Elmossallamy 2020
Elmossallamy 2020
Elmossallamy 2020
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
1
Abstract—Recently there has been a flurry of research on amble unused spectra exist. However, the unreliable stochastic
the use of reconfigurable intelligent surfaces (RIS) in wireless nature of wireless propagation remains inevitable. Conven-
networks to create smart radio environments. In a smart radio tional wisdom regards the wireless channel as an uncontrol-
environment, surfaces are capable of manipulating the prop-
agation of incident electromagnetic waves in a programmable lable stochastic link with inherent unreliability. Therefore, the
manner to actively alter the channel realization, which turns the best we can do is to understand it, model it, and combat
wireless channel into a controllable system block that can be its unpredictability with sophisticated signal processing at
optimized to improve overall system performance. In this article, the transmitter and the receiver. This traditionally includes
we provide a tutorial overview of reconfigurable intelligent diversity techniques, beamforming, and adaptive coding and
surfaces (RIS) for wireless communications. We describe the
working principles of reconfigurable intelligent surfaces (RIS) modulation to squeeze as much usable capacity as possible.
and elaborate on different candidate implementations using Recently, with the advent of reconfigurable intelligent surfaces
metasurfaces and reflectarrays. We discuss the channel models (RIS) and the emergence of the concept of the smart radio
suitable for both implementations and examine the feasibility of environment [1], [2], we might be able to also control, at least
obtaining accurate channel estimates. Furthermore, we discuss partially, the wireless channel itself.
the aspects that differentiate RIS optimization from precoding for
traditional MIMO arrays highlighting both the arising challenges The idea of controlling the ambient environment to pro-
and the potential opportunities associated with this emerging vide more favorable propagation characteristics represents
technology. Finally, we present numerical results to illustrate the a paradigm shift in how we think about wireless systems
power of an RIS in shaping the key properties of a MIMO design. Instead of treating reflection and scattering in the
channel. environment as uncontrollable phenomena whose effects can
only be modeled stochastically, they become part of the system
I. I NTRODUCTION parameters that may be optimized, which can overcome many
of the challenges of wireless communications.
N the relentless pursuit to increase the capacity of wireless
I networks and support higher and higher data rates, wireless
system designers have sought to optimize every aspect of
In general, the practically achievable rate over a wireless
link is limited by the order of the modulation and the number
of spatial streams. Both are decided according to the current
communication systems. This has started with more spec-
channel realization. The order of modulation is adapted ac-
trum efficient waveforms and multiplexing techniques (e.g.
cording to the signal strength perceived at the receiver, which
OFDM), leveraging the spatial domain (e.g. MIMO), and
is a consequence of the channel gain. To keep error rates low
inching ever closer to the maximum theoretical capacity using
and avoid re-transmissions, a user at the cell edge will be
more advanced adaptive modulation and coding techniques.
forced to use lower order modulation, and thus suffer from
On the network side, cellular networks have gotten denser with
lower rates. On the other hand, the number of spatial streams
more aggressive frequency reuse and inter-cell coordination
is adapted according to the number of the usable eigenmodes
techniques have been developed to deal with the resulting
of the channel. A line-of-sight (LOS) link may enjoy high
interference. However, the capacity of the network is still
channel gain, but will probably suffer from a spatially-sparse
ultimately limited by the unreliability of wireless propagation
low-rank channel, limiting the number of spatial streams
and the available spectrum.
and consequently the achievable rate. The two scenarios are
To tackle the shortage of spectrum, communications systems
illustrated in Fig. 1. These scenarios can arise in any wireless
have been steadily moving into higher frequency bands where
network, but future communication systems are expected to be
Mohamed A. ElMossallamy, Hongliang Zhang, and Zhu Han are with the especially affected. In particular, the propagation characteris-
Electrical and Computer Engineering Department, University of Houston, TX, tics at the higher frequency bands used in future generations
USA. Zhu Han is also with the Department of Computer Science and Engi-
neering, Kyung Hee University, Seoul, South Korea (emails: [email protected], of communications systems, e.g. 30-100 GHz, will give rise to
[email protected], [email protected]). these scenarios more often. Reconfigurable intelligent surfaces
Lingyang Song is with Department of Electronics, Peking University, can be used to alter the channel realization in these scenarios
Beijing, China (email: [email protected]).
Karim G. Seddik is with the Electronics and Communications Engineering and improve overall system performance dramatically.
Department, American University in Cairo, New Cairo, Egypt (email: ksed- Wireless communication systems have traditionally relied
[email protected]). on frequency bands extending from a few hundred megahertz
Geoffrey Ye Li is with the School of Electrical and Computer En-
gineering, Georgia Institute of Technology, Atlanta, GA, USA (email: to a few gigahertz. This was mainly driven by the favor-
[email protected]). able propagation characteristics at these bands and the ease
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
2
RIS
The above challenges can be overcome by leveraging the
power of RISs. In low-received power scenarios, the RISs can
AP function as centralized beamformer to increase channel gains
RIS or create a propagation path around major obstacles to restore
RIS a link in an outage. While in LOS spatially-sparse scenarios,
the RIS can be used to emulate a rich scattering environment
(B) to enhance the channel condition number and improve spatial
multiplexing capability. Different technologies can be used to
implement an RIS [5], [6]. In its most simple form, an RIS can
be implemented as a dynamic reflectarray [5], whose elements
are omnidirectional antennas with controllable termination that
RIS (A) can be changed dynamically to backscatter and phase shift the
incident waveform. A more elaborate implementation would
be using a dynamically tunable metasurface [6], a 2D planar
form of metamaterials that have been shown to possess great
RIS
electromagnetic wave manipulation capabilities. Relying on
Fig. 1. A smart radio environment with multiple RISs. User A is far away the metasurface implementation, an RIS element can not only
from the AP and suffers from low received signal strength, while user B has scatter and phase-shift the signal but can also act as an
amble received power but a low-rank ill-conditioned channel. The RISs can anomalous mirror with a controllable reflection angle and even
be optimized to help in both scenarios.
polarization manipulation abilities. The implementation can
also greatly affect the link-budget as reflected and scattered
waves suffer from different path loss scaling. However, as the
of implementing efficient and cheap transceivers. Although RIS element gains more wave manipulation capabilities, the
there exists a huge swath of spectrum at the millimeter-wave optimization task gets more complicated.
range (30-100 GHz), it has been traditionally sought that In this article, we provide a comprehensive tutorial on re-
these frequencies are not suitable for wireless communica- configurable intelligent surfaces for wireless communications.
tion, especially outdoor cellular communications. Subsequent We first discuss the different implementations of reconfig-
intensive measurement campaigns [3] have shown the potential urable intelligent surfaces and elaborate on their differences
of millimeter-wave bands when combined with directional with respect to wave manipulation capabilities, channel mod-
high gain antenna arrays. This has led to immense interest eling, and the link budget. We then discuss the challenges of
in millimeter wave wireless communications. optimizing the RIS parameters to benefit the overall system
This foray into higher frequency bands necessitated a reeval- performance, and also highlight the potential opportunities if
uation of channel modeling techniques. Communicating at the challenges can be overcome. Finally, we provide numerical
millimeter-wave frequencies poses a lot of challenges. The results to illustrate the power of an RIS in shaping some key
channel at these frequencies is significantly more hostile properties of a MIMO channel to enhance performance and
than at sub 6 GHz frequencies. Diffraction ceases to be a simplify transceiver designs.
reliable propagation mechanism with line-of-sight, first-order The rest of this article is organized as follows. In Section II,
reflections and scattering becoming much more dominant. This we introduce different implementations to realize reconfig-
means shadowing will have severe detrimental effects on the urable intelligent surfaces. In Section III, we discuss how
average received power. Indeed, channel models developed to incorporate the effects of the RISs in the channel model,
for millimeter-wave include a third state, in addition to line- how the implementation technology can influence the choice
of-sight and non-line-of-sight, to explicitly model an outage of the appropriate model, and how different implementations
event when received power is too weak to establish a link [3]. can affect the link budget. In Section IV, we review the
Although adaptive beam steering techniques can improve the state-of-the-art of RIS-assisted system optimization techniques
reliability of millimeter-wave links, communications at these and discuss the difference between the optimization of RIS
frequencies remain very challenging. configurations and precoding for traditional MIMO arrays
Another challenge on millimeter-wave channels is that they highlighting both the arising challenges and the potential
are spatially sparse which means that only a small number opportunities associated with this emerging technology. In
of propagation paths exist between the transmitter and the Section V, we present numerical results showing the power of
receiver. This is in contrast to the rich scattering assumption RIS in improving the spatial multiplexing capability of MIMO
usually utilized in sub 6 GHz channels. Although this sparsity channels and simplifying transceivers. Finally, we identify
could be leveraged in channel estimation and precoding, some future directions in Section VI and conclude the article
especially for hybrid analog/digital architectures [4], it also in Section VII.
limits the number of spatial data streams that can be supported
by the channel. Hence, the spatial multiplexing capability
II. R ECONFIGURABLE I NTELLIGENT S URFACES
of millimeter-wave channels is limited. In many cases, e.g.
LOS, only a single viable propagation path exists and spatial In a smart radio environment [1], shown in Fig. 1, one or
multiplexing is not feasible. more RISs can be used to influence wireless propagation in a
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
3
(a) A 48-element reflectarray-based RIS. Each element is a traditional antenna (b) A 4-element metasurface-based RIS. Each element/tile is
connected to a phase shifter. a dynamic metasurafce with numerous tightly-packed meta-
atoms and can apply an arbitrary quasi-continuous phase
gradient.
Fig. 2. Different Implementations of an RIS.
manner that is beneficial to the overall system performance. RIS is similar to a backscatter tag, the elements of an RIS
For example, by increasing received power through beamform- work collectively, based on knowledge of the propagation
ing or by influencing the channel rank and condition number to environment, over a very large area to induce significantly
facilitate spatial multiplexing. In essence, any passive surface more potent effects on the incident waves.
that can be dynamically reconfigured to manipulate incident The reflectarray-based RIS can be thought of as providing
electromagnetic waves and change the channel conditions powerful centralized analog beamforming capabilities in ad-
change can be called an RIS. This definition is valid regardless vantageous locations that can be utilized by communication
of the particular implementation. Two main implementations endpoints. This can also lead to much simpler transmitters
have been investigated in the literature; based on traditional re- and receivers by shifting complexity to the RIS and con-
flectarrays, or metasurfaces. Regardless of the implementation, trollers. Finally, note that the dimension of each element in
an RIS should be passive, that is it does not emit any power a reflectarray-based RIS is comparable to the wavelength, e.g.
𝜆
of its own and only aims to manipulate existing transmitted 2 , and individually act as a diffuse scatterer.
waves. In this aspect, RIS is similar to backscatter technology
and different from relaying. In this section, we elaborate on B. Metasurface-based Implementation
these RIS implementations.
A more sophisticated implementation of an RIS might
be done using metasurfaces [8]. A metasurface is the two-
A. Reflectarray-based Implementation dimensional planar form of metamaterials, which are man-
As shown in Fig. 2, the simplest way to implement a made synthetic materials with electromagnetic properties not
reconfigurable intelligent surface is to use a passive reflec- found in naturally occurring materials. They were originally
tarray whose elements’ antenna termination can be controlled conceived for applications in the optical domain to allow
electronically to backscatter and phase-shift the incident signal cheap and robust planar optical components replacing more
[5]. Each element individually has a very limited effect on the expensive custom-made lenses.
propagated waves, but a sufficiently large number of elements A metasurface is comprised of a large number of closely
can effectively manipulate the incident wave in a controllable spaced deeply subwavelength resonating structures called pix-
manner. To be effective, this implementation would require a els or meta-atoms [8]. Both individual meta-atoms and the
vastly large number of antenna elements, probably thousands space between adjacent meta-atoms are much smaller than
[7]. the wavelength in size. The very small size of these closely-
Each element in the reflectarray-based RIS is similar to a packed atoms and their large number offer a vast number of
tag in backscatter communications systems. However, there degrees of freedom in manipulating the incident electromag-
are two main differences. The first one is that the reflections netic waves. In particular, by deliberately designing its meta-
are used to communicate information from the reflector to atoms, a metasurface can impose arbitrary quasi-continuous
the receiver in backscatter communications, while the RIS [9] amplitude/phase profiles on the incident wave-fronts and
aims only to help the ongoing transmission and does not exercise fine-grained control over the scattered electric field.
communicate information of its own in an RIS-assisted com- Earlier metasurfaces designs have been based on static
munication scenario. The other one is the size and scale of preset meta-atoms designs that cannot be modified after fab-
the RIS. Although a single element in a reflectarray-based rication, which is good enough for making custom lenses
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
4
√ √
for optical applications. However, later designs rely on semi- receiver through the RIS. 𝐺 𝑑 and 𝐺 𝑟 represent the large-
conductor components, which can be reconfigured in real- scale gain
√ associated√with Henv and HRIS , respectively. Finally,
time to change the underlying meta-atom structure and hence Heff B 𝐺 𝑑 Henv + 𝐺 𝑟 HRIS represents the overall effective
the electromagnetic behavior of the metasurface [6]. This RIS-augmented channel seen by the transceivers.
reconfigurability is achieved by integrating components that It is worth discussing how the effective channel in RIS-
can be tuned either electrically, mechanically, or thermally. assisted communications differs from the effective channel
Electrically tunable metasurfaces are especially attractive since in traditional precoded multi-antenna systems. In traditional
they can be cheaply manufactured using well-understood semi- precoded systems, the equivalent channel is given by the
conductor technologies, and can be tuned fast enough to adapt product Heff = Henv P, where P is the applied precoder. While
to the time-varying wireless channel. For instance, by incorpo- in RIS-assisted communications, the RIS effect is additive,
rating varactor diodes or liquid crystals within the meta-atoms. cf. (1). Furthermore, elements of a traditional precoder can
This dynamic tunability is paramount in wireless applications take any value satisfying some constraints, e.g. power, while
to allow adapting to the changing channel realization. the additive term in RIS effective channel depends on the
A metasurface-based RIS is comprised of several tiles, propagation environment and is only partially controllable
where each tile is an individually reconfigurable metasurface via the RIS. Although this makes the RIS optimization task
whose dimensions are much larger than the wavelength. In more challenging in general, the additive effect of the RIS
a sense, each individual element in a metasurface-based RIS is more potent than the mixing and steering effects of the
has functionality akin to a reflectarray on its own right. In traditional multiplicative precoders [14], giving the system
particular, you can consider each tile in a reconfigurable designer more control over the effective channel. Next, we
metasurface as the limit of a reflectarray as both the an- present two techniques to model the channel through the RIS,
tenna size and antenna spacing diminish and the applied HRIS .
amplitude/phase profile becomes approximately continuous
along the surface. This allows a large degree of flexibility
in manipulating the incident wave-front. For example, each A. Dyadic Backscatter Channel Model
tile can reflect the incident wave-front in a different direction. The first technique models the RIS at the scatterer level and
However, the majority of empirical works in the literature assumes each element in the RIS is a regular omnidirectional
[7], [10]–[13] employ reflectarray-based RIS. Even with the antenna subject to the effects of fading and uses the so-called
simpler reflectarray-based implementation, impressive results dyadic backscatter channel [15] to model the channel through
have been reported. In theory, the increased flexibility associ- the RIS. Using this model, the channel through the RIS, HRIS ,
ated with the metasurface-based RIS leads to even better re- can be written as
sults; however, only real-world experimentation will tell if the
increased sophistication would lead to practical performance HRIS = FQG, (2)
gains.
where F is the 𝑁 × 𝐿 channel from the RIS to the receiver,
III. C HANNEL M ODEL G is the 𝐿 × 𝑀 channel from the transmitter to the RIS. The
In this section, we discuss how to incorporate the existence 𝐿 × 𝐿 matrix, Q, represents the interaction of the RIS with
of an RIS in the propagation environment into the channel the transmitted waveform. Assuming no coupling between the
model. Accurate channel models are essential for the analytical RIS elements, the interaction matrix can be written as
study of RIS-assisted communications as well as simulations.
This is paramount in the evaluation of the utility of the Q = diag 𝛽1 e𝑖 𝜃1 , 𝛽2 e𝑖 𝜃2 , . . . , 𝛽 𝐿 e𝑖 𝜃𝐿 , (3)
RIS, especially as it compares to other technologies, such
as relaying. The implementation method of the RIS, whether where 𝛽𝑖 ∈ [0, 1] and 𝜃 𝑖 ∈ [0, 2𝜋) and can be controlled by
its reflectarrays or metasurfaces, may play a role in deciding changing the complex antenna load similar to backscatter tags
which modeling technique is more appropriate. We present [16]. Note that the phase shifts can either be continuous [10],
two modeling techniques used in the literature and discuss or discrete [5], [17] based on implementation, e.g., varactors
their underlying assumptions and rationale. vs. switched loads.
Consider a scenario where an 𝑀-antenna transmitter com- From (2), for any assumed statistical distribution for the
municates with an 𝑁-antenna receiver and an 𝐿-element RIS entries of G and F, the overall channel distribution will be
exists in the environment. Assuming a narrowband flat-fading given by their product distribution. In general, this kind of
scenario, the received signal can be written as cascaded fading is known to have more detrimental effects on
p p performance compared to regular fading. However, increasing
y = 𝐺 𝑑 Henv x + 𝐺 𝑟 HRIS x + n,
(1) the number of backscattering elements, 𝐿, does improve the
= Heff x + n, fading characteristics [15]. More importantly, it is probably
where x is the transmitted vector, n is the white Gaussian not an adequate model for the metasurface-based RIS, where
noise vector at the 𝑁 receiver’s antennas. Henv represents each element is not a typical antenna but rather a planar surface
the uncontrollable channel between the transmitter and the with dimensions much larger the wavelength. Hence, typical
receiver, excluding the effects of the RIS, and HRIS repre- statistical characterization of the received signal envelope by
sents the controllable channel between the transmitter and the an antenna, i.e. fading, may not be appropriate.
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
5
(a) Large-scale path loss of diffusely scattered electromagnetic (b) Large-scale path loss of reflected electromagnetic waves.
waves.
Fig. 3. Implementation effects on large-scale pathloss.
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
6
Hence, in the case of the metasurface-based RIS, the path straightforward strategy is to use the RIS to maximize the
loss through a single RIS element will be proportional to the effective channel gain, which is equivalent to maximizing
overall distance [21], [23], 𝑑sr + 𝑑rd , i.e., the received power or minimizing the transmit power for a
1 given SNR constraint. If the relevant channels are known,
PLreflected ∝ , (5) then co-phasing all paths is optimal, which can be solved
(𝑑sr + 𝑑rd ) 𝑛
analytically or formulated as a semidefinite program (SDP)
where 𝑛 is the path loss exponent, e.g. 2 in free space, while [32], [33] in case of continuous phase shifts. If phase shifts
in the case of the reflectarray-based RIS, the path loss will be are discrete, quantizing the continuous solution is near-optimal
proportional to the product of the distances, 𝑑sr and 𝑑rd , i.e., or a greedy iterative search can be utilized as indicated in
1 [37]. When the relevant channels are not known, the RIS
PLscattered ∝ . (6) phase shifts can be optimized based on feedback from the
(𝑑sr × 𝑑rd ) 𝑛
intended receiver, which can be implemented as a beam search
It is evident that the difference in path loss between the two procedure [31], or similar algorithms [7]. Furthermore, when
cases can be immense. limited channel information or receiver position information
Before we conclude this section, it is worth mentioning that can be obtained, they can be used to configure the RIS. In [57],
path loss scaling through the RIS has been a subject of debate a small subset of the relevant channels is used to infer good
in recent literature [26]–[28]. In particular, the assumption RIS configurations based on a deep neural network when the
that a metasurface acts approximately as a specular reflector channels are generated using realistic ray-tracing simulation.
is only valid under specific conditions related to physical In [35], a deep neural network was also used but relying on
size of the RIS and distances to communication endpoints, position data rather than partial channel estimates.
i.e., 𝑑sr and 𝑑rd . In particular, the RIS physical size has In multi-user scenarios, the problem of configuring RIS
to be large enough relative to the distances, 𝑑sr and 𝑑rd , complicates considerably. In particular, maximizing the chan-
such that the communication endpoints are within the near- nel gains is no longer the only goal since interference must
field of the RIS. In this case, it is valid to assume that the be taken into account. Two heuristic algorithms have been
average received power through the RIS scales with the sum proposed to minimize transmit power under users’ SINR
of the distances to the RIS as predicted by (5). This has constraints in a downlink broadcast scenario in [50], one
been analyzed theoretically in [26], [27], [29] and observed is based on alternating optimization similar to [32] and the
in empirical measurements conducted in [27]. In envisioned other maximizes the weighted channel gains of all users,
RIS deployments [1], where an RIS covers large sections of where weights are proportional to SINR constraints. In [43], a
buildings’ facades outdoors and walls indoors, it is safe to heuristic algorithm based on majorization-minimization (MM)
assume that the transceivers of interest would be in the near- is used to maximize the sum achievable rate under user
field of the RIS [30], especially in dense networks. For a rate constraints, when a zero-forcing precoder is applied to
concrete example, consider an RIS with physical dimensions cancel all inter-user interference. The same approach is gen-
of 1.5𝑚 × 1.5𝑚 deployed on the facade of a building and a eralized to accommodate energy efficiency maximization in
transmitted signal at 28 GHz. The near-field of the RIS extends [44]. Motivated by the fact that the Transmitter-RIS-Receiver
2
up to 2𝐷𝜆 meters away, where 𝐷 is the largest dimension of the link resembles a hybrid beamforming scenario, where the
RIS and 𝜆 the wavelength. Hence, transceivers within 420𝑚 transmitter applies precoding in the digital domain and the
of the RIS are within its near-field, and path loss through the RIS applies further analog beamforming using only phase
RIS is well approximated by (5) [27]. shifts, a hybrid beamforming framework has been proposed
in [52] to maximize the achievable sum-rate in multi-user
IV. RIS- ASSISTED O PTIMIZATION downlink setting. Weighted sum-rate maximization has been
The presence of RISs in the propagation environment tackled recently in [45], where efficient suboptimal solutions
provides the system designer with vast abilities to alter the are found based on fractional programming and alternating
wireless channel realization to achieve different objectives in direction method of multipliers (ADMM). A deep reinforce-
various scenarios. In this section, we provide a brief overview ment learning (DRL) framework to jointly optimize transmitter
of the state of the art and discuss some of the aspects that precoding and RIS phase shifts has been developed in [51] and
differentiate RIS optimization from precoding for traditional found to achieve comparable performance to the alternating
MIMO arrays. optimization techniques proposed earlier in the literature.
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
7
on devising algorithms to configure the RIS given channel be very hard in practical scenarios.
knowledge. The rationale for this assumption is usually to The other approach is to equip a small subset of the RIS
explore the upper bounds of what can be achieved in practical elements with low-power active receivers that are capable of
systems which would have to rely on imperfect estimates processing pilots and estimating the relevant channels. Hence,
of the channels. However, different from traditional wireless the transmitter and the receiver can send pilots to help the
systems where channel acquisition is a straightforward matter RIS acquire partial information about the channel. Note that
and could be accounted for by penalizing the achievable rates the directly estimated channels constitute a tiny subset of
to account for pilot overhead and SNR degradation from the elements of the RIS and do not provide enough accurate
imperfect channel estimation, RIS-assisted wireless systems information on their own to facilitate beamforming or other
would encounter more difficulties given the passive nature channel manipulations. However, by leveraging tools from
of the RIS and the massive number of channel parameters compressed sensing and machine learning, the limited avail-
to be estimated. Hence, it should be noted that the gains able information can be used to compute sufficiently accurate
reported for these schemes principally relies on the ability estimates of the channels at all elements, or directly use the
to estimate the channels between the RIS elements and the partial channel information to infer good RIS configurations.
communication endpoints. Such channel estimation may prove Encouraging results in this direction have been reported in
to be tremendously difficult in practice. [57], where it has been shown that only 1% of the RIS
The majority of published work on RIS assume a elements need to be active in practical propagation scenarios
reflectarray-based implementation, where each element is an to support a sufficiently accurate prediction of the channels at
antenna terminated by a configurable load. As such, RIS the remaining elements. This can be attributed to the fact that
elements are neither regular wireless transmitters nor receivers. the channels at all elements result from the interactions of a
They do not possess the capability to send pilots or even limited number of scatterers and reflectors in the environment
process pilots to estimate the channel. Adding a receive chain and this limited number of receivers can be regarded as sensors
for each element in the RIS is prohibitive. A large number providing descriptors of the environment that carry sufficient
of receive chains, one for each element, would consume a information to guide the configuration of the RIS. Note that
formidable amount of power, and severely limits the attractive- this technique is somewhat related to [35], where the small
ness of RIS assisted communications. Without receive chains, number of “environment descriptors” can be thought of as
the RIS would not be able to process pilots sent from the two providing a spatial fingerprint indicating the position of the
communicating ends to estimate the channel. Moreover, the user.
effect of a single RIS element on the received signal is minus- The discussion above highlighted the problem of acquiring
cule; hence, turning the RIS element individually on and off phase information to optimize the state of the RIS. Indeed,
to measure their effects on the received signal is impractical. in all currently available prototypes and testbeds [7], [10],
Next, we present two approaches from the literature to tackle [11], [13], [31], [59], receiver-side measurements, e.g. signal
the problem of optimizing the RIS configuration with limited strength or received power, is used to guide the configura-
channel information. tion of the RIS. In these cases, there is a single user and
The first approach is to forgo channel estimation altogether; the problem of optimizing the RIS state is more akin to
instead, the optimization of the RIS can be based on feedback analog beam training. In a sense, the RIS can be thought
from the receiver. In [31], a three-way beam-searching proto- of as a large beamformer disjoint from the transmitter. One
col has been designed to jointly optimize the beam-direction of attractive aspect of some of these techniques is that they
an IEEE 802.11ad access point and an RIS towards a common require no modifications to the used wireless standard. In
receiver. This can be done using a predefined codebook of particular, the RIS training process piggybacks on normal
beam-directions; however, the size of the codebook will be data-carrying transmissions and gradually improves the quality
proportional to the number of elements. If there are a large of the received signal without requiring dedicated training
number of RIS elements, training may take a long time [58]. signaling. Finally, it is worth noting that these simple proof-of-
This can be a problem in a dynamic environment where the concept testbeds have used traditional antennas to implement
channel coherence time is limited. Nevertheless, relying on a reflectarray-based RIS and only optimized the phase shifts;
feedback from the receiver is a promising solution and the it remains to be seen if using metasurfaces to implement the
majority of empirically oriented studies have built prototype RIS would lead to higher performance gains in real life.
systems that rely on receiver feedback to guide the optimiza- Before we conclude the discussion of channel state in-
tion of the RIS configuration without explicit knowledge of formation, we would like to point out that the presence of
the channels involved. In [13], the received signal strength the RIS does not affect the channel estimation procedure
indicator (RSSI) is used to judge the current configuration between the transmitter and the receiver in the environment.
of the RIS and a simple greedy algorithm is devised to In particular, for any configuration of the RIS, the transmitter
sequentially adjust the phase shifts of the reflectarray element- and the receiver can estimate the channel between them
by-element without requiring any additional signaling. In [7], regularly as if the RIS does not exist. To elaborate, it is
RSSI has been also leveraged to guide the optimization of the straightforward to estimate the overall channel, Heff , cf. (1);
RIS; however, a simpler two-state phase shifter has been used however, estimating the channels between the RIS and the two
and a more robust optimization algorithm has been devised to link ends individually or isolating the effect of the RIS is much
avoid measuring the effect of a single element, which might more challenging [60].
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
8
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
9
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
10
VII. C ONCLUSION
Research in RIS-assisted wireless communications is still in
its infancy; many practical aspects are not well understood and
still need to be thoroughly investigated. However, the potential
for this new technology is immense. The idea of being able to
Fig. 5. Comparison of the error performance of zero-forcing decoder. The
change the propagation environment is not only conceptually
RIS are used to orthogonalize MIMO channels and achieve a unity condition interesting but also highly beneficial in a variety of scenarios.
number. In this article, we have discussed two candidate implementa-
tions for the RIS based on reflectarrays or metasurfaces. We
elaborated on channel modeling and how it can be affected
VI. F UTURE R ESEARCH D IRECTIONS
by the RIS implementation in terms of channel distribution
In this section, we briefly introduce some potential ap- and large-scale path loss. Furthermore, we presented some of
plications of RIS in wireless networks and discuss potential the challenges that need to be addressed in the optimization
research directions. of RIS-assisted networks. Optimization of the RIS will have
Centralized Beamforming for IoT Devices: The Internet- to be performed given limited information about the channel
of-Things (IoT) is an important component in future wireless and many optimization objectives can be pursued making
networks. However, some IoT devices are constrained in the problem very complex. In the end, practical optimization
both size and energy consumption. Future 5G and beyond techniques are essential to reap the system performance gains
cellular networks will operate in the millimeter-wave channels resulting from the ability to shape the wireless propagation.
where highly directional antenna gains are essential to achieve
reliable high-rate communications. Some IoT devices will be R EFERENCES
too small in size to support the antenna arrays required to [1] M. D. Renzo, M. Debbah, D.-T. Phan-Huy, A. Zappone, M.-S.
achieve enough beamforming gain to establish a link with Alouini, C. Yuen, V. Sciancalepore, G. C. Alexandropoulos, J. Hoydis,
H. Gacanin, J. d. Rosny, A. Bounceur, G. Lerosey, and M. Fink, “Smart
a distant base-station. RISs can be used to supply these radio environments empowered by reconfigurable AI meta-surfaces: an
devices with large beamforming gain, much larger than can idea whose time has come,” EURASIP J. Wireless Commun. and Netw.,
be afforded by them given their limited size. Note that both vol. 2019, no. 1, May 2019.
[2] Y.-C. Liang, R. Long, Q. Zhang, J. Chen, H. V. Cheng, and
the RIS and the base station will be fixed in place and H. Guo, “Large intelligent surface/antennas (LISA): Making reflective
probably with limited surrounding scatterers, which simplifies radios smart,” arXiv:1906.06578 [cs, math]. [Online]. Available:
the beamforming optimization between the RIS and the base- http://arxiv.org/abs/1906.06578
[3] M. R. Akdeniz, Y. Liu, M. K. Samimi, S. Sun, S. Rangan, T. S.
station. Rappaport, and E. Erkip, “Millimeter wave channel modeling and
Experimental Validation of Path Loss Scaling: Path cellular capacity evaluation,” IEEE J. Sel. Areas Commun., vol. 32, no. 6,
loss through the RIS will be an important factor in deter- pp. 1164–1179, Jun. 2014.
[4] O. E. Ayach, S. Rajagopal, S. Abu-Surra, Z. Pi, and R. W. Heath,
mining the practicality of RIS-assisted communications [64]. “Spatially sparse precoding in millimeter wave MIMO systems,” IEEE
This is especially important since the RIS is passive and Trans. Wireless Commun., vol. 13, no. 3, pp. 1499–1513, Mar. 2014.
only co-phasing gains are possible. Transparent relaying and [5] S. V. Hum and J. Perruisseau-Carrier, “Reconfigurable reflectarrays and
array lenses for dynamic antenna beam control: A review,” IEEE Trans.
backscatter channels are known to suffer from aggravated Antennas Propag., vol. 62, no. 1, pp. 183–198, Jan. 2014.
path loss due to additional spreading of the signal at the [6] Q. He, S. Sun, and L. Zhou, “Tunable/Reconfigurable metasurfaces:
relay node or backscatter tag. Traditionally, smooth surfaces Physics and applications,” Research, vol. 2019, pp. 1–16, Jul. 2019.
[7] V. Arun and H. Balakrishnan, “RFocus: Practical beamforming for
much larger than the wavelength are modeled as reflectors small devices,” arXiv:1905.05130 [cs], May 2019, arXiv: 1905.05130.
[21], [23], which gives metasurface-based RIS a huge advan- [Online]. Available: http://arxiv.org/abs/1905.05130
tage over reflectarray-based RIS. However, this still needs to [8] H. Chen, A. J. Taylor, and N. Yu, “A review of metasurfaces: physics
and applications,” Reports on Progress in Physics, vol. 79, no. 7, Jun.
be validated experimentally. Measurements for 0.35𝜆-spaced 2016.
reflectarray-based RIS were reported in [27]; however, it is [9] C. Huang, S. Hu, G. C. Alexandropoulos, A. Zappone, C. Yuen,
still not clear whether the tighter packing of scatterers in R. Zhang, M. Di Renzo, and M. Debbah, “Holographic MIMO
surfaces for 6G wireless networks: Opportunities, challenges,
a metasurface-based RIS would improve on these results, and trends,” arXiv:1911.12296 [cs, math]. [Online]. Available:
especially for the same overall physical size. http://arxiv.org/abs/1911.12296
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
11
[10] X. Tan, Z. Sun, J. M. Jornet, and D. Pados, “Increasing indoor spectrum [32] Q. Wu and R. Zhang, “Intelligent reflecting surface enhanced wireless
sharing capacity using smart reflect-array,” in IEEE Int. Conf. on network: Joint active and passive beamforming design,” in IEEE Global
Commun. (ICC), Kuala Lumpur, Malaysia, May 2016. Commun. Conf. (GLOBECOM), Abu Dhabi, UAE, Dec. 2018.
[11] A. Welkie, L. Shangguan, J. Gummeson, W. Hu, and K. Jamieson, [33] P. Wang, J. Fang, X. Yuan, Z. Chen, H. Duan, and H. Li, “Intelligent
“Programmable radio environments for smart spaces,” in ACM Workshop reflecting surface-assisted millimeter wave communications: Joint
on Hot Topics in Netw., Palo Alto, CA, 2017, pp. 36–42. active and passive precoding design,” arXiv:1908.10734 [eess].
[12] P. d. Hougne, M. Fink, and G. Lerosey, “Optimally diverse commu- [Online]. Available: http://arxiv.org/abs/1908.10734
nication channels in disordered environments with tuned randomness,” [34] E. Basar, “Transmission through large intelligent surfaces: A new
Nature Electronics, vol. 2, no. 1, pp. 36–41, Jan. 2019. frontier in wireless communications,” in European Conf. on Netw. and
[13] Z. Li, Y. Xie, L. Shangguan, R. I. Zelaya, J. Gummeson, W. Hu, and Commun. (EuCNC), Valencia, Spain, Jun. 2019, pp. 112–117.
K. Jamieson, “Towards programming the radio environment with large [35] C. Huang, G. C. Alexandropoulos, C. Yuen, and M. Debbah, “Indoor
arrays of inexpensive antennas,” in USENIX Symp. on Netw. Syst. Des. signal focusing with deep learning designed reconfigurable intelligent
and Implementation (NSDI), Boston, MA, Feb. 2019, pp. 285–300. surfaces,” in IEEE Int. Workshop on Signal Process. Adv. in Wireless
[14] R. W. Heath and A. Lozano, Foundations of MIMO Communication, Commun. (SPAWC), Cannes, France, Jul. 2019. [Online]. Available:
1st ed. Cambridge University Press, Dec. 2018. http://arxiv.org/abs/1905.07726
[15] J. D. Griffin and G. D. Durgin, “Gains for RF tags using multiple [36] A. Taha, M. Alrabeiah, and A. Alkhateeb, “Deep learning for large
antennas,” IEEE Trans. Antennas Propag., vol. 56, no. 2, pp. 563–570, intelligent surfaces in millimeter wave and massive MIMO systems,” in
Feb. 2008. IEEE Global Commun. Conf. (GLOBECOM), Waikoloa, HI, Dec. 2019.
[16] M. ElMossallamy, Z. Han, M. Pan, R. Jantti, K. Seddik, and G. Y. [37] Q. Wu and R. Zhang, “Beamforming optimization for intelligent reflect-
Li, “Noncoherent frequency shift keying for ambient backscatter over ing surface with discrete phase shifts,” in IEEE Int. Conf. on Acoust.,
OFDM signals,” in IEEE Int. Conf. on Commun. (ICC), Shanghai, China, Speech and Signal Process. (ICASSP), Brighton, United Kingdom, May
May 2019. 2019, pp. 7830–7833.
[17] H. Zhang, B. Di, L. Song, and Z. Han, “Reconfigurable intelligent [38] Y. Yang, S. Zhang, and R. Zhang, “IRS-enhanced OFDM: Power
surfaces assisted communications with limited phase shifts: How many allocation and passive array optimization,” Dec. 2019.
phase shifts are enough?” IEEE Trans. Veh. Technol., to be published. [39] Y. Yang, B. Zheng, S. Zhang, and R. Zhang, “Intelligent
[18] R. W. Heath, N. González-Prelcic, S. Rangan, W. Roh, and A. M. reflecting surface meets OFDM: Protocol design and rate
Sayeed, “An overview of signal processing techniques for millimeter maximization,” arXiv:1906.09956 [cs, math]. [Online]. Available:
wave mimo systems,” IEEE J. Sel. Topics Signal Process., vol. 10, no. 3, http://arxiv.org/abs/1906.09956
pp. 436–453, Apr. 2016. [40] C. You, B. Zheng, and R. Zhang, “Intelligent reflecting surface with
[19] P. Almers, E. Bonek, A. Burr, N. Czink, M. Debbah, V. Degli-Esposti, discrete phase shifts: Channel estimation and passive beamforming,”
H. Hofstetter, P. Kyösti, D. Laurenson, G. Matz, A. F. Molisch, C. Oest- arXiv:1911.03916 [cs, math], Nov. 2019. [Online]. Available:
ges, and H. Özcelik, “Survey of channel and radio propagation models http://arxiv.org/abs/1911.03916
for wireless MIMO systems,” EURASIP J. on Wireless Commun. and [41] B. Zheng and R. Zhang, “Intelligent reflecting surface-enhanced ofdm:
Netw., vol. 2007, no. 1, Dec. 2007. Channel estimation and reflection optimization,” IEEE Wireless Com-
[20] E. Björnson, L. Sanguinetti, H. Wymeersch, J. Hoydis, and T. L. mun. Lett., vol. 9, no. 4, Apr. 2020.
Marzetta, “Massive MIMO is a reality–What is next?: Five promising [42] S. Abeywickrama, R. Zhang, and C. Yuen, “Intelligent
research directions for antenna arrays,” Digital Signal Process., vol. 94, reflecting surface: Practical phase shift model and beamforming
pp. 3–20, Nov. 2019. optimization,” arXiv:1907.06002 [cs, eess, math]. [Online]. Available:
[21] T. Rappaport, Wireless Communications: Principles and Practice, http://arxiv.org/abs/1907.06002
2nd ed. Upper Saddle River, NJ: Prentice Hall PTR, 2001. [43] C. Huang, A. Zappone, M. Debbah, and C. Yuen, “Achievable rate
[22] T. Rappaport and S. Sandhu, “Radio-wave propagation for emerging maximization by passive intelligent mirrors,” in IEEE Int. Conf. on
wireless personal-communication systems,” IEEE Antennas Propag. Acoust., Speech and Signal Process. (ICASSP), Apr. 2018, pp. 3714–
Mag., vol. 36, no. 5, pp. 14–24, Oct. 1994. 3718.
[23] K. R. Schaubach, N. J. Davis, and T. S. Rappaport, “A ray tracing [44] C. Huang, G. C. Alexandropoulos, A. Zappone, M. Debbah, and
method for predicting path loss and delay spread in microcellular C. Yuen, “Energy efficient multi-user MISO communication using low
environments,” in Veh. Technol. Soc. Conf. (VTC), vol. 2, Denver, CO, resolution large intelligent surfaces,” in IEEE Globecom Workshops (GC
May 1992, pp. 932–935. Wkshps), Abu Dhabi, UAE, Dec. 2018.
[24] E. Ben-Dor, T. S. Rappaport, Y. Qiao, and S. J. Lauffenburger, [45] H. Guo, Y.-C. Liang, J. Chen, and E. G. Larsson, “Weighted sum-
“Millimeter-wave 60 GHz outdoor and vehicle AOA propagation mea- rate maximization for intelligent reflecting surface enhanced wireless
surements using a broadband channel sounder,” in IEEE Global Com- networks,” in IEEE Global Commun. Conf. (GLOBECOM), Waikoloa,
mun. Conf. (GLOBECOM), Houston, TX, Dec. 2011. HI, Dec. 2019.
[25] A. Díaz-Rubio, V. S. Asadchy, A. Elsakka, and S. A. Tretyakov, “From [46] C. Huang, A. Zappone, G. C. Alexandropoulos, M. Debbah, and
the generalized reflection law to the realization of perfect anomalous C. Yuen, “Reconfigurable intelligent surfaces for energy efficiency in
reflectors,” Science Advances, vol. 3, no. 8, Aug. 2017. wireless communication,” IEEE Trans. Wireless Commun., vol. 18, no. 8,
[26] O. Özdogan, E. Björnson, and E. G. Larsson, “Intelligent pp. 4157–4170, Aug. 2019.
reflecting surfaces: Physics, propagation, and pathloss modeling,” [47] Q.-U.-A. Nadeem, A. Kammoun, A. Chaaban, M. Debbah, and
IEEE Wireless Commun. Lett., to be published. [Online]. Available: M.-S. Alouini, “Asymptotic max-min SINR analysis of reconfigurable
http://arxiv.org/abs/1911.03359 intelligent surface assisted MISO systems,” arXiv:1903.08127 [cs,
[27] W. Tang, M. Z. Chen, X. Chen, J. Y. Dai, Y. Han, M. Di Renzo, math]. [Online]. Available: http://arxiv.org/abs/1903.08127
Y. Zeng, S. Jin, Q. Cheng, and T. J. Cui, “Wireless communications with [48] ——, “Intelligent reflecting surface assisted wireless communication:
reconfigurable intelligent surface: Path loss modeling and experimental Modeling and channel estimation,” arXiv:1906.02360 [cs, eess, math].
measurement,” arXiv:1911.05326 [cs, eess, math], Nov. 2019. [Online]. [Online]. Available: http://arxiv.org/abs/1906.02360
Available: http://arxiv.org/abs/1911.05326 [49] Q. Wu and R. Zhang, “Beamforming optimization for wireless network
[28] E. Basar, M. Di Renzo, J. de Rosny, M. Debbah, M.-S. Alouini, aided by intelligent reflecting surface with discrete phase shifts,” IEEE
and R. Zhang, “Wireless Communications Through Reconfigurable Trans. on Commun., to be published.
Intelligent Surfaces,” arXiv:1906.09490 [cs, eess, math], Jun. 2019, [50] ——, “Intelligent reflecting surface enhanced wireless network via
arXiv: 1906.09490. [Online]. Available: http://arxiv.org/abs/1906.09490 joint active and passive beamforming,” IEEE Trans. Wireless Commun.,
[29] S. W. Ellingson, “Path loss in reconfigurable intelligent surface-enabled vol. 18, no. 11, pp. 5394–5409, Nov. 2019.
channels,” arXiv:1912.06759 [cs, eess, math]. [Online]. Available: [51] C. Huang, R. Mo, and C. Yuen, “Reconfigurable intelligent surface
http://arxiv.org/abs/1912.06759 assisted multiuser MISO systems exploiting deep reinforcement
[30] S. Hu, F. Rusek, and O. Edfors, “Beyond massive MIMO: The potential learning,” IEEE J. Sel. Areas Commun., to be published. [Online].
of data transmission with large intelligent surfaces,” IEEE Trans. Signal Available: http://arxiv.org/abs/2002.10072
Process., vol. 66, no. 10, pp. 2746–2758, May 2018. [52] B. Di, H. Zhang, L. Song, Y. Li, Z. Han, and H. V. Poor,
[31] X. Tan, Z. Sun, D. Koutsonikolas, and J. M. Jornet, “Enabling indoor “Hybrid Beamforming for Reconfigurable Intelligent Surface based
mobile millimeter-wave networks based on smart reflect-arrays,” in IEEE Multi-user Communications: Achievable Rates with Limited Discrete
Conf. on Comput. Commun. (INFOCOM), Honolulu, HI, Apr. 2018, pp. Phase Shifts,” arXiv:1910.14328 [cs, eess, math], Oct. 2019, arXiv:
270–278. 1910.14328. [Online]. Available: http://arxiv.org/abs/1910.14328
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
12
[53] Y. Cao and T. Lv, “Delay-constrained joint power control, user Hongliang Zhang (S’15-M’19) received the B.S.
detection and passive beamforming in intelligent reflecting surface and Ph.D. degrees at the School of Electrical Engi-
assisted uplink mmwave system,” arXiv:1912.10030 [eess]. [Online]. neering and Computer Science at Peking University,
Available: http://arxiv.org/abs/1912.10030 in 2014 and 2019, respectively. Currently, he is a
[54] B. Zheng, C. You, and R. Zhang, “Intelligent reflecting Postdoctoral Fellow in the Electrical and Computer
surface assisted multi-user OFDMA: Channel estimation and Engineering Department at the University of Hous-
training design,” arXiv:2003.00648 [cs, math]. [Online]. Available: ton, Texas. His current research interest includes
http://arxiv.org/abs/2003.00648 cooperative communications, Internet-of-Things net-
[55] S. Hua, Y. Zhou, K. Yang, and Y. Shi, “Reconfigurable intelligent works, hypergraph theory, and optimization theory.
surface for green edge inference,” arXiv:1912.00820 [cs, eess, math]. He has served as a TPC Member for many IEEE
[Online]. Available: http://arxiv.org/abs/1912.00820 conferences, such as Globecom, ICC, and WCNC.
[56] T. Jiang and Y. Shi, “Over-the-air computation via intelligent reflecting He is currently an Editor for IET Communications. He also serves as a
surfaces,” arXiv:1904.12475 [cs, eess, math]. [Online]. Available: Guest Editor for IEEE IoT-J special issue on Internet of UAVs over Cellular
http://arxiv.org/abs/1904.12475 Networks.
[57] A. Taha, M. Alrabeiah, and A. Alkhateeb, “Enabling large intelligent
surfaces with compressive sensing and deep learning,” arXiv:1904.10136
[cs, eess, math], Apr. 2019, arXiv: 1904.10136. [Online]. Available:
http://arxiv.org/abs/1904.10136
[58] J. Wang, Z. Lan, C. Pyo, T. Baykas, C. Sum, M. Rahman, J. Gao,
R. Funada, F. Kojima, H. Harada, and S. Kato, “Beam codebook based
beamforming protocol for multi-Gbps millimeter-wave WPAN systems,”
IEEE J. Sel. Areas Commun., vol. 27, no. 8, pp. 1390–1399, Oct. 2009.
[59] L. Dai, B. Wang, M. Wang, X. Yang, J. Tan, S. Bi, S. Xu, F. Yang,
Z. Chen, M. Di Renzo, and L. Hanzo, “Reconfigurable intelligent
surface-based wireless communication: Antenna design, prototyping
and experimental results,” arXiv:1912.03620 [cs, math]. [Online].
Available: http://arxiv.org/abs/1912.03620
Lingyang Song (S’03-M’06-SM’12-F’19) received
[60] Z.-Q. He and X. Yuan, “Cascaded channel estimation for large intelligent
his Ph.D. from the University of York, United King-
metasurface assisted massive mimo,” IEEE Wireless Commun. Lett., to
dom, in 2007, where he received the K. M. Stott
be published.
Prize for excellent research. He worked as a research
[61] E. Basar and I. F. Akyildiz, “Reconfigurable intelligent surfaces for
fellow at the University of Oslo, Norway, until
Doppler effect and multipath fading mitigation,” arXiv:1912.04080 [cs,
rejoining Philips Research UK in March 2008. In
eess, math]. [Online]. Available: http://arxiv.org/abs/1912.04080
May 2009, he joined the Department of Electronics,
[62] W. Yang, J. Gibson, and T. He, “Coefficient rate and lossy source
School of Electronics Engineering and Computer
coding,” IEEE Trans. Inf. Theory, vol. 51, no. 1, pp. 381–386, Jan.
Science, Peking University, and is now a Boya
2005.
Distinguished Professor. His main research interests
[63] O. Roy and M. Vetterli, “The effective rank: A measure of effective
include wireless communication and networks, sig-
dimensionality,” in European Signal Process. Conf., Poznan, Poland,
nal processing, and machine learning. He was the recipient of the IEEE
Sep. 2007, pp. 606–610.
Leonard G. Abraham Prize in 2016 and the IEEE Asia Pacific (AP) Young
[64] E. Björnson, O. Özdogan, and E. G. Larsson, “Intelligent reflecting
Researcher Award in 2012. He has been an IEEE Distinguished Lecturer since
surface vs. decode-and-forward: How large surfaces are needed to beat
2015.
relaying?” arXiv:1906.03949 [cs, math], Jun. 2019. [Online]. Available:
http://arxiv.org/abs/1906.03949
[65] J. Hu, H. Zhang, B. Di, L. Li, L. Song, Y. Li, Z. Han, and H. V.
Poor, “Reconfigurable intelligent surfaces based RF sensing: Design,
optimization, and implementation,” arXiv:1912.09198 [cs, eess], Dec.
2019. [Online]. Available: http://arxiv.org/abs/1912.09198
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCCN.2020.2992604, IEEE
Transactions on Cognitive Communications and Networking
13
2332-7731 (c) 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Auckland University of Technology. Downloaded on May 23,2020 at 01:44:48 UTC from IEEE Xplore. Restrictions apply.