IRS-Empowered Wireless Communications: State-of-the-Art, Key Techniques, and Open Issues
Ming Zeng, Ebrahim Bedeer, Xingwang Li, Quoc-Viet Pham, Octavia A. Dobre, Paul Fortier, Leslie A. Rusch
11 IRS-Empowered Wireless Communications:State-of-the-Art, Key Techniques, and OpenIssues
Ming Zeng, Ebrahim Bedeer, Xingwang Li, Quoc-Viet Pham, Octavia A. Dobre,
Fellow, IEEE , Paul Fortier, Leslie A. Rusch,
Fellow, IEEE
Abstract
In this article, we overview intelligent reflecting surface (IRS)-empowered wireless communicationsystems. We first present the fundamentals of IRS-assisted wireless transmission. On this basis, weexplore the integration of IRS with various advanced transmission technologies, such as millimeter wave,non-orthogonal multiple access, and physical layer security. Following this, we discuss the effects ofhardware impairments and imperfect channel-state-information on the IRS system performance. Finally,we highlight several open issues to be addressed.
Index Terms
Intelligent reflecting surface, millimeter wave, non-orthogonal multiple access, physical layer secu-rity, hardware impairments.
I. I
NTRODUCTION
Recently, intelligent reflecting surface (IRS) has drawn great attention as a promising physicallayer transmission technology for next-generation communication systems [1]–[4]. An IRS is
M. Zeng, P. Fortier, and L. A. Rusch are Universit´e Laval, Quebec, QC, G1V 0A6, Canada (e-mail: ming.zeng, paul.fortier,[email protected])E. Bedeer is with University of Saskatchewan, Saskatoon, S7N5A9, Canada, (email: [email protected])X. Li is with Henan Polytechnic University, Jiaozuo, 454000, China (e-mail: [email protected])Q. V. Pham is with Pusan National University, Busan, 46241, South Korea (e-mail: [email protected])O. A. Dobre is with Memorial University, St. John’s, NL, A1B 3X9, Canada (e-mail: [email protected]). a r X i v : . [ c s . I T ] J a n a planar surface equipped with massive low-cost passive reflecting elements; each can inducea phase and/or amplitude change to the impinging signals to achieve fine-grained reflectivebeamforming. By judiciously deploying an IRS in the environment, an extra communicationlink that goes through the IRS can be built between the transmitter (Tx) and receiver (Rx), andthus, better support diverse user requirements, such as extended coverage, increased data rate,minimized power consumption, and enhanced secure transmissions [1]–[4]. Not only theoreticallyattractive, IRS also possesses various advantages in terms of practical implementation. It is oflow hardware and energy cost, and can be easily deployed on the environment objects, e.g., thefacades of buildings. Moreover, IRS can operate in full-duplex mode without self-interferenceand noise amplification.To fully reap the benefits provided by IRS, it is necessary to investigate the integrationof IRS with other transmission technologies for next-generation communication systems, suchas millimeter wave (mmWave), non-orthogonal multiple access (NOMA), and physical layersecurity (PLS). The integration of IRS into mmWave is natural, since IRS can establish additionalline-of-sight (LoS) links to extend the coverage of mmWave, which suffers from severe signalattenuation and poor diffraction [5]–[7]. The application of IRS into NOMA is also promising,as IRS can be utilized to introduce desirable channel gain differences among users as well asto suppress the inter-user interference; both can lead to performance improvement of NOMAsystems [8], [9]. Lastly, IRS can be used to enhance the signal strength at the legitimate userswhile nulling the signal reception at the eavesdroppers, and thus, improve PLS of wirelesssystems [10]–[12].The aim of this article is to provide a comprehensive survey on IRS-assisted wireless trans-mission. We cover both conventional IRS-assisted systems and more advanced ones where IRSis integrated with other candidate technologies, such as mmWave, NOMA and PLS. For allthe considered scenarios, we not only present the state-of-the-art research progress, but alsoidentify the challenges and key techniques for resource allocation. Motivated by the practicalchallenges of implementing IRS, we further discuss the effects of hardware impairments (HWIs)and imperfect channel-state-information (CSI) on the IRS system performance. Finally, severalopen issues are highlighted. II. IRS-A
SSISTED W IRELESS T RANSMISSION
A. Motivation
Conventional network optimization in wireless communication systems has been limited totransmission control at the transceivers, with little attention has been paid to the wirelesspropagation environment. Indeed, the wireless propagation environment has long been perceivedas an uncontrollable and randomly behaving entity between the transceivers. Aside from beinguncontrollable, the environment usually has an adverse impact on the communication efficiency,owing to the signal attenuation, fading, and interference introduced. As a result, the propagationenvironment itself becomes a major limiting factor that hinders further performance improve-ment of wireless networks. Recently, there is an increasing demand for novel communicationparadigms that can smartly tune the propagation environment either to increase the commu-nication efficiency or to simplify the transceiver architecture. In this regard, IRS has receivedgreat attention owing to its ability to reconfigure the propagation environment via softwarecontrolled reflection [1]–[4]. As shown in Fig. 1, an IRS is a planar surface consisting of manylow-cost passive reflecting elements; each can induce a phase and/or amplitude change to theincident signal to achieve fine-grained reflective beamforming. When the direct link betweenthe transceivers fails due to unfavourable channel conditions, IRS can be deployed to build acascaded link and resumes the communication, as illustrated in Fig. 1(a). Even when the directlink exists, IRS can still be used to add an extra communication link between the transceiversto improve the system performance, as presented in Fig. 1(b).
B. State-of-the-Art and Key Techniques
The simplest IRS-assisted communication system consists of three nodes, namely the Tx,Rx, and IRS, which is referred to as the point-to-point communication. Depending on howmany antennas the transceivers are equipped with, such a system can be further classified intothree categories, i.e., single-input single-output (SISO), multiple-input single-output (MISO) andmultiple-input multiple-output (MIMO) systems. Existing works on MISO systems have revealedthat IRS can achieve squared power gain under asymptotically large number of reflecting elements[1]. Such a gain shows great potential of IRS and its superiority over conventional massiveMIMO systems, where only a linear gain is achieved [1]. It is further shown that by judiciouslycontrolling the IRS reflection coefficients, the IRS-assisted MIMO channel can be substantiallyenhanced in terms of channel power, condition number, or rank. BS User(a) without direct link (b) with direct link BS User
Fig. 1: IRS-assisted wireless transmission: a) without direct link; b) with direct link.In addition to point-to-point communication, IRS-assisted multi-user systems have also beenstudied [1], [2]. The performance analysis of multi-user systems is undoubtedly much morechallenging than that of a point-to-point system, owing to the existence of inter-user interfer-ence. To evaluate the fundamental capacity limits of IRS-assisted multi-user SISO systems, [2]characterizes the capacity and rate regions subject to the constraint on IRS reconfiguration times.Presented simulation results show that the capacity and rate regions can be greatly improvedusing IRS. The more general MISO scenario is studied in [1], and it is shown that employingIRS can significantly improve the coverage, energy consumption and achievable rate of wirelessnetworks.The works mentioned above focus on single-cell systems. The use of IRS in multi-cell systemsis however investigated in [3], [4]. Compared with the single-cell counterpart, new issues emergein multi-cell systems, such as where to deploy the IRSs and how to coordinate available resourcesamong different base stations (BSs). The authors in [3] consider a large-scale deployment ofIRSs in wireless networks, and characterize the achievable spatial throughput averaged over bothchannel fading and random locations of the deployed BSs/IRSs. It is unveiled that deployingdistributed IRSs can greatly boost the received signal power but only cause marginal extrainterference in the network. Note that SISO is assumed in [3]. The authors in [4] study the weighted sum rate maximization problem for a MIMO multicell system. Numerical results revealthat employing IRSs can notably enhance the cell-edge performance.In IRS-assisted systems, the IRS phase shifts need to be optimized in addition to the conven-tional transceiver optimization. As the IRS-assisted user channels are cascaded, the variables to beoptimized are often coupled, and thus, resulting in non-trivial joint resource optimization. More-over, the IRS optimization needs to satisfy the highly non-convex constant modulus constraint,since the IRS can only reflect the incident signal without amplifying it. Existing works oftenapply the block coordinate descent (BCD) method (also referred to as the alternating optimization(AO) when there are only two types of variables) to resolve the coupling among the optimizationvariables [1], [4]. On this basis, existing approaches developed for systems without IRS oftencan be borrowed for the transceiver optimization. Meanwhile, the semidefinite relaxation (SDR)technique is widely used for addressing the passive beamforming at the IRS. The frameworkcombining BCD/AO with SDR is shown to be effective in handling the joint resource allocation ofvarious IRS-assisted systems [1]. Nonetheless, it still suffers from two drawbacks: i) the obtainedsolution can only be considered as a lower bound; and ii) the complexity may be too high, sincethe high complexity SDR operation needs to be performed many times until convergence. Toobtain a tight upper bound, a potential solution is to apply the successive convex approximation(SCA) technique to construct a convex approximation for the joint optimization. This howevercould be non-trivial due to the coupling among the variables. Nevertheless, for certain cases, byexploiting the closed form solutions existing for the transceivers optimization under given IRSphase shifts, the problem could be simplified and it becomes relatively easy to apply the SCAtechnique. Additionally, the following two approaches could be adopted to replace SDR andthus, lower the complexity; one is SCA while the other is the complex circle manifold (CCM)method. For SCA, the majorization-minimization (MM) algorithm appeals to be quite promisingand the key then will be to find the appropriate surrogate function [4]. The usage of CCM onthe other hand is motivated by the complex forms of the IRS phases, and the main challengelies in how to design a gradient descent algorithm based on the manifold space [4].III. I
NTEGRATION OF
IRS
WITH A DVANCED T RANSMISSION T ECHNOLOGIES
To further exploit the potential of IRS, it is of interest to investigate the integration of IRSwith other advanced transmission technologies, including mmWave, NOMA, and PLS.
A. IRS-assisted mmWave Communication1) Motivation: mmWave communication has drawn considerable attention recently owing toits ability to provide ultra-wide bandwidth [5]–[7]. Nonetheless, mmWave communication suffersfrom severe signal attenuation and poor diffraction, which significantly limits its applications inmobile cellular systems. MIMO represents an effective technology to enhance the mmWavesignal strength owing to the high beamforming gain. However, the property of poor diffractionstill makes mmWave vulnerable to blocking by obstacles that break the LoS links. To addressthis, IRS can be deployed to create additional LoS links, and thus, extend the coverage ofmmWave systems [5]–[7].
2) State-of-the-Art and Key Techniques:
Earlier works on IRS-assisted mmWave MIMO focuson full digital beamforming at the BS, and have shown that deploying IRS can alleviate theblockage effect and enhance the performance of mmWave systems in coverage and throughput[5]. To lower the number of RF chains, one can either adopt the hybrid analog/digital beam-forming structure (shown in Fig. 2 [6]) or the lens antenna array (illustrated in Fig. 3 [7]). Theformer consists of two parts, namely the baseband digital beamforming under limited numberof RF chains and the RF band analog beamforming via a network of phase shifters. Likewise,the latter also comprises of two main components, namely the electromagnetic (EM) lens andthe matching antenna array with elements located in the focal region of the lens. The EMlenses provide controllable phase shifting to obtain angle-dependent energy focusing property.It is shown that using IRS can enhance the performance for both hybrid beamforming and lensantenna array based mmWave systems [6], [7].The BCD framework for microwave can be applied to mmWave as well. However, the channelmodel should be updated using the appropriate mmWave channel models, e.g., the geometricchannel model [5], [6]. Considering the poor diffraction and penetration abilities of mmWave,the direct links between the users and BS under mmWave are often assumed blocked [6], [7].Furthermore, the Tx-IRS and IRS-Rx channels can be approximated as a rank-one matrix/vector,as they are LoS dominated. This rank-one structure can be exploited for further simplifyingthe passive beamforming design at the IRS [5] and the active beamforming at the BS [6].Additionally, the analog beamforming under hybrid beamforming in general can be handled usingSDR as the phase shift optimization at the IRS, since both are limited by the same constant-modulus constraint. To lower the complexity, beam search based on pre-defined codebooks can
Baseband signalprocessing
PAPAPA
Digital precoding Analog Beamforming
User 1User 2User 3
Fig. 2: Structure of hybrid analog/digital beamforming.
Dimension-Reduced
DigitalProcessing
User 1 User 2User k
SelectingNetwork Lens
RF Chains
Fig. 3: Structure of lens antenna array.also be adopted. In terms of lens antenna array, the key lies in how to perform an appropriateantenna/beam selection to significantly lower the RF chain cost, without sacrificing the systemperformance too much [7]. Interference-aware beam selection could be of interest.
B. IRS-assisted NOMA Transmission1) Motivation:
NOMA is envisioned as a promising radio access technique for next-generationcommunication systems [8], [9]. By enabling multiple users to access the same time/frequency resources, NOMA can achieve higher spectral-efficiency and energy efficiency and better supportmassive connectivity when compared to orthogonal multiple access (OMA). Nevertheless, toobtain a decent performance gain of NOMA over OMA, the users are required to have a largechannel gain disparity. Such a requirement may be violated in conventional NOMA systems,since the user channels are determined by the highly stochastic propagation environments. Toovercome this, IRS can be utilized to introduce desirable channel gain differences among theusers via constructively or destructively adding the user signals (shown in Fig. 4). Meanwhile,IRS can also be used to suppress the inter-user interference, and thus, lead to improved throughputor fairness of NOMA systems.
2) State-of-the-Art and Key Techniques:
A body of research works has emerged very recentlywhich investigate the design of IRS-assisted NOMA systems [8], [9]. The authors in [8] inves-tigate the sum rate maximization for an IRS-assisted multi-user SISO-NOMA system, requiringto jointly optimize the channel assignment, decoding order of NOMA users, power allocation,and reflection coefficients. A three-step resource allocation algorithm is proposed and presentednumerical results demonstrate the superiority of IRS-assisted NOMA over conventional NOMAwithout IRS and IRS-assisted OMA in terms of system throughput. To further enhance the systemperformance, [9] aims to maximize the minimum rate of all users for an IRS-assisted MIMO-NOMA system, by jointly optimizing the transmit beamforming at the BS and the phase shifts atthe IRS. An efficient algorithm based on the framework of BCD with SDR is proposed to addressthe formulated non-convex problem. It is shown that the IRS-assisted MIMO-NOMA system cangreatly boost the rate performance, when compared with conventional NOMA without IRS andOMA with/without IRS.In IRS-assisted NOMA systems, resource allocation becomes more complicated due to theextra need for optimizing the decoding order [9]. Moreover, the optimization of the decodingorder is coupled with that of the IRS phase shifts, since the optimal decoding order cannot bedetermined without knowing the IRS phase shifts while the IRS phase shifts cannot be properlyconfigured without fixing the decoding order. To decompose the coupling among them, a viablesolution could be to iteratively update the decoding order and IRS phase shifts, by fixing theother. However, the complexity may become prohibited when the number of iterations requiredfor convergence is large. An alternative is to exhaustively search all decoding orders and on thisbasis, optimize the IRS phase shifts. Likewise, the resulting complexity of exhaustive searchmay be too high, especially when the number of users is large. To lower the complexity, one
BS User 1 User 2
Subtract User 2's signalSIC
Fig. 4: Illustration of IRS-assisted NOMA transmission.can greedily set the decoding order by fixing the IRS phase matrix to certain values, such asall zeros or ones. When the IRS phase matrix is set to all zeros, it means only the direct linkis considered. Clearly, this may be highly suboptimal due to the neglect of the effect of thecascaded Tx-IRS-Rx link. In contrast, when the IRS phase matrix is set to all ones, both thedirect link and the cascaded Tx-IRS-Rx link are considered. Nevertheless, the resulting decodingorder may still be quite different from the optimal one for the optimized IRS phase matrix. Toaddress this, the authors in [9] propose a combined-channel strength-based user ordering scheme,where users are ordered based on their maximally achievable combined channel strengths viaoptimizing the IRS phase shifts. Numerical results show that the proposed user ordering schemeachieves near-optimal performance with much lower complexity. Except for user ordering, extraconstraints should be imposed on users’ achievable rates in IRS-assisted NOMA systems, toensure the success of successive interference cancellation (SIC). That is, each user’s achievablerate cannot exceed the minimum rates decodable at all users that need to decode its signal. Suchextra constraints may make it more challenging to identify the initial feasible IRS phase shiftsrequired for the BCD based optimization framework. C. IRS-assisted PLS Systems1) Motivation:
Owing to the broadcasting nature of wireless transmission media, wireless sys-tems are vulnerable to impersonation attacks and eavesdropping. Encryption techniques representan effective way to ensure communication confidentiality; however, they may not be suitablefor some Internet-of-things applications that cannot afford their complexity and/or have stringentdelay requirements. Through exploiting the randomness nature of wireless propagation channels,PLS can help secure wireless communication confidentiality without consuming much of theresources. To ensure a non-negative secrecy rate, it is however often required that the legitimateusers experience better channel conditions than the eavesdroppers. Clearly, such a requirementdoes not always hold in conventional PLS systems. A simple workaround is to deploy IRS inPLS systems to reconfigure the channels for the legitimate users and eavesdroppers. In particular,IRS can be used to enhance the signal strength at the legitimate users while nulling the signalreception at the eavesdroppers, thereby enhancing the secrecy transmission rates.
2) State-of-the-Art and Key Techniques:
The research on IRS-assisted PLS systems is still atan earlier stage, and existing works mainly focus on simple scenarios with one legitimate user andone eavesdropper (shown in Fig. 5), e.g., [10]–[12]. Both [10] and [11] consider the case wherethe BS is equipped with multiple antennas, while the legitimate user and eavesdropper are withsingle antenna. [10] aims to maximize the secrecy transmission rate under the maximum transmitpower constraint, whereas [11] studies the transmit power minimization subject to secrecy rateconstraint at the legitimate user. Presented simulation results show that the proposed schemeswith IRS outperform their counterparts without IRS in terms of secrecy rate and transmit power.In addition, the authors in [12] analyse the secrecy outage probability of an IRS-assisted PLSsystem, where all network nodes are equipped with single antenna. Numerical results illustratethat deploying the IRS can lower the secrecy outage probability as well.Introducing PLS into IRS-assisted systems often leads to the original non-convex objectivefunction being more complicated, and thus harder to handle. As in conventional IRS systems,the BCD method can be used to decompose the coupling among the optimization variables,thereby making the problem more tractable. On this basis, there exist several ways to optimizethe transmit beamforming. First, a closed-form solution may be derived for the optimal trans-mit beamformer, e.g., [10], [11]. Second, the SCA technique, e.g., difference-of-convex (DC)programming, could be used, considering the expression of the secrecy rate. Last, fractional BS Legitimate userEavesdropper
Fig. 5: Illustration of IRS-assisted PLS systems.programming may also be employed, by removing the log( · ) operation in the objective function.For the IRS phase shifts optimization, semi-closed form solutions may exist in certain cases,e.g., [10]. Besides, SDR has been widely used to obtain a high quality solution, e.g. [11]. Last,the SCA technique, e.g., the MM algorithm can also be adopted.A summary of resource allocation for conventional and advanced IRS-assisted systems is givenin Fig. 6.IV. N ON -I DEAL T RANSMISSION OF
IRS-A
SSISTED C OMMUNICATION S YSTEMS
IRS-assisted communication systems, like any other communication systems, suffer from non-ideal transmission conditions (mainly result from HWIs and/or imperfect CSI) that can deterioratetheir performance if not properly taken into consideration. HWIs in IRS-assisted communicationssystems can result from the finite-resolution of the phase shifters at the IRS reflecting elementsand/or the RF front-end mismatches at the transceivers. In the rest of this section, we will discussHWIs and imperfect CSI and their effect on the performance of IRS-assisted systems. Resource Allocation
Conventional IRS-assisted systemsIRS-assisted mmWaveIRS-assisted NOMAIRS-assisted PLS
BCD/AOSCA IRS: SDR/CCMMM algorithmFull-digital beamformingOptimization variablesChallengesSolutions BS beamforming/powerIRS phase shiftsCoupling of variablesConstant modulus constraint of IRS phaseBS: existing solution without IRSHybrid analog/digital beamformingLens antenna array BCD/AO+SDR; Rank-1 channel for simplificationExtra variable SIC decoding orderExtra challenges Coupling of decoding order with phase optimizationSIC decoding constraintsSolutions Analog: SDR/beam searchAntenna/beam selectionmmWave channel modelExhaustive search of decoding orderBCD including decoding orderHigh complexityLow complexity Set decoding order under fixed phase valuesSet decoding order under optimized phase valuesExtra challengesSolutions More complicated objective functionBCD/AO Transmit beamformingIRS phase optimization Closed-formSCA, e.g., DC programmingFractional programmingSemi-closed formSDRSCA, e.g., MM algoirthm
Fig. 6:
Summary of research allocation for conventional and advanced IRS-assisted systems.
A. Hardware Impairments1) Finite-Resolution Phase Shifters:
IRS reflecting elements need to adjust their phase shiftsin real time to compensate for the time-varying nature of the wireless channel. Such phaseshift adjustments of IRS reflecting elements can be achieved via using positive intrinsic-negative(PIN) diodes, micro-electromechanical system based switches, or field-effect transistors [13]. Alarge number of studies on IRS consider that the phase shifts of the IRS reflecting elementscan change continuously, which is hard to achieve in practice. For instance, the phase shiftlevels of the PIN diode is typically adjusted to two levels (0 and π radians) by changing theapplied biasing voltage between two levels. To achieve M different phase shift levels, log M PIN diodes are required for each reflecting elements. Alternatively, a single varactor diode can be used to achieve more than two phase shifts; however, it requires a high number of biasingvoltages which will increase the complexity of the IRS controller. Manufacturing IRS reflectingelements to support a high number of phase shifts will increase its cost, and hence, will not bea scalable solution given that IRS typically has a very high number of reflecting elements. Theeffect of the finite-resolution phase shifters of the IRS reflecting elements is investigated in anumber of recent works. It is shown that for an asymptotically large number of IRS reflectingelements, a 1-bit phase shifter approaches the same squared power gain when compared to theideal continuous phase shifters. However, as the number of IRS reflecting elements decreases,the power loss increases and it depends on the number of the available phase shift levels.
2) RF Chain Impairments:
RF font-end impairments in IRS-assisted communication systemsinclude in-phase and quadrature imbalance (IQI) at the transceiver, phase noise of IRS, and othertransmission non-linearities. A general model to capture such RF impairments in IRS-assistedcommunication systems is the extended error vector magnitude (EEVM) proposed in [14]. In theEEVM model, the RF chain impairments are effectively modeled as multiplicative and additiverandom variables terms, with a certain mean and variance. This model has been applied to bothsingle- and multi-antenna transmitters in IRS-assisted communication systems. It is shown thatthe aggregate RF front-end impairments at both Tx and Rx can be accurately modeled as azero-mean complex Gaussian process with a variance that depends on the transmit power, theeffective channel gain of the Tx-IRS-Rx, and the square of the error vector magnitude at boththe Tx and Rx. Analysis of IRS-assisted communication systems employing the EEVM modelshows that at high signal-noise-ratio, the performance (in terms of spectral efficiency or outageprobability) is independent of the number of IRS reflecting elements, and is mainly constrainedby the level of the RF chain impairments at the Tx when compared to the phase distortion atthe IRS elements. Such analysis reveals that modest IRS elements with low-resolution phaseshifters can be used as affordable deployment solutions for IRS without significantly degradingthe performance.
B. Imperfect CSI
Estimating the indirect channel (between the Tx and IRS) and the reflection channel (betweenthe IRS and Rx) is challenging given the fact that IRS consists of passive reflecting elements withlimited signal processing capabilities. The indirect channel estimation problem can be partiallysolved with the knowledge of the angle of arrival given that the IRS elements are mounted on buildings, and hence, considered of fixed location. The reflection channel estimation is challeng-ing given the expected end users’ mobility, and errors in the estimation of the reflection channelcan significantly deteriorates the performance of IRS-assisted communication systems if theireffect is not properly considered in the IRS system design. In general, imperfect CSI estimationcan be addressed through a worst case/robust design, considering knowledge of the channelstatistics rather than the instantaneous channel coefficients, or allowing a controlled outage in theIRS system performance. In a robust design, the transmit power will be typically increased, whencompared to its counterpart assuming perfect CSI, to compensate for the estimation inaccuracy. Ifthe statistics of the CSI is known, it is possible to improve the performance of IRS systems (e.g.,achievable rate) on average rather than instantaneously. However, recent studies in the literaturereported that the the performance of IRS systems under statistical CSI knowledge deteriorateswith increasing the number of IRS elements serving a particular end user [15].V. O PEN I SSUES
In this section we highlight several open issues that are worthy of investigation.
A. Unmanned Aerial Vehicles-integrated IRS systems
Unmanned aerial vehicles (UAVs) have shown several benefits, as relays or flying BSs,to improve the performance of communication systems, and hence, they are currently beingconsidered as a key enabler of next-generation wireless systems. UAVs typically have a strongLoS and favorable propagation conditions to terrestrial BSs which will lead radio waves fromterrestrial BSs to interfere with UAVs in adjacent cells. IRS is expected to be a promisingcandidate solution to mitigate the inter-cell interference problem of future UAVs networks.This is because IRS can efficiently control the direction of travel of radio waves, through jointbeamforming with terrestrial BSs, to reduce the power leaked to UAVs in adjacent cells.
B. Machine Learning-empowered IRS Systems
Due to the coupling of optimization variables and non-convex nature of the underlying prob-lems, joint resource allocation in IRS systems is challenging to solve, and often sophisticatedsolutions with high complexity are required to obtain near-optimal performance. However, thetime-varying and highly dynamic nature of wireless networks requires the proposed solutionsto be of low complexity and execute easily. Such a dilemma is non-trivial to overcome using conventional optimization based methods. A promising way to tackle this is to employ machinelearning techniques, which have been shown as an effective tool to obtain near-optimal solutionsfor non-convex and sophisticated optimization problems under highly dynamic wireless environ-ments. Machine learning also holds the potential to learn the channel indirectly from the dataduring training, without the need for explicit CSI. Accordingly, machine learning-empoweredIRS systems are of practical interest. C. Sensing and Localization
Next-generation wireless communication systems will operate at higher frequencies (mmWaveor Terahertz band) to support applications that require sensing of the surrounding environmentand accurate localization. Such high frequencies have a limited number of propagation paths(mainly due to large penetration losses, high values of path loss, and low scattering) whichmay reduce the accuracy of sensing and localization. Hence, IRS as a controlled and dynamicscattering is considered a promising solution to the sensing and localization problem for next-generation wireless communication systems. One of the main research challenges is that for suchhigh frequencies and large size IRS, users are no longer in the far-field and conventional sensingand localization models are no longer valid. That said, sensing and localization models in thenear-field that exploit the information in the wavefront curvature need to be developed.VI. C
ONCLUSION
In this article, we surveyed IRS-empowered wireless networks. We first showed that judiciouslydeploying IRS can substantially improve the spectral efficiency, energy efficiency and coverageof wireless networks. On this basis, we validated that IRS can be further used to enhancethe performance of mmWave, NOMA, and PLS systems. However, the promised gains of IRSare often obtained under ideal assumptions on channel estimation and hardware configuration.Motivated by this, we further discussed the effects of HWIs and imperfect CSI on the performanceof IRS. Lastly, we identified three open issues for future research.R
EFERENCES [1] Q. Wu and R. Zhang, “Intelligent reflecting surface enhanced wireless network via joint active and passive beamforming,”
IEEE Trans. Wireless Commun. , vol. 18, no. 11, pp. 5394–5409, Nov. 2019.[2] X. Ma et al. , “Capacity and optimal resource allocation for IRS-assisted multi-user communication systems,” [Online].Available: arXiv:2001.03913, May 2020. [3] J. Lyu and R. Zhang, “Hybrid active/passive wireless network aided by intelligent reflecting surface: System modeling andperformance analysis,” [Online]. Available: arXiv:2004.13318, Sep. 2020.[4] C. Pan et al. , “Multicell MIMO communications relying on intelligent reflecting surfaces,” IEEE Trans. Wireless Commun. ,vol. 19, no. 8, pp. 5218–5233, Aug. 2020.[5] P. Wang et al. , “Intelligent reflecting surface-assisted millimeter wave communications: Joint active and passive precodingdesign,”
IEEE Trans. Veh. Technol. , pp. 1–1, Oct. 2020.[6] J. Qiao and M. Alouini, “Secure transmission for intelligent reflecting surface-assisted mmwave and terahertz systems,”
IEEE Wireless Commun. Lett. , vol. 9, no. 10, pp. 1743–1747, Jun. 2020.[7] Y. Wang et al. , “Energy efficiency optimization in IRS-enhanced mmwave systems with lens antenna array,” in
Proc. IEEEGLOBECOM , Dec. 2020, pp. 1–6.[8] J. Zuo et al. , “Resource allocation in intelligent reflecting surface assisted NOMA systems,”
IEEE Trans. Commun. , pp.1–1, Aug. 2020.[9] G. Yang, X. Xu, and Y.-C. Liang, “Intelligent reflecting surface assisted non-orthogonal multiple access,” [Online].Available: arXiv:1907.03133, Dec. 2019.[10] H. Shen, W. Xu, S. Gong, Z. He, and C. Zhao, “Secrecy rate maximization for intelligent reflecting surface assistedmulti-antenna communications,”
IEEE Commun. Lett. , vol. 23, no. 9, pp. 1488–1492, Sep. 2019.[11] Z. Chu, W. Hao, P. Xiao, and J. Shi, “Intelligent reflecting surface aided multi-antenna secure transmission,”
IEEE WirelessCommun. Lett. , vol. 9, no. 1, pp. 108–112, Jan. 2020.[12] L. Yang et al. , “Secrecy performance analysis of RIS-aided wireless communication systems,”
IEEE Trans. Veh. Technol. ,vol. 69, no. 10, pp. 12 296–12 300, Oct. 2020.[13] P. Nayeri, F. Yang, and A. Z. Elsherbeni,
Reflectarray Antennas: Theory, Designs and Applications . Wiley Online Library,2018.[14] T. Schenk,
RF imperfections in high-rate wireless systems: impact and digital compensation . Springer Science & BusinessMedia, 2008.[15] H. Guo et al.et al.