Intelligent and Converged Networks ISSN 2708-6240

2022, 3(1): 1−32 DOI: 10.23919/ICN.2022.0005

Reconfigurable intelligent surfaces for wireless communications:

Overview of hardware designs, channel models,
and estimation techniques
(invited paper)

Mengnan Jian, George C. Alexandropoulos, Ertugrul Basar, Chongwen Huang,

Ruiqi Liu*, Yuanwei Liu, and Chau Yuen

Abstract: The demanding objectives for the future sixth generation (6G) of wireless communication networks have
spurred recent research efforts on novel materials and radio-frequency front-end architectures for wireless
connectivity, as well as revolutionary communication and computing paradigms. Among the pioneering candidate
technologies for 6G belong the reconfigurable intelligent surfaces (RISs), which are artificial planar structures with
integrated electronic circuits that can be programmed to manipulate the incoming electromagnetic field in a wide
variety of functionalities. Incorporating RISs in wireless networks have been recently advocated as a revolutionary
means to transform any wireless signal propagation environment to a dynamically programmable one, intended for
various networking objectives, such as coverage extension and capacity boosting, spatiotemporal focusing with
benefits in energy efficiency and secrecy, and low electromagnetic field exposure. Motivated by the recent increasing
interests in the field of RISs and the consequent pioneering concept of the RIS-enabled smart wireless environments, in
this paper, we overview and taxonomize the latest advances in RIS hardware architectures as well as the most recent
developments in the modeling of RIS unit elements and RIS-empowered wireless signal propagation. We also present a
thorough overview of the channel estimation approaches for RIS-empowered communications systems, which
constitute a prerequisite step for the optimized incorporation of RISs in future wireless networks. Finally, we discuss
the relevance of the RIS technology in the latest wireless communication standards, and highlight the current and
future standardization activities for the RIS technology and the consequent RIS-empowered wireless networking
approaches.

Key words: reconfigurable intelligent surfaces; beyond 5G; 6G; channel modeling; hardware architecture; channel
estimation; smart wireless environment; standardization
1 Introduction of Release 18 in the 3rd Generation Partnership Project
Being on the verge of the final decision on the content (3GPP), known also as the first 5th generation (5G)

• Mengnan Jian and Ruiqi Liu are with the Wireless Research Institute, ZTE Corporation, Beijing 100029, China, and also with the
State Key Laboratory of Mobile Network and Mobile Multimedia Technology, Shenzhen 518055, China. E-mail: {jian.mengnan;
richie.leo}@zte.com.cn.
• George C. Alexandropoulos is with the Department of Informatics and Telecommunications, National and Kapodistrian University
of Athens, Athens 15784, Greece. E-mail: [email protected].
• Ertugrul Basar is with the Communications Research and Innovation Laboratory (CoreLab), Department of Electrical and
Electronics Engineering, Koç University, Istanbul 34450, Turkey. E-mail: [email protected].
• Chongwen Huang is with the College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027,
China. E-mail: [email protected].
• Yuanwei Liu is with the School of Electronic Engineering and Computer Science, Queen Mary University of London, London, E1
4NS, UK. E-mail: [email protected].
• Chau Yuen is with the Engineering Product Development (EPD) Pillar, Singapore University of Technology and Design, Singapore
487372, Singapore. E-mail: [email protected].
* To whom correspondence should be addressed.
Manuscript received: 2021-12-20; revised: 2022-03-06; accepted: 2022-03-21

© All articles included in the journal are copyrighted to the ITU and TUP. This work is available under the CC BY-NC-ND 3.0 IGO license:
https://creativecommons.org/licenses/by-nc-nd/3.0/igo/.
 2 Intelligent and Converged Networks, 2022, 3(1): 1−32

advanced standard, a number of promising topics have Communication through RISs appears as a
been already discussed[1], such as the evolution of revolutionary paradigm in which the end-to-end
downlink multiple-input multiple-output (MIMO) and wireless propagation channel among end terminals can
non-terrestrial networks, as well as the inclusion of be controlled by network operators[8−18]. Software-
artificial intelligence (AI) and machine learning in controlled RISs are able to manipulate their impinging
network management[2, 3], integrated access and signals over-the-air to improve the performance of
backhaul[4, 5], duplex operation[6], and uplink/mobility target communication systems, particularly suffering
enhancements. Despite the massive roll-out of 5G from severe attenuation and blockage. In the past
networks worldwide, which provide higher flexibility
decade, there has been a growing interest in novel
and spectrum/energy efficiency compared to their 4th
communication paradigms that exploit the implicit
generation (4G) counterparts, it is inevitable that the
randomness of the propagation environment. In this
6th Generation (6G) of wireless networks, expected
context, index modulation that exploits the indices of
around 2030, and beyond will require more radical
building blocks of target communication systems to
communication paradigms, especially at the physical
convey information, appeared as an intriguing
layer and multi-access edge computing.
alternative to well-established MIMO modes and
As per the wide consensus, 5G can be regarded as an
waveforms used in standards[19]. RIS-empowered
evolution of 4G, in terms of physical-layer technologies
communications take this paradigm one step beyond,
adopted, while a true revolution can be realized by 6G
by enabling Internet-of-everything/Internet-of-sensing by exploiting the wireless channel itself, not only to
applications. According to the recent white paper convey information as in index modulation schemes,
released by one of the major industrial bodies[7], some but also to boost various communication objectives,
dominant applications of 6G will be truly immersive such as the error performance, capacity, secrecy,
virtual reality, mobile holography, and digital replica. outage, and energy efficiency. In other words, smart
From the perspective of communication engineering, radio environments enabled by software-defined RISs
these science-fiction-type applications will require user can be regarded as one-step beyond software-defined
experienced and peak data rates of 1 Gbps and wireless communication networks.
10 Gbps, which are 10- and 50-fold higher than that of An RIS is a specifically designed man-made surface
5G, respectively, along with substantially higher of electro-magnetic (EM) material that is electronically
spectrum and energy efficiency. Despite the significant controlled with integrated electronics and has unnatural
efforts and developments in the physical layer, it is wireless communication capabilities. In the most
inevitable that we will need to resort more radical commonly considered case, the large number of small-
communication paradigms to level up these major key sized, low-cost, and almost passive elements that
performance indicators. Similar to the standardization comprise an RIS can simply modify the incident
of 5G which had long past, promising candidate signal over-the-air to improve the signal coverage
technologies are raising to satisfy the challenging and/or quality[20]. This technology is different from
future requirements of 6G wireless networks, including existing massive MIMO transceivers[21] and active
TeraHertz (THz)/visible light communications, AI- beamformers[22], as well as the cooperative
empowered communications, wireless power transfer, relaying[23−25] and backscatter communication[26]
integrated non-terrestrial networks, cell-free massive paradigms. The most distinguishable property of RISs
MIMO systems, index modulation, advanced is their cheap and nearly passive panel of unit cells,
waveforms, and smart radio environments each being capable of applying a controllable phase
enabled/assisted by reconfigurable intelligent surfaces shift to the impinging field (e.g., via two-state
(RISs). reflection coefficients in one of the first RIS designs
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 3

in Ref. [27]). When used as pure tunable reflectors, states or how to perform channel estimation in
RISs do not require radio frequency (RF) signal highly dynamic environments is quite challenging.
processing, amplification, filtering, mixing, or Unfortunately, nearly passive reflective RIS
up/down-conversion. In light of these, RIS-empowered architectures like Refs. [27, 66] cannot perform any
communication can be regarded as an alternative/ form of parameter estimation at their side, and the
complementary technology to massive MIMO and problem becomes the estimation of the cascaded
active relaying systems. channel between the transmitter and the receiver
Remarkable developments in the field of RIS- through the RIS. Furthermore, the increasing size of
empowered communications have been witnessed in the RIS increases the channel estimation burden and
the past 3 years, and the underlying RIS technology has dedicated channel-bearing signaling might be required
become mature enough for both academia and industry between the network controller and the RIS to adjust its
to explore its promising future use-cases. Particularly, configuration in real-time. At this point, standalone
within the context of RIS wireless systems, researchers RIS architectures equipped with AI modules might
have focused on joint active and passive beamforming provide promising solutions to configure the RIS on
problems[9, 12, 28], channel modeling under different the go along with efficient channel estimation
conditions[29−33] and performance analyses[34−36], protocols, which have low signaling overhead for RIS-
physical layer security[37−39], non-orthogonal multiple empowered systems (e.g., Ref. [67]).
access schemes (NOMA)[40−42], non-coherent The last years, there has been increasing interest in
modulation [43, 44] , vehicular networks [45, 46], hybrid
transceiver hardware architectures (see, e.g.,
passive and sensing designs[47−49], active and joint Refs. [13, 22] and references therein), and researchers
reflective-transmissive designs[50, 51], deep learning have explored effective schemes to provide
solutions[52−56], reflection modulation schemes[57−59], reconfigurability, such as analog beamforming
real-word experiments with prototyping[30, 60, 61], along networks, large arrays of inexpensive antennas
with many other interesting application areas. equipped with external phase shifters, reconfigurable
As in any other communication technology, channel loads with varactor diodes, or even reconfigurable
modeling is an integral part of the RIS research to
reflecting elements with PIN diodes. Even more
understand the most convincing use cases of RISs in
recently[68], active RIS systems including a single
future wireless networks. Within this perspective, one
power amplifier have been proposed for offering
has to not only consider the physical propagation
reflection amplification. Their aim is to challenge the
effects in different environments, such as different path
double path loss effect of passive RIS systems, and
loss and scattering behavior of large indoor office or
promising results were reported in terms of both
street canyon environments, but also the effect of the
achievable rate performance and energy efficiency.
RIS on the end-to-end channel model, such as near-
field effects for RISs in close proximity and potential Considering the higher cost, complexity, and power
spatial correlation[62, 63], as well as mutual coupling[64, 65] consumption of these active RIS architectures,
among RIS elements. The architecture of the RIS, such interesting new problems arise in terms of system
as reflect array based or metasurface-based, plays an optimization and hybrid RIS system design.
important role to the underlying channel models. Furthermore, practical analyses on mutual coupling as
Moreover, modified line-of-sight (LOS) probabilities well as the correlation between reflection amplitudes
as well as path loss exponents can be derived and phases are gaining momentum.
considering potential locations of RISs in practical In line with the latest rich research and development,
wireless environments. as well as the ongoing booming interests in RIS-
When considering smart wireless environments empowered wireless networking, this survey article
with almost passive RISs, the knowledge of channel presents various up-to-date advances in the RIS
 4 Intelligent and Converged Networks, 2022, 3(1): 1−32

technology, focusing particularly on the taxonomy of another note, active RIS systems comprise a natural
the available RIS hardware architectures and their evolution of conventional massive MIMO systems, by
operation modes, the diverse channel modeling packing more and more software-controlled antenna
approaches including multi-user NOMA systems, as elements onto a two-dimensional (2D) surface of finite
well as the plethora of the channel acquisition size[71]. In Ref. [72], the authors named an active RIS
techniques. In particular, in Section 2, we present a as a large intelligent surface (LIS), where the element
taxonomy of the available RIS hardware designs and spacing reduces when the element number increases. A
their available implementations, while Section 3 compact integration of a large number of tiny active
includes the overview of the up-to-date channel RIS elements with reconfigurable processing networks
models for wireless communications empowered by can realize a practical continuous antenna aperture. The
RISs. Section 4 discusses the available channel active RIS structure can be used to transmit and receive
estimation approaches in RIS-empowered wireless signals across the entire surface utilizing the hologram
communication systems, while the relevance of the RIS principle[69, 70]. The discrete photonic antenna array is
technology in the latest standards together with the another practical implementation of active RISs. It
topic’s road ahead are discussed in Section 5. The integrates active optical-electrical detectors, converters,
concluding remarks of the paper are included in and modulators for performing transmission, reception,
Section 6. and conversion of optical or RF signals[69]. In
Ref. [73], RISs consisting of passive RIS elements and
2 RIS hardware architectures a single receiving RF chain for baseband measurements
In this section, we present the up-to-date RIS hardware were presented. Very recently in Ref. [74], an active
architectures and some indicative fabrication RIS structure where each reflecting element is
methodologies. Compared with the currently used equipped with a power amplifier was proposed.
transceiver technologies in wireless networks, the most (2) Passive RISs: Passive RIS acts like a passive
distinctive characteristics of the RIS concept lie in metal mirror or wave collector which can be
making the environment controllable by providing the programmed to change an impinging EM field in a
possibility of fully shaping and controlling the EM customizable way[9, 11]. Compared with its active
response of the environmental objects that are counterpart, a passive RIS is usually composed of low-
distributed throughout the network. An RIS structure is cost and almost passive elements that do not require
usually intended to operate as a signal source or wave dedicated power sources. Their circuitry and embedded
collector with reconfigurable characteristics, especially sensors can be powered with energy harvesting
for application scenarios where it is used as a passive modules, an approach that has the potential of making
reflector with the objective of improving the them truly energy neutral. Regardless of their specific
communication performance. Existing RISs can be implementations, what makes the passive RIS
roughly categorized as contiguous and discrete RISs, technology attractive from an energy consumption
while on the basis of power consumption, RIS also can standpoint is their capability to shape radio waves
be classified into the active and passive types. We also impinging upon them, forwarding the incoming signal
discuss the different operation modes that an RIS can without employing any power amplifier nor RF chain,
be used, identifying the hardware requirements per and even without applying sophisticated signal
distinctive mode. processing. Moreover, passive RISs can work in full
duplex mode without significant self interference or
2.1 Hardware designs increased noise level, and require only low-rate control
(1) Active RISs: The term active RIS is adopted when link or backhaul connections. Finally, passive RIS
energy-intensive RF circuits and consecutive signal structures can be easily integrated into the wireless
processing units are embedded in the surface[69, 70]. On communication environment, since their extremely low
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 5

power consumption and hardware costs allow them to to reconfigure the reflection characteristics of
be deployed into building facades, room and factory its elements, enabling programmable manipulation
ceilings, laptop cases, or even human clothing[9, 11]. of incoming EM waves in a wide variety of
(3) Discrete RISs: A discrete holographic multiple functionalities. It is essential to achieve a fine-grained
input multiple output surface (HMIMOS) is usually control over the reflected EM field for quasi-free
composed of many discrete unit cells made of low- space beam manipulation so as to realize accurate
power and software-tunable metamaterials. The means beamforming. Meta-atoms of sub-wavelength size are a
to electronically modify the EM properties of the unit favorable choice, although inevitable strong mutual
cells range from off the shelves electronic components coupling and well-defined gray-scale-tunable EM
to using liquid crystals, microelectromechanical properties exist.
systems or even electromechanical switches, and other Conversely, in rich scattering environments[15], the
reconfigurable metamaterials. This structure is wave energy is statistically equally spread throughout
substantially different from the conventional MIMO the wireless medium. The ensuing ray chaos implies
antenna array. One embodiment of a discrete surface is that rays impact the RIS from all possible, rather than
based on discrete “meta-atoms” with electronically one well-defined, directions. The goal becomes the
steerable reflection properties[8]. As mentioned earlier, manipulation of as many ray paths as possible, which is
another type of discrete surface is the active one based different from the common goal of creating a directive
on photonic antenna arrays. beam. This manipulation has two kinds of aims,
(4) Continuous RISs: Integrating a virtually infinite including tailoring those rays to create constructive
number of elements into a limited surface area, a interference at a target location and stirring the field
contiguous RIS can thus form a spatially continuous efficiently. These manipulations can be efficiently
transceiver aperture[69, 70]. Compared with discrete realized with RISs equipped with half-wavelength-
RISs, contiguous RISs have some essential differences sized meta-atoms, enabling the control of more rays
from the perspectives of implementation and hardware, with a fixed amount of electronic components (PIN
as will be described in the sequel in the fabrication diodes, etc.).
methodologies subsection. The comparison between (2) Metasurface-based antennas: As an efficient
discrete RISs and continuous RISs is illustrated in realization of extreme massive antenna arrays, dynamic
Fig. 1. metasurface antennas (DMAs) have been recently
proposed[71] to pave the way. DMAs have beam
2.2 Operation modes
tailoring capabilities and facilitate processing of the
(1) Reflecting RISs: The concept of the RIS- transmitted and received signals in the analog domain.
empowered smart wireless environments[8, 9, 11] initially DMAs work in a dynamically configurable manner
considered only passive RISs with almost zero power with simplified transceiver hardware. Additionally,
consumption unit elements[13]. Their envisioned compared with conventional antenna arrays, DMA-
prominent role lies on the capability of the surface based architectures require much less power and cost.
Interface and
Continuous surface Discrete surface communication layer
Metallic patches Computing layer
Shielding layer
I I I I Sensing and actuation
I
C C C C C Metamaterial layer
S A S A S
C C C C C
S A S A S
Varactor S A S A S
Continuous strip

(a) Continuous RIS (b) Discrete RIS

Fig. 1 Schematic view of RIS hardware structures[20], including (a) continuous and (b) discrete implementations.
 6 Intelligent and Converged Networks, 2022, 3(1): 1−32

In this way, eliminating the need for complicated completion based channel estimation technique with
corporate feed and/or active phase shifters becomes relatively short training requirements, requiring much
possible. Another promising advantage of DMAs is fewer receiving RF chains than RIS elements.
that they can comprise massive numbers of tunable The receiving RIS hardware architecture, which
metamaterial-based antenna elements fitting into small is illustrated in Fig. 3, connects the outputs of the
physical areas and providing wide range of operating RIS elements to a single reception RF chain. The
frequencies. hardware includes a low noise amplifier, a mixer
In Ref. [71], a DMA architecture that consists of downconverting the signal from RF to baseband, and
multiple separate waveguide-fed element arrays with an analog-to-digital converter. Each impinging EM
each connected to a single input/output port was training signal at the meta-elements is received in
discussed. A large number of radiating elements can be the RF domain with one of the M available RIS
accommodated in waveguides, and the sub-wavelength configurations included in the random sampling unit.
spaced character allows each input/output port to feed a This configuration is selected by random spatial
multitude of possibly coupled radiators. For 2D sampling, which is essential for efficient channel
waveguides, a scattered wave from each element reconstruction[76].
propagates in all directions. Since the proposed (4) Simultaneously reflecting and sensing RISs: A
waveguide is typically designed to be single mode and hybrid meta-atom, which can simultaneously reflect a
the wave can only propagate along one line, its analysis portion of the impinging signal in a programmable way
is much easier than 2D waveguides. Furthermore, while another portion of it can be fed to a sensing unit,
ensuring isolation between different ports is easier in has been recently designed in Ref. [47]. The so-called
1D waveguides than in multiple ports of a 2D hybrid RISs are realized by adding a waveguide to
waveguide. A common implementation of 1D couple to each meta-atom. Each waveguide can be
waveguides is based on microstrips, as depicted in connected to an RF chain. This makes it possible to
Fig. 2. locally process a portion of the received signals in the
(3) Receiving RISs: The DMA principle was also digital domain. However, the coupling of the RIS
conceptualized in the parallel work[73] for enabling elements to the waveguides makes it impossible for the
RISs with baseband reception capability in order to incident wave to be perfectly reflected. Actually, the
perform explicit or implicit channel estimation with coupling level determines the ratio of the reflected
minimal number of receiving RF chains. In fact, Ref. [75] energy to the absorbed energy. Its footprint can be
presented a signal direction estimation approach with a reduced and the coupling to the sampling waveguide
single RF chain, while Ref. [73] designed a matrix can be mitigated by keeping this waveguide near
Observed
signals

Metamaterial Output
elements port

Propagation inside
the waveguide

Fig. 2 Schematic illustration of a microstrip for the DMA design in Ref. [71]. The microstrip’s output port is connected to a
receiving RF chain for baseband processing. The port can be alternatively fed by the output of a transmit RF chain to realize
analog transmit beamforming.
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 7

Random spatial (5) Simultaneously transmitting and reflecting

sampling from
the M available Online RISs: The concept of simultaneously transmitting
RIS configurations configuration
controller and reflecting RISs (STAR-RISs) was recently
presented in Ref. [77]. This new type of RIS variant is
Baseband
also referred to as the intelligent omni-surface
Reception (IOS)[78], allowing wireless signals incident on the
RF chain
surface to be simultaneously reflected and transmitted.
Fig. 3 Block diagram of the receiving RIS hardware
STAR-RISs can assist in achieving a full-space
architecture in Ref. [73] including a single active reception
RF chain for explicit channel estimation at the RIS side. reconfigurable wireless environment that has a 360-
degree coverage[79]. In Fig. 5, the schematic illustration
cutoff. Such hybrid RISs were recently considered in of the STAR-RIS element is presented, where the
Refs. [48, 49] for facilitating explicit and implicit tunable surface impedance can be configured to
channel estimation as well as flexible network adjust the intensities and distributions of the surface
management with reduced overhead compared to electric and magnetic currents. This, in principle,
cascaded channel estimation with passive RISs. enables an independent control of both the reflecting
In Fig. 4, the designed hybrid RIS (HRIS) of phase ϕR and the transmission phase ϕT appearing at an
Ref. [47] is schematically illustrated. It consists of STAR-RIS.
mushroom structures, each loaded with a varactor The key to achieving tunability for both the reflected
diode. This reconfigurable capacitance results in a and transmitted (refracted) signals is that each element
reconfigurable resonance frequency, and consequently, has to support both electric and magnetic currents. As
a reconfigurable effective impedance. This element is reported in Ref. [80], several prototypes have been
selected since it provides a simple mechanism to implemented which support simultaneous reflection
realize high reflectivity with reconfigurable phases. and transmission, e.g., the transparent dynamic
To address each meta-atom independently, as required metasurface designed by researchers at NTT
for forming desired refection patterns, the via of the DOCOMO[81] and the PIN diode empowered by the
mushroom structure extends through the bottom STAR-IOS prototype[82, 83]. Similar to the approach of
conductive plate. An annular slot separates the via from studying conventional reflecting RISs, the hardware
the ground plane beneath the substrate. This annular modeling of STAR-RISs comes with additional
slot allows for coupling the incident wave to another challenges, because transmitted (refracted) signals need
layer. In previous designs, an RF choke (in the form of to be considered. The authors of Ref. [77] studied a
a radial stub) was placed at the end of this via to general hardware model and near/far-field channel
diminish RF coupling. In Ref. [47], however, this models for STAR-RISs, where the reflection and
annular slot was exploited to sense the incident wave. transmission coefficients of the unit elements were
This sensing information can be then used for various assumed to be independently adjustable. However, in
network management tasks. practical cases, the unit elements at STAR-RISs have
Diode
2.6 mm Side view
4 mm

0.5 mm
2 mm h
4 mm
h=1.5 mm

Coupled
incident signal

Via

Fig. 4 Portion of the hybrid RIS design in Ref. [47] with integrated sensing capability.
 8 Intelligent and Converged Networks, 2022, 3(1): 1−32

Reflected and nanoimprint lithography, as well as direct laser

field writing or printed circuit board processes at
ϕ0

Magnetic
microwaves. Usually, these fabrication techniques will
Incident field current
ϕ0+ϕR be ascribed to produce two typical apertures,
Electric current
continuous or discrete apertures. In approximately
continuous microwave aperture, the meta-particle
ϕ0+ϕT
structure uses the varactor loading technique to
broaden its frequency response range, and achieves
Tunable surface
impedances Transmitted controllable reflection phase. It is a continuous
field monolayer metallic structure, and comprises a large
number of meta-particles. Each meta-particle contains
Fig. 5 Schematic illustration of the STAR-RIS element,
where ϕ0, ϕR, and ϕT denote the phases of the incident, two metallic trapezoid patches, a central continuous
reflected, and transmitted signals, respectively. strip, and varactor diodes. By independently and
continuously controlling the bias voltage of the
to be passive-lossless to maintain good scalability.
varactors, the surface impedance of continuous RISs
Therefore, the authors of Refs. [84, 85] considered the
can be dynamically programmed, and thus manipulate
constraints that should be imposed on the reflection
the reflection phase, amplitude states, and the phase
and transmission coefficients of a given passive STAR-
distribution over a wide range of frequency bands[87]. It
RIS element.
should be highlighted that this impedance pattern is a
(6) Amplifying RISs: An amplifying RIS design that
map of the hologram, and can be calculated directly
incorporates a single power amplifier (PA), enabling
from the field distribution of the provided reference
reflection amplification, was recently proposed in
wave and reflected object wave.
Ref. [68]. Its motivation lies on the double path loss
In contrast to continuous apertures, another instance
problem and the lossy reflection with passive reflective
RISs. In fact, only when the passive RISs are of RISs is a realization based on discrete apertures that
positioned close to terminals can they perform are usually realized with software-defined metasurface
efficiently. Otherwise, they cannot go beyond being antennas. A general logical structure was proposed in
a supportive technology for future wireless Ref. [8]. Its general unit cell structure contains a
communications systems. According to this very recent metamaterial layer, sensing and actuation layers,
amplifying RIS design, the impinging signal is shielding layer, computing layer, as well as an interface
received by a portion of the RIS. Then, the received and communications layer with different objectives.
EM field is phase configured and fed to the single Specifically, the meta-material layer is implemented by
power amplifier, who in turns feeds it to the remaining graphene materials for delivering the desired EM
portion of the RIS that reflects it with the controllable behavior through a reconfigurable pattern, while the
phase configuration. The amplifying design works objective of sensing and actuation layer is to modify
solely in the RF domain similar to the waveguide-
the behavior of the meta-material layer. The shielding
based approach in Ref. [47]. This is different from the
layer is made of a simple metallic layer for decoupling
full-duplex multi-antenna decode-and-forward relays,
the EM behavior of the top and bottom layers to avoid
as down-conversion and baseband processing is
mutual interference. The computing layer is used to
performed[86].
execute external commands from the interface layer or
2.3 Fabrication methodologies sensors. Finally, the interface and communications
There are various fabrication techniques for RISs layer aim at coordinating the actions of the computing
including electron beam lithography at optical layer and updating other external wireless entities via
frequencies, focused-ion beam milling, interference the reconfigurable interface.
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 9

While the development of RISs is in its infancy, paths available when the LOS component (or another
basic prototyping work on different kinds of this strong path in case of non-LOS propagation) is
technology is available already. Various discrete RISs blocked. However, in real situations such as in street
have been developed by the start-up company canyon environments, all propagation paths are within
“Greenerwave”, which shows the basic feasibility and the same beam and are blocked simultaneously, leading
effectiveness of the RIS concept using discrete to a significant performance decrease[93]. The
metasurface antennas. In contrast, another start-up increasing demand for higher frequency bands has
company named as “Pivotalcommware”, with the recently stimulated the proposal of new models for
investment of Bill Gates, is developing initial above 100 GHz[94], providing first results on angular
commercial products of a contiguous RIS based on characterization of the sub-THz channel[95]. On the
low-cost contiguous metasurfaces; this further verifies scenario side, going beyond the classical enhanced
the feasibility of the RIS concept as well as the mobile broadband, the industrial environment
advancement of holographic technologies. The ZTE motivated new studies were also addressed by 3GPP.
Corporation, a network infrastructure vendor, also Very recently, there exists increasing research interest
demonstrated recently a series of RIS panels which are in propagation channel models that incorporate the
dual-polarized and capable of 2-bit phase control, change of the wireless environment itself via the
empowered by PIN diodes as well as liquid inclusion of RISs. In the following, we overview the
crystals[88, 89]. Continuing prototyping development is key latest advances in channel models for wireless
highly desired to prove the RIS concept with brand communications empowered by RISs.
new holographic beamforming technologies, Dating back to 1992 to the concept of time
potentially unveiling new technical issues which will reversal[96] in conjunction with its very recent
require further research efforts. In addition, one of the feasibility demonstrations in high frequency bands with
main goals of the ongoing European Union H2020 large available transmission bandwidths[97−99], as well
RISE-6G project[90], which is the first collaborative as the recent interests for RIS-empowered smart
project planning to design, prototype, and trial radical wireless environments[8−18], it becomes apparent that
technological advances based on RISs to forge a new controlling waves in complex media is a major topic of
generation of dynamically programmable wireless interest. In Ref. [100], the authors experimented with
propagation environments[16, 17], is to provide low- an RIS inside a cavity, emulating a rich scattering
power and cost-effective RIS hardware designs environment, and showcased that by maximizing the
operating from below 6 GHz to millimeter wave Green’s function between two antennas at a chosen
(mmWave) and sub-THz frequencies. time, yields diffraction-limited spatiotemporal
focusing. In addition, by changing the photons’ dwell
3 RIS-empowered signal propagation
time inside the designed cavity, the relative distribution
modeling
of the spatial and temporal degrees of freedom were
The accurate modeling of wireless channels is one of modified. Such RIS-enabled tunable rich scattering
the most critical factors enabling wireless technologies, conditions were recently considered in Ref. [15] for
and serves as a core means for validating them before analog multipath shaping. In particular, the RIS was
and in conjunction with field trials. As such, it has deployed to create pulse-like channel impulse
attracted great interest in the scientific community of responses that can maximize the channel capacity. In
wireless communications in recent years. From the Ref. [56], the authors used a coupled dipole based
viewpoint of characterization, many measurement simulator, which faithfully models the underlying wave
campaigns[91, 92] have studied wireless channels from a physics in RIS-empowered communications inside
static point of view. From these measurement cavities, to simulate fast fading conditions in rich
campaigns, we learned that there are some propagation scattering environments.
 10 Intelligent and Converged Networks, 2022, 3(1): 1−32

A stochastic cascaded model for RIS-empowered end-to-end channel model including frequency
wireless communications was initially presented in selectivity for RIS-enabled wireless communications
Refs. [9, 12], where the impact of the RIS in the signal was presented in Ref. [105], together with a framework
propagation was encapsulated through the reflection for the potential of RISs for performing equalization. It
coefficients of its state-tunable unit elements. In this has been theoretically shown that the RIS can be
generic model which has been extensively adopted up adjusted according to the incoming signals to maximize
to date, typical channel models can be incorporated the magnitude of the first channel tap, making the ISI
(e.g., Rayleigh, Ricean, and Nakagami-m) and the RIS terms negligible, hence, maximizing the achievable rate
is mainly assumed to be capable of controlling the performance. In a similar manner, RISs have been used
phases of its unit-amplitude reflection coefficients. in Ref. [46] to confront the Doppler effect in mobile
Considering realistic circuitry for resonant RIS RIS-empowered communications.
elements and capitalizing on the transmission line By partitioning the RIS units into several subsets,
model, Ref. [101] presented a practical model for its referred to as tiles, a physics-based model was
RIS-induced reflection coefficient, where the phase presented in Ref. [106] that also incorporated
configuration depends on the amplitude. Very recently polarization. The impact of each tile on the wireless
in Ref. [102], the authors considered a transfer function channel is modeled using concepts from the radar
approach and introduced a physics-compliant model literature, according to which it is an anomalous
for the Lorentzian frequency response of the RIS reflector. In this case, for a given phase shift, the tile’s
elements, which captures the parameters one can impact can be derived by solving the corresponding
externally control to modify the surface’s reflection integral equations for the electric and magnetic vector
profile. Based on this RIS frequency response model, fields. Using this model, each RIS tile is then
an optimization framework for setting the RIS optimized in two stages. In the offline stage, the RIS
controllable parameters in order to maximize the units of each tile are jointly designed to support
achievable average sum-rate in the uplink of wideband different transmission modes, with every transmission
RIS-empowered multi-user MIMO systems with mode effectively corresponding to a given
orthogonal frequency division multiplexing (OFDM) configuration of the phase shifts that the units of the
signaling was presented. A multipath signal tile apply to an incident EM wave. In the online
propagation channel model based on the cascaded optimization stage, the best transmission mode of each
channel model was considered in Ref. [103], where the tile is selected such that a desired quality of service
authors presented an optimization framework for RIS- (QoS) criterion is maximized.
enabled equalization, aiming to reduce the inter-symbol A channel gain expression for planar arrays of
interference (ISI). arbitrary sizes, taking the varying distances to the
A two-path propagation model was proposed in antenna elements, polarization mismatches, and the
Ref. [104] for RIS-assisted wireless networks by effective areas into account, was presented in Ref. [29]
considering both the direct path from the transmitter to to investigate the spectral efficiency behavior and
the receiver and the assisted path established by the asymptotic power scaling laws in RIS-empowered
RIS. The proposed propagation model unveiled that the communications. It was proved for this deterministic
phase shifts of RISs can be optimized by appropriate propagation model (i.e., free-space LOS) that an RIS
configuration for multipath fading mitigation. In cannot achieve a higher signal-to-noise ratio (SNR)
particular, four types of RISs with different than MIMO setups when the array sizes are equal,
configuration capabilities were introduced and their despite the fact that the SNR in the RIS setup grows
performances on improving received signal power in with the square of its number of elements in the far-
virtue of the assisted path to resist fast fading were field. A closed-form expression showcasing how large
compared through extensive simulation results. An an RIS must be to beat conventional massive MIMO or
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 11

half-duplex massive MIMO relaying was derived. the desired surface transformation. Closed-form
Using physical optics techniques in Ref. [107], a far- expressions are obtained in two asymptotic regimes
field pathloss model was presented to explain why an that are representative of far-field and near-field
RIS, consisting of many elements that individually act deployments. A summary of the available path loss
as diffuse scatterers, can beamform the signal in a models was recently provided in Ref. [110] with
desired direction with a certain beamwidth. some of them being experimentally validated inside
In Ref. [108], an analytical channel model for point- an anechoic chamber in Ref. [30]. Very recently,
to-point RIS-assisted free-space optical (FSO) systems pathloss models considering THz frequencies are
was developed, which is based on the Huygens-Fresnel presented[111, 112]. In Ref. [112], a near-field channel
principle for the intermediate and the far-field. The model was proposed that accounts for the size of the
model determines the reflected electric field and RIS elements in the path loss calculation, as well as in
captures the impact of the size, position, and the spherical wavefront of the radiated waves. It is
orientation of the RIS, as well as its phase shift profile shown that a typical THz RIS is likely to operate in the
on the end-to-end channel. The accuracy of the Fresnel zone, where conventional beamforming is
proposed analytical model was validated via suboptimal and hence can reduce the power gain.
simulations and it was shown that, in contrast to Capitalizing on the circuit theory of
models based on the far-field approximation, it is valid communications [64] and its applications in tunable
even for intermediate distances, which are relevant in impedance systems (e.g., in beamforming with single-
practice. The model was exploited to devise RIS RF electronically steerable parasitic array radiators[65]),
sharing protocols in multi-link FSO systems, whose an end-to-end mutual-coupling-aware channel model
performance was analytically investigated. for free-space signal propagation empowered by an
Considering an RIS modeled as a sheet of EM RIS was presented in Ref. [113]. The model is based
material of negligible thickness and leveraging the on the mutual impedances between all existing
general scalar theory of diffraction as well as the radiating elements, i.e., the transmit/receive antennas
Huygens-Fresnel principle in Ref. [109], closed-form and the passive scatterers which are all modeled as
expressions for the power reflected from an RIS were dipoles. It yields a one-to-one mapping between the
presented as functions of the size of the RIS, the voltages fed into the ports of a transmitter and the
distance between the transmitter/receiver and the RIS, voltages measured at the ports of a receiver, while
and the phase shift matrix configured by the RIS. With accounting for the generation and propagation of the
the aid of the stationary phase method, the authors EM fields, the mutual coupling among the sub-
identified sufficient conditions under which an RIS acts wavelength unit cells of the RIS, and the interplay
as an anomalous mirror, indicating that the received between the amplitude and phase response of the unit
power decays as a function of the reciprocal of the sum cells of the RIS.
of the distances between the transmitter/receiver and Very recently in Ref. [33], the authors introduced a
the RIS. It is shown that in short distances, an RIS acts physics-based end-to-end model of RIS-parametrized
as an anomalous mirror, and in the far distance regime wireless channels with adjustable fading, which is
acts as a scatterer. Based on the vector generalization based on a first-principles coupled-dipole formalism.
of the Green’s theorem, a free-space pathloss model for The proposed model, whose MATLAB source code
homogenized RISs, made of sub-wavelength scattering has been made available online, incorporates the
elements and being capable to operate either in notions of space and causality, dispersion (i.e.,
reflection or transmission mode, is introduced in Ref. frequency selectivity) and the intertwinement of each
[32]. The model is formulated in terms of a computable RIS element’s phase and amplitude response, as well as
integral that depends on the transmission distances, the any arising mutual coupling effects including long-
polarization of the radio waves, the size of the RIS, and range mesoscopic correlations. This model has been
 12 Intelligent and Converged Networks, 2022, 3(1): 1−32

used for simulating Ricean fading channel and was random matrix and replica theories. Under mild
considered for a prototypical problem of RIS-enabled assumptions, asymptotic expressions for the
over-the-air channel equalization in rich-scattering eigenvalues and the eigenvectors of the channel
wireless communications. covariance matrices were derived. It was found that,
In Ref. [31], the authors capitalized on the 5G when the channel close to an RIS is correlated (e.g., for
mmWave channel model with random number of instance due to small angle spread), the communication
clusters/scatterers[114] and extended it to include an link benefits significantly from the RIS optimization,
RIS. The proposed model is valid for RIS-empowered resulting in gains that are surprisingly higher than the
narrowband communications in indoor and outdoor nearly uncorrelated case. Furthermore, when the
environments and includes many physical desired reflection from the RIS departs significantly
characteristics, such as LOS probability, shadowing from geometrical optics, the surface can be optimized
effects, and shared clusters. In addition, the model to provide robust communication links. Building on the
incorporates realistic gains and array responses for RIS properties of the eigenvectors of the covariance
elements in addition to the existing channel models. matrices, the authors of Ref. [36] were able to find the
Based on this model, an open-source channel simulator optimal response of the RISs in closed form, bypassing
MATLAB package has been finalized, which can be the need for brute-force optimization.
used in channel modeling of RIS-based systems with In Ref. [118], the authors considered three-
tunable operating frequency, terminal locations, dimensional small-scale fading in the far-field and
number of RIS elements, and environments. assumed that it can be modeled as a zero-mean
In Ref. [115], the authors proved that channel fading spatially-stationary and correlated Gaussian scalar
in RIS-empowered communications will always be random field satisfying the Helmholtz equation in the
spatially correlated[62, 115−117] thus discouraging the frequency domain. This model yields a physically
common adoption of the independent and identically meaningful spatial correlation function, whose power
distributed Rayleigh fading model. It was shown spectral density (in the spatial-frequency or
that the asymptotic SNR limit is equal, but the wavenumber domain) is impulsive with support on the
convergence rate and rank of the spatial correlation surface of a sphere of radius κ = 2π/λ (with λ being the
matrices are different. A channel model for isotropic wavelength), and can be uniquely described by a
scattering was presented, whose properties were spectral factor that specifies directionality and
characterized, including the rank and channel physically characterizes the propagation environment
hardening. The derived channel properties were also in its most general form. This structure of the spatial
applied for massive MIMO arrays with the same form correlation function leads to the two-dimensional
factor, called holographic MIMO[20, 72]. Kronecker- Fourier plane-wave spectral representation for the
product covariance matrices for the incoming and field, which is given by a superposition of a continuum
outgoing channels to/from RISs were recently of plane-waves having zero-mean statistically-
proposed in Ref. [36], which relate the field’s mean independent Gaussian-distributed random amplitudes.
direction of arrival/departure and its angle spread It is shown that the small-scale fading has a singularly-
with the spatial correlation in planar arrays. The integrable band-limited spectrum in the wavenumber
presented model enables the investigation of the role of domain that is defined on a disk of radius κ . This is a
the channel conditions near RISs in wireless direct result of the Helmholtz equation, which acts as a
communications empowered by multiple RISs. Based two-dimensional linear space-invariant physical filter
on this model, an asymptotic closed-form expression that projects the number of observable field
was derived for the mutual information of a multi- configurations to a lower-dimensional space. The band-
antenna transmitter-receiver pair in the presence of limited nature of the two-dimensional Fourier plane-
multiple RISs, in the large-antenna limit, using the wave spectral representation is exploited to statistically
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 13

characterize the small-scale fading over holographic the best- and worst-case of channel statistics for
MIMO arrays of compact size. Recently, the authors in characterizing the effective channel gains for RIS-
Ref. [119] extended the results of Ref. [118] to aided NOMA networks, which can be used as upper
downlink multi-user communications, by modeling the and lower bounds. Additionally, Ref. [122]
communication channel in the wavenumber domain investigated channel models for both ideal and non-
using the Fourier plane wave representation. Based on ideal RISs. For further evaluation, Ref. [123] estimated
this channel model, maximum-ratio transmission and the diversity gains of the RIS-aided channels. Focusing
zero-forcing pre-coding schemes are devised on RIS channel modeling and optimization algorithms,
capitalizing on the sampled channel variance, which the authors in Ref. [124] considered RISs to deal with
depends on the number and spacing of the patch multiple blocked users. In Refs. [125, 126], the authors
antennas in the holographic MIMO system. Analysis of proposed an NOMA solution with RIS partitioning to
spectral efficiency is also presented. enhance the spectrum efficiency.
The channel characteristics in multi-user RIS- Approximate distributions for the channel statistics
empowered NOMA communication systems has lately in multi-cell RIS-empowered NOMA networks have
attracted significant research efforts[120]. For the easy been recently presented. In Ref. [127], the authors
understanding of the readers, a simplified NOMA considered RISs as linear materials and introduced a
system with two users is illustrated in Fig. 6. In path loss model for the RIS reflected channels. In
Ref. [41], a basic channel model was presented, where addition, based on a Poisson cluster process model, the
the RIS was deployed in order to ensure that additional reflection angle distributions were evaluated and
cell-edge users can be served via spatial division further used to study spatial effects. Under Rayleigh
multiple access beams, by aligning the cell-edge users’ fading channels, the curve fitting channel model for
effective channel vectors with the predetermined RISs has been investigated by Ref. [128] to mimic the
spatial directions. In Ref. [121], the authors presented a integrated RIS channels as a Gamma distribution. The
model where the signals reflected through the scatters authors in Ref. [129] exploited the central limit
and the RIS elements are jointly studied as multipath theorem to evaluate the channel models and expanded
components of the overall received envelope. the channel model approach to other distributions, such
Approximate distributions of the RIS-empowered as Nakagami-m channels, Rician channels, and Weibull
channels were derived for both continuous and discrete channels with implications for their multivariate
phase shift scenarios. The authors in Ref. [42] derived versions[62, 115−117].

Base station
RIS
The m-th user Um
User m in a NOMA pair
detection
The n-th user Un
Subtract user in a NOMA pair
SIC

m signal

User n
The direct
detection links to Um

BS The direct
links to Un
User n
The BS-RIS
A two-user
User m links
NOMA
detection
cluster User m The RIS-User
links

Fig. 6 Illustration of a two-user RIS-empowered NOMA system.

 14 Intelligent and Converged Networks, 2022, 3(1): 1−32

4 Channel estimation in RIS-empowered of both the BS-RIS channel and the RIS-UE channel is
systems relatively short, where rapid estimation is needed for
both channels.
It is indispensable to obtain accurate channel state
There are many literatures investigating the
information (CSI) to achieve intelligent control of the
framework design of CE for RIS-empowered wireless
electromagnetic environment. The channel estimation
systems. In Ref. [132], a general framework for
(CE) for RIS involves the estimation of multiple
cascaded channel estimation in RIS-empowered MIMO
channels simultaneously: the direct channel between
systems was introduced by leveraging combined
the base station (BS) and each user equipment (UE),
bilinear sparse matrix factorization and matrix
the channel between the RIS and BS, and the channel
completion. A control protocol enabling linear square
between the RIS and each UE[130, 131]. Moreover, RIS estimation for the involved channel vectors that
channel estimation faces several challenges, including activates a part of RIS elements in the MISO case was
the significantly increased channel estimation presented in Ref. [133]. In Ref. [134], a two-timescale
dimension, cascaded channel measurement results, channel estimation schedule was introduced which
and the near-field channel estimation problem. This utilizes the semi-static characteristic of the BS-RIS
task becomes more complex when the deployed channel. It estimates the high-dimensional BS-RIS
RISs are equipped with massive elements with non- channel in a relatively long period and performs
linear hardware impairments[20, 71]. The unavoidable multiple low-dimensional RIS-UE CEs under one BS-
hardware error decreases the accuracy of channel RIS channel estimation. In Ref. [135], two anchor
estimation, and the beam squint phenomenon occurs nodes were deployed near the RIS to estimate the BS-
in wideband channels. Many literatures on channel RIS-UE channel and two cascaded channel estimation
estimation have been published to solve the frameworks are presented accordingly. The first
aforementioned problems, and RIS channel estimation approach takes advantage of the fact that all users share
schemes have been designed from different the BS-RIS channel. Within the coherence time, a part
perspectives. of the BS-RIS channel is estimated based on anchor-
In different application scenarios, the channel assisted training and feedback, and then the cascaded
characteristics of the BS-RIS and RIS-UE channels can channel is estimated using user pilot signals more
be different. For example, in the scenario where the frequently. The second approach estimates the relevant
RIS is deployed on the facade, the BS-RIS channel is channels of the anchor point A1 and A2 first, namely
semi-static with the coherence time being extended and the BS-RIS-A1 and A1-RIS-A2 channels, then
main energy is concentrated on the LOS path. In this estimates the A2-RIS-UE channel to restore the
case, channel estimation can be done in a longer cascaded BS-RIS-UE channel. In Ref. [136], a
periodicity. Conversely, due to the mobility of the UE, transmission protocol for RIS CE and configuration
the coherence time of the RIS-UE channel is optimization was proposed for RIS-empowered OFDM
significantly shorter, and the channel estimation needs systems. However, the presented RIS reflection pattern
to be done in shorter periodicity than that of the BS- is implemented by ON/OFF switching, which can be
RIS channel. In another possible scenario where RIS is costly in practice, and the estimation accuracy suffers
deployed on glasses of high speed trains, the BS-RIS since only a portion of RIS elements are activated
channel is dynamic while the RIS-UE channel can during the CE period. Recently, a novel transmission
be regarded as semi-static. How to perform high- protocol with quasi-periodic training symbols was
dimensional BS-RIS fast-changing channel estimation proposed in Ref. [137] to perform channel estimation
becomes the major challenge. Under scenarios where and passive beamforming simultaneously. This
the RIS is deployed on mobile objects such as buses or adaptive protocol adopts RIS elements grouping
unmanned aerial vehicles (UAV), the coherence time and the time interval of adjacent training symbols can
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 15

be adjusted to balance data transmission delay and angle domain was analyzed in Ref. [144]. It feeds back
effective transmission rate for supporting delay- user-independent CSI and user-specific CSI through
sensitive transmissions. dimension reduction method and designs a dynamic
The introduction of RIS makes the channel codebook through angle information to realize BS-RIS-
dimension increase sharply, thus it is urgent to explore UE cascaded channel estimation. On this basis, Ref.
a short-pilot and low-overhead channel estimation and [145] analyzed the double-structured sparsity of the
feedback scheme, which has become one of the major angle domain cascaded channel, using the common
issues that must be considered in the design of the BS-RIS channel and the sharing environmental
actual RIS channel estimation algorithm. RIS element scatters. It reveals that the angular cascaded channels
grouping is a possible way to reduce the pilot length. In related to different users have identical non-zero rows
Ref. [138], a novel RIS elements grouping method was and partially identical non-zero columns, and the
proposed as each group consists of a set of adjacent channel estimation is realized through the DS-OMP
RIS elements that share a common reflection algorithm. In addition, to solve the high-dimensional
coefficient. The training overhead is greatly reduced as CE problem and avoid huge pilot overhead, Ref. [146]
only the combined channel of each group needs to be studied the achievable rate of the RIS-empowered
estimated. Similarly, a novel hierarchical training SISO communication and selected the optimal number
reflection design was proposed in Ref. [139] to of RIS elements according to the statistical CSI to
progressively estimate the RIS-empowered channels balance the power gain and the channel estimation
over multiple blocks by exploiting RIS-elements overhead. It helps maximize the upper limit of the
grouping and partition. Moreover, the pilot length can average achievable rate and realize low-complexity and
be reduced if only the aggregated channel from the BS low-rate channel adaptation.
to the users, instead of every individual channel If all RIS elements are connected to the baseband
coefficient, was estimated in Ref. [140]. In the two- through a digital architecture, the excessive hardware
timescale scheme[141], the BS beamforming is designed complexity and power loss will affect the power cost-
based on the instantaneous aggregated CSI and the benefit of RIS seriously. Some researches adopt the
length of the pilot sequences only needs to be larger hybrid RIS hardware structure that requires a small
than the number of users instead of the number of RIS number of active elements with sensing and signal
elements. processing ability. In Ref. [54], a new RIS architecture
The design of the RIS reflection pattern affects the based on sparse channel sensors was studied where a
channel estimation performance as it controls the small number of active components are connected to
reflection signal. The authors in Ref. [142] claimed that the baseband. RIS channel estimation and
they designed an optimal CE scheme, where the RIS reconstruction with significantly reduced pilot
elements follow an optimal series of activation overhead are performed through compressed sensing,
patterns. In Ref. [143], the authors designed the joint and deep learning tool is adopted to design the
optimal training sequence and reflection pattern to reflection matrix to achieve a rate near the upper limit.
minimize the channel estimation MSE for the RIS- The CSI acquisition under the hardware design of the
empowered wireless system by recasting the uniformly distributed active sensor was analyzed in
corresponding problem into the convex semi-definite Ref. [147], and the rank-deficient structure of the RIS
programming problem. However, the authors simply channel was utilized to design the residual neural
estimated cascaded channels instead of separate network. The single-scale enhanced deep residual
channels. network reduces the RIS hardware complexity
The feature utilization can be of great help to significantly. The multi-scale enhanced deep residual
simplify the estimation procedure. The single-structured network uses the correlation between multi-scale
sparsity of RIS-UE channels in the hybrid space and models to ensure that the proportion of active elements
 16 Intelligent and Converged Networks, 2022, 3(1): 1−32

is adjusted to adapt to different scenarios and problem based on the joint received signal nuclear
achieve multi-scale super-resolution. For comparison, norm and the channel gain-norm objective was
Ref. [148] studied the channel estimation performance proposed in Ref. [151], and the cascaded channel
of the fully passive RIS architecture and the hybrid estimation was transformed into a sparse matrix
RIS architecture. For fully passive RIS, the angle of recovery problem. The channel estimation method
departure (AOD) of BS and the angle of arrival (AOA) based on the alternating direction multiplier achieves
of UE are estimated in the first stage, and the other high accuracy CE performance.
channel parameters are estimated subsequently. For An important advantage of high-frequency
hybrid RIS, alternate uplink and downlink training is transmission is that the available bandwidth is large.
assumed to estimate the segmented channels. However, the spatial wideband effect may occur in
RIS has numerous applications in the millimeter- millimeter-wave and THz frequency bands, affecting
wave frequency band, while a prominent feature of the channel estimation accuracy and bringing about
high-frequency scenario is that the transmission signal dispersion problems in beam regulation[152].
faces severer attenuation. The path with multiple Additionally, the large surface of RIS will further
reflections is usually ignored, and only the direct path aggravate the delay between the signals received by
and a small amount of single-reflection path can reach RIS elements, which may cause different RIS elements
the receiver. High-frequency channel usually has to receive different symbols. Correspondingly,
sparse characteristic, being significantly helpful for different subcarriers may point towards different
simplifying the RIS CE. In Ref. [149], a two-stage directions for transmitting and cause dispersion
channel estimation for parameter decoupling was problems. It is proved in Ref. [153] that the cross-
considered. The AOD of BS-RIS and the degree of correlation function between the spatial steering vector
arrival (DOA) of the RIS-UE channel are estimated in and the cascaded channel has two peaks in the angle
the first stage. The parameters estimated in the first domain, that is, the frequency-dependent actual angle
stage are used to estimate the azimuth, elevation, and and the frequency-independent error angle. The
path gain of the RIS cascaded channel in the second existence of the error angle affects estimation
stage. Low-overhead CE is achieved through ESPRIT performance severely, and a two-stage orthogonal
or compressive sensing algorithm. The inherent sparse matching pursuit (TS-OMP) algorithm is designed
quasi-static BS-RIS channel assumption was adopted accordingly. The influence of beam squint on
in Ref. [131], and the long-term statistical information transmission was discussed in Ref. [154], transforming
of the slow-varying channel components were utilized the wideband CE into the recovery problem of angle,
to carry out jointly BS-RIS and RIS-UE channel delay, gain, and other parameters. The Newton OMP
calibration and estimation. The cascaded channel can algorithm is proposed to detect the channel parameters,
be inferred through Bayesian posterior estimation, and and the Cramer-Rao lower bound (CRLB) is derived to
the approximate message passing (AMP) algorithm is provide strong support for performance evaluation. In
used to reduce the complexity of estimation Ref. [155], the problem that different incident angles
approximation. Moreover, Ref. [150] considered the bring different path phases due to beam squint while
ill-conditioned low-rank channel where RIS is the RIS phase shift matrix has the same value for all
deployed near the UE, transforming the RIS CE frequencies was analyzed. The phase shift matrix is
problem into a dictionary learning problem. The designed based on the SVD decomposition of the
reliable and robust bilinear adaptive vector AMP average channel covariance matrix to reduce the
algorithm is used to estimate the BS-RIS-UE channel, negative effect of beam squint effect on system
and the sparsity of the RIS phase shift matrix is utilized performance.
in the training phase to eliminate the restored channel Another direction to solve the high-dimensional CE
permutation ambiguity. A multi-objective optimization problem is channel splitting. The high-dimensional
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 17

channel can be equally split into multiple low- Ref. [161]. Alternate least square (ALS) and vector
dimensional channels, greatly reducing channel approximation message passing (VAMP) algorithms
estimation complexity. An ON-OFF based minimum are designed to reconstruct the unknown segmented
mean squared error (MMSE) channel estimation channels. This paper gave feasibility conditions and
method was designed in Ref. [156], splitting the RIS computational complexity evaluation, showing
into multiple elements and setting them as ON state at promising CE performance gains. In Ref. [162], the
different times. CE is performed for each RIS element adaptive PARAFAC decomposition scheme was
separately, achieving low complexity channel adopted to get the initial estimate and track the time-
estimation. It is assumed in Ref. [142] that there is no varying channel for the uplink of RIS-enabled multi-
prior knowledge about the channel and linear square user (MU) MIMO systems. Moreover, the generalized
estimation is used to estimate channel through each approximate message passing (GAMP) is utilized to
RIS element (or group). It aims to minimize CRLB recover the RIS-UE channels with reducing the pilot
under the constraints of attenuation and phase overhead.
quantization models, and the optimal design is guided The mobility of the UE brings challenges to channel
by the minimum variance unbiased estimation criteria. estimation and it is inevitable to study the time-varying
Also, it shows that the estimation performance depends channel in a mobile scenario. If the mobility cannot be
on the activation mode of RIS. Actually, the ON/OFF well handled, it becomes challenging to avoid channel
method[130, 133, 156] generally yields sub-optimal CE estimation deviations, leading to channel matching
results. In Ref. [157], the channel passing through each regardless of link adaptation or beamforming and
RIS element was modeled as a keyhole channel, and resulting in a gap between the actual achievable
the rank-one feature was considered to simplify the capacity and the ideal channel capacity. In Ref. [163],
estimation. This paper designed a channel splitting the cascaded channel was represented as a mobile
expansion scheme and used eigenvalue decomposition state-space model. A Kalman filter based scheme is
to perform cascaded channel estimation for RIS- designed to track the time-varying channel and
empowered communications, greatly reducing the improve estimation accuracy using the channel time
estimation time overhead. Similarly, a sub-channel correlation and prior information. The time-varying
estimation scheme is designed[158] based on multiple cascaded channel estimation problem was exploited in
rounds of pilot training. It appropriately configures the Ref. [164] by designing a deep learning channel
RIS phase shift matrix as hadamard matrix and DFT extrapolation scheme. It decomposes the neural
matrix and uses sub-channel CSI to design joint RIS- network into the time-domain section and the antenna-
transmitter precoding. domain section. In the time domain, the dynamic
In Ref. [159], a tensor modeling approach was channel is accurately described by combining the
considered to transform the CE problem of RIS- neural network and the cyclic neural network. In the
empowered communications to a 3D tensor through the antenna domain, the neural network is modified, the
parallel factor (PARAFAC) model and perform structure of the feedforward neural network is changed,
channel estimation using cascaded channel decoupling. and the enhanced feedforward neural network is
It derives a closed-form solution based on the Khatri- designed to enhance the network performance.
Rao decomposition of the cascaded channel. An Moreover, the doppler effect was considered in
iterative alternate estimation scheme promotes the Ref. [165], and a wideband CE scheme was designed
tensor signal model to the BS and multiple UEs based on doppler shift adjustment for multipath and
scenarios. Similarly, inspired by the promising results single path propagations. A quasi-static channel
of the PARAFAC decomposition[160] in CE for relay estimation mechanism is proposed for multipath
systems, two iterative estimation algorithms through scenarios, adjusting doppler distortion through the joint
PARAFAC of cascaded channels were proposed in RIS phase shift matrix design and time-frequency
 18 Intelligent and Converged Networks, 2022, 3(1): 1−32

domain conversion. RIS with some active elements for partial channels
The fixed position information of BS and RIS helps sensing. They presented an approach based on
design a low-complexity channel estimation method by compressive sensing to recover the full channels from
obtaining critical information such as the AOA and the samples. The rich-scattering environment was
AOD. The introduction of position information can considered in Ref. [169] and a deep neural network
also enhance other channel estimation algorithms, was trained as surrogate forward model to capture the
providing additional information, improving channel stochastic dependence of such RIS-empowered
estimation accuracy, and reducing channel estimation wireless channels. The link between the RIS
complexity. In Ref. [166], the position information of configuration and the key statistical channel parameter
the external positioning system was used to design the is captured to determine the communication rate with
directional beam. The atomic norm optimization the help of a DNN. Recently, in Ref. [73], an RIS
method extracts the channel angle and other parameters architecture consisting of any number of passive
to achieve the estimation performance with a RIS elements, a signal controller, and a single RF
predetermined beam codebook. It effectively chain for baseband measurements was presented. This
accelerates beam alignment and the channel parameter novel architecture is proposed to estimate channels
estimation process. The RIS-empowered SISO multi- at the RIS side via matrix completion for sparse
carrier system is considered in Ref. [167], and the channels.
large-scale, multi-element, and low-cost characteristics However, these machine learning based
of RIS are considered for joint 3D positioning. A low- methods[52, 54, 73] either require extensive offline
dimensional parameter space search algorithm is training or are built on the assumption that RISs have
designed to reduce the dimensionality of the four- some active elements to realize analog or hybrid
dimensional parameter estimation problem. It performs reception. Intuitively, architectures based on the latter
two one-dimensional searches for times-of-arrival and assumption would mean an increase in the RIS
a two-dimensional search for angles and achieves sub- hardware complexity and power consumption. In
meter positioning and synchronization accuracy. In Ref. [170], the approximate optimal MMSE channel
Ref. [168], the influence of RIS on the positioning was estimation scheme was studied, and the best linear
analyzed from the perspective of how to design RIS LMMSE estimation based on the majorization-
phase shift matrix to improve positioning accuracy. minimization algorithm was analyzed. A data-driven
This paper verified the positioning channel model of nonlinear solution based on deep learning is proposed,
the three-dimensional RIS-empowered communication achieving denoising and approximating the optimal
system and derived the CRLB. It designed RIS passive MMSE through convolutional neural networks. In
beamforming to minimize the CRLB, solved the Ref. [171], the number of active users was decreased to
problem through alternate optimization and gradient reduce pilot overhead, and a two-stage scheme for
descent methods, and realized centimeter-level overall CE was designed. In the first stage, the AOA at
positioning targets. BS is obtained through energy detection, the effective
The rapidly developed machine learning approaches path gain is obtained through the least squares (LS)
have been lately presented[52, 54, 73] to reduce the pilot method, and the AOA of the RIS-UE channel is
overhead and computational complexity in RIS- derived correspondingly. In the second stage, partial
empowered systems. Considering an indoor scenario, a CSI obtained in the first stage is used to handle the
deep neural network (DNN) is designed to unveil the spatio-temporal correlation between adjacent users.
mapping between the measured coordinate information The mapping of the channels of active and inactive
at a certain user location and the configuration of RIS users can be realized. The channel extrapolation
elements to maximize the received signal strength of a method of activating part of RIS elements was studied
given user[52]. In Ref. [54], the authors considered the in Ref. [172]. The proposed three-stage CSI acquisition
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 19

of BS-UE comprises direct channel estimation, partial blockage to achieve array blocking coefficient
RIS cascaded channel estimation, and overall cascaded calculation and effective sparseness channel parameter
channel prediction through deep neural networks. The estimation.
cumulative estimation error of DNN training in the Evaluating the impact of CSI error on beamforming
prediction stage can be overcome effectively through performance is the critical point of RIS-empowered
the BS antenna mapping relationship. Moreover, in communication system design. Robust beamforming
Ref. [56], an ML-based receiver was designed by based on non-ideal cascaded channel was investigated
treating the RIS reflection pattern as hyperparameter in Ref. [176]. For the bounded CSI error model, the BS
and a Bayesian ML framework was optimized to precoding matrix and the RIS phase shift matrix are
jointly tune the RIS and the multi-antenna receiver. jointly designed to minimize the total transmission
The analysis of RIS channel estimation and feedback power under the constant modulus constraint and the
mainly focuses on the design of the ideal RIS worst-case QoS constraint. For the statistical CSI error
hypothesis, but there are inevitable errors in practical model, modulus constraint and rate interruption
applications. For example, the ideal RIS hypothesis probability constraint are considered to achieve better
assumes that each RIS element has a constant performance with minimum transmit power, low
amplitude, variable phase, giving the same response to convergence speed and complexity. In Ref. [177], the
different frequencies. The discrete phase shift model channel hardening effect was analyzed, assuming that
and the relationship between the response under actual the interference channel is the spatially correlated
RIS hardware were analyzed in Ref. [173]. LS channel Rician fading. The UE uplink data rate and the RIS
estimation scheme based on non-ideal RIS hardware is system performance limit are derived. It concludes that
designed, and an alternate iterative algorithm is the hardware impairments and noise from estimation
proposed to obtain the RIS time-varying phase shift errors and the non-line of sight (NLOS) path could be
matrix to minimize the NMSE. In Ref. [174], RIS ignored with massive RIS elements. The statistical CSI
hardware impairments were considered, and a linear of the BS was utilized in Ref. [178] to design the best
LMMSE estimator was designed. The relationship transmission covariance matrix and RIS phase shift
between channel estimation performance and matrix. The achievable data rate is optimized through
impairment level, the number of reflective elements, iterative water-filling optimization and the RIS phase
and pilot power is given. It proves that the hardware shift matrix is optimized by the projected gradient
impairment of the transceiver under a high SNR limits ascent method.
the CE performance. The non-ideal CSI with the In Table 1, we analyze the complexity of different
random phase noise is studied in Ref. [130], and only types of channel estimation algorithms, hoping to
large-scale statistics are utilized for BS precoding provide a basis for channel estimation algorithm
and RIS reflecting matrix design. A closed-value selection in actual system design. As can be observed,
expression of the achievable spectrum efficiency of the the AI-based channel estimation schemes can exploit
uplink transmission is derived and optimized using the the massive channel measurement data and achieve
maximum ratio combining. The random blocking good channel estimation performance with acceptable
problem of RIS due to environmental influences was complexity. With the continuous development of AI
considered in Ref. [175], and the array diagnosis in the technology, AI-based channel estimation can have
case of passive RIS was discussed. The non-ideal greater development prospects in the future. Moreover,
channel estimation is modeled as joint antenna we found that in-depth analysis of the special
diagnosis, and channel estimation is solved by properties of RIS-empowered communication can
constructing a two-timescale non-convex optimization effectively reduce the complexity of channel
problem. A batch processing algorithm is selected to estimation, which is also a brighting way for future
identify induced attenuation and phase shift caused by channel estimation design.
 20 Intelligent and Converged Networks, 2022, 3(1): 1−32

Table 1 Complexity analysis for RIS-empowered MISO communication systems.

Source Algorithm Complexity
[133] Least Squares O((N M)2 T p )

[179] Bilinear Alternating Least Squares (BALS) O(2N + 4N MT p )

[180] Orthogonal Matching Pursuit (OMP) O(N MT p )

O(KNT ) + O(LG K MT Lk3 )
[145] Double-Structured Orthogonal Matching Pursuit (DS-OMP)
[142] Cramer–Rao lower bound ON/OFF O(N 3 M 3 )

[149] Two-Stage RIS-Aided Channel Estimation (TRICE) -CS O(LKT (L̄T L̄R + L + L2 ) + 2L3 + LKS L̄Sv L̄Sh )

[181] TRICE - Beamspace ESPRIT O(KT2 KS + KS3 + 3L3 + L)

[131] Matrix Calibration based CE O(Imax (NKT + 2NKLG + NNG (LG )2 ))

[182] DoA with 1-bit Quantization O(4BstN 2 M 2 )

[151] Alternating Direction Method of Mmultiplier (ADMM) O(Imax N M 2 T p )

[134] Coordinate Descent-based Channel Estimation O(M 3 + MLN 2 + MLNImax )

[183] Convolutional Blind Denoising Network (CBDNet) O(N MK 2 st(Ld D2l + Le El2 ))

[147] Single-scale Enhanced Deep Residual (EDSR) O(4N MK 2 (2B + κ2 )stD2 )

[147] Multiple-scale Enhanced Deep Residual (MDSR) O(8N MK 2 (B + 16)stD2 )

[184] Alternating Least Squares (ALS) O(2M 3 + 4M 2 P(K + N) − MP(K + N))

[161] Vector Approximate Message Passing (VAMP) O((K + N)(5M 2 − M))

[165] Multi-path CE Incorporating Doppler Shift Adjustment O(M 2 NO + MNO log NO )

[170] Denoising Convolutional Neural Network (DnCNN) O(N 2 (M + 1)2 + 9NN f (M + 1)(4 + (D − 1)N f ))
[170] Fast and Flexible Denoising Network (FFDNet) O(N 2 (M + 1)2 + 9NN f (M + 1)(4.5 + 0.5(D − 2)N f ))

5 RIS standardization: Current status and During its 55th meeting, the CCSA technical
road ahead committee 5 - working group 6 approved a proposal to
establish a study item (SI) on RIS[187]. The SI aims to
In this section, some highlights on the early
conclude in June 2022 with a technical report as the
standardization efforts on RIS are summarized and
major deliverable. The scope of this SI includes
presented. Reference [185] provides some details
channel modeling, channel estimation and feedback,
related to standardization activities as well.
As extensive research efforts are invested on RISs beamforming with RISs, AI and RISs, as well as
globally, the standardization work has also taken off, network protocols for RIS-assisted networks.
mainly on a regional-level. In China, one of the largest In June 2021, a new industry specification group
single markets for wireless devices, RIS-related (ISG) on RIS was approved by the European
standardization work has already been initiated in two Telecommunications Standards Institute (ETSI).
different organizations, namely, the FuTURE Mobile During the second plenary meeting of ISG RIS, three
Communication Forum and the China Communications new work items were approved, which focus on
Standards Association (CCSA). In the FuTURE forum, channel and system modeling, use cases and
a working group (WG) to study approaches to integrate deployment scenarios, and impact to current standards,
RISs into next generation wireless networks was respectively. All work item descriptions and related
established in December 2020[186]. This WG studies the contributions can be found on the ETSI portal[188]. The
use cases and key technologies to support RISs towards draft group reports of the 3 WIs are targeted to be
commercialization, and a white paper is planned to released in the middle of 2022 and stable report shall
summarize the technical trends for further promoting be made available by the end of 2022. The ISG will
RIS-assisted wireless systems. Compared to the focus on studies and generating informative reports till
FuTURE forum, CCSA is a more formal standards the year of 2023, after which normative specifications
developing organization (SDO) which can produce may be considered.
normative specifications on top of informative reports. The regional efforts to standardize RISs have some
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 21

echos in global SDOs. During the international Release 20 to Release 22. With the COVID-19
telecommunication union radiocommunication sector pandemic and its profound impact to the global
(ITU-R) WP 5D meeting in October 2020, countries economy and society, the future schedule of 3GPP may
submitted their candidate proposals for next generation be delayed. To standardize RIS as a whole new
wireless technologies, which will be considered for technology, some preparation work needs to be done
inclusion into the IMT Future Technology Trends and one of the most important work is channel
report. According to the draft report[189], RISs are modeling[191]. Due to the novel structure of
depicted as critical components for the physical layer metamaterials, new channel models are required to
of next generation networks. In 3GPP, there was a first model the channel between the antenna of the base
proposal submitted to 3GPP during the March 2021 station or the UE and reflective elements or regions on
meeting[190] by ZTE Corporation, after which more and RIS.
more companies are also showing interests in In 3GPP, there are two possible ways to standardize
supporting RIS to become a key component of 5G- RISs. One is to first initiate an SI on use cases,
Advanced networks and beyond, especially the deployment scenarios and channel models in Release
network operators. 19 and then to initiate a WI after the completion of the
Under the current status quo, the following future corresponding SI. This would allow RISs to be
progress is predicted as depicted in Fig. 7. In regional deployed as a component of 5G networks, which will
SDOs, the standardization of RISs is under way already help mature this technology and manufacturing
as described in the previous part of this section. For processes. In this case, RISs will be part of 5G
future standard activities, it is unlikely for RIS to be standards and, therefore, will be automatically
established as a dedicated focus group or a WI in the inherited to 6G standards. Another possible approach is
ITU since the ITU mainly focuses on regulatory, to standardize RISs as part of 6G standards together
spectrum, and business aspects. However, RIS related with other new features for 6G. This implies that the
channel modeling is possible to be studied in the ITU. deployment of RISs will be delayed to around 2030s
and is subject to uncertainties. Since the RIS
As far as the 3GPP is concerned, the current focus is
technology can be seen as generic and band-agnostic,
starting the second stage of 5G by Release 18, and
standardizing it as early as possible would enable
there is no formal plan for 6G yet. According to past
opportunities for RISs to be combined with other
experience, each generation of wireless communication
technologies, thus contributing to greener, safer, and
standards usually lasts for a decade, indicating that it is
more reliable wireless networks. On the other hand,
reasonable to expect that possible discussions on
there are still some challenging problems to be tackled,
requirements for 6G systems will begin after 2027.
which requires a certain amount of time and efforts.
Based on the current standardization progress on 5G,
Considering the current research status and the
we estimate the first 6G release will be between
maturity of this technology, the year to start studying
Study Work
RISs in 3GPP is likely to be 2023.
3GPP phase phase
The current cellular and local area networks are
Report on
ITU 6G trends designed to maximize the spectrum efficiency and
Regional Study phase
Work
SDOs phase energy efficiency, resulting in advanced and
Industry Prototyping&testing in 5G networks complicated physical layer designs and higher layer
Academia Vision, theoretical research&prototyping mechanisms and signalling. To study and standardize
2000 2020 2021 2022 2023 2024 2025 2026 RIS, it is best to start with simple scenarios and
Year practical models. For instance, considering only fixed
Fig. 7 Current status and forecasted timeline for the RIS panels would avoid mobility issues and help to
standardization of RISs. stabilize the network topology. What is more, although
 22 Intelligent and Converged Networks, 2022, 3(1): 1−32

carrier aggregation is widely used in different References

generations of cellular networks, single carrier
[1] 3GPP, 3GPP highlights, https://www.3gpp.org/ftp/
operation can be prioritized and studied first. Some key
Information/Highlights/2021Issue03/files/downloads/
aspects of standardization also relate to specific
Highlight Issue 3 final web.pdf, 2021.
frequency bands on which RIS will be deployed, for [2] AI and ML – Enablers for beyond 5G networks, White
example, the channel modeling and estimation, and it is paper, 5G PPP Technology Board, https://5g-ppp.eu/wp-
more efficient to consider lower frequency bands as a content/uploads/2021/05/AI-MLforNetworks-v1-0.
starting point. pdf, 2021.
[3] Y. Gao, J. Chen, Z. Liu, B. Zhang, Y. Ke, and R. Liu,
6 Conclusion Machine learning based energy saving scheme in
wireless access networks, in Proc. 2020 IEEE IWCMC,
Due to the increased potential of RISs for 6G wireless
Limassol, Cyprus, 2020, pp. 1573–1578.
communication networks, as witnessed by the various [4] Innovations in 5G backhaul technologies, Tech. Rep., 5G
recent proof of concepts ranging from passive Americas, 2020.
reflectarrays and dynamic metasurface antennas, there [5] L. Chen, X. Diao, J. Chen, and R. Liu, QoS assurance in
has been lately surprisingly increasing research and IAB network, in Proc. 2020 IEEE IWCMC, Limassol,
development activities on the RIS topic from both Cyprus, 2020, pp. 1860–1865.
[6] G. C. Alexandropoulos, Low complexity full duplex
academia and industry working in antenna design,
MIMO systems: Analog canceler architectures,
metamaterials, electromagnetics, signal processing, and
beamforming design, and future directions, ITU J. Future
wireless communications. The incorporation of RISs in Evolv. Technol., vol. 2, no. 2, pp. 1–19, 2021.
wireless networks has been recently advocated as a [7] Samsung, Samsung’s 6G white paper lays out the
revolutionary means to transform any wireless company’s vision for the next generation of
signal propagation environment to a dynamically communications technology, https://news.samsung.com/
programmable one for various networking objectives, global/samsungs-6gwhite-paper-lays-out-the-companys-
such as coverage extension, environmental perception vision-for-the-next-generationof-communications-
Technology, 2012.
and positioning, and spatiotemporal focusing. In this
[8] C. Liaskos, S. Nie, A. Tsioliaridou, A. Pitsillides, S.
paper, we provided an overview of the latest advances
Ioannidis, and I. F. Akyildiz, A new wireless
in RIS hardware architectures and their operational communication paradigm through software-controlled
considerations, as well as the most recent metasurfaces, IEEE Commun. Mag., vol. 56, no. 9, pp.
developments in the modeling of RISs and RIS- 162–169, 2018.
empowered wireless signal propagation. We also [9] C. Huang, A. Zappone, G. C. Alexandropoulos, M.
presented the up-to-date channel estimation approaches Debbah, and C. Yuen, Reconfigurable intelligent surfaces
for energy efficiency in wireless communication, IEEE
for RIS-empowered communications systems, which
Trans. Wireless Commun., vol. 18, no. 8, pp. 4157–
constitute a prerequisite step for the optimized
4170, 2019.
incorporation of RISs in future wireless networks. We [10] E. Basar, M. D. Renzo, J. D. Rosny, M. Debbah, M. -S.
discussed the relevance of the RIS technology in the Alouini, and R. Zhang, Wireless communications
latest standards, and highlighted the current and future through reconfigurable intelligent surfaces, IEEE Access,
standardization activities for RISs and RIS-empowered vol. 7, pp. 116753–116773, 2019.
wireless networking. [11] M. D. Renzo, M. Debbah, D. -T. Phan-Huy, A. Zappone,
M. -S. Alouini, C. Yuen, V. Sciancalepore, G. C.
Acknowledgment Alexandropoulos, J. Hoydis, H. Gacaninet, et al., Smart
radio environments empowered by reconfigurable AI
This work was supported by the EU H2020 Industrial
meta-surfaces: An idea whose time has come, EURASIP
Leadership Project (No. 101017011) and the Scientific J. Wireless Commun. Netw., vol. 2019, no. 1, pp. 1–20,
and Technological Research Council of Turkey 2019.
(TUBITAK) (No. 120E401). [12] Q. Wu and R. Zhang, Intelligent reflecting surface
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 23

enhanced wireless network via joint active and passive [23] M. Dohler and Y. Li, Cooperative Communications:
beamforming, IEEE Trans. Wireless Commun., vol. 18, Hardware, Channel and PHY. West Sussex, UK: Wiley,
no. 11, pp. 5394–5409, 2019. 2010.
[13] G. C. Alexandropoulos, G. Lerosey, M. Debbah, and M. [24] K. Ho-Van, P. C. Sofotasios, G. C. Alexandropoulos, and
Fink, Reconfigurable intelligent surfaces and S. Freear, Bit error rate of underlay decode-and-forward
metamaterials: The potential of wave propagation control cognitive networks with best relay selection, J. Commun.
for 6G wireless communications, IEEE ComSoc TCCN Netw., vol. 17, no. 2, pp. 162–171, 2015.
Newslett., vol. 6, no. 1, pp. 25–37, 2020. [25] G. C. Alexandropoulos, A. Papadogiannis, and P. C.
[14] E. Basar and I. Yildirim, Reconfigurable intelligent Sofotasios, A comparative study of relaying schemes
surfaces for future wireless networks: A channel with decode-and-forward over Nakagami-m fading
modeling perspective, IEEE Wireless Commun., vol. 28, channels, J. Comp. Netw. Commun., vol. 2011, no. 1, p.
no. 3, pp. 108–114, 2021. 560528.1–560528.14, 2011.
[15] G. C. Alexandropoulos, N. Shlezinger, and P. D. Hougne, [26] J. -P. Niu and G. Y. Li, An overview on backscatter
Reconfigurable intelligent surfaces for rich scattering communications, J. Commun. Inf. Netw., vol. 4, no. 2, pp.
wireless communications: Recent experiments, 1–14, 2019.
challenges, and opportunities, IEEE Commun. Mag., vol. [27] N. Kaina, M. Dupré, G. Lerosey, and M. Fink, Shaping
59, no. 6, pp. 28–34, 2021. complex microwave fields in reverberating media with
[16] E. Calvanese Strinati, G. C. Alexandropoulos, H. binary tunable metasurfaces, Scientific Reports, vol. 4, p.
Wymeersch, B. Denis, V. Sciancalepore, R. D’Errico, A. 6693, 2014.
Clemente, D. -T. Phan-Huy, E. D. Carvalho, and P. [28] B. Di, H. Zhang, L. Song, Y. Li, Z. Han, and H. V. Poor,
Popovski, Reconfigurable, intelligent, and sustainable Hybrid beamforming for reconfigurable intelligent
wireless environments for 6G smart connectivity, IEEE surface based multi-user communications: Achievable
Commun. Mag., vol. 59, no. 10, pp. 99–105, 2021. rates with limited discrete phase shifts, IEEE J. Sel.
[17] E. C. Strinati, G. C. Alexandropoulos, V. Sciancalepore, Areas Commun., vol. 38, no. 8, pp. 1809–1822, 2020.
M. D. Renzo, H. Wymeersch, D. -T. Phan-Huy, M. [29] E. Björnson and L. Sanguinetti, Power scaling laws and
Crozzoli, R. D’Errico, E. D. Carvalho, P. Popovski, et al., near-field behaviors of massive MIMO and intelligent
Wireless environment as a service enabled by reflecting surfaces, IEEE Open J. Commun. Soc., vol. 1,
reconfigurable intelligent surfaces: The RISE-6G pp. 1306–1324, 2020.
perspective, in Proc. 2021 Joint EuCNC & 6G Summit, [30] W. Tang, M. Z. Chen, X. Chen, J. Y. Dai, Y. Han, M. D.
Porto, Portugal, 2021, pp. 562–567. Renzo, Y. Zeng, S. Jin, Q. Cheng, and T. J. Cui, Wireless
[18] E. Basar and H. V. Poor, Present and future of communications with reconfigurable intelligent surface:
reconfigurable intelligent surface-empowered Path loss modeling and experimental measurement, IEEE
communications, IEEE Signal Process. Mag., vol. 38, no. Trans. Wireless Commun., vol. 20, no. 1, pp. 421–439,
6, pp. 146–152, 2021. 2021.
[19] E. Basar, M. Wen, R. Mesleh, M. D. Renzo, Y. Xiao, and [31] E. Basar, I. Yildirim, and F. Kilinc, Indoor and outdoor
H. Haas, Index modulation techniques for next- physical channel modeling and efficient positioning for
generation wireless networks, IEEE Access, vol. 5, pp. reconfigurable intelligent surfaces in mmWave bands,
16693–16746, 2017. IEEE Trans. Commun., vol. 69, no. 12, pp. 8600–8611,
[20] C. Huang, S. Hu, G. C. Alexandropoulos, A. Zappone, C. 2021.
Yuen, R. Zhang, M. D. Renzo, and M. Debbah, [32] F. H. Danufane, M. D. Renzo, J. D. Rosny, and S.
Holographic MIMO surfaces for 6G wireless networks: Tretyakov, On the path-loss of reconfigurable intelligent
Opportunities, challenges, and trends, IEEE Wirel. surfaces: An approach based on Green’s theorem applied
Commun., vol. 27, no. 5, pp. 118–125, 2020. to vector fields, IEEE Trans. Commun., vol. 69, no. 8, pp.
[21] T. L. Marzetta, E. G. Larsson, H. Yang, and H. Q. Ngo, 5573–5592, 2021.
Fundamentals of Massive MIMO. Cambridge, UK: [33] R. Faqiri, C. Saigre-Tardif, G. C. Alexandropoulos, N.
Cambridge University Press, 2016. Shlezinger, M. F. Imani, and P. D. Hougne, PhysFad:
[22] A. F. Molisch, V. V. Ratnam, S. Han, Z. Li, S. L. H. Physics-based end-to-end channel modeling of RIS-
Nguyen, L. Li, and K. Haneda, Hybrid beamforming for parametrized environments with adjustable fading, https:/
massive MIMO: A survey, IEEE Commun. Mag., vol. 55, /arxiv.org/pdf/2202.02673.pdf, 2022.
no. 9, pp. 134–141, 2017. [34] Q. Nadeem, A. Kammoun, A. Chaaban, M. Debbah, and
 24 Intelligent and Converged Networks, 2022, 3(1): 1−32

M. -S. Alouini, Asymptotic max-min SINR analysis of Commun. Netw., doi. 10.3389/frcmn.2021.672857.
reconfigurable intelligent surface assisted MISO systems, [47] I. Alamzadeh, G. C. Alexandropoulos, N. Shlezinger, and
IEEE Trans. Wireless Commun., vol. 19, no. 12, pp. M. F. Imani, A reconfigurable intelligent surface with
7748–7764, 2020. integrated sensing capability, Scientific Reports, vol. 11,
[35] D. Selimis, K. P. Peppas, G. C. Alexandropoulos, and F. p. 20737, 2021.
I. Lazarakis, On the performance analysis of RIS- [48] H. Zhang, N. Shlezinger, I. Alamzadeh, G. C.
empowered communications over Nakagami-m fading, Alexandropoulos, M. F. Imani, and Y. C. Eldar, Channel
IEEE Commun. Let., vol. 25, no. 7, pp. 2191–2195, 2021. estimation with simultaneous reflecting and sensing
[36] A. L. Moustakas, G. C. Alexandropoulos, and M. reconfigurable intelligent metasurfaces, in Proc. 2021
Debbah, Capacity optimization using reconfigurable IEEE 22nd International Workshop on SPAWC, Lucca,
intelligent surfaces: A large system approach, in Proc. Italy, 2021, pp. 536–540.
2021 IEEE GLOBECOM, Madrid, Spain, 2021, pp. 1–6. [49] G. C. Alexandropoulos, N. Shlezinger, I. Alamzadeh, M.
[37] M. Cui, G. Zhang, and R. Zhang, Secure wireless F. Imani, H. Zhang, and Y. C. Eldar, Hybrid
communication via intelligent reflecting surface, IEEE reconfigurable intelligent metasurfaces: Enabling
Wireless Commun. Lett., vol. 8, no. 5, pp. 1410–1414, simultaneous tunable reflections and sensing for 6G
2019. wireless communications, https://arxiv.org/abs/2104.
[38] J. Chen, Y. -C. Liang, Y. Pei, and H. Guo, Intelligent 04690, 2021.
reflecting surface: A programmable wireless environment [50] Z. Zhang, L. Dai, X. Chen, C. Liu, F. Yang, R. Schober,
for physical layer security, IEEE Access, vol. 7, pp. and H. V. Poor, Active RIS vs. passive RIS: Which will
82599–82612, 2019. prevail in 6G? https://arxiv.org/abs/2103.15154v1, 2021.
[39] G. C. Alexandropoulos, K. Katsanos, M. Wen, and D. B. [51] Y. Liu, X. Mu, J. Xu, R. Schober, Y. Hao, H. V. Poor,
D. Costa, Safeguarding MIMO communications with and L. Hanzo, STAR: Simultaneous transmission and
reconfigurable metasurfaces and artificial noise, in Proc. reflection for 360 coverage by intelligent surfaces, https:/
2021 IEEE ICC, Montreal, Canada, 2021, pp. 1–6. /arxiv.org/abs/2103.09104v2, 2021.
[40] Z. Ding and H. V. Poor, A simple design of IRS-NOMA [52] C. Huang, G. C. Alexandropoulos, C. Yuen, and M.
transmission, IEEE Commun. Lett., vol. 24, no. 5, pp. Debbah, Indoor signal focusing with deep learning
1119–1123, 2020. designed reconfigurable intelligent surfaces, in Proc.
[41] T. Hou, Y. Liu, Z. Song, X. Sun, Y. Chen, and L. Hanzo, 2019 IEEE 20th International Workshop on SPAWC,
Reconfigurable intelligent surface aided NOMA Cannes, France, 2019, pp. 1–6.
networks, IEEE J. Sel. Areas Commun., vol. 38, no. 11, [53] G. C. Alexandropoulos, S. Samarakoon, M. Bennis, and
pp. 2575–2588, 2020. M. Debbah, Phase configuration learning in wireless
[42] I. Al-Nahhal, O. A. Dobre, and E. Basar, Reconfigurable networks with multiple reconfigurable intelligent
intelligent surface-assisted uplink sparse code multiple surfaces, in Proc. 2020 IEEE Globecom Workshops,
access, IEEE Commun. Lett., vol. 25, no. 6, pp. Taiwan, Province of China, 2020, pp. 1–6.
2058–2062, 2021. [54] A. Taha, M. Alrabeiah, and A. Alkhateeb, Enabling large
[43] K. Chen-Hu, G. C. Alexandropoulos, and A. G. Armada, intelligent surfaces with compressive sensing and deep
Non-coherent modulation for RIS-empowered multi- learning, IEEE Access, vol. 9, pp. 44304–44321, 2021.
antenna OFDM communications, https://arxiv.org/abs/ [55] L. Wang, N. Shlezinger, G. C. Alexandropoulos, H.
2103.05547, 2021. Zhang, B. Wang, and Y. C. Eldar, Jointly learned symbol
[44] K. Chen-Hu, G. C. Alexandropoulos, and A. G. Armada, detection and signal reflection in RIS-aided multi-user
Non-coherent MIMO-OFDM uplink empowered by the MIMO systems, in Proc. 2021 55th Asilomar Conf.
spatial diversity in reflecting surfaces, arXiv preprint Signals, Sys., and Comp., Pacific Grove, CA, USA, 2021,
arXiv: 2202.02035, 2022. pp. 715–721.
[45] D. Dampahalage, K. B. Shashika Manosha, N. Rajatheva, [56] K. Stylianopoulos, N. Shlezinger, P. D. Hougne, and G.
and M. Latva-aho, Intelligent reflecting surface aided C. Alexandropoulos, Deep-learning-assisted
vehicular communications, in Proc. 2020 IEEE configuration of reconfigurable intelligent surfaces in
Globecom Workshops, Taiwan, Province of China, 2020, dynamic rich-scattering environments, arXiv preprint
pp. 1–6. arXiv: 2202.07884, 2022.
[46] E. Basar, Reconfigurable intelligent surfaces for Doppler [57] S. Guo, S. Lv, H. Zhang, J. Ye, and P. Zhang, Reflecting
effect and multipath fading mitigation, Frontiers modulation, IEEE J. Sel. Areas Commun., vol. 38, no. 11,
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 25

pp. 2548–2561, 2020. Smith, Dynamically reconfigurable holographic

[58] S. Lin, B. Zheng, G. C. Alexandropoulos, M. Wen, M. D. metasurface aperture for a mills-cross monochromatic
Renzo, and F. Chen, Reconfigurable intelligent surfaces microwave camera, Opt. Express, vol. 26, no. 5, pp.
with reflection pattern modulation: Beamforming design 5281–5291, 2018.
and performance analysis, IEEE Trans. Wireless [70] A. Pizzo, T. L. Marzetta, and L. Sanguinetti, Spatial
Commun., vol. 20, no. 2, pp. 741–754, 2021. characterization of holographic MIMO channels, IEEE J.
[59] J. Yuan, M. Wen, Q. Li, E. Basar, G. C. Sel. Areas Commun., vol. 38, no. 9, pp. 1964–1979,
Alexandropoulos, and G. Chen, Receive quadrature 2020.
reflecting modulation for RIS-empowered wireless [71] N. Shlezinger, O. Dicker, Y. C. Eldar, I. Yoo, M. F.
communications, IEEE Trans. Veh. Technol., vol. 70, no. Imani, and D. R. Smith, Dynamic metasurface antennas
5, pp. 5121–5125, 2021. for uplink massive MIMO systems, IEEE Trans.
[60] L. Dai, B. Wang, M. Wang, X. Yang, J. Tan, S. Bi, S. Commun., vol. 67, no. 10, pp. 6829–6843, 2019.
Xu, F. Yang, Z. Chen, M. D. Renzo, et al., [72] S. Hu, F. Rusek, and O. Edfors, Beyond massive MIMO:
Reconfigurable intelligent surface-based wireless The potential of positioning with large intelligent
communications: Antenna design, prototyping, and surfaces, IEEE Trans. Signal Process., vol. 66, no. 7, pp.
experimental results, IEEE Access, vol. 8, pp. 1761–1774, 2018.
45913–45923, 2020. [73] G. C. Alexandropoulos and E. Vlachos, A hardware
[61] J. Y. Dai, W. Tang, M. Z. Chen, C. H. Chan, Q. Cheng, architecture for reconfigurable intelligent surfaces with
S. Jin, and T. J. Cui, Wireless communication based on minimal active elements for explicit channel estimation,
information metasurfaces, IEEE Trans. Microw. Theory in Proc. 2020 IEEE ICASSP, Barcelona, Spain, 2020, pp.
Tech., vol. 69, no. 3, pp. 1493–1510, 2021. 9175–9179.
[62] G. C. Alexandropoulos, N. C. Sagias, F. I. Lazarakis, and [74] L. Dong, H. -M. Wang, and J. Bai, Active reconfigurable
K. Berberidis, New results for the multivariate intelligent surface aided secure transmission, IEEE
Nakagami-m fading model with arbitrary correlation Trans. Veh. Technol., vol. 71, no. 2, pp. 2181–2186,
matrix and applications, IEEE Trans. Wireless Commun., 2022.
vol. 8, no. 1, pp. 245–255, 2009. [75] G. C. Alexandropoulos, I. Vinieratou, and H.
[63] G. C. Alexandropoulos and M. Kountouris, Maximal Wymeersch, Localization via multiple reconfigurable
ratio transmission in wireless Poisson networks under intelligent surfaces equipped with single receive RF
spatially correlated fading channels, in Proc. 2015 IEEE chains, IEEE Wireless Commun. Lett., doi. 10.1109/
GLOBECOM, San Diego, CA, USA, 2015, pp. 1–6. LWC. 2022.3156427.
[64] M. T. Ivrlač and J. A. Nossek, Toward a circuit theory of [76] E. Vlachos, G. C. Alexandropoulos, and J. Thompson,
communication, IEEE Trans. Circuits Syst. I: Reg. Wideband MIMO channel estimation for hybrid
Papers, vol. 57, no. 7, pp. 1663–1683, 2010. beamforming millimeter wave systems via random
[65] G. C. Alexandropoulos, V. I. Barousis, and C. B. spatial sampling, IEEE J. Sel. Topics Signal Process.,
Papadias, Precoding for multiuser MIMO systems with vol. 13, no. 5, pp. 1136–1150, 2019.
single-fed parasitic antenna arrays, in Proc. 2014 IEEE [77] J. Xu, Y. Liu, X. Mu, and O. A. Dobre, STAR-RISs:
GLOBECOM, Austin, TX, USA, 2014, pp. 3897–3902. Simultaneous transmitting and reflecting reconfigurable
[66] L. Subrt and P. Pechac, Intelligent walls as autonomous intelligent surfaces, IEEE Commun. Lett., vol. 25, no. 9,
parts of smart indoor environments, IET Commun., vol. pp. 3134–3138, 2021.
6, no. 8, pp. 1004–1010, 2012. [78] S. Zhang, H. Zhang, B. Di, Y. Tan, M. D. Renzo, Z. Han,
[67] V. Jamali, G. C. Alexandropoulos, R. Schober, and H. V. H. V. Poor, and L. Song, Intelligent omni-surfaces:
Poor, Lowto-zero-overhead IRS reconfiguration: Ubiquitous wireless transmission by reflective-refractive
Decoupling illumination and channel estimation, arXiv metasurfaces, IEEE Trans. Wireless Commun., vol. 21,
preprint arXiv: 2111.0942, 2021. no. 1, pp. 219–233, 2022.
[68] R. A. Tasci, F. Kilinc, E. Basar, and G. C. [79] Y. Liu, X. Mu, J. Xu, R. Schober, Y. Hao, H. V. Poor,
Alexandropoulos, An amplifying RIS architecture with a and L. Hanzo, STAR: Simultaneous transmission and
single power amplifier: Energy efficiency and error reflection for 360◦ coverage by intelligent surfaces, arXiv
performance analysis, https://arxiv.org/pdf/2111. preprint arXiv: 2103.09104, 2021.
09855, 2021. [80] J. Xu, Y. Liu, X. Mu, J. T. Zhou, L. Song, H. V. Poor,
[69] O. Yurduseven, D. L. Marks, T. Fromenteze, and D. R. and L. Hanzo, Simultaneously transmitting and reflecting
 26 Intelligent and Converged Networks, 2022, 3(1): 1−32

(STAR) intelligent omni-surfaces, their modeling and [94] T. S. Rappaport, Y. Xing, O. Kanhere, S. Ju, A.
implementation, arXiv preprint arXiv: 2108.06233, 2021. Madanayake, S. Mandal, A. Alkhateeb, and G. C.
[81] NTT DOCOMO, DOCOMO conducts world’s first Trichopoulos, Wireless communications and applications
successful trial of transparent dynamic metasurface, above 100 GHz: Opportunities and challenges for 6G and
https://www.docomo.ne.jp/english/info/media_center/ beyond, IEEE Access, vol. 7, pp. 78729–78757, 2019.
pr/2020/0117_00.html, 2020. [95] L. Pometcu and R. D’Errico, An indoor channel model
[82] C. Zhang, W. Yi, Y. Liu, Z. Ding, and L. Song, STAR- for high data-rate communications in D-band, IEEE
IOS aided NOMA networks: Channel model Access, vol. 8, pp. 9420–9433, 2019.
approximation and performance analysis, arXiv preprint [96] M. Fink, Time reversal of ultrasonic fields—part I: Basic
arXiv: 2107.01543, 2021. principles, IEEE Trans. Ultrason. Ferroelectron.
[83] H. Zhang, S. Zeng, B. Di, Y. Tan, M. D. Renzo, M. Frequency Contr., vol. 39, no. 5, pp. 555–566, 1992.
Debbah, L. Song, Z. Han, and H. V. Poor, Intelligent [97] G. C. Alexandropoulos, R. Khayatzadeh, M. Kamoun, Y.
omni-surfaces for full-dimensional wireless Ganghua, and M. Debbah, Indoor time reversal wireless
communications: Principle, technology, and communication: Experimental results for localization and
implementation, arXiv preprint arXiv: 2104.12313, 2021. signal coverage, in Proc. 2019 IEEE ICASSP, Brighton,
[84] J. Xu, Y. Liu, X. Mu, R. Schober, and H. V. Poor, UK, 2014, pp. 7844–7848.
STAR-RISs: A correlated T&R phase-shift model and [98] A. Mokh, J. D. Rosny, G. C. Alexandropoulos, R.
practical phase-shift configuration strategies, arXiv Khayatzadeh, A. Ourir, M. Kamoun, A. Tourin, and M.
preprint arXiv: 2112.00299, 2021. Fink, Time reversal precoding at subTHz frequencies:
[85] Y. Liu, X. Mu, R. Schober, and H. V. Poor, Experimental results on spatiotemporal focusing, in Proc.
Simultaneously transmitting and reflecting (STAR)-RISs: 2021 IEEE CSCN, Thessaloniki, Greece, 2021, pp.
A coupled phase-shift model, arXiv preprint arXiv: 78–82.
2110.02374, 2021. [99] A. Mokh, J. D. Rosny, G. C. Alexandropoulos, R.
[86] I. Yildirim, F. Kilinc, E. Basar, and G. C. Khayatzadeh, A. Ourir, M. Kamoun, A. Tourin, and M.
Alexandropoulos, Hybrid RIS-empowered reflection and Fink, Time reversal for multiple access and mobility:
decode-and-forward relaying for coverage extension, Algorithmic design and experimental results, presented at
IEEE Commun. Lett., vol. 25, no. 5, pp. 1692–1696, IEEE WCNC, Austin, TX, USA, 2022.
2021. [100] P. D. Hougne, F. Lemoult, M. Fink, and G. Lerosey,
[87] I. F. Akyildiz, C. Han, and S. Nie, Combating the Spatiotemporal wave front shaping in a microwave
distance problem in the millimeter wave and terahertz cavity, Phys. Rev. Lett., vol. 117, no. 13, p. 134302,
frequency bands, IEEE Commun. Mag., vol. 56, no. 6, 2016.
pp. 102–108, 2018. [101] S. Abeywickrama, R. Zhang, Q. Wu, and C. Yue,
[88] Demonstration of ZTE RIS Panels, http://www. Intelligent reflecting surface: Practical phase shift model
risforum.com/LiveVideoServer/risPC/video/zxtx_926. and beamforming optimization, IEEE Trans. Commun.,
mp4?v=20211101-1545, 2021. vol. 68, no. 9, pp. 5849–5863, 2020.
[89] ZTE builds efficient way to 5G-Advanced and 6G with [102] K. D. Katsanos, N. Shlezinger, M. F. Imani, and G. C.
RIS solution, https://www.youtube.com/watch?v=ml00 Alexandropoulos, Wideband multi-user MIMO
XFnIzn8, 2022. communications with frequency selective RISs: Element
[90] EU H2020 RISE-6G, https://cordis.europa.eu/project/ response modeling and sum-rate maximization, arXiv
id/101017011, 2020. preprint arXiv: 2202.02169, 2022.
[91] T. S. Rappaport, G. R. MacCartney, M. K. Samimi, and [103] H. Zhang, L. Song, Z. Han, and H. V. Poor, Spatial
S. Sun, Wideband millimeter-wave propagation equalization before reception: Reconfigurable intelligent
measurements and channel models for future wireless surfaces for multi-path mitigation, in Proc. 2021 IEEE
communication system design, IEEE Trans. Commun., ICASSP, Toronto, Canada, 2021, pp. 8062–8066.
vol. 63, no. 9, pp. 3029–3056, 2015. [104] R. Zhou, X. Chen, W. Tang, X. Li, S. Jin, E. Basar, Q.
[92] EU METIS, https://metis2020.com/, 2020. Cheng, and T. J. Cui, Modeling and measurements for
[93] R. J. Weiler, M. Peter, W. Keusgen, and M. Wisotzki, multi-path mitigation with reconfigurable intelligent
Measuring the busy urban 60 GHz outdoor access radio surfaces, https://arxiv.org/pdf/2109.11820, 2021.
channel, in Proc. 2014 IEEE ICUWB, Paris, France, [105] E. Arslan, I. Yildirim, F. Kilinc, and E. Basar, Over-the-
2014, pp. 166–170. air equalization with reconfigurable intelligent surfaces,
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 27

https://arxiv.org/pdf/2106.07996, 2021. 1964–1979, 2020.

[106] M. Najafi, V. Jamali, R. Schober, and H. V. Poor, [119] L. Wei, C. Huang, G. C. Alexandropoulos, W. E. I. Sha,
Physics-based modeling of large intelligent reflecting Z. Zhang, M. Debbah, and C. Yuen, Multi-user
surfaces for scalable optimization, in Proc. 2020 54th holographic MIMO surface: Channel modeling and
Asilomar Conf. Signals, Sys., and Comp., Pacific Grove, spectral efficiency analysis, arXiv preprint arXiv:
CA, USA, 2020, pp. 559–563. 2112.02803, 2021.
[107] O. Özdogan, E. Björnson, and E. G. Larsson, Intelligent [120] Y. Liu, X. Mu, X. Liu, M. D. Renzo, Z. Ding, and R.
reflecting surfaces: Physics, propagation, and pathloss Schober, Reconfigurable intelligent surface (RIS) aided
modeling, IEEE Wireless Commun. Lett., vol. 9, no. 5, multi-user networks: Interplay between NOMA and RIS,
pp. 581–585, 2020. https://arxiv.org/abs/2011.13336, 2021.
[108] H. Ajam, M. Najafi, V. Jamali, B. Schmauss, and R. [121] J. Xu and Y. Liu, A novel physics-based channel model
Schober, Modeling and design of IRS-assisted multi-link for reconfigurable intelligent surface-assisted multi-user
FSO systems, arXiv preprint arXiv: 2108.00291, 2021. communication systems, IEEE Trans. Wireless Commun.,
[109] M. D. Renzo, F. H. Danufane, X. Xi, J. D. Rosny, and S. vol. 21, no. 2, pp. 1183–1196, 2021.
Tretyakov, Analytical modeling of the path-loss for [122] Z. Sun and Y. Jing, On the performance of multi-antenna
reconfigurable intelligent surfaces – Anomalous mirror or IRS-assisted NOMA networks with continuous and
scatterer? arXiv preprint arXiv: 2001.10862, 2020. discrete IRS phase shifting, IEEE Trans. Wireless
[110] S. W. Ellingson, Path loss in reconfigurable intelligent Commun., doi: 10.1109/TWC.2021.3117494.
surface-enabled channels, arXiv preprint arXiv: [123] Y. Cheng, K. H. Li, Y. Liu, K. C. Teh, and G. K.
1912.06759, 2019. Karagiannidis, Nonorthogonal multiple access (NOMA)
[111] A. -A. A. Boulogeorgos and A. Alexiou, Pathloss with multiple intelligent reflecting surfaces, IEEE Trans.
modeling of reconfigurable intelligent surface assisted Wireless Commun., vol. 20, no. 11, pp. 7184–7195, 2021.
THz wireless systems, arXiv preprint arXiv: 2102.08757, [124] M. Fu, Y. Zhou, Y. Shi, and K. B. Letaief,
2021. Reconfigurable intelligent surface empowered downlink
[112] K. Dovelos, S. D. Assimonis, H. Q. Ngo, B. Bellalta, and non-orthogonal multiple access, IEEE Trans. Commun.,
M. Matthaiou, Intelligent reflecting surfaces at terahertz vol. 69, no. 6, pp. 3802–3817, 2021.
bands: Channel modeling and analysis, arXiv preprint [125] G. Yang, X. Xu, Y. -C. Liang, and M. D. Renzo,
arXiv: 2103.15239, 2021. Reconfigurable intelligent surface-assisted non-
[113] G. Gradoni and M. D. Renzo, End-to-end mutual orthogonal multiple access, IEEE Trans. Wireless
coupling aware communication model for reconfigurable Commun., vol. 20, no. 5, pp. 3137–3151, 2021.
intelligent surfaces: An electromagnetic-compliant [126] A. Khaleel and E. Basar, A novel NOMA solution with
approach based on mutual impedances, arXiv preprint RIS partitioning, IEEE J. Sel. Top. Signal Process., vol.
arXiv: 2009.02694, 2020. 16, no. 1, pp. 70–81, 2021.
[114] Study on channel model for frequencies from 0.5 to 100 [127] C. Zhang, W. Yi, Y. Liu, K. Yang, and Z. Ding,
GHz, 3GPP TR 38.901 V16.1. 0, 2020. Reconfigurable intelligent surfaces aided multi-cell
[115] E. Björnson and L. Sanguinetti, Rayleigh fading NOMA networks: A stochastic geometry model, IEEE
modeling and channel hardening for reconfigurable Trans. Commun., vol. 70, no. 2, pp. 951–966, 2021.
intelligent surfaces, IEEE Wireless Commun. Lett., vol. [128] C. Zhang, W. Yi, Y. Liu, and Q. Wang, Multi-cell
10, no. 4, pp. 830–834, 2021. NOMA: Coherent reconfigurable intelligent surfaces
[116] G. C. Alexandropoulos, N. C. Sagias, and K. Berberidis, model with stochastic geometry, in Proc. 2021 IEEE
On the multivariate Weibull fading model with arbitrary ICC, Montreal, Canada, 2021, pp. 1–6.
correlation matrix, IEEE Ant. Wireless Prop. Lett., vol. 6, [129] Z. Xie, W. Yi, X. Wu, Y. Liu, and A. Nallanathan,
pp. 93–95, 2007. STAR-RIS aided NOMA in multi-cell networks: A
[117] G. C. Alexandropoulos, P. T. Mathiopoulos, and N. C. general analytical framework with gamma distributed
Sagias, Switch-and-examine diversity over arbitrary channel modeling, https://arxiv.org/abs/2108.06704, 2021.
correlated Nakagami-m fading channels, IEEE Trans. [130] Q. -U. -A. Nadeem, H. Alwazani, A. Kammoun, A.
Veh. Technol., vol. 59, no. 4, pp. 2080–2087, 2010. Chaaban, M. Debbah, and M. -S. Alouini, Intelligent
[118] A. Pizzo, T. L. Marzetta, and L. Sanguinetti, Spatially- reflecting surface-assisted multiuser MISO
stationary model for holographic MIMO small-scale communication: Channel estimation and beamforming
fading, IEEE J. Sel. Areas Commun., vol. 38, no. 9, pp. design, IEEE Open J. Commun. Soc., vol. 1, pp.
 28 Intelligent and Converged Networks, 2022, 3(1): 1−32

661–680, 2020. CSI, arXiv preprint arXiv: 2108.07622, 2021.

[131] H. Liu, X. Yuan, and Y. J. A. Zhang, Matrix-calibration- [142] T. L. Jensen and E. D. Carvalho, An optimal channel
based cascaded channel estimation for reconfigurable estimation scheme for intelligent reflecting surfaces
intelligent surface assisted multiuser MIMO, IEEE J. Sel. based on a minimum variance unbiased estimator, in
Area. Comm., vol. 38, no. 11, pp. 2621–2636, 2020. Proc. 2020 IEEE ICASSP, Barcelona, Spain, 2020, pp.
[132] Z. He and X. Yuan, Cascaded channel estimation for 5000–5004.
large intelligent metasurface assisted massive MIMO, [143] J. Kang, Intelligent reflecting surface: Joint optimal
IEEE Wireless Commun. Lett., vol. 9, no. 2, pp. 210–214, training sequence and refection pattern, IEEE Commun.
2020. Lett., vol. 24, no. 8, pp. 1784–1788, 2020.
[133] D. Mishra and H. Johansson, Channel estimation and [144] D. Shen and L. Dai, Dimension reduced channel
low-complexity beamforming design for passive feedback for reconfigurable intelligent surface aided
intelligent surface assisted MISO wireless energy wireless communications, IEEE Trans. on Commun., vol.
transfer, in Proc. 2019 IEEE ICASSP, Brighton, UK, 69, no. 11, pp. 7748–7760, 2021.
[145] X. Wei, D. Shen, and L. Dai, Channel estimation for RIS
2019, pp. 4659–4663.
[134] C. Hu, L. Dai, S. Han, and X. Wang, Two-timescale assisted wireless communications—part II: An improved
channel estimation for reconfigurable intelligent surface solution based on double-structured sparsity, IEEE
aided wireless communications, IEEE Trans. on Commun. Lett., vol. 25, no. 5, pp. 1403–1407, 2021.
[146] N. K. Kundu and M. R. Mckay, Large intelligent surfaces
Commun., vol. 69, no. 11, pp. 7736–7747, 2021.
with channel estimation overhead: Achievable rate and
[135] X. Guan, Q. Wu, and R. Zhang, Anchor-assisted channel
optimal configuration, IEEE Wireless Commun. Lett.,
estimation for intelligent reflecting surface aided
vol. 10, no. 5, pp. 986–990, 2021.
multiuser communication, arXiv preprint arXiv:
[147] Y. Jin, J. Zhang, X. Zhang, H. Xiao, B. Ai, and D. W. K.
2102.10886, 2021.
Ng, Channel estimation for semi-passive reconfigurable
[136] B. Zheng and R. Zhang, Intelligent reflecting surface-
intelligent surfaces with enhanced deep residual
enhanced OFDM: Channel estimation and reflection
networks, IEEE Trans. Veh. Technol., vol. 70, no. 10, pp.
optimization, IEEE Wireless Commun. Lett., vol. 9, no. 4,
11083–11088, 2021.
pp. 518–522, 2020.
[148] R. Schroeder, J. He, and M. Juntti, Passive RIS vs. hybrid
[137] S. Lin, B. Zheng, G. C. Alexandropoulos, M. Wen, F. -J.
RIS: A comparative study on channel estimation, in Proc.
Chen, and S. Mumtaz, Adaptive transmission for
2021 IEEE 93rd VTC-Spring, Helsinki, Finland, 2021, pp.
reconfigurable intelligent surface-assisted OFDM
1–7.
wireless communications, IEEE J. Sel. Areas Commun.,
[149] K. Ardah, S. Gherekhloo, A. L. F. D. Almeida, and M.
vol. 38, no. 11, pp. 2653–2665, 2020.
Haardt, TRICE: A channel estimation framework for
[138] Y. Yang, B. Zheng, S. Zhang, and R. Zhang, Intelligent
RIS-aided millimeter-wave MIMO systems, IEEE Signal
reflecting surface meets OFDM: Protocol design and rate
Process. Lett., vol. 28, pp. 513–517, 2021.
maximization, IEEE Trans. Commun., vol. 68, no. 7, pp. [150] J. Mirza and B. Ali, Channel estimation method and
4522–4535, 2020. phase shift design for reconfigurable intelligent surface
[139] C. You, B. Zheng, and R. Zhang, Channel estimation and
assisted MIMO networks, IEEE Trans. Cogn. Commun.
passive beamforming for intelligent reflecting surface: Netw., vol. 7, no. 2, pp. 441–451, 2021.
Discrete phase shift and progressive refinement, IEEE J. [151] H. Liu, J. Zhang, Q. Wu, H. Xiao, and B. Ai, ADMM
Sel. Areas Commun., vol. 38, no. 11, pp. 2604–2620, based channel estimation for RISs aided millimeter wave
2020. communications, IEEE Commun. Lett., vol. 25, no. 9, pp.
[140] T. V. Chien, H. Q. Ngo, S. Chatzinotas, M. D. Renzo, 2894–2898, 2021.
and B. Ottersten, Reconfigurable intelligent surface- [152] M. Jian and R. Liu, Baseband signal processing for
assisted cell-free massive MIMO systems over spatially- terahertz: Waveform design, modulation and coding, in
correlated channels, arXiv preprint arXiv: 2104.08648, Proc. 2021 IEEE Int. Wireless Commun. Mob. Comput.,
2021. Harbin, China, 2021, pp. 1710–1715.
[141] K. Zhi, C. Pan, H. Ren, K. Wang, M. Elkashlan, M. D. [153] S. Ma, W. Shen, J. An, and L. Hanzo, Wideband channel
Renzo, R. Schober, H. V. Poor, J. Wang, and L. Hanzo, estimation for IRS-aided systems in the face of beam
Two-timescale design for reconfigurable intelligent squint, IEEE Trans. Wireless Commun., vol. 20, no. 10,
surface-aided massive MIMO systems with imperfect pp. 6240–6253, 2021.
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 29

[154] Y. Liu, S. Zhang, F. Gao, J. Tang, and O. A. Dobre, Wireless Commun. Lett., vol. 10, no. 4, pp. 790–794,
Cascaded channel estimation for RIS assisted mmWave 2020.
MIMO transmissions, arXiv preprint arXiv: 2106.10405, [166] J. He, H. Wymeersch, and M. Juntti, Leveraging location
2021. information for RIS-aided mmWave MIMO
[155] Y. Chen, D. Chen, and T. Jiang, Beam-squint mitigating communications, IEEE Wireless Commun. Lett., vol. 10,
in reconfigurable intelligent surface aided wideband no. 7, pp. 1380–1384, 2021.
mmWave communications, in Proc. 2021 IEEE WCNC, [167] K. Keykhosravi, M. F. Keskin, G. Seco-Granados, and H.
Nanjing, China, 2021, pp. 1–6. Wymeersch, SISO RIS-enabled joint 3D downlink
[156] B. Al-Nahhas, Q. -U. -A. Nadeem, and A. Chaaban, localization and synchronization, in Proc. 2021 IEEE
Intelligent reflecting surface assisted MISO downlink: ICC, Montreal, Canada, 2021, pp. 1–6.
Channel estimation and asymptotic analysis, in Proc. [168] Y. Liu, E. Liu, R. Wang, and Y. Geng, Reconfigurable

2020 IEEE GLOBECOM, Taiwan, Province of China, intelligent surface aided wireless localization, in Proc.
2020, pp.1–6. 2021 IEEE ICC, Montreal, Canada, 2021, pp. 1–6.
[169] L. Wang, N. Shlezinger, G. C. Alexandropoulos, H.
[157] S. E. Zegrar, L. Afeef, and H. Arslan, Reconfigurable
Zhang, B. Wang, and Y. C. Eldar, Jointly learned symbol
intelligent surface (RIS): Eigenvalue decomposition-
detection and signal reflection in RIS-aided multi-user
based separate channel estimation, in Proc. 2021 IEEE
MIMO systems, in Proc. 2021 55th Asilomar Conf.
32nd PIMRC, Helsinki, Finland, 2021, pp. 1–6.
[158] Z. Zhou, N. Ge, Z. Wang, and L. Hanzo, Joint transmit
Signals, Sys., Comp., Pacific Grove, CA, USA, 2021, pp.
715–721.
precoding and reconfigurable intelligent surface phase
[170] N. K. Kundu and M. R. McKay, Channel estimation for
adjustment: A decomposition-aided channel estimation
reconfigurable intelligent surface aided MISO
approach, IEEE Trans. Commun., vol. 69, no. 2, pp.
communications: From LMMSE to deep learning
1228–1243, 2020.
solutions, IEEE Open J. Commun. Soc., vol. 2, pp.
[159] G. T. D. Araujo, A. L. D. Almeida, and R. Boyer,
471–487, 2021.
Channel estimation for intelligent reflecting surface
[171] E. Shtaiwi, H. Zhang, A. Abdelhadi, and Z. Han, RIS-
assisted MIMO systems: A tensor modeling approach,
assisted mmWave channel estimation using
IEEE J. Sel. Areas Commun., vol. 15, no. 3, pp. 789–802,
convolutional neural networks, in Proc. 2021 IEEE
2021.
WCNCW, Nanjing, China, 2021, pp. 1–6.
[160] R. A. Harshman and M. E. Lundy, PARAFAC: Parallel
[172] S. Gao, P. Dong, Z. Pan, and G. Y. Li, Deep multi-stage
factor analysis, Comput. Stat. & Data Analysis, vol. 18,
CSI acquisition for reconfigurable intelligent surface
no. 1, pp. 39–72, 1994.
aided MIMO systems, IEEE Commun. Lett. , vol. 25, no.
[161] L. Wei, C. Huang, G. C. Alexandropoulos, C. Yuen, Z.
6, pp. 2024–2028, 2021.
Zhang, and M. Debbah, Channel estimation for RIS- [173] W. Yang, H. Li, M. Li, Y. Liu, and Q. Liu, Channel
empowered multi-user MISO wireless communications, estimation for practical IRS-assisted OFDM systems, in
IEEE Trans. Commun., vol. 69, no. 6, pp. 4144–4157, Proc. 2021 IEEE WCNCW, Nanjing, China, 2021, pp.
2021. 1–6.
[162] J. Yuan, G. C. Alexandropoulos, E. Kofidis, T. L. Jensen, [174] Y. Liu, E. Liu, R. Wang, and Y. Geng, Channel
and E. D. Carvalho, Tensor-based channel tracking for estimation and power scaling of reconfigurable intelligent
RIS-empowered multi-user MIMO wireless systems, surface with non-ideal hardware, in Proc. 2021 IEEE
arXiv preprint arXiv: 2202.08315, 2022. WCNC, Nanjing, China, 2021, pp. 1–6.
[163] Z. Mao, M. Peng, and X. Liu, Channel estimation for [175] B. Li, Z. Zhang, Z. Hu, and Y. Chen, Joint array
reconfigurable intelligent surface assisted wireless diagnosis and channel estimation for RIS-aided mmWave
communication systems in mobility scenarios, China MIMO system, IEEE Access, vol. 8, pp. 193992–194006,
Commun., vol. 18, no. 3, pp. 29–38, 2021. 2020.
[164] M. Xu, S. Zhang, J. Ma, and O. A. Dobre, Deep learning- [176] G. Zhou, C. Pan, H. Ren, K. Wang, and A. Nallanathan,
based time-varying channel estimation for RIS assisted A framework of robust transmission design for IRS-aided
communication, arXiv preprint arXiv: 2108.05661, 2021. MISO communications with imperfect cascaded
[165] S. Sun and H. Yan, Channel estimation for channels, IEEE Trans. Signal Process., vol. 68, pp.
reconfigurable intelligent surface-assisted wireless 5092–5106, 2020.
communications considering doppler effect, IEEE [177] M. Jung, W. Saad, Y. Jang, G. Kong, and S. Choi,
 30 Intelligent and Converged Networks, 2022, 3(1): 1−32

Performance analysis of large intelligent surfaces (LISs): [184] L. Wei, C. Huang, G. C. Alexandropoulos, and C. Yuen,
Asymptotic data rate and channel hardening effects, Parallel factor decomposition channel estimation in RIS-
IEEE Trans. Wireless Commun., vol. 19, no. 3, pp. assisted multi-user MISO communication, in Proc. 2020
2052–2065, 2020. IEEE 11th SAM, Hangzhou, China, 2020, pp. 1–5.
[178] J. Zhang, J. Liu, S. Ma, C. -K. Wen, and S. Jin, Large [185] M. Jian, R. Liu, and Y. Chen, Standardization for
system achievable rate analysis of RIS-assisted MIMO reconfigurable intelligent surfaces: Progresses,
wireless communication with statistical CSIT, IEEE challenges and the road ahead, in Proc. 2021 IEEE/CIC
Trans. Wireless Commun., vol. 20, no. 9, pp. 5572–5585, Int. Conf. Commun. China, Xiamen, China, 2021, pp.
2021. 337–342.
[179] G. T. D. Araújo and A. L. F. D. Almeida, PARAFAC- [186] FuTURE Forum, http://www.future-forum.org/en/index.
based channel estimation for intelligent reflective surface asp, 2021.
assisted MIMO system, in Proc. 2020 IEEE 11th SAM, [187] China Communications Standards Association, Meeting
Hangzhou, China, 2020, pp. 1–5. detail of TC5 WG6 meeting 55, http://www.ccsa.org.cn/
[180] P. Wang, J. Fang, H. Duan, and H. Li, Compressed meetingDetail, 2020.
channel estimation for intelligent reflecting surface- [188] ETSI, Work item descriptions of new WIs, https://
assisted millimeter wave systems, IEEE Signal Process. portal.etsi.org, 2022.
Lett., vol. 27, pp. 905–909, 2020. [189] ITU-R, Future technology trends of terrestrial IMT
[181] J. Zhang and M. Haardt, Channel estimation and training Systems towards 2030 and beyond, https://www.itu.
design for hybrid multi-carrier mmWave massive MIMO int/md/R19-WP5D-210607-TD-0400, 2021.
systems: The beamspace ESPRIT approach, in Proc. [190] ZTE Corporation, Sanechips, Support of reconfigurable
2017 25th EUSIPCO, Kos, Greece, 2017, pp. 385–389. intelligent surface for 5G Advanced, https://www.3gpp.
[182] W. Han, P. Chen, and Z. Cao, Deep neural network- org/ftp/TSG_RAN/TSG_RAN/TSGR_91e/Docs/RP-2106
based quantized signal reconstruction for DOA 18.zip, 2021.
estimation, arXiv preprint arXiv: 2005.01102, 2020. [191] R. Liu, Q. Wu, M. D. Renzo, and Y. Yuan, A path to
[183] Y. Jin, J. Zhang, B. Ai, and X. Zhang, Channel smart radio environments: An industrial viewpoint on
estimation for mmWave massive MIMO with reconfigurable intelligent surfaces, IEEE Wireless
convolutional blind denoising network, IEEE Commun. Communications, vol. 29, no. 1, pp. 202–208, 2022.
Lett., vol. 24, no. 1, pp. 95–98, 2020.

Ruiqi Liu received the BS and MS Chau Yuen received the BEng and PhD
degrees (with honors) in electronic degrees from Nanyang Technological
engineering from Tsinghua University in University (NTU), Singapore, in 2000 and
2016 and 2019, respectively. He is now a 2004, respectively. He was a postdoctoral
master researcher in the Wireless Research researcher with Lucent Technologies Bell
Institute of ZTE Corporation, responsible Labs at Murray Hill in 2005. From 2006 to
for long-term research as well as 2010, he was with the Institute for
standardization. His research interests Infocomm Research (I2R), Singapore.
include reconfigurable intelligent surfaces, wireless positioning, Since 2010, he has been with the Singapore University of
quantum communication, and visible light communication. He is Technology and Design. He received the IEEE Marconi Prize
the author or co-author of one book and one book chapter. He Paper Award in Wireless Communications and EURASIP Best
has authored and submitted more than 500 technical documents Paper Award for Journal on Wireless Communications and
to 3GPP TSG RAN4, where he served as the co-rapporteur of Networking on 2021, IEEE Asia Pacific Outstanding Young
the work item (WI) on NR RRM enhancement and the feature Researcher Award on 2012, and IEEE VTS Singapore Chapter
lead of multiple features. He currently serves as the vice chair of Outstanding Service Award on 2019. He serves as an editor for
ISG RIS in the ETSI. He actively participates in organizing IEEE Transactions on Communications and IEEE Transactions
committees, technical sessions, workshops, symposia, and on Vehicular Technology, awarded as the top editor during
industrial panels in IEEE conferences as the chair, organizer, 2009−2015. He also served as the guest editor for several special
moderator, or panelist. He serves as the editor of ITU Journal of issues, including IEEE Journal on Selected Areas in
Future and Evolving Technologies (ITU J-FET) and the Communications, IEEE Wireless Communications Magazine,
associate editor of IET Quantum Communication. He is the IEEE Communications Magazine, IEEE Vehicular Technology
standardization officer for IEEE ComSoc ETI on reconfigurable Magazine, IEEE Transactions on Cognitive Communications
intelligent surfaces (ETI-RIS) and the standards liaison officer and Networking, and Elsevier Applied Energy. He is currently an
for IEEE ComSoc Signal Processing and Computing for IEEE fellow and distinguished lecturer of IEEE Vehicular
Communications Technical Committee (SPCC-TC). Technology Society.
 Mengnan Jian et al.: Reconfigurable intelligent surfaces for wireless communications: Overview of hardware designs… 31

Mengnan Jian received the BS degree in George C. Alexandropoulos received the

information engineering from Beijing engineering diploma, MASc, and PhD
Institute of Technology, Beijing, China, in degrees in computer engineering and
2016 and the MS degree from Tsinghua informatics from University of Patras,
University, Beijing, China, in 2019. She is Greece in 2003, 2005, and 2010,
currently an engineer in ZTE Cooperation. respectively. He has held research
Her research interests include orbital positions at various Greek universities and
angular momentum and reconfigurable research institutes, as well as at the
intelligent surfaces. Mathematical and Algorithmic Sciences Lab, Paris Research
Center, Huawei Technologies France, and he is currently an
Yuanwei Liu received the BS and MS assistant professor with the Department of Informatics and
degrees from Beijing University of Posts Telecommunications, National and Kapodistrian University of
Athens (NKUA), Greece. His research interests span the general
and Telecommunications in 2011 and
areas of algorithmic design and performance analysis for
2014, respectively, and the PhD degree in wireless networks with emphasis on multi-antenna transceiver
electrical engineering from Queen Mary hardware architectures, active and passive reconfigurable
University of London, UK, in 2016. He metasurfaces, millimeter wave communications, as well as
was with the Department of Informatics, distributed machine learning algorithms. He currently serves as
King’s College London, from 2016 to an editor for IEEE Transactions on Wireless Communications,
2017, where he was a postdoctoral researcher. He has been a IEEE Communications Letters, IEEE Wireless Communications
Letters, ELSEVIER Computer Networks, Frontiers in
senior lecturer (associate professor) with the School of Communications and Networks, and the ITU Journal on Future
Electronic Engineering and Computer Science, Queen Mary and Evolving Technologies, as well as a guest editor for various
University of London, since 2021. His research interests include IEEE journals. He is a senior member of the IEEE
Communications, Signal Processing, and Information Theory
non-orthogonal multiple access, 5G/6G networks, machine
Societies as well as a professional engineer of the Technical
learning, and stochastic geometry. Chamber of Greece. He has participated and/or technically
He is a Web of Science Highly Cited Researcher 2021. He is managed more than 10 European Union (EU) research and
currently a senior editor of IEEE Communications Letters, an innovation projects, as well as several Greek and international
research projects. He is currently NKUA’s principal investigator
editor of the IEEE Transactions on Wireless Communications for the EU H2020 RISE-6G research and innovation project
and the IEEE Transactions on Communications. He serves as dealing with RIS-empowered smart wireless environments. He
the leading guest editor for IEEE JSAC special issue on next has received the best PhD thesis award in 2010, the IEEE
generation multiple access, a guest editor for IEEE JSTSP Communications Society Best Young Professional in Industry
Award in 2018, the EURASIP Best Paper Award of the Journal
special issue on signal processing advances for non-orthogonal on Wireless Communications and Networking in 2021, the IEEE
multiple access in next generation wireless networks. He has Marconi Prize Paper Award in Wireless Communications in
served as a TPC member for many IEEE conferences, such as 2021, and the Best Paper Award from the IEEE GLOBECOM
GLOBECOM and ICC. He received IEEE ComSoc Outstanding in 2021. More information is available at
www.alexandropoulos.info.
Young Researcher Award for EMEA in 2020. He received the
2020 Early Achievement Award of the IEEE ComSoc Signal Ertugrul Basar received the BS degree
Processing and Computing for Communications (SPCC) from Istanbul University, Turkey, in 2007,
Technical Committee. He received 2021 Early Achievement and the MS and PhD degrees from Istanbul
Award of IEEE Communication Theory Technical Committee Technical University, Turkey, in 2009 and
2013, respectively. He is currently an
(CTTC). He received the Exemplary Reviewer Certificate of
associate professor with the Department of
IEEE Wireless Communications Letters in 2015, IEEE Electrical and Electronics Engineering,
Transactions on Communications in 2016 and 2017, and IEEE Koç University, Istanbul, Turkey and the
Transactions on Wireless Communications in 2017 and 2018. director of Communications Research and Innovation
He has served as the publicity co-chair for VTC 2019-Fall. He is Laboratory (CoreLab). His primary research interests include
MIMO systems, index modulation, intelligent surfaces,
the leading contributor for “Best Readings for Non-Orthogonal
waveform design, visible light communications, and signal
Multiple Access (NOMA)” and the primary contributor for processing for communications. He currently serves as a senior
“Best Readings for Reconfigurable Intelligent Surfaces (RIS)”. editor of IEEE Communications Letters and the editor of IEEE
He serves as the chair of Special Interest Group (SIG) in SPCC Transactions on Communications and Frontiers in
Communications and Networks. He is a senior member of IEEE
Technical Committee on the topic of signal processing and the academic chair of IEEE ComSoc Emerging
techniques for next generation multiple access (NGMA), the Technologies Initiative on Reconfigurable Intelligent Surfaces.
vice-chair of SIG Wireless Communications Technical He is the author/co-author of around 160 journal and conference
Committee (WTC) on the topic of Reconfigurable Intelligent publications and 8 patents (granted/applied), which received
around 8000 citations. The most recent recognition of his
Surfaces for Smart Radio Environments (RISE), and the tutorials research includes the IEEE Communications Society Best
and invited presentations officer for Reconfigurable Intelligent Young Researcher Award for the Europe, Middle East, and
Surfaces Emerging Technology Initiative. Africa Region in 2020.
 32 Intelligent and Converged Networks, 2022, 3(1): 1−32

Chongwen Huang received the BSc

degree from Nankai University, Binhai
College, in 2010 and the MSc degree from
University of Electronic Science and
Technology of China (UESTC, Chengdu)
in 2013. He joined the Institute of
Electronics, Chinese Academy of Sciences
(IECAS, Beijing) as a research engineer in
2013. He received the PhD degree from Singapore University of
Technology and Design (SUTD, Singapore) and
CentraleSupélec University (Paris, France), in 2019. From 2019
to 2020, he was a postdoctoral researcher in SUTD. Since Sep.
2020, he joined into Zhejiang University as a tenure-track young
professor.
He is the recipient of IEEE Marconi Prize Paper Award in
Wireless Communications, and IEEE ComSoc Asia-Pacific
Outstanding Young Researcher Award in 2021. He was also a
recipient of Singapore Government PhD scholarship, and PHC
Merlion PhD Grant 2016−2019 for studying in CentraleSupélec,
France. He has served as an editor of IEEE Communications
Letter, Elsevier Signal Processing, EURASIP Journal on
Wireless Communications and Networking, and Physical
Communication since 2021. In addition, He also has served the
chair of some wireless communications flagship conference,
including session chair of 2021 IEEE WCNC, 2021 IEEE VTC-
Fall, and symposium chair of IEEE WCSP 2021, etc. His main
research interests are focused on holographic MIMO
surface/reconfigurable intelligent surface, B5G/6G wireless
communication, mmWave/THz communications, deep learning
technologies for wireless communications, etc.