1y ago

17 Views

1 Downloads

1.61 MB

16 Pages

Transcription

1A State-of-the-Art Survey on Reconfigurable Intelligent Surface AssistedNon-Orthogonal Multiple Access NetworksZhiguo Ding, Lu Lv, Fang Fang, Octavia A. Dobre, George K. Karagiannidis, Naofal Al-Dhahir, Robert Schober,and H. Vincent PoorAbstract—Reconfigurable intelligent surfaces (RISs) and nonorthogonal multiple access (NOMA) have been recognized askey enabling techniques for the envisioned sixth generation (6G)of mobile communication networks. The key feature of RISs isto intelligently reconfigure the wireless propagation environmentwhich was once considered to be fixed and untunable. The keyidea of NOMA is to utilize the users’ dynamic channel conditionsto improve spectral efficiency and user fairness. Naturally, thetwo communication techniques are complementary to each otherand can be integrated to cope with the challenging requirementsenvisioned for 6G mobile networks. This survey provides acomprehensive overview of the recent progress on the synergisticintegration of RISs and NOMA. In particular, the basics of thetwo techniques are introduced first, and then the fundamentalsof RIS-NOMA are discussed for two communication scenarioswith different transceiver capabilities. Resource allocation is ofparamount importance for the success of RIS assisted NOMAnetworks, and various approaches, including artificial intelligence(AI) empowered designs, are introduced. Security provisioning inRIS-NOMA networks is also discussed as wireless networks areprone to security attacks due to the nature of the shared wirelessmedium. Finally, the survey is concluded with detailed discussionsof the challenges for the practical implementation of RIS-NOMA,future research directions, and emerging applications.I. I NTRODUCTIONWith the rapid rollout of the fifth generation (5G) mobilenetworks, the focus of the research community is currentlyshifting towards the design of beyond 5G (B5G) and the sixthgeneration (6G) networks [1]–[4]. However, the envisionedperformance metrics of B5G and 6G are much more demanding than those for the previous generations of wireless systems.For example, B5G and 6G systems are expected to supportusers with extremely diverse data rate requirements, wherean augmented reality (AR) or virtual reality (VR) user mightdemand a peak data rate of several Tbps, but an Internet-ofThings (IoT) node could be served with a data rate in thekbps range. Another example is that the connection densitysupported in B5G and 6G systems is expected to be 100 timeshigher than that of 5G, e.g., more than 107 devices per squarekilometre are to be connected [1]. To meet these demandingand diverse requirements for B5G and 6G, more intelligenttechniques that can efficiently increase transmission reliability,improve system throughput, and support massive connectivityare needed.Reconfigurable intelligent surfaces (RISs) and nonorthogonal multiple access (NOMA) have been recognized astwo very promising communication techniques to meet theaforementioned challenges for the design of B5G and 6Gsystems. On the one hand, an RIS can intelligently reconfigurea mobile user’s propagation environment, which ensures thatthe user’s data rate and reception reliability can be significantlyimproved [5]–[13]. Note that these performance improvementsare realized in a low-cost, energy-efficient, and spectrallyefficient manner, since no additional spectrum needs to beacquired and the success of RISs is due to their ability to createfavourable radio propagation conditions. On the other hand,NOMA can effectively increase spectral efficiency, supportmassive connectivity, improve user fairness, and reduce transmission latency, by encouraging dynamic spectrum sharingamong mobile users and opportunistically exploiting theirheterogeneous channel conditions and quality-of-service (QoS)requirements [14]–[22].Naturally, these two important enabling technologies arecomplementary to each other, where the use of NOMA canimprove the spectral efficiency and connectivity of RIS systems, and the use of RISs ensures that the users’ propagationenvironments can be effectively and intelligently customizedfor the implementation of NOMA. In particular, many forms ofNOMA have been developed to encourage dynamic spectrumsharing among mobile users by opportunistically exploitingthe users’ heterogeneous channel conditions. Conventionally,a user’s channel conditions are viewed as a type of fixedand non-tunable phenomenon that is solely determined by theuser’s propogation environment. Therefore, if the users’ channel conditions are not ideal for the application of NOMA, e.g.,the users have similar channel conditions, the performancegain of NOMA over conventional orthogonal multiple access(OMA) can be quite limited. RISs enable a paradigm shiftfor the design of intelligent NOMA, since the use of an RISensures that the propagation environment can be effectivelyand intelligently customized for the needs of NOMA.This paper provides a comprehensive overview of theopportunities and challenges that arise in connection withthe integration of RISs and NOMA in the envisioned nextgeneration mobile networks. In particular, we will focus onthe following aspects: The basics of the two considered communication techniques, namely RISs and NOMA, are reviewed first. Inparticular, the capability of NOMA to facilitate spectrumsharing among mobile users to improve spectrum efficiency is described, and the key idea behind RISs, i.e.,reconfiguration of the users’ propagation environment, isillustrated. Then, the fundamentals of RIS-NOMA are presented,where the benefits of the integration of RIS and NOMAare unveiled. Furthermore, the quasi-degradation criterionis used as a metric to illustrate how the use of RISguarantees that NOMA can realize the same performanceas dirty paper coding, but with lower computationalcomplexity. Dynamic resource allocation is crucial for realizingthe performance gains enabled by RIS-NOMA and isreviewed. In particular, RIS-NOMA offers additional

24RIS3g0gh2User1Base stationFig. 1. 0An illustration example for RIS transmission.degrees-of-freedom (DoFs) in the spatial, frequency, andtime domains. We will show how these DoFs can beeffectively exploited to improve the system performanceby applying matching and game theoretic techniques.Artificial intelligence (AI)-empowered RIS-NOMA ispresented. In particular, the benefits of using AI toolsfor realizing long-term performance gains in RIS-NOMAnetworks are illustrated, where conventional optimizationtools serve as benchmarks.Security provisioning via RIS-NOMA is also covered inthis survey. The application of RIS-NOMA to enhancephysical layer security with respect to passive eavesdropping is studied first, where the impact of using anRIS on the secrecy performance is illustrated. Then, theapplication of RIS-NOMA to covert communications isinvestigated.While RIS-NOMA can offer various performance gains,its practical implementation faces many challenges,which are discussed here as well. In particular, the designof channel estimation schemes for RIS-NOMA networksis considered. This is a challenging problem, due tothe passive nature of RIS arrays and the multi-usernature of NOMA. Low-complexity solutions for practicaldeployment of RISs in NOMA networks are introduced.RIS-NOMA has widespread applications in various communication network architectures, which are presented inthis survey. In particular, the exploitation of unmannedaerial vehicles (UAVs) equipped with RISs to supportaerial radio access networks is described. Other applications of RIS-NOMA, such as mobile edge computing (MEC) and simultaneously wireless information andpower transfer (SWIPT), are presented in the survey aswell.The remainder of this survey is organized as follows. InSection II, the basics of RIS and NOMA are reviewed, andthe fundamentals of RIS-NOMA are elaborated in SectionIII. Existing resource allocation approaches for RIS-NOMAare surveyed in Section IV, where AI empowered approachesare also discussed. Security provisioning and the practicalimplementation of RIS-NOMA are considered in Sections Vand VI, respectively. The survey is concluded with a discussionof various future research directions and emerging applicationsfor RIS-NOMA.051015202530Fig. 2. Illustration of the performance of RIS transmission. The distancebetween the BS and the user is 15 m. For Case 1, the RIS, is located at avertex of an equilateral triangle whose other two vertices are the user and theBS, and for Case 2, the RIS is located at the center of the aforementionedequilateral triangle. Both small scale Rayleigh fading and large scale path loss0are considered. The path loss is modelled by dcα, where α denotes the pathloss exponent and c0 denotes the reference power gain at the distance of 1m. The path loss exponent for the channel between the BS and the user is4.5, i.e., the direct link suffers severe blockage, the path loss exponents forthe channels related to the RIS are 2.2, and c0 30 dB [23]–[25]. Theuser’s target data rate is 4 bit per channel use (BPCU). The noise power is 80 dBm.II. BASICS OF RIS S AND NOMAIn this section, the basics of the two considered communication techniques, RISs and NOMA, are briefly reviewed.A. RISsUnlike conventional information transmitters and receivers,RISs themselves do not have any information to send, butare deployed to assist information transceivers [5]–[7]. Thebasic idea of RIS systems can be illustrated by using thesimple example shown in Fig. 1, where a single-antenna basestation (BS) communicates with a single-antenna user via anRIS equipped with N reflecting elements. For this illustrativeexample, the signal received by the user can be expressed asfollows:y (h g0H Θg)s w,(1)where s denotes the information symbol, h denotes the channelgain between the BS and the user, Θ is an N N diagonalmatrix, g0 denotes the channel vector between the RIS andthe user, g denotes the channel vector between the BS andthe RIS, and w denotes the noise. In the RIS literature, Θis termed the phase shifting matrix, since the RIS does notchange the amplitudes of the reflected signals, but alters theirphases only [8]–[10]. Denote the i-th main diagonal elementof Θ by ejθi , where θi is the phase shift caused by the i-threflecting element on the RIS.In Fig. 2, the performance achieved for the considered RISassisted single-user scenario is illustrated by using the outagerate as the performance metric. In particular, the outagerate is defined as follows: R 1 P log 1 ρ g0H Θg 2 R ,where P log 1 ρ g0H Θg 2 R denotes the probabilityof the outage event log 1 ρ g0H Θg 2 R , ρ denotesthe transmit signal-to-noise ratio (SNR), and R denotes theuser’s target data rate. For the considered simple example, thephase shifts are obtained by first generating N sets of randomphases and selecting the best set yielding the largest data rate.

3Note that finding the optimal phase shifts are the key stepfor the design of RIS systems, and the design principles foroptimizing the RIS phase shifts will be provided in detail inthe following sections.As can be observed from Fig. 2, the use of an RIS withN 20 reflecting elements can already offer a significantperformance gain over the scheme without RIS, e.g., witha transmit power of 20 dBm, the data rate achieved by thescheme without RIS is 1.5 bit per channel use (BPCU), but theRIS-assisted scheme can realize a data rate of 2.3 BPCU. Byincreasing the number of reflecting elements, the performancegain achieved with the RIS can be further improved. Forexample, Fig. 2 shows that the RIS with N 50 reflectingelements yields two times the data rate compared to thescheme without RIS. It is also interesting to observe thatthe performance of the RIS assisted system is affected bythe location of the RIS, where a more detailed discussionsregarding how to optimize the deployment of RISs will beprovided in Section VI.User 21 We note that there are various forms of NOMA, which exploit not onlythe users’ different channel conditions, but also the users’ heterogeneous QoSrequirements [14], [15], [26]–[30]. h2 2 α12 h2 2 α22 1ρ"!"log 1 ρ h2 2 α22Decode User1’s signalDecode User2’s signalh2!log 1 h1 2 α12 h1 2 α22 1ρ"Decode User 1’ssignal directlyh1User 1Base station(a) A two-user NOMA example765432B. NOMAAs a multiple access technique which allows multiple usersto concurrently use scarce bandwidth resources, NOMA isfundamentally different from conventional OMA, which permits users to only individually occupy orthogonal bandwidthresource blocks. Instead, the key principle of NOMA transmission is to encourage spectrum sharing among users, e.g.,multiple users are allowed to be served at the same time andfrequency [14], [15]. The key idea behind NOMA can beillustrated based on power-domain NOMA1 which encouragesusers with different channel conditions to share the spectrum[31], [32]. The success of power-domain NOMA is based onits capability to efficiently exploit the users’ dynamic channelconditions, as explained in the following.Consider a simple downlink scenario with one BS andtwo users, as illustrated in Fig. 3(a), where each node isequipped with a single antenna. Assume that user 1 has a weakchannel gain, denoted by h1 , and user 2 has a strong channelgain, denoted by h2 , i.e., h1 2 h2 2 . In conventionalOMA, each user occupies a different orthogonal bandwidthresource, such as a time slot, whereas in NOMA, both usersare served simultaneously in the same resource. In particular,the BS sends a superimposed signal, i.e., x α1 s1 α2 s2 ,where si denotes user i’s signal and αi denotes the powerallocation coefficient for user i’s signal. Unlike conventionalpower allocation polices which allocate more power to userswith strong channel conditions, NOMA allocates more powerto the weak user, i.e., α1 α2 , in order to ensure that bothusers are connected. The two users adopt different detectionstrategies, depending on their channel conditions. As the weakuser, user 1 treats user 2’s signal as noise and tries to directly h 2 α2decode its own message with a rate of log 1 h 12 α2 1 1 .12ρAs the strong user, user 2 applies successive interferencecancellation (SIC), i.e., it decodes user 1’s signal first with!log 1 10051015202530(b) Outage sum rates achieved by NOMAFig. 3. An illustrative example for NOMA. The distance between the BSand user 1 is 15 m, and the distance between the BS and user 2 is 10 m. Bothsmall scale Rayleigh fading and large scale path loss are considered, with apath loss exponent of 4. User 1’s target data rate is R1 1.5 BPCU, anduser 2’s target date rate is denoted by R2 . The other simulation parameters ones used for are the same as the2 2Fig. 2. h αa rate of log 1 h 22 α2 1 1 . If the first stage of SIC is22ρcarried out successfully, user 2 can remove user 1’s signal and decode its own signal with a rate of log 1 ρ h2 2 α22 .On the other hand, the data rates achieved by OMA are givenby 21 log 1 hi 2 , for i {1, 2}, where the factor 12 is dueto the fact that each user can use the resource half of the timeonly.In Fig. 3(b), the outage sum rate is used as a metric forperformance evaluation, where the outage sum rate is definedas R1 (1 P1 ) R2 (1 P2 ), with Ri as the user i’s targetdata rate,!! h1 2 α12 R1 ,(2)P1 P log 1 h1 2 α22 ρ1and h2 2 α12P2 1 P log 1 h2 2 α22 log 1 ρ h2 2 α22 R2 .1ρ! R1 ,(3)As can be observed from the figure, NOMA can yield asignificant performance gain over OMA. For example, forR2 5 BPCU and a transmit power of 15 dBm, the outagesum rate realized by NOMA is 4.6 BPCU, which is more thanthree times the data rate achieved by OMA. An importantobservation from the figure is that the performance gainof NOMA over OMA diminishes when the two users havesimilar target data rates. Thus, NOMA cannot only exploit the

47User 110-9RIS6543User 2Base station2Fig. 4.Illustration for a two-user RIS-NOMA network.1users’ heterogeneous channel conditions, but also other userheterogeneities, such as users’ different mobility profiles andQoS requirements [14], [15], [26]–[30].III. F UNDAMENTALS OF RIS-NOMAAlthough the concepts of RISs and NOMA have beendeveloped separately in the literature, the two communicationtechniques are naturally complementary to each other. In thissection, the fundamentals of RIS-NOMA are described andthe benefits of RIS-NOMA are illustrated based on a simpletwo-user example. The two users are denoted by user 1and user 2, respectively, as shown in Fig. 4. In particular,two scenarios, namely single-input single-output (SISO) andmultiple-input single-output (MISO) transmission, are studiedin the following subsections, respectively.A. SISO-RIS-NOMAConsider a SISO scenario, where each node in the networkis equipped with a single antenna. For this simple scenario,the use of NOMA is clearly motivated since it can ensure thatthe two users are served simultaneously, whereas RIS-OMAcan support only a single user at a time. The application ofRISs to NOMA is motivated in the following.Without RIS, the performance gain of NOMA over OMAis critically depending on the two users’ channels, denoted byh1 and h2 , respectively, as shown in the following. If powerdomain NOMA is used and the users’ channels are the same,the achievable sum rate is given by! h1 2 α12log 1 log 1 ρ h2 2 α22(4)122 h1 α2 ρ log 1 ρ h2 2 ,which is the same as the sum rate achieved by OMA.Conventionally, the users’ channel conditions are viewed asbeing fixed to be solely determined by the users’ propagationenvironment. However, the use of RISs opens up opportunitiesfor intelligently reconfiguring the users’ propagation environment in order to facilitate the application of NOMA, which canlead to significant performance gains of NOMA over OMA.In particular, with RIS, the two users’ effective channel gainsare given by [5], [6]h̃i hi gH Θgi ,i {1, 2},(5)where gi denotes the channel vector between the RISs anduser i, and g and Θ are defined in Section II. By intelligentlytuning Θ, RIS-NOMA can introduce more DoFs for systemdesign. This can be exploited to not only generate a significantdifference in the users’ channel gains but also to customize012345678910Fig. 5. Illustration of the benefit of using SISO-RIS-NOMA with N 100.The BS and the two users are located at the vertices of an equilateral trianglewith side length d 15 m, and the RIS is located at the center of thetriangle. For illustrative purposes, small scale fading is omitted for the users’direct channels to the BS, such that the users have the same channel gains,h1 h2 , because their distances to the BS are the same. The other simulationparameters are the same as the ones used for Fig. 2. The 10 cases are generatedby using 10 random phase shifts.the users’ effective channel gains according to the users’ QoSrequirements, as shown in Fig. 5, where it is assumed thath1 h2 . As discussed before, this situation is not suitable forthe application of power-domain NOMA. As can be observedfrom the figure, the use of RISs can effectively generatesufficient difference between the users’ effective channel gains.For example, for case 9 shown in Fig. 5, user 1’s channel gainis about 2 times larger than user 2’s channel gain, which meansthat user 1 can act as the strong user and user 2 can act asthe weak user in power domain NOMA. If user 2 has a morestringent target data rate and needs to be treated as the stronguser, Fig. 5 shows that this can be realized by simply adjustingΘ.The fundamentals of SISO-RIS-NOMA have been thoroughly analyzed in the literature from various perspectives. Forexample, in [33], the energy efficiency of downlink SISO-RISNOMA has been studied, where SISO-RIS-OMA has beenused as a benchmark. As shown in [33], when the users havedifferent data rate requirements, the use of RIS-NOMA canrealize a reduction in the transmit power by almost a factor oftwo, even if the users have similar channel conditions, whereasthe performance gain of RIS-NOMA diminishes if the usershave similar target data rate requirements. Another performance criterion used in the literature is reception reliability,where the works in [34], [35] demonstrate that downlink SISORIS-NOMA can efficiently utilize the spatial DoFs to improvethe reception reliability, compared to RIS-OMA. In particular,the use of SISO-RIS-NOMA can ensure that the diversitygain achieved by each user in the system is proportional tothe number of reflecting elements of the RIS, even thoughthe multiple users share the same bandwidth resource. Theapplication of SISO-RIS-NOMA for uplink transmission hasbeen studied in [36], where the sum rate is used as the metric.In particular, the authors of [36] have shown that the use ofRIS-NOMA can yield significant performance improvementwhen the number of RIS elements is large. Another importantfinding in [36] is that the use of NOMA is particularly usefulin crowded scenarios, e.g., the number of users a performancegain of 5 BPCU over OMA is achievable for the two-usercase, and it can be increased to 8 BPCU when there are 6

5h2 (Case 1)h2 (Case 2)h1 (fixed)NQD Region(a) NQD without RISsNDQ Region without RISNDQ Region with RIS(b) An RIS with 10 reflecting elementsFig. 6. Improvement of the quasi-degradation region by using RIS, withh1 [1 1]T and the two users’ target data rates being 1 BPCU. G and gi ,i {1, 2}, are randomly generated. NQD represents non-quasi-degradation.users.B. MISO-RIS-NOMAConsider an MISO scenario, where the BS is equippedwith multiple antennas and each user has a single antenna.The benefit of the integration of RISs and NOMA can beclearly illustrated by using the quasi-degradation criterion, aninsightful condition under which the use of MISO-NOMAresults in the same performance as dirty paper coding (DPC)but with less complexity [37], [38].The quasi-degradation criterion can be illustrated by considering a simple two-user downlink scenario, where the BShas two antennas, and each user has a single antenna. Thetwo users’ channel vectors are denoted by hi and assumedto be areal-valued vectors. Furthermore, Tassume that user 1’schannel vector is fixed, i.e., h1 1 1 . The shaded regionshown in Fig. 6(a) is obtained by highlighting the non-quasidegradation (NQD) region, i.e., those realizations of h2 whichdo not meet the quasi-degradation condition [37], [38]. Forexample, Case 1 shown in the figure, i.e., the two users’channel vectors are almost aligned, is in the quasi-degradationregion. This is an ideal case for using NOMA, since theusers’ channel vectors occupy a single spatial direction andone beam can be used to serve both users simultaneously. Case2, i.e., the two users have orthogonal channels, is in the NQDregion. This means that the use of NOMA for this case mayresult in a throughput loss compared to DPC, even though theuse of NOMA can still offer other advantages, especially inoverloaded scenarios in which the number of users exceeds theavailable resources. By using RISs and tuning the RIS phaseshifting matrix, it is possible to reconfigure the users’ channelvectors, such that they satisfy the quasi-degradation criterion.For example, Fig. 6(b) reveals that by using an RIS with 10reflecting elements, the NQD region is diminishing, comparedto the case without an RIS shown in Fig. 6(a). In otherwords, if there are a sufficient number of reflecting elementson the RIS, it can be ensured that the use of RIS-NOMAyields the same performance as DPC, regardless of what theusers’ original channel conditions are. Furthermore, since bothNOMA users occupy only a single spatial dimension, theremaining dimensions can be used to accommodate additionalusers.A more formal study of the impact of RISs on the quasidegradation criterion was provided in [39], where a newexpression of the quasi-degradation criterion for MISO-RISNOMA was developed. While the quasi-degradation criterionsheds light on whether NOMA can realize the same performance as DPC, it still has a few limitations. For example,the expressions for the quasi-degradation criterion reportedin [39] have been developed by fixing the SIC decodingorder, whereas recent studies have shown that opportunisticallychoosing SIC decoding orders can yield significant performance improvement in NOMA networks [40], [41]. Furthermore, the criterion has been investigated mainly for the simpletwo-user case, and its generalization to the more importantoverloaded cases, where the number of users exceeds thenumber of antennas at the BS, is still unknown.Compared to SISO-RIS-NOMA and MISO-RIS-NOMA,the fundamentals of RIS-NOMA for more general setups,e.g., when there are multiple RISs or/and when each node isequipped with multiple antennas, have been less investigatedin the literature, because of their challenging nature. Onepromising approach is to use RIS-NOMA as a type of addons, and combine it with conventional spatial division multipleaccess (SDMA). Figure 7 illustrates this approach, whereconventional SDMA is used to form orthogonal beams andserve multiple primary users. Assume that there are additionalsecondary users to be connected. Conventionally, it is notpossible to serve these additional users directly with theexisting spatial beams, since the secondary users’ channelsmay not be perfectly aligned with the primary users’ channelsdue to the random radio propagation environment. The useof RISs opens the possibility to reconfigure the propagationenvironment and align the secondary users’ channel vectorswith the existing SDMA beams. As a result, the additionalsecondary users can be served without changing the SDMAlegacy system, as shown in Fig. 7. In the next section, moresophisticated designs for RIS-NOMA will be considered byexploiting optimization based resource allocation.IV. DYNAMIC R ESOURCE A LLOCATION FOR RIS-NOMADriven by the benefits of RIS-NOMA systems, suchas their high spectral/energy efficiency, resource allocation/optimization has attracted significant research attention

6- Primary Users- Secondary Users- RISsFig. 7.Illustration of the benefit of RIS-NOMA, where primary usersare served via conventional SDMA and the use of RIS-NOMA ensuresthat additional secondary users can be served by an existing SDMA beamsimultaneously.to further improve the communication performance. In particular, by jointly optimizing the communication resources,e.g., beamforming, power, and sub-channels, and the reflectingcoefficients of each RIS, the strength of the received signalcan be significantly improved. In this section, we review traditional resource allocation methods, the related optimizationproblems, and their solutions in the context of RIS-NOMAsystems, including sum rate maximization, spectral/energyefficiency maximization, power minimization and physicalsecure communications issues, as summarized in Table I.Subsequently, we discuss AI/ML-based resource allocation inRIS-NOMA.A. Throughput/Data Rate MaximizationDue to the non-convexity of the optimization problemsformulated for throughput maximization, the alternating optimization algorithm proposed in [42] is widely adopted todesign the transmit beamforming at the BSs and the RISphase shifts in an alternating manner. Let us start with asimple system model, i.e., a SISO system. The average sumrate of a two-user downlink SISO-RIS-NOMA network wasmaximized by alternatingly applying a scheme with two phaseshift adjustments (namely, dynamic phase adjustment andone-time phase adjustment) and a power allocation algorithm[43]. Considering a wireless powered RIS-NOMA system withmultiple devices/users, the authors in [44] first proved thatdynamic phase shifts are not needed for downlink NOMAand uplink NOMA systems, which simplifies the problem andreduces the signalling overhead. In addition to the optimizationof the beamforming at the BS and the phase shifts of the RIS,the channel assignment and decoding order of the NOMAusers were optimized to maximize the system throughput[45]. To compare three multiple access technologies, namelyNOMA, frequency division multiple access (FDMA), andtime division multiple access (TDMA), the weighted sumrate was studied for RIS systems in [46]. In this work, ajoint optimization algorithm for the deployment location andthe reflection coefficients of the RIS as well as the powerallocation was proposed and shown to achieve near-optimalperformance. Furthermore, the use of a BS equipped withmultiple antennas was considered in RIS-NOMA systems [47],where a scheme for jointly optimizing active beamforming atthe BS and the passive beamforming at the RIS was proposedin order to achieve the maximum sum rate. In addition todownlink case, the sum rate maximization problem was alsostudied for uplink RIS-NOMA systems [36].RIS-NOMA is highly compatible with other key technologies for 6G communication systems, including mmWave [48],[49] and massive MIMO [50]. The authors in [48] consideredthe design of mmWave networks and proposed an optimizationalgorithm to maximize the system sum rate, where the powerallocation and reflection coefficients are optimized alternately.The results shown in [48] illustrate the effectiveness of integrating RISs in mmWave-NOMA systems. Recall that thesmall wavelength of mmWave signals enables the deploymentof a large number of antennas, which motivates the design ofmmWave massive MIMO systems. [49] considered the application of RIS-NOMA in mmWave massive MIMO systems,where the effects of the power leakage and the per-antennapower constraint have been joint investigated. In this work, twomulti-beam selection strategies were proposed to maximizethe system weighted sum rate by jointly optimizing the activebeamforming at the BS and the passive beamforming at theRIS. Regarding imperfect SIC in NOMA, the capabilitiesof RISs to manipulate wave polarization in dual-polarizedMIMO-NOMA networks was investigated in [50], where theproposed n

gain of NOMA over conventional orthogonal multiple access (OMA) can be quite limited. RISs enable a paradigm shift for the design of intelligent NOMA, since the use of an RIS ensures that the propagation environment can be effectively and intelligently customized for the needs of NOMA. This paper provides a comprehensive overview of the

Related Documents: