co-site reflection sensing while maintaining the base station transmission rates.
Sensing requires lower signal-to-noise ratios (SNRs) than traditional communication, which simplifies the
implementation of a full-duplex to some extent. However, remote target sensing is still a challenge.

3.2.3 Joint waveform design

Communication networks maximize spectral efficiency, whereas sensing systems focus on improving es-
timation (e.g., positioning and location tracking) resolution and accuracy. This is a major challenge in
joint waveform designs for integrated sensing and communication. Cyclic-prefixed orthogonal frequency
division multiplexing (CP-OFDM) has been adopted as a waveform design scheme for communication.
Previous studies reported similarities between OFDM and continuous waveforms (CW), making them
candidates for integrated sensing and communication. Different applications and scenarios result in dif-
ferent choices. For example, Figure 9 shows the estimation errors of the two waveforms with a center
frequency of 26 GHz and a bandwidth of 400 MHz. When the SNRs are similar, the two waveforms have
close estimation errors. For applications that require a low communication rate but high sensing preci-
sion, CW waveforms may be preferred. In contrast, OFDM waveforms are more suitable for applications
that require a high communication rate but low sensing precision. However, it is still worth investigating
optimal waveform designs to balance the performance of sensing and communication systems.

3.2.4 Channel modeling and analysis

Sensing performance greatly depends on the environment, so its channel model should be based on the
geographical environment (or geometry-based channel model), which is different from the conventional
statistical channel model in wireless communication. This makes it difficult to concurrently evaluate
both communication and sensing performance and simulate their interference and influence in the same
channels. In the case of communication and sensing systems for vehicles (Figure 10), the vehicles perform
both tasks, where channels H1 and H2 are highly correlated in angle information (or spatial domain) but
have a low correlation with channel H3 . H1 refers to the communication channel, H2 is the monostatic
sensing channel, and H3 is the bi-static channel (Figure 8). Therefore, the channel correlation between
sensing and communication systems is a challenge regarding the improvement of mutually beneficial
performance.
In conclusion, the main design principle of integrated sensing and communication is to seamlessly inte-
grate and effectively utilize both functions in the same system (by sharing the same frequency, signaling,
hardware, etc.) in a reciprocal and symbiotic way. This technique would not only alleviate the spectrum
scarcity but further enhance the performance of a single sensing or communication system. Furthermore,
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:13

Transmitter

Information Received
Source Signal Signal

Noise

Figure 11 (Color online) End-to-end intelligent communication system.

the advancement of millimeter-wave, massive MIMO, and advanced radar techniques will also open up
more possibilities for it. However, realizing the full potential of integrated sensing and communication in
practical wireless networks is still challenging, and there are still many important open research problems,
including joint waveform designs, optimal signal processing algorithm designs, network architectures and
transmission protocol design, and security and privacy issues.

3.3 Integrated intelligence and communications

The existing 5G already supports three major application scenarios, including enhanced mobile broadband
(eMBB), ultra-reliable low-latency communication (URLLC), and massive machine-type communications
(mMTC), and 6G will evolve toward higher throughput, lower latency, higher reliability, several connec-
tions, and higher spectrum utilization. Also, AI is driving a new round of technological revolution with the
continuous development of computing power, algorithms, and data sets. Therefore, integrating commu-
nication and AI is a major trend for future wireless network development. AI has shown a great potential
to enable unknown environment modeling, channel prediction, intelligent signal generation and process-
ing, network status tracking and intelligent scheduling, and network optimization and deployment. It is
expected to promote the evolution of future communication paradigms and network architecture changes.
Currently, the integration of communication and AI focuses on air interface technologies and AI datasets
of wireless networks. In this regard, we discuss the preliminary research and ideas around these issues.

3.3.1 Intelligent wireless network air interface

• AI- and ML-based physical layer air-interface techniques. Traditional communication net-
works are designed by first dividing the whole system into several separate functional modules and then
optimizing each module independently. Therefore, to apply AI/ML to wireless communication, the basic
functional modules of the traditional physical layer are replaced with AI/ML methods, such as neural
networks. These methods can improve performance, and they do not change the framework of existing
communication systems and seamlessly integrate intelligence and communication. As shown in Figure 11,
unlike modular optimization based on AI and ML, end-to-end intelligent air-interface technology directly
uses neural networks to process and transmit signals at both the transmitter and the receiver. This
method treats physical layer communications as an end-to-end signal reconstruction problem by jointly
optimizing the design of the transmitter and receiver. For additive white Gaussian noise (AWGN) or
other known fading channels, the entire encoding and decoding neural networks can be trained based on
the stochastic gradient descent algorithm. For common channels, gradient estimation at the transmitter
is necessary to address the unknown channel response.
• AI- and ML-based link-layer air-interface techniques. Existing wireless resource manage-
ment method generally requires explicit mathematical modeling of communication systems and designs
resource management strategies through optimization or heuristic methods. However, this method has
several limitations. First, it is challenging to accurately model wireless networks. The solution based
on approximated models inevitably causes some performance losses. Second, network planning and opti-
mization adapt to dynamic environments, resulting in high system complexity and long communication
delay.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:14

Build the virtual

environment based on
the environment model Virtual
environment

Training

Channel,
Data environment data Environment Digital
collection modeling
twins

Training &
inference

Real system
Predict the environment
information based on the
environment model

Figure 12 Framework of intelligent training and inference for radio resource management.

Several challenges must be well-considered and solved when applying intelligent wireless resource man-
agement methods. First, communication systems must support real-time status collection, action delivery,
benefit calculation, and other key functions. Second, it is important to fully evaluate the influence of
high complexity and active exploration in the training stage on system performance and stability. Third,
the topology and terminals in communication systems vary with time; thus, the systems must be scalable
enough to accommodate changes. Given these challenges, a training and inference framework shown in
Figure 12 is proposed for intelligent radio resource management. In this framework, channel condition
and environmental data are first collected through channel estimation, sensing, and other methods for
environment modeling. Based on this environment model, a virtual environment is built to support
background training. Moreover, to avoid the loss of performance and stability stems from early training
decisions, the policies generated from training are used as the initial inputs.

3.3.2 Wireless AI research dataset

High-quality data are a prerequisite for wireless AI research. However, collective building and sharing of
wireless AI research datasets are still a major challenge. Typical wireless AI research datasets may include
the following five types: the channel dataset revealing the channel states, environment dataset showing
environment information, experience dataset describing network states, decisions, and performance, user
profile dataset (user properties and behaviors), and pretrained model dataset representing trained neural
network models and parameters.
Each dataset is collected in real-time. Through training and inference of pretraining models, the data
can assist various communication tasks to achieve an intelligent network. Figure 13 shows how a wireless
communication system collects and utilizes each dataset. First, a channel model can be trained based
on historical channel data. By learning the changing pattern of the channel state, frequency, and spatial
domains, the model outputs a compressed representation of the channel. This representation data are
then combined with some information, such as environmental data and user profiles, to explicitly describe
the ongoing communication scenario. After that, neural network models of different tasks leverage the
data to perform intelligent decision-making tasks, such as intelligent air-interface configuration, resource
management and scheduling, and intelligent environment perception. The status, decisions, and perfor-
mance indicators of these tasks are collected to form an experience dataset, which is fed back into the
model training process in real-time.
In conclusion, we envision that intelligence (or AI) will be natively integrated into the 6G network at the
beginning of designing the air interface and network architecture. Two designs are involved in integrated
intelligence and communication: the AI for network, which leverages AI or ML as the tools to optimize
wireless networks, and the network for AI that enables the wireless network to support and optimize
AI applications, creating opportunities for post-Shannon communication research and innovations [2].
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:15

Experience dataset

Data cleaning
Data cleaning Environment Experience data
dataset

Environment
profiling Air interface KPI
configuration feedback
Data cleaning Channel
dataset Radio resource
User location scheduling
Input Decision
output
Neural network Environment
Pre-trained Compressed channel sensing
Real-time
channel model representation
channel data Input Inference Downstream tasks
Model pre-training Concatenation
User profile dataset
Data cleaning

Model online training

Figure 13 (Color online) Composition and application of wireless AI research datasets.

However, implementing networks with native AI demands more compact chips with stronger computation
power (such as nanophotonic chips), algorithms more suitable for network collaboration (such as federated
learning), and new interfaces between networks and devices that aid the generation and exchange of
intelligence among all layers, which is a decade-long journey.

3.4 Ultra-massive MIMO

After almost a decade of development, the massive MIMO technology has achieved remarkable progress in
theoretical research and system designs [7], and it has moved toward large-scale commercial use, making
it one of the most important technical techniques to improve system spectrum efficiency. For the future
development of 6G, improving the spectrum efficiency would remain an important performance metric.
Ultra-large MIMO technology will be the further evolution and upgrade of 5G massive MIMO technology.
Higher operating frequency bands also provide convenient conditions for the deployment of ultra-large
MIMO. By deploying ultralarge-scale antenna arrays, applying new materials, and introducing new tools,
higher spectrum efficiency, wider and more flexible network coverage, higher positioning accuracy, and
higher energy efficiency can be achieved.
The power consumption, size, and weight of devices increase with the scale of antenna arrays, posing
great challenges to network operators during deployment and applications. However, with continuous
improvement of antenna and chip integration and the emergence of new materials and processes, the
scale of antenna arrays can be increased under controllable size, weight, and power consumption. In the
6G era, the trends of ultralarge-scale antenna development include the larger arrays, smaller size, higher
integration, and ability to support ultra-high-frequency and multiple frequency bands.

3.4.1 Ultra-massive MIMO baseband processing algorithm

Baseband processing in ultra-massive MIMO involves precoding processing, channel estimation, and re-
source scheduling. The complexity of baseband processing is positively related to the scale of antenna ar-
rays. Theoretically, the maximum-ratio-transmit preprocessing algorithm can achieve good performance
with several antennas [8]. However, the multiple antenna scheme poses serious challenges in baseband
processing. In addition, using a larger channel bandwidth may go beyond the computing capability of
devices. Therefore, leveraging the distributed computing and parallel processing algorithms can be an
effective way to reduce the computational complexity and communication requirements of devices [9].

3.4.2 Channel state information feedback

The accuracy of channel state information highly affects the preprocessing performance of a transmitter
equipped with ultralarge-scale antennas. In general, time division duplex (TDD) systems utilize channel
reciprocity to obtain the channel state, whereas frequency division duplex (FDD) systems depend on the
measurement and feedback of terminals for the same purpose. However, the limited capacity and delayed
transmission in the feedback channel compromise the accuracy of the channel state information obtained
in FDD systems. As the scale of the antenna arrays increases, the form types and deployment modes
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:16

become more diverse. However, obtaining channel state information is increasingly challenging. Recent
research [10] has shown reciprocity between the uplink and the downlink FDD channels, including delay
and angle information. This is a breakthrough in compressing feedback overhead and reducing computing
complexity. However, antenna calibration is the prerequisite to acquire channel state information through
reciprocity. Highly integrated antennas make it difficult to design self-calibration circuits. A feasible
solution is to measure the signals of ultra-massive MIMO arrays through terminals or network nodes and
communicate with a base station to calibrate the antenna.

3.4.3 Intelligent design for ultra-massive MIMO

Enabling large-scale antenna systems needs several decision and optimization procedures, but the tra-
ditional algorithms may not guarantee its efficiency and performance. Intelligent ultra-massive MIMO
can be a feasible solution by transforming modules, such as channel measurement, beam management,
precoding, multi-user MIMO (MU-MIMO) user detection and scheduling, signal processing and channel
state information feedback, and transmission point selection and planning. However, targeted research
is required to determine ML algorithms that best suit wireless communication systems. Problems, such
as dataset collection and generalization capabilities, must also be addressed.

3.4.4 Channel measurement and modeling

Channel modeling is the basis of wireless communication system simulations and performance analyses.
For a 5G system, channel modeling is performed between a terminal and a base station under the
assumption of a far-field. Thus, a received electromagnetic wave can be modeled as a relatively stable
plane wave. In an ultra-massive MIMO system, antennas of a base station may be deployed in a wide
area, thereby forming an ultralarge-scale antenna array. In this case, when the terminal is in the near-
field of the antenna array, the received nonstationary electromagnetic wave cannot approximate a plane
wave. Therefore, far-field modeling is not applicable. Thus, there is a need for novel channel modeling
considering near-field scenarios. Moreover, as new antenna forms emerge, the propagation characteristics
of electromagnetic waves and the impact on channels should also be considered in the modeling.

3.4.5 Distributed ultra-massive MIMO

Compared with centralized antenna systems, distributed ultra-massive MIMO facilitates the construction
of the ultralarge-scale antenna arrays, providing higher spatial resolution and better performance. As its
network architecture is nearly scale-free, its transmission shifts from network-centered to user-centered,
achieving a consistent user experience. In addition, it can shorten the distance between network nodes
and end-users, effectively reducing system power consumption. Practical and cost-efficient deployment
would be a priority to implement distributed ultra-massive MIMO, followed by real-time information
exchange and time-frequency synchronization issues among nodes.
In addition to effectively improving spectral efficiency, ultra-massive MIMO can enhance network cov-
erage performance and positioning accuracy. In terms of coverage, 6G will further expand to mmWave
or even THz bands. Despite larger bandwidth, the sensitivity of transmission and atmospheric loss will
also significantly increase. However, with the ultra-narrow beam of ultralarge-scale antennas, communi-
cation distances and coverage can be effectively increased. Considering positioning accuracy, meter and
even submeter level accuracies have already been achieved in 5G, which would be further improved in
6G. Ultra-massive MIMO features ultrahigh spatial resolution, which is inherently suitable for precise
positioning [11]. In the real environment, there may be a little or weak line-of-sight (LoS) path. Precise
spatial angle information from ultra-massive MIMO and temporal information of traditional positioning
technologies would be combined to achieve higher accuracy in complex environments [12].
In summary, empowered by new materials and advanced techniques (e.g., ultralarge aperture arrays,
AI, and sensing), ultra-massive MIMO techniques would achieve higher spectral efficiency, wider and more
flexible network coverage, higher positioning precision, and higher energy efficiency in a wider frequency
range. However, to actualize this technique in practice, several research challenges, such as complicated
channel measurement and modeling, heavy signal processing load, high reference signal overhead, and
limited front-haul capacity, should be addressed. Antenna arrays and RF chips that feature low power
consumption, low costs, and high integration would be the key to the commercialization of ultra-massive
MIMO [1].
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:17

3.5 Reconfigurable intelligent surface

Traditional network designs and technological evolution focus on better adaptation to the limitations
of wireless transmission channels. For example, beamforming is employed to resist transmission losses
and interference, and modulation and coding schemes and MIMO are utilized to improve transmission
capacity [13, 14]. However, with the continuous expansion of future communication demand and network
complexity, the increase in network construction (including energy overheads [15]) has become key in
future communication [16]. RIS, with a “reconfigurable/programmable wireless transmission channel”
as the core, opens a new direction for future networks owing to its intrinsic characteristics, including
low cost, low energy consumption, high reliability, and large capacity [17]. Authors in [18] provided
an overview of the fundamental physical layer issues of the RIS-assisted wireless systems, including the
reflection principle, channel estimation, and system designs for RIS-aided wireless communication where
RIS acts as a reflector and RIS-based information transmission where RIS functions as a transmitter.
Moreover, several potential applications of RIS in 6G have been proposed.
Generally, RIS is the use of an artificial electromagnetic surface to realize the intervention of electro-
magnetic wave transmissions, such as ideal absorption and specific reflection [19, 20]. Compared with
traditional meta-surfaces with fixed functions, RIS can enable real-time programming of electromagnetic
characteristics [21, 22]. It includes several elements. In actual applications, control signals are applied
according to the characteristics of the elements to realize a dynamic change in the electromagnetic proper-
ties of each element and then control the amplitude, phase, polarization, and frequency of electromagnetic
waves. Current research is in the following directions.

3.5.1 Lumped circuit based RIS

The first one is the diode/varactor diodes. This method controls electromagnetic signals by adjusting the
capacitance or inductance in the equivalent circuit [23–25]. For instance, multi-level phase adjustment
can be achieved by increasing the number of diodes or multi-level bias voltages. The signal amplitude
can be changed by adjusting the load impedance [26]. This kind of RIS can quickly respond to electro-
magnetic signals, and has low reflection loss and power consumption. It has high maturity and low cost
in implementation.
The second one is the micro-electro-mechanical system (MEMS) assisted method. This method uses
MEMS to adjust electromagnetic signals by controlling electromagnetic units, such as signal transmis-
sion delay change [27], electromagnetic units [28] rotation, and piezoelectric effect based capacitance
change [29]. This kind of RIS has good linearity, low power consumption and high monolithic integra-
tion.

3.5.2 New material based RIS

This is achieved by adjusting material parameters. For example, nematic liquid crystals can enable
voltage-controlled birefringence in different bias electric fields and then achieve electromagnetic absorp-
tion [30]. By introducing yttrium iron garnets into the backing material, the magnetic permeability can
be adjusted through the magnetic field. In addition, graphene is also considered a new material for
RIS [31]. Metals can help reduce the surface equivalent impedance, and improve its ability to adjust
signal amplitude, phase and frequency [32, 33].
Based on the above characteristics, such RIS can be widely applied to various wireless transmission
scenarios and frequency bands, including typical cellular scenarios, such as outdoor dense urban areas
(e.g., for blind area coverage and capacity enhancement), suburban areas (e.g., for overcoming two-path
loss), and indoor and outdoor-to-indoor (O2I) coverage.
Currently, research on RIS is still in the initial stage, and the following areas need to be investigated.

3.5.3 RIS modeling and simulation

Due to the diverse implementation and performance of RIS, effective and accurate modeling is needed
to evaluate the performance of actual system design and algorithm. Electromagnetism-based full-wave
simulations, such as the finite-difference time-domain (FDTD), finite-integral time-domain (FITD), finite
element (FE), and method of moments (MoM), are necessary to analyze the performance of RIS elements
and integrated systems. It is also helpful to extract simplified models, which can be applied to system
simulations, and identify and analyze potential factors that affect performance. For instance, MoM
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:18

vh dBi
40 dBi
vh-3D
40
20
100 20
Azimuth (degree)

30
0
20 0
200 −20 10
−20
−40 20
300 −40
0 10
−60 0 −60
10 20 30 40 50 60 70 80 90 −20 −10
Zenith angle (degree)
vv dBi
40 dBi
vv-3D
40
20
100 20
Azimuth (degree)

0 40
0
200 −20 20
−20
−40
300 20 −40
20
−60 0 0
−20 −60
10 20 30 40 50 60 70 80 90 −20
Zenith angle (degree)

Figure 14 (Color online) RIS-based composite channel.

features high accuracy and computing efficiency and can be perfectly combined with classical high-
frequency analysis methods, such as physical optics (PO) and uniform theory of diffraction (UTD).
Therefore, they are suitable for the study of metasurfaces. For lumped-circuit-based RIS products, full-
wave simulations can also be used based on the multiport scattering network model [34, 35]. In addition,
it is of great value and research significance to extract an effective simplified model for the physical and
electromagnetic characteristics of RIS. The model should reflect the RIS error on electromagnetic signals
and the mutual coupling between electromagnetic units.

3.5.4 RIS channel modeling

For wireless communications, accurate channel modeling ensures precise and universal evaluation. In 5G
research, various channel models, including stochastic and hybrid models, have been developed to reflect
new scenario features, such as large-scale antennas, massive MIMO, and large bandwidth.
For the large-scale characteristics of a purely statistical model, a large-scale loss model in the LoS
scenario with RIS has been proposed [36]. For small-scale fading, Rayleigh and Rician distributions are
employed to model channels (such as LoS and NLoS). However, this method cannot accurately reflect
important channel features in the performance evaluation, such as LoS or NLoS statistical features,
after introducing RIS, multi-path delay, polarization, and reciprocity. Numerous measurements and
simulations are needed to optimize the channel modes.
Considering RIS varying the certainty of propagation channels, Ref. [37] followed the scheme of map-
based hybrid channel models and adopted the direct modeling of specific propagation environment and
communication subject and ray tracing technology to analyze the channel with RIS. In this model, based
on the high-frequency electromagnetic scattering theory, a RIS panel is modeled as an antenna wave
source driven by multiple incident waves, and then a channel composed of gNB-UE, gNB-RIS, and RIS-
UE is obtained (Figure 14). Because this model can analyze the characteristics of the main components of
a channel, it can accurately and effectively evaluate the algorithm and system performance. In addition,
it fits the concept of a “reconfigurable/programmable wireless transmission channel” in RIS applications.
It has great research and application prospects.

3.5.5 Algorithms of RIS

The acquisition of channel state information is critical for beamforming, secure transmission, passive in-
formation transmission, and sensing positioning. However, RIS is characterized by several passive ele-
ments with spacing less than a half wavelength, all passive elements, and no complex signal processing
capability, making it difficult to obtain segmented channels. Therefore, for effective channel estimation,
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:19

it is necessary to minimize the complexity of channel estimation to avoid complex signal processing op-
erations. This can be achieved by replacing RIS passive reflection elements with elements that contain
certain sensing and signal processing functions and designing the number and layout pattern of active
elements based on the tradeoff between segmented channel estimation and transmission performance [38].
The position-information-based RIS channel estimation method uses a fixed position of the base station
and RIS, as well as the array characteristics of RIS, to obtain the signal angle of arrival and other
key information. This method can reduce channel estimation complexity. Leveraging the sparsity of
high-frequency channels and the programmable property of RIS, RIS channels can be obtained through
compressed sensing, matrix filling, and other methods [39]. The cooperative solution of channel estima-
tion and beam matching can be used to estimate RIS channels. However, segmented transmission poses
challenges to the conventional codebook solution. There is a need to design a codebook and matching
scheme of base-station-RIS and RIS-UE channels for random access design and beam selection. Due to
limited feedback in downlink channel estimation, the tradeoff between channel estimation accuracy and
nonideal channel feedback should be considered. To solve the problem of channel estimation when the
channel model is unknown, recently developed AI can be applied to channel estimation [40], which can
reduce the demand for channel modeling and has high robustness and anti-noise performance.

3.5.6 Challenges to industrialization

As a promising technology, RIS has attracted research interest from academia and the industry world-
wide. Various verifications and tests have demonstrated its feasibility, which introduces a new paradigm
for mobile communication. Based on the system mentioned in [41], the feasibility and high performance
of single-input single-output (SISO) or MIMO multi-modulation transmission systems have been verified.
NTT DoCoMo has also demonstrated that its prototype materials can be used on bands higher than
mmWave (traditionally, 30–300 GHz). Moreover, EU’s VISORSURF and ARIADNE projects have stud-
ied the software/hardware platforms, protocol stacks, and so on. Massachusetts Institute of Technology
built a RIS test system named RFocus. Tests showed that it can increase the median received signal
strength by 95 times and the median channel capacity by 20 times.
The following challenges should be addressed for the sustainable development and industrialization of
RIS.
• High-precision beam control. Refined high-gain beamforming is crucial for RIS applications in
mmWave or even higher frequency bands. Based on beamforming, high-precision control of signal phases
is required in RIS. However, this single-dimensional method has high requirements in product costs, power
consumption, and complexity. Therefore, joint optimization based on phase and amplitudes [21, 42] are
expected to further reduce the control overhead. Moreover, increasing the number of elements to provide
a higher gain increases the size of the reflecting surface, and reducing the spacing between elements causes
severe mutual coupling.
• RF linearity. When RIS adjusts electromagnetic signals, aperture illumination results in RF
intermodulation distortion (IMD), resulting in adjacent channel interference and harmonic wave. MEMS-
based methods are expected to further improve this phenomenon [43].
In conclusion, wide-area seamless coverage and stable high throughput are the basis for applications
in wireless communication networks. RIS will further facilitate the development of high-quality networks
in the following aspects.
• Coverage enhancement. In cellular networks, coverage holes may occur in outdoor areas due
to obstructions by buildings and plants. For cell edges, limited transmit power and strong interference
reduce user received signal quality and uplink experience. For O2I and indoor deep coverage in complex
environments, it is difficult to ensure the quality of the communications due to penetration loss and other
factors, especially for high-frequency networks. RIS (such as simplified static passive RIS) can improve
transmission quality in these areas (for example, introducing stable reflection or projection paths).
• High capacity. Currently, high-order modulation and multi-stream MIMO technologies are mainly
employed to increase capacity. However, MIMO has high requirements for channel conditions (like stable
multi-path transmission). RIS-based applications (such as dynamic active and controllable RIS) enable
base stations to control the propagation between RIS and UEs to construct stable multi-path conditions
(for example, increasing reflection path in LoS transmission) to support multi-layer MIMO transmis-
sion. With the extensive applications of RIS, the stability of the communication environment would be
effectively improved, and the stable system performance and user experience are guaranteed.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:20

FD@BS FD@relay FD@BS&UE

FD_UE

DL UL
FD_BS BS
FD_BS
FD_Relay

HD_UE HD_UE UE
HD_UE HD_UE

Figure 15 (Color online) Typical full-duplex architectures.

RIS will pave the way for more applications, such as positioning and simultaneous wireless information
and power transfer (SWIPT). For positioning, in addition to effectively improving signal quality, RIS can
be flexibly deployed in the service area to aid positioning and improve positioning accuracy. For SWIPT,
RIS effectively transmits the mixed signal of information and energy, improves the service quality of
weak users (such as large-scale IoT users) [44, 45], and enables these users to extend the service cycle
through energy collection. RIS will provide new momentum for future communication networks (5G-
Advanced/6G) by building a controllable wireless environment.
In summary, as a promising paradigm for designing wireless networks, the RIS technique can dynami-
cally tune the signal reflection amplitude or phase via a digitally-controlled reflecting surface to further
enhance wireless communication performance, such as creating effective LoS links, improving channel
rank, and reshaping channel realizations. Moreover, it can improve the radio-sensing and computing per-
formance of mobile networks. However, for practical use, RIS faces some challenges, including physical
modeling and design of surface materials, channel modeling, channel state information collection, beam-
forming design, passive transmission, AI-enabled design, and industrial challenges, which are discussed
in this subsection.

3.6 Co-frequency co-time full-duplex

Improving spectral efficiency and alleviating spectrum scarcity are the objectives of advancing wire-
less communication theories, technologies, systems, and standards. Self-interference cancellation (SIC)
enables simultaneous transmission and reception in the same frequency band, creating new duplex tech-
nologies. For example, the uplink and downlink data transmission can be performed by different carriers
in the same frequency band and by the same carriers simultaneously, thus called full-duplex. Theoret-
ically, full-duplex technology doubles spectral efficiency for both FDD and TDD modes. For FDD, the
unified duplex mode provides unified spectrum resource usage and management and implements a virtual
spectrum without separate planning for the uplink and downlink. For TDD, full-duplex prevents time
division that reduces delay and overhead, which is more favorable for low-delay services. It can also allo-
cate more spectrum resources to uplink services for better uplink coverage. This technology has a great
potential to improve network capabilities and satisfy the requirements of low latency, wide coverage, and
high throughput.
Typical full-duplex architectures are shown in Figure 15, which can be deployed depending on the
maturity of SIC technology. In the early stage, SIC is mainly implemented on the base station side,
and full-duplex is employed in cellular hotspots and relay transmission. In cellular hotspot areas, a full-
duplex base station can schedule uplink and downlink UEs simultaneously, improving spectral efficiency
and flexibility. For a relay transmission, received and forwarded signals can be simultaneously transmitted
in the same frequency band, significantly reducing end-to-end delay. As SIC develops, full-duplex will be
applied to dense networking, macrobase stations, and other innovative scenarios, such as device-to-device
(D2D). Moreover, future full-duplex terminals will improve the bidirectional data transmission rate and
reduce feedback delay in UEs.
With over 115 dB self-interference cancellation for 20 MHz bandwidth signals of two transmitters
and two receivers (2T2R) antennas, existing SIC technology can be used in low-power and small-scale
antennas for single-site full-duplex operations, such as Wi-Fi and microcellular isolated sites. With
integrated access and backhaul, full-duplex is implemented in specific scenarios by extending the distance
between TX and RX antennas and fully utilizing the difference between the access and the backhaul
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:21

Transmitted Self-
signal interference Intended
signal

Receive
Transmit
antennas
antennas

Interference suppression in space domain

Interference cancellation in RF domain

d1

d 1 dk
d1
d 1 dk
dk
dk

RF chains RF chains
(PA/DAC) (LNA/ADC)

Interference cancellation in digital domain

Baseband

Figure 16 (Color online) SIC technology.

transmission directions. 6G communication systems will continue to evolve regarding high bandwidth,
large-scale antennas, and dense networking, posing challenges on interference cancellation and networking.

3.6.1 SIC technology

SIC is a key full-duplex technology. Interference cancellation involves multiple domains, including
space/RF/digital domains (Figure 16). Space-domain SIC can be implemented through antenna po-
sition optimization, spatial nulling beamforming, and high isolation of transmitter (TX) and receiver
(RX) antennas. In the RF domain, SIC is implemented by constructing cancellation signals whose phases
are opposite to those of the received self-interfering signals. This effectively cancels the interfering sig-
nals in the RF analog domain. Finally, digital-domain SIC further reconstructs signals and eliminates
residual linear and nonlinear self-interferences. Research is required in the following directions to satisfy
the requirements of high power, large bandwidth, and large-scale antennas.
• Low-complexity algorithm for RF signal reconstruction and interference cancellation.
Constrained by the existing architecture, especially the delay line and interference cancellation matrix in
the RF domain, the complexity of traditional RF interference cancellation links increases with the square
of the number of antennas. Therefore, supporting large-scale antennas (> 32T32R) and large bandwidth
(> 100 MHz) is a key challenge to implementing full-duplex. This can be solved by redesigning the
antenna structure or optimizing the antenna position for high-isolation transceiver antennas. Moreover,
low-complexity RF signal reconstruction and interference cancellation can be implemented through mirror
antennas or heterogeneous multi-port networks.
• Implementation of high-level interference cancellation. The transmission power of macrobase
stations can reach 23–46 dBm, which requires SIC of up to 118–141 dB (taking −95 dBm low noise as an
example). High-level digital cancellation increases the requirements on the linearity of RF components.
First, the linearity of RF components, such as the power amplifier, should be improved. Second, nonlinear
effect modeling based on ML can be achieved in the digital domain for better interference cancellation.
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:22

Self-interference

Cell-1 Cell-2
Inter-cell
interference

BS-1 BS-2

Intra-cell
Inter-cell interference
UE-2 UE-3
UE-1 interference UE-4

Intended DL signal
Intended UL signal

Figure 17 (Color online) Full-duplex networking.

• SIC equipment minimization. Small and lightweight SIC components, low power consumption,
and low costs play an important role in implementing full-duplex in terminals, such as phones and
laptops. In this regard, miniaturized high-isolation transceiver antennas and chip-based RF interference
cancellation modules are essential for full-duplex terminals.

3.6.2 Networking technology

Another key technology of full-duplex is networking. A key challenge in full-duplex networking is the
interference cancellation among base stations and UEs (Figure 17). Complex multi-point to multi-point
interference is formed between base stations, and the high transmit power of the base station reduces the
uplink signal-to-interference-plus-noise ratio (SINR). In addition, when uplink and downlink transmissions
occur simultaneously in one cell, the uplink user transmission interferes with the downlink user reception.
Especially in the case of MU-MIMO, multiple uplink users cause complex interference to downlink users,
reducing the downlink SINR. To address the above-mentioned challenges, powerful network coordination
capabilities and joint uplink and downlink optimization are required to improve the user experience of
the entire network. Therefore, research is needed in the following directions.
• Advanced inter-base station interference cancellation technology. It is necessary to study
the mechanism that enables channel estimation and interference cancellation between base stations, such
as inter-base station channel estimation technology combining centralized radio access network (C-RAN)
and air interface, and interference suppression technologies, such as beamforming of base stations with
large-scale antenna arrays and downlink power control.
• Uplink and downlink coordinated scheduling. Realizing the joint optimization of uplink and
downlink user performance through user pairing, base station and user power control, beam management,
etc. For complex networking scenarios, AI can be used to realize intelligent scheduling based on location
to reduce the huge overhead of channel and interference information acquisition.
In conclusion, compared with traditional duplex modes, such as FDD and TDD, the co-frequency
co-time full-duplex technique facilitates simultaneous transmission and reception at the same carrier fre-
quency, achieving much higher spectral efficiency as well as flexibly and optimally configuring transmission
resources. Considering the commercialization of co-frequency co-time full-duplex, the following techni-
cal challenges should be addressed: suppression of high-power dynamic self-interference signal, novel
interference-elimination mechanism under full-duplex systems, novel networking designs that support the
coexistence of TDD or FDD systems [4].

3.7 Holographic radio technology

The 6G network will be a common ultra-broadband mobile network with ultrahigh data rates, ultrahigh
data density (uHDD), and ultralow latency, which can support data exchange and computing collabo-
ration among high-performance super-intelligent terminals. Intelligent driving and intelligent industrial
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:23

Holographic spatial wave

field synthesis and
UTC-PD modulation
Optical
coupled
transform
antenna
and optical
array or
correlation
holographic
processing
metasurface
Spatial spectral holography
(holographic channel
measurement and reconstruction)

Figure 18 (Color online) Conceptual block diagram of holographic radio.

revolution have core requirements for 6G that create ubiquitous mobile ultra-broadband (uMUB), ultra-
broadband with low latency (uBBLLC), and uHDD. These services require an end-to-end collaborative
design of communication, sensing, and computing, causing challenges in computing capabilities, energy
efficiency, and latency. Therefore, 6G, as a new generation of mobile communication systems, needs new
theories and paradigms, as well as innovative technological breakthroughs [45,46]. In addition, compared
with 5G, 6G will have denser network deployment. The reduction of cell size and more antennas would
result in more severe intra- and inter-cell interference. Traditional interference cancellation technology
is no longer optimal, and innovative interference-processing methods are needed. A potential innovative
approach is the holographic radio technology, which leverages interference as a useful resource to develop
high-efficiency and high-precision holographic communication systems [46].
Holography, based on electromagnetic wave interferometry, records the electromagnetic field in space.
It reconstructs the target electromagnetic field with the information recorded by the interference of
the reference and signal waves. In holography, it is key that reference waves strictly cohere, and the
holographic recording sensor must be able to record the continuous wavefront phase of the signal wave
to accurately record high-resolution holographic electromagnetic fields [47]. Since both RF and light
waves are electromagnetic waves, holographic radio is very similar to optical holography. Intelligent
holographic radio (IHR) uses the holographic interferometry principle of electromagnetic waves to realize
dynamic reconstruction and real-time precision control of electromagnetic space. Through RF spatial
spectrum holography and holographic space wave-field synthesis technology, IHR realizes 3D holographic
imaging level, pixel-based ultrahigh-resolution spatial multiplexing, and provides nearly continuous and
nearly infinite multiplexing space, which can meet the 6G ultrahigh spectral efficiency, high data density,
and capacity requirements [48]. As Fourier optics and optical holography are mature, optical signal
processing techniques, such as fast Fourier transform (FFT) and inverse fast Fourier transform (IFFT),
can be employed to process RF signals. In the uplink, spatial spectrum holography employs the microwave
photonic technique to perform coherent optical up-conversion and processing on the received RF signals
and complete RF-to-optical mapping and correlation processing. This process is similar to real-time 3D
RF “light”-field imaging. Holographic channel measurement and spatial reconstruction of RF holography
are realized simultaneously by spatial spectrum holography, which can provide accurate feedback for space
wave field synthesis and modulation in the downlink. In the downlink, space wave-field synthesis and
modulation perform space-time precoding based on RF holographic space established by spatial spectrum
holography in the uplink to realize complex and accurate structural distribution of the electromagnetic
wave field in the target space. The entire process is similar to the “light”-field projection of real-time RF
holography (see Figure 18).
Among the three capabilities of holographic radio, RF holography, spatial spectrum holography, and
spatial wave field synthesis, the last two can modulate and adjust the electromagnetic fields of the whole
physical space in a full closed loop. This greatly improves the spectral efficiency and network capacity
and integrates imaging, sensing, and positioning into wireless communication [49]. In addition, IHR
applies to very-near-field communication and wireless energy transmission in distributed scenarios owing
to its evanescent wave property. The flexibility of IHR can be increased by deploying several holographic
meta-surfaces (Huygens’ meta-surface antennas) at ultra-low cost and power consumption and using dis-
tributed antenna systems as signal sources. As an interdisciplinary technology, IHR integrates wireless
communication, microwave photonics, optical computing, computational electro-magnetics and computa-
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:24

tional holography, RF tomography, and integrated photonics. It is a broad field open for study, especially
its unexplored techniques.
Holographic radio can not only realize RF holography, spatial spectrum holography, and spatial wave
field synthesis but also precisely modulate and adjust the electromagnetic field of the entire physical space
in a full closed loop through spatial spectrum holography and spatial wave field synthesis, which greatly
improve the spectral efficiency and network capacity and integrate imaging, sensing, positioning, and
wireless communication [49]. In addition, IHR can be applied to near-field communication and wireless
energy transmission in distributed scenes owing to its evanescent wave characteristics. By deploying sev-
eral ultralow-cost and -power-consumption holographic super surfaces (Huygens super surface antenna)
and distributed antenna systems as signal sources, a more flexible IHR solution can be realized. As an
interdisciplinary technology, IHR integrates wireless communication, microwave photonics, optical com-
puting, computational electromagnetics and computational holography, RF tomography, and integrated
photonics. However, many technologies still need to be researched [50].

3.7.1 Filled-aperture active antenna array

As mentioned above, holographic radio accurately reconstructs high-resolution holographic electromag-
netic fields by recording the continuous wavefront phase of the signal waves. A filled-aperture antenna
array is needed to receive and measure the continuous wavefront phase. To realize a continuous aperture
antenna array, when the traditional discrete interval antenna array is used, the number of oscillators is
nearly infinite, which is unrealistic, and the size, weight, and power (SWaP) of the system are unaccept-
able. Another scheme is to integrate several antenna elements into a compact space in the form of a
spatial continuous electromagnetic aperture, but this method is limited to passive super surfaces because
ultradense elements cannot realize RF feed networks.
This can be overcome using an ultra-broadband tightly coupled array (TCA) based on the current
sheet. In TCAs, a uni-traveling-carrier photodiode (UTC-PD) is bonded to the antenna elements using
the flip-chip technique to couple antenna oscillators [51]. The patch oscillator is directly integrated into
the electro-optic modulator (EOM) [52]. The output current of UTC-PD directly drives the antenna
oscillator, so the entire active antenna array has a very large bandwidth (about 40 GHz). Moreover,
this innovative solution eliminates the need for an ultradense RF feed network. It is achievable and
has obvious advantages in SWaP. Although UTC-PD tightly coupled antenna arrays can transform the
limited beam space of traditional antenna arrays into nearly infinite and continuous multiplexing space,
there is still a long way to go before practical applications. The main future research directions include
high-power UTC-PD and advanced electro-optic modulator technology, electromagnetic theory and anal-
ysis models of integration and packaging of photodiode and EOM tightly coupled holographic antenna
arrays, transparent fusion and seamless integration of UTC-PD tightly coupled antenna array and optical
signal processing or optical computing transparent fusion and seamless integration, etc. In addition, RIS
antenna, another form of holographic radio antenna, is an important research direction.

3.7.2 Holographic radio channel modeling

Unlike traditional large-scale MIMO, holographic radio uses Fresnel Fraunhofer interferometry, diffrac-
tion, and spatial correlation models, instead of the traditional Rayleigh propagation model, to model and
compute holographic radio space. Accurate calculations of communication performance require detailed
electromagnetic numerical calculations, i.e., algorithms and tools of computational electromagnetics and
computer-generated hologram, instead of zero-forcing, maximum ratio combining (MRC), and minimum
mean-square error (MMSE) in MIMO. Potential research directions on holographic radio theory and
modeling include holographic radio theory and modeling based on FFT, the convolution theorem, green
function, and sparse sampling, and compressed sensing Holographic radio theory and modeling based on
coded aperture correlations.

3.7.3 Integration of holographic radio and AI

Holographic radio can not only realize RF holography, spatial spectrum holography, and space wave field
synthesis but also precisely modulate and adjust the electromagnetic field of the entire physical space in
a fully closed loop. It can greatly improve spectral efficiency and network capacity and integrate imaging,
positioning, and wireless communication [49]. The generation and perception of extremely wide spectra
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:25

and holographic radio frequency would produce massive data, which can provide big data and radio
frequency spectrum datasets, enabling effective ML training. However, low-latency, high-reliability, and
scalable artificial intelligence architectures should be developed. 6G systems will integrate full-spectrum,
artificial intelligence, and RF holography. If only traditional electronic signal processing and computing
are adopted, its SWaP and latency will be a big challenge. A hierarchical and heterogeneous optoelectronic
computing and signal processing architecture would be an optimal solution for providing the required
energy efficiency, latency, and flexibility in 6G [53–55]. Holographic radio achieves ultrahigh coherence and
high parallelism of signals through coherent optical conversion of UTC-PD-integrated TCAs, facilitating
direct processing of signals in the optical domain. Considering that optical computing is more suitable
for linear computing [56], and more than 90% of holographic radio signal processing is linear computing,
performing most of the linear computing in real-time in the optical domain is key to providing high energy
efficiency and low latency over 6G air interfaces. Therefore, to integrate holographic radio and AI, it is
necessary to study an effective hybrid optoelectronic computing architecture and strengthen the research
on large-scale photonic integration [11] and electro-optic hybrid integration technology.
In summary, the holographic radio can dynamically reconstruct electromagnetic space with real-time
precision control and provide a quasi-continuous and -infinite multiplexing space. With the provided
ultrahigh-resolution spatial multiplexing, holographic radio can be used in wireless access, which requires
ultrahigh capacity/traffic density and ultralow latency, for example, enabling wireless industrial buses
operating in ultrahigh traffic density in smart factories. However, there are still some research challenges
to be addressed before commercialization, including the study of the fundamental theories and modeling,
hardware and physical layer design issues, integration of filled-aperture active antenna arrays, and high-
performance optical computing based on microwave photonics.

3.8 Ultralow-power IoT communications

6G networks will support trillions of devices, which are driven by the surge of IoT devices covering several
applications, such as smart cities, healthcare, and smart farms. An important category of applications
needs to deploy power-constrained devices that can work for prolonged periods, even several decades.
In these applications, devices are inaccessible, or it is difficult or expensive to reach them once they are
installed. A typical use case is smart farms, where numerous devices need to be deployed, and they cannot
be maintained or replaced. However, one major drawback of existing IoT solutions is the batteries need to
be frequently replaced due to their communication power consumption. Therefore, there is an urgent need
to develop novel communication technologies and system-level operational paradigms to enable batteries
with life spans of up to the life cycle of an IoT device, thereby facilitating cellular networks to provide
ubiquitous IoT connectivity for trillions of everyday objects.
Take backscatter communication (BackCom) technology as an example of extremely-low-power com-
munication [57, 58]. A BackCom device can convey its data by reflecting and modulating an ambient or
incident RF signal by switching antenna impedance. Without power-hungry and expensive radio compo-
nents, BackCom devices can operate at sub-milliwatt power consumption, and they can be manufactured
at extremely low cost. In addition, based on BackCom, symbiotic radio (SR), proposed in [59–61], can be
used to enable the BackCom and primary systems to benefit from each other. The key idea of SR is to
leverage cognitive radio to make BackCom and primary systems work at the same spectrum and create
multipath to boost the performance of primary systems by BackCom devices. Thus, BackCom has at-
tracted significant interest from both the academic and industry communities. However, to accommodate
a practical BackCom system in cellular networks, the following technology issues must be addressed.
First, with power constraints and low-end hardware, BackCom devices cannot employ traditional
physical-layer technology, and thus, it is necessary to develop new physical-layer technology for BackCom
systems. The main study points may include: to investigate new modulation schemes and the corre-
sponding SR technology so that both RF sources (e.g., cellular signal) and backscattered signals can be
separated at the receiver side, avoiding their interference; to develop low-complexity coding techniques
that can provide reliable transmission for resource-constraint BackCom devices; to evaluate the sources
of excitation signals, including dedicated excitation devices, relays, and ambient signals.
Second, it is necessary to investigate new networking and media-access-control layer (MAC-layer)
protocols to integrate BackCom with short ranges into wide-area cellular systems. Key research directions
include: (1) to introduce efficient multiple access methods to boost the resource utilization efficiency of
uplink and downlink to develop an appropriate network architecture so that the communication distance
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:26

can be extended; (2) to design a low-complexity and robust security protocol to thwart security attacks,
such as eavesdropping and jamming.
Third, low-complexity signal processing algorithms and circuits also need to be considered to further
reduce power consumption. For example, low-power baseband chips and oscillators can be used to
well balance the tradeoff between power consumption and communication rates, and a passive envelope
detector can be devised to wake up BackCom devices with extremely low power consumption.
In summary, BackCom can reduce power consumption in communication to the sub-milliwatt level,
making it a competitive technology for IoT applications in 6G networks. However, since BackCom is a
passive communication manner, it is difficult to directly apply traditional communication algorithms to
BackCom. There are several important open research problems, including new air-interface technology,
low-complexity signal processing schemes, lightweight networking, and MAC-layer protocol designs.

4 Conclusion
With the scaled commercial deployment of 5G networks, the global industry has initiated the exploration
of 6G. However, in general, 6G studies are still at the early stage. In this article, we present our vision
on 6G with a particular focus on the key candidate enabling technologies, including THz communication,
integrated sensing and communications, integrated intelligence and communication, ultramassive MIMO,
reconfigurable intelligent surface, co-frequency co-time full-duplex, holographic radio, and ultralow-power
IoT communication. Considering the industrialization and standardization promotion, potential appli-
cation scenarios and challenges are analyzed, the main research directions are highlighted, and future
research opportunities are identified. This study can provide a good reference for both academia and the
industry to prompt the rapid development of 6G.
Acknowledgements This work was supported by National Key R&D Program of China (Grant No. 2020YFB1806601).

References
1 IMT-2030 (6G) Promotion Group. White paper on 6G vision and candidate technologies. 2021. http://www.caict.ac.cn/
english/news/202106/t20210608 378637.html
2 Tong W, Zhu P Y. 6G: The Next Horizon: From Connected People and Things to Connected Intelligence. Cambridge:
Cambridge University Press, 2021
3 You X H, Wang C-X, Huang J, et al. Towards 6G wireless communication networks: vision, enabling technologies, and new
paradigm shifts. Sci China Inf Sci, 2021, 64: 110301
4 Yuan Y F, Zhao Y J, Zong B Q, et al. Potential key technologies for 6G mobile communications. Sci China Inf Sci, 2020, 63:
183301
5 Feng D Q, Lai L F, Luo J J, et al. Ultra-reliable and low-latency communications: applications, opportunities and challenges.
Sci China Inf Sci, 2021, 64: 120301
6 Wang Z Q, Han K F, Jiang J M, et al. Symbiotic sensing and communications towards 6G: vision, applications, and technology
trends. 2021. ArXiv:2106.15197
7 Marzetta T L. Noncooperative cellular wireless with unlimited numbers of base station antennas. IEEE Trans Wireless
Commun, 2010, 9: 3590–3600
8 Ngo H Q, Larsson E G, Marzetta T L. Energy and spectral efficiency of very large multiuser MIMO systems. IEEE Trans
Commun, 2013, 61: 1436–1449
9 Jeon C, Li K, Cavallaro J R, et al. On the achievable rates of decentralized equalization in massive MU-MIMO systems.
In: Proceedings of IEEE International Symposium on Information Theory Conference, 2017
10 Zhong Z, Fan L, Ge S. FDD massive MIMO uplink and downlink channel reciprocity properties: full or partial reciprocity?
In: Proceedings of IEEE Global Communications Conference, 2020
11 Garcia N, Wymeersch H, Slock D T M. Optimal precoders for tracking the AoD and AoA of a mmWave path. IEEE Trans
Signal Process, 2018, 66: 5718–5729
12 Witrisal K, Meissner P, Leitinger E, et al. High-accuracy localization for assisted living: 5G systems will turn multipath
channels from foe to friend. IEEE Signal Process Mag, 2016, 33: 59–70
13 Tse D, Viswanath P. Fundamentals of Wireless Communication. Cambridge: Cambridge University Press, 2005
14 3rd Generation Partnership Project (3GPP). NR and NG-RAN Overall description. TS 38.300. https://portal.3gpp.org/
desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3191
15 I C L, Han S, Bian S. Energy-efficient 5G for a greener future. Nat Electron, 2020, 3: 182–184
16 Tariq F, Khandaker M R A, Wong K K, et al. A speculative study on 6G. IEEE Wireless Commun, 2020, 27: 118–125
17 Basar E, Di Renzo M, de Rosny J, et al. Wireless communications through reconfigurable intelligent surfaces. IEEE Access,
2019, 7: 116753
18 Liang Y-C, Chen J, Long R Z, et al. Reconfigurable intelligent surfaces for smart wireless environments: channel estimation,
system design and applications in 6G networks. Sci China Inf Sci, 2021, 64: 200301
19 Glybovski S B, Tretyakov S A, Belov P A, et al. Metasurfaces: from microwaves to visible. Phys Rep, 2016, 634: 1–72
20 Chen H T, Taylor A J, Yu N. A review of metasurfaces: physics and applications. Rep Prog Phys, 2016, 79: 076401
21 Cui T J, Qi M Q, Wan X, et al. Coding metamaterials, digital metamaterials and programmable metamaterials. Light Sci
Appl, 2014, 3: e218
22 Shadrivov I V, Neshev D N. Tunable Metamaterials. Singapore: World Scientific, 2017. 9: 387–418
23 Zhu B, Feng Y, Zhao J, et al. Switchable metamaterial reflector/absorber for different polarized electromagnetic waves. Appl
Phys Lett, 2010, 97: 051906
24 Tao Z, Wan X, Pan B C, et al. Reconfigurable conversions of reflection, transmission, and polarization states using active
metasurface. Appl Phys Lett, 2017, 110: 121901
25 Zhao J, Cheng Q, Chen J, et al. A tunable metamaterial absorber using varactor diodes. New J Phys, 2013, 15: 043049
 Wang Z Q, et al. Sci China Inf Sci May 2022 Vol. 65 151301:27

26 Yang H H, Chen X B, Yang F, et al. Design of resistor-loaded reflectarray elements for both amplitude and phase control.
Antenn Wirel Propag Lett, 2017, 16: 1159–1162
27 Cheng C-C, Lakshminarayanan B, Abbaspour-Tamijani A. A programmable lens-array antenna with monolithically integrated
MEMS switches. IEEE Trans Microwave Theor Techn, 2009, 57: 1874–1884
28 Phillion R, Okoniewski M. Improving the phase resolution of a micromotor-actuated phased reflectarray. In: Proceedings of
IEEE Microsystems and Nanoelectronics Research Conference, 2008
29 Chicherin D, Dudorov S, Lioubtchenko D, et al. MEMS-based high-impedance surfaces for millimeter and submillimeter wave
applications. Microw Opt Technol Lett, 2006, 48: 2570–2573
30 Shrekenhamer D, Chen W C, Padilla W. Liquid crystal tunable metamaterial absorber. Phys Rev Lett, 2013, 110: 77403
31 Ju L, Geng B, Horng J, et al. Graphene plasmonics for tunable terahertz metamaterials. Nat Nanotech, 2011, 6: 630–634
32 Lee S H, Choi M, Kim T T, et al. Switching terahertz waves with gate-controlled active graphene metamaterials. Nat Mater,
2012, 11: 936–941
33 Jadidi M M, Sushkov A B, Myers-Ward R L, et al. Tunable terahertz hybrid metal-graphene plasmons. Nano Lett, 2015, 15:
7099–7104
34 Perruisseau-Carrier J, Bongard F, Golubovic-Niciforovic R, et al. Contributions to the modeling and design of reconfigurable
reflecting cells embedding discrete control elements. IEEE Trans Microwave Theor Techn, 2010, 58: 1621–1628
35 Salti H, Fourn E, Gillard R, et al. Minimization of MEMS breakdowns effects on the radiation of a MEMS based reconfigurable
reflectarray. IEEE Trans Antenn Propagat, 2010, 58: 2281–2287
36 Tang W, Chen M Z, Chen X, et al. Wireless communications with reconfigurable intelligent surface: path loss modeling and
experimental measurement. IEEE Trans Wireless Commun, 2021, 20: 421–439
37 Dou J W, Chen Y J, Zhang N, et al. On the channel modeling of intelligent controllable electro-magnetic-surface. Chin J
Radio Sci, 2020. doi: 10.12265/j.cjors.2020195
38 Jian M, Zhao Y. A modified off-grid SBL channel estimation and transmission strategy for RIS-assisted wireless communication
systems. In: Proceedings of IEEE International Wireless Communications and Mobile Computing Conference, 2020
39 He Z Q, Yuan X J. Cascaded channel estimation for large intelligent metasurface assisted massive MIMO. IEEE Wireless
Commun Lett, 2020, 9: 210–214
40 Zhang C, Patras P, Haddadi H. Deep learning in mobile and wireless networking: a survey. IEEE Commun Surv Tutor, 2019,
21: 2224–2287
41 Tang W, Dai Y J, Chen M Z, et al. Design and implementation of MIMO transmission through reconfigurable intelligent
surface. In: Proceedings of IEEE 21st International Workshop on Signal Processing Advances in Wireless Communications,
2020. 1–5
42 Nayeri P, Yang F, Elsherbeni A Z, et al. Theory, Designs, and Applications. Hoboken: John Wiley & Sons, 2018
43 Hum S V, McFeetors G, Okoniewski M. Integrated MEMS reflectarray elements. In: Proceedings of IEEE European Conference
on Antennas Propagation, 2006
44 Bi S, Ho C K, Zhang R. Wireless powered communication: opportunities and challenges. IEEE Commun Mag, 2015, 53:
117–125
45 Wu Q Q, Zhang R. Towards smart and reconfigurable environment: intelligent reflecting surface aided wireless network. IEEE
Commun Mag, 2020, 58: 106–112
46 Zong B Q, Fan C, Wang X Y, et al. 6G technologies: key drivers, core requirements, system architectures, and enabling
technologies. IEEE Veh Technol Mag, 2019, 14: 18–27
47 Goodman J W. Introduction to Fourier Optics. New York: McGraw Hill, 1968
48 Bjornson E. The role of academia in beyond-5G wireless research. IEEE ComSoc Technology News, 2020.
https://www.comsoc.org/publications/ctn/role-academia-beyond-5g-wireless-research
49 Xu B, Qi W, Zhao Y, et al. Holographic radio interferometry for target tracking in dense multipath indoor environments.
In: Proceedings of IEEE 9th International Conference on Wireless Communications and Signal, 2017. 1–6
50 Saad W, Bennis M, Chen M. A vision of 6G wireless systems: applications, trends, technologies, and open research problems.
IEEE Network, 2020, 34: 134–142
51 Konkol M R, Ross D D, Shi S, et al. High-power photodiode-integrated-connected array antenna. J Lightwave Technol, 2017,
35: 2010–2016
52 Murata H, Wijayanto Y N, Okamura Y. Integration of patch antenna on optical modulators. IEEE Photonic Soc Newslett,
2014, 28: 4–7
53 Zong B, Zhao X, Wang J, et al. Photonics defined radio: a new paradigm for future mobile communication of B5G/6G.
In: Proceedings of 6th International Conference on Photonics, Optics and Laser Technology, 2018. 1–6
54 Murakowski J, Schneider G J, Shi S, et al. Photonic probing of radio waves for k-space tomography. Opt Express, 2017, 25:
15746–15759
55 Barber Z W, Harrington C, Krishna M R, et al. Spatial-spectral holographic real-time correlative optical processor with > 100
Gb/s throughput. Appl Opt, 2017, 56: 5398–5406
56 Prucnal P R, Shastri B J. Neuromorphic Photonics. Boca Raton: CRC Press, 2017
57 Han K F, Huang K B. Wirelessly powered backscatter communication networks: modeling, coverage, and capacity. IEEE
Trans Wireless Commun, 2017, 16: 2548–2561
58 Zhang Q, Guo H, Liang Y C, et al. Constellation learning-based signal detection for ambient backscatter communication
systems. IEEE J Sel Areas Commun, 2019, 37: 452–463
59 Long R, Liang Y C, Guo H, et al. Symbiotic radio: a new communication paradigm for passive Internet of Things. IEEE
Internet Things J, 2020, 7: 1350–1363
60 Liang Y C, Zhang Q, Larsson E G, et al. Symbiotic radio: cognitive backscattering communications for future wireless
networks. IEEE Trans Cogn Commun Netw, 2020, 6: 1242–1255
61 Guo H, Liang Y C, Long R, et al. Cooperative ambient backscatter system: a symbiotic radio paradigm for passive IoT. IEEE
Wireless Commun Lett, 2019, 8: 1191–1194