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

net/publication/347982266

A Comprehensive Survey of 6G Wireless Communications

Preprint · December 2020

CITATIONS READS
0 3,075

7 authors, including:

Yang Zhao Wenchao Zhai

Agency for Science, Technology and Research (A*STAR) China Jiliang University
34 PUBLICATIONS 1,339 CITATIONS 24 PUBLICATIONS 255 CITATIONS

SEE PROFILE SEE PROFILE

Jun Zhao Zhang Tinghao

Nanyang Technological University Nanyang Technological University
377 PUBLICATIONS 10,484 CITATIONS 13 PUBLICATIONS 118 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Yang Zhao on 18 February 2021.

The user has requested enhancement of the downloaded file.

 1

A Comprehensive Survey of 6G Wireless

Communications
Yang Zhao, Graduate Student Member, IEEE, Wenchao Zhai, Member, IEEE, Jun Zhao, Member, IEEE,
Tinghao Zhang, Graduate Student Member, IEEE, Sumei Sun, Fellow, IEEE, Dusit Niyato, Fellow, IEEE, and
Kwok-Yan Lam, Senior Member, IEEE

Abstract—While fifth-generation (5G) communications are be- the high computation, low latency, etc., which are required
ing rolled out worldwide, sixth-generation (6G) communications by the 6G’s applications [3]–[6]. Compared with 5G, 6G has
have attracted much attention from both the industry and the stricter requirements in power consumption, lower latency,
academia. Compared with 5G, 6G will have a wider frequency
band, higher transmission rate, spectrum efficiency, greater higher reliability, privacy and security, etc. Also, it provides
connection capacity, shorter delay, broader coverage, and more the better service of quality (QoS) and broader coverage than
robust anti-interference capability to satisfy various network wireless communications in the past. 6G is considered to be a
requirements. This survey presents an insightful understanding revolutionary generation of wireless communication because
of 6G wireless communications by introducing requirements, of the growing roles of intelligence, autonomy will bring new
features, critical technologies, challenges, and applications. First,
we give an overview of 6G from perspectives of technologies, vigor and vitality into the communications.
security and privacy, and applications. Subsequently, we in-
troduce various 6G technologies and their existing challenges Radar-assisted Drone Taxi
Drone network Mitola radio BBU Pool
in detail, e.g., artificial intelligence (AI), intelligent surfaces,
THz, space-air-ground-sea integrated network, cell-free massive
MIMO, etc. Because of these technologies, 6G is expected to
outperform existing wireless communication systems regarding Edge AI
the transmission rate, latency, global coverage, etc. Next, we Fog-RAN
discuss security and privacy techniques that can be applied to
F-AP
protect data in 6G. Since edge devices are expected to gain F-AP
LIFI F-AP
popularity soon, the vast amount of generated data and frequent F-UE F-UE
C-UE
data exchange make the leakage of data easily. Finally, we predict Software-deÀned materials
F-UE
real-world applications built on the technologies and features Pervasive/collective AI C-UE
of 6G; for example, smart healthcare, smart city, and smart
manufacturing will be implemented by taking advantage of AI.
Vi Re
C-UE
UE

rtu a
al lity
Cellular BS

/A (
ug VA
Index Terms—6G, Wireless Communications, Survey, Index
m R
Cellular BS

en )
te
Modulation (IM), Artificial Intelligence (AI), Intelligent Reflect-
d
Visible light communication Remote surgery
ing Surfaces (IRS), Artificial Internet of Things (AIoT). Robo Taxi

I. I NTRODUCTION F-UE: Fog user equipment,¬ ¬C-UE: Cellular UE,¬ ¬F-AP:¬ Fog access point

As 5G communication networks are being deployed com- Fig. 1: The vision of 6G.
mercially [1], the academic and industry start developing 6G
wireless communication systems. Currently, the rapid growth To be more specific, Table I lists details of key performance
of data-centric intelligent systems has brought significant indicators (KPI) of 6G. Various candidate technologies have
challenges to 5G wireless systems. For example, the haptic been proposed to overcome the bottleneck of existing wire-
Internet-based telemedicine requires that the delay of air less communication systems to meet 6G’s requirements. For
interface is less than 0.1 millisecond (ms) [2]. But the existing example, artificial intelligence (AI) is expected to enable a
delay is only 1ms, which is not satisfying. 5G’s ubiquitous significant paradigm shift in 6G wireless networks, including
mobile ultra-broadband, ultrahigh data density, and ultrahigh- machine learning, deep learning, etc. [7]. The rapid consump-
speed-with-low-latency communications cannot fully satisfy tion of the spectrum resources makes the efficient utilization
of the spectrum highly important. In order to improve the
Yang Zhao, Jun Zhao, Tinghao Zhang, Dusit Niyato and Kwok-Yan Lam are
with School of Computer Science and Engineering, Nanyang Technological spectral efficiency, various techniques will be introduced in
University, Singapore. (Email: s180049@e., junzhao, tinghao001@e., dniyato, 6G, including existing technologies like index modulation
kwokyan.lam}@ntu.edu.sg). (IM) and free duplex. Moreover, new technologies such as
Wenchao Zhai is with College of Information Engineering, Jiliang Univer-
sity, Hangzhou, China. (Email: zhaiwenchao@cjlu.edu.cn). intelligent reflecting surface (IRS) may enhance the spectrum
Sumei Sun is with Institute for Infocomm Research (I2R), Agency for efficiency by leveraging the passive reflecting elements with-
Science, Technology and Research (A*STAR), Singapore, and Infocomm out using the external power source. Furthermore, because the
Technology Cluster, Singapore Institute of Technology, Singapore. (Email:
sunsm@i2r.a-star.edu.sg). consumption of the spectrum frequency is fast in the existing
*Part of this paper is published in the FICC conference. wireless communication systems, 6G also explores the sub-
terahertz (sub-THz), visible light communication (VLC), and TABLE I: Requirements and Features of 6G [5,18]–[20].
terahertz (THz), which use frequencies ranging from 100 GHz
Requirements 6G
to 3 THz [4,8]–[15]. Besides, to obtain the target of the
global coverage, 6G wireless communications use the space- Service types MBRLLC/mURLLC/HCS/MPS
air-ground-sea integrated network (SAGSIN) to connect the Service level Tactile
communication systems from the sky to the deep sea [16]. Device types Smart implants/ CRAS/
Also, the blockchain is used to manage and share the spec- XR and BCI equipment/
trum resources, which may eliminate the central authority
Sensors and DLT devices
because blockchains are immutable, untampered decentralized
databases [17]. In addition to the spectral efficiency, the Jitters 1 µsec
improvement of the efficiency of the energy consumption is Individual data rate 100 Gbps
quite necessary as well. Therefore, the simultaneous wire- Peak DL data rate ≥ 1 Tbps
less information and power transfer (SWIPT) technology is Latency 0.1 msec
leveraged to prolong the battery’s span to reduce the power Mobility up to 1000 km/h
consumption.
Reliability up to 99.99999%
Additionally, security and privacy technologies are being
- sub-THz band
eye-catching due to the explosive growth of data generated by Frequency bands
- Non-RF, e.g, optical, VLC, laser · · ·
edge devices. As the 6G generation is coming, the number of
edge devices will increase rapidly. Therefore, more security Multiplexing Smart OFDMA plus IM
and privacy related techniques will be explored to protect Power consumption Ultra low
data’s confidentiality. Finally, when above technologies are Processing delay ≤ 10ns
getting mature, emerging applications illustrated by Fig. (1), Maximum rate 100Gb s−1
such as the smart healthcare and the holography radio, will
Security and privacy Very high
be launched to enable 6G to be a part of our life. More new
Network orientation Service-centric
services will also be supported, for instance, smart wearable
devices, computing reality devices, autonomous vehicles, im- Wireless power transfer/
plants, sensing, and 3D mapping [3]. Wireless charging Support (BS to devices power transfer)
Contributions. The contributions of this survey are sum- Smart city components Integrated
marized below. Autonomous V2X Fully
• Existing papers on 6G pay more attention to predicting Localization precision 1 cm on 3D
technologies that may be used in the future, and few
of them give a summary. Our paper surveys almost
all of the existing visions of 6G and summarizes them Section II gives an overview of emerging technologies that
in detail, including both physical and network layer enable the paradigm shift in 6G wireless networks. From
technologies. Furthermore, we give a deep insight into Section V to Section VIII, we present each technology of 6G
current technologies potentially utilized in 6G according in detail. Section IX discusses potential security and privacy
to how these technologies satisfy the requirements of problems and their corresponding solutions existing in the 6G
future 6G network, thus facilitating the understanding of wireless communications. Section X lists some applications
their respective advantages comprehensively. that will be enabled by 6G technologies. Section XI concludes
• We highlight some technologies that may be essential this paper.
for 6G, including index modulation, artificial intelligence, Notations. Boldfaced letters denote vector and boldfaced
intelligent surfaces, simultaneous wireless information, capital letters represent matrix. ||·||F represents the Frobenious
power transfer, etc. These technologies’ advantages and norm of a matrix. | · | is the magnitude of a complex number.
challenges are discussed in our manuscript, which gives The subscripts (·)T , (·)∗ , and (·)H define the operation of
future work directions for 6G applications. transpose, conjugate, and Hermitian transpose, respectively. IN
• Not only the principles of the technologies themselves are represents an N × N identity matrix. Table II summarizes
presented, but their relevant variations are also provided. abbreviations used in this paper.
Therefore, readers can have a comprehensive understand-
ing of these technologies and their evolution in solving II. A N OVERVIEW OF 6G T ECHNOLOGIES
sophisticated problems.
• Applications of potential 6G technologies are also investi- 6G is expected to outperform 5G in multiple specifications.
gated in our manuscript. By summarizing the applications We highlight six of them in Fig. (1), including the frequency,
in various scenarios, we help recognize the 6G network’s individual data rate, peak data rate, spectral efficiency, mo-
differences from its 5G counterpart. Therefore, it is easy bility, and latency. From Fig. (1), 6G requires the lower
to have an exact blueprint for the development of a future latency, higher frequency and data transmission rate, faster
network. mobility, and better spectral efficiency than 5G. In the follow-
Organization. The rest of paper is organized as follows. ing, we present an overview of promising technologies that

2
 TABLE II: Table of Abbreviations.
5G Fifth generation Frequency
6G Sixth generation
IM Index Modulation
IoE Internet of Everything Up to visible light
Simultaneous Wireless Information
SWIPT Individual data rate Up to millimetre
and Power Transfer Latency
AI Artificial Intelligence 100 Gbps wave
≤ 0.1 ms
IRS Intelligent Reflecting Surfaces
LIS Large Intelligent Surfaces 0.1~1 Gbps
THz Terahertz Communications ≤ 1 ms
SR Symbiotic Radio
VLC Visible Light Communication
D2D Device-to-Device Communication
CFmMM Cell-Free massive MIMO Tens of Gbps
SAGSIN Space-Air-Ground-Sea Integrated Network 500+ km/h
30 bps/Hz
≥ 1 Tbps
NIB Network in Box
AIoT Artificial Internet of Things Up to 1000
100 bps/Hz Mobility
mMIMO massive multiple input multiple output km/h
FD Full-duplex Peak data rate
APM Amplitude Phase Modulation
PSK Phase-Shift Keying
QAM Quadrature Amplitude Modulation
ML Maximum Likelihood Spectral Efficiency
SD-IM Spatial-Domain IM 5G 6G
SSK Spatial Shift Keying
FD-IM Frequency-Domain IM Fig. 2: Comparison between 5G and 6G communications.
TD-IM Time-Domain IM
CD-IM Channel-Domain IM
SM Spatial modulation
OFDM Orthogonal Frequency Division Multplexing (IM), simultaneous wireless information and power trans-
TDD Time Division Duplex fer (SWIPT), visible light communication (VLC), device-
RF Radio Frequency
MMSE Minimum Mean Square Error to-device (D2D), and network in box [19,21]–[27]. These
CSI Channel State Information technologies have potentials to lay the foundation for future’s
QIMMA Quadrature IM Multiple Access communications and applications.
NOMA Non-Orthogonal Multiple Access
BICM Bit-Interleaved Coded Modulation
Artificial Intelligence. Artificial intelligence (AI) provides
UM-MIMO Ultra-Massive MIMO intelligence and automation to wireless networks by imitating
LIM Large Intelligent Metasurface human thought processes and intelligent behaviors. As 6G
SDS Software-Defined Surface is envisioned to offer full automation, AI becomes one of
SDM Software-Defined Metasurface
EM Electromagnetic the critical and highly valued technologies in the future’s 6G
QoS Quality of Service wireless communication systems. AI is also considered to be
BS Base Station a modern tool for data analysis. By learning from big data,
UE User Equipment
SINR Signal-to-Interference-Plus-Noise Ratio
machine learning models are built to make correct decisions
BER Bit-Error Rate automatically. Thus, AI technologies, for example, machine
HD Half-Duplex learning, are expected to play an essential role in networks
RFID Radio Frequency Identification when they are too dynamic and complicated to be analyzed
TS Time Switching
PS Power Splitting by humans. In particular, scenarios involve complex processes
SIC Successive Interference Cancellation such as the joint optimization in network design, resource
SER Symbol Error Rate management, and resource allocation.
PEP Pairwise Error Probability
SNR Signal-to-Noise Ratio
Intelligent Surfaces. There are two kinds of intelligent
surfaces, including the large intelligent surfaces (LIS) and in-
telligent reflecting surfaces (IRS). Both of them are considered
contribute to the development of 6G wireless communications; as promising candidate technologies for 6G. The concept of
Subsequently, we discuss security and privacy problems and deploying antenna arrays as the LIS in massive MIMO systems
introduce potential applications. is initially proposed by Hu et al. [28,29]. Compared with the
beamforming technology, where many antennas are required to
make the signals focused, the LIS is electromagnetically active
A. Technologies in the physical environment and impose little restrictions on
In this section, we give a brief introduction to the how antennas spread. Therefore, the LIS can avoid the adverse
most eye-catching technologies pertaining to 6G, includ- effects of antenna correlations. However, due to the surfaces’
ing artificial intelligence (AI), intelligent surfaces, terahertz active property, the LIS consumes many power resources and
(THz), symbiotic radio (SR), free duplex, cell-free mas- are not power efficient.
sive MIMO (CFmMM), space–air–ground-sea integrated net- To overcome the shortness of power consumption, the IRS is
work (SAGSIN), blockchain-based network, index modulation proposed by Wu et al. [30] to replace the active antennas used

3
by the LIS. The IRS uses passive reflecting elements to reflect databases which are constructed based on the theory of the
signals intelligently such that the data rate and the network hash tree, and they are tamper-proof and hard to reverse [17].
coverage can be improved. Thus, the IRS is considered to Blockchains have attributes like auditability, data integrity,
tune the wireless environments to increase the spectrum and and transparency [37]. Thus, blockchains can be used to
energy efficiency while consuming little power. Therefore, by manage spectrum resources without using a centralized au-
controlling the incident waves’ reflection characteristics, the thority. Besides, blockchains are also suitable to protect data’s
signal quality can be improved significantly. security and privacy or control access. Fan et al. [38] propose
Terahertz. Terahertz (THz) is considered as one of the an efficient and secure blockchain-based privacy-preserving
key technologies for the 6G wireless communications [4,9]– scheme, combining access policy, and encryption technology
[13]. 5G defines below 100 GHz as the millimeter-wave to guarantee data privacy. Kotobi et al. [39] use blockchain as
bands, whereas 100 GHz - 3 THz is categorized as THz a decentralized database to improve the access protocols and
band in 6G [31]. The above 90 GHz band is merely used secure spectrum sharing in mobile cognitive radio networks.
for scientific service, which has not been fully explored. Yang et al. [40] present a trusted authentication architecture
Therefore, it is envisioned to support the increased wireless based on blockchain for could radio over a fiber network.
network capacity [15]. The THz also enables the ultra-high Index Modulation. Index modulation (IM) conveys the
bandwidth and ultra-low latency communication paradigms, source information bits through the classical APM signals
which caters to the needs of many emerging applications such and the index selection of resource entities. Therefore, IM
as autonomous driving, Internet of Things. However, as the can improve the transmission rate and is potentially to be
frequency increases, the path loss increases as well, which used in 6G. Chau et al. [41] suggest that information bits
makes it challenging for long-range communications. Thus, can be transmitted through the index of the antennas in
the THz is particularly suitable for high bit-rate short-range MIMO systems. They name such a technique as space shift
communications [14]. keying (SSK) and combine the SSK with classical linear
Symbiotic Radio. Symbiotic radio (SR) [32] is a new modulation, amplitude-phase modulation (APM); for example,
technique that maintains the advantages and make up the dis- space modulation (SM) is proposed based on the same idea
advantages of cognitive radio (CR) and ambient backscattering in SSK. In SM technology, the source information bits are
communications (AmBC). SR is considered as one of the divided into two parts: the index of the transmit antennas and
promising solutions to build 6G as a spectrum and energy- the other parts for the APM. Therefore, SM can significantly
efficient communication system. In addition, IRS technology increase the transmission rate by sending the spare information
can further contribute to improving the performance of the bits through the traditional APM transceiver’s antenna index.
transmission by enhancing the backscattering link signal. Apart from the antennas, other resource entities can also be
Free Duplex. 6G will eliminate the difference between indexed to transmit the additional information bits. These
FDD and TDD, and the frequency sharing is based on the resource entities include time slots, sub-carriers, and channel
requirements, which we call free duplex [33]–[35]. Hence, state. Also, this modulation class, where the spare information
the spectrum resource allocation in 6G will be more efficient bits are transmitted through the index of resources, is called
and effective. Free duplex technology will help 6G increase index modulation. Orthogonal frequency division multiplexing
its transmission rate, throughput, and reduce its transmission access (OFDMA) plus IM will be essential technologies to
latency. significantly increase the throughput to support more users to
Cell-Free Massive MIMO. Cellular wireless networks are access the 6G network.
based on cellular topology. An area is then divided into Simultaneous Wireless Information and Power Transfer.
multiple cells according to the topology, and each cell is served 6G is supposed to be a complex network where a large
by one base station. The drawback of the cell-based wireless variety of smart devices are accessed to the system and are
network is that if the device is at the edge the cell, its signal is required to communicate with others at anytime, and the
pretty weak. By tackling the low communication ability at the lifetime of the battery-charging modules is also required to
edge of the cell, the cell-free massive multiple-input-multiple- fulfill the constraints of ultra-low power consumption as listed
output (CFmMM) concept is proposed, which removes the in Table I. To prolong the life span of various devices in the
cells of the network. The idea of CFmMM means that one network, simultaneous wireless information and power transfer
device is no longer attached to a single base station; instead, (SWIPT) technology is proposed. SWIPT enables sensors
all base stations coherently serve the device in an area. to be charged exploiting wireless power transfer; thereafter,
Space-Air-Ground-Sea Integrated Network. 6G wireless battery-free devices can be supported in 6G, reducing the
communication networks will integrate space-air-ground-sea network’s power consumption substantially [42]. Zhang et
networks to achieve the global coverage [16], i.e., 6G will con- al. [43] describe SWIPT technology from the perspective
struct a space-air-ground-sea integrated network (SAGSIN). of a scientific hypothesis and engineering practice in detail.
As expected in the white paper released in January 2020, Subsequently, in [44]–[47], performance on the outage prob-
the 6G network should cover environments, including sky ability, throughput, and sum rate for non-orthogonal multiple
(10,000km) and sea (20 nautical miles) [36]. By integrating access (NOMA) networks with SWIPT are derived. Bariah et
the three networks, SAGSIN can cope with various users and al. [48,49] first conduct error probability analysis of the
services’ growing traffic demands. NOMA-based networks with SWIPT. The error probability
Blockchain-based Network. Blockchains are distributed performance of SWIPT is severely affected by the power

4
splitting factor, which provides a principle for designing the be fully realized as smart manufacturing will achieve high-
SWIPT system in practical engineering. precision manufacturing [64]. Intelligent robots connected by
Visible Light Communication. Visible light communica- ubiquitous 6G network enable manufacturing systems to carry
tion (VLC) uses visible light between 400 and 800 THz out complex and dangerous tasks without risking people’s
(780–375 nm) to communicate information. It provides ultra- life [63]. Moreover, the smart home that equips with intelligent
high bandwidth (THz), zero electromagnetic interference, IoT devices will provide a comfortable living environment to
free abundant unlicensed spectrum, and very-high-frequency people [65], and 6G allows the smart home to ensure the
reuse [50]. Therefore, VLC helps to develop the short range residents’ security. In terms of traffic and transportation, the
communications of 6G networks [51]. Besides, 1G-5G wire- sophisticated sensing and planning algorithms can be deployed
less networks have utilized micro-wave communications over for traffic optimization [65]. Other applications such as smart
the sub-6 GHz band, whose resources are almost used up [13]. grid [66] and unmanned aerial vehicle [67] will also be
Correspondingly, since more smart devices are integrated enhanced with the aid of 6G.
into the network, and an explosive growth appears for the
area traffic capacity, the high capacity becomes an essential
requirement of 6G. To fulfill the requirement of capacity or Supervised
Learning
data rate, VLC technologies are potential to be used in 6G.
Device-to-Device. Device-to-Device (D2D) communication
Deep
represents the direct communication between devices without Learning
going through base stations, and the communication can be
done under licensed (i.e., cellular network) or unlicensed
spectrum (i.e., WiFi) [52,53]. D2D improves the throughput, Deep
Unsupervised Reinforcement
energy efficiency, delay, and fairness of the communica- Learning Learning
tion [54,55]. As the number of edge devices is explosively
increasing, D2D is gaining more attention and getting more Machine Learning
widely implemented [56].
Artificial Intelligence

B. Security and Privacy Fig. 3: The relationship of AI, ML and DL [68].

In 6G, every edge device is envisioned to connect to the
Internet, and AI applications will be widely used by edge
devices. Most AI applications are data-driven, which raises the III. A RTIFICIAL I NTELLIGENCE
concern of security and privacy of the collected data [57]. For Artificial Intelligence (AI) provides intelligence to wire-
example, wearable devices may collect clients’ health status less networks by simulating human thought processes and
everyday, but these data are private and sensitive. Clients’ intelligent behaviors [62,69]. Machine learning is the widely
privacy may be compromised if the data are leaked. Moreover, recognized the start-of-the-art technology to achieve AI, which
Akhtar et al. [58] investigate the adversarial attacks against learns from datasets to find out their pattern. Commonly used
deep learning models and show how the attacks happen in machine learning methods include deep learning, supervised
practice, whereas some other security and privacy issues are learning, unsupervised learning, and deep reinforcement learn-
identified in [59]. Lovén et al. [7] identify the challenges of ing [70]. Depending on the labels of the training datasets and
the edge AI and aim to improve edge computing and AI-based types of training tasks, different machine learning approaches
approaches security via security systems. Zhou et al. [60] are applied. Deep learning is a branch of ML, which uses
combine deep learning and edge computing, developing robust the principle of backpropagation for building models. Besides,
mobile crowdsensing that can conduct data validation and the supervised learning, unsupervised learning, and deep re-
local processing. Sattiraju et al. [61] analyze the feasibility of inforcement learning also can leverage deep learning methods
utilizing machine learning techniques such as recurrent neural for training. The relation among aforementioned machine
networks in the physical layer and propose an unsupervised learning methods are illustrated in Fig. (3). In 6G, more
learning algorithm to improve the physical layer security. advanced training approaches are introduced, including model-
driven deep learning, federated learning, and explainable ar-
tificial intelligence. Specifically, model-driven deep learning
C. Applications uses the professional experience to build a model instead
Empowered by 6G, a wide range of AI applications will of pure data-driven training [71,72]. Federated learning is a
evolve into “connected intelligence” [62], hence facilitating new machine learning technique that enables users to train
every aspect of our daily life. For example, advanced AI models with local data without sending raw data to the central
approaches can be employed in network management or auton- server. Since existing deep learning training is still a black-
omy to save manpower [7,18]. In addition, 6G revitalizes smart box, scientists are unable to explain its working principles.
healthcare by providing real-time health monitoring [63], However, to apply deep learning training methods in real-
high-precision medical treatment [9], and reliable privacy world applications, the explanation is required to gain trust.
protection [20]. With the advent of 6G, Industry 4.0 will Hence, the explainable artificial intelligence is motivated.

5
 Connected Autonomous Un
Unlicensed Spectrum
Electric Vehicles Access
Cellular Network

Smart city Fleet

Conventional UE Autonomous Wi-Fi Access Points
Charging
Car

Smart Smart Smart Smart 5G Fair-Access

Traffic Grid Health Home Beamforming Mechanism

Macro Cell
Macro Cell

Co-Tier Interference
Small Cell

Small Cell

Fig. 4: An AI-Enabled 6G wireless network and related applications.

With the increasing popularity of AI technologies, 6G en-

ables various applications in communications such as network Base Station
planning, network optimization, network operation and main-
tenance, and autonomous vehicles [8,18,51,69,71,73]–[78] as FL Global Model
shown in Fig. (4). In particular, AI, residing in new local
“clouds” and “fog” environments, helps create many novel FL Local Model
applications using sensors that can be embedded into every Laptop
corner of our life [10]. For example, Table III lists applications User 1
and their corresponding deep learning methods and tools in the
physical layer, MAC layer, and network layer. Deep learning,
which is considered as the vital ingredient of AI technologies, User 5
has been widely used in wireless networks [70]. It will play an User 2
essential role in various areas, including semantic communi- User 3 User 4
cations, holistic management of communication, computation,
caching, control resources areas, etc., which may contribute
to the paradigm-shift of 6G. In the following, we introduce Fig. 5: Federated Learning.
popular AI methods and their applications in detail.
Supervised Learning. The supervised learning trains the
machine learning model using labelled training data [18]. labels. Commonly used unsupervised learning techniques in-
Thus, supervised learning is particularly suitable for tasks clude clustering, autoencoders, deep belief nets, generative
which can be trained using the historical data. There are some adversarial networks, and the expectation–maximization algo-
well developed algorithms that are used in 6G for classification rithm. In the physical layer, unsupervised learning techniques
or regression, for example, support vector machines, linear are applicable to optimal modulation, channel-aware feature-
regression, logistic regression, linear discriminant analysis, extraction, etc. In addition, unsupervised learning technologies
naive Bayes, k-nearest neighbors, and decision tree, etc. Su- can be used for anomaly detection [115], localization [116],
pervised learning techniques can be used in both physical traffic control [117], and parameter optimization [118], etc. in
layer and network layer in 6G. In physical layer, we can network layer.
utilize supervised learning for channel states estimation where Deep Reinforcement Learning. Deep reinforcement learn-
transmitted signals known by the receiver can be used for train- ing (DRL) leverages Markov decision models to select the
ing [111,112]. Besides, supervised learning techniques can be next “action” based on the state transition models [70]. DRL
deployed for data-driven tasks such as channel decoding [113], technique is considered as one of the promising solutions
caching [114], and delay mitigation [18] and so on in the to maximize the long term cumulative reward by sequential
network layer. decision-making [18]. The agent observes the state of the
Unsupervised Learning. The unsupervised learning is environment, chooses an action from available actions, obtains
leveraged to find undefined patterns in the dataset without rewards, and then the state changes to the next state during

6
 TABLE III: Machine Learning Techniques Overview.

ML Types Description ML Structures References

Support vector machine (SVM),

K-nearest neighbors (KNN),
Multilayer Perceptron (MLP),
Bayesian Learning (BL),
Supervised learning refers to algorithms that Deep Neural Network (DNN), [79], [18], [80], [21],
Supervised Learning determine a predictive model using a Long Short Term Memory (LSTM), [81], [57], [82], [83],
set of labeled data. Linear regression (LR), [84], [69], [85], [86]
Support Vector Regression (SVR),
Gaussian Process (GP),
Regression,
Self-supervised learning
K-means, EM,
Unsupervised learning refers to algorithms that [18], [21], [79], [81],
Clustering,
Unsupervised Learning can automatically find structure in the data with no [87], [22], [88],
PCA,
labels and with minimal human supervision. [89], [61]
Generative Adversarial Network (GAN)
Q-learning,
Deep Q-learning,
Reinforcement learning refers to algorithms that
Advantage Actor Critic (A2C), [90], [91], [92], [93],
seek to intelligent agents to learn
Reinforcement Learning Asynchronous Actor-Critic Agents (A3C), [94], [95], [96], [97],
correct decisions by trial and error and
Dueling DQN, [98], [99], [100]
pursue a long term reward.
Double DQN,
Proximal Policy Optimization (PPO)
Federated learning is a privacy-preserving machine
[80], [101], [102], [103],
learning technique, which allows users to hold
Federated Learning Decentralized Learning [104], [105], [106], [107],
data locally and send trained models to the
[108], [109], [110]
central server for obtaining the global model.

the learning process [119]. In DRL, a neural network is used vacy [120]. The idea of FL is illustrated in Fig. (5), where
to estimate states instead of mapping every solution. DRL users’ devices train local models and then send trained local
is used to solve resource allocation problems in 6G [69,78]. models to the base station for aggregation. Since users’ data
As 6G wireless networks serve a wider variety of users in are still maintained in the devices, the privacy of their data
the future, the radio-resource will become extremely scarce. can be well preserved. As 6G heads towards a distributed
Hence, efficient solutions for radio-resource allocation are architecture, FL technologies can contribute to enabling the
urgent and challenging [69]. shift of AI moving from a centralized cloud-based model to the
Model-Driven Deep Learning. The model-driven approach decentralized devices based [9,74,121,122]. In addition, since
is to train an artificial neural network (ANN) with prior infor- the edge computing and edge devices are gaining popularity,
mation based on domain knowledge [71,72,75]. The model- AI computing tasks can be distributed from a central node to
driven approach is more suitable for most communication multiple decentralized edge nodes. Thus, FL is one of the es-
devices than the pure data-driven deep learning approach, sential machine learning methods to enable the deployment of
because it does not require tremendous computing resources accurately generalized models across multiple devices [123].
and considerable time to train what the data-driven method Explainable Artificial Intelligence. There will be a large
needs [72]. The approach to apply model-driven deep learning scale of applications such as autonomous driving and remote
proposed by Zappone et al. [72] includes two steps: first, surgery in 6G era. Since these applications are closely related
we can use theoretical models derived from wireless com- to humans’ life, a mistake may incur miserable disasters.
munication problems as prior expert information; second, we Therefore, it is very necessary to make AI explainable for
can subsequently tune ANN with small sets of live data even building trust between humans and machines. Currently, most
though initial theoretical models are inaccurate. AI approaches in PYH and MAC layers of 5G wireless
Federated Learning. Federated Learning (FL) aims to networks are inexplicable [124]. Some AI applications like
train a machine learning model with training data remaining autonomous driving and remote surgery, are considered to be
distributed at clients in order to protect data owners’ pri- widely used in 6G, which requires explainability to enable

7
 Surface with reflecting elements and receive electromagnetic fields. Buildings, streets, and walls
are expected to be electronically active after decorating with
LISs [125]. As radio frequency circuits and signal processing
units are embedded in the surface, the entire surface of the LIS
can be used to transmit and receive communication signals.
The LIS has the following main favorable features [125]:
(i) Generate perfect LoS indoor and outdoor propagation
environments. (ii) They put little restriction on the spread of
antenna elements. Hence, mutual coupling effects and antenna
correlations can be easily avoided, such that sub-arrays are
large and the channel is well-conditioned for propagation.
Thus, the LIS can be realized via THz Ultra-Massive MIMO
RF transmitter
Receiving user (UM-MIMO). The LIS is very useful for applications with
low-latency such as wireless virtual/augmented reality and
(a) Intelligent Reflecting Surface. vehicular communications.
RF signal generator IRS. The IRS is considered as a promising candidate to
improve the signal quality at the receiver by modifying the
phase of incident waves [126]–[133]. IRSs are made of electro-
magnetic (EM) material that are electronically controlled with
integrated low-cost passive reflecting elements, so that they
can contribute to forming the smart radio environment [134].
The highly probabilistic wireless channel is tuned into a de-
terministic space by using the software-controlled propagation
of the EM waves in the smart radio environment realized
by the IRS. The IRS helps to enhance the communication
Surface with radiating elements
between a source and a destination by reflecting the incident
Receiving user wave [134]–[136]. By adjusting the reflection coefficients, the
IRS enables the reflected signals being coherently added to
the receiver without adding additional noise [136]. Besides,
the IRS can modify the signal phase and increase signal
(b) Large Intelligent Surface. power [125]. In particular, by utilizing local tuning, graphene-
based plasmonic reconfigurable metasurfaces can obtain some
Fig. 6: Compare IRS with LIS.
benefits, including the beam focusing, beam steering, and
control on wave vorticity [137]. Unlike the LIS, the IRS uses
trust. AI decisions should be explainable and understood by a passive array architecture for the reflecting purpose [138].
human experts to be considered as reliable. Existing methods, In the following, we list some features of the IRS [134,138] :
including visualization with case studies, hypothesis testing, • They comprise low-cost passive elements which are con-
and didactic statements, can improve the explainability of deep trolled by the software programming.
learning. • They do not require a specific energy source to support
the transmission.
IV. I NTELLIGENT S URFACES • They do not need any backhaul connections to exchange
traffics.
• The IRS is a configurable surface, so that points on its
Intelligent surfaces are new and revolutionary technologies surface can shape the wave impinging upon it.
for significantly improving the performance of the wireless • They are fabricated with the low profile, lightweight,
communication networks. Currently, two types of intelligent and conformal geometry such that they can be easily
surfaces attract researchers’ attention, including the large intel- deployed.
ligent surface (LIS) and intelligent reflecting surface (IRS). As • They work in the FD mode.
shown in Fig. (6), the LIS is useful for constructing an intel- • No self-interference.
ligent and active environment with integrated electronics and • The noise level does not increase.
an external signal generator [125,126]; the IRS utilizes many
low-cost passive reflecting elements to reflect beamforming. Other Similar Technologies. Besides LIS and IRS, some
In the following, we present an overview of the LIS, IRS, and other similar technologies have appeared in recent years.
other similar technologies. • Large intelligent metasurface (LIM): LIM uses a special
LIS. The concept of deploying antenna arrays as the metallic material called meta-atom to form its surface,
LIS in massive MIMO systems is originally proposed by which is more flexible in manipulating electromagnetic
Hu et al. [127]. The LIS is electromagnetically active in the waves [139].
physical environment, where each part of an LIS can send • Smart reflect arrays: smart reflect arrays put more empha-

8
 sis upon the reflection function other than the transmis- considerably fast. The reflect coefficient has two parts: phase
sion, reception and waveguiding functions [140]–[142]. shift θi and amplitude attenuation αi , i.e.,
• Software-defined surface (SDS) / software-defined meta-
φi = αi e1jθi (i = 1, ..., NI ) (4)
surface (SDM): the SDS/SDM introduces the thought
of software-defined radio into smart surfaces and by with 1j is the imaginary unit.
controlling the smart surface units through programmable According to whether the models for the phase shift and
fashion [143,144]. amplitude attenuation are continuous or not, the IRS can be
As the above mentioned technologies have almost similar classified into continuous mode and discrete mode. For the
functions with LIS or IRS, and in our survey, we put stress phase shift, φi can take any value in the range of [0, 2π)
on the investigation of these intelligent surfaces’ quality of in the continuous mode, however, φi can only take a finite
service (QoS) enhancement, therefore, we use the name IRS number of values in [0, 2π) in the discrete mode. Similarly
in our following manuscript for general purpose. for the amplitude attenuation, when there is no attenuation for
The structure of IRS is depicted in Fig. (7). The base the coefficients, αi = 1, and when the amplitude attenuation
station (BS) equipped with NT antennas is communicating is in the continuous mode, αi can take any value in (0, 1] or a
with one or multiple single-antenna cell-edge user equipment finite number of values between 0 and 1 in the discrete mode.
(UE). Since the channel fading of the BS-UE direct link is too However, when the phase shift or the amplitude attenuation are
large to guarantee the QoS, an IRS of NI units is utilized to in the discrete mode, the optimization problem of maximizing
facilitate the communication. the capacity or minimizing the transmitting power as discussed
later are non-convex. The generally-used method is to solve the
IRS Controller problem considering the phase shift and amplitude attenuation
BS-IRS-UE link in continuous mode, and choose the discrete values that are
BS-UE direct link close to the solved continuous solutions. The problem with no
amplitude attenuation also makes the constraint non-convex,
and in such a case the optimization can be done first by
IRS neglecting this non-convex constraint and then by normalizing
the phase shifts to fulfill the constraint of unit modulus [145]–
[147].
Nadeem et al. [150] propose a channel estimation protocol
based on the MMSE principle. The protocol first estimates the
channel coefficients between BS and UEs by turning off the
Base Station (BS) Cell-edge UEs IRS, and then estimates the BS-IRS-UE link by turning on the
Fig. 7: Structure of IRS system. IRS unit in turn. Finally, MMSE approach is utilized to obtain
a comprehensive estimation of all channels. This protocol is
complicated and time consuming due to the fact that each IRS
Suppose there are (is) NUE UE(s) in the IRS-aided wireless
should be turned on once a time while the others turned off,
communication systems (NUE ≥ 1). The received signal for
especially when the number of IRS elements is large. He et
the nth UE is given by
al. [151] propose a three-state channel estimation algorithm
yI,n = (hI,n ΦHI,B + hTB,n )x + wn , n = 1, 2, ..., NUE , (1) for MIMO systems. The algorithm first obtains the channel
matrices of the BS-IRS and IRS-UE links via matrix factoriza-
where hI,n ∈ C1×NI and hB,n ∈ C1×NT are the channel
tion. Then, by exploiting the IRS state matrix, the ambiguity
fading vectors between the IRS and the nth UE and the BS and
of the solutions to matrix factorization is eliminated. In the
the nth UE, respectively, HI,B ∈ CNI ×NT is the channel fading
final stage, the properties of channel matrices are utilized to
matrix between the BS and the IRS, wn is the additive white
recover the missing entries. The proposed estimation algorithm
Gaussian noise at the nth UE, x ∈ CNT ×1 is the transmitting
is also time consuming because the IRS units should also be
signal vector of the BS, and Φ ∈ CNI ×NI is a diagonal
turned on in turns to eliminate the ambiguity in the second
matrix whose diagonal elements are the corresponding reflect
stage. To address such a problem, Zheng et al. [152] propose
coefficients of the IRS, i.e.,
a novel method to estimate the channel state information
Φ = diag(φ1 , φ2 , ..., φNI ) (2) for IRS-enhanced OFDM systems. The authors first design
a reflection pattern of the IRS for channel estimation by
with φi (i = 1, ..., N + I) is the reflect coefficient of the IRS’s
exploiting pilot signals, and then perform a joint channel
ith element. By defining ϕ = [φ1 , ..., φNI ]T , the received
estimation and reflection optimization based on the strongest
signal can be rewritten as
signal path resolved in the first stage. This novel method avoids
yI,n = (ϕdiag(hI,n )HI,B + hTB,n )x + wn , n = 1, 2, ..., NUE . the requirement of large numbers of pilot signals and operation
(3) for each IRS units, therefore, reducing the complexity of the
Compared with Eq. (1), Eq. (3) is more convenient to be receiver for a large scale.
processed. Besides the fundamental application as described above,
However, hTB,n is often omitted since the distance between there exist some other IRS-assisted communication systems
the BS and UE is long and the channel fading degrades or IRS technologies in the available references. For instance,

9
 TABLE IV: Developments of IRS Technology.

Developments Key IRS Technologies

- Maximization of transmission rate with power constraint

IRS-aided multi-user systems [140]–[142]
- Minimization of transmitting power with received SINR constraint

IRS-aided OFDM systems [148] Maximization of capacity with power constraint

IRS systems with hardware impairment [149] Capacity performance is investigated

IRS-aided SM systems [143] BER performance of IRS-SM/IRS-SSK is investigated

- MMSE principle [150]

Channel estimation - Matrix factorization, ambiguity elimination, matrix completion [151]

- Joint optimization of channel estimation and IRS reflection [152]

HMIMOS systems [153] Constructive interference is utilized to promote performance

- Lorentzian resonance response φi = (1j + e1jηi )/2 [154]

Phase-dependent amplitude model for IRS schemes
- Lumped circuit model αi = (1 − αmin )[(1 + sin(θi − η))/2]k + αmin [155,156]

Yang et al. [148] discuss the application of the IRS in OFDM model by exploiting the Lorentzian resonance response [158],
systems where multiple paths exist in the wireless com- where the coefficient is expressed as φi = (1j + e1jηi )/2
munication environment. The authors establish an objective with ηi ∈ [0, 2π). In addition, Abeywickrama et al. consider
of maximizing the capacity with the maximum transmitting a phase-dependent amplitude model in [159,160] according
power limited, where the coefficients of the IRS and the power to the lumped circuit model [155,156]. The amplitude of
allocation for each subcarrier are unknown parameters. By coefficient is represented as αi = (1 − αmin )[(1 + sin(θi −
exploiting the idea of alternating optimization, the optimal η))/2]k + αmin , where constant parameters αmin ∈ [0, 1],
power allocation and the IRS coefficients can be derived. η ≥ 0, and k ≥ 0 are set depending on the specific
Hu et al. [149] investigate the capacity performance with circuit implementation. The recent development of the IRS
hardware impairments exist in the IRS. The authors also give technology in available references are listed in Table IV.
a conclusion that by splitting the IRS into several subgroups Challenges of IRS Technologies. Although the advantages
with each subgroup consists of a number of the IRS’ units, the of the IRS is attractive, there still exists some challenges in
performance degradation can be mitigated to a certain degree. the application of the IRS technology [161] :
Basar [143] investigate the bit-error rate (BER) performance • Seeking for the optimal coefficients of the IRS and the
of the IRS-aided SM systems. In the manuscript, optimal beamforming vector is painstaking, it consumes a lot of
(exhaustive search) and suboptimal (greedy) detectors for the time and hardware resources in solving the optimization
IRS-assisted SM or SSK schemes are formulated in detail, problems.
and their theoretical average bit error probabilities are derived • Although available references refer to the methods for
as well, to give a benchmark of the IRS-IM/SSK wireless channel estimation, the derivation of CSI is still a com-
communication systems. plex work in engineering, especially when the number of
Furthermore, Huang et al. [153] propose the scheme where IRS element is large.
the IRS is combined together with MIMOS to form the
HMIMOS systems. Because the interference is regarded as
useful resources for developing holographic communication V. I NDEX M ODULATION
systems [73], the multi-user interference can be decomposed Index Modulation (IM) is a kind of modulation scheme
into constructive and destructive parts using simple geometric that sends extra information bits through resource entities’
relations [157]. Constructive part is considered as beneficial index. On the one hand, high capacity can be achieved due to
communication resources, which can be used to improve QoS the spare information bits’ transmission. On the other hand,
of 6G communication systems. To be more specific, active because the sending of these spare bits does not consume
HMIMOS using the IRS are equipped with RF circuits and any power and spectrum resources, IM has both high spectral
signal processing units, whereas passive HMIMOS only use efficiency and high energy efficiency simultaneously compared
the IRS for reflecting signals. with its non-IM-aided counterpart. The structure of the IM
The above models all consider that the phase and amplitude system is depicted in Fig. (8). As shown in the figure, at
of the IRS units are independent to each other. However, in the transmitter, the information bits are divided into two parts
some cases, the amplitudes of IRS units are relevant to their after the operation of serial to parallel conversion, one part
own phases. Di et al. [154] construct a reflecting coefficient for the classical amplitude phase modulation (APM) such as

10
phase-shift keying (PSK) or quadrature amplitude modulation designing the constellation used in IM aided fast-OFDM
(QAM), etc., and the other part for the activation of the systems, trade-offs between spectral efficiency and energy
index resources exploited for transmitting the modulated APM efficiency is obtained , thus being able to serve huge numbers
signals. After the up-conversion, passband modulated signals of devices efficiently in 5G beyond networks. Althunibat et
are transmitted through the activated resource entities. At the al. [170] propose a quadrature index modulation multiple
receiver, down-conversion is first performed to convert the access (QIMMA), which allows each user to activate two of
received signals to baseband. An index demodulator is carried the orthogonal resources when communicating with the base
out to detect the information bits used for APM and index station. Such an access scheme can serve larger numbers of
activation of resource entities. Finally, a parallel to serial users compared with its non quadrature counterpart, which is
conversion is conducted to recover the original information beneficial for future network where the number of users grows
sequence. According to the categories of the resource entities exponentially.
used for index selection, IM can be classified into spatial- Security is another important requirement of 6G network.
domain IM (SD-IM), frequency-domain IM (FD-IM), time- Yang et al. [171] encrypt the SM signals through random
domain IM (TD-IM), and channel-domain IM (CD-IM). As mapping patterns to avoid being eavesdropped for SM assisted
stated in Table I, IM is one of the main multiplexing architec- multi-hop wiretap ad-hoc networks. By activating the SM map-
tures of 6G network [19], and a lot of IM based technologies ping patterns through random CQI patterns over the legitimate
have been studied to satisfy requirements of 6G network. link, the security performance of the conceived scheme is
By exploiting SM technology in VLC systems, the overall enhanced a lot, and thus satisfying the security requirement
spectral efficiency can be improved to satisfy the requirements of 6G network.
of high reliability, high power efficiency and high spectral The error probability performance of IM relevant schemes
efficiency for 6G network. Nahhal et al. [162] introduce a is also studied by a lot of references to meet the requirement of
flexible generalized spatial modulation (FGSM) scheme for high reliability. Shi et al. [172] investigate an OFDM scheme
VLC systems. The proposed scheme changes the modulation with all index modulation (OFDM-AIM). By replacing the
sizes and the number of activated transmit LEDs adaptively PAM constellation modulator with subblock modulator of an
to improve the average SER under a predefined spectral OFDM symbol, the diversity order can be enhanced. There-
efficiency threshold of VLC systems. Gao et al. [163] propose fore, the BER performance is improved significantly. They also
a light emitting dionode (LED) grouping SM scheme, where consider the legitimate subblock realizations in OFDM-AIM
the channel correlation effect can be decreased and thus as the chromosomes of genetic algorithm [173]. By utilizing
improve the SER performance significantly. Kumar et al. [164] the ABEP as the population fitness, subblock realizations with
optimize the collaborative constellation of GSM to enhance good BER performance can be obtained. Yu et al. [174]
the power efficiency for VLC systems. And in [165], by investigate a phase rotation based precoding SM scheme.
using different modulation sizes according to which LED is By optimizing the phase rotation-based precoding matrix, the
activated, adaptive SM is investigated to improve the spectral minimum Euclidean distance of the SM signal constellation
efficiency. is improved, and thus enhancing the BER performance to a
In addition, Huang et al. [166] combine IM with frequency large scale. In addition, Zhang et al. [175] introduce a soft-
diverse array (FDA) and propose a IM-FDA scheme. The input soft-output detector based on expectation propagation
proposed scheme can transmit information and sense target framework for SM schemes. By calculating the LLRs of the
locations simultaneously. Such a feature makes IM-FDA an transmitted bits iteratively, the proposed detector can conduct
attractive technique for the future sensor network which re- trade-offs between BER performance and detector complexity.
quires the knowledge of each sensor’s location. And Nusenu et However, all the schemes metioned above need the knowl-
al. [167] propose a space-frequency increment IM by utilizing edge of CSI when detecting the transmitted information bits,
the indices of the activated transmit FDA antennas and the In [176], a multi-set STSK scheme transmitted in mmWave
corresponding frequency increments. A low-complexity de- channel is proposed, and a neural network is trained to jointly
coding algorithm is investigated and performance is evaluated, detect the transmitted source information. For the proposed
in order to provide theoretical analysis for future 5G beyond detection algorithm, the channel estimation can be eliminated
network. and outperforms the classical ML algorithm in the scenario
Purwita et al. [168] introduce a signle carrier generalized with channel impairments. And Satyanarayana et al. [177]
time slot IM for single-carrier frequency domain equalization also propose a deep learning based soft detector. For the
(SC-FDE) systems. By changing the number of activated time proposed scheme, a neural network is trained to obtain the soft
slots, the performance limitations due to the multipath optical values for transmitted signals without CSI information, which
wireless channel can be mitigated, leading to more than 3dB is promising for 6G wireless network where perfect CSI is
gain than its non-IM-aided SC-FDE technique. The proposed difficult to evaluate. The summaries of IM-based technologies
scheme is more attractable for future indoor optical wireless for future network are listed in Table V.
communications with limited dynamic range of LEDs. Challenges of IM Techniques. Although IM technology
The ability of serving huge numbers of UEs is also a key has the advantage of high data rate and low energy consump-
requirement for future network. To improve such an ability, tion, it still meets some challenges as follows [178,179] :
Nguyen et al. [169] introduce a combined hybrid OFDM- • For IM system, the non-activated resource entities are
based modulation scheme for narrowband IoT. By properly kept empty in transmission, and therefore the utilization

11
 Index Modulator Up Converter

Index Selector

Source Serial-to-Parallel Time slots,

Converter subcarriers,
antennas,
channel states

Modulator

Fading
Channel

Index Detector Down Converter

Index Detector

Destination Time slots,

Parallel-to-Serial
Converter subcarriers,
antennas,
channel states

Demodulator

Fig. 8: The structure of IM-aided systems.

efficiency is unsatisfactory. et al. [185] first propose the concept of SWIPT, which de-
• The complexity of detection algorithm is very high. liver information and energy concurrently from a theoretical
• Since the APM symbols are conveyed only through the perspective. The performance is evaluated by assessing the
selected resources, once the detection of the index is reception reliability and information transmission rate.
incorrect, the demodulation is almost destructive in the Fig. (9) depicts the architecture of SWIPT schemes. The
decoupled detection algorithm. generally-used structures for resource allocation are time
• In MBM system, the number of mirrors is large to switching (TS) structure, power splitting (PS) structure, and
achieve high data rate, therefore, huge training sequence antenna switching (AS) structure. In TS structure, a frame
should be required for acquiring the channel states, which consisting of several time slots are classified into information
consumes a lot of resource in transmission. transfer slots and power transfer slots, and a switcher is used
• Since hybrid IM technologies outperform its signal do- at the receiver to decide which state is at work [186,187].
main counterpart in error performance and transmission The receiver splits the received signal into two streams with
rate, current references are not enough in investigation power ratio ρ(t) : 1 − ρ(t), where the two parts are used for
into hybrid IM technologies. harvesting energy and decoding the information, respectively.
The architecture of SWIPT system is depicted in Fig. (9).
VI. S IMULTANEOUS W IRELESS I NFORMATION AND Assume the received block consists of Nf time slots, therefore
P OWER T RANSFER (SWIPT) the splitting vector can be defined as ρ = [ρ1 , ..., ρNf ]T . For
the TS structure, in the first αNf time slots, the received signal
power are used for energy harvesting overall, and vice versa,
In the future 6G network, various mobile wireless devices all the signal power are exploited for information decoding for
will be accessed to the network. The high transmission rate the remaining (1 − α)Nf time slots. Therefore, the element of
leads to an increase in power consumption, and thus, requires the splitting vector for TS structure can be expressed as
the long lifetime of battery-powered devices. SWIPT is such a
novel technology promisingly used in the 6G network where 
1, k = 1, ..., αNf
energy harvesting through RF signals can be made to fulfill ρk = (5)
0, k = αNf + 1, ..., Nf .
the requirement of high power efficiency [180]–[183]. The idea
of energy harvesting can date back to 1969, when Brown et In PS structure, a power splitting factor is utilized to decide
al. [184] conduct a series of experiments and demonstrate that how much signal flows to the decoder circuit for signal
a helicopter at a height of 50 feet operating on 2.45GHz can processing or the battery circuit for energy harvesting. And
provide a direct current power supply of about 270W. The subsequently, the two power streams can function separately
energy is harvested either from the ambient wireless signals but simultaneously through the overall frame duration [188].
or from a specific controlled power source. In 2008, Varshney At the receiver, the ratio of the power used for harvesting

12
 TABLE V: Summary of IM-based Technologies for future network.

IM-based technologies Features

- Change the modulation sizes and the number of activated transmit LEDs adaptively [162]
- LED grouping SM [163]
SM in VLC systems
- Optimize the collaborative constellation of GSM [164]
- Use different modulation sizes according to which LED is activated [165]
- Transmit information and sense target locations simultaneously [166]
IM combined FDA systems
- Utilize the indices of the activated transmit FDA antennas and the corresponding frequency increments [167]
IM assisted SC-FDE systems Change the number of activated time slots in SC-FDE systems [168]
- Design a constellation properly used in IM aided fast-OFDM systems [169]
Ability of serving huge numbers of UE
- Allow each user to activate two of the orthogonal resources [170]
Security requirement Encrypt the SM signals through random mapping patterns [171]
- Replace the PAM constellation modulator with subblock modulator [172]
- Consider the legitimate subblock realizations in OFDM-AIM as the chromosomes of genetic algorithm [173]
Error probability performance
- Optimize the phase rotation-based precoding matrix [174]
- Soft-input soft-output detector based on expectation propagation framework for SM schemes [175]
- Jointly detect the transmitted source information exploiting neural network [176]
Detector without CSI
- Obtain the soft values for transmitted signals exploiting neural network [177]

energy and decoding information can be set as a constant the capacity of the system and the harvested energy can
ρ, leading to the element of the splitting vector be ρk = be promoted considerably. A lot of references have discuss
ρ, (k = 1, ..., Nf ). In addition, the power splitting factor the energy efficiency of SWIPT-based technologies. Ojo et
can be designed to balance according to different system al. [190] investigate the energy efficiency performance of
requirements. a SWIPT-enabled cooperative relay network with interfering
In AS architecture, the RXs are divided into two subsets for exists in the systems. Two different scenarios where the relay
energy harvesting and information decoding separately [189]. can harvest only from the source and that harvest energy from
Compared to TS and PS structures, AS is easier to imple- both the source and the interference are considered. A sub-
ment and thus appealing for practical SWIPT deployment, optimal PS factor is derived and a new energy-saving power
however, the performance will degrade significantly in case splitting relaying protocol is investigated, which provides PS
of hardware impairments. In one frame duration, all split- based relaying protocol for the future relay networks. Fang et
ting factors remain unchanged for a given received antenna, al. [191] consider the SWIPT scheme with linear precoder
ρk = ρ, (k = 1, ..., Nf ) for energy harvesting RXs and MIMO system. By optimizing the precoder to maximize
ρk = 1 − ρ, (k = 1, ..., Nf ) for information decoding RXs. the harvested energy with the power constraint, high power
The overall performance can be increased by optimizing the efficiency is achieved, which is feasible for both separated
combination of the signal power and the AS ratio. and co-located receivers. Park et al. [192] study the effect
of high power amplifiers’ (HAPs’) nonlinearity on the multi-
tone SWIPT systems. Authors conclude that HAP has adverse



effects on both energy harvesting and information decoding.

    To address this problem, Jang et al. [193] give a frequency-

splitting SWIPT, which separates the power and transmitted
signals in frequency domain. The proposed scheme can reduce
   the harmonic distortion of the high power amplifier, thus
 


decreasing the effects amplifiers’ non-linearity. Frequency-
splitting SWIPT outperforms its PS counterpart in terms of
Fig. 9: Architecture of SWIPT system. the harvested power, and is more beneficial to the 6G network
where a lot of low-power wearable devices are accessed to the
As the goal of SWIPT scheme is to decide the splitting network. Zargari et al. [194] study an IRS-assisted SWIPT
factor to balance the tradeoff between transmission rate and technology for MISO systems. By introducing an energy
energy harvesting, the architecture of SWIPT is depicted in efficiency indicator to trade off between transmission rate and
Fig. (9). In most application scenarios, the research works harvested energy, a joint optimization is performed to obtain
focus on different resource allocation schemes considering the beamforming vector, the PS ratio of each user, and the
data rate, harvested energy power, transmit power, and re- phase shifts of IRS. The optimization problem is divided into
ceived SNR. By performing a joint optimization of resource two subproblems and a suboptimal solution is achieved for
allocation with power control and resource scheduling, both

13
the system. The proposed scheme has the advantages of both communication network, UAV communication network, ter-
low power consumption and high BER performance, which is restrial communication network, and maritime communication
attractive for the 6G network. network to build a space–air–ground-sea integrated network
Sun et al. [195] also consider an IRS-assisted SWIPT (SAGSIN) to support the global coverage and ubiquitous
technology, where the security issue is emphasized since the connection as shown in Fig. (10). These networks may work
broadcast nature makes the system be vulnerable to eavesdrop- independently or collaboratively.
ping. The secrecy rate is optimized by jointly designing the In the previous generations of communication systems,
beamforming vector at the BS and reflective beamforming at satellite-air-ground integrated network (SAGIN) has been
the IRS, subject to the transmit power constraint and energy hotly discussed by the academic. Specifically, SAGIN is
harvest constraints. The proposed scheme is useful for future essential to support applications like IoT, big data, and cloud
network because both low power consumption and high secu- computing [200]. Radhakrishnan et al. [201] summarize the
rity are guaranteed, which can satisfy the requirements of 6G inter-satellite communication from viewpoints of the physical
network. Additionally, Ma et al. [196] investigate the security and network layer. Besides, Niephaus et al. [202] analyze
issues of SWIPT networks. A SWIPT system enabled by full- the QoS provisioning in converged satellite and terrestrial
duplex jamming is provided to improve the security, where networks. They emphasize the technical challenges related
the full-duplex energy-limited receiver is used to generate to the convergence of satellite and terrestrial networks in
jamming signals. By optimizing the PS factor, the security detail, where the satellite networks act as a complement to the
throughput is maximized. Zhu et al. [197] employ artificial existing terrestrial infrastructures. Hamdi et al. [203] study
noise generation to improve the security of transmission. The a satellite-based hybrid sensor network in the perspective
minimum harvested energy among users is maximized with of detection and tracking in a mobile scenario. However,
the constraints of secrecy rate and transmit power. Thakur et they only consider two networks, either space-ground or air-
al. [198] investigate the security performance of SWIPT ground integrated networks. Different from the space-air-
scheme for cognitive radio network. The intercept and secrecy ground integrated network in 5G, SAGSIN in 6G also provides
outage probability is derived with imperfect CSI and the coverage for the underwater and deep-sea communications. In
influence of power splitting factor. The secrecy performance is the following, we introduce each component of the SAGSIN
also studied in detail. All the SWIPT based scheme mentioned in detail.
above consider security as one of the main problems in system
design, and is beneficial for 6G network where high security A. Satellite Communication Network
is required to protect the individual data of each UE. The
5G communication network focuses on the terrestrial cov-
summarys of SWIPT relavant schemes for future network are
erage, which leaves some areas uncovered. Satellite communi-
listd in Table VI.
cation in 6G is envisioned to integrate with terrestrial commu-
Challenges of SWIPT. The SWIPT-based systems
nication to provide full coverage and high throughput for the
metioned above exemplify the general usage of SWIPT tech-
regions where terrestrial wireless communication cannot reach,
nology in wireless communication networks. It can degrade the
such as rural areas. Satellite communication can be catego-
power consumption and takes advantage at scenarios where it
rized into Geosynchronous Earth Orbit (GEO) and non-GEO
is difficult to charge the battery such as volcano and marine
based on the satellites’ altitude. Chini et al. [204] conclude
areas. However, there still exist some challenges of SWIPT
that a GEO satellite is at an altitude of about 35,800 km.
technology [183,199]:
GEO satellite communication uses high frequencies, which
• Existing studies mainly focus on low-power relays and aggravates the path loss. Non-GEO satellites have Low Earth
sensors due to the fact that the level of the harvested Orbit (LEO) and Medium Earth Orbit (MEO). LEO is at an
energy is too low to satisfy the requirements of large altitude of 500 to 2000 km, and MEO is at an altitude of
wireless devices. 8,000 to 12,000 km. Non-GEO satellites have lower end-to-end
• In 6G, the dense deployment of wireless devices leads to delay compared to GEO satellites. Different orbits decide that
complex electromagnetic interference, which is desirable satellites are suitable for different scenarios. GEO is stationary
to power transfer while adverse to information decoding. to the Earth, while LEO is moving with the Earth. Therefore,
• Intense RF radiation has adverse impact on human health. GEO can provide a designated region with continuous service.
However, to achieve high QoS of SWIPT-based system, However, due to the long distance, it suffers a relatively high
denser sensors or relays should be deployed for energy signal delay. Compared with GEO, LEO has a shorter delay
harvesting, which causes more intense RF exposure. because its orbit is much lower than GEO, which is more
convenient for communicating with ground terminals, such as
VII. S PACE -A IR -G ROUND -S EA I NTEGRATED N ETWORK GPS communication and satellite phones.

In the 5G network, users only have access to a single B. UAV Communication Network
wireless communication system, for example, the terrestrial A unmanned aerial vehicle (UAV) is an unmanned aircraft,
wireless communication system, which is experiencing ex- which means it works automatically. Each UAV can be viewed
plosive growth in both the number of users and services as a node. Different UAVs construct a UAV network. Since
available. To overcome 5G’s bottleneck, 6G integrates satellite the UAV communication network is distributed and federated

14
 TABLE VI: Summary of SWIPT Technologies for future network.

Developments Key SWIPT Technologies

- Energy harvesting and information perform decoding at different time slots [186,187]

SWIPT - A power splitting factor is used for energy harvesting and information decoding [188]

- RXs are divided into two subsets for energy harvesting and information decoding [189]

- Energy-saving power splitting relaying protocol is investigated [190]

- Optimize the precoder to maximize the harvested energy [191]

SWIPT with efficiency
- Separate the power and transmitted signals in frequency domain [193]

- An energy efficiency indicator is used to trade off between transmission rate and harvested energy [194]

- The secrecy rate is optimized by jointly designing the beamforming vector and the reflective beamforming [195]

- A SWIPT system enabled by full-duplex jamming is provided to improve the security [196]
SWIPT with security
- Employ artificial noise generation to improve the security of transmission [197]

- The intercept and secrecy outage probability is derived with imperfect CSI [198]

tion network, it has shipped and involves the ashore sta-

tions and satellites. There are self-organized ad hoc networks
which are consisted of vessels. This network’s function is
to accelerate the communication between mobile stations
and aid navigation, and switch the route [207]. Maritime
communication also has several challenges to be overcome.
Xia et al. [208] conclude eight existing challenges. They
are ubiquitous connectivity & service continuity, capacity &
scalability, simplicity & reliability, device heterogeneity, traffic
non-uniformity, service-centricity, interoperability, and radio
spectrum internationality.
However, SAGSIN is still facing many challenges as listed
in Table VII.

VIII. OTHER P OTENTIAL T ECHNOLOGIES

We have presented some key technologies that will be
Fig. 10: Space-Air-Ground-Sea Integrated Network. enabled by 6G in detailed in the previous section. To meet
the KPIs of 6G, we introduce more advanced technologies
that may contribute to the paradigm shift. In this section,
learning is an emerging distributed learning framework, some we briefly introduce some potential technologies in 6G, in-
studies leverage federated learning to assist UAV communi- cluding terahertz communication, symbiotic radio, device-to-
cation [205,206]. In [205], Lim et al. propose a collaborative device communication, free-duplex, cell-free massive MIMO,
learning scheme based on federated learning and involve UAVs blockchain-based network, and visible light communication.
for applications in the IoV paradigm. In [206], Zeng et al.
apply federated learning algorithms to a UAV swarm and also
present a joint power allocation and scheduling design, which A. Terahertz Communications
increases the convergence rate of federated learning.
The terahertz (THz) frequency band, which ranges from 0.1
to 10 THz, is the unexplored span of radio spectrum [31,210].
C. Maritime Communication Network Besides, THz communications provide new communication
It is envisioned that deep sea communication will be paradigms with the ultra-high bandwidth and ultra-low la-
supported by the robust underwater link in 6G [207]. That tency [31]. It is envisioned the data rate should be as high
is, the underwater links will support communication among as Tbps to satisfy 6G applications’ requirements of high
ships, submarines, and sensors. In a maritime communica- throughput and low latency [210].

15
 TABLE VII: A Summary of Challenges in SAGSIN [207]–[209].
Networks Challenges

- The design of the physical layer transmission to have accurate modelling of the satellite channels;

- The application of massive MIMO;

Satellite Communication Network
- Resource allocation in cooperative satellite transmission;

- The physical layer transmission and media access control (MAC) protocols;

- UAV deployment is too flexible to reach efficient result;

- How to keep users’ sessions if the UAV is out of service;

UAV Communication Network - Designing low power consumption is tricky;

- Deficiencies like node mobility, network partitioning, intermittent links are remained to be ameliorated;

- Optimization of throughput, delay, etc.

- The services over the sea are limited so that ubiquitous connectivity and service continuity are not realized.

- The distribution of maritime traffic is sparse on high seas. However, it is more dense close to the shore.

- The maritime communication system should adapt to various applications.

- The maritime communication network should be able to support different types of devices.
Maritime Communication Network
- Maritime communication network should be deployed with low costs and high security.

- The maritime communication system should be extendible for the growing capacity.

- Offer services for various information systems.

- Standards and regulations of international frequency band are not addressed.

Traditionally, the THz frequency band limits the widespread quadratically with the operating frequency [213]. This feature
use of THz. THz transceiver design is regarded as the most limits the use of THz to short-distance transmission such
critical factor in facilitating THz communications [210]. Re- as indoor communications [214]. Meanwhile, THz band can
cent technology advancements in THz transceivers, such as satisfy the requirement of ultra-high data rate; therefore, ultra-
photonics-based devices and electronics-based devices, over- broadband applications, for example, virtual reality (VR) and
come the THz gap, and enable some potential use cases wireless personal area networks can also exploit THz band
in 6G [31]. The electronic technologies such as silicon- to transmit signals [213]. THz technique can also be used in
germanium BiCMOS, III-V semiconductor, and standard sil- secure wireless communications. Since THz signals possess
icon CMOS related technologies (III and V represents the a narrow beam, it’s very hard to wiretap the information
old numbering of the periodic system groups), have been for the eavesdropper when locating outside the transmitter
vastly advanced, such that mixers and amplifiers can operate at beam [214].
around 1THz frequency [210,211]. The photonic technologies
are possiblely be used in the practical THz communication B. Cell-Free Massive MIMO
systems [210,211]. In addition, the combination of electronic-
based transmitter and photonics-based receiver is feasible.
Recent nanomaterials may help to develop novel devices that The conventional cellular network divides the network into
can used for THz communications [211]. In addition, a novel cells according to the topology theory. Each cell uses few ac-
approach to generate the THz frequency is discovered by cess points which are equipped with large co-located antenna
Chevalier et al. [212]. They build a compact device that can arrays [215]. Due to the serious path loss, the cell center
use the nitrous oxide or laughing gas to produce a THz laser. and the cell edge’s performances are different. At the cell
The frequency of the laser can be tuned over a wide range at edge, the spectral efficiency and energy efficiency are very
room temperature. low. In 6G, cell-free massive multiple-input-multiple-output
(CFmMM) is envisioned to be widely deployed. The CFmMM
Due to the high transmission RF, signals transmitted through uses a large number of cheap access points (APs), and each of
THz frequency band suffer from a high pass loss. Accord- them is equipped with few antennas instead of the traditional
ing to the Friis’ law, the pass loss in free space increases way of deploying few access points with large co-located

16
 Access
Point PTx PRx

STx SRx

User Equipment
(a) System model for SR. PTx uses
active radio to transmit messages
(a) Traditional cellular system. to PRx, and STx exploits backscat-
tering radio to transmit messages
to SRx riding over the RF signals
User Equipment
from PTx.

PTx PRx

CPU
Access Point

STx SRx

(b) System model for CR. PTx

transmits messages to PRx, and
STx transmits messages to SRx by
sharing the same radio spectrum
(b) Cell-free massive MIMO. with the primary system. STx re-
frains its transmit power to en-
Fig. 11: Compare the cellular system with the cell-free massive sure that the caused interference
MIMO. level to PRx is below a tolerable
threshold. Solid lines denote trans-
missions from active radios, while
antenna arrays [64]. The central processing units coordinate dash lines denote reflections from
backscattering radios.
these APs. Since the number of APs increases, each user’s
equipment (UE) may use multiple APs and antennas nearby Fig. 12: Compare SR and CR’s system models [32].
with reduced path loss. If the number of APs is large enough,
there is no cell concept because UEs can connect to several
neighbouring APs anywhere. Thus, there will be less path loss. network to increase the achievable uplink sum rate of the
Each UE can choose the cell with the strongest signals in CFmMM system. In addition, as the number of mobile phones
the traditional cellular network, but it will try to access all increases, telecommunication operators start exploring ways to
the neighboring APs to seek support in the cell-free network. leverage the wireless computation to solve the machine learn-
The proposed CFmMM systems ensure that UEs can achieve ing problems. To conquer the federated learning problems,
similar performance regardless of position and low-complexity UEs compute local updates with local training data to train
signal processing. Thus, the number of APs should be more local models, and then they send trained models to the central
than that of users. Therefore, a CFmMM system is able to server, while the central processor can be used to aggregate
provide uniform service throughout its coverage area instead the local models to obtain the global model. Contrary to the
of being partitioned into small cells. In order to illustrate their traditional deep learning approaches, users can maintain their
differences clearly, we compare the structures of conventional data locally to support training to prevent from the privacy
cellular system and the CFmMM system in Fig. (11). leakage [216].
As AI technologies will be widely utilized in 6G, for
example, deep learning and federated learning, novel tech-
niques and schemes based on CFmMM are proposed to solve C. Symbiotic Radio
machine learning problems [216,217]. Instead of traditional Symbiotic radio (SR) potentially enables the spectrum-
optimization techniques, Bashar et al. [217] have developed and energy-efficient massive access technology of 6G [26,32,
a new deep learning based algorithm which uses a neural 77,218,219]. It is a cognitive backscattering communication

17
system. SR reserves the advantages and solves the exist- To efficiently utilize the frequency resources, 6G is expected
ing challenges of cognitive radio (CR) [220] and ambient to implement the free duplex [33]–[35]. Free duplex means
backscatter communications (AmBC) [221]. Specifically, CR that there is no difference between the time division duplex
means that the radio network is smart enough to learn from (TDD) and frequency division duplex (FDD), so that the
the environment and history, so that it may allocate spectrum utilization of spectrum resources is free in terms of time,
resources dynamically to improve the efficiency of the radio frequency, and space. In the previous generations of commu-
spectrum utilization and avoid congestion. The secondary nication systems, the frequency is allocated based on TDD
transmission and primary transmission exist at the same time, or FDD. Specifically, before 5G, the frequency resources
which may result in interfering with each other. AmBC enable are allocated statically according to TDD or FDD. Then,
that communications among smart devices using the ambient transmission resources are allocated dynamically in 5G, which
radio frequency instead of active RF transmission [221,222]. In is called flexible duplex. Fig. (13) shows the evolution of the
the wireless communication systems with low power, AmBC duplex technologies. Free duplex is built on the full duplex
is very effective in enhancing the energy efficiency. However, technology. In the following, we present a detailed explanation
the backscatter transmission may have direct-link interference on the development of the duplex technology.
due to the nature of the spectrum sharing, which may affect
the performance of the transmission [223].
SR can enhance the spectrum-, energy-, and cost-efficiency
for wireless networks. Specifically, SR has two spectrum Flexible
Fixed Duplex Free Duplex
Duplex
sharing systems, i.e., the primary system and the secondary
system. Different from the interfering spectrum sharing in
CR, SR achieves mutually beneficial spectrum sharing. By 1G .... 4G 5G 6G
leveraging the joint decoding, SR obtains highly reliable
backscattering communications, which makes up the disad- Fig. 13: Evolution of Duplex Techniques.
vantage of the AmBC. The primary and secondary systems
work collaboratively at both the transmitter and receiver’s Duplex represents the ability of communication supported
sides to improve the spectrum and energy efficiency further. by two systems, including transmission and reception [224].
Fig. (12) illustrates system models of SR and CR, respectively. Based on the capability of systems’ data flow, the transmission
SR utilizes the backscattering radio technology to support the and reception can be simultaneous or asynchronous, which are
transmission from the transmitter (STx) to the receiver (SRx) called full-duplex (FD) and half-duplex (HD), respectively. To
in the secondary system, while the active RF chains are used be more specific, if systems are in FD mode, they can transit
by CR at both the primary receiver (PTx) and STx shown in and receive simultaneously; otherwise, they choose to transmit
Fig. (12b). Fig. (12a) illustrates that the backscattering link or receive in different time slots. That means that the system
from PTx-STx-SRx has to go through the double fading, so using HD has to spend half of time transmitting and the other
the strength of the direct link from PTx to SRx is much half of time receiving, which greatly reduces the throughput
stronger. The secondary system’s performance will be limited, and efficiency compared with leveraging FD. Fig. (14) and
and the primary system’s performance will be improved. In Fig. (15) illustrate differences between full-duplex and half-
addition, the backscattering radios can backscatter ambient duplex.
signals within a specific frequency band to capture the RF
signals that the secondary system uses.
As introduced in Section IV, IRS is also a promising tech-
nology to improve the efficiency of the energy in the wireless
communication system. Specifically, IRS has the possibility to
enhance the backscattering link’s signal, which contributes to
improving the performance of the transmission. IRS-assisted
SR is capable to adjust the reflecting elements intelligently,
so SR system can capture the direction of the backscattring
link to further improve the strength of the backscattering link. Fig. 14: Full-Duplex.
Also, the reflecting coefficients can be tuned to maximize the
primary system’s objectives. Besides, the number of reflecting In the 4G/5G wireless systems, transmission and reception
elements can be increased as well to enhance the strength of cannot be done at the same frequency or a same time interval,
the received SNR of the backscattering link. Thus, AmBC because HD does not support performing the transmission and
backscatters all the ambient signals; on the contrary, IRS- reception at the same time. Currently, 5G’s spectrum is limited
assisted SR is able to capture the intended PTx signal and to the time division duplex (TDD) or frequency division duplex
avoid the undesired interference at the SRx. (FDD), and most of spectral resources are TDD. TDD and
FDD are orthogonal transmission which may decrease the
efficiency of the utilization of spectrum. FD technique has
D. Free Duplex been enabled by 5G, but it has not been adopted by 3GPP
yet.

18
 such as anomalous reflection, frequency shifting, absorption,
wavefront shaping, nonreciprocity, and focusing, etc. These
functionalities can contribute to FD-enabled wireless transmis-
sion as follows :
• Anomalous Reflection.
• Frequency Shifting.
• Absorption.
• Wavefront Shaping.
• Nonreciprocity.
• Focusing.
Fig. 15: Half-Duplex.
Many applications can leverage FD, including : (1) Appli-
cation which are in low transmission power scenarios such
as device-to-device (D2D) (introduced in Section VIII-G) and
FD will be utilized completely in 6G wireless systems. FD
vehicle-to-vehicle (V2V) applications. By using FD, V2V
technologies have the possibility to double current efficiency
communication can be more reliable with low latency. (2)
in sharing spectrum and increasing the throughput of the
Sensing-based semi-persistent scheduling (SPS) achieves a rel-
networks and communication systems. Both FD and its related
atively better performance. (3) Scenarios equipped transceiver
techniques such as in-band full-duplex (IBFD) technologies
devices with unlimited complexity and cost, such as wireless
improve the efficiency of communication by allowing de-
relay and wireless Backhaul. Scenarios which use spatial
vices to transmit and receive a signal in the same frequency
freedom and narrow beams.
band [64]. Compared with HD, FD technology leverages self-
interference cancellation technology to increase the utility of
spectral resources, improve the throughput, and reduce the E. Blockchain-based Network
transmission delay between transceiver and receiver links. The
difference between TDD and FDD will be eliminated, and the Blockchain is a chain of blocks that constitute a distributed
true FD mode based on the communication requirements. The database. It is initially designed for cryptocurrencies (e.g., bit-
arrival of data packets follows Poisson distribution, so that the coin). However, nowadays, blockchain can do more than just
utility of resources fluctuates dynamically. in cryptocurrencies but run Turing-complete programs such as
By using FD technology, devices can transmit and receive smart contracts in a distributed way (e.g., Ethereum) [229].
signals at the same time. Yuan et al. [225] propose to im- Blockchain provides a secure and distributed database for
prove the receiver with self-interference cancellation to realize storing records of transactions, and each node includes the
self-interference cancellation, which uses 20% time-frequency previous block’s cryptographic hash, a time stamp, and trans-
spectrum resources of traditional solution. However, the hard action data [77,230]. Besides, blockchain-like mechanisms are
part is to eliminate the self-interference and transmit signal expected to provide the distributed authentication and control
generates over 100dB higher noise than the receiver noise by leveraging digital actions provided by smart contracts [51].
floor. Thus, the new scheduling algorithms and cost saving With these unique features, blockchain is envisioned to
circuits should be designed for 6G networks. Currently, three support numerous applications in 6G. In particular, by combin-
types of self-interference cancellation techniques are proposed, ing with federated learning, blockchain-based AI architectures
which include digital cancellation, analog cancellation, and are shifting AI processing to the edge [231]. Recently, a
passive suppression. blockchain radio access network (B-RAN) has been pro-
Except for self-cancellation techniques, Shen et al. [226] posed with the prototype [232,233]. Thus, blockchain can
and Xu et al. [227] propose to utilize intelligent reflecting help to form a secure and decentralized environment in 6G.
surface (IRS) introduced in Section IV to assist FD to improve Blockchain can provide a secure architecture for 6G wireless
the system performance and mitigate the interference. The networks [90]. Blockchain makes the consensus through min-
cost for IRS is much cheaper than relay and it does not ers instead of a central authority as shown in Fig. (16), and
consume energy by using its soft-controlled functionalities of it includes a wide range of applications in 6G. Blockchain
electromagnetic (EM) waves. IRS can assist FD application can help enhance authentication security by the approach
in two ways : (1) IRS acts as a bridge to facilitate FD of distributed ledger technologies [51]. Besides, blockchain
transmission. Different from RF relays, IRS supports co-time has applications in solving the problem of the low spectrum
and co-frequency FD transmission scenarios, including line- utilization and spectrum monopoly when deployed in spec-
of-sight (LOS) and non-LOS. (2) Employ IRS to generate trum sharing system [77]. By integrating wireless networks
reflect/transmission waves of FD signals for specific purposes, and blockchain, the central administrator can be eliminated,
including WPT, artificial noise, and cooperative jamming. improving network security and reducing costs. As D2D and
By integrating IRS and FD, the energy and cost saving IoE are gaining popularity in 6G, the cooperation among
FD-enabled IRS systems are proposed. Besides, the self- devices is getting more frequent, so that the distributed way
interference cancellation is unnecessary because IRS does not of resource management and spectrum control is required.
require any RF components, and then IRS is free of inter- Since blockchain guarantees transparency, it is easy to track
ference [228]. IRS enables electromagnetic (EM) functions the spectrum’s real-time utilization, which enables to allocate

19
the spectrum dynamically and efficiently. Thus, if 6G wireless The laser diode (LD)-phosphor conversion lighting tech-
network couples with blockchain when building, it will be nology can provide better performance in efficiency, bright-
simple to track the resource management and spectrum shar- ness, and larger illumination range compared with traditional
ing. IoT and D2D enable applications such as smart farming, lighting techniques [210]. Thus, it is considered as the most
healthcare, and machine-to-machine communication, etc. promising technology for 6G. The speed LD-based VLC
system is possible to reach 100Gbps, which meets the re-
quirements of ultra-high data density (uHDD) services in 6G.
Besides, the upcoming new light sources based on microLED
will overcome the limitation of low speed in the short-range
Miner network communication [51]. As massive parallelization of microLED
arrays, spatial multiplexing techniques, CMOS driver arrays,
Smart
and THz communications develop, VLC’s data rate is expected
Contract to reach Tbps in the short-range indoor scenario by the year
2027 [31,51]. Fig. (17) illustrates the indoor VLC scenario.
blockchain VLC is envisioned to be utilized in various applications in
6G. By integrating with SAGSIN, short-range network, and
cellular communication, VLC can be used to provide a better
coverage [14]. Also, traditional electromagnetic-wave signals
Resource Allocation and sharing
cannot achieve high data transmission speed using laser beams
in the free space and underwater. Still, VLC has a ultra-high
bandwidth and high data transmission speed [237]. Therefore,
Fig. 16: Blockchain-based network.
VLC is useful in cases where traditional RF communication
is less active, for example, indoor communication, underwater
communication, underground communication, and in-cabin
internet service [9,238]. Furthermore, VLC is envisioned to
F. Visible Light Communication
be widely used in vehicle-to-vehicle communications, which
depend on the head and tail lights of cars for communica-
The visible light communication (VLC) is considered as tions [9,210,237]. Besides, VLC serves as a potential solution
one of the techniques that will be used in 6G, and it operates to build gigabit wireless networks underwater.
at the rarely explored THz frequencies [234]. Specifically, 6G However, VLC is still facing many challenges. For example,
moves to higher frequencies because of the spectral congestion the light source’s bandwidth restricts the speed of VLC [234],
in frequencies that 5G uses and the increasing requirements so new materials and mechanisms should be developed to
for higher data rates. VLC has high data rates, a large increase the light source’s bandwidth. Besides, Si-based detec-
frequency spectrum, high-speed transmission, and robustness tors used by VLC systems are more sensitive to infrared waves
against interference [235]. Hence, VLC contributes to the than visible light. Moreover, no application-specific integrated
development of short-range communications in 6G [236]. For circuits for VLC baseband processing. Furthermore, the data
the short-range communication, either data-modulated white processing for future systems will be a much more complex
laser diodes or light-emitting diodes are used as transmitters, to process.
while photodetectors are utilized as receivers. Besides, VLC is
considered as a complementary technology for radio frequency
G. Device-to-Device Communication
communication because it can utilize an unlicensed spectrum
for communication [10].
As the Internet of Everything (IoE) will be introduced in
6G, the number of edge devices’ booming increase imposes
LED Lamp unpredictable pressure on the communication between cen-
tralized servers and edge devices. Device-to-Device (D2D)
communication enables devices to communicate directly with-
LED Lamp out going through infrastructures like base stations or access
points, which guarantees ultra-low latency, speeds up the
transmission, improves communication quality, and offloads
Desktop traffic from the conventional cellular network in end-to-end
Television Smart Phone communication [52,53]. Thus, D2D communication will gain
much more attention in increasing data transmission speed,
decreasing end-to-end latency, and reducing the cost of the
communication [239].
Specifically, D2D communication can operate in licensed
bands and unlicensed bands (i.e., Bluetooth and Wi-Fi). Be-
Fig. 17: Visible Light Communication. cause of the decentralized features that D2D possesses, D2D

20
obtains its advantages in multiple areas, such as mobile edge key sharing in traditional cryptography, physical layer security
computing, IoT, IoE, etc., which also share decentralized methods that can guarantee wireless communication security
features. With the aid of D2D communication, coordination are proposed.
between the centralized server and user’s devices has signifi-
cantly changed. Users who are close to each other exchange
data without forwarding through the BS nearby [240]. Hence, A. Attack Types
D2D communication can reduce energy consumption and
upgrade its quality of service (QoS) demands. The attacks to the physical layer of the wireless communi-
D2D is encountering challenges, including severe interfer- cation system can be classified into two categories: the passive
ence, prohibitively high cost, a vast amount of signaling, com- attack and the active attack.
plex resource management, and energy consumption. To facili- Passive Attack. The passive attack means that attackers
tate the D2D communication, more advanced and state-of-the- attempt to learn the information from the transmission without
art technologies such as AI and IRS can be utilized [239]. To affecting the system resources, as shown in Fig. (18). It can
be more specific, D2D communication will become AI-Driven be divided into two distinct categories: one is to obtain the
and intelligent in 6G. One application is to leverage AI to content of the message directly; the other one is to analyze
manage resources intelligently. AI-driven D2D communication the data flow. The eavesdropping attack is a way to obtain the
will enable three kinds of applications, including NOMA- effective data sent from the original station to the destination
based D2D cognitive networking, intelligent D2D-enhanced station without affecting the normal data communication. By
mobile edge computing, and D2D-enabled intelligent network monitoring the data transmission, the eavesdropping attack
slicing. Network slicing (NS) helps to manage and share damages data confidentiality and results in privacy leakage.
resources in 5G network. A large number of D2D clusters Suppose some approaches, such as encryption, make the
can extend the network’s flexibility, which enables to provide attacker unable to obtain the true content of the message from
services according to users’ needs. In 6G, the large number of the intercepted parts. In that case, the attacker may obtain the
D2D clusters can provide both physical and virtual resources, message format, determine the location and identity of both
which are critical to building NS. AI will be employed to sides of the communication, the number of communication
support and manage D2D clusters. The process of distributing times, and the length of the message, which may also be
resources will be intelligent and automatic, which means sensitive to both sides of the communication. Since the passive
that AI will monitor the underlying resource and network attack does not make any modification to the message, it is
slices to intelligently achieve resource mapping. Combined difficult to detect.
with NOMA, D2D will support the cognitive network. Delta-
orthogonal multiple access (D-OMA) technique is proposed to Sender Receiver
Message
obtain large scale concurrent access. D-OMA which serves as
the new multiple access scheme for 6G is leveraged to tackle
terminal devices, including the high complexity and increased
energy consumption at terminal devices.

H. Network in Box Observing

Message

More and more kinds of technologies will be embedded

Attacker
in 6G, such as the autonomous vehicles, factory automation,
etc. To satisfy the real-time and reliable features of the Fig. 18: Passive Attack.
network, Network in Box (NIB) technique has attracted much
more attention from the industrial automation because NIB Active Attack. The active attack is a deliberate action of
offers a device that can provide seamless connectivity between actively accessing the required data, causing a direct impact on
different services. On the other hand, NIB covers the wireless legitimate users as illustrated in Fig. (19). The active attack can
environments, including ground, air, and marine well agreed be divided into three categories: The first one is to intercept
to the vision of the IoE in 6G. the data sent by the original station, interrupting the effective
data so that the destination station cannot receive the data
IX. S ECURITY AND P RIVACY transmitted by the original station. It affects data availability.
The second one is to tamper with the data sent from the
In the 6G wireless communication systems, edge devices original station to the destination station, thereby affecting the
with the high mobility are getting popular. A huge amount of information received by the destination station. It affects data
data are generated by mobile applications containing sensitive integrity. The third one is to forge data and send it to the
information, which closely relates to users’ privacy. Because of destination station even if the original station does not send
the wireless communication system’s broadcast nature, eaves- data, which affects data authenticity. The primary approach to
droppers may monitor and steal users’ confidential information deal with the active attack is to detect and recover the damage
during the transmission. To avoid the necessary but complex caused by it.

21
 Sender Receiver Bayesian classifier to analyze amplitude, phase, and frequency,
thus providing authentication for the communication system.
Dan et al. [251] propose a method using the radio frequency as
physical fingerprints to authenticate a Wi-Fi device’s identity.
The approach of radio frequency fingerprint authentication
is effective in the aspect of preventing network intrusion,
Modifying
especially when cryptography based authentication techniques
Message are challenging to implement in some specific systems.
Power Allocation Approaches. In Wyner’s wiretap channel
Attacker model, the eavesdropper’s channel’s quality is worse than that
of the legitimate user’s to achieve communication confiden-
Fig. 19: Active Attack. tiality. But in some cases, the eavesdropper is close to the
original station than the legitimate user, which means that the
eavesdropper has better channel quality than the legitimate
B. Technologies for 6G Physical Layer Security user. Hence, the eavesdropper may monitor and receive infor-
In 6G, two basic requirements need to be satisfied when mation that is confidential to the legitimate users. To prevent
it comes to the physical layer security: confidentiality and the information leakage, Goel et al. [252] propose to add the
authentication. Confidentiality means that eavesdroppers have artificial noise to the channel to deteriorate the eavesdropper’s
no access to the effective data. The authentication ensures that channel, thus achieving minimum guaranteed secrecy capacity.
attackers cannot forge data and send them to the destination Using this method, if the original station has more antennas
station. Although traditional cryptographic technologies such than the eavesdropper, the transmitter can use part of the power
as the public key cryptography can enforce security, they to generate the artificial noise and inject it into the channel
are all implemented in the upper layer. Wyner shows that with multiple antennas. The information signal is transmitted
secure communication can be achieved in the physical layer in the range space in the legitimate channel. At the same
just by technologies adopted in noise and interference [241], time, the artificial noise is generated in the null space in the
which becomes the foundation of research on the security of eavesdropper’s channel, so the artificial noise only impairs the
the wireless communication theory. Csiszár and Köner [242] eavesdropper’s channel but not the intended receiver’s channel.
generalize Wyner’s results by adopting a non-degraded discrete However, this design heavily depends on the obtainment of
memoryless broadcast channel. Since then, more and more accurate channel knowledge.
researchers focus on this field and propose a large number of Signal Processing Approaches. In the 6G wireless com-
approaches. To be more specific, existing methods can be cate- munication systems, edge devices with high mobility are
gorized into channel approaches, power allocation approaches, getting popular. A huge amount of data are generated by
and signal processing approaches. The intelligent reflecting mobile applications containing sensitive information, which
surface (IRS) is a kind of metasurfaces that can improve may closely relate to users’ privacy. Also, technologies are
the efficiency of the spectrum utilization by reconfiguring the proposed, for example, the intelligent reflecting surfaces. Be-
reflection angle of signals regardless of the incidence angle. cause of the wireless communication systems’ broadcast na-
However, attacks may happen in IRS, and many have been ture, eavesdroppers may monitor and steal users’ confidential
investigated [243]–[245]. information during data transmission. To avoid the necessary
Channel Approaches. In recent years, many researchers but complex key sharing in traditional cryptography, physical
focus on the fundamental issues of secure channel capacity. layer security methods that can guarantee the security of the
Their work’s main idea is to distinguish the quality of signals wireless communication are proposed.
received by legitimate users and by unauthorized receivers.
Wyner has shown that reliable and secure transmission can
X. A PPLICATIONS
be achieved in degraded broadcast channels. The perfect
secrecy capacity is the gap between the attacker’s capac-
ity and legitimate user’s capacity for discrete memoryless Generally, the Internet of Things (IoT) refers to a network
channels [241]. [246,247] introduce the Gaussian channels of connected devices (also know as smart objects) that can
based on Wyner’s work and generalize the conclusion to collect and exchange data over the Internet [253]. Over the
Gaussian channels. Klinc et al. [248] propose an effective past few years, there is an emerging trend of employing
coding scheme for Gaussian wiretap channel based on LOPC artificial intelligence (AI) for IoT as AI has made lots of
codes, which is encodable in linear time. It can combine with remarkable achievements in the big-data era [254]. To date,
cryptography techniques, providing improved data security artificial intelligence of things (AIoT) has been widely used
protection in communication channels. The radio frequency in various kinds of areas. As AIoT becomes ubiquitous, 6G
recognition system proposed by Sperandio et al. [249] can will play a vital role in AIoT applications. First, 6G will
recognize the identities of transmitters from the received support a large number of smart devices to work cooperatively
signal. Each participant has its intrinsic physical properties, in an AIoT system while maintaining the low latency. Second,
so the system aims to process the extracted features and by integrating space-air-ground-sea networks, 6G provides a
obtain a fingerprint for each party. Cobb et al. [250] use a wider coverage for AIoT applications. Third, advanced AI

22
technologies will be implemented in 6G-based AIoT. Last but traditional healthcare systems require patients to visit the
not least, 6G enables AIoT to better protect users’ security and hospitals, which are time-consuming and labor-intensive [268].
privacy. In a word, with the help of 6G wireless network, both Therefore, healthcare systems improvement is significant for
device-device communication and device-server communica- people’s wellbeing. An efficient healthcare system is expected
tion will be greatly enhanced. For example, autonomous cars to carry out health monitoring, disease diagnosis, and medical
with high running speeds reply on low-latency communication treatment remotely with high efficiency. Therefore, researchers
to react quickly when avoiding of collision. To this end, have resorted to the smart healthcare. Smart healthcare is a
automobiles exchange information (e.g. speed and location) to healthcare service system that combines a variety of tech-
each other rapidly through vehicle-to-vehicle communication nologies such as wearable sensors and AIoT [269]. Smart
(V2V communication), a typical application of device-device healthcare can not only provide convenience for the people
communication. The example of device-server communication but also save lives in emergency. Even though time and space
is smart healthcare. The wearable and in-body sensors will are barriers of current healthcare systems, 6G wireless network
continuously collect patients’ information and upload the data enables smart healthcare to overcome these barriers. Therefore,
to the cloud serve for real time monitoring through low- 6G allows healthcare systems to complete more useful and so-
latency communication [255]. In fact, many researchers have phisticated tasks as illustrated in Fig. (20). That is, patients can
demonstrated the versatility of 6G-based AIoT systems on a be accurately diagnosed and treated by professional doctors
wide range of scenarios. In the following, we focus on several even if they are at home.
typical applications for further illustration and summarize
them in Table VIII. Local doctor
H2H Service
A. AI in Network Management
As 6G network becomes complex, it may utilize deep MRI
learning instead of human operators to improve the flexi- Pathology

bility and efficiency in the network management [18]. AI

technologies are applicable to both the physical and net-
work layers. In physical layer, AI techniques have involved CT Scan
in design and resource allocation in wireless communica- Local Server
BSR Sensor IWD
tions [5]. For example, unsupervised learning are applica-
ble to interference cancellation, optimal modulation, channel- Remote Doctor

aware feature-extraction, and channel estimation, etc. [18].

Deep reinforcement learning is possible to be employed for Prescription
link preservation, scheduling, transmission optimization, on-
Data Scientist
demand beamforming, and energy harvesting [18,121]. In
addition, AI technologies can be used to the network layer
as well. Supervised learning techniques can tackle problems
Remote Server
such as resource allocation, fault prediction [18]. Besides,
unsupervised learning algorithms can help in routing, traffic
control, parameter prediction, resource allocations, etc. [18]. Fig. 20: Smart Healthcare.
Reinforcement learning can be important for traffic prediction,
packet scheduling, multi-objective routing, security, and clas- Health Monitoring and Disease Detection. Continuous
sification [18,121]. health status monitoring is crucial in healthcare service, espe-
cially for those patients who suffer from sudden disease (e.g.,
B. AI in Autonomy cardiovascular diseases). In 6G-based healthcare systems,
AI technologies are potential to enable 6G wireless systems data-driven models that can detect abnormal health status will
to be autonomous [7,75,77]. Agents with intelligence can de- be uploaded to the cloud in advance. Patients’ physiological
tect and resolve network issues actively and autonomously. AI- data can be collected continuously by noninvasive sensors and
based network management contributes to monitoring network will be transmitted to the cloud. The models on the cloud will
status in real-time and keep network health. Also, AI tech- analyse the data and send a warning message if the diseases
niques can provide intelligence at the edge devices and edge occur suddenly. 6G technology will play a critical role as
computing, which enables edge devices and edge computing to both real-time monitoring and disease notification require high
learn to solve security problems autonomously [7,231,256]. In data rates to react quickly. What’s more, 6G will prompt the
addition, autonomous applications such as autonomous aerial development of the hospital-to-home (H2H) service to replace
vehicles and autonomous robots are envisioned to be available the traditional ambulance service [63].
in 6G [5]. Medical Treatment. The merits of 6G will support doctors
to perform more effective medical treatment. For example,
C. Smart Healthcare augmented reality (AR) and virtual reality (VR) can be applied
The number of Chronic patients and the aging of popula- in medical treatment because of high data rates and low latency
tion are increasing dramatically year by year [255]. Besides, in 6G communication. To be more specific, AR enables doc-

23
 TABLE VIII: Summarization of Applications in Artificial Intelligence of Things.
Typical applications Reference Requirements
Ultra-low latency
Smart Healthcare [63], [9], [19], [20], [257] High bandwidth
High security
Ultra-low latency
Ultra-high reliability
Smart Manufacturing [64], [10], [258], [51], [259], [260]
Ultra-high bandwidth
Very high intelligence
Ultra-low latency
Smart Home [65], [261], [260], [262] Ultra-high security
Very high intelligence
Ultra-low latency
Ultra-high bandwidth
Ultra-high security
Intelligent Transportation System [65], [263], [64], [237], [67], [264], [265], [266], [237]
High intelligence
High mobility
Long distance communication
Ultra-low latency
Smart Grid [66], [5]
Ultra-high security
Ultra-low latency
Ultra-high bandwidth
Unmanned Aerial Vehicle [67], [65], [9], [267], [3]
Ultra-high mobility
Ultra-long distance communication

tors to observe the inside of the patient’s body clearly without the connected sensors and machinery, thus leading to a preciser
making any incision, and doctors can try out various surgical and smarter manufacturing system [10].
plans in a simulated environment before real surgeries [63]. High-precision Manufacturing. It is vital for the manu-
6G also revitalizes the medical robots [9]. For example, the facturing system to maintain high precision during operations.
medical robots will take care of the patients and provide timely Numerically speaking, implementing high-precision manufac-
help for them when the hospital is too busy and most nurses turing requires very high reliability (up to about 109 ) and
are unavailable. Besides, the medical robots will help doctors extremely low latency (0.1 to 1ms round trip time) [258].
in surgeries as the 6G wireless network allows the robots to In addition, massive amounts of data and transmissions are
carry out complex tasks with high precision. Specifically, the involved in industrial control networks, hence requiring a very
size of the medical robots can be very small so that the doctors low delay jitter (about 1µs) [258].
will control them to enter the human’s body to take pictures, The above requirements can hardly be met if using 5G
deliver drugs, or remove diseased tissues. technology. In contrast, the emergence of 6G paves the way for
Privacy and Security in Healthcare. Security and privacy high-precision manufacturing because of its superior features.
are challenges of the 6G technology and the key concerns for For example, a new 6G-based architecture that integrates
patients [19]. Edge computing can be used to protect patients’ different resources has been proposed to satisfy the tight
privacy [63]. The data will be delivered to different edge physical constraints [51]. Some researchers also resort to
nodes because the memory of edge nodes is small. Therefore, other advanced IoT approaches, such as blockchain [259] and
healthcare data do not need to be stored in only one place, edge computing [260], to improve the performance of the
hence increasing communication security. Besides, selectively manufacturing systems.
uploading the data to the cloud helps to improve security as it
Intelligent Robots. It is very common in a modern man-
is easier for the cloud to protect lesser data [63]. Last but not
ufacturing system to apply intelligent robots to deal with
least, blockchains or federated learning are also possible ap-
dull and tedious work. Robots can also replace humans to
proaches to address the privacy problem in healthcare systems
carry out dangerous industrial operations or tasks requiring
in the future [20,257].
extremely high precision. According to Industry 4.0, the
robots are required to react quickly when interacting with
D. Smart Manufacturing humans and machinery in a dynamic environment [270]. To
The Industry 4.0 has envisioned a digital transformation of this end, it is essential to apply 6G technology into robotics
manufacturing through cyber physical systems and IoT ser- communication [63].
vices, and its main goal is to reduce the human intervention in The intelligent robots connected by 6G wireless network
industrial processes by using efficient control approaches and robots will be able to conduct complex cooperative opera-
communication technologies [51]. 6G will finally realize this tions [260,271]. For example, intelligent robots on the edge
revolution by investigating smart manufacturing [64]. Smart side can take videos of the industrial process and then up-
manufacturing refers to a IoT-connected manufacturing system load the data to the cloud, while the learning algorithms on
that applies a variety of control and data analytics approaches the cloud will make decisions to control the robots. Aided
to improve manufacturing performance. The advantages of 6G by 6G technology, robots will be competent enough even
will boost the communication and computing capabilities of though the whole control loop requires ultrahigh data rates.

24
As a result, senors, robots, machinery, and 6G will form an F. Intelligent Transportation System
efficient distributed intelligent network which has terabytes
of computing capacity [73]. Furthermore, there are some The intelligent transportation system (ITS) utilizes advanced
manufacturing processes which are hazardous but still require communication, control, and sensing technologies to provide
high precision, such as nuclear power plants and oil pipelines. safer and efficient traffic and traffic management. In a ITS, au-
In this case, nano-robots can be used in those dangerous tonomous driving vehicles require reliability above 99.99999%
environments [272]. and latency below 1 ms, while the vehicle speed in some
cases can be as high as to 1000 km/h [10,51]. However, the
ITS fails to meet these requirements in such high mobility
scenarios because of the insufficient capability of the current
E. Smart Home
communication technologies. In contrast, 6G network will
A smart home contains different kinds of IoT devices and significantly improve the capability of the ITS and make it
AI-driven in-network services to remotely control the house- satisfy the strict requirements.
hold systems like lighting, furniture, and thermostats [260]. Traffic Management. An effective traffic management ap-
Current smart homes have been able to control the furniture proach can force down the traffic jams, reduce passengers’
and house environment according to people’s commands. In waiting time, and preserve the road security. To provide real
the future, 6G will allow the household systems to be smarter, time transportation planning, the traffic information has to
like providing adaptive real-time control without much human be collected in high data rates. Besides, the global optimal
intervention. In addition to the convenience, residents’ safety solution will be obtained only if the coverage of the mobile
and privacy will also be well protected using 6G technologies communication network is large enough. Thus, it is necessary
such as federated learning. to apply 6G technology in traffic management owing to its high
Intelligent Furniture. Intelligent furniture in the smart speed Internet, low latency, and extensive coverage [65]. The
home will facilitate people’s life as well as save energy. To ITS empowered by 6G will keep guiding the drivers so as to
begin with, an intelligent light will switch off when nobody minimize the travel time. It is also promising to investigate 6G
occupies the room, and the light intensity can be continuously for traffic signal control problems since traffic signal control
tuned according to sunlight intensity [65]. Over time, this will involves a massive amount of real time traffic data and requires
save substantial electrical energy while people do not have to sophisticated algorithms to make decisions [263]. Last but not
pay extra attention to this matter. Similarly, the air conditioner least, 6G technology will enhance the public traffic security.
can work based on the indoor temperature and occupancy Police can utilize vehicle surveillance to track a suspected
detection. Furthermore, the running data of furniture can be vehicle [65]. The parameters and components of the vehicles
recorded on the cloud side so that individual preferences can will be continuously monitored to ensure safe driving.
be learned by AI algorithms. Under this circumstance, the Autonomous Vehicles. Autonomous vehicles are key ap-
intelligent furniture will be tuned with the considering of plications in the ITS [273]. Compared with the traffic man-
individual preferences, thus leading to a comfortable home. agement, the implementation of autonomous vehicles requires
Noted that real-time control, occupancy detection, and indi- even higher data rates [64]. 6G will help autonomous vehicles
vidual preferences estimation will generate a large amount of overcome the physical barriers and realize full automation. Ad-
data and requires higher capacity requirements. Therefore, 6G vanced communication methods, such as dedicated short-range
is essential for the implementation of intelligent furniture. communication (DSRC), vehicle-to-network (V2N), vehicle-
Emergency Detection. 6G will aid the smart homes in to-infrastructure (V2I), vehicle-to-Pedestrian (V2P), Vehicle-
emergency detection so as to keep residents’ safe. For ex- to-Home (V2H), and vehicle-to-everything (V2X), have great
ample, fall detection of the elderly is a major public health potential to formulate a comprehensive autonomous vehicle
challenge [261]. If an old man suddenly falls down in a smart network [237]. Besides, novel AI approaches, like real-time
home, the data collected by embedded intelligent sensors and intelligent edge, are indispensable for vehicle networks imple-
video surveillance will be sent to the cloud by 6G wireless mentation as they enable the autonomous vehicles to react to
network immediately. The well-trained prediction model on the unfamiliar environment in real time [67]. In addition to
the cloud side should detect this emergency by analysing the improving the speed, users’ privacy will be better protected
data and then send distress signals to the man’s relatives and in the next generation autonomous vehicles. For example,
the ambulance. Other kinds of emergencies, such as forced a Efficient and Privacy-preserving Truth Discovery (EPTD)
entry and fire, can be detected easily by a similar method. method is developed to strengthen the privacy protection [264].
Privacy Protection. Privacy sensitive data are frequently Also, researchers solve the security problems of vehicular Ad-
transmitted in the smart home, hence requiring a reliable hoc Networks (VANETs) by designing a privacy preserving
privacy protection approach. Some researchers tackled the machine learning-based collaborative intrusion detection sys-
privacy problems through edge-native solutions [260], which tem [264,265].
means data storage and processing will be done within the Airport and Waterway Transportation. In addition to the
residents’ premises. Mao et al. [262] propose an AI based land transportation, the applicability of 6G technology will
adaptive security specification method for 6G IoT networks to expand into airport and waterway transportation. By integrat-
address the privacy problems, and the proposed method has ing with satellite communication, 6G can provide localiza-
been evaluated in a smart room. tion services, broadcast, Internet connectivity, and weather

25
 time. In addition, by employing 6G technology, the scale of
the smart grid can be greatly extended without sacrificing the
Mobile Network
control precision and increasing the communication latency.

V2N
H. Unmanned Aerial Vehicle
The operation of unmanned aerial vehicle (UAV) requires
V2H
either manual control or autonomous control by intelligent
V2P
algorithms. Both two control approaches need exchange large
Pedestrian
amounts of data every second. Moreover, UAV is usually
V2X
V2V
expected to carry out tasks in a long distance at high altitudes,
which requires the wireless network to have high data rates
and large coverage. Even though UAV cannot be applied
V2I successfully in the 5G network, 6G technology will facilitate
the implementation of UAV due to the high capabilities of
6G [67].
Road Side Unit
UAV will be used as an aerial base station (BS) in 6G
wireless communication owing to its aerial superiority. By us-
ing Drone-to-Infrastructures (D2I) and Drone-to-Drone (D2D)
communication, the 6G network will maintain high data rate
Fig. 21: Vehicle-to-Everything (V2X). while extending the wireless coverage [65]. Besides the cover-
age, UAV has several advantages over fixed BS infrastructures,
such as strong line-of-sight links, easy deployment, and high
information to cellular users [18]. Satellite communication mobility [9,267]. Therefore, UAV can be used as a flexible
has potential benefits, such as providing readily connection and low-latency BS in the areas where the infrastructures
to moving objects, and it is expected to be used in wireless are absent or heavily loaded. For instance, natural disasters
network architectures in the future. Specifically, the global can destroy the ground communication infrastructures, and
coverage provided by 6G enables the ships and airplanes to be it is very difficult to establish new infrastructures in such a
connected with a variety of IoT devices, which leads to a more hazardous place [3]. In this case, UAV-supported aerial base
intelligent transportation system. With the ubiquitous connec- station will provide stable wireless connectivity.
tivity, 6G can keep updating the weather condition to the
captains, which ensures the transportation safety [65]. Besides, I. Smart and Autonomous Communication Systems
the satellite communication can be applied to connect the
land, air, and sea together into an integrated 6G system [266]. In this section, we present some potential use cases of AI in
For instance, AANETs, which is proposed in [274], demon- 6G such as AI in network management and AI in autonomy.
strated the feasibility of satellite communication in transport As 6G network becomes complex, it may utilize deep learning
network. Moreover, optical wireless communication, such as instead of human operators to improve the flexibility and
free space optics (FSO) [237] and visible light communication efficiency in the network management [18]. AI technologies
(VLC) [64], will also be useful in airport and waterway are applicable to both the physical and network layers. In
transportation. physical layer, AI techniques have involved in design and re-
source allocation in wireless communication [5]. For example,
unsupervised learning are applicable to channel-aware feature-
G. Smart Grid extraction, optimal modulation, interference cancellation, and
Smart grid is an IoT-based electricity network that utilizes channel estimation, etc. [18]. Deep reinforcement learning is
advanced communication and AI methods to deliver power in possible to be employed for link preservation, scheduling,
more efficient ways [66]. Researchers have been working on transmission optimization, on-demand beamforming, and en-
integrating 5G into the smart grid [5], yet few people have ergy harvesting, etc. [18,121]. In addition, AI technologies
considered applying 6G communication into this area. In the can be used to the network layer as well. Supervised learning
near future, however, 6G will be an indispensable technology techniques can tackle problems such as resource allocation,
for the further development of smart grid due to the fact that fault prediction, etc. [18]. Besides, unsupervised learning algo-
smart grid systems will demand more extensive computations rithms can help in routing, traffic control, parameter prediction,
and higher data rates. resource allocations, etc. [18]. Reinforcement learning can
In a smart grid system, all activities and electrical equipment be important for traffic prediction, packet scheduling, multi-
should be supervised in order to make sure the system runs objective routing, security, and classification, etc. [18,121].
smoothly and safely. 6G enables smart grid systems that In addition, AI technologies have potentials to enable 6G
contain a great number of IoT devices to conduct real-time wireless systems to be autonomous [7,75,77]. Agents with
remote monitoring and control. Moreover, because of ultra intelligence can detect and resolve network issues actively
reliability and low latency of 6G, the electricity network will and autonomously. AI-based network management contributes
be able to detect the fault quickly and then take actions in to monitoring network status in real-time and keep network

26
healthy. Also, AI techniques can provide intelligence at the or retransmissions is a promising method to enhance 6G
edge devices and edge computing, which enables edge devices networks’ energy efficiency. In [278], the authors tackle the
and edge computing to learn to solve security problems energy conversation problem in 6G networks through intel-
autonomously [7,231,256]. Besides, autonomous applications ligent resource management. In [51], the authors anticipate
such as autonomous aerial vehicles and autonomous robots are that effective energy-efficient communication strategies will
envisioned to be available in 6G [5]. be developed in 6G networks. The strategies are expected
to achieve battery-free communications, targeting communi-
J. Intelligent Vehicle-to-Everything Communications cation efficiency in the order of 1 pJ/bit [279]. Last but not
The vehicular network builds the bridge between human least, Dean Bubley envisions that ”energy budget” should be
beings and transportation [79], for example, vehicle-to-vehicle tied closely to costs (including externalities) in a wide variety
shown in Fig. (21). As the number of vehicles is increasing of fields [280].
rapidly, the over crowded vehicular network fails to achieve
high latency and low reliability. While traditional vehicular M. Holography Radio
networks attach great attention to vehicle-to-vehicle (V2V) and Holography radio is the highest level of interference ex-
vehicle-to-infrastructure (V2I) communications, the 6G vehic- ploitation and it improves spectrum efficiency and network
ular network will realize space-air-ground-sea even underwater capacity by controlling the entire physical space and the
vehicles. full loop of the electromagnetic field through spatial spec-
Although 5G technology has spanned over cognitive radio tral holography and spatial wave field synthesis [4,153,281].
(CR), network function virtualization (NFV), and reactive Specifically, unwanted signals are treated as noises, and people
vehicular network control, they cannot meet the requirements try to reduce the interference caused by these noises. However,
of 6G communications. 6G requires to evolve to network in- in 6G, the interference is regarded as useful resources for
telligentization, intelligent radio, and self-learning with proac- developing holographic communication systems [73]. Inter-
tive exploration. Aided by advanced automation techniques ference exploitation is that communication system obtains
and sensitive collision avoidance ability, vehicular networks’s gains through decomposing interference. According to [157],
performance can be significantly enhanced in 6G era. the multi-user interference can be decomposed into construc-
tive and destructive parts using simple geometric relations.
K. Radio Access Network Constructive part is considered as beneficial communication
resources, which can be used to improve QoS of 6G communi-
A radio access network (RAN) connects a device to other
cation systems. In recent years, MIMO is gaining its popularity
parts of a network by radio access technology. RAN is of ut-
because of its high throughput. Thus, 6G is expected to realize
most importance as it can be applied to different kinds of AIoT
holographic MIMOs (HMIMOS) by combining MIMOS with
applications such as smart healthcare, connected vehicles, and
LIS or IRS. HMIMOS can be categorized as active HMIMOS
public surveillance. To integrate RAN into 6G network, Lee et
and passive HMIMOS based on the power consumption, which
al. [275] elaborate three critical performance requirements
are supported by LIS and IRS, respectively [153]. To be more
for 6G RAN, flexibility, massive interconnectivity, and energy
specific, active HMIMOS using LIS are equipped with RF
efficiency.
circuits and signal processing units, whereas passive HMIMOS
Moreover, several variants of RAN have great application
only use IRS for reflecting signals.
potential in the 6G era. For example, open-radio access
network (O-RAN) alliance, which combines extensible RAN XI. C ONCLUSION
(xRAN) forum and cloud RAN (C-RAN) alliance, is proposed
in [276]. O-RAN adopts virtualized network elements and In this paper, we highlight some promising technologies
open interfaces to integrate intelligence into RAN, thus en- in 6G networks. We present a detailed explanation of artifi-
abling itself to support 6G technology. Fadlullah et al. [277] cial intelligence, intelligent reflecting surfaces, SWIFT, THz
propose an all-photonic RAN that contains two key elements: communications, blockchain, space-air-ground-sea integrated
a photonic engine (PE) and an all-photonic arrayed antenna network, free-duplex technologies and how these technologies
unit (AAU). All-photonic RAN is envisioned to tackle sensing will be applied in 6G. In addition, we discuss the potential
tasks and services when deploying in 6G networks. security and privacy problems brought by these technologies.
Moreover, we envision that 6G will enable a large sum of new
application in facilitating our daily life.
L. Energy-efficient Networks
As a wide range of IoT devices will be involved in the R EFERENCES
future, the implementation of 6G networks will further exac- [1] J. Wills, “5G technology: Which country will be the first to adapt?”
erbate e-waste management challenges. Therefore, it is vital https://www.investopedia.com/articles/markets-economy/090916/
to consider energy consumption for 6G networks. 5g-technology-which-country-will-be-first-adapt.asp, April 23, 2020.
[2] L. Matti and L. Kari, “Key drivers and research challenges for 6G
Ali et al. [119] demonstrates several ways to build energy- ubiquitous wireless intelligence,” 6G Flagship, Oulu, Finland, White
efficient 6G networks. For example, the MAC layer is the best Paper, 2019.
layer of the system to perform energy conversation. Besides, [3] M. Z. Chowdhury, M. Shahjalal, S. Ahmed, and Y. M. Jang, “6G
wireless communication systems: Applications, requirements, technolo-
long battery life is also a crucial factor for designing energy- gies, challenges, and research directions,” IEEE Open Journal of the
efficient 6G Networks. In addition, transmit power control Communications Society, vol. 1, pp. 957–975, 2020.

27
 [4] W. Saad, M. Bennis, and M. Chen, “A vision of 6G wireless systems: [27] X. You, C. Wang, J. Huang, X. Gao, Z. Zhang, M. Wang, Y. Huang,
Applications, trends, technologies, and open research problems,” IEEE C. Zhang, Y. Jiang, J. Wang, M. Zhu, B. Sheng, D. Wang, Z. Pan,
network, vol. 34, no. 3, pp. 134–142, 2019. P. Zhu, Y. Yang, Z. Liu, P. Zhang, X. Tao, S. Li, Z. Chen,
[5] T. M. Ho, T. D. Tran, T. T. Nguyen, S. Kazmi, L. B. Le, C. S. X. Ma, C. I, S. Han, K. Li, C. Pan, Z. Zheng, L. Hanzo,
Hong, and L. Hanzo, “Next-generation wireless solutions for the smart X. Shen, Y. J. Guo, Z. Ding, H. Haas, W. Tong, P. Zhu, G. Yang,
factory, smart vehicles, the smart grid and smart cities,” arXiv preprint J. Wang, E. G. Larsson, H. Ngo, W. Hong, H. Wang, D. Hou,
arXiv:1907.10102, 2019. J. Chen, Z. Chen, Z. Hao, G. Li, R. Tafazolli, Y. Gao, V. Poor,
[6] M. B. Mollah, S. Zeadally, and M. A. K. Azad, “Emerging wireless G. Fettweis, and Y. Liang, “Towards 6G wireless communication
technologies for Internet of Things applications: Opportunities and networks: Vision, enabling technologies, and new paradigm
challenges,” in Encyclopedia of Wireless Networks. Springer Inter- shifts,” SCIENCE CHINA Information Sciences. [Online]. Avail-
national Publishing Cham, 2019, pp. 1–11. able: https://engine.scichina.com/publisher/ScienceChinaPress/journal/
[7] L. Lovén, T. Leppänen, E. Peltonen, J. Partala, E. Harjula, P. Poram- SCIENCECHINAInformationSciences///10.1007/s11432-020-2955-6
bage, M. Ylianttila, and J. Riekki, “EdgeAI: A vision for distributed, [28] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: the potential
edge-native artificial intelligence in future 6G networks,” The 1st 6G of data transmission with large intelligent surfaces,” IEEE Transactions
Wireless Summit, pp. 1–2, 2019. on Signal Processing, vol. 66, no. 10, pp. 2746–2758, 2018.
[8] K. David, J. Elmirghani, H. Haas, and X.-H. You, “Defining 6G: [29] ——, “The potential of using large antenna arrays on intelligent
Challenges and opportunities [from the guest editors],” IEEE Vehicular surfaces,” in IEEE 2017 Vehicular Technology Conference (VTC).
Technology Magazine, vol. 14, no. 3, pp. 14–16, 2019. IEEE, 2017, pp. 1–6.
[9] F. Tariq, M. R. Khandaker, K.-K. Wong, M. A. Imran, M. Bennis, and [30] Q. Wu and R. Zhang, “Intelligent reflecting surface enhanced wireless
M. Debbah, “A speculative study on 6G,” IEEE Wireless Communica- network via joint active and passive beamforming,” IEEE Transactions
tions, vol. 27, no. 4, pp. 118–125, 2020. on Wireless Communications, vol. 18, no. 11, pp. 5394–5409, 2016.
[10] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi, [31] H. Sarieddeen, N. Saeed, T. Y. Al-Naffouri, and M.-S. Alouini, “Next
“Toward 6G networks: Use cases and technologies,” IEEE Communi- generation terahertz communications: A rendezvous of sensing, imag-
cations Magazine, vol. 58, no. 3, pp. 55–61, 2020. ing, and localization,” IEEE Communications Magazine, vol. 58, no. 5,
[11] M. Xiao, S. Mumtaz, Y. Huang, L. Dai, Y. Li, M. Matthaiou, G. K. pp. 69–75, 2020.
Karagiannidis, E. Björnson, K. Yang, I. Chih-Lin, and A. Ghosh, [32] Y.-C. Liang, Q. Zhang, E. G. Larsson, and G. Y. Li, “Symbiotic
“Millimeter wave communications for future mobile networks,” IEEE radio: Cognitive backscattering communications for future wireless
Journal on Selected Areas in Communications, vol. 35, no. 9, pp. 1909– networks,” IEEE Transactions on Cognitive Communications and Net-
1935, 2017. working, vol. 6, no. 4, pp. 1242–1255, 2020.
[12] J. G. Andrews, T. Bai, M. N. Kulkarni, A. Alkhateeb, A. K. Gupta, [33] Y. Lu and X. Zheng, “6G: A survey on technologies, scenarios,
and R. W. Heath, “Modeling and analyzing millimeter wave cellular challenges, and the related issues,” Journal of Industrial Information
systems,” IEEE Transactions on Communications, vol. 65, no. 1, pp. Integration, p. 100158, 2020.
403–430, 2016. [34] Y. Zhao, G. Yu, and H. Xu, “6G mobile communication net-
[13] L. Zhu, Z. Xiao, X.-G. Xia, and D. O. Wu, “Millimeter-wave com- work: vision, challenges and key technologies,” arXiv preprint
munications with non-orthogonal multiple access for B5G/6G,” IEEE arXiv:1905.04983, 2019.
Access, vol. 7, pp. 116 123–116 132, 2019. [35] M. W. Akhtar, S. A. Hassan, R. Ghaffar, H. Jung, S. Garg, and M. S.
[14] S. J. Nawaz, S. K. Sharma, S. Wyne, M. N. Patwary, and M. Asaduz- Hossain, “The shift to 6g communications: vision and requirements,”
zaman, “Quantum machine learning for 6G communication networks: Human-centric Computing and Information Sciences, vol. 10, no. 1,
State-of-the-art and vision for the future,” IEEE Access, vol. 7, pp. pp. 1–27, 2020.
46 317–46 350, 2019. [36] W. paper, “5G evolution and 6G,” in NTT DOCOMO, INC., 2020.
[15] Y. Corre, G. Gougeon, J.-B. Doré, S. Bicaı̈s, B. Miscopein, E. Faus- [37] H. Xu, P. V. Klainea, O. Oniretia, B. Caob, M. Imrana, and L. Zhang,
surier, M. Saad, J. Palicot, and F. Bader, “Sub-THz spectrum as enabler “Blockchain-enabled resource management and sharing for 6G com-
for 6G wireless communications up to 1 Tbit/s,” 2019. munications,” arXiv preprint arXiv:2003.13083, 2020.
[16] M. Giordani and M. Zorzi, “Satellite communication at millimeter [38] K. Fan, Y. Ren, Y. Wang, H. Li, and Y. Yang, “Blockchain-based
waves: A key enabler of the 6G era,” in 2020 International Conference efficient privacy preserving and data sharing scheme of content-centric
on Computing, Networking and Communications (ICNC). IEEE, 2020, network in 5G,” IET Communications, vol. 12, no. 5, pp. 527–532,
pp. 383–388. 2017.
[17] S. Underwood, “Blockchain beyond bitcoin,” 2016. [39] K. Kotobi and S. G. Bilen, “Secure blockchains for dynamic spectrum
[18] M. J. Piran and D. Y. Suh, “Learning-Driven wireless communications, access: A decentralized database in moving cognitive radio networks
towards 6G,” in 2019 International Conference on Computing, Elec- enhances security and user access,” ieee vehicular technology maga-
tronics and Communications Engineering (ICCECE). IEEE, 2019, pp. zine, vol. 13, no. 1, pp. 32–39, 2018.
219–224. [40] H. Yang, H. Zheng, J. Zhang, Y. Wu, Y. Lee, and Y. Ji, “Blockchain-
[19] S. Dang, O. Amin, B. Shihada, and M.-S. Alouini, “What should 6G based trusted authentication in cloud radio over fiber network for 5G,”
be?” Nature Electronics, vol. 3, no. 1, pp. 20–29, 2020. in 2017 16th International Conference on Optical Communications and
[20] S. Nayak and R. Patgiri, “6G: Envisioning the key issues and chal- Networks (ICOCN). IEEE, 2017, pp. 1–3.
lenges,” arXiv preprint arXiv:2004.04024, 2020. [41] Y. A. Chau and S. H. Yu, “Space modulation on wireless fading
[21] N. Kato, B. Mao, F. Tang, Y. Kawamoto, and J. Liu, “Ten challenges in channels,” in IEEE 54th Vehicular Technology Conference. VTC Fall
advancing machine learning technologies toward 6G,” IEEE Wireless 2001. Proceedings. IEEE, 2001, pp. 1668–1671.
Communications, 2020. [42] H. Wang, W. Wang, and X. Chen, “Wireless information and energy
[22] R. Shafin, L. Liu, V. Chandrasekhar, H. Chen, J. Reed, and J. C. transfer in interference aware massive MIMO systems,” in 2014 IEEE
Zhang, “Artificial intelligence-enabled cellular networks: A critical path Global Communications Conference. IEEE, 2014, pp. 2556–2561.
to beyond-5G and 6G,” IEEE Wireless Communications, vol. 27, no. 2, [43] R. Zhang, R. G. Maunder, and L. Hanzo, “Wireless information and
pp. 212–217, 2020. power transfer: from scientific hypothesis to engineering practice,”
[23] J. Tan and L. Dai, “THz precoding for 6G: Applications, challenges, IEEE Communications Magazine, vol. 53, no. 8, pp. 99–105, 2015.
solutions, and opportunities,” arXiv preprint arXiv:2005.10752, 2020. [44] Y. Liu, Z. Ding, and M. Elkashlan, “Cooperative non-orthogonal
[24] S. Elmeadawy and R. M. Shubair, “Enabling technologies for 6G multiple access with simultaneous wireless information and power
future wireless communications: Opportunities and challenges,” arXiv transfer,” IEEE Journals on Selected Areas in Communications, vol. 34,
preprint arXiv:2002.06068, 2020. no. 4, pp. 938–953, 2016.
[25] G. Gui, M. Liu, F. Tang, N. Kato, and F. Adachi, “6G: Opening new [45] Z. Yang, Z. Ding, P. Fan, and N. Al-Dhahir, “The impact of power
horizons for integration of comfort, security and intelligence,” IEEE allocation on cooperative non-orthogonal multiple access networks with
Wireless Communications, 2020. SWIPT,” IEEE Transactions on Wireless Communications, vol. 16,
[26] S. Chen, Y.-C. Liang, S. Sun, S. Kang, W. Cheng, and M. Peng, no. 7, pp. 4332–4343, 2017.
“Vision, requirements, and technology trend of 6G: how to tackle the [46] J. Gong and X. Chen, “Achievable rate region of non-orthogonal mul-
challenges of system coverage, capacity, user data-rate and movement tiple access systems with wireless powered decoder,” IEEE Journals
speed,” IEEE Wireless Communications, vol. 27, no. 2, pp. 218–228, on Selected Areas in Communications, vol. 35, no. 12, pp. 2846–2859,
2020. 2017.

28
[47] Y. Alsaba, C. Y. Leow, and S. K. A. Rahim, “Full-duplex cooperative [70] Q. Mao, F. Hu, and Q. Hao, “Deep learning for intelligent wireless
non-orthogonal multiple access with beamforming and energy harvest- networks: A comprehensive survey,” IEEE Communications Surveys &
ing,” IEEE Access, vol. 6, pp. 19 726–19 738, 2018. Tutorials, vol. 20, no. 4, pp. 2595–2621, 2018.
[48] L. Bariah, S. Muhaidat, and A. A. Dweik, “Error probability analysis [71] A. Zappone, M. Di Renzo, M. Debbah, T. T. Lam, and X. Qian,
of NOMA-based relay networks with SWIPT,” IEEE Access, vol. 23, “Model-aided wireless artificial intelligence: Embedding expert knowl-
no. 7, pp. 1223–1226, 2019. edge in deep neural networks for wireless system optimization,” IEEE
[49] S. Li, L. Bariah, S. Muhaidat, P. Sofotasios, J. Liang, and A. Wang, Vehicular Technology Magazine, vol. 14, no. 3, pp. 60–69, 2019.
“Error analysis of NOMA-based user cooperation with SWIPT,” in [72] H. He, S. Jin, C.-K. Wen, F. Gao, G. Y. Li, and Z. Xu, “Model-driven
2019 15th International Conference on Distributed Computing in deep learning for physical layer communications,” IEEE Wireless
Sensor Systems (DCOSS). IEEE, 2019, pp. 507–513. Communications, 2019.
[50] H. Haas, “LiFi is a paradigm-shifting 5G technology,” Reviews in [73] B. Zong, C. Fan, X. Wang, X. Duan, B. Wang, and J. Wang, “6G
Physics, vol. 3, pp. 26–31, 2018. technologies: Key drivers, core requirements, system architectures, and
[51] E. C. Strinati, S. Barbarossa, J. L. Gonzalez-Jimenez, D. Ktenas, enabling technologies,” IEEE Vehicular Technology Magazine, vol. 14,
N. Cassiau, L. Maret, and C. Dehos, “6G: The next frontier: From no. 3, pp. 18–27, 2019.
holographic messaging to artificial intelligence using subterahertz and [74] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, and Y.-J. A. Zhang, “The
visible light communication,” IEEE Vehicular Technology Magazine, roadmap to 6G: AI empowered wireless networks,” IEEE Communi-
vol. 14, no. 3, pp. 42–50, 2019. cations Magazine, vol. 57, no. 8, pp. 84–90, 2019.
[52] U. N. Kar and D. K. Sanyal, “A critical review of 3GPP standardization [75] H. Gacanin, “Autonomous wireless systems with artificial intelligence:
of device-to-device communication in cellular networks,” SN Computer A knowledge management perspective,” IEEE Vehicular Technology
Science, vol. 1, no. 1, p. 37, 2020. Magazine, vol. 14, no. 3, pp. 51–59, 2019.
[53] P. K. Malik, D. S. Wadhwa, and J. S. Khinda, “A survey of device [76] R.-A. Stoica and G. T. F. de Abreu, “6G: the wireless communica-
to device and cooperative communication for the future cellular net- tions network for collaborative and AI applications,” arXiv preprint
works,” International Journal of Wireless Information Networks, pp. arXiv:1904.03413, 2019.
1–22, 2020. [77] Z. Zhang, Y. Xiao, Z. Ma, M. Xiao, Z. Ding, X. Lei, G. K. Kara-
[54] A. Asadi, Q. Wang, and V. Mancuso, “A survey on device-to-device giannidis, and P. Fan, “6G wireless networks: Vision, requirements,
communication in cellular networks,” IEEE Communications Surveys architecture, and key technologies,” IEEE Vehicular Technology Mag-
& Tutorials, vol. 16, no. 4, pp. 1801–1819, 2014. azine, vol. 14, no. 3, pp. 28–41, 2019.
[55] Y. Zhang, E. Pan, L. Song, W. Saad, Z. Dawy, and Z. Han, “Social [78] L. Zhang, Y.-C. Liang, and D. Niyato, “6G visions: Mobile ultra-
network aware device-to-device communication in wireless networks,” broadband, super Internet-of-Things, and artificial intelligence,” China
IEEE Transactions on Wireless Communications, vol. 14, no. 1, pp. Communications, vol. 16, no. 8, pp. 1–14, 2019.
177–190, 2014. [79] F. Tang, Y. Kawamoto, N. Kato, and J. Liu, “Future intelligent and
[56] S. Zhang, H. Zhang, and L. Song, “Beyond D2D: Full dimension UAV- secure vehicular network toward 6G: Machine-learning approaches,”
to-everything communications in 6G,” IEEE Transactions on Vehicular Proceedings of the IEEE, vol. 108, no. 2, pp. 292–307, 2019.
Technology, 2020. [80] Y. Xiao, G. Shi, Y. Li, W. Saad, and H. V. Poor, “Towards self-learning
edge intelligence in 6G,” arXiv preprint arXiv:2010.00176, 2020.
[57] Y. Sun, J. Liu, J. Wang, Y. Cao, and N. Kato, “When machine learning
meets privacy in 6G: A survey,” IEEE Communications Surveys & [81] F. Jameel, N. Sharma, M. A. Khan, I. Khan, M. M. Alam, G. Mas-
Tutorials, vol. 22, no. 4, pp. 2694–2724, 2020. torakis, and C. X. Mavromoustakis, “Machine learning techniques
for wireless-powered ambient backscatter communications: Enabling
[58] N. Akhtar and A. Mian, “Threat of adversarial attacks on deep learning
intelligent IoT networks in 6G era,” in Convergence of Artificial
in computer vision: A survey,” IEEE Access, vol. 6, pp. 14 410–14 430,
Intelligence and the Internet of Things. Springer, 2020, pp. 187–211.
2018.
[82] W. Guo, “Explainable artificial intelligence for 6G: Improving trust be-
[59] B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas,
tween human and machine,” IEEE Communications Magazine, vol. 58,
“Communication-efficient learning of deep networks from decentral-
no. 6, pp. 39–45, 2020.
ized data,” in Artificial Intelligence and Statistics. PMLR, 2017, pp.
[83] J. Du, C. Jiang, J. Wang, Y. Ren, and M. Debbah, “Machine learning
1273–1282.
for 6G wireless networks: Carrying forward enhanced bandwidth,
[60] Z. Zhou, H. Liao, B. Gu, K. M. S. Huq, S. Mumtaz, and J. Rodriguez, massive access, and ultrareliable/low-latency service,” IEEE Vehicular
“Robust mobile crowd sensing: When deep learning meets edge com- Technology Magazine, vol. 15, no. 4, pp. 122–134, 2020.
puting,” IEEE Network, vol. 32, no. 4, pp. 54–60, 2018.
[84] S. Han, T. Xie, I. Chih-Lin, L. Chai, Z. Liu, Y. Yuan, and C. Cui,
[61] R. Sattiraju, A. Weinand, and H. D. Schotten, “AI-assisted PHY “Artificial-Intelligence-Enabled air interface for 6G: Solutions, chal-
technologies for 6G and beyond wireless networks,” arXiv preprint lenges, and standardization impacts,” IEEE Communications Magazine,
arXiv:1908.09523, 2019. vol. 58, no. 10, pp. 73–79, 2020.
[62] K. B. Letaief, W. Chen, Y. Shi, J. Zhang, , and Y. A. Zhang, “The [85] Y. Liu, S. Bi, Z. Shi, and L. Hanzo, “When machine learning meets
roadmap to 6G: AI empowered wireless networks,” IEEE Communi- big data: A wireless communication perspective,” IEEE Vehicular
cations Magazine, vol. 57, no. 8, pp. 84–90, 2019. Technology Magazine, vol. 15, no. 1, pp. 63–72, 2019.
[63] S. Nayak and R. Patgiri, “6G communication technology: A vision on [86] R. W. Liu, J. Nie, S. Garg, Z. Xiong, Y. Zhang, and M. S. Hossain,
intelligent healthcare,” arXiv preprint arXiv:2005.07532, 2020. “Data-driven trajectory quality improvement for promoting intelligent
[64] N. Rajatheva, I. Atzeni, E. Bjornson, A. Bourdoux, S. Buzzi, J.- vessel traffic services in 6G-enabled maritime IoT systems,” IEEE
B. Dore, S. Erkucuk, M. Fuentes, K. Guan, Y. Hu, X. Huang, Internet of Things Journal, 2020.
J. Hulkkonen, J. M. Jornet, M. Katz, R. Nilsson, E. Panayirci, K. Rabie, [87] H. Yang, A. Alphones, Z. Xiong, D. Niyato, J. Zhao, and K. Wu,
N. Rajapaksha, M. Salehi, H. Sarieddeen, T. Svensson, O. Tervo, “Artificial-intelligence-enabled intelligent 6G networks,” IEEE Net-
A. Tolli, Q. Wu, and W. Xu, “White paper on broadband connectivity work, vol. 34, no. 6, pp. 272–280, 2020.
in 6G,” arXiv preprint arXiv:2004.14247, 2020. [88] Z. Huang and X. Cheng, “A general 3d space-time-frequency non-
[65] S. Nayak and R. Patgiri, “6G communications: A vision on the potential stationary model for 6G channels,” IEEE Transactions on Wireless
applications,” arxiv, 04 2020. Communications, 2020.
[66] X. Fang, S. Misra, G. Xue, and D. Yang, “Smart grid — the new and [89] M. Giordani, M. Polese, M. Mezzavilla, S. Rangan, and M. Zorzi,
improved power grid: A survey,” Communications Surveys & Tutorials, “Toward 6G networks: Use cases and technologies,” IEEE Communi-
IEEE, vol. 14, pp. 944–980, 01 2012. cations Magazine, vol. 58, no. 3, pp. 55–61, 2020.
[67] M. Wang, T. Zhu, T. Zhang, J. Zhang, S. Yu, and W. Zhou, “Security [90] Y. Dai, D. Xu, S. Maharjan, Z. Chen, Q. He, and Y. Zhang, “Blockchain
and privacy in 6G networks: New areas and new challenges,” Digital and deep reinforcement learning empowered intelligent 5G beyond,”
Communications and Networks, vol. 6, 07 2020. IEEE Network, vol. 33, no. 3, pp. 10–17, 2019.
[68] S. Zhang and D. Zhu, “Towards artificial intelligence enabled 6g: [91] F. Tang, Y. Zhou, and N. Kato, “Deep reinforcement learning for
State of the art, challenges, and opportunities,” Computer Networks, dynamic uplink/downlink resource allocation in high mobility 5G
p. 107556, 2020. hetnet,” IEEE Journal on Selected Areas in Communications, vol. 38,
[69] M. Elsayed and M. Erol-Kantarci, “AI-Enabled future wireless net- no. 12, pp. 2773–2782, 2020.
works: Challenges, opportunities, and open issues,” IEEE Vehicular [92] Z. Xiong, Y. Zhang, D. Niyato, R. Deng, P. Wang, and L.-C. Wang,
Technology Magazine, vol. 14, no. 3, pp. 70–77, 2019. “Deep reinforcement learning for mobile 5G and beyond: Funda-

29
 mentals, applications, and challenges,” IEEE Vehicular Technology Transactions on Wireless Communications, vol. 17, no. 5, pp. 3128–
Magazine, vol. 14, no. 2, pp. 44–52, 2019. 3140, 2018.
[93] R. Zhao, X. Wang, J. Xia, and L. Fan, “Deep reinforcement learning [115] S. Rajendran, V. Lenders, W. Meert, and S. Pollin, “Crowdsourced
based mobile edge computing for intelligent Internet of Things,” wireless spectrum anomaly detection,” IEEE Transactions on Cognitive
Physical Communication, vol. 43, p. 101184, 2020. Communications and Networking, vol. 6, no. 2, pp. 694–703, 2019.
[94] Y. Al-Eryani, M. Akrout, and E. Hossain, “Multiple access in cell- [116] H. Yang, X. Xie, and M. Kadoch, “Machine learning techniques and a
free networks: Outage performance, dynamic clustering, and deep case study for intelligent wireless networks,” IEEE Network, vol. 34,
reinforcement learning-based design,” IEEE Journal on Selected Areas no. 3, pp. 208–215, 2020.
in Communications, 2020. [117] H. Huang, W. Xia, J. Xiong, J. Yang, G. Zheng, and X. Zhu,
[95] H. He, R. Wang, S. Jin, C.-K. Wen, and G. Y. Li, “Beamspace channel “Unsupervised learning-based fast beamforming design for downlink
estimation in terahertz communications: A model-driven unsupervised MIMO,” IEEE Access, vol. 7, pp. 7599–7605, 2018.
learning approach,” arXiv preprint arXiv:2006.16628, 2020. [118] R. Nikbakht, A. Jonsson, and A. Lozano, “Unsupervised learning for
[96] C. Huang, R. Mo, C. Yuen et al., “Reconfigurable intelligent surface parametric optimization,” IEEE Communications Letters, 2020.
assisted multiuser MISO systems exploiting deep reinforcement learn- [119] S. Ali, W. Saad, N. Rajatheva, K. Chang, D. Steinbach, B. Sliwa,
ing,” arXiv preprint arXiv:2002.10072, 2020. C. Wietfeld, K. Mei, H. Shiri, H.-J. Zepernick et al., “6G white
[97] Y. Zhang, Z. Mou, F. Gao, J. Jiang, R. Ding, and Z. Han, “UAV- paper on machine learning in wireless communication networks,” arXiv
enabled secure communications by multi-agent deep reinforcement preprint arXiv:2004.13875, 2020.
learning,” IEEE Transactions on Vehicular Technology, vol. 69, no. 10, [120] J. Konečnỳ, H. B. McMahan, F. X. Yu, P. Richtárik, A. T. Suresh, and
pp. 11 599–11 611, 2020. D. Bacon, “Federated learning: Strategies for improving communica-
[98] S. P. Rout, “6G wireless communication: Its vision, viability, appli- tion efficiency,” arXiv preprint arXiv:1610.05492, 2016.
cation, requirement, technologies, encounters and research,” in 2020 [121] R. Shafin, L. Liu, V. Chandrasekhar, H. Chen, J. Reed, and J. Zhang,
11th International Conference on Computing, Communication and “Artificial intelligence-enabled cellular networks: A critical path to
Networking Technologies (ICCCNT). IEEE, 2020, pp. 1–8. beyond-5G and 6G,” arXiv preprint arXiv:1907.07862, 2019.
[99] J. Du, F. R. Yu, G. Lu, J. Wang, J. Jiang, and X. Chu, “MEC-assisted [122] Z. Yang, M. Chen, K.-K. Wong, H. V. Poor, and S. Cui, “Federated
immersive VR video streaming over terahertz wireless networks: A learning for 6G: Applications, challenges, and opportunities,” arXiv
deep reinforcement learning approach,” IEEE Internet of Things Jour- preprint arXiv:2101.01338, 2021.
nal, vol. 7, no. 10, pp. 9517–9529, 2020.
[123] T. Cousik, R. Shafin, Z. Zhou, K. Kleine, J. Reed, and L. Liu, “CogRF:
[100] L. Wang, K. Wang, C. Pan, W. Xu, N. Aslam, and L. Hanzo, “Multi- A new frontier for machine learning and artificial intelligence for 6G
agent deep reinforcement learning based trajectory planning for multi- RF systems,” arXiv preprint arXiv:1909.06862, 2019.
UAV assisted mobile edge computing,” IEEE Transactions on Cognitive
[124] W. Guo, “Explainable artificial intelligence (XAI) for 6G: Improving
Communications and Networking, 2020.
trust between human and machine,” arXiv preprint arXiv:1911.04542,
[101] Y. Lu, X. Huang, K. Zhang, S. Maharjan, and Y. Zhang, “Low-
2019.
latency federated learning and blockchain for edge association in
[125] A. Faisal, H. Sarieddeen, H. Dahrouj, T. Y. Al-Naffouri, and M.-S.
digital twin empowered 6G networks,” IEEE Transactions on Industrial
Alouini, “Ultra-massive MIMO systems at terahertz bands: Prospects
Informatics, 2020.
and challenges,” arXiv preprint arXiv:1902.11090, 2019.
[102] M. Aledhari, R. Razzak, R. M. Parizi, and F. Saeed, “Federated learn-
ing: A survey on enabling technologies, protocols, and applications,” [126] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The potential
IEEE Access, vol. 8, pp. 140 699–140 725, 2020. of data transmission with large intelligent surfaces,” IEEE Transactions
on Signal Processing, vol. 66, no. 10, pp. 2746–2758, 2018.
[103] K. Yang, T. Jiang, Y. Shi, and Z. Ding, “Federated learning via over-
the-air computation,” IEEE Transactions on Wireless Communications, [127] ——, “The potential of using large antenna arrays on intelligent
vol. 19, no. 3, pp. 2022–2035, 2020. surfaces,” in 2017 IEEE 85th Vehicular Technology Conference (VTC
[104] L. U. Khan, W. Saad, Z. Han, E. Hossain, and C. S. Hong, “Federated Spring). IEEE, 2017, pp. 1–6.
learning for Internet of Things: Recent advances, taxonomy, and open [128] Q.-U.-A. Nadeem, A. Kammoun, A. Chaaban, M. Debbah, and M.-S.
challenges,” arXiv preprint arXiv:2009.13012, 2020. Alouini, “Large intelligent surface assisted MIMO communications,”
[105] M. Chen, Z. Yang, W. Saad, C. Yin, H. V. Poor, and S. Cui, “A joint arXiv preprint arXiv:1903.08127, 2019.
learning and communications framework for federated learning over [129] M. Jung, W. Saad, and G. Kong, “Performance analysis of large in-
wireless networks,” IEEE Transactions on Wireless Communications, telligent surfaces (LISs): Uplink spectral efficiency and pilot training,”
2020. arXiv preprint arXiv:1904.00453, 2019.
[106] L. U. Khan, I. Yaqoob, M. Imran, Z. Han, and C. S. Hong, “6G wireless [130] E. De Carvalho, A. Ali, A. Amiri, M. Angjelichinoski, and R. W.
systems: A vision, architectural elements, and future directions,” IEEE Heath Jr, “Non-stationarities in extra-large scale massive MIMO,” arXiv
Access, vol. 8, pp. 147 029–147 044, 2020. preprint arXiv:1903.03085, 2019.
[107] Y. Qu, C. Dong, J. Zheng, Q. Wu, Y. Shen, F. Wu, and A. Anpalagan, [131] S. Hu, K. Chitti, F. Rusek, and O. Edfors, “User assignment with
“Empowering the edge intelligence by air-ground integrated federated distributed large intelligent surface (LIS) systems,” in 2018 IEEE
learning in 6G networks,” arXiv preprint arXiv:2007.13054, 2020. 29th Annual International Symposium on Personal, Indoor and Mobile
[108] Z. M. Fadlullah and N. Kato, “HCP: heterogeneous computing platform Radio Communications (PIMRC). IEEE, 2018, pp. 1–6.
for federated learning based collaborative content caching towards 6G [132] S. Hu, F. Rusek, and O. Edfors, “Cramér-rao lower bounds for posi-
networks,” IEEE Transactions on Emerging Topics in Computing, 2020. tioning with large intelligent surfaces,” in 2017 IEEE 86th Vehicular
[109] Z. Zhao, C. Feng, H. H. Yang, and X. Luo, “Federated-learning- Technology Conference (VTC-Fall). IEEE, 2017, pp. 1–6.
enabled intelligent fog radio access networks: Fundamental theory, key [133] S. Gong, X. Lu, D. T. Hoang, D. Niyato, L. Shu, D. I. Kim, and
techniques, and future trends,” IEEE Wireless Communications, vol. 27, Y.-C. Liang, “Toward smart wireless communications via intelligent
no. 2, pp. 22–28, 2020. reflecting surfaces: A contemporary survey,” IEEE Communications
[110] Z. Yang, M. Chen, W. Saad, C. S. Hong, and M. Shikh-Bahaei, “Energy Surveys & Tutorials, vol. 22, no. 4, pp. 2283–2314, 2020.
efficient federated learning over wireless communication networks,” [134] E. Basar, M. Di Renzo, J. De Rosny, M. Debbah, M.-S. Alouini, and
IEEE Transactions on Wireless Communications, 2020. R. Zhang, “Wireless communications through reconfigurable intelligent
[111] Q. Bai, J. Wang, Y. Zhang, and J. Song, “Deep learning-based channel surfaces,” IEEE Access, vol. 7, pp. 116 753–116 773, 2019.
estimation algorithm over time selective fading channels,” IEEE Trans- [135] Ö. Özdogan, E. Björnson, and E. G. Larsson, “Intelligent reflecting
actions on Cognitive Communications and Networking, vol. 6, no. 1, surfaces: Physics, propagation, and pathloss modeling,” IEEE Wireless
pp. 125–134, 2019. Communications Letters, vol. 9, no. 5, pp. 581–585, 2019.
[112] Z. Qin, H. Ye, G. Y. Li, and B.-H. F. Juang, “Deep learning in physical [136] Y.-C. Liang, R. Long, Q. Zhang, J. Chen, H. V. Cheng, and H. Guo,
layer communications,” IEEE Wireless Communications, vol. 26, no. 2, “Large intelligent surface/antennas (LISA): Making reflective radios
pp. 93–99, 2019. smart,” Journal of Communications and Information Networks, vol. 4,
[113] T. Wang, C.-K. Wen, H. Wang, F. Gao, T. Jiang, and S. Jin, “Deep no. 2, pp. 40–50, 2019.
learning for wireless physical layer: Opportunities and challenges,” [137] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis,
China Communications, vol. 14, no. 11, pp. 92–111, 2017. and I. Akyildiz, “A new wireless communication paradigm through
[114] K. N. Doan, T. Van Nguyen, T. Q. Quek, and H. Shin, “Content-aware software-controlled metasurfaces,” IEEE Communications Magazine,
proactive caching for backhaul offloading in cellular network,” IEEE vol. 56, no. 9, pp. 162–169, 2018.

30
[138] Q. Wu and R. Zhang, “Towards smart and reconfigurable environment: [162] M. Al-Nahhal, E. Basar, and U. Uysal, “Flexible generalized spatial
Intelligent reflecting surface aided wireless network,” IEEE Communi- modulation for visible light communications,” IEEE Transactions on
cations Magazine, vol. 58, no. 1, pp. 106–112, 2019. Vehicular Technology, pp. 1–1, 2020.
[139] Z.-Q. He and X. Yuan, “Cascaded channel estimation for large intelli- [163] X. Gao, Z. Bai, P. Gong, and D. O. Wu, “Design and performance
gent metasurface assisted massive MIMO,” IEEE Wireless Communi- analysis of led-grouping based spatial modulation in the visible light
cations Letters, vol. 9, no. 2, pp. 210–214, 2019. communication system,” IEEE Transactions on Vehicular Technology,
[140] X. Tan, Z. Sun, and J. M. Jornet, “Increasing indoor spectrum sharing vol. 69, no. 7, pp. 7317–7324, 2020.
capacity using smart reflect-array,” in 2016 IEEE International Con- [164] C. R. Kumar and R. K. Jeyachitra, “Dual-mode generalized spatial
ference on Communications (ICC). IEEE, 2016, pp. 1–6. modulation mimo for visible light communications,” IEEE Communi-
[141] X. Tan, Z. Sun, and D. Koutsoni, “Enabling indoor mobile millimeter- cations Letters, vol. 22, no. 2, pp. 280–283, 2018.
wave networks based on smart reflect-arrays,” in 2018 IEEE 29th [165] J. Wang, J. Zhu, S. Lin, and J. Wang, “Adaptive spatial modulation
Annual International Symposium on Personal, Indoor and Mobile based visible light communications: Ser analysis and optimization,”
Radio Communications (PIMRC). IEEE, 2018, pp. 1–6. IEEE Photonics Journal, vol. 10, no. 3, pp. 1–14, 2018.
[142] S. Nie, J. M. Jornet, and I. F. Akyildiz, “Intelligent environments [166] G. Huang, S. Ouyang, Y. Ding, and V. Fusco, “Index modulation for
based on ultra-massive MIMO platforms for wireless communication frequency diverse array,” IEEE Antennas and Wireless Propagation
in millimeter wave and terahertz bands,” in 2019 IEEE International Letters, vol. 19, no. 1, pp. 49–53, 2020.
Conference on Acoustics, Speech and Signal Processing (ICASSP). [167] S. Y. Nusenu, S. Huaizong, Y. Pan, and A. Basit, “Space-frequency
IEEE, 2019, pp. 7849–7853. increment index modulation approach for fifth generation and beyond
[143] E. Basar, “Reconfigurable intelligent surface-based index modulation: wireless communication systems,” IEEE Transactions on Vehicular
A new beyond MIMO paradigm for 6G,” IEEE Transactions on Technology, vol. 69, no. 6, pp. 6286–6298, 2020.
Communications, vol. 68, no. 5, pp. 3187–3196, 2020. [168] A. A. Purwita, A. Yesilkaya, M. Safari, and H. Haas, “Generalized
[144] C. Liaskos, A. Tsioliaridou, and S. Nie, “An interpretable neural net- time slot index modulation for optical wireless communications,” IEEE
work for configuring pro-grammable wireless environments,” in 2019 Transactions on Communications, vol. 68, no. 6, pp. 3706–3719, 2020.
IEEE 20th International Workshop on Signal Processing Advances in [169] N. H. Nguyen, B. Berscheid, and H. H. Nguyen, “Fast-ofdm with index
Wireless Communications (SPAWC). IEEE, 2019, pp. 1–6. modulation for nb-iot,” IEEE Communications Letters, vol. 23, no. 7,
[145] A. L. Yuille and A. Rangarajan, “The concave-convex procedure,” pp. 1157–1160, 2019.
Neural Computation, vol. 15, no. 4, pp. 915–936, 2003. [170] S. Althunibat, R. Mesleh, and K. Qaraqe, “Quadrature index mod-
[146] T. Lipp and S. Boyd, “Variations and extension of the convex-concave ulation based multiple access scheme for 5g and beyond,” IEEE
procedure,” Optimization and Engineering, vol. 17, no. 2, pp. 263–287, Communications Letters, vol. 23, no. 12, pp. 2257–2261, 2019.
2016. [171] Y. Yang, M. Ma, S. Aı̈ssa, and L. Hanzo, “Physical-layer secret key
[147] S. Boyd and L. Vandenberghe, “Convex optimization,” Cabridge Uni- generation via cqi-mapped spatial modulation in multi-hop wiretap
versity Press, 2014. ad-hoc networks,” IEEE Transactions on Information Forensics and
[148] Y. Yang, S. Zhang, and R. Zhang, “Irs-enhanced OFDM: Power Security, vol. 16, pp. 1322–1334, 2020.
allocation and passive array optimization,” in 2019 IEEE Global
[172] Y. Shi, X. Lu, K. Gao, J. Zhu, and S. Wang, “Subblocks set design aided
Communications Conference (GLOBECOM). IEEE, 2019, pp. 1–6.
orthogonal frequency division multiplexing with all index modulation,”
[149] S. Hu, F. Rusek, and O. Edfors, “Capacity degradation with modeling
IEEE Access, vol. 7, pp. 52 659–52 668, 2019.
hardware impairment in large intelligent surface,” in 2018 IEEE Global
[173] ——, “Genetic algorithm aided ofdm with all index modulation,” IEEE
Communications Conference (GLOBECOM). IEEE, 2018, pp. 1–6.
Communications Letters, vol. 23, no. 12, pp. 2192–2195, 2019.
[150] Q. U. A. Nadeem, H. Alwazani, and A. Kammoun, “Intelligent
[174] Z. Yu, Z. Bai, K. Pang, X. Hao, X. Yang, and R. Li, “Optimization of
reflecting surface assisted multi-user MISO communication: channel
phase rotation-based precoding for spatial modulation system,” in 2020
estimation and beamforming design,” IEEE Open Journal of the
IEEE 20th International Conference on Communication Technology
Communications Society, vol. 1, pp. 661–680, 2020.
(ICCT), 2020, pp. 236–240.
[151] Z. He and X. Yuan, “Cascaded channel estimation for large intelligent
metasurface assisted massive MIMO,” IEEE Wireless Communication [175] Z. Zhang, C. Gong, H. Li, Y. Dong, X. Wang, and X. Dai, “Soft-input
Letters, vol. 9, no. 2, pp. 6210–6214, 2020. soft-output detection via expectation propagation for massive spatial
[152] B. Zheng and R. Zhang, “Intelligent reflecting surface-enhanced modulation mimo systems,” IEEE Communications Letters, pp. 1–1,
OFDM: Channel estimation and reflection optimization,” IEEE Wireless 2020.
Communications Letters, vol. 9, no. 4, pp. 518–522, 2019. [176] S. Katla, L. Xiang, Y. Zhang, E. El-Hajjar, and et al., “Deep learning
[153] C. Huang, S. Hu, G. C. Alexandropoulos, A. Zappone, C. Yuen, assisted detection for index modulation aided mmwave systems,” IEEE
R. Zhang, M. Di Renzo, and M. Debbah, “Holographic MIMO surfaces Access, vol. 8, pp. 202 738–202 754, 2020.
for 6G wireless networks: Opportunities, challenges, and trends,” IEEE [177] K. Satyanarayana, M. El-Hajjar, A. A. M. Mourad, and et al., “Soft-
Wireless Communications, vol. 27, no. 5, pp. 118–125, 2020. decoding for multi-set space-time shift-keying mmwave systems: A
[154] B. Di, H. Zhang, L. Song, Y. Li, Z. Han, and H. V. Poor, “Hybrid deep learning approach,” IEEE Access, vol. 8, pp. 49 584–49 595, 2020.
beamforming for reconfigurable intelligent surface based multi-user [178] T. Mao, Q. Wang, Z. Wang, and S. Chen, “Novel index modulation
communications: Achievable rates with limited discrete phase shifts,” techniques: A survey,” IEEE Communications Surveys & Tutorials,
IEEE Journal on Selected Areas in Communications, vol. 38, no. 8, vol. 21, no. 1, pp. 315–348, 2018.
pp. 1809–1822, 2020. [179] M. Wen, B. Zheng, K. J. Kim, and et al., “A survey on spatial
[155] D. M. Pozar, Microwave engineering. John wiley & sons, 2011. modulation in emerging wireless systems: Research progresses and
[156] S. Koziel, “Surrogate-based modeling and optimization: applications in applications,” IEEE Journal on Selected Areas in Communications,
engineering,” New York, NY: Springer, 2013. vol. 37, no. 9, pp. 1949–1972, 2019.
[157] X. Meng, F. Liu, J. Zhou, and S. Yang, “Interference exploitation pre- [180] Q. Gu, G. Wang, and R. Fan, “Rate-energy tradeoff in simultaneous
coding for intelligent reflecting surface aided communication system,” wireless information and power transfer over fading channels with
IEEE Wireless Communications Letters, 2020. uncertain distribution,” IEEE Transactions on Vehicular Technology,
[158] D. R. Smith, O. Yurduseven, L. P. Mancera, P. Bowen, and N. B. vol. 67, no. 4, pp. 3663–3668, 2018.
Kundtz, “Analysis of a waveguide-fed metasurface antenna,” Physical [181] G. Pan, H. Lei, and Y. Yuan, “Performance analysis and optimization
Review Applied, vol. 8, no. 5, p. 054048, 2017. for SWIPT wireless wensor networks,” IEEE Transactions on Commu-
[159] S. Abeywickrama, R. Zhang, Q. Wu, and C. Yuen, “Intelligent re- nications, vol. 65, no. 5, pp. 2291–2302, 2017.
flecting surface: practical phase shift model and beamforming opti- [182] M. Babaei, U. Aygolu, and E. Basar, “Ber analysis of dual-hop
mization,” in 2020 IEEE International Conference on Communications relaying with energy harvesting in Nakagami-m fading channel,” IEEE
(ICC). IEEE, 2020, pp. 1–6. Transactions on Communications, vol. 17, no. 7, pp. 4352–4361, 2018.
[160] ——, “Intelligent reflecting surface: practical phase shift model and [183] T. D. P. Perera, D. N. K. Jayakody, S. K. Sharma, S. Chatzinotas, and
beamforming optimization,” IEEE Transactions on Communications, J. Li, “Simultaneous wireless information and power transfer (SWIPT):
vol. 68, no. 9, pp. 5849–5863, 2020. Recent advances and future challenges,” IEEE Communications Sur-
[161] S. Gong, X. Lu, D. T. Hoang, and et al., “Toward smart wireless veys & Tutorials, vol. 20, no. 1, pp. 264–302, 2017.
communications via intelligent reflecting surfaces: A contemporary [184] W. C. Brown, “Experiments involving a microwave beam to power and
survey,” IEEE Communications Surveys Tutorials, vol. 22, no. 4, pp. position a helicopter,” IEEE Transactions on Aerospace and Electronic
2283–2314, 2020. Systems, vol. AES-5, no. 5, pp. 692–702, 1969.

31
[185] L. R. Varshney, “Transporting information and energy simultaneously,” [207] J. Zhang, M. M. Wang, T. Xia, and L. Wang, “Maritime IoT: An
in 2008 IEEE International Symposium on Information Theory (ISIT). architectural and radio spectrum perspective,” IEEE Access, vol. 8, pp.
IEEE, 2008, pp. 1612–1616. 93 109–93 122, 2020.
[186] J. Tang, J. Luo, J. Ou, and et al., “Decoupling or learning: Joint power [208] T. Xia, M. M. Wang, J. Zhang, and L. Wang, “Maritime Internet of
splitting and allocation in mc-noma with swipt,” IEEE Transactions on Things: Challenges and solutions,” IEEE Wireless Communications,
Communications, vol. 68, no. 9, pp. 5834–5848, 2020. vol. 27, no. 2, pp. 188–196, 2020.
[187] R. F. Buckley and R. W. Heath, “System and design for selective ofdm [209] L. Gupta, R. Jain, and G. Vaszkun, “Survey of important issues in UAV
swipt transmission,” IEEE Transactions on Green Communications and communication networks,” IEEE Communications Surveys & Tutorials,
Networking, pp. 1–1, 2020. vol. 18, no. 2, pp. 1123–1152, 2015.
[188] J. Wang, G. Wang, Z. Lin, and e. , “Swipt in mimo af relay systems [210] Y. Yuan, Y. Zhao, B. Zong, and S. Parolari, “Potential key technologies
with direct link,” in 2019 IEEE 89th Vehicular Technology Conference for 6G mobile communications,” Science China Information Sciences,
(VTC2019-Spring), 2019, pp. 1–6. vol. 63, pp. 1–19, 2020.
[189] I. Krikidis, S. Sasaki, S. Timotheou, and Z. Ding, “A low complexity [211] C. Han, Y. Wu, Z. Chen, and X. Wang, “Terahertz communications
antenna switching for joint wireless information and energy transfer in (TeraCom): Challenges and impact on 6G wireless systems,” arXiv
mimo relay channels,” IEEE Transactions on Communications, vol. 62, preprint arXiv:1912.06040, 2019.
no. 5, pp. 1577–1587, 2014.
[212] P. Chevalier, A. Armizhan, F. Wang, M. Piccardo, S. G. Johnson,
[190] F. K. Ojo and M. F. Mohd Salleh, “Energy efficiency optimization for
F. Capasso, and H. O. Everitt, “Widely tunable compact terahertz gas
swipt-enabled cooperative relay networks in the presence of interfering
lasers,” Science, vol. 366, no. 6467, pp. 856–860, 2019.
transmitter,” IEEE Communications Letters, vol. 23, no. 10, pp. 1806–
[213] C. Han and Y. Chen, “Propagation modeling for wireless communica-
1810, 2019.
tions in the terahertz band,” IEEE Communications Magazine, vol. 56,
[191] B. Fang and X. Zhu, “Linear precoder design maximizing energy
no. 6, pp. 96–101, 2018.
harvesting in swipt systems with finite-alphabet inputs,” in 2019 IEEE
90th Vehicular Technology Conference (VTC2019-Fall), 2019, pp. 1–5. [214] N. Khalid and O. B. Akan, “Wideband THz communication channel
[192] J. J. Park, J. H. Moon, H. H. Jang, and D. I. Kim, “Performance analysis measurements for 5G indoor wireless networks,” in 2016 IEEE Inter-
of power amplifier nonlinearity on multi-tone swipt,” IEEE Wireless national Conference on Communications (ICC). IEEE, 2016, pp. 1–6.
Communications Letters, pp. 1–1, 2020. [215] H. Yang and T. L. Marzetta, “A macro cellular wireless network
[193] H. H. Jang, K. W. Choi, and D. I. Kim, “Novel frequency-splitting swipt with uniformly high user throughputs,” in 2014 IEEE 80th Vehicular
for overcoming amplifier nonlinearity,” IEEE Wireless Communications Technology Conference (VTC2014-Fall). IEEE, 2014, pp. 1–5.
Letters, vol. 9, no. 6, pp. 826–829, 2020. [216] T. T. Vu, D. T. Ngo, N. H. Tran, H. Q. Ngo, M. N. Dao, and R. H.
[194] S. Zargari, A. Khalili, and R. Zhang, “Energy efficiency maximization Middleton, “Cell-free massive MIMO for wireless federated learning,”
via joint active and passive beamforming design for multiuser miso irs- IEEE Transactions on Wireless Communications, vol. 19, no. 10, pp.
aided swipt,” IEEE Wireless Communications Letters, pp. 1–1, 2020. 6377–6392, 2020.
[195] W. Sun, Q. Song, L. Guo, and J. Zhao, “Secrecy rate maximization for [217] M. Bashar, A. Akbari, K. Cumanan, H. Quoc Ngo, A. G. Burr, P. Xiao,
intelligent reflecting surface aided swipt systems,” in 2020 IEEE/CIC and M. Debbah, “Deep learning-aided finite-capacity fronthaul cell-free
International Conference on Communications in China (ICCC), 2020, massive MIMO with zero forcing,” IEEE ICC 2020, 2020.
pp. 1276–1281. [218] S. J. Nawaz, S. K. Sharma, B. Mansoor, M. N. Patwary, and
[196] R. Ma, H. Wu, J. Ou, S. Yang, and Y. Gao, “Power splitting-based swipt N. M. Khan, “Non-coherent and backscatter communications: Enabling
systems with full-duplex jamming,” IEEE Transactions on Vehicular ultra-massive connectivity in 6g wireless networks,” arXiv preprint
Technology, vol. 69, no. 9, pp. 9822–9836, 2020. arXiv:2005.10937, 2020.
[197] Z. Zhu, N. Wang, W. Hao, Z. Wang, and I. Lee, “Robust beamforming [219] L. Bariah, L. Mohjazi, S. Muhaidat, P. C. Sofotasios, G. K. Kurt,
designs in secure mimo swipt iot networks with a non-linear channel H. Yanikomeroglu, and O. A. Dobre, “A prospective look: Key enabling
model,” IEEE Internet of Things Journal, pp. 1–1, 2020. technologies, applications and open research topics in 6g networks,”
[198] A. Thakur, A. Kumar, N. Gupta, and P. Chatterjee, “Secrecy analysis arXiv preprint arXiv:2004.06049, 2020.
of reconfigurable underlay cognitive radio networks with swipt and im- [220] J. Mitola and G. Q. Maguire, “Cognitive radio: making software radios
perfect csi,” IEEE Transactions on Network Science and Engineering, more personal,” IEEE personal communications, vol. 6, no. 4, pp. 13–
pp. 1–1, 2020. 18, 1999.
[199] M. A. Hossain, R. Md Noor, K. A. Yau, I. Ahmedy, and S. S. Anjum, [221] V. Liu, A. Parks, V. Talla, S. Gollakota, D. Wetherall, and J. R. Smith,
“A survey on simultaneous wireless information and power transfer “Ambient backscatter: Wireless communication out of thin air,” ACM
with cooperative relay and future challenges,” IEEE Access, vol. 7, pp. SIGCOMM Computer Communication Review, vol. 43, no. 4, pp. 39–
19 166–19 198, 2019. 50, 2013.
[200] Y. Kawamoto, H. Nishiyama, N. Kato, and N. Kadowaki, “A traffic [222] N. Van Huynh, D. T. Hoang, X. Lu, D. Niyato, P. Wang, and D. I. Kim,
distribution technique to minimize packet delivery delay in multilay- “Ambient backscatter communications: A contemporary survey,” IEEE
ered satellite networks,” IEEE Transactions on Vehicular Technology, Communications surveys & tutorials, vol. 20, no. 4, pp. 2889–2922,
vol. 62, no. 7, pp. 3315–3324, 2013. 2018.
[201] R. Radhakrishnan, W. W. Edmonson, F. Afghah, R. M. Rodriguez-
[223] R. Long, Y. Liang, H. Guo, G. Yang, and R. Zhang, “Symbiotic radio:
Osorio, F. Pinto, and S. C. Burleigh, “Survey of inter-satellite com-
A new communication paradigm for passive internet of things,” IEEE
munication for small satellite systems: Physical layer to network layer
Internet of Things Journal, vol. 7, no. 2, pp. 1350–1363, 2020.
view,” IEEE Communications Surveys & Tutorials, vol. 18, no. 4, pp.
[224] M. Amjad, F. Akhtar, M. H. Rehmani, M. Reisslein, and T. Umer,
2442–2473, 2016.
“Full-duplex communication in cognitive radio networks: A survey,”
[202] C. Niephaus, M. Kretschmer, and G. Ghinea, “QoS provisioning in
IEEE Communications Surveys & Tutorials, vol. 19, no. 4, pp. 2158–
converged satellite and terrestrial networks: A survey of the state-of-
2191, 2017.
the-art,” IEEE Communications Surveys & Tutorials, vol. 18, no. 4, pp.
2415–2441, 2016. [225] Z. Yuan, Y. Ma, Y. Hu, and W. Li, “High-efficiency full-duplex V2V
[203] M. Hamdi, N. Boudriga, and M. S. Obaidat, “Bandwidth-effictive communication,” in 2020 2nd 6G Wireless Summit (6G SUMMIT).
design of a satellite-based hybrid wireless sensor network for mobile IEEE, 2020, pp. 1–5.
target detection and tracking,” IEEE Systems Journal, vol. 2, no. 1, pp. [226] D. Xu, X. Yu, Y. Sun, D. W. K. Ng, and R. Schober, “Resource
74–82, 2008. allocation for IRS-assisted full-duplex cognitive radio systems,” arXiv
[204] P. Chini, G. Giambene, and S. Kota, “A survey on mobile satellite preprint arXiv:2003.07467, 2020.
systems,” Int. J. Satellite Communications Networking, vol. 28, pp. [227] H. Shen, T. Ding, W. Xu, and C. Zhao, “Beamforming design with
29–57, 01 2009. fast convergence for IRS-Aided full-duplex communication,” IEEE
[205] W. Y. B. Lim, J. Huang, Z. Xiong, J. Kang, D. Niyato, X.-S. Hua, Communications Letters, 2020.
C. Leung, and C. Miao, “Towards federated learning in UAV-enabled [228] G. Pan, J. Ye, J. An, and M.-S. Alouini, “When Full-Duplex transmis-
Internet of Vehicles: A multi-dimensional contract-matching approach,” sion meets intelligent reflecting surface: Opportunities and challenges,”
arXiv preprint arXiv:2004.03877, 2020. arXiv preprint arXiv:2005.12561, 2020.
[206] T. Zeng, O. Semiari, M. Mozaffari, M. Chen, W. Saad, and M. Bennis, [229] G. Wood, “Ethereum: A secure decentralised generalised transaction
“Federated learning in the sky: Joint power allocation and scheduling ledger,” Ethereum project yellow paper, vol. 151, no. 2014, pp. 1–32,
with UAV swarms,” arXiv preprint arXiv:2002.08196, 2020. 2014.

32
[230] M. B. Mollah, J. Zhao, D. Niyato, K.-Y. Lam, X. Zhang, A. M. [254] A. Poniszewska-Maranda and D. Kaczmarek, “Selected methods of
Ghias, L. H. Koh, and L. Yang, “Blockchain for future smart grid: artificial intelligence for Internet of Things conception,” 10 2015, pp.
A comprehensive survey,” IEEE Internet of Things Journal, 2020. 1343–1348.
[231] P. Porambage, T. Kumar, M. Liyanage, J. Partala, L. Lovén, M. Yliant- [255] L. Mucchi, S. Jayousi, S. Caputo, E. Paoletti, P. Zoppi, S. Geli,
tila, and T. Seppänen, “Sec-EdgeAI: AI for edge security vs security and P. Dioniso, “How 6G technology can change the future wireless
for edge AI.” healthcare,” 03 2020, pp. 1–6.
[232] X. Ling, J. Wang, T. Bouchoucha, B. C. Levy, and Z. Ding, “Blockchain [256] M. B. Mollah, M. A. K. Azad, and A. Vasilakos, “Secure data sharing
radio access network (B-RAN): Towards decentralized secure radio and searching at the edge of cloud-assisted Internet of Things,” IEEE
access paradigm,” IEEE Access, vol. 7, pp. 9714–9723, 2019. Cloud Computing, vol. 4, no. 1, pp. 34–42, 2017.
[233] Y. Le, X. Ling, J. Wang, and Z. Ding, “Prototype design and test [257] T. Hewa, G. Gür, A. Kalla, M. Ylianttila, A. Braeken, and M. Liyanage,
of blockchain radio access network,” in 2019 IEEE International “The role of blockchain in 6G: Challenges, opportunities and research
Conference on Communications Workshops (ICC Workshops). IEEE, directions,” 03 2020.
2019, pp. 1–6. [258] G. Berardinelli, N. Mahmood, I. Rodriguez Larrad, and P. Mogensen,
[234] S. Ariyanti and M. Suryanegara, “Visible light communication (VLC) “Beyond 5G wireless irt for industry 4.0: Design principles and
for 6G technology: The potency and research challenges,” in 2020 spectrum aspects,” 12 2018, pp. 1–6.
Fourth World Conference on Smart Trends in Systems, Security and [259] R. Sekaran, R. Patan, A. Raveendran, F. Al-Turjman, R. MANIKAN-
Sustainability (WorldS4). IEEE, 2020, pp. 490–493. DAN, and L. Mostarda, “Survival study on blockchain based 6G-
[235] M. A. Arfaoui, M. D. Soltani, I. Tavakkolnia, A. Ghrayeb, M. Sa- enabled mobile edge computation for IoT automation,” IEEE Access,
fari, C. Assi, and H. Haas, “Physical layer security for visible light vol. PP, 08 2020.
communication systems: A survey,” IEEE Communications Surveys & [260] E. Peltonen, M. Bennis, M. Capobianco, M. Debbah, A. Ding, F. Gil-
Tutorials, 2020. Castiñeira, M. Jurmu, T. Karvonen, M. Kelanti, A. Kliks, T. Leppänen,
[236] “6G white paper.” http://jultika.oulu.fi/Record/isbn978-952-62-2354-4. L. Lovèn, T. Mikkonen, A. Rao, S. Samarakoon, K. Seppänen, P. Sroka,
[237] M. Z. Chowdhury, M. Shahjalal, M. Hasan, and Y. M. Jang, “The S. Tarkoma, and T. Yang, “6G white paper on edge intelligence,” arXiv
role of optical wireless communication technologies in 5G/6G and preprint arXiv:2004.14850, 2020.
IoT solutions: Prospects, directions, and challenges,” Applied Sciences, [261] N. Noury, A. Fleury, P. Rumeau, A. Bourke, G. ÓLaighin, V. Rialle,
vol. 9, no. 20, p. 4367, 2019. and J.-E. Lundy, “Fall detection – principles and methods,” Annual
[238] M. Katz and I. Ahmed, “Opportunities and challenges for visible International Conference of the IEEE Engineering in Medicine and
light communications in 6G,” in 2020 2nd 6G Wireless Summit (6G Biology Society, vol. 2007, pp. 1663–6, 02 2007.
SUMMIT). IEEE, 2020, pp. 1–5. [262] B. Mao, Y. Kawamoto, and N. Kato, “AI-based joint optimization of
[239] S. Zhang, J. Liu, H. Guo, M. Qi, and N. Kato, “Envisioning device- QoS and security for 6G energy harvesting Internet of Things,” IEEE
to-device communications in 6G,” IEEE Network, vol. 34, no. 3, pp. Internet of Things Journal, vol. PP, pp. 7032–7042, 03 2020.
86–91, 2020. [263] G. Zheng, X. Zang, N. Xu, H. Wei, Z. Yu, V. Gayah, K. Xu, and
[240] Y. Chen, B. Ai, H. Zhang, Y. Niu, L. Song, Z. Han, and H. V. Poor, Z. Li, “Diagnosing reinforcement learning for traffic signal control,”
“Reconfigurable intelligent surface assisted device-to-device commu- arXiv preprint arXiv:1905.04716, 2019.
nications,” arXiv preprint arXiv:2007.00859, 2020. [264] Y. Sun, J. Liu, J. Wang, Y. Cao, and N. Kato, “When machine learning
[241] A. D. Wyner, “The wire-tap channel,” Bell system technical journal, meets privacy in 6G: A survey,” IEEE Communications Surveys &
vol. 54, no. 8, pp. 1355–1387, 1975. Tutorials, vol. PP, 07 2020.
[242] I. Csiszár and J. Korner, “Broadcast channels with confidential mes- [265] T. Zhang and Q. Zhu, “Distributed privacy-preserving collaborative
sages,” IEEE transactions on information theory, vol. 24, no. 3, pp. intrusion detection systems for VANETs,” IEEE Transactions on Signal
339–348, 1978. and Information Processing over Networks, vol. PP, 02 2018.
[243] H. Shen, W. Xu, S. Gong, Z. He, and C. Zhao, “Secrecy rate [266] O. Dizdar, Y. Mao, W. Han, and B. Clerckx, “Rate-Splitting multiple
maximization for intelligent reflecting surface assisted multi-antenna access: A new frontier for the PHY layer of 6G,” arxiv, 06 2020.
communications,” IEEE Communications Letters, vol. 23, no. 9, pp. [267] P. Mehta, R. Gupta, and S. Tanwar, “Blockchain envisioned UAV
1488–1492, 2019. networks: Challenges, solutions, and comparisons,” Computer Com-
[244] S. Hong, C. Pan, H. Ren, K. Wang, and A. Nallanathan, “Artificial- munications, vol. 151, 02 2020.
noise-aided secure MIMO wireless communications via intelligent [268] A. Aygun, H. Ghasemzadeh, and R. Jafari, “Robust interbeat interval
reflecting surface,” arXiv preprint arXiv:2002.07063, 2020. and heart rate variability estimation method from various morphologi-
[245] X. Yu, D. Xu, Y. Sun, D. W. K. Ng, and R. Schober, “Robust and cal features using wearable sensors,” IEEE Journal of Biomedical and
secure wireless communications via intelligent reflecting surfaces,” Health Informatics, vol. PP, 12 2019.
IEEE Journal on Selected Areas in Communications, 2020. [269] S. Tian, W. Yang, J. M. Le Grange, P. Wang, W. Huang, and Z. Ye,
[246] H. Yamamoto, “A coding theorem for secret sharing communication “Smart healthcare: making medical care more intelligent,” Global
systems with two THz wiretap channels,” IEEE Transactions on Health Journal, vol. 3, 10 2019.
Information Theory, vol. 37, no. 3, pp. 634–638, 1991. [270] G. Torfs, H. Li, S. Agneessens, J. Bauwelinck, L. Breyne, O. Caytan,
[247] S. Leung-Yan-Cheong and M. Hellman, “The Gaussian wire-tap chan- W. Joseph, S. Lemey, H. Rogier, A. Thielens, D. V. Ginste, X. Yin, and
nel,” IEEE transactions on information theory, vol. 24, no. 4, pp. 451– P. Demeester, “ATTO: Wireless networking at fiber speed,” Journal of
456, 1978. Lightwave Technology, vol. 36, no. 8, pp. 1468–1477, 2017.
[248] D. Klinc, J. Ha, S. W. McLaughlin, J. Barros, and B.-J. Kwak, [271] J. Monserrat, D. Martin-Sacristan, F. Bouchmal, O. Carrasco, J. Flo-
“LDPC codes for the Gaussian wiretap channel,” IEEE Transactions on res de Valgas, and N. Cardona, “Key technologies for the advent of
Information Forensics and Security, vol. 6, no. 3, pp. 532–540, 2011. the 6G,” 04 2020, pp. 1–6.
[249] C. Sperandio and P. G. Flikkema, “Wireless physical-layer security via [272] I. Akyildiz, A. Kak, and S. Nie, “6G and beyond: The future of wireless
transmit precoding over dispersive channels: optimum linear eaves- communications systems,” IEEE Access, vol. PP, 07 2020.
dropping,” in MILCOM 2002. Proceedings, vol. 2. IEEE, 2002, pp. [273] N. Bizon, L. Dascalescu, and N. M. Tabatabaei, Autonomous vehicles:
1113–1117. Intelligent transport systems and smart technologies, 01 2014.
[250] W. E. Cobb, E. D. Laspe, R. O. Baldwin, M. A. Temple, and Y. C. Kim, [274] J. Zhang, T. Chen, S. Zhong, J. Wang, W. Zhang, X. Zuo, R. Maunder,
“Intrinsic physical-layer authentication of integrated circuits,” IEEE and L. Hanzo, “Aeronautical Ad Hoc networking for the Internet-
Transactions on Information Forensics and Security, vol. 7, no. 1, pp. Above-The-Clouds,” Proceedings of the IEEE, 05 2019.
14–24, 2011. [275] Y. L. Lee, D. Qin, L.-C. Wang, G. Hong, and Sim, “6G massive
[251] D. Shan, K. Zeng, W. Xiang, P. Richardson, and Y. Dong, “PHY- radio access networks: Key issues, technologies, and future challenges,”
CRAM: Physical layer challenge-response authentication mechanism 2019.
for wireless networks,” IEEE Journal on selected areas in communi- [276] S. Niknam, A. Roy, H. S. Dhillon, S. Singh, R. Banerji, J. H. Reed,
cations, vol. 31, no. 9, pp. 1817–1827, 2013. N. Saxena, and S. Yoon, “Intelligent O-RAN for beyond 5G and 6G
[252] S. Goel and R. Negi, “Guaranteeing secrecy using artificial noise,” wireless networks,” 2020.
IEEE transactions on wireless communications, vol. 7, no. 6, pp. 2180– [277] Z. Fadlullah and N. Kato, “HCP: Heterogeneous computing platform
2189, 2008. for federated learning based collaborative content caching towards
[253] G. Fortino and P. Trunfio, Internet of Things Based on Smart Objects, 6G networks,” IEEE Transactions on Emerging Topics in Computing,
Technology, Middleware and Applications, 01 2014. vol. PP, 04 2020.

33
[278] K. Sheth, K. Patel, H. Shah, S. Tanwar, R. Gupta, and N. Kumar, “A
taxonomy of ai techniques for 6G communication networks,” Computer
Communications, vol. 161, pp. 279 – 303, 2020.
[279] M. L. Memon, N. Saxena, A. Roy, and D. R. Shin, “Backscatter
communications: Inception of the battery-free era—a comprehensive
survey,” Electronics, vol. 8, p. 129, 2019.
[280] “Predictions for 2030: telecoms, networks, spectrum &
5G/6G,” https://disruptivewireless.blogspot.com/2020/01/
predictions-for-next-decade-looking-out.html.
[281] R. J. Williams, E. De Carvalho, and T. L. Marzetta, “A communication
model for large intelligent surfaces,” in 2020 IEEE International
Conference on Communications Workshops (ICC Workshops). IEEE,
2020, pp. 1–6.

34
View publication stats