《全球6G技術大會：智能超表面白皮書（英文版）（42頁）.pdf》由會員分享，可在線閱讀，更多相關《全球6G技術大會：智能超表面白皮書（英文版）（42頁）.pdf（42頁珍藏版）》請在三個皮匠報告上搜索。

1、1AbstractIn the emerging study of 6G technology,Reconfigurable Intelligent Surface(RIS)is a new type of artificial electromagnetic surface that can intelligently anddynamicallycontroltheelectromagneticcharacteristicsofeachunitoftheelectromagnetic surface through digital programming.It will realize a

2、ccurate andefficient control of electromagnetic waves and change the electromagnetic wavepropagation environment,bringing a new 6G airspace transmission paradigm to thefuture.This white paper introduces RIS from multiple perspectives such as applicationscenarios,requirements,key technologies and alg

3、orithms,current research status,field verification,and proposes suggestions on industry maturity,developmentdirection,and standardization research direction.2ContentsAbstract.1Chapter I Overview.31.1.Concept.31.2.PotentialApplication Scenarios and Requirements.41.3.Current Research at Home and Abroa

4、d.5References.9Chapter II Key Technologies andAlgorithms of RIS.112.1.RIS Structure Design and Adjustment.112.3.Channel Estimation and Feedback.142.3.1.Analysis on difficulties of channel estimation.172.3.2.RIS-based channel estimation and feedback algorithm.182.3.3.Future Research Direction of RIS

5、Channel Estimation and Feedback202.4.Beamforming Design.202.4.1.Common Parameter Optimization Algorithms.232.4.2.Application Scenarios.24References.26Chapter III RIS Realization and Prototype Validation.283.1.Experimental Validation 1.283.2.Experimental Validation 2.333.3.Experimental Validation 3.3

6、7References.38Chapter IV Technical Challenges and Standardization.394.1.Technical Challenges.394.2.Standardization Research Directions.40Acknowledgment.41Acronyms.423Chapter I Overview1.1.ConceptAbout every ten years,there is a new generation of cellular mobile communication systems,in which new key

7、 technologies are introduced to improve service quality.The 5th-generationmobile communication(5G)aim to provide mobile communication infrastructures in a higherstandard.It supports multiple scenarios,such as Enhanced Mobile Broadband(eMBB),MassiveMachine Type Communications(mMTC),and Ultra-reliable

8、 and Low-latency Communications(uRLLC).In the meantime,mobile communication network development is changed towardsoftware-defined development.That is,software is used to configure and optimize the network inreal time.However,there are many randomnesses and uncertainties in the wireless environment,b

9、ringing lots of uncontrollable factors to mobile communication networks.As the research on6th-generationmobilecommunication(6G)commences,6Gisconsideredtoprovidefull-coverage,full-spectrum,all-application,and strong-security services to meet peoples variousincreasing communication requirements.6G wil

10、l provide greater capacity,ultra-low latency,highreliability,high security,and full-space coverage.To explore and break through the restriction thatthere are many uncontrollable factors in the wireless environment and reshape the wirelesstransmission environment is the new direction for 6G developme

11、nt.In the last two years,the reprogrammable metasurface technology has drawn great attentionin the field of mobile communication.The reprogrammable metasurface technology was firstproposed and verified in lab in 2014 by Cui Tiejun,Academician of CAE in Southeast University1.Figure1-1showsitsbasicstr

12、ucture.Thereprogrammablemetasurfaceadoptsatwo-dimensionalthinartificialelectromagneticsurfacestructurewithreprogrammableelectromagnetic features,and it can be used for various frequency bands from microwaves tovisible light 2.The reprogrammable metasurface is composed of delicately designed andregul

13、arly arranged electromagnetic units.Such units are usually made of metal,medium,andadjustable elements.The adjustable elements in the electromagnetic units allow you to modify theelectromagneticparametersettingsrelatedtoreflectedelectromagneticwavesthroughreprogramming.Such parameters include the ph

14、ase and amplitude.This technology connects thephysical electromagnetic world of the metasuface and the digital world of information technology3,and is particularly attractive to mobile communication applications.4Figure 1-1 Reprogrammable MetasurfaceThe Reconfigurable Intelligent Surface(RIS)is comp

15、osed of and enabled by thereprogrammable metasurface,and can be used to adjust the electromagnetic signals of mobilecommunication in real time.The primary application for RIS is the RIS-based wireless relay.TheRIS allows people to adjust the wireless channel environment to significantly improve thet

16、ransmission performance between communication devices.Wireless signals from the transmittingterminal to the receiving terminal may be attenuated and scattered to a certain degree due to theabsorption of objects and natural signal diffusion in the space.This increases the computationalcomplexity and

17、reduces the performance for signal recovery on the receiving terminal.Intraditional mobile communication systems,people cannot control the wireless environment.Theonly thing that can be done is to create channel characteristic models through massive channelsounding and delicately design algorithms f

18、or the transceivers.However,with RIS in thecommunication system,flexible control of the electromagnetic transmission can proactivelyimprove the wireless transmission environment.RIS-based wireless relays can implementredirection and beamforming for electromagnetic signals in the wireless channels to

19、 utilizewireless signal energy more efficiently and improve the system performance.1.2.PotentialApplication Scenarios and RequirementsCurrently,the potential application for RIS is the relay of wireless transmission.It adjusts thephase and/or amplitude of electromagnetic waves to form desired reflec

20、ted beams,enhancingtransmission.When the RIS is operating as a relay,it improves the wireless signal transmission throughreflection/transparent transmission.The adjustments made by the electromagnetic components inthe RIS can change the transmission direction of the reflected or transmitted electrom

21、agneticwaves on the RIS to form the desired beam pattern.Specifically,the application scenarios for RISto be adopted as a relay are as follows:1)Coverage Holes FillingIn wireless cellular networks,there can be regional weak coverage areas or coverage holes.They may occupy small areas,but will hamper

22、 user experience.Besides,the mmWave frequencyband used for 5G is vulnerable to the blockage of obstacles,causing deteriorated signal quality.Moving vehicles,pedestrians,growing vegetation,and leaves can be potential obstacles.RIS can resolve all of the above coverage issues.Deploy the RIS at proper

23、locations(such as5the surface of a wall)can artificially build a line of sight(LOS)link,improving the weak coverageand wireless signal transmission robustness.See Figure 1-2.Figure 1-2 Coverage Hole Filling by RIS2)Regional Traffic IncreaseFor hotspot areas,RIS can reduce the multi-path effect for w

24、ireless channels and increase thenumber of channel subspace,to increase the traffic and expand capacity.Unlike coverage holefilling,when RIS is used for regional traffic increase,the base station must be able to implementchannel estimation for the cascading channels between the RIS and the UE.Theref

25、ore,it is morecomplicated to design the transmission scheme.3)Indoor Coverage EnhancementFor the scenarios of outdoor base stations covering indoor areas,the transparent RIS materialcan be used to design a transparent glass film that does not block light and reduces loss of wirelesssignals when they

26、 penetrate glass,to improve indoor network coverage quality.For common weakcoverage scenarios such as the stairway corner and the deep sunken area of stadium stand,deploythe RIS at proper locations such as the walls of the stairway corner or stadium.A LOS link fromthe base station to the weak covera

27、ge area can be planned based on the requirement to improve thecoverage of weak coverage and wireless signal transmission robustness.1.3.Current Research at Home andAbroadMassive low-cost electromagnetic intelligent control units of the RIS can control thereflections of wireless signals,in order to r

28、earrange the wireless transmission environment.Therefore,this technology is an emerging technology that is being widely discussed.It has greatpotential to improve the transmission rate,coverage area,and energy efficiency of mobile systems.Through adjustments of the reflection phase shift of the RIS

29、electromagnetic unit for wirelesschannels,the signals transmitted through RIS reflections and signals transmitted through otherpaths can be superposed in a co-phase way,to enhance the receive signal quality.Compared withtraditional relays,RIS-based wireless relays can implement full-duplex transmiss

30、ions withoutintroducing interference.In addition,the RIS features lightweight,low cost,and low powerconsumption,and has great potential.In 4,a single-cell wireless communication system with the RIS-assisted relay is studied,inwhich a RIS-assisted multi-antenna access point communicates with multiple

31、 single-antenna UEs.6Joint optimization is implemented for the transmitted beams from the active antenna arrays of theaccess point and the reflected beams of the RIS,to minimize the total transmit power of the accesspoint while maintaining a specified signal-to-noise ratio(SNR)of receive signals on

32、the UE.Significantly different from communications with passive signal scattering,RIS-based wirelessrelays are mainly used to enhance the current communication link performance,but not to transmitinformation of the relay itself through reflection.In the RIS-assisted communication,the signalstransmit

33、ted in the direct path and the signals that are reflected carry the same information,andtherefore the coherent signals can be superposed at the receiver to enhance the transmission andmaximize the receiver power.In 5,the RIS is used for the downlink of a multi-antenna basestation serving multiple UE

34、s,and a power optimization model is proposed for the RIS-assistedcommunication system.This model utilizes the number of RIS-based reflection units and theirphase control capabilities.Under maximum power and minimum service quality,it maximizes theenergy efficiency to optimize phase shift distributio

35、n of the RIS electromagnetic units anddownlink transmit power.In a mobile communication system with RIS-assisted relay,the channel status information(CSI)is critical for the passive beamforming of RIS.In 6,the performance of an RIS-assistedlarge-scale antenna system in different propagation scenario

36、s is evaluated by formulating a tightupper bound of the ergodic spectral efficiency.It is demonstrated that the ergodic spectralefficiency is related to the reflected phase shift distribution of RIS.In addition,considering ahardware non-ideal factor,an optimal phase shift design based on the upper b

37、ound of the ergodicspectral efficiency and statistical channel state information is proposed to maximize the ergodicspectral efficiency under the condition that the phase quantization bits of RIS are limited.In 7,the beamforming optimization is studied for the RIS-assisted mobile communication syste

38、m underdiscrete phase shift.Assume that the electromagnetic units can assist the communication betweenthe multi-antenna access point and the single-antenna UE with limited phase shift status.Asolution that is closest to the optimal and least complex is proposed based on the alternativeoptimization t

39、echnology.That is,joint optimization is implemented for the continuouslytransmitted beams of the access point and the discretely reflected beams from the RIS,tominimize the transmit power for the access point while maintaining a certain SNR on the UE.Thesimulation shows that RIS with discrete phase

40、shifts enables the same performance as that of thecontinuous phase shift when there are massive reflective electromagnetic units.In 4,5,6,and7,the reflective phase shift design is implemented for the RIS electromagnetic units under theassumption that the perfect CSI is known.This assumption helps us

41、 understand the upper bound ofthe system performance.There is also research on how the RIS can improve the security of mobile communication.In8,the adaptive adjustment of the reflective phase shift of the RIS can enhance the desired signalsand suppress the undesired signals.Joint design of the beamf

42、ormings for the transmitted signals ofthe access point and the passively reflected signals on the RIS can maximize the encryption oflegitimate communication links.In 9,how the RIS enhances physical-layer encryption of thecommunication is explored.In a Multi-Input Single-Output(MISO)broadcast system,

43、the basestation transmits independent data flows to multiple legitimate receivers,and encryption isconducted against multiple eavesdroppers.Joint optimization is implemented for the beamformerof the base station and the reflected phase shift distribution of the RIS.Under actual constraints,the minim

44、um encryption is maximized,and the path tracing algorithm based on alternative7optimization reduces the computational complex.In the preceding publications,theories have been used to study the RIS-assisted relaysmobile communication system,and multiple optimization algorithms have been proposed toim

45、prove system performance.In fact,there have been a few early studies on the systemrealizations of the RIS-based wireless relay.In 10,an RIS-based wireless relay named RFocus moves the beamforming function fromwireless terminals to the wireless environment.In typical indoor scenarios,every reflective

46、electromagnetic unit of RFocus is configured by software controllers to maximize receive signalpower of the receiver.In theoretical analysis and actual measurement,RFocus can improve signalintensity by 10.5 times and channel capacity by 2 times in average.In 11,there are experimentsproving good robu

47、stness of the RFocus in case of electromagnetic units failures.The relativeperformance improvement will not drop to 0 but will drop by 50%even though one third of theseunits fail.In 10 and 11,the feasibility of the RIS applied into actual communications systems ispractically verified.In 12,a large a

48、rray composed of 36 low-cost antenna units is deployed in an indoorhousehold environment to adjust the wireless environment,with a channel decompositionalgorithm designed to quickly estimate the wireless channel environment,and the phasedistribution of the large array is configured in real time to a

49、lign phases of multiple sub-channels.This system realizes flexible reprogrammable wireless channels.This experiment indicates thatreconfiguring wireless environment can improve the system throughput by 24%in average.Inaddition,the Shannon capacity is improved by 51.4%compared with that of the baseli

50、nesingle-antenna links,and by 12.23%to 18.95%compared with that of the baseline antenna links.In 13,an RIS-based scattering MIMO system is proposed to improve the spatial multiplexinggains by enhancing the scattering effect with the low-cost RIS.It pairs with active access points tocreate virtual pa

51、ssive access points.In the experiment,the configuration is optimized to enablevirtual access points to provide signals with the same power of real active access points andimprove the coverage range of each single access point.On the other hand,algorithms of lowcomputational complexity are designed t

52、o optimize the RIS for the scattering MIMO for each UE.The RIS-based scattering MIMO system has reduced interference,and the power requirement fordistributed MIMO systems is reduced.In the experiment,the RIS increases the averagethroughput by 2 times in an existing commercial MIMO-Wi-Fi network afte

53、r deployment in thesystem.In 14,the concept of smart space is proposed.In the space,the wireless environment isreprogrammable to have required link quality in the wireless space.Low-cost devices areembedded into the wall of a building to improve the wireless link quality by passively or activelyrefl

54、ecting wireless signals.The experiment has proved that it is feasible to use passive elements tochange the wireless channels.The 22 MIMO channel matrix condition number is increased by1.5 dB,and the signal intensity is improved by 26 dB.The RIS,a new technology that can flexiblycontrol the incoming

55、electromagnetic waves,can be deployed on the surface of the scatterer of alarge space to intelligently control the wireless environment,enhance the coverage of5G-advanced networks,and implement transformation of the communication paradigm.The RISreflection mode is an important application direction

56、for the RIS,and many institutions in Chinahave carried out research on the RIS in the form of relays.8Southeast University has built an measurement system for the RIS free space path loss(FSPL)to verify the theoretical model of the RIS FSPL.The system can measure the FSPL of multipletypes of RISs to

57、 verify the theoretical formula for the RIS FSPL.In addition,a series of live network tests have been carried out,producing lots of inspiringresults.China Mobile,together with the Cui Tiejun academician team of Southeast University andHangzhou Qiantang Information Co.,Ltd.,has completed the technica

58、l verification of the RIS onlive networks in Nanjing 15.The preliminary test result shows that the RIS can flexibly adjustthe signal beams in the wireless environment based on the UE distribution,to significantlyimprove the signal intensity,network capacity,and user rate in weak coverage areas of th

59、e livenetwork.In the outdoor test scenarios,the cell edge coverage is improved by 3 to 4 dB in average,and the cell edge user throughput is improved by about over 10 times.In the test scenarios thatoutdoor base stations cover indoor areas,the indoor coverage is improved by about 10 dB,and theuser th

60、roughput is improved by about 2 times.ZTE,together with China Unicom,has completed the technical verification test of the RISreflective panel in 5G intermediate frequency(IF)network outfield in Shanghai 16.The testresult shows that at the non-line-of-sight(NLOS)cell edge of the 5G IF base stations,t

61、hereference signal received power(RSRP)of 5G UEs can reach 10 dB,and the performance of 5Gcell edge users can be improved by over 40%.Therefore,the RIS reflection technology will be ascientific and feasible innovative approach for the in-depth coverage of 5G IF base stationnetworks.ZTE,together with

62、 China Telecom,has completed the far-distance technical verification testof the RIS reflective panel in 5G high frequency network outfield in Shanghai 17.The test resultshows that the reference signal receive intensity of 5G UEs is improved by 12.5 dB,and the 5Ghigh frequency weak coverage UE perfor

63、mance in improved by 296%for NLOS coverage holesor week coverage over 150 meters away from 5G high frequency(26 GHz frequency band)basestations.The RIS reflection technology will be a scientific and feasible innovative approach forthe in-depth coverage of 5G high frequency base station networks.In a

64、ddition,ZTE,together with China Mobile Beijing Branch,has published the RIScascading technology prototype verification results for the commercial networks in 2.6 GHz.Two-level cascaded RIS panels are used to reflect 5G signals in controlled circumstance,toactively explore the feasibility of RIS in 5

65、G networks 18.The early test results preliminarily show the feasibility of the RIS,but there are fourchallenges for its standardization and actual engineering application.Challenge one:The basictheory is not complete.The reflection and transparent transmission characteristics of the RIS are tobe det

66、ermined,the channel transmission model is incomplete,and there is no modeling for actualtransmission environment.Challenge two:The key technology needs to be created.There arethings to be studied and standardized,such as co-channel and adjacent-frequency interferencecharacteristics,interference coor

67、dination among operators,and beamforming and channelestimation algorithms.Challenge three:The component maturity and reliability are not sufficient.The current RIS in the industry is only a prototype,with insufficient adjustable angle andcomponent adjustment speed,and there are massive units for qui

68、ckly locating and identifying unitfailures.Challenge four:The deployment scenarios are limited.The RIS has a large size and alarge surface,and active and wired control will reduce its application scenarios.Further optimized9engineering design can improve its deployment flexibility.The intelligent re

69、flector is of relatively high technological and industrial maturity,andthree-phase development can be adopted for the intelligent reflector.Phase one:To realize thepassive static reflection surface that can be used for quick deployment and satisfies the needs forexpanding network coverage and filing

70、 coverage holes;Phase two:To realize the semi-staticcontrollable reflector.Adjustment and control on component units implements beam selection toexpand the coverage of the metasurface beams and improve the cell capacity and rate;Phase three:To realize the dynamic intelligent reflector.Coding algorit

71、hms dynamically track UE location andmatch the channel environment,to intelligently adjust and control the transmissions ofelectromagnetic waves.References1.C T J,Q M Q,W X,et al.Coding metamaterials,digital metamaterials and programmablemetamaterialsJ.Light Science&Applications,2014(3):e218.2.Zhang

72、 Lei,Liu Shuo,and Cui Tiejun,Theory and application of coding metamaterials J,Chinese Optics,2017,10(1):1-12.3.C T J,L S,Z L.Information metamaterials and metasurfacesJ.Journal of MaterialsChemistry C,2017(5):3644-3668.4.W Q,Z R.Itelligent reflecting surface enhanced wireless network via joint activ

73、e and passivebeamformingJ.IEEE Transactions on Wireless Communications,2019,18(11):5394-5409.5.H C,Z A,A G C,et al.Reconfigurable intelligent surfaces for energy efficiency in wirelesscommunicationJ.IEEE Transactions on Wireless Communications,2019,18(8):4157-4170.6.H Y,T W K,J S,et al.Large intelli

74、gent surface assed wireless communication exploitingstatistical CSIJ.IEEE Transactions on Vehicular Technology,2019,68(8):8238-8242.7.W Q,Z R.Beamforming optimization for intelligent reflecting surface with discrete phaseshiftsC/2019 IEEE International Conference on Acoustics,Speech and Signal Proce

75、ssing(ICASSP).Brighton,U.K.,2019:7830-7833.8.8C M,Z G,Z R.Secure wireless communication via intelligent reflecting surfaceJ.IEEEWireless communications letters,2019,8(5):1410-1414.9.9C J,L Y C,P Y,et al.Intelligent reflecting surface:A programmable wireless environmentfor physical layer securityJ.IE

76、EE Access,2019(7):82599-82612.10.V A,H B.RFocus:Practical beamforming for small devicesJ.arXiv preprint arXiv:1905.05130,2019.11.V A,H B.RFocus:Beamforming using thousands of passive antennasC/NSDI 20.SantaClara,CA,USA,2020.12.L Z Q,X Y X,S L F.Towards programming the radio environment with large ar

77、rays ofinexpensive antennasC/NSDI 19.Boston,MA,USA,2019.13.MD,ZC,DS,etal.ScatterMIMO:EnablingvirtualMIMOwithsmartsurfacesC/MobiCom 20.London,U.K.,2020.14.A W,S L F,J G,et al.Programmable radio environments for smart spacesC/HotNets-XVI.Palo Alto,CA,USA,2017.15.Interdisciplinary innovation.China Mobi

78、le,together with the Cui Tiejun academician team,hascompletedthefirsttechnicalexperimentoftheRISE.https:/ Mobile,together with China Telecom,has completed the first technical verification ofthe RIS under the 5G IF network outfield1017.E.https:/ with China Telecom,has completed the first technical ve

79、rification of the RISunder the 5G high-frequency outfield E.https:/ 5G pilot city innovation summit,discussing 5G+smart city construction hot topicsE.http:/ II Key Technologies andAlgorithms of RIS2.1.RIS Structure Design andAdjustmentThe RIS is a two-dimensional electromagnetic meta-material based

80、on theories such as thetraditional periodic structure electromagnetic theory,Huygens-Fresnel principle,and generalizedreflection/refraction principles.With such theories and principles,a specially designed thin PCBstructure can be used to have full control of the amplitude,phase,polarization,spread,

81、andmomentum of electromagnetic waves.The electromagnetic metasurface can have all thecharacteristics of electromagnetic meta-materials.In addition,it features multiple functions,simple structure,easy-to-integrate,flexible manufacturing,and low cost,and therefore is widelyapplied in the electromagnet

82、ic wave control field.With the same metasurface,the adjustable material or component can be controlled toimplement multiple functions,greatly expanding the application scenarios of the metasurface andreducing the application cost of the metasurface.In order to enable the metasurface to beapplicable

83、to more scenarios(such as applications in information transmission),high-speedcontrol circuits can be introduced to the adjustable metasurface for quickly switching functions ofthe metasurface.In the meantime,artificial atoms or artificial units with powerful functions can beintroduced to the metasu

84、rface to implement specific functions,enabling evolution from adjustablemetasurface to intelligent metasurface.Thus,features that can be realized for metasurface aredetermined by:Coding strategy(unit control strategy);Unit structural characteristics(sizeand shape);Characteristics of adjustable mater

85、ials or components;Unit layout in thespace;Control circuit characteristics.The metasurface can be used to construct many new functions that are primarily done by thecoding strategy,such as full adjustment and control of the amplitude,phase,polarization,frequency spectrum,and momentum of electromagne

86、tic waves.The unit structure design ofmetasurface can be relatively fixed,to cover typical scenarios with only several types of unitstructures of metasurface.Peripheral control circuits control adjustable materials or adjustablecomponents to implement the coding strategy.However,introducing active c

87、omponents andattached control circuits on the metasurface will cause certain impact on the design,usage andoriginal characteristics of the metasurface.Therefore,joint optimization and design should beimplemented on the peripheral control circuit,and active components or adjustable materialsduring ne

88、w metasurface design.The following sections will analyze the basic performance of theadjustable components and adjustable metasurface,in order to analyze their application scenariosin wireless communications systems.The excellent performance of the passive metasurface is determined by its unit struc

89、ture andlayout(periodic or non-periodic structure).However,once the metasurface is made,its functionsare fixed.The non-adjustable characteristics limit its application in the wireless communicationssystem.Introducing switching diodes or other adjustable materials to the passive metasurface unitsand

90、using the different operating statuses of the adjustable materials or components to control themetasurface unit resistance can help control the electromagnetic wave characteristics,thusdevelopingadjustablemetasurfaceandintelligentmetasurface.Therefore,theoperatingperformance of these adjustable mate

91、rials or components will have vital impact on functions that12can be realized for the metasurface.Adjustable materials or components used on the metasurface have certain characteristics.Forexample,such substances can have drastic or rapid physical property changes under externalenvironmental stimula

92、tion.When there are changes in the external electronic field,magnetic field,temperature,humidity,pressure,or light intensity,the properties of such materials will changeaccordingly,thus causing changes to the functions of the metasurface and realizing differentmodulation mechanisms.Examples:The nema

93、tic liquid crystals and graphene in the adjustableelectrical structure,the ferrite magnetic bar or ferrite magnetic plate in the adjustable magneticstructure,the Si,GaAs,or semiconductor optoelectronic materials in the adjustable opticalstructure,and phase shift material,phase shift materials(PCM)su

94、ch as VO2 in the adjustablethermal structure can be used.Since the current wireless communications system is based on theelectrical control mechanism,the adjustable electrical structure and the corresponding metasurfaceshould be preferentially considered.Adjustableelectricalmaterialsoradjustableelec

95、tricalcomponentshaveexcellentperformance and can be flexibly controlled,and therefore are widely used in metasurface unit andmetasurface design.Future wireless communications systems use the metasurface to realize morenew functions,such as to flexibly control electromagnetic environment,change chann

96、el quality,suppressinter-UEinterference,convergingtransmissionenergy,andsimplifyingsystemarchitecture.In the meantime,the future wireless communications system will possibly work ondifferent frequency bands and multiple frequency bands,or on ultra-large bandwidth.Therefore,select the proper metasurf

97、ace to satisfy different needs.2.2.Channel Measurement and ModelingThe channel characteristics of the RIS will determine the performance limit and optimizationsolution of the RIS communications system.Therefore,it is important to measure the channels ofthe RIS and build an easy-to-use and sufficient

98、ly accurate channel model for the RIS.The RIS wireless communication system has different path losses in different scenarioapplications.Common expressions for the FSPL of the RIS-assisted path are not simple enough.In the RIS-assisted far-field beamforming scenarios,the FSPL of the reflection-assist

99、ed pathis proportional to(d1*d2)2,where d1 and d2 are the distance between the base station and theRIS and the that between the base station and the UE,respectively.In the meantime,thebeamforming gain is proportional to the square of the number of electromagnetic units.In theRIS-assisted near-field

100、signal broadcast scenario,the overall FSPL of the reflection-assisted pathis proportional to(d1+d2)2.That is,the FSPL is proportional to the square of the sum of the twotransmission paths,and there is no beamforming gain in this scenario.The RIS FPSL model needsto be further simplified and optimized

101、 in the future.Currently,large-scale attenuation characteristics(path loss and shadow attenuation),fastattenuation characteristics(multi-path effect,delay expansion and angle expansion,Dopplerspectrum,and relativity),spatial consistency,and so on are considered in wireless channelmeasurement and mod

102、eling.The commonly used channel models include the deterministic model,random model,or mixed channel model.The deterministic models include the geometric opticalmodel and ray tracing model.The scenario-map-based model and the point cloud model arecommon simplified ray tracing models.In recent years,

103、the random channel model that can be13classified into GBCM and CBCM has been widely used in academia and industry.GBCM and thedigital-map-based hybrid channel model are the channel models that are mainly adopted in thecurrent 3GPP and ITU standard research.The following must be further considered fo

104、r the RISchannel measurement and modeling based on the current wireless channel models:(1)Physical model abstraction of RISTo implement model abstraction for different RIS physical implementation solutions,thefollowing needs to be considered:RIS unit modelPolarization model(single/dual polarization,

105、polarization leakage/twist,and anisotropy)Amplitude/phase adjustment and control modelInsertion loss modelNon-ideal factors RIS panel modelBS-RIS far-field plane wave feeding excitation modelBS-RIS near-field spherical wave feeding excitation modelRIS-UE near-field modelRIS codebook model(2)RIS chan

106、nel cascadingSelection and optimization of the RIS channel cascading linkFor the merge of absolute-latency-based RIS cascading links,considering the balancebetween simplicity and usability and between complications and accuracy is the key for creatingan effective channel model.In recent preliminary

107、researches,people are attempting to explain the path loss model of theRIS.In 1-2,the physical and electromagnetic characteristics of the RIS are considered,andFSPL models of RIS-assisted wireless communications are created for different scenarios.Themeasurement in the lab has further proved the prop

108、osed path loss mode.The measurement resultsare well matched with the modeling results.In 3,the general scalar diffraction theory andHuygens-Fresnel principle are used to propose the closed expression for calculating the RISreceive power and to determine the conditions for the RIS to serve as the ano

109、malous specularreflection.In 4,a physical and electromagnetic compatible communications model is introducedfortheRIS-assistedwirelesssystem.Themodelfeaturesend-to-end,electromagneticcompatibility,mutual-coupling sensing,and unit amplitude and phase coupling.This model iscompatible with the tradition

110、al communications theory frame.For the multi-path attenuation(small-scale attenuation),the typical method for modelingestimation is to use the well-known distribution(such as Rayleigh attenuation and Ricianattenuation).In 5,a controllable smart electromagnetic surface reflection model based on raytr

111、acing is proposed in addition to the map-based hybrid channel model,to create a channelmodeling method for the smart electromagnetic surface to be deployed in complex scenarios.Inthis method,multiple virtual logic base stations are assumed to exist on the electromagneticsurface,in order to simplify

112、the model algorithm and reduce calculations while maintaining thehigh accuracy.142.3.Channel Estimation and FeedbackThe RIS,one of the candidate key technologies for 6G,can be applied in the electromagneticactive wireless environment of physical surfaces.The electromagnetic characteristics of the ar

113、rayson the metasurface are altered through intelligent methods to form an electric field withcontrollable polarization,amplitude,and phase,where the energy is concentrated in a designatedthree-dimensional space for transmission and reception.This increases the energy efficiency andreduces interferen

114、ce,to implement electromagnetic environment sensing,communications,andcontrol in a completely new way.The RIS can control the scattering,reflection,and refraction ofradio waves,to neutralize the negative effect of multi-path fading.Without complicated codingand decoding processes and RF processing,d

115、irectional reflection can be implemented for theincoming electromagnetic waves.In traditional communications systems,the channel conditions are completely determined bythe environment,and optimization can only be carried out on the Tx/Rx nodes.However,the RISenables controllable channel conditions,a

116、nd therefore joint optimization can be implemented onthe Tx/Rx nodes and channels,to flexibly control the transmission of electromagnetic waves.TheRIS is of positive significance for solving NLOS-related problems,increasing coverage range,expandingtransmissionfreedom,reducingelectromagneticpollution

117、,andenablingultra-large-scale terminal access,environment sensing,and positioning,and it provides certainsupport for the unification of future communications.It is necessary to obtain accurate CSI toassist in realizing the preceding functions.This section analyzes the related information ofscenario

118、requirements,algorithm design,and engineering implementation of the RIS channelestimation and feedback.The RIS can operate on different frequency bands.The channel characteristics vary underdifferent frequencies.Generally,the low-frequency bands have strong penetration but severemulti-path effect,th

119、e mmWave channels feature sparsity and have a significantly shortertransmission distance than that of the low-frequency bands,and terahertz channels featureabundant scattering paths.For terahertz channels,reflections are no longer ample,and the energyis prone to molecular absorption.In channel estim

120、ation algorithm designing,the channelcharacteristics under different frequency bands,such as sparsity,need to be considered and fullyutilized,and the channel estimation method needs to be designed according to the specificfrequency bands.The RIS is greatly different from the traditional MIMO,relays,

121、and other technologies,especially in channel presentation and estimation.The antenna spacing generally equals a halfwavelength for the traditional MIMO,which is usually used for far-field situations.However,theantenna spacing can be less than a half wavelength for the RIS.When there are massive RISe

122、lements and the distance between the base station and the RIS is small,the distances from thebase station to different metasurface arrays do not need to be the same.In this situation,thenear-field model of the channel needs to be analyzed.Traditional MIMOs acting as signaltransmitters or receivers d

123、o not affect the electromagnetic wave propagation environment,whilethe RIS may change the electromagnetic field of the propagation environment,bringing anonlinear impact on the electromagnetic wave propagation.In relay transmission,the segmental15channel CSI must be obtained,while the RIS imposes di

124、fferent requirements for the channel CSIin different transmission scenarios,and the cascading channel CSI obtained is sufficient to supportcommunication in some transmission scenarios.The RIS channel estimation and feedback also has many challenges.The RIS features passivereflection,and does not nee

125、d massive RF links to be configured,but embraces a large number ofarrays,making it difficult to obtain segmental channels.Compared with MIMOs and relaysconfigured with strong signal processing capabilities,general RISes are only equipped with simpleonboard signal processing units.For effective chann

126、el estimation,new algorithms and protocolsmust be designed to simplify the RIS channel estimation to the greatest extent and avoid complexonboard signal processing operations.Reasonable channel models are essential for the analysisand research of RIS communication.The current dual-polarization backs

127、cattering channel model,spatial scattering channel model,large-scale path loss analysis,and other studies are still in need,while the near-/far-field channels of the RIS have different propagation properties.Theintroduction of the RIS affects the electromagnetic field,posing new challenges to thecha

128、racterization and simplification of RIS channels.For typical application scenarios of the RIS,the BS and the RIS remain fixed in their places.Channels between these two are high-dimensional,varying slowly and quasistatic,while thosebetween UEs with the BS and RIS are time-varying and low-dimensional

129、 because users aremoving.Channels in these two cases are estimated separately.Although the mobile channels forUEs still need to be frequently estimated,those with large dimensions,between the BS and theRIS,no longer need to be estimated frequently,so the average pilot overhead is reduced.To estimate

130、 the channels between the BS and the RIS under the limitation that the RIS cannotreceive or transmit signals,round-trip pilot Rx/Tx methods were mentioned in some references.Insuch cases,the BS must support the duplex,transmit downlink pilots to the RIS,and receiveuplink pilots reflected.BS-RIS chan

131、nels can be estimated based on the received and transmittedpilots.After BS-RIS channels are estimated,the only thing that needs to do is to transmit pilotsthrough UEs within each hourly scale,so as to estimate the channels from the BS to UEs andthose from the RIS to UEs.This helps to complete the RI

132、S-UE channel estimation with arelatively small pilot overhead.In high-frequency scenarios,transmission signals attenuate more severely,the path signalsafter multiple reflections are quite weak and can generally be ignored,and only a small number ofpaths can reach the receiver,so the high-frequency c

133、hannels are generally considered to be sparse.Cascading channels are estimated by building a combined sparse matrix and designing matrixcompletion issues considering that the RIS is programmable and that the channels arerank-deficient.In this solution,the RIS is fully passive,and the channel estimat

134、ion is based on thecombination of bilinear sparse matrix decomposition and matrix completion,so that cascadingchannels can be estimated accurately.However,this method featuring high computationalcomplexity can only obtain cascading channels(but not decompose segmental channels),so itdoes not apply t

135、o the designs and scenarios where the segmental channel CSI must be known.For high-frequency communication scenarios such as millimeter wave and terahertz,the pilotoverhead is reduced based on the sparsity of multi-UE channels in the angle domain.On the onehand,as all UEs communicate with the BS thr

136、ough the same RIS,all non-zero elements of themulti-UE angle-domain cascading channel are in the same row.On the other hand,since all UEs16share some scatterers between the RIS and UEs,some non-zero elements of the multi-UEangle-domain cascading channel are in the same column.This can be called stru

137、ctured sparsity.Based on this nature,a support set detection algorithm based on dual-structured sparsity wasintroduced in 6.The multi-UE joint estimation is employed to get the common row where thenon-zero elements of the angle-domain cascading channels are located and the common columnwhere the non

138、-zero elements of the angle-domain cascading channel are located.It is possible toestimate each UEs non-common columns where the non-zero elements of the UEs angle-domaincascading channels are located.Because the multi-UE joint estimation employed can suppressnoise to a certain extent,the estimation

139、 accuracy is improved under the same low pilot overhead.In order to make full use of information about collected data or to solve the channelestimation issues due to unknown channel models,a potential idea is proposed to use new AImethods for channel estimation.Among them,the data-driven channel est

140、imation has lessstringent requirements for model accuracy and does not even need to know the transmission model.In recent years,the rapidly developing AI provides new processing paradigms for traditionalwireless communication,bringing new solutions to the channel estimation of the RIS wirelesscommun

141、ication.This type of method creates databases through offline training,and,during theonline training stage,just needs to obtain signals received by pilots to output the channelestimation results,with high robustness and anti-noise capability.However,existing AI-basedsolutions generally use data info

142、rmation directly for channel estimation,but have seldom divedinto data+model dual-driven strategies that may simplify the training process and achieve bettertraining results.The channel estimation and beam matching collaboration solutions may be designed to adaptto more practical application scenari

143、os.The codebook-based channel estimation solutions may beused for the RIS channel estimation.The characteristics of the RIS segmental transmissionchallenge the traditional codebook-based solutions.It is urgent to find out how to design thecodebooks of BS-RIS and RIS-UE channels and match them well.I

144、n addition,codebook-basedchannel estimation and transmission solutions face more issues in algorithm and protocol design,such as random access design,beam selection method,etc.In RIS-assisted communication systems,the CSI acquisition is the prerequisite for jointprecoding between the BS and the RIS.

145、However,since the RIS is all composed of passive andnear-passive units,it cannot actively receive/transmit or process pilot signals,posing a seriouschallenge to the channel estimation.In addition,the massive RIS units lead to a huge pilotoverhead required for channel estimation.A low-overhead channe

146、l estimation based oncompressed sensing was introduced in 6 according to a hybrid active/passive RIS structure,andthe estimation accuracy was improved using the Denoising Convolutional Neural Network(DnCNN).Specifically,the hybrid RIS structure means that only a few RIS units are active andcan recei

147、ve/transmit or process signals and most RIS units are still passive,so as to ensure lowcosts and low power consumption.In the channel estimation stage,a few RIS units can receivepilot signals,and thanks to the channel sparsity in the angle time delay domain,the compressedsensing algorithm may be emp

148、loyed to recover high-dimensional channels from a small number ofreceived pilots.In order to further improve the estimation accuracy,6 denoised the channelsestimated by the compressed sensing algorithm in the DnCNN,thus realizing the accurate channelestimation with a low pilot overhead.However,the i

149、ntroduction of active units increased the RIS17cost and power consumption.In order to achieve the channel estimation under passive RISes,mosttraditional algorithms directly estimate cascading channels from the BS to UEs via the RIS,so thatthe RIS channel issues can be converted into channel estimati

150、on in a large-scale MIMO system.The AI-based channel estimation in traditional MIMO systems can also be transplanted to the RISchannel estimation.However,it may be difficult to model accurately RIS channels for different applicationscenarios,and only measured channel data can be collected.In such ca

151、ses,the traditional channelestimation cannot work as expected.To this end,we can generate deep learning technologies toestimate channels.For example,based on the measured channel data,we can first use generativeadversarial networks(GANs)to learn the channel distribution(including the channels from t

152、he BSto the RIS and those from the RIS to UEs),and then in the channel estimation stage,by optimizingthe GAN inputs,output the estimate of the current channel to maximize the correlation betweenthe received pilot signals and the channel estimation.More efforts are needed to explore channelestimation

153、 based on generation models in future studies,because it is essential for performancecomparison and unified evaluation of different algorithms.The performance indicators currentlyconsidered include channel estimation accuracy,BS/RIS/terminal computational complexity,timedelay,RS,signaling overhead,e

154、tc.2.3.1.Analysis on difficulties of channel estimationTo design a RIS-assisted communication system,obtaining CSI in the RIS wirelesscommunication system is an essential issue to be tackled.In wireless communication systemscollaborated by traditional active devices(such as repeaters),the CSI can be

155、 estimated based onthe training sequence sent by active devices.However,in the RIS collaborative-basedcommunication systems,the passive device RIS composed of a large number of passive reflectionunits does not have the capability of active Tx/Rx and signal processing.Therefore,the CSIcontaining mass

156、ive unknown parameters is to be estimated in the BS or UEs when they are in theactive state,which without doubt brings tremendous difficulties and challenges to channelestimation.Most of the previous efforts assumed that all individual channels between the RIS with theBS and UEs had ideal CSI,and su

157、ch CSI was known to the BS and RIS.The RIS-based channelestimation and feedback have the following challenges:The RIS can only passively reflect signals and has no signal transmission/processingcapability,RF links,or sensing and signal amplification capability.In fact,it is difficultto estimate its

158、channels with the BS and UEs.The RIS is usually composed of a large number of reflecting elements.The traditionalone-off channel estimation,puting the cascading channels of all RIS reflectingelements at one time,requires a long pilot length,and such length increases as thereflecting elements augment

159、,resulting in excessive pilot overheads.In addition,due tomassive RIS units,high-dimensional channel matrix,complex channel estimation,andhigh feedback overhead,Channel interoperability may not necessarily be established,making it difficult toestimate the channel uplink and downlink.18The system per

160、formance is sensitive to the accuracy of channel estimation,i.e.the BSand RIS beamforming are also affected.2.3.2.RIS-based channel estimation and feedback algorithm1)Bistatic reflection-based channel estimation and feedbackFirst of all,the entire channel estimation time is divided into several subs

161、tages 7.In the firstsubstage,all reflection units in the RIS are closed,and the BS only needs to estimate the directchannel from the BS to UEs.Therefore,the wireless communication system based on RIScollaboration can be simplified into a traditional wireless communication system without RIS.Thedirec

162、t channel can be estimated through classic solutions,such as the least square(LS)orMinimum Mean Square Error(MMSE).In the following sub-stages,each individual reflectionunit of the RIS is opened in turn and the remaining are kept closed.The BS estimates thecascading channels from the BS to reflectio

163、n units and those from reflection units to UEs.Finally,the channel estimation of the entire system is completed based on the estimation results of allsubstages through the LS/MMSE.2)MMSE-based unbiased channel estimation and feedbackAll reflection units of the RIS remain active throughout the traini

164、ng stage.First of all,theentire cascading channel estimation time is also divided into several substages.In each substage,the optimal phase shift matrix of the RIS is a discrete Fourier transform(DFT)matrix 8.Therefore,the cascading channel estimation is based on all pilot signals received in all su

165、bstages.However,both of the above methods result in significant pilot overhead.Because theunknown channel coefficient in the wireless communication system based on RIS coordination ismuch larger than that of the conventional wireless communication system,the huge pilot overheaddefinitely impairs the

166、 system performance.Therefore,in the wireless communication systembased on RIS collaboration,the channel estimation is quite necessary to reduce the pilot overhead.3)Multi-time scale-based channel estimation and feedbackAfter being deployed,the RIS remains static relative to the BS.It can be conside

167、red that theRIS-BS channel is quasistatic,with a long coherence time.Due to user mobility,UE-BS andUE-RIS channels are dynamic and fast-changing,with a short coherence time.Therefore,thechannel estimation with multi-dimensional time scales can be set for different channels.BS-RIS channels are quasis

168、tatic,but not highly time-varying,with a relatively long channelestimation period.UE-BS and UE-RIS channels are dynamic,with a shorter estimation period.This solution can effectively reduce the pilot overhead,and for these two types of time scales,different settings may be configured according to di

169、fferent RIS deployment scenarios to adaptmore flexibly to complicated and ever-changing environments.In 9,the authors introduced a two-timescale channel estimation framework.The key idea ofthe framework was based on the high-dimensional and quasistatic nature of BS-RIS channels andthe mobile and low

170、-dimensional nature of RIS-UE channels.In order to estimate the quasistatictransmitter-RIS channel,they proposed a dual-link pilot transmission solution,i.e.the transmittersends downlink pilots and receives uplink pilots reflected from the RIS.In addition,they also19introduced a BS-RIS channel recov

171、ery algorithm based on coordinate transformation.Thenumerical results showed that the proposed two-timescale channel estimation framework couldachieve accurate channel estimation with lower pilot overhead.4)Compressed sensing-based channel estimation and feedbackThe BS and RIS are generally installe

172、d at the same level.Under HF communicationconditions,with limited scatterers around the BS and RIS,the angle-domain channel appears to besparse to some degree.In order to reduce the pilot overhead of the RIS channel estimation,theDFT dictionary helps to convert the space-domain channel to the angle

173、domain and model thechannel estimation issues into a sparse signal recovery issue,so that the compressed sensingalgorithm can be employed to reduce the pilot overhead.The compressed sensing effectivelyreduces the pilot overhead,whether it is single-structured sparse for segmental channels ordual-str

174、uctured sparse for cascading channels.Reference 10 proposed a method of channel estimation and joint beamforming design forRIS-assisted mmWave systems.The inherent sparsity of the mmWave channel was considered toreduce the training overhead.First of all,the sparse representation of cascading channel

175、s wasgiven,and then a channel estimation method based on compressed sensing was proposed to workout the joint beamforming design according to the estimated channels.5)AI-based channel estimation and feedbackCompared with traditional channel estimation algorithms,AI-based channel estimation canobtain

176、 highly accurate CSI with just a small number of dynamic RIS units.Moreover,AI does notneed any RIS geometric knowledge,nor does it take into account the channel sparsity,so thesolution is more universal.By creating databases through offline training,and obtaining a smallnumber of pilot signals in t

177、he online stage,the channel estimation can output results with theanti-noise capability superior to those of traditional channel estimation algorithms.In order toachieve these gains,AI needs to collect sufficient data sets.The current AI-based channelestimation and feedback mainly rely on strong dat

178、a drivers to obtain reliable gains,while thestudies on dual-driven(data-driven and model-driven)models that can simplify the trainingprocess are quite limited.6)Block training-based channel estimation and feedbackThe channel estimation for block training is also named the channel estimation for laye

179、redtraining.It divides RIS elements into multiple groups and estimates the cascading channels of allelements in each group.This solution greatly reduces the pilot overhead because the number ofpilot symbols required is only related to the number of groups rather than all RIS elements.However,this ma

180、y undermine the beamforming accuracy as the reflection coefficients are thesame for all RIS elements in each group,resulting in lower freedom of the RIS beamforming.In order to compensate for the beamforming accuracy loss,11 proposed to break down eachgroup of RIS elements further into smaller subgr

181、oups.Reflection coefficient vectors of all RISunits over a given pilot symbol period are decomposed into Kronecker products of two vectors,called base reflection vector and intragroup reflection vector respectively.Intragroup reflectionvectors are common to all groups;while base reflection vectors a

182、re designed to effectivelyestimate the valid channels of each group.This method takes into account the beamforming20accuracy and the complexity of channel estimation as well.7)Location-aided channel estimation and feedbackGenerally,once deployed,the RIS remains unchanged,so the RIS location informat

183、ion canbe applied in the RIS-BS channel estimation.The relatively fixed location between the RIS andthe BS helps to obtain information such as the angle of arrival(AOA)through low-complexitychannel estimation algorithms.In the cascading channel estimation,the RIS location informationis regarded as t

184、he known information for channel estimation,to obtain additional gains.8)Matrix theory-based channel estimation and feedbackFirst,the sparse matrix is employed to export the channel matrices between the BS and theRIS and between the RIS and UEs based on the received signals.Second,the on/off state m

185、atrixcontaining all RIS reflection units helps to clear ambiguity matrices decomposed from the abovematrices Third,the whole RIS reflection unit undergoes the channel estimation finally byrecovering missing items based on the matrix characteristics.2.3.3.Future Research Direction of RIS Channel Esti

186、mation and Feedback1)Channel estimation of multi-cell communication systemThe current RIS-based channel estimation is mostly based on the channel estimation of singleand multiple UEs in single-cell scenarios,while there are few studies on the single-and multi-UEchannel estimation in multi-cell scena

187、rios.Considering the inter-cell interference exists amongmultiple cells,it is a big challenge to overcome such interference.Existing models seldomconsider the multi-RIS cases.The RIS has the advantage that it may have the wirelessenvironment under control after large-scale deployment.The future RIS

188、deployment will bedefinitely diversified and multi-dimensional,so the channel estimation in multi-RIS scenarios willbe a top topic for future research.2)Channel estimation in highly dynamic scenariosRISes are usually deployed in fixed locations such as buildings and street lights.For somespecial app

189、lications,RISes must be mounted on dynamic locations,such as unmanned aerialvehicles(UAVs),high-speed railways(HSRs),etc.In these dynamic scenarios,RIS-BS channelsbecome dynamic,while RIS-US channels are quasistatic.As the RIS-BS channel matrix ishigh-dimensional and dynamic,the channel estimation b

190、ecomes more difficult,and the challengegets bigger.It should be explored that whether the existing sparsity-based compressed sensingalgorithms and block estimation algorithms are applicable to dynamic channels.Future researcheswill also focus on solving the RIS-based channel estimation in dynamic sc

191、enarios.2.4.Beamforming DesignThe RIS can adjust the beamforming to transmit signals toward a specific direction,so as toincrease the power of the required signals and weaken the interference simultaneously.Analysisresults from numerous references showed that by increasing the number of RIS reflecti

192、ngelements,the interruption and capacity of the communication system could be significantly21improved.The current massive beamforming design efforts are based on the instantaneous CSI ofperfect/imperfect cascading channels(i.e.,channels between the BS-each RIS reflecting element-the BS).In order to

193、obtain these instantaneous CSI,a series of estimation methods have beenproposed in the academic community.However,in the case of an extremely large number of RISreflecting elements,the instantaneous CSI estimation overhead of cascading channels istremendous,seriously reducing the frequency spectrum

194、efficiency of the system.Considering thatthe statistical CSI changes much more slowly than the instantaneous CSI,the RIS beamformingdesign based on statistical CSI is beneficial to improving the frequency spectrum efficiency of thesystem.It can be seen from the components of the RIS system that the

195、RF processing units andintelligent controller units at the BS and receiver are crucial for intelligent communication andtransmission.In particular,the deployment of large-scale antenna arrays and the use of cheap RFunits will greatly reduce the system overhead.However,it is impractical that the hard

196、ware systemis always working in a perfect and stable state due to the environmental noise,I/Q imbalance,phase noise(PN),amplifier nonlinearity,working point drifting of relevant core components(e.g.low-precision A/D converters),etc.in actual physical communication scenarios.Although somemeasures suc

197、h as transmitter signal correction and receiver signal compensation can relieve theimpact of hardware impairments on the system,the impact of residual hardware impairments orhardware limitations on the system cannot be ignored,otherwise,the beamforming algorithm inthe ideal hardware state would seri

198、ously distort the training pilots and the signals expected toreceive.The RIS beamforming design with hardware limitations is an essential research direction inthis field because it can achieve communication transmission stability and performanceoptimization from a practical perspective.Affected by a

199、ctual factors such as transmitter PN andchannel estimation errors,reflector PN,receiver distorting noise,etc.,traditional methods forbeamforming optimization and design of RIS systems are not applicable because these actualfactors will also undermine the performance of traditional algorithms.It is q

200、uite necessary toconsider the impact of hardware impairment in advance from the perspective of system modeling,but taking these factors into consideration will exacerbate the complexity and difficulty inoptimizing the original parameter coupling.Therefore,the key to improving the RIS systemperforman

201、ce is to design a joint beamforming and phase optimization algorithm that is lesscomplex,reliable,and analytic.The RIS configured with active electromagnetic units has certain signal detection capabilities.Due to the limited digital processing capabilities of the RIS,conventional pilot training meth

202、odscannot be directly employed to estimate RIS channels.In actual application systems,generally,thetransmitting terminal can only obtain some imperfect CSI.To solve this issue,robust active Tx/Rxbeams and RIS passive beams must be jointly designed to improve the system performance underimperfect CSI

203、 conditions.As the demands of wireless communication users increase,signal transmission requireshigher bandwidth,and the frequency band required for wireless communication needs to beincreased to obtain sufficient bandwidth resources.However,when the frequency band increases,the corresponding path l

204、oss increases,and then a larger antenna array is needed to provide higherbeamforming gains.Since each antenna in the array has to be connected to a phase shifter,the22power consumption and cost of the phase shifter will get unaffordable as the array size increases.Considering that the RIS features l

205、ow cost,low power consumption,and intelligent phase control,it is possible to use the RIS auxiliary base stations for joint beamforming.Because most of the existing RISes can only achieve low-precision phase shift,there arenon-convex limitations on joint beamforming between the BS and the RIS,and a

206、highly complexexhaustive search algorithm is required to obtain the optimal solution.For example,in the case of1-bit quantification,with 64 antennas,the search space would be up to 264,which is not availablein practice.As the RIS phase shift is low-bit quantitative,the phase shift of each unit is qu

207、antifiedin a fixed set,that is,the quantification of low-bit quantitative phase shift for each unit can beregarded as a classification issue.Given the excellent performance of machine learning algorithmsin solving non-convex classification issues,we can use the deep learning methods and input thecha

208、nnel matrix to predict the classification results of the phase shift for each unit of the RIS.Afteroffline training,the performance close to the best practices can eventually be achieved with lowcomplexity.Compared with large-scale MIMOs,RISes usually have more units and larger channeldimensions,and

209、 thus face greater difficulty in obtaining channel information.Especially in mobilescenarios,users movement causes channels to change rapidly as time varies and requires repeatedbeam training to obtain accurate channel information,which will result in huge pilot overheadsand seriously limit the actu

210、al deployment of the RIS.For this issue,beam training assisted bybeam tracking can be employed to update the channel information quickly at a low overheadaccording to the past channel information and the time-varying rules of channels.Time-varying channels are affected by many environmental factors,

211、such as users movementspeed,movement direction,and whether the link is blocked.These factors are usually difficult tomodel,so unsupervised reinforcement learning can be selected,and dynamic environmentalcharacteristics can be adaptively captured using historical beam training information,tosignifica

212、ntly reduce pilot overheads.Specifically,the RIS beam tracking process is first modeledas a Markov Decision Process(MDP).Assuming that the optimal beam at the previous momenthas been obtained,the action space can be defined as the distance between the optimal beam atthis moment and the optimal beam

213、at the previous moment within the beam domain.The reward isdefined as the combined rate reachable by the optimal beam predicted based on the current action.Based on Q-learning strategies,online learning is selected specifically for environmentalinformation,to intelligently change the beam selection

214、strategy,and greatly reduce pilot overheadsunder time-varying channels of the RIS.The RIS consists of numerous controllable elements.Each of them can independently changethe phase of the incoming wave and reflect it back to the environment.However,with theintroduction of the RIS containing numerous

215、controllable components,the communication systembecomes more complex.The major challenge is the absence of closed-form solutions to thepreceded vector at the BS and the precoded matrix at the RIS in the RIS-assisted MIMO system.Therefore,many convex optimization methods are employed to iteratively o

216、ptimize the precodedvector and matrix,and fixed-design signal detection methods are used at clients.On this basis,theentire system is optimized by separately optimizing individual modules,but this does notguarantee the global optimization of the entire system.To solve this issue,the RIS-assistedcomm

217、unication system can be jointly optimized based on the concept of joint optimization of23end-to-end communication systems.The RIS passively reflects signals to assist the communication between the BS and UEs.TheRIS reflection coefficient is regulated by the RIS controller from the BS.Therefore,the p

218、recodingat the BS side is called the active beamforming,while the reflected beamforming at the RIS side isnamed the passive beamforming.The key to improving the RIS system performance is the designof a combined active and passive beamforming.Usually,such joint optimization is non-convexand involves

219、highly coupled variables,bringing some challenges to the solutions.RIS parameters are always challenged by the following factors:Due to tremendous RIS units,numerous configured parameters,and high computationalcomplexity,plus the fact that the RIS itself has no computing power,the computingloads of

220、the BS or the edge server are increased.Current optimization algorithms are mostly based on perfect instantaneous CSI.Obtaining these CSI leads to an extremely large overhead for estimating cascadingchannels.Therefore,it is necessary to consider some optimization algorithms with someCSI or statistic

221、al CSI.As the reflector phase is coupled to channels,if both parameters are optimizedsimultaneously,there are great difficulties in achieving the joint optimization design ofactive and passive beamforming under certain optimization objectives and constraints.Therefore,different algorithms are propos

222、ed to solve issues for different target functions andapplication scenarios.This section summarizes several optimization algorithms for the parameterdesign of the RIS.2.4.1.Common Parameter Optimization Algorithms1)SDRAt present,the RIS modeling at the communication layer is usually designed as anN-d

223、imensional diagonal matrix,represented by),.,(11NjNjeediag.N indicates thenumber of reflection units in the RIS.The nthdiagonal element in the matrix represents theamplitude and phase shift of the nthreflection unit in the RIS.Most of the current referencesconsider the ideal phase shift,i.e.,2,0,1nn

224、,which means each RIS unit has aunit-modulus constraint(UMC)that is non-convex.The SDR is a typical algorithm for UMCprocessing and is widely used in precoding research of the RIS.The general process based on the SDR algorithm is to turn optimization issues into a convexsemidefinite programming(SDP)

225、issue to be solved by convex optimization solvers,such as CVX.Generally,however,the solution to the issue is not that to Rank 1,so it should be restored byGaussian randomization 12.Currently,studies have shown that this algorithm can obtainapproximately the optimal solution and ensure high-quality s

226、olutions within the polynomial time.However,due to tremendous units in the RIS,this method is too complex to obtain desirablesolutions in practice.242)MMMMrepresents“Majorize-Minimization”or“Minorize-Maximization”,dependingonwhether the required optimization is minimization or maximization.The Major

227、ize-Minimization isto obtain the minimum value of the upper bound functions of each target function at each iteration.The Minorize-Maximization is to get the maximum value of the lower bound functions of eachtarget function at each iteration.The MM algorithm is an iterative optimization method that

228、uses the function convexity toobtain the maximum or minimum value.When it is difficult to optimize target function f(),thealgorithm does not directly solve the optimal solution of the target function.Instead,it seeks aseries of easy-to-optimize substitutes,substitute function g(),for target function

229、 f()and thensolves substitute function g().The optimal solution of substitute function g()is close to that off().After each iteration,a new substitute function for the next iteration is created based on thesolution.The new substitute function is then optimized to find the solution for the next itera

230、tion.The solution increasingly closer to the optimal solution of the target function is obtained afterseveral iterations.The MM algorithm can reach a desirable compromise between the computing amount of eachiteration and the total number of iterations.The algorithm has the characteristic of monotono

231、usimprovement.For example,in Majorize-Minimization,it will monotonously reduce the value ofthe true target after each interaction,which means that the target value is convergent.3)QuantitationUnder the assumption of limited phase shift,quantitation relaxes each phase shift variable toa continuous on

232、e.Each of the continuous variables resulting from relaxation will be quantified tothe discrete value closest to it.However,quantitation may degrade system performance and onnon-convex unit-modulus constraint will survive the continuous relaxation.4)Deep learningThe latest studies optimized RIS param

233、eters with the assistance of deep learning.13developed a DRL-based algorithm to obtain a joint design by observing the predefined rewardsand interacting with the environment through trial and error under continuous states and actions,instead of using complex mathematical formulas and numerical optim

234、ization techniques.If the direct transmission between a base station and a user is completely blocked,the joindesign produced in research on the transmission beamforming of the base station and RISreflectivity matrix can be used as the output of a DRL neural network to maximize the utilizationof the

235、 DRL multi-user downlink MISO system and its speed.Specifically,the policy-based deepdeterministic policy gradient(DDPG)is used to tackle the CW formation matrix and phase shift.2.4.2.Application Scenarios1)Single-user transmissionIssues in RIS single-user transmission are typically non-convex and h

236、ard to find the optimalsolution.To represent the theoretical performance gain brought by RIS,12 assumed that the CSIof all channels in the AP was known and addressed the unit-modulus constraint with SDR.Nevertheless,this approach can only generate an approximate solution,which has no guarantee of25b

237、eing optimal.Prior RIS research has mainly hypothesized an ideal phase-shift model in which aunified reflection amplitude is used without considering the phase shift of individual elements.However,when the reflection amplitude depends on phase shift,such a reflection design isgenerally no longer opt

238、imal and can lead to performance degradation.14 considered a practicalphase-shift model in which the amplitude and phase shift cannot be separately adjusted,compounding the problem.For this reason,alternative optimization(AO)and punishment-basedoptimization techniques were employed to efficiently fi

239、nd the second-best solution.In 2020,Yu etal.proposed a branch-and-bound(BnB)algorithm to address the major issues in beamformeroptimization15.This first globally optimal algorithm developed by a RIS-assisted MISO systemin the literature ensured the acquisition of the optimal solution in single-user

240、circumstances and,though with high complexity,could address discrete phase shift.In addition,approximate optimalsolutions were obtained with a manifold optimization-based algorithm taking the proposed BnBalgorithmasthebaseline,especiallyinlarge-scalewirelesssystems.ForRIS-assistedpoint-to-point MISO

241、 communication systems,16 put forward two new algorithms to optimizeAP beamforming and RIS phase shift,with the unit-modulus constraint addressed respectively byfixed-point iteration and the multi-optimization technique.Different from the results of 12,bothproposed algorithms ensure the partial opti

242、mal solution of the beamformer at AP and RIS phaseshifts and are superior to the most advanced SDR approaches in spectral efficiency andcomputational complexity.2)Multi-user transmissionCompared with single-user transmission,multi-user systems significantly improve the SNRperformance of all users in

243、 a network by jointly optimizing both base station transmissionbeamforming and RIS reflection beamforming.To solve joint optimization problems,AO isgenerally used to divide a joint optimization problem into two sub-problems,i.e.,active basestation transmission beamforming and passive RIS beamforming

244、,and then obtain their optimalsolutions respectively.This approach is advantageous in that,for a given vector of passive RISbeamforming,it converts the problem of designing active base station transmission beamformingto a conventional optimization problem,for which numerous optimization tools are re

245、adilyavailable to solve its solution.Notwithstanding,for a given vector of base station beamforming,the passive RIS beamforming remains a knotty problem to be tackled.The main challengesinclude the unit-modulus constraint and the discrete nature of the inherent feasible set of the RISreflection unit

246、.17 researched a multi-user MISO system comprising of several antenna basestations(BSs),each communicating with a single-antenna UE.The RIS was used to assist wirelesstransmission and inhibit inter-cell interference.Efficient algorithms based on alternativeoptimization can be utilized to maximize us

247、ers minimum weighted SNR by jointly optimizing thecoordinated transmission beamforming of BSs and the RIS reflection beamforming.To raise the system summation rate as far as possible,Huang et al employed the combinationof MM and alternative optimization in the context of a multi-user MIMO system.The

248、y managedto increase the system summation rate by over 40%without increasing any energy consumption18.In addition,Wang et al proposed in 2020 a self-defined block-coordinate acceleratedproximal gradient(APG)algorithm 19 to jointly optimize the transmission beamformer on basestations and continuous o

249、r discrete phase shifts on the RIS,thus maximizing the summation rate.To minimize the total AP transmission power,8 put forward a punishment-based26optimization algorithm and a two-phase optimization algorithm for obtaining the second-bestsolution in the context of multi-user transmission.The two al

250、gorithms offer different tradeoffsbetweencomplexityandperformance.Thepunishment-basedapproachdeliversbetterperformance with higher complexity.Under the constraint of user signal to interference plus noiseratio(SINR),12 generalized the single-user case to the multi-user setting,which morecommonly fou

251、nd,to minimize the total AP transmission power.Similar to the single-user case,thealgorithm used alternative optimization,but the difference was that for AP transmissionbeamforming,the principle of minimum mean square error(MMSE)was applied to address themulti-user interference,instead of MRT being

252、used in the single-user case where there was nointerference.Moreover,inspired by the maximization of combined channel gain in single-usersituations,researchers divided the design of joint beamforming into two beamforming sub-issuesand utilized the two-phase algorithm,which was less complex than the

253、alternative optimizationalgorithm,to optimize the phase shift and transmission beamforming respectively.References1.W.Tang,M.Z.Chen,X.Chen,J.Y.Dai,Y.Han,M.Di Renzo,Y.Zeng,S.Jin,Q.Cheng,and T.J.Cui,“Wireless communications with reconfigurable intelligent surface:Path lossmodeling and experimental mea

254、surement,”IEEE Trans.Wireless Commun.,vol.20,no.1,pp.421-439,Jan.2021.2.W.Tang,X.Chen,M.Z.Chen,J.Y.Dai,Y.Han,M.Di Renzo,S.Jin,Q.Cheng,and T.J.Cui,“Path loss modeling and measurements for reconfigurable intelligent surfaces in themillimeter-wave frequency band,”arXiv:2101.08607,Jan.2021.3.M.Di Renzo,

255、F.H.Danufane,X.Xi,J.de Rosny,and S.A.Tretyakov,“Analytical Modelingof the Path-loss for Reconfigurable Intelligent Surfaces-Anomalous Mirror or Scatterer?”IEEE21stInternationalWorkshoponSignalProcessingAdvancesinWirelessCommunications(SPAWC),May 2020.4.G.Gradoni and M.Di Renzo,“End-to-End Mutual-Cou

256、pling-Aware Communication Modelfor Reconfigurable Intelligent Surfaces:An Electromagnetic-Compliant Approach Based onMutual Impedances”,IEEE Wireless Commun.Lett.,vol.10,no.5,pp.938-942,May 2021.5.Dou Jianwu,Chen Yijian,Zhang Nan,et al.On the Channel Modeling of IntelligentControllable Electro-Magne

257、tic-Surface J.Chinese Journal of Radio Science,2021,36(3):368-377.6.Z.He and X.Yuan,“Cascaded channel estimation for large intelligent metasurface assistedmassive MIMO,IEEE Wireless Commun.Lett.,vol.9,no.2,pp.210-214,Feb.2020.7.D.Mishra and H.Johansson,“Channel estimation and low-complexity beamform

258、ing designfor passive intelligent surface assisted MISO wireless energy transfer,”in Proc.IEEE Int.Conf.Acoust.,Speech Signal Process.(IEEE ICASSP19),Brighton,UK,May 2019,pp.46594663.8.Q.Nadeem,H.Alwazani,A.Kammoun,A.Chaaban,M.Debbah,and M.S.Alouini,“Intelligent reflecting surface-assisted multi-use

259、r MISO communication:Channel estimationand beamforming design,”IEEE Open J.Commun.Soc.,vol.1,pp.661680,May 2020.9.Hu C,Dai L.Two-Timescale Channel Estimation for Reconfigurable Intelligent Surface27Aided Wireless CommunicationsJ.2019.10.Wang P,Fang J,H Duan,et al.Compressed Channel Estimation and Jo

260、int Beamforming forIntelligent Reflecting Surface-Assisted Millimeter Wave SystemsJ.2019.11.You C,Zheng B,Zhang R.Channel Estimation and Passive Beamforming for IntelligentReflecting Surface:Discrete Phase Shift and Progressive RefinementJ.2019.12.Q.Wu and R.Zhang,“Intelligent reflecting surface enh

261、anced wireless network:Joint activeand passive beamforming design,”in Proc.IEEE GLOBECOM,Dec.2018.13.Huang C,Member,IEEE,et al.Reconfigurable Intelligent Surface Assisted Multiuser MISOSystems Exploiting Deep Reinforcement LearningJ.IEEE Journal on Selected Areas inCommunications,2020,38(8):1839-185

262、0.14.S.Abeywickrama,R.Zhang,Q.Wu and C.Yuen,Intelligent Reflecting Surface:PracticalPhaseShiftModelandBeamformingOptimization,inIEEETransactionsonCommunications,vol.68,no.9,pp.5849-5863,Sept.2020.15.X.Yu,D.Xu and R.Schober,Optimal Beamforming for MISO Communications viaIntelligent Reflecting Surface

263、s,2020 IEEE 21st International Workshop on SignalProcessing Advances in Wireless Communications(SPAWC),2020,pp.1-5,16.X.Yu,D.Xu and R.Schober,MISO Wireless Communication Systems via IntelligentReflecting Surfaces:(Invited Paper),2019 IEEE/CIC International Conference onCommunications in China(ICCC),

264、2019,pp.735-740.17.H.Xie,J.Xu and Y.Liu,Max-Min Fairness in IRS-Aided Multi-Cell MISO Systems viaJoint Transmit and Reflective Beamforming,ICC 2020-2020 IEEE InternationalConference on Communications(ICC),2020,pp.1-6.18.C.Huang,A.Zappone,M.Debbah and C.Yuen,Achievable Rate Maximization by PassiveInt

265、elligent Mirrors,2018 IEEE International Conference on Acoustics,Speech and SignalProcessing(ICASSP),2018,pp.3714-3718.19.Li Q,Cui X,Wu S X,et al.Sum Rate Maximization for Multiuser MISO Downlink withIntelligent Reflecting SurfaceJ.2019.28Chapter III RIS Realization and Prototype Validation3.1.Exper

266、imental Validation 1To explore the actual performance of RIS,its application scenarios,as well as possibleproblems in its indoor and outdoor deployment,China Mobile completed in Nanjing the testingvalidation of the feasibility of the RIS new technology in a 5G live network environment in June2021,in

267、 collaboration with the team of Cui Tiejun,an academician of CAE,at SoutheastUniversity and the Hangzhou Qiantang Information Co.,Ltd.The RIS new technology validatedwill enable adjustable electromagnetic unit components and flexible control of the beamformingdirection.1.Testing EnvironmentTo ensure

268、 that the test result would truly reflect the improving effect of RIS at cell edges andin weak coverage areas,the working frequency was locked for all UEs used in the validation.Further,the validation was conducted in several scenarios possible for actual application,including under-tower shadow zon

269、e,indoor coverage from outdoors,and outdoor traversing.1)Under-tower shadow zoneAn under-tower shadow zone,i.e.,a weak coverage zone,often exists under a base stationtower,due to the limitations of the downtilt angle of the base station antenna and antenna direction.A RIS was installed in a location

270、 with good reception inside the test cell and the RIS panelparameters were adjusted so that signals from the base station would be reflected to theunder-tower shadow zone.The CDF distribution differences in indicators,such as RSRP,SINR,and throughput of users in the zone,were compared to examine the

271、 overall improving effect ofRIS on signal coverage in the under-tower shadow zone and validate whether the RIS could satisfythe high-speed services for users in the zone.Figure 3-1 shows the site of the test case.Theselected cell base station mainly covered urban roadways.The RIS,installed at the di

272、agonallyopposite side of the road in front of a building,received signals from the base station(46 metershigh)on the 13th floor of the building and reflected the signals to the weak signal coverage area atthe back of the building.The straight-line distance between the base station and the RIS was ab

273、out120 meters and the transmission in between traveled on a line-of-sight(LOS)path.Figure 3-1 Image of the Site for the Under-Tower Shadow Zone Test292)Indoor coverage from outdoorsDuring propagation,electromagnetic signals will be significantly attenuated by factors suchas reinforced concrete build

274、ing walls,glass curtain walls,and aluminum alloy building materials.Therefore,in most cases,indoor environments are typical weak coverage settings.Capable oftuning the reflection beam shape,the RIS can enhance indoor signal coverage by convergingbeams for greater penetration.In this test case of ind

275、oor weak coverage,the overall effect of RISin the indoor coverage from outdoors scenario was examined by comparing the transmissionperformance difference before and after RIS installation at multiple points.The indoor scenario ofthe test case was set up in an office building with glass curtain walls

276、.Stable transmission testswere conducted at multiple points inside the building.The RIS reflection panel was installed onthe opposite of the road under the building.The straight-line distance between the base station andthe RIS was about 65 meters and the transmission in between traveled on a line-o

277、f-sight(LOS)path.See Figure 3-2 for the image of the site for the test case.Figure 3-2 Image of the Site for the Indoor Coverage from Outdoors Test3)Outdoor traversingRIS can flexibly control the reflection beams and the beam shape and thus can be used forfilling coverage holes,improving coverage at

278、 cell edges,and other scenarios.Given that,theoutdoor traversing test was conducted in an urban area.The improving effect of the installed RISon base station coverage was investigated by comparing the differences in indicators of UEtraversing in the cell under overlapped coverage before and after th

279、e installation.A pole-mountedsite(10 meters high)in an urban area with a dense population was selected for the test case.Occluded by high buildings in the test cell,the signals of the pole-mounted site only covered alimited range.(The road perpendicular to the one where the pole-mounted site stood w

280、as a weakcoverage area.)The RIS was deployed at a crossroad 70 meters away from the pole-mounted siteto receive signals from the pole-mounted site and reflect the signals to the weak coverage road inthe test cell.UE traversing was conducted on the weak coverage road at a constant speed beforeand aft

281、er the RIS deployment respectively.The transmission path between the RIS and the UE andthat between the UE and the base station were good LOS paths.See Figure 3-3 for the image ofthe site for the test case.30Figure 3-3 Image of the Site for the Outdoor Traversing Test2.Base Station and Cell Configur

282、ationTwo outdoor base stations were involved in the above test cases.One was on the 13th floor ofa building with glass curtain walls and the other was a pole-mounted site in an urban area with adense population.See the table below for information on the base stations and cells.Table 3-1 Base Station

283、 Configuration InformationTest CaseTypeTransmissionPowerRRUTypeBaseStationModelDowntiltAngleof InstallationDirectionAngleof InstallationUnder-towershadow zoneOutdoorbasestation327 W64channelsHuawei9/1060Indoor coveragefrom outdoorsOutdoortraversingOutdoorbasestation327 W64channelsHuawei6/3200Table 3

284、-2 Cell Configuration InformationTest CaseSectorNumberDownlinkFrequencyPointDownlinkBandwidthPhysical CellIdentificationCellDuplexModeTimeslotRatioUnder-towershadow zone1504990100301TDD8:2Indoor coveragefrom outdoorsOutdoortraversing250499010013TDD8:23.RIS Panel ConfigurationThe field performance te

285、st was jointly conducted by China Mobile,Southeast University,andHangzhou Qiantang Information Co.,Ltd.using a 160 cm*80 cm panel.31Figure 3-4 RIS Prototype for Field Test4.Validation Conclusions1)Under-tower shadow zoneRespectively before and after RIS deployment,a UE was carried to traverse the un

286、der-towershadow zone at a constant speed along the same route.This resulted in the RSRP,SINR,andthroughput dot figures before and after the RIS deployment,as shown in Figure 3-5.Figure 3-5 RSRP,SINR,Throughput Dot Figures Before andAfter RIS Deployment inUnder-Tower Shadow Zone Test CaseWith the RIS

287、 deployed,the RSRP and throughput were found to significantly improve andthere were relatively fewer weak,low-quality overage areas.No significant changes wereobserved in SINR.According to analysis,this may be because the RIS also magnified theinterference signals of adjacent cells while reflecting

288、the base station signals.Moreover,relativelysubstantial performance improvement was observed at the back of the office building as well,although the UE-RIS transmission path was NLOS.Reasons for this phenomenon may be that asetting full of refraction and diffraction formed by nearby obstacles made t

289、he base stationdownlink signals reflected by the RIS receivable by the UE within the area.The test results indicated that before RIS deployment,both the RSRP of cell edge users and32the average RSRP of users were relatively lower at-102.18 dBm and-94.93 dBm,respectively.After RIS deployment,RSRP was

290、 increased to some extent.The figure was increased by 4.03 dBfor the RSRP of cell edge users and 3.8 dB for the average RSRP of users.The RIS deploymentraised the average user throughput from 91.50 Mbps to 109.00 Mbps,up about 19%.However,itmade no significant difference to SINR.2)Indoor coverage fr

291、om outdoorsThe indoor scenario was set up on the 2nd and 4th floors of an office building with glasscurtain walls.For the office area on the 2nd floor,eight points were selected,covering threetypical scenarios:conference room,office,and studio.The 4th floor was a supermarket,wherestable transmission

292、 tests were conducted at four locations.See Figure 3-6 for a diagram of the testpoint locations.Figure 3-6 Schematic Plan for Fixed-Point Transmission Test in the Office BuildingIt was found during the test process that the UE could not receive downlink signalstransmitted from the base station when

293、it was behind the indoor reinforced concrete walls.Thissuggested that the RIS version used in the test had lower penetration capability to cover indoorenvironments from outdoors such that it was not able to further penetrate an interior wall afterpenetrating a glass curtain wall.The following data w

294、ere obtained by averaging the data recordedafter one minute of stable signal transmission at each point that can receive downlink signals.Preliminary conclusions were drawn by analysis of the fixed-point transmission data beforeand after RIS deployment.Specifically,after RIS deployment,RSRP,SINR,and

295、 downlinktransmission rate were increased at most points.Despite the loss after signals penetrating the glasscurtain wall,RSRP increases of 317 dB were stilled found,averaging at 10 dB across all points,and the transmission rate increases of 5137 Mbps were also observed,averaging at 78.19 Mbpsacross

296、 all points.The increment varied significantly across the points probably due to signalfluctuations and limited range of RIS coverage.Although the gain was substantially improved,thesignals failed to continue to penetrate an interior wall(15 dB of interior wall penetration loss orabove)as the indoor

297、 coverage had poor overall quality and the basic level was lower in the testsetting(about-100 dB after RIS deployment).In addition,the gain was not significantly improvedin the studio on the 2nd floor.Since the studio was on the rightmost side of the building and theRIS reflection beams were limited

298、 in width,it was likely that the test area was beyond thecoverage of RIS reflection beams.3)Outdoor traversing33The outdoor traversing test was conducted by carrying a UE to traverse the test cell at aconstant speed along the same route,respectively before and after RIS deployment.This resultedin th

299、e RSRP,SINR,and throughput dot figures shown in Figure 3-7.Figure 3-7 RSRP,SINR,Throughput Dot Figures in the Outdoor Traversing TestThe dot figures show that RIS deployment did not lead to significant changes to performanceindicators in areas where RSRP was greater than-80 dBm,whereas in areas wher

300、e RSRP wassmaller than-90 dBm,RSRP,SINR,and throughput after RIS deployment were significantlybetter than those before RIS deployment.Further,the test data suggest that RIS deployment had a significant influence on cell edgeusers,with RSRP,SINR,and throughput increasing by about 3.3 dB,1.45 dB,and 7

301、9 Mbps,respectively.In contrast,the average RSRP for users was improved by 1.25 dB,meaning nosignificant influence on the average gain for users.In addition,to examine the improving effect ofRIS on the cell coverage,a remote base station test was conducted in the scenario.Specifically,respectively b

302、efore and after RIS deployment,a tester carrying a UE kept moving until the UEcould not receive signals.A comparison between the two locations of communication interruptionshowed that the longest reception distance before RIS deployment was 150 meters and wasincreased by about 60 meters to 210 meter

303、s after.Results of the test case demonstrated that RISwas of substantial practical value in improving performance for cell edge users and extending cellcoverage.3.2.Experimental Validation 2From 2018 to 2021,NTT DOCOMO,a Japanese company,completed several validationexperiments on RIS prototypes usin

304、g different prototype systems,confirming potential applicationscenarios of intelligent metamaterial surfaces such as filling outdoor coverage holes,interferencecontrol,and outdoor-to-indoor coverage enhancement.34Figure 3-8 Onboard Millimeter-Wave Metamaterial Reflector(left);Test Scenario and Resul

305、ts(right)1In 2018,DOCOMO worked with Metawave Corp.to conduct an outdoor field validationexperiment on the coverage enhancement capability of metamaterial reflector in a 28 GHzmillimeter-wave system 1.The metamaterial reflector used in the experiment,manufactured byMetawave Corp.,was deployed on the

306、 top of the test vehicle,as shown in Figure 3-8.Themetamaterial reflector was 80 cm 80 cm 5 cm and weighed around 700 g.With plane waveincidence taken into account,its reflection beam width was 18 degrees.The base station wasdeployed on the roof of a building in the experimental area.The roads under

307、 the building wereblind zones,to which the metamaterial reflector was used to reflect 28 GHz millimeter-wavesignals in the experiment.Results showed that using a metamaterial reflector to fill coverage holescould improve the SNR by over 15 dB and increase the rates by over 500 Mbps.Thus,theexperimen

308、t verified the capability of metamaterial surfaces to fill coverage holes.Figure 3-9 Appearance(Left)and Structure(Right)of a Millimeter-Wave Transparent DynamicMetasurface235Figure 3-10 Imagined Application Scenarios of Transparent Dynamic Metasurface2DOCOMO cooperated with AGC in 2020 to develop a

309、 transparent dynamic metasurface formillimeter waves and conducted an experiment in DOCOMO R&D Center to verify the dynamictuning capability and transparency effect of the metasurface2.The metamaterial surface used inthe experiment,designed by DOCOMO and manufactured by AGC,comprised of two substrat

310、es,one of which contained transparent electromagnetic material.As shown in Figure 3-9,themetamaterialsurfacecouldswitchbetweenthreemodes,i.e.,transparenttransmission,semi-reflection,and reflection,through adjustments to the gap between the two substrates.Theobserved results showed that when the meta

311、surface was in transparent transmission mode,28 GHzmillimeter waves were only attenuated by about 1 dB.However,this figure was more than 10 dBwhen the metasurface was in reflection mode.Figure 3-10 gives an example of the application anddeployment of the transparent metamaterial surface.Transparent

312、dynamic metasurfaces can beflexibly configured to transparent transmission or reflection mode to meet the communication orinterference shielding needs across various scenarios,such as office,warehouse,and confidentialareas.The experiment proved the capability of metamaterial surfaces being applied i

313、n interferencecontrol scenarios.Meanwhile,metamaterial surfaces with transparent design can be deployed at a large scale onwindows,building surfaces,and other positions to realize the required functions since they do notinfluence the appearance of the surrounding environment.Figure 3-11 Fixed-Focus

314、Metasurface Lens(Left)and Metasurface Lens with Focus ControlFunction(Right)336Figure 3-12 Common Glass,Fixed Metasurface Lens,and Dynamic Metasurface Lens(from Leftto Right)3Figure 3-13 Metasurface Lens Combined with a Repeater installed on the Focus to EnhanceIndoor Coverage 3In 2021,DOCOMO contin

315、ued its partnership with AGC and developed metasurface lensesfor millimeter waves,shown in Figure 3-11,validating the focusing and dynamic focus controlcapabilities of metasurfaces3.The metasurface lenses were also manufactured with transparencymaterial and capable of focusing signals and dynamicall

316、y changing the focus.The experimentalresult showed that,compared with common glass,metasurface lenses improved the SNR at thefocus by around 24 dB.The dynamic metasurface lens designed and manufactured by DOCOMOand AGC were capable of focus control and could switch between 1-2 focus,as shown in Figu

317、re3-12.When the metasurface lens operated in single-focus mode,the gain at the focus was about 11dB.When it operated in dual-focus mode,the gains at the focuses were 6 dB respectively.Millimeter-wave outdoor-to-indoor coverage enhancement is a typical application scenario of themetamaterial lens.The

318、 metamaterial lens can be used in combination with a repeater or other relaydevices to improve indoor millimeter-wave coverage.As shown in Figure 3-13,a repeater can bedeployed at the focus of the metamaterial lens so that the communication link between therepeater and the outdoor millimeter-wave ba

319、se station will be efficient and stable.At the sametime,after being magnified and forwarded by the repeater,the coverage of the outdoormillimeter-wavebasestationcanbeexpandedtotheindoorareas,enhancingthe37outdoor-to-indoor coverage.3.3.Experimental Validation 3In June 2021,China Unicom and ZTE valid

320、ated the RIS technology in a 5G fieldenvironment in Shanghai.In the 5G live network environment,the gNB,RIS,and UE weredeployed as shown in Figure 3-14.The C-band gNG was deployed on the roof of a prototypebuilding,whereas the UE was deployed in the channel.The communication path between the gNBand

321、the UE was an NLOS path.Due to limitations of the on-site environment,the RIS wasdeployed about 60 meters away from the gNB and about 52 meters away from the UE,to ensurethat a LOS transmission path existed between the RIS and the gNB and to shorten the distancebetween the RIS and the UE.Nevertheles

322、s,conditions remained to be met for the LOS pathtransmission between the RIS and the UE.Before the RIS was deployed,the rank fluctuatedbetween 1-2 and the throughput was about 250 Mbps because there was no LOS transmission pathbetween the gNB and the UE,and the RSRP and the SINR were relatively lowe

323、r around the UE inthe channel.With the RIS deployed,although the RSRP and the SINR around the UE were notsignificantly increased,significant improvements were observed for the wireless channel rank andthroughput.The rank fluctuated between 3-4 and the throughput rose to 350 Mbps.The test resultssugg

324、est that performance was improved by over 40%at the edge of the NLOS coverage cell of the5G IF base station.Figure 3-14 Deployment of the RIS Field Test Conducted by China Unicom and ZTE38Figure 3-15 Image of the RIS Field Test Conducted by China Unicom and ZTEThe RIS field test confirmed that the t

325、echnology can reconstruct the wireless transmissionenvironment between a 5G IF base station and commercial endpoints and boost the in-depthcoverage capability of the 5G network at extremely low energy consumption cost.It thus willprovide a revolutionary wireless technology to enable the new service

326、of 5G-Advanced ultra-highexperience.References1.D.Kitayama,D.Kurita,K.Miyachi,Y.Kishiyama,S.Itoh and T.Tachizawa,5G RadioAccess Experiments on Coverage Expansion Using Metasurface Reflector at 28 GHz,2019IEEE Asia-Pacific Microwave Conference(APMC),2019,pp.435-437.2.NTT DOCOMO,“DOCOMO Conducts World

327、s First Successful Trial of TransparentDynamic Metasurface,”https:/www.nttdocomo.co.jp/english/info/media_center/pr/2020/0117_00.html.3.NTT DOCOMO,“DOCOMO andAGC Use Metasurface Lens to Enhance Radio SignalReception Indoors,”https:/www.nttdocomo.co.jp/english/info/media_center/pr/2021/0126_00.html.3

328、9Chapter IV Technical Challenges and Standardization4.1.Technical ChallengesSo far,the industry has planned the RIS test validation and trial application in somescenarios to promote the key technological validation and comprehensive performance assessmentof RIS.However,its not uncommon to find that

329、severe challenges remain in the hardwarerealization,engineering deployment,theoretical and solution design,control solutions,networkarchitecture,and networking of RIS.For hardware realization,RIS material and devices are not quite mature and are relativelycostly.The performance of adjustable compone

330、nts cannot meet control needs and their structuraldesign remains to be optimized.With the limitations of the control rates of adjustable components,the RIS has not been able to be dynamically controlled at a high frequency band.The RIS has acertain degree of reflection loss,thus having difficulties

331、in achieving ultra-long distance coverage.Currently,the RIS more commonly used on low-bit adjustable elements is faced with the problemof increased large-angel beamforming grating lobes,which will degrade other aspects of usercommunication performance.As a result,in multi-user communication situatio

332、ns,networking mayproduce interference to other cells and even the networks of other operators.Further,ametamaterial panel comprises hundreds of cycle units.Some faulty tuning components on themetasurface will cause metamaterial units to fail to deliver functions as expected and it will bedifficult t

333、o troubleshoot and repair the faulty units.The increased number of faulty metamaterialunits will lead to magnetic-wave tuning performance degradation of the entire metamaterial.Thisincludes problems such as lower beam gain,beam pointing deviation,increase in minor lobes,etc.In terms of engineering deployment,the larger size of the RIS panel makes it necessary tocommunicate with the property manage