ACCEPTED FROM OPEN CALL

Reconfigurable Intelligent Surfaces:

Potentials, Applications, and Challenges for
6G Wireless Networks
Sarah Basharat, Syed Ali Hassan, Haris Pervaiz, Aamir Mahmood, Zhiguo Ding, and Mikael Gidlund

Abstract trum in high frequency bands, thus resolving spec-

trum scarcity issues at microwave frequencies.
Reconfigurable intelligent surfaces (RISs), with Although mMIMO and mmWaves significantly
the potential to realize smart radio environments, improve spectral efficiency (SE), high hardware
have emerged as an energy-efficient and a cost-ef- cost and complexity are major hurdles in their
fective technology to support the services and practical implementation. Moreover, mmWaves
demands foreseen for coming decades. By lever- are highly vulnerable to signal blockage and atten-
aging a large number of low-cost passive reflect- uation. Therefore, reliable and efficient communi-
ing elements, RISs introduce a phase-shift in the cation is still not guaranteed.
impinging signal to create a favorable propagation While 5G is yet to be realized fully, research-
channel between the transmitter and the receiver. ers have already started looking for energy and
In this article, we provide a tutorial overview of spectral-efficient solutions for sixth-generation (6G)
RISs for sixth-generation (6G) wireless networks. systems. In addition to energy and spectral efficien-
Specifically, we present a comprehensive discus- cy, the new paradigm is smart and reconfigurable
sion on performance gains that can be achieved wireless environments [2]. Recently, a cost-effective
by integrating RISs with emerging communication and energy-efficient technology, reconfigurable
technologies. We address the practical implemen- intelligent surface (RIS) [3–5], also called intelligent
tation of RIS-assisted networks and expose the cru- reflecting surface (IRS) [2, 6–14] or passive holo-
cial challenges, including the RIS reconfiguration, graphic MIMO surface (HMIMOS) [15], has been
deployment and size optimization, and channel proposed in this regard. RIS empowers smart wire-
estimation. Furthermore, we explore the integra- less environments by overcoming the stochastic
tion of RIS and non-orthogonal multiple access nature of the propagation channel, thereby improv-
(NOMA) under imperfect channel state infor- ing quality-of-service (QoS) and connectivity. The
mation (CSI). Our numerical results illustrate the wireless environment that used to be a dynamic
importance of better channel estimation in RIS-as- uncontrollable factor is now considered to be a
sisted networks and indicate the various factors part of the network design parameter.
that impact the size of RIS. Finally, we present Specifically, an RIS is a software-controlled
promising future research directions for realizing planer surface consisting of a large number of low-
RIS-assisted networks in 6G communication. cost passive reflecting elements. Each element, of
size smaller than the wavelength, has the capabil-
Introduction ity to alter the phase of the impinging signal, cre-
The global mobile traffic volume is anticipated to ating a favorable wireless environment between
reach 5016 exabytes per month (Eb/mo) in 2030, the transmitter and the receiver. The RIS reflec-
compared with 7.462 EB/mo in 2010 [1]. This tion adaption is programmed and controlled via
clearly depicts the importance of the evolution a smart controller, such as a field-programmable
and advancement of mobile communication tech- gate array (FPGA), which acts as a gateway to
nologies. To date, fifth-generation (5G) communi- communicate and coordinate with the BS through
cation, which is expected to realize the targeted a separate wireless or wired link. In particular, RIS
1000x increase in network capacity with new and receives a signal from the base station (BS), and
advanced services, is being rolled out in the world. then reflects the incident signal by inducing phase
However, 5G systems will not be able to fully sup- changes, adjusted by the controller. Consequent-
port the growing demand for wireless communi- ly, the reflected signal can be added coherently
cation in 2030. The core 5G technologies include with the direct signal from the BS to either boost or
massive multiple-input multiple-output (mMIMO) attenuate the overall signal strength at the receiv-
and millimeter-wave (mmWave) communications. er. Although RIS is theoretically passive, since it
mMIMO technology exploits the spatial domain reflects the signal without power amplification,
by deploying numerous antennas to enable par- however, in practice, it has minimal power require-
allel transmission to multiple users using the same ment for the operation of a smart controller and
frequency-time resource block. The mmWaves, reconfiguration of the elements for controllable
on the other hand, offer plenteous spare spec- reflections.

Digital Object Identifier: Sarah Basharat and Syed Ali Hassan are with the National University of Sciences and Technology (NUST); Haris Pervaiz is with Lancaster
10.1109/MWC.011.2100016 University; Aamir Mahmood and Mikael Gidlund are with Mid Sweden University; Zhiguo Ding is with The University of Manchester.

184 1536-1284/21/$25.00 © 2021 IEEE IEEE Wireless Communications • December 2021

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 184 1/11/22 11:42 AM
 Specifically, an RIS is
a software-controlled
RIS self-interference
(FD mode)
Relay planer surface consist-
Passive Active
Energy Efficient Low energy ing of a large number
Cost Effective consumption
Free from self- Low hardware cost Source%3Cm of low-cost passive
Reflects signal interference
`
Suffers from noise
amplification &
reflecting elements.
Receives/transmits signal
self-interference (FD
mode) Receiver Each element, of size
smaller than the wave-
RIS reflected link
Source-Relay-Receiver link
length, has the capabil-
Direct link
RIS
RIS control link Direct link ity to alter the phase
RIS aided Communication Conventional Relaying of the impinging signal,
RIS controller
creating a favorable
MIMO Relay BackCom wireless environment
self-interference Relay
(FD mode) Active Passive between the transmitter
High energy Energy efficient
consumption Cost effective and the receiver.
High hardware cost Short operating
MIMO relay
Suffers from noise Reflects signal range
amplification & Suffers from self-
Receives/transmits signal self-interference (FD interference
mode)
Backscatter tag

Source-MIMO Relay-Receiver link Backscatter reflected link

Direct link Direct link

Massive MIMO Relaying Backscatter Communication

FIGURE 1. RIS comparison with existing related technologies.

As illustrated in Fig. 1, the RIS concept can be technologies for realizing 6G wireless net-
viewed to operate similarly as other related wireless works, namely, non-orthogonal multiple access
technologies such as conventional relaying, back- (NOMA), simultaneous wireless information
scatter communication (BackCom), and mMIMO and power transfer (SWIPT), unmanned aeri-
relaying. We now present the major differences al vehicles (UAVs), BackCom, mmWaves, and
and competitive strengths of RIS that make it stand multi-antenna systems.
out among these technologies. First, compared • For practical implementation of RIS-assisted
to conventional relaying that requires additional networks, we identify three crucial challenges,
power for signal transmission, amplification, and including RIS reconfiguration for controllable
regeneration, RIS passively reflects the impinging reflections, deployment and size optimization,
signal by inducing intelligent phase-shifts, without and channel estimation.
the need for an additional radio frequency (RF) • We present a novel case study for a RIS-as-
source. Moreover, RIS operates in full-duplex (FD) sisted NOMA network with imperfect chan-
mode, free from noise amplification and self-inter- nel state information (CSI) to highlight the
ference. Second, compared to traditional BackCom impact of channel estimation errors on the
such as radio frequency identification (RFID) tag performance of RIS-assisted NOMA networks.
that loads its own information on the incident sig- We further determine the various factors that
nal and then backscatters the modulated signal to affect the size of RIS, that is, the number of
the receiver, RIS reflects the incident signal to assist RIS elements.
communication between the source and receiver • To provide effective guidance for future
without sending any information of its own. Back- research, we introduce five promising research
Com also requires sophisticated signal process- directions for realizing RIS-assisted networks.
ing for self-interference cancellation in order to
decode the tag message. Finally, unlike mMIMO
relaying, RIS can be implemented at a much low
IntegrAtIng rIs WIth
hardware cost and power consumption. Although emergIng communIcAtIon technologIes
signal-to-noise ratio (SNR) achieved through RIS is Current research contributions have revealed RIS
less than the equal-sized mMIMO counterpart [6], to be a cutting-edge technology, opening new
however, the SNR of an RIS-assisted system can promising research opportunities on the road
be improved by increasing the reflecting elements, toward 6G. In this section, we elaborate on the
since the cost per reflecting element of RIS is much performance gains that can be achieved by inte-
less than the cost per antenna element in mMIMO grating RIS with emerging communication tech-
relaying. Architecturally, RIS is lightweight with con- nologies.
formal geometry and can be easily mounted on
ceilings, walls, and building facades. rIs And nomA
Inspired by the RIS potential to realize smart NOMA has emerged as a promising technol-
wireless environments and its compatibility with ogy for future generation networks to support
other technologies, the main contributions of this massive connectivity. Power domain NOMA
study can be summarized as follows: (PD-NOMA) enables multiple users to share
• We provide a comprehensive discussion on the same resource (e.g., time, frequency, code)
integrating RIS with emerging communication block, hence improving both spectral and energy

IEEE Wireless Communications • December 2021 185

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 185 1/11/22 11:42 AM
 SWIPT is an effective efficiency. In a downlink NOMA system, super-
solution to power position coding (SC) is used at the BS to multi- RIS
massive devices in plex the data of the users with different channel
a wireless-powered gains, and successive interference cancellation
(SIC) is employed at the receiver to decode the
Internet-of-things (IoT) message signals. Although NOMA provides suf- Reflected IF
network. In practice, ficient performance gains over OMA, the strin- Direct IF
the significant power gent demands on data rate and connectivity for
loss over long dis- B5G/6G systems compel to shift to smart and BS IRs
Reflected EF

tances reduces the reconfigurable wireless networks; hence, the ERs

RIS-assisted NOMA system has been proposed EF (Energy flow)
energy harvested at by the research community. IF (Information flow Direct EF

the energy receiver, To fully exploit the gains of the RIS-assisted

which degrades the NOMA system, appropriate phase-shift matrix RIS-assisted SWIPT system
performance of SWIPT and beamforming vectors need to be designed.
systems. However, the The two types of phase-shifting designs, coher- Controllable UAV-RIS link RIS
ent phase-shifting and random phase-shifting,
limitations of practical achieve different trade-offs between performance UAV
SWIPT systems can be and complexity for an RIS-assisted NOMA sys- Blocked Direct link

compensated via RIS. tem [7]. For coherent phase-shifting, each RIS
element introduces a phase that matches the ` Controllable
phase of fading channels from RIS-to-BS and RIS-UE link

users. Despite the superior performance, it might

not be feasible to implement coherent shifting
design because of hardware limitations of phase
shifters and excess system overhead of acquiring
CSI. Although the random phase-shifting design Ground user
RIS-assisted UAV communication
reduces the system complexity and overhead
of acquiring CSI, it degrades the system perfor- FIGURE 2. RIS-assisted SWIPT and UAV commu-
mance. Furthermore, the RIS phase-shift matrix nication.
and beamforming vectors can be jointly opti-
mized to minimize the transmit power of the BS
using semi-definite relaxation (SDR). However, RIS. Such a performance is quite appealing for
for large-scale RIS-assisted networks, the com- practical applications.
putational complexity of the SDR technique is
extremely high and the probability of returning rIs And uAvs
a rank-one optimum solution is extremely small. RIS can even be applied to UAV-enabled commu-
Therefore, a difference-of-convex (DC) method nication systems to improve the propagation envi-
has been proposed that overcomes the limitation ronment and enhance communication quality, as
of the SDR method and outperforms in terms of shown in Fig. 2. In a dense urban environment, the
minimizing the BS transmit power [8]. line-of-sight (LoS) links between the UAV and the
ground users may be blocked, which deteriorates
rIs And sWIPt the channel gains. However, the RIS-assisted UAV
SWIPT is an effective solution to power massive system can enable virtual LoS paths by reflecting
devices in a wireless-powered Internet-of-things the signal received from UAV to the ground users.
(IoT) network. In practice, the significant power The received signal power at the ground user can
loss over long distances reduces the energy har- be significantly enhanced by the joint optimiza-
vested at the energy receiver, which degrades the tion of RIS beamforming and UAV trajectory. In
performance of SWIPT systems. However, the lim- this regard, Li et al. [3] proposed an iterative algo-
itations of practical SWIPT systems can be com- rithm for achievable rate maximization under UAV
pensated via RIS, as illustrated in Fig. 2. Through mobility and the RIS’s phase-shift constraints, with
intelligent signal reflections, RIS boosts the signal the results confirming significant improvement in
strength both at the information receiver (IR) and achievable rate with the aid of RIS.
the energy receiver (ER), thereby improving the RIS can also enhance the cellular communica-
energy efficiency of SWIPT system. tion of UAVs [10], which suffers from down-tilted
Recently, the technology in [9] unleashed the BS antennas, that is, the main lobes of antennas
benefits of a RIS-assisted SWIPT network, where a are optimized to serve the ground users, while
multi-antenna BS serves several multi-antenna infor- UAV communication is supported by side lobes
mation users, while satisfying the energy require- only. Through intelligent and optimized signal
ments of the energy users. The authors proposed reflections, controlled via cellular BS, RIS can
a weighted sum-rate maximization problem for direct the impinging BS signal toward a specific
the joint optimization of a BS’s transmit precoding UAV. The RIS reflected signal combines coherent-
matrices and RIS phase-shifts. The proposed prob- ly with the direct BS-UAV signal, thus improving
lem is challenging to solve for an optimal solution, the received signal strength at the UAV. Even a
owing to highly coupled optimization parameters. small-sized RIS, deployed on a building facade,
Therefore, the authors developed an iterative solu- can improve the cellular communication of the
tion using a block coordinate descent (BCD) algo- UAV flying substantially above the BS. Moreover,
rithm that converges rapidly and outperforms the RIS location, that is, RIS distance from BS distance
benchmark approaches, that is, fixed RIS phase- and RIS deployment height, is a critical factor in
shifts and conventional SWIPT networks without such applications as performance gain achieved

186 IEEE Wireless Communications • December 2021

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 186 1/11/22 11:42 AM
 through RIS is maximized when RIS location is design. However, the efficacy of the proposed For an energy-efficient
selected optimally. design under imperfect CSI needs further inves- solution, RIS can be
tigation. Moreover, the proposed design for the applied to MISO sys-
RIS and BackCom hybrid precoding and phase-shifts can be extend- tems to improve the
BackCom is a promising solution toward an ener- ed to a system with the direct links between the
gy-efficient and sustainable IoT network. Despite BS and the users. network performance
the extensive research on the improvement of at significantly low
reliability and throughput of a BackCom system, RIS and Multi-Antenna Systems hardware cost and
its short operation range remains a key barrier Multi-antenna systems aim at actively improving energy consumption.
toward the large-scale deployment that needs to the signal quality by employing a large number Different from the
be addressed. Recently, in [11], the authors elab- of antennas and exploiting the spatial domain for
orated on the potential of RIS-assisted monostatic transmit beamforming. However, the convention- conventional systems,
and bistatic BackCom systems, where the RIS is al multiple-input single-output (MISO) systems an RIS-aided MISO sys-
employed to assist the communication between suffer from wireless channel randomness, limiting tem can guarantee the
the tag and the reader. The authors proposed their performance. Therefore, for an energy-effi- users’ QoS, with fewer
a joint optimization framework, under transmit cient solution, RIS can be applied to MISO sys- BS antennas, by utiliz-
power minimization, to optimize RIS phase-shifts tems to improve the network performance at
and the source transmit beamforming. The appli- significantly low hardware cost and energy con- ing the smart passive
cation of RIS to BackCom can significantly reduce sumption. Different from the conventional sys- reflections of RIS.
the transmit power, which can be mapped to tems, an RIS-aided MISO system can guarantee
improve the operational range. the users’ QoS, with fewer BS antennas, by utiliz-
Furthermore, the ability of RIS to steer the sig- ing the smart passive reflections of RIS.
nals in different directions to reduce the inter-user Recently, the authors in [5] presented a hybrid
interference can be utilized to improve the detec- beamforming design for a multi-user RIS-assisted
tion performance of the ambient BackCom sys- MISO system, where the communication is estab-
tems. In this regard, Jia et al. [12] proposed a deep lished via RIS only due to the presence of obstacles
reinforcement learning (DRL) based approach, between the BS and the various users. The authors
namely, the deep deterministic policy gradient proposed a two-step sum-rate maximization algo-
(DDPG) algorithm, to jointly optimize the RIS and rithm to design continuous digital beamforming
reader beamforming for a RIS-assisted ambient at the BS and discrete analog beamforming at the
BackCom system with no knowledge of channels RIS. The observations indicated in [5] provide use-
and ambient signals. The results in [12] demon- ful insights into the design of RIS-based systems.
strate the significant improvement in detection per- First, the system performance is influenced by the
formance of ambient BackCom with the aid of RIS. size of RIS and the number of quantization bits
for discrete phase-shifts. Second, the RIS-based
RIS and mmWaves hybrid beamforming design can greatly reduce the
mmWave communication, with the capabili- requirement of dedicated hardware while provid-
ty to support multi-gigabits of data rate, is per- ing the satisfactory sum-rate.
ceived as a potential solution for the looming
capacity crunch. However, the high directivity
of mmWaves makes it vulnerable to blockage
Practical Implementation of
instants, especially in indoor and dense urban RIS-Assisted Networks
environments. As RIS has the capability to intro- In this section, we identify and discuss the crucial
duce effective additional paths, an RIS-enhanced challenges for the practical implementation of
mmWave system can overcome the limitations RIS-assisted networks.
of a conventional mmWave system. When the
direct links from the BS to users are severely RIS Reconfiguration for Controllable Reflections
blocked, optimizing the system parameters can The RIS phase-shift per element can be tuned
provide satisfactory performance gains. Recently, for controllable reflections through three main
in [13], the authors utilized the alternative opti- approaches, namely, mechanical actuation, func-
mization and successive convex approximation tional materials, and electronic devices. Besides
(SCA) to jointly optimize the beamforming vec- the phase control, the reflection amplitude can
tors and power allocation for the RIS-assisted be adjusted by varying the load impedance in
mmWave-NOMA system. The results confirm the each element. Thus, the reflection amplitude and
RIS’s ability to enhance the coverage range of the phase-shift can be realized in the range of [0, 1]
mmWave-NOMA system, especially when the and [0, 2p), respectively. The continuous varia-
direct BS to users’ links are blocked. tion of reflection coefficients is usually beneficial
Furthermore, the hybrid precoding design for from the communication performance perspec-
a multi-user RIS-assisted mmWave communica- tive. However, for practical RISs with a massive
tion system is presented in [4], where the direct number of reflecting elements, it is desirable to
links from the BS and the users are assumed to be implement only a finite number of discrete phase-
blocked. The authors jointly optimized the hybrid shift and amplitude levels, since high-resolution
precoding at the BS and phase-shifts at the RIS to elements increase the hardware cost, design
minimize the mean square error (MSE) between complexity, and control overhead. In addition,
the transmitted and the received symbols. The the optimization of reflection amplitudes and
gradient-projection (GP) method, based on alter- phase-shifts becomes more challenging with dis-
nating optimization (AO), is adopted to address crete variables. Furthermore, most of the works
the non-convex constraint for the analog precod- on RIS assume an ideal phase-shift model with
ing and the phase-shifts. The results illustrate the full signal reflection at each element regardless of
significant performance gains of the proposed the phase-shift at each element, which is difficult

IEEE Wireless Communications • December 2021 187

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 187 1/11/22 11:42 AM
 In general, the distribut-
ed RISs have a greater
probability of establish- Power
RIS
ing LoS links with the RIS Controller

cluster 1

cluster M
BS and the users than
the centralized RIS.
However, in distributed Frequency

e
m
deployment, the com-

Ti
Power U1 UK
munication between
UK Cluster M
the RISs and the BS, BS U1 UK
and the coordination U1 U1 UK
Cluster 1
among the multiple e Frequency
m
Ti
RISs, greatly increase
the signalling overhead.
Moreover, how to RIS control link BS-UE link BS-RIS link RIS-UE link Unblocked Cluster Blocked Cluster

optimally select the FIGURE 3. An illustration of RIS-assisted downlink NOMA network.

active number of RIS
elements in both cen-
tralized and distributed to realize in practice because of the strong cou- ly, the direct channels from the BS to users and
pling among the reflection amplitude and phase- the RIS reflected channels are estimated sepa-
placement is another shift. Therefore, realistic phase-shift models, with rately. However, the ON/OFF-based scheme is
design challenge. phase-dependent amplitude response, have to be not cost-effective for large-scale RISs because of
conceived for accurate performance analysis. training overhead. Furthermore, the RIS reflected
signal suffers from substantial power loss as each
rIs dePloyment And sIze oPtImIzAtIon time only one element is turned on. This weak-
The RIS deployment in a hybrid network, with ens the received signal strength, which degrades
both active BSs and passive RISs, is another cru- the channel estimation accuracy. To reduce the
cial problem. The deployment strategy significant- training overhead for practical scale RISs, the RIS
ly influences the distribution of the RIS associated elements are grouped into sub-surfaces [14]; as a
propagation channels, that is, BS-RIS and RIS-us- result, only the RIS reflected channel associated
ers’ channels. Therefore, an effective RIS deploy- with the sub-surface needs to be estimated. How-
ment strategy needs to be adopted to guarantee ever, the design of low-overhead channel estima-
the performance enhancements promised by the tion protocols for future networks with a large
RIS technology. Moreover, the RIS deployment number of users remains an open problem. More-
must also consider the practical factors, name- over, most of the existing works on RIS assume
ly, deployment cost, user distribution, and avail- the availability of perfect CSI at the BS, which is
able space, along with propagation conditions. practically not feasible. Therefore, the channel
For the single-user design, the given number of estimation errors should be taken into account for
RIS elements can be either grouped as a single accurate performance analysis.
RIS or partitioned into multiple cooperative RISs
for reaping the cooperative passive beamform-
ing gain. Furthermore, for the multi-user design,
rIs-AssIsted nomA netWork
the given number of RIS elements can be either under ImPerfect csI: A cAse study
grouped as a single RIS placed in the vicinity of The integration of RIS with NOMA can provide a
the BS (centralized deployment) or partitioned potential multiple access solution for future net-
into multiple RISs placed close to users’ hot spots works. RIS-assisted NOMA networks can play a
(distributed deployment). In general, the distribut- significant role in improving network coverage
ed RISs have a greater probability of establishing and capacity in urban areas where high-rise build-
LoS links with the BS and the users than the cen- ings and structures disrupt wireless services. As
tralized RIS. However, in distributed deployment, illustrated in Fig. 3, we consider an RIS-assisted
the communication between the RISs and the BS, downlink PD-NOMA network, where a macro BS
and the coordination among the multiple RISs, is equipped with a single transmit antenna and
greatly increase the signalling overhead. More- an RIS equipped with N reconfigurable passive
over, how to optimally select the active number reflecting elements serve uniformly distributed sin-
of RIS elements in both centralized and distribut- gle-antenna users, grouped into clusters. The RIS
ed placement is another design challenge. is connected to a smart controller, which commu-
nicates with the BS and changes the phase of the
chAnnel estImAtIon In rIs-AssIsted netWorks incident signal based on coherent phase-shifting
The channel estimation for RIS-assisted networks design [7]. The total system bandwidth is equally
is performed at the BS station, and the acquired divided into orthogonal frequency resource blocks,
CSI is communicated to the RIS controller via such that each NOMA cluster is allocated a single
the control link, which adjusts the phase-shifts. frequency resource block. The available BS power
One practical approach for RIS’s channel esti- is equally divided among the resource blocks, such
mation is by employing an element-by-element that the power allocated to a single resource block
ON/OFF-based channel estimation scheme, that is Ps. In the considered network, the direct links
is, only one RIS element is turned on each time from the BS to users are blocked for some clus-
while all other elements are set OFF, consequent- ters, due to the blocking objects, hence such users

188 IEEE Wireless Communications • December 2021

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 188 1/11/22 11:42 AM
 exploit RIS to establish communication. On the 9
other hand, some clusters have strong and direct
N=100
BS-to-user links such that the signal received via RIS
N=150
is negligible because of the increased path loss. 8
N=200
To better explore the role of RIS, we focus on N=250
a single blocked cluster with K users, where we

Total spectral efficiency (bps/Hz)

7 N=300
assumed the user 1 and user K to be the stron- N=350
gest and the weakest users, respectively, based
on their overall channel gains. Following the fixed 6
power allocation approach for NOMA users, the
highest channel gain user, that is, the strongest
user, gets the minimum share of transmit power 5
while the lowest channel gain user, that is, the
weakest user, gets the maximum share of the 4
transmit power. All wireless links, that is, BS to
RIS links and RIS to users’ links, are modeled as
Rayleigh fading channels with path loss and per- 3
turbed by additive white Gaussian noise (AWGN).
We modeled the channel estimates using the 2
minimum mean square error (MMSE) channel
estimation error model, whose quality of estima-
tion is indicated by the variance of the channel 1
estimation error, s 2e; the smaller the error variance 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
the better the estimation. The channel estimation Channel estimation error variance, 2e 10 -3
error is regarded as interference in the system,
which adversely affects the system performance. FIGURE 4. The total SE versus channel estimation error variance for different
number of RIS elements (N) with transmit power (Ps) = 30 dBm, and users
Impact of Imperfect CSI and RIS Elements (K) = 3.
The SE, that is, the ratio of achievable rate and
bandwidth, in the RIS-assisted NOMA network 1.2
highly depends on the quality of channel estima- Transmit Power,P
29 (dBm)
28
37

32

31
34 s

30
36

33
tion and the number of RIS elements. In Fig. 4
35
(bps/Hz)

we demonstrate the impact of varying channel

estimation error variance, and the number of RIS 1.1
elements on total SE, that is, the SE achieved by 27
K users. For any value of RIS elements, as error
Per user target spectral efficiency,

variance increases, the total SE decreases, since 1

the channel estimation error acts as a source of
26

28
interference. However, even with imperfect CSI,
29
30
31
32

high SE can be achieved by deploying a higher

34

0.9
33

number of RIS elements.

The transmit power consumption for a RIS-as- 27 25
sisted NOMA network depends on the number
of RIS elements, and the target spectral efficiency. 0.8
In this regard, Fig. 5 shows the impact of varying
per user target spectral efficiency and the number 24
26
28

of RIS elements on BS transmit power. First, it is

29

0.7
30

observed that for a given number of RIS elements,

31
32

the required transmit power increases with the 23

25
27

increase in target spectral efficiency. Second, the

transmit power scales down with the increase in 0.6
the number of RIS elements. For instance, for the 100 150 200 250 300
same per-user target spectral efficiency of 1 bps/ Number of RIS elements,N
Hz, a transmit power of 32 dBm is required for 150
RIS elements, while this value reduces to about 26 FIGURE 5. Required transmit power for varying per user target SE and number
dBm for 300 elements, which indicates 6 dB gain of RIS elements with channel estimation error variance (s2e) = 0.0005, and
by doubling the RIS elements. From this, we can users (K) = 3.
conclude that RIS passive reflection adds power
gain, which can be either utilized to improve SE or ferent frequency slots. It can be observed that for
reduce total power consumption. the same number of users and transmit power,
the number of RIS elements required to achieve
Factors Affecting the Size of RIS the target SE are greater for RIS-OMA than for
The minimum number of RIS elements required RIS-NOMA. This clearly depicts that the superi-
to guarantee the per user target SE varies with ority of NOMA compared to OMA still remains
many factors. In Fig. 6, we compare the number after introducing RIS. The number of RIS elements
of RIS elements required for RIS-NOMA and RIS- also increases with the increase in the number
OMA with a varying number of users and BS of users in the cluster, owing to the increase in
transmit power. Here, for RIS-OMA, we consider required orthogonal sub-bands. Moreover, the
frequency division multiple access (FDMA), where required number of RIS elements can be reduced
multiple users transmit data simultaneously at dif- by increasing the BS transmit power.

IEEE Wireless Communications • December 2021 189

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 189 1/11/22 11:42 AM
 350 RIS-Assisted Optical Wireless Communica-
RIS-NOMA, Ps = 40dBm
tion: Optical wireless communication (OWC)
is a promising solution for next-generation high
RIS-NOMA, Ps = 30dBm
300 data rate applications at relatively low hardware
RIS-OMA, P s = 40dBm cost and complexity than the RF counterpart.
RIS-OMA, P s = 30dBm Nonetheless, the performance of OWC is sub-
Number of RIS elements,N

250 ject to the existence of LoS between the trans-

ceivers. To relax this constraint, RIS can be
applied to OWC to mitigate the LoS blockages
200 by directing the optical beam in a desired direc-
tion. Thus, the integration of RIS and OWC can
enable a plethora of applications for both indoor
150 and outdoor scenarios.
RIS-Assisted mMIMO Network: mMIMO,
which is the extension of MIMO technology, is
100 one of the key enablers for dramatically improv-
ing the transmission gain and spectral efficiency.
However, high hardware cost and power con-
50 sumption are the fundamental limitations toward
the practical implementation of mMIMO sys-
tems. Nevertheless, RIS can be integrated with
0 mMIMO to provide required performance gains
1 2 3 4 in an energy-efficient and cost-effective fashion.
Number of users in cluster,K The low complexity algorithms for beamforming
designs and resource allocation for RIS-assisted
FIGURE 6. Number of RIS elements versus the number of users for RIS-NOMA mMIMO systems need to be studied to achieve
and RIS-OMA with per user target SE (t) = 1.2 b/s/Hz, and channel estima- maximum performance.
tion error variance (s 2e) = 0.0001.
Conclusion
In this article, we overviewed the potentials of
Research Directions RIS technology for 6G wireless networks. We
In this section, we present promising future first discussed the performance gains that can be
research directions that we consider to be of achieved by integrating RIS with emerging com-
great importance to unlock the full potential of munication technologies, such as NOMA, SWIPT,
RISs for 6G networks. UAVs, BackCom, mmWaves, and multi-antenna
RIS-Assisted Terahertz (THz) Communication: systems. Despite the great potential, RIS encoun-
THz communication, with ultra-wide bandwidth, ters new challenges to be efficiently integrated
is considered to be a promising candidate for 6G into the wireless network. In this regard, we
communication. Because of the ultra-high fre- exposed the crucial challenges for the practical
quency, the THz signal may undergo severe signal implementation of RIS-assisted networks. A case
attenuation and communication interruptions. To study for RIS-assisted NOMA networks under
this end, RIS can be applied to THz communi- imperfect CSI has also been presented to demon-
cation for better coverage performance. Howev- strate the importance of better channel estima-
er, the RIS-assisted THz communication poses a tion for RIS-assisted networks and to indicate the
major challenge of exploiting the unique propaga- various factors affecting the size of RIS. Finally,
tion properties of THz in the RIS-assisted network, to provide effective guidance for future research,
which needs to be addressed. we highlighted promising research directions for
Aerial RIS Empowered Communication: The RIS-assisted networks.
aerial RIS, carried by UAV or balloon, can realize
full-space reflections to serve a relatively larger num- References
[1] M. W. Akhtar et al., “The Shift to 6G Communications:
ber of users than the ground RIS, fixed at a location. Vision and Requirements,” Human-Centric Computing and
The aerial RIS is more likely to enjoy the LoS channel Information Sciences, vol. 10, 2020.
conditions, thus mitigating the blockages. The high [2] Q. Wu et al., “Towards Smart and Reconfigurable Envi-
mobility of UAVs can be exploited to further expand ronment: Intelligent Reflecting Surface Aided Wireless
Network,” IEEE Commun. Mag., vol. 58, no. 1, 2020, pp.
the coverage of aerial RIS. However, in practice, 106–12.
the aerial RIS brings new challenges, including the [3] S. Li et al., “Reconfigurable Intelligent Surface Assisted UAV
three-dimensional (3D) placement and the channel Communication: Joint Trajectory Design and Passive Beam-
estimation, that are thus worthy of investigation. forming,” IEEE Wireless Commun. Lett., vol. 9, no. 5, 2020,
pp. 716–20.
RIS-Assisted Physical Layer Security: The RIS’s [4] C. Pradhan et al., “Hybrid Precoding Design for Reconfig-
signal manipulation capability can even enhance urable Intelligent Surface Aided mmWave Communication
the physical layer security (PLS) of the communi- Systems,” IEEE Wireless Commun. Letters, vol. 9, no. 7, 2020,
cation links, by simultaneously boosting the signal pp. 1041–45.
[5] B. Di et al., “Hybrid Beamforming for Reconfigurable Intelli-
beam at the intended user and suppressing the gent Surface based Multi-User Communications: Achievable
beam at the unintended user. The RIS-assisted Rates with Limited Discrete Phase Shifts,” IEEE JSAC, vol. 38,
PLS requires the information of the channels from no. 8, 2020, pp. 1809–22.
the eavesdropper to the RIS and the BS, which is [6] E. Björnson et al., “Power Scaling Laws and Near-Field Behav-
iors of Massive MIMO and Intelligent Reflecting Surfaces,”
difficult to obtain in practice. This, therefore, calls IEEE Open J. Commun. Society, vol. 1, 2020, pp. 1306–24.
for a sophisticated channel estimation and RIS [7] Z. Ding et al., “On the Impact of Phase Shifting Designs on
passive beamforming design for RIS-assisted PLS IRSNOMA,” IEEE Wireless Commun. Lett., vol. 9, no. 10,
under imperfect CSI. 2020, pp. 1596–1600.

190 IEEE Wireless Communications • December 2021

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 190 1/11/22 11:42 AM
 [8] M. Fu et al., “Intelligent Reflecting Surface for Downlink Haris Pervaiz [S’09, M’09] ([email protected]) is an
Non-Orthogonal Multiple Access Networks,” IEEE Globecom assistant professor with the School of Computing and Commu-
Wkshps, 2019, pp. 1–6. nications (SCC), Lancaster University, U.K. From April 2017 to
[9] C. Pan et al., “Intelligent Reflecting Surface Aided MIMO October 2018, he was a research fellow with the 5G Innova-
Broadcasting for Simultaneous Wireless Information and tion Centre, University of Surrey, U.K. From 2016 to 2017, he
Power Transfer,” IEEE JSAC, 2020. was an EPSRC Doctoral Prize Fellow with the SCC, Lancaster
[10] D. Ma et al., “Enhancing Cellular Communications for University. He received his Ph.D. degree from Lancaster Univer-
UAVs via Intelligent Reflective Surface,” IEEE WCNC, 2020, sity, U.K., in 2016. His current research interests include green
pp. 1–6. heterogeneous wireless communications and networking, 5G
[11] X. Jia et al., “Intelligent Reflecting Surface-Aided Backscatter and beyond, millimeter wave communication, and energy and
Communications,” arXiv preprint arXiv:2004.09059, 2020. spectral efficiency.
[12] X.Jia et al., “IRS-Assisted Ambient Backscatter Communica-
tions Utilizing Deep Reinforcement Learning,” arXiv preprint Aamir Mahmood [M’18, SM’19] ([email protected])
arXiv:2103.07083, 2021. is an assistant professor of communication engineering at Mid
[13] J. Zuo et al., “Intelligent Reflecting Surface Enhanced Mil- Sweden University, Sweden. He received the M.Sc. and D.Sc.
limeter-Wave NOMA Systems,” IEEE Commun. Lett., 2020, degrees in communications engineering from Aalto University
pp. 1–1. School of Electrical Engineering, Finland, in 2008 and 2014,
[14] Y. Yang et al., “Intelligent Reflecting Surface Meets OFDM: respectively. He was a research intern at Nokia Researcher
Protocol Design and Rate Maximization,” IEEE Trans. Com- Center, Finland and a visiting researcher at Aalto University
mun., vol. 68, no. 7, July 2020, pp. 4522 - 35. from 2014 to 2016. His research interests include network time
[15] C. Huang et al., “Holographic MIMO Surfaces for 6G Wire- synchronization, resource allocation for URLLC, and RF interfer-
less Networks: Opportunities, Challenges, and Trends,” IEEE ence/coexistence management.
Wireless Commun., vol. 27, no. 5, 2020, pp. 118–25.
Zhiguo Ding [S’03, M’05, SM’17, F’20] (zhiguo.ding@manches-
Biographies ter.ac.uk) is currently a professor at the University of Manches-
S arah B asharat ([email protected]) ter. His research interests are 5G networks, signal processing
received her B.E. and M.S. degrees in electrical engineering and statistical signal processing. He has been serving as an edi-
from National University of Sciences and Technology (NUST), tor for IEEE TCOM, IEEE TVT, and served as an editor for IEEE
Pakistan, in 2019 and 2021, respectively. Her research interests WCL and IEEE CL. He received the EU Marie Curie Fellowship
include B5G and 6G communications, non-orthogonal multiple 2012-2014, IEEE TVT Top Editor 2017, 2018 IEEE ComSoc
access (NOMA), and reconfigurable intelligent surfaces (RISs). Heinrich Hertz Award, 2018 IEEE VTS Jack Neubauer Memorial
Award, and 2018 IEEE SPS Best Signal Processing Letter Award.
Syed Ali Hassan [S’08, M’11, SM’17] ([email protected].
pk) received his Ph.D. in electrical engineering from Georgia Mikael Gidlund [M’98, SM’16] ([email protected]) is a
Tech, Atlanta, in 2011, his M.S. in mathematics from Georgia professor of computer engineering at Mid Sweden University,
Tech in 2011, and his M.S. in electrical engineering from the Sweden. He has worked as a senior principal scientist and global
University of Stuttgart, Germany, in 2007, and a B.E. in electrical research area coordinator of wireless technologies, ABB Corpo-
engineering (highest honors) from the National University of Sci- rate Research, Sweden, project manager and senior specialist
ences and Technology (NUST), Pakistan, in 2004. Currently, he with Nera Networks AS, Norway, and a research engineer and
is working as an associate professor at NUST, where he is head- project manager with Acreo AB, Sweden. His current research
ing the IPT research group, which focuses on various aspects of interests include wireless communication and networks, wireless
theoretical communications. sensor networks, access protocols, and security.

IEEE Wireless Communications • December 2021 191

Authorized licensed use limited to: ULAKBIM UASL - Omer Halisdemir Universitesi. Downloaded on October 10,2023 at 18:58:48 UTC from IEEE Xplore. Restrictions apply.
BASHARAT_LAYOUT.indd 191 1/11/22 11:42 AM