PaperlayerView exact revision on arXiv

Source-linked AI summary

Reconfigurable Intelligent Surfaces: Principles and Opportunities

Yuanwei Liu, Xiao Liu, Xidong Mu, Tianwei Hou, Jiaqi Xu, Marco Di Renzo, Naofal Al-Dhahir

arXiv:2007.03435v3eess.SP

TL;DR

The paper evaluates multi-antenna assisted RIS systems and discusses integrating machine learning with RISs. It also surveys optimization challenges and reports potential gains in channel gains, QoS, coverage range, and energy dissipation.

  • Problem

    RIS-enhanced wireless networks present new research challenges, including sophisticated information-theoretic investigations and optimization issues.

  • Method

    The paper develops performance evaluation techniques for multi-antenna assisted RIS systems and discusses amalgamating machine learning with RISs.

  • Results

    RIS-enhanced wireless networks can achieve tuned channel gains, improved QoS, enhanced coverage range, and reduced energy dissipation.

  • Takeaways & Limitations

    Machine learning and big-data analytics are discussed as approaches for optimizing RIS-enhanced wireless networks.

  • Takeaways & Limitations

    Information-theoretic investigations remain important and sophisticated, and further mathematical tools are expected to be employed.

Abstract

from arXiv · show

Reconfigurable intelligent surfaces (RISs), also known as intelligent reflecting surfaces (IRSs), or large intelligent surfaces (LISs), have received significant attention for their potential to enhance the capacity and coverage of wireless networks by smartly reconfiguring the wireless propagation environment. Therefore, RISs are considered a promising technology for the sixth-generation (6G) of communication networks. In this context, we provide a comprehensive overview of the state-of-the-art on RISs, with focus on their operating principles, performance evaluation, beamforming design and resource management, applications of machine learning to RIS-enhanced wireless networks, as well as the integration of RISs with other emerging technologies. We describe the basic principles of RISs both from physics and communications perspectives, based on which we present performance evaluation of multi-antenna assisted RIS systems. In addition, we systematically survey existing designs for RIS-enhanced wireless networks encompassing performance analysis, information theory, and performance optimization perspectives. Furthermore, we survey existing research contributions that apply machine learning for tackling challenges in dynamic scenarios, such as random fluctuations of wireless channels and user mobility in RIS-enhanced wireless networks. Last but not least, we identify major issues and research opportunities associated with the integration of RISs and other emerging technologies for applications to next-generation networks.

I. INTRODUCTION

RISs are presented as a reconfigurable approach to improving wireless-network coverage, quality, efficiency, and connectivity across diverse applications. The paper surveys their principles, system evaluation, optimization, machine-learning integration, and opportunities with emerging technologies.

  • RIS fundamentals: RISs are nearly-passive electromagnetic devices that reconfigure incident signals through arrays of reflecting elements.They can be deployed on structures including building facades, indoor walls, aerial platforms, roadside billboards, vehicle windows, and clothing.
  • RIS benefits: RISs can form virtual line-of-sight links and improve throughput, coverage, QoS, and received SINR, especially when direct links are blocked.Their phase-controlled reflections can bypass obstacles, compensate for power loss, and mitigate interference.
  • RIS benefits: Compared with conventional amplify-and-forward and decode-and-forward relays, RISs shape signals without employing a power amplifier, making deployment more energy-efficient.RISs control the phase shift of each reflecting element rather than actively amplifying or decoding signals.
  • Paper scope: The paper addresses a need to categorize heterogeneous RIS research and provides a comprehensive treatment of operating principles, performance evaluation, optimization, machine learning, and emerging-technology integration.Its contributions include physics- and communications-based principles, multi-antenna evaluation, beamforming and resource-allocation design, ML-based optimization, and research opportunities.
  • Applications: RISs support applications spanning cellular, indoor, WiFi, SWIPT, unmanned, and intelligent IoT networks.Examples include bypassing obstacles, reducing UAV movement and energy consumption, supporting massive connectivity, and improving indoor QoS.
  • Paper scope: The survey also compares existing research, summarizes advantages and limitations, and identifies open research problems and potential solutions for RIS-enhanced networks.The discussion covers information-theoretic investigations, joint beamforming and resource allocation, and ML for dynamic wireless conditions.

II. RIS: FROM PHYSICS TO WIRELESS COMMUNICATIONS

RISs are configurable two-dimensional structures that modify electromagnetic-wave responses, allowing wireless channels and received-signal strength to be controlled. The paper surveys their physical principles, operating categories, performance, beamforming, resource management, and learning-based extensions.

  • RIS fundamentals: RISs are two-dimensional material structures with programmable macroscopic physical characteristics and reconfigurable electromagnetic-wave responses.Their configurable response distinguishes RIS-aided networks from conventional wireless networks.
  • RIS fundamentals: Controlling the wireless channels between transmitters and receivers can enhance the desired signal strength at terminal devices.
  • Paper scope: The paper organizes its broader survey around RIS physics and communications, multi-antenna performance analysis, beamforming and resource allocation, machine learning, and discussions of emerging technologies.
  • RIS categories: RISs may use metamaterial or patch-array structures and can be tuned electrically, mechanically, or thermally.They are also categorized as passive-lossy, passive-lossless, or active according to energy consumption and amplification characteristics.
  • RIS operating principles: RISs can operate through waveguide, reflection, or refraction, converting incident or guided waves into desired free-space propagating waves.Surface equivalence principles model reflected and refracted fields using equivalent electric and magnetic currents.
  • Metasurface operation: Metasurface elements can be modeled as uncoupled magnetic dipoles whose tunable polarizabilities enable beamforming, with each element acting as a micro-antenna.A compact waveguide metasurface occupies less space than conventional antenna arrays and can transmit toward wider angles.

2) Refracting RIS:

Refracting RIS analysis combines electromagnetic equivalence principles, wave propagation models, and tunable metasurface implementations. The section contrasts ray- and wave-optics perspectives, emphasizing wave optics for power-flow analysis and practical RIS design.

  • Electromagnetic foundations: Love’s field equivalence principle provides the theoretical foundation for analyzing RIS radiation patterns through equivalent surface currents.Equivalent currents reproduce the correct fields in the considered region, while the Huygens-Fresnel principle quantifies scattered fields.
  • RIS operating mechanisms: Holographic waveguide RISs couple three-dimensional free-space waves to two-dimensional surface waves, embedding pre-designed information into the radiated field.The metasurface therefore acts as a hologram that shapes the resulting three-dimensional radiation.
  • Ray- and wave-optics perspectives: Ray optics models EM waves as geometrical rays, whereas wave optics represents electric and magnetic fields using local amplitude, direction, and phase.Both perspectives are used in RIS research, but they rely on intrinsically different assumptions and physical interpretations.
  • Ray- and wave-optics perspectives: Wave-optics designs can provide increasingly improved performance over ray-optics designs, while also enabling analysis of local and overall RIS power consumption.Ray optics is easier to adopt for quick predictions but fails when RIS power flow must be considered.
  • Tunability: RIS electromagnetic characteristics can be reconfigured by tuning surface impedance, with electrical control favored because FPGA chips can quantize and control voltage conveniently.Other tuning mechanisms include thermal excitation, optical pumping, and physical stretching.

E. RIS Operating Principles

RIS operating principles include anomalous reflection and beamforming, both interpretable as wavefront transformations. Near-field focusing and far-field steering differ in how power density varies with distance and in the locations or angles they target.

  • Operating functions: Anomalous reflection transforms one plane wave into another, whereas beamforming transforms a plane wave into a desired wavefront.Under ray optics, anomalous reflection directs an incident beam toward a far-field terminal, while beamforming is also called focusing.
  • Near field v.s. far field: The commonly used near-field/far-field boundary is z = 2L^2/λ, where L is RIS aperture size and z is distance to the field point.The boundary is motivated by inspection of how power density varies with distance and depends on the specific RIS configuration.
  • Near field v.s. far field: In the far field, power density follows spherical dissipation with distance, whereas near-field focusing produces a small focal area and can achieve high focusing gain.The far-field scaling is proportional to L^2Ω/z^2, while the near-field area is governed by the diffraction-related first term.
  • Near field v.s. far field: Near-field RIS enhancement targets users at specific locations, while far-field enhancement targets users at specific angles relative to the surface.The two regimes therefore use different spatial interpretations of the desired enhancement.
  • Generalized laws of refraction and reflection: Introducing a phase discontinuity makes reflection and refraction angles depend on incidence angle, wavelength, refractive indexes, and phase-gradient parameters.Tuning the phase gradient dΦ/dx enables anomalous reflection, although a constant gradient need not hold for all desired wave transformations.

3) Co-phase condition:

RIS phase configurations support beam steering and focusing, but practical performance depends on physical modeling, hardware constraints, and difficult joint optimization. The surveyed communication effects include channel modeling, performance analysis, and benchmark comparisons.

  • 3) Co-phase condition:: Near-field focusing accounts for non-negligible incident and reflected wavefront curvature to form a pencil beam toward the terminal.The source and terminal are considered close to the RIS, with line-of-sight links assumed for the co-phase condition.
  • 3) Co-phase condition:: The co-phase condition configures each RIS element so reflected contributions combine coherently toward a specified observer direction.Element positions, source position, and observer direction determine the phase shift φ_mn.
  • 3) Co-phase condition:: In complex wireless systems, RIS configuration requires optimization because the RIS role cannot be reduced to only beam steering or focusing.The resulting issues are discussed through communication models, performance analysis, and benchmark schemes.
  • 3) Co-phase condition:: RIS performance is constrained by theoretical assumptions and hardware implementation, including discretization of an ideally continuous surface profile.Relevant hardware parameters include phase quantization, the maximum number of integrable elements, and the coated fraction of the environment.
  • 3) Co-phase condition:: Open issues include experimentally validated outdoor path-loss models, exact channel-gain distributions, and information-theoretic characterization of attainable performance.Existing approximations are accurate only in the high-SNR regime, while globally optimal solutions remain difficult for coupled non-convex designs.
  • 3) Co-phase condition:: For electrically small RISs, received power generally increases with surface size, whereas electrically large RISs have a received power that does not grow indefinitely with size.The corresponding asymptotic path-loss descriptions use product-of-distances and sum-of-distances models, respectively.

B. Performance Analysis

Performance analysis of RIS-enhanced networks covers channel distributions, multi-antenna designs, signal enhancement, and signal cancellation. The literature reports gains in spectral and energy efficiency while also highlighting analytical and CSI-related challenges.

  • B. Performance Analysis: Existing RIS performance analysis addresses single-user and multi-user networks, including multi-antenna systems and effective channel gains after passive beamforming.The survey organizes contributions around performance analysis and compares important RIS-enhanced network designs.
  • B. Performance Analysis: RIS advantages are increasingly reported in spectral-efficiency and energy-efficiency enhancement, but exact cascade-channel distributions remain difficult to evaluate.The latter challenge limits closed-form performance analysis for RIS-enhanced networks.
  • B. Performance Analysis: RISs can enhance desired signals by co-phasing reflected waves, while signal-cancellation designs destructively combine reflected and direct signals.Cancellation formulations may use interference-channel knowledge to adjust phase and amplitude coefficients of BS-RIS-user links.
  • B. Performance Analysis: RIS research considers continuous and discrete phase shifts, imperfect CSI, fairness-oriented designs, and stochastic-geometry models for user locations.Discrete phase-shift models address the practical fact that RIS phase responses may not be continuous.
  • B. Performance Analysis: RIS-assisted interference cancellation can eliminate inter-cluster interference without active beamforming weights and detection vectors.Related applications include coordinated multi-point networks, secure beamforming, and RIS-based jamming without internal energy.

C. Benchmark Schemes

The survey benchmarks RIS-enhanced networks against random-phase surfaces and relay systems. RIS performance becomes more competitive as the number of tunable elements increases, while several analytical and application limitations remain open.

  • C. Benchmark Schemes: Benchmark schemes use non-configurable random-phase surfaces and full-duplex or half-duplex amplify-and-forward or decode-and-forward relays.These references provide comparison points for quantifying RIS performance enhancement.
  • C. Benchmark Schemes: A sufficiently large number of tunable RIS elements can allow an RIS-enhanced network to outperform a decode-and-forward relay-aided network.The comparison assumes blocked BS-user links and optimal power splitting for the relay systems.
  • C. Benchmark Schemes: The throughput gap between RIS-enhanced and relay-aided networks becomes smaller as the number of RIS elements increases.Figure 11 compares spectral efficiency against the number of RIS elements for RIS-enhanced, full-duplex-relay, and half-duplex-relay networks.
  • C. Benchmark Schemes: When N = 23 and transmit power P = 25 dBm, the proposed RIS-enhanced network outperforms both full-duplex and half-duplex relay-aided networks.The cited comparison uses network throughput, with half-duplex relaying evaluated using an equal time-split ratio.
  • C. Benchmark Schemes: Open research problems include outdoor path-loss experiments, exact channel distributions beyond high-SNR approximations, and simultaneous enhancement and mitigation of desired and interference signals.Outdoor models should be validated in environments containing reflecting and scattering objects.

IV. RIS BEAMFORMING AND RESOURCE ALLOCATION

RIS beamforming and resource allocation research jointly designs passive surface configurations with active transmission parameters under non-convex constraints. The surveyed results span information-theoretic regions, capacity-achieving schemes, interference management, robust design, and efficiency optimization.

  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: The section surveys RIS performance limits, joint transmit and passive beamforming, resource allocation, and mathematical tools used to solve these designs.The joint problems are motivated by the need to optimize RIS parameters and network resources together.
  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: Capacity achievement in RIS-aided SIMO systems with finite constellations requires joint information encoding across transmitted signals and RIS configurations.A practical layered-encoding and successive-cancellation strategy is reported to outperform conventional max-SNR transmission.
  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: Deploying an RIS improves NOMA capacity regions and OMA rate regions, with further enlargement from finer phase resolution and more reflection elements.The regions are characterized through Pareto-boundary optimization using the rate-profile technique.
  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: RIS beamforming can simultaneously enhance desired-signal strength and mitigate interference in multi-user scenarios.Semidefinite relaxation is used for passive beamforming, while successive-refinement algorithms achieve near-optimal performance with lower complexity under discrete phase shifts.
  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: User pairing matters because NOMA may perform worse than TDMA for symmetric deployments and rate requirements.This result shows that multiple-access performance depends on deployment and rate conditions, not only on the use of an RIS.
  • IV. RIS BEAMFORMING AND RESOURCE ALLOCATION: RISs can achieve better energy-efficiency performance than traditional active relay-assisted communication, but high channel error can degrade communication performance.Robust designs address bounded or statistical CSI errors through SDP-based transformations and related approximation procedures.

2) Approaches for passive beamforming design:

Passive beamforming design in RIS-assisted networks is shaped by the feasible reflection model and the coupling of transmit and passive beamforming variables. Existing methods trade solution quality, feasibility, and computational complexity when addressing the resulting non-convex optimization problem.

  • Feasible reflection models: RIS implementations are modeled through continuous or discrete amplitude and phase-shift constraints, with fixed-amplitude cases representing practical finite-resolution settings.Continuous amplitude and phase shifts provide theoretical upper bounds, whereas continuous amplitude and phase control is difficult to realize in practice.
  • Optimization challenges: The joint beamforming problem is generally non-convex because the transmit vector w and passive beamforming vector θ are coupled.The passive design additionally faces unit-modulus constraints and, in some implementations, a discrete feasible set.
  • Existing approaches: Alternating optimization decouples transmit and passive beamforming, making the transmit subproblem conventional while leaving passive design non-trivial.This approach exploits the tractability of transmit beamforming under a fixed passive beamforming vector.
  • Existing approaches: Semidefinite relaxation converts the unit-modulus problem into a convex SDP by dropping the rank-one constraint, but randomization can yield suboptimal or infeasible solutions.The resulting approximation may degrade performance and cannot guarantee convergence of the AO-based iterative algorithm.
  • Existing approaches: Quantization maps continuous phase solutions to nearby discrete values, yet low-resolution shifts can cause substantial performance loss and retain non-convex unit-modulus constraints.Continuous relaxation therefore does not eliminate all difficulties of the original passive beamforming problem.
  • Existing approaches: Iterative algorithms seek locally optimal or high-quality suboptimal passive beamforming solutions at acceptable computational complexity, creating a performance-versus-complexity tradeoff.The surveyed examples include successive refinement, alternating DC, conjugate-gradient, fixed-point, manifold-optimization, and sequential rank-one relaxation methods.

C. Resource Management in RIS-enhanced Networks

Resource management in RIS-enhanced networks jointly addresses association, subchannel assignment, power allocation, and passive beamforming. The main challenge is obtaining efficient solutions to large, coupled, and often combinatorial optimization problems.

  • Subchannel assignment: A common RIS reflection matrix across non-frequency-selective subchannels makes subchannel assignment difficult, motivating dynamic allocation of resource blocks and RIS phase shifts across user groups.Dynamic passive beamforming is used to coordinate these varying assignments.
  • Resource-management problems: Joint resource management can include subchannel assignment, power allocation, and passive beamforming in multi-channel downlink RIS-NOMA systems.These variables must be coordinated across users and RIS configurations.
  • User-RIS association: User-RIS association determines overall network performance, and infinitely large RISs can provide automatic interference cancellation that enables a greedy-search solution to the max-min SINR problem.The association problem is therefore central in multi-RIS multi-user systems.
  • Multi-cell management: Multi-cell optimization jointly considers user-BS association, user-RIS association, and subchannel assignment, while RIS deployment can improve desired power and mitigate interference for cell-edge users.The combined association and allocation structure becomes more sophisticated than in single-cell settings.
  • Matching and heuristic methods: Matching theory decomposes high-dimensional association into efficiently solvable two-dimensional subproblems and can achieve near-optimal performance in RIS-enhanced NOMA.Its preference lists must be updated as channel conditions fluctuate.
  • Matching and heuristic methods: Relaxation and heuristic methods reduce complexity but may remain non-convex, lose performance relative to the original integer problem, or produce unstable results sensitive to strategy design.Large-scale networks require low-complexity algorithms that balance performance and computational cost.

D. Discussions and Outlook

The outlook emphasizes practical RIS design beyond passive beamforming, including deployment, dynamic configuration, robust optimization, and machine learning. Key boundaries arise from path loss, channel coupling, limited CSI, mobility assumptions, and dynamic environments.

  • RIS deployment design: RIS deployment location must be optimized because the reflection link experiences greater path loss than the direct link.Jointly optimizing deployment, passive beamforming, and AP/BS resource allocation remains non-trivial.
  • RIS deployment design: Asymmetric RIS locations are preferable for NOMA, whereas symmetric locations are preferable for OMA in the reported deployment study.The preferred geometry depends on the multiple-access scheme.
  • RIS deployment design: LoS-oriented deployment can be ineffective in multi-user systems because low-rank, ill-conditioned channel matrices limit achievable capacity despite small path loss.Deployment must balance path loss against the non-LoS channel components.
  • Dynamic RIS configuration: Dynamic RIS configuration is practically valid because RIS adjustment can occur multiple times within one channel coherence duration.The paper gives a 220-microsecond adjustment example versus coherence blocks lasting tens of milliseconds.
  • Machine learning integration: Machine learning is surveyed for RIS challenges including channel estimation, beamforming, resource allocation, user mobility, and random wireless-channel fluctuations.The paper identifies ML-enabled architectures and research opportunities for dynamic RIS-enhanced networks.
  • Open challenges: Conventional RIS studies commonly assume static users, perfectly known environments, unlimited instantaneous CSI, and no learning from limited user feedback.These assumptions motivate ML methods for highly dynamic stochastic environments with resource-hungry feedback.
  • Machine learning integration: Deep learning can estimate RIS-related CSI from sampled channel knowledge, including compressive estimation with low training overhead and improved robustness or NMSE.The surveyed results also report improved BER and an unsupervised mechanism outperforming a conventional optimization approach.

C. Reinforcement Learning for RIS-enhanced Communication Systems

Reinforcement learning enables BS/RIS agents to learn control policies from environment states, historical experience, and user feedback. The surveyed applications jointly optimize RIS and transmission decisions across dynamic channel, demand, and deployment settings.

  • RL foundations: RL is organized into value-based, policy-based, and actor-critic algorithms, with DQN suited to RIS phase-shift design when phase shifts are discrete.The suitability follows from the discrete action structure of RIS phase control.
  • Joint beamforming: Unlike alternating optimization, RL can simultaneously design BS transmit beamforming and RIS passive beamforming through interaction with the environment.Reported applications target throughput, achievable rate, or secrecy rate under different system conditions.
  • Joint beamforming: DDPG jointly optimizes continuous transmit beamforming and RIS phase shifts with low complexity while using sum rate as an instant training reward.The cited application maximizes throughput in MISO systems.
  • Dynamic wireless settings: DRL has been applied to achievable-rate maximization, secrecy-rate maximization, imperfect-CSI settings, and time-varying user QoS requirements.Reported mechanisms include direct optimization from sampled channel knowledge, return-distribution modeling, and prioritized experience replay.
  • ML-enabled architecture: ML-empowered RIS networks collect and process user information to predict behavior and requirements, then adapt phase shifts, resource allocation, and interference cancellation.RISs can learn from both the environment and user feedback to adapt rapidly to dynamic conditions.
  • DRL deployment control: An RL model can jointly control RIS deployment and phase shifts alongside user power allocation, with the BS selecting actions from states containing RIS, user, and power information.Actions include changing RIS positions and phase shifts and varying allocated power according to Q-values and rewards.
  • DRL deployment control: The reported deployment study finds an optimal RIS position for energy efficiency and improved performance over random or alternative placement strategies.The comparison distinguishes the RIS-barycenter and RIS-random deployment strategies.

E. Other ML Techniques for RIS-enhanced Communication Systems

The section surveys supervised, unsupervised, and federated learning approaches for RIS-enhanced wireless networks, emphasizing applications to dynamic environments and multi-RIS deployment. It also identifies unresolved challenges in modeling joint discrete–continuous decisions and coordinating multiple RISs.

  • Supervised learning: Supervised learning methods can address RIS-network problems when sufficient training data are available, offering low complexity and fast convergence.Examples include regression, decision trees, random forests, KNN, SVMs, and Bayesian classification for tasks such as spectrum sensing, QoE prediction, channel selection, and networking association.
  • Unsupervised learning: Unsupervised methods do not rely on prior knowledge and can support RIS tasks including deployment, user clustering, state detection, data aggregation, and interference cancellation.The surveyed examples include K-means, expectation-maximization, PCA, and ICA.
  • Federated learning: Federated learning enables decentralized multi-RIS deployment and design by exchanging local model parameters instead of raw training data.Each RIS can act as a distributed learner, preserving the inaccessibility of private data while learning a shared policy through an aggregating unit.
  • Reinforcement learning: Reinforcement learning can adapt RIS control and deployment policies from environmental feedback while incorporating farsighted system evolution.This orientation targets dynamic or uncertain wireless environments and can account for changing positions and phase shifts.
  • Open challenges: Current ML models commonly represent either discrete or continuous state spaces, whereas RIS networks require joint discrete–continuous design.The section also highlights challenges in deep-learning layer design, RL state–action construction, reward design, and cooperation among multiple RISs.
  • Integration opportunities: RIS-enhanced systems have been reported to obtain tuned channel gains, improved QoS, enhanced coverage, and reduced energy dissipation across diverse wireless applications.The surveyed integration directions include NOMA, physical-layer security, SWIPT, UAV-enabled networks, and autonomous driving.

B. PLS and RIS

RISs are surveyed as tools for improving physical-layer security, SWIPT, and UAV-enabled communications. The section reports security and energy-transfer gains while emphasizing challenges from imperfect eavesdropper CSI, EM modeling, and UAV energy constraints.

  • PLS and RIS: RIS-assisted secrecy designs have reported significant secrecy improvements in MISOSE and multi-user MISOME settings.These studies include continuous and discrete phase shifts, imperfect CSI, artificial-noise covariance, and alternating-optimization methods.
  • PLS and RIS: RISs can enhance secrecy by destructively combining direct and reflected signals at eavesdroppers while strengthening intended-user reception.The surveyed designs jointly optimize transmit beamforming and RIS phase shifts, sometimes with artificial noise.
  • PLS and RIS: Joint transmit and passive beamforming is difficult because obtaining AP–eavesdropper and RIS–eavesdropper CSI is challenging.Eavesdroppers may remain silent to conceal their positions, motivating robust designs under imperfect CSI.
  • PLS and RIS: Deploying RISs can increase information-leakage risk because eavesdroppers may receive both direct and RIS-reflected signals.The risk can worsen with multiple cooperative eavesdroppers; protected zones and careful RIS deployment are proposed for further investigation.
  • SWIPT and RIS: RIS-assisted SWIPT studies report improved energy-harvesting efficiency, an enlarged wireless power-transfer range, and fewer required energy beams.The surveyed work jointly optimizes transmit and passive beamforming or minimizes transmit power under QoS constraints.
  • UAV-enabled networks: UAV-RIS integration can form virtual LoS links and improve average data rate, downlink LoS probability, coverage, service quality, and UAV endurance.RIS phase-shift control can reduce the need for UAV movement, lowering total energy consumption.

E. Autonomous Driving/Connected Vehicles and RIS

The section examines RISs for autonomous-driving and connected-vehicle networks, where urban channel complexity makes reliable real-time service difficult. It also summarizes broader RIS research coverage and identifies open challenges in CSI acquisition, coupled optimization, and dynamic environments.

  • Autonomous driving and connected vehicles: Reliable real-time V2I service remains challenging because urban terrain creates complex communication channels, while autonomous-driving safety requires dependable service at each timeslot.The section identifies reliability in RIS-enhanced autonomous driving as an open problem.
  • Autonomous driving and connected vehicles: RISs can improve vehicular-network performance and establish virtual LoS connections that enhance reliability between base stations, roadside units, and autonomous vehicles.RISs may be installed on surfaces such as building facades, highway poles, advertising panels, vehicle windows, and clothing.
  • Reported benefits: Reported RIS-network benefits include tuned channel gains, improved QoS, enhanced coverage range, and reduced energy dissipation.These benefits are discussed across diverse wireless communication applications.
  • Open challenges: Dynamic RIS phase-shift control couples network and beamforming designs, creating challenges for emerging application areas.The paper characterizes RIS-enhanced-network research as being at an early stage with ample opportunities for further contributions.
  • Paper scope: The paper surveys RIS operating principles, multi-antenna performance evaluation, beamforming and resource allocation, machine learning, and integration with key 6G technologies.The integration discussion covers NOMA, physical-layer security, SWIPT, UAV-terrestrial networks, and autonomous or connected vehicles.
  • Open challenges: Most current studies assume perfect CSI, although obtaining and exchanging CSI in passive RIS systems requires non-negligible training overhead.Deep learning is identified as a possible way to exploit CSI structures beyond linear correlations.
  • Open challenges: RIS networks involve rapidly fluctuating topologies, vulnerable links, heterogeneous mobility, and multi-objective optimization with trade-offs among delay, throughput, BER, and power.The paper points to near-real-time ML-aided Pareto optimization as a possible direction for high-dynamic adaptation.
Loading 2007.03435v3…