PaperlayerView exact revision on arXiv

Source-linked AI summary

A Survey on Channel Estimation and Practical Passive Beamforming Design for Intelligent Reflecting Surface Aided Wireless Communications

Beixiong Zheng, Changsheng You, Weidong Mei, Rui Zhang

arXiv:2110.01292v3cs.ITcs.ETeess.SP

TL;DR

IRS is proposed as a low-power, programmable means of reshaping wireless propagation, but practical integration requires reliable channel acquisition, reflection design, and hardware-aware modeling. This paper surveys research addressing these issues across channel models, system setups, CSI regimes, architectures, and applications. It also identifies trade-offs among estimation overhead, complexity, applicability, and beamforming performance, while discussing limitations in specific tracking and optimization approaches.

  • Problem

    IRS promises controllable wireless propagation, but its practical integration is challenged by channel estimation, passive beamforming under imperfect CSI, and hardware constraints or imperfections.

  • Method

    The paper provides a comprehensive survey of channel-estimation methods, passive reflection optimization under multiple CSI regimes, hardware effects, and emerging IRS architectures and applications.

  • Results

    The survey finds that IRS design choices depend on channel and signal models, CSI availability, and hardware constraints, with methods exhibiting different overhead, complexity, applicability, and performance trade-offs.

  • Takeaways & Limitations

    Practical IRS research should evaluate channel acquisition and passive beamforming jointly with realistic CSI conditions, hardware impairments, deployment settings, and emerging architectures.

Abstract

from arXiv · show

Intelligent reflecting surface (IRS) has emerged as a key enabling technology to realize smart and reconfigurable radio environment for wireless communications, by digitally controlling the signal reflection via a large number of passive reflecting elements in real-time. Different from conventional wireless communication techniques that only adapt to but have no or limited control over dynamic wireless channels, IRS provides a new and cost-effective means to combat the wireless channel impairments in a proactive manner. However, despite its great potential, IRS faces new and unique challenges in its efficient integration into wireless communication systems, especially its channel estimation and passive beamforming design under various practical hardware constraints. In this paper, we provide a comprehensive survey on the up-to-date research in IRS-aided wireless communications, with an emphasis on the promising solutions to tackle practical design issues. Furthermore, we discuss new and emerging IRS architectures and applications as well as their practical design problems to motivate future research.

I. INTRODUCTION

6G targets substantially higher demands for connectivity, rates, efficiency, reliability, latency, positioning, and mobility than 5G, while existing technologies face cost, energy, blockage, and propagation challenges. IRS offers a low-power, programmable way to reshape wireless propagation, but its deployment motivates practical research across channel estimation, passive beamforming, hardware constraints, and emerging architectures.

  • 6G requirements: 6G targets 1 Tbps peak data rates, 10–100 times 5G energy efficiency, and 5 times 5G spectral efficiency.
  • 6G requirements: 6G also targets approximately 10^7 devices/km^2, reliability of at least 99.99999%, air-interface latency of at most 100 µs, centimeter-level 3D positioning, and mobility up to 1,000 km/h.
  • Motivation for IRS: Massive MIMO, mmWave communication, and network densification improve 5G performance but face high energy consumption, hardware cost, blockage, absorption loss, and scalability difficulties.
  • Motivation for IRS: IRS is a digitally controlled metasurface whose many passive elements tune incident-signal phase shifts and/or amplitudes with ultra-low power to reshape wireless propagation.
  • Research scope: IRS research spans channel estimation, passive beamforming under different CSI conditions, discrete reflection, hardware imperfections, and applications across smart, terrestrial, and non-terrestrial deployments.

B. Main Practical Issues in IRS-aided Wireless Communications

IRS integration faces practical challenges in acquiring channel state information, designing passive reflections under imperfect information, and modeling hardware constraints and impairments. These issues vary with deployment and system configuration, motivating dedicated design methods and broader architectural studies.

  • Emerging architectures and applications: Emerging IRS architectures and applications introduce additional design problems that require dedicated investigation.
  • IRS channel estimation/acquisition: Accurate CSI acquisition is indispensable for controlling the IRS-aided propagation environment but is difficult because passive IRS elements lack active baseband processing and transmit/receive capability.
  • IRS channel estimation/acquisition: IRS channel-estimation requirements differ across single- and multi-user, single- and multi-IRS, antenna configurations, mobility conditions, and narrowband or broadband systems.
  • Practical passive beamforming/reflection design: Passive beamforming must be jointly designed with active transceivers under imperfect instantaneous or statistical CSI to achieve robust communication performance despite CSI errors.
  • Hardware constraints and imperfections: Practical IRS hardware issues include discrete phase shifts and amplitudes, phase-dependent amplitude, mutual coupling, and other imperfections that can limit signal reflection capability.

C. Objective and Organization

The paper surveys practical IRS channel estimation and passive beamforming design, including hardware constraints, emerging architectures, and application challenges. Its channel-estimation review is organized by IRS architecture, system setup, and signal-processing method.

  • The survey emphasizes practical IRS channel estimation and passive beamforming/reflection optimization under hardware constraints and imperfections.
  • It discusses emerging IRS architectures and applications, together with their practical design challenges, to motivate future research.
  • Accurate CSI is essential for passive beamforming, but estimating it is difficult because IRS elements generally lack transmitting and receiving capabilities.
  • IRS channel estimation: IRS channel-estimation studies are classified by architecture, system setup, and signal-processing method.
  • IRS channel estimation: The review covers semi-passive and fully-passive architectures, single- and multi-user systems, broadband systems, double/multi-IRS systems, and separate, cascaded, and combined estimation.
  • IRS channel estimation: Signal-processing categories include classical channel estimation, compressed sensing, matrix factorization, and deep learning.

1) Separate Channel Estimation with Semi-Passive IRS:

Semi-passive IRSs use a small number of sensing devices to separately estimate links from the BS and users, whereas fully-passive IRSs estimate only cascaded channels at an endpoint. These approaches trade hardware requirements, applicability, and training overhead.

  • Separate Channel Estimation with Semi-Passive IRS: Semi-passive IRSs separately estimate BS-IRS and user-IRS channels using dedicated low-power sensing devices and transmitted pilots.The sensing devices may have low-resolution ADCs and are substantially fewer than the IRS reflecting elements.
  • Separate Channel Estimation with Semi-Passive IRS: Separate estimation works naturally with TDD reciprocity but may not apply to FDD unless the IRS has active sensors that transmit and receive pilots.Such active sensors increase hardware cost and power consumption.
  • Separate Channel Estimation with Semi-Passive IRS: Because measurements come from few low-cost sensors, signal processing must interpolate or extrapolate full reflecting-element CSI using low-rank, sparse, or spatially correlated channel structure.Quantization error, ambient noise, and circuit nonlinearity can distort the measurements used for CSI construction.
  • Separate Channel Estimation with Semi-Passive IRS: Representative separate-estimation methods use L-shaped sensing arrays, randomly distributed sensors, compressed sensing, deep learning, and angular-domain sparsity.
  • Cascaded Channel Estimation with Fully-Passive IRS: Fully-passive IRSs cannot generally estimate BS-IRS and user-IRS channels separately; instead, an endpoint estimates their cascaded channel.Cascaded estimation applies to both TDD and FDD systems, with reciprocity in TDD and additional feedback in FDD.
  • Cascaded Channel Estimation with Fully-Passive IRS: Cascaded estimation can sequentially estimate each IRS-element channel by designing pilot sequences and IRS reflection patterns, but the cascaded channel has more coefficients and higher training overhead.

3) Comparison/Combination of Separate and Cascaded Channel Estimation:

Separate and cascaded estimation have complementary strengths, motivating hybrid designs. A hybrid method could estimate the quasi-static, high-dimensional BS-IRS channel separately and track dynamic, low-dimensional IRS-user channels through cascaded estimation.

  • Comparison: Separate and cascaded channel estimation have distinct advantages and disadvantages, including trade-offs involving accuracy, hardware cost, energy consumption, and CSI errors.
  • Combination: The BS-IRS channel is generally high-dimensional and quasi-static, whereas IRS-user channels are more dynamic and low-dimensional.
  • Combination: A hybrid design can estimate the quasi-static BS-IRS channel at IRS sensing devices and track dynamic IRS-user channels at the BS using cascaded estimation.
  • Combination: This combined approach has the potential to reduce real-time training overhead and improve applicability to both TDD and FDD systems.Its effective design and practical realization remain open problems.

B. Channel Estimation for Different IRS System Setups

IRS channel estimation must be tailored to system setup because training requirements vary across users, antennas, mobility, and bandwidth. Single-user and multi-user designs therefore address different overhead, coupling, and CSI-feedback challenges.

  • Single-User System with Single IRS: Training overhead for resolving the full cascaded channel generally scales with the number of IRS elements N, causing potentially long estimation delays.Reducing this overhead is a central problem in single-user IRS channel estimation.
  • Single-User System with Single IRS: Single-user IRS channel estimation is commonly classified into SISO, MISO, and MIMO setups according to the downlink antenna configuration.Early schemes estimate direct or cascaded CSI independently, whereas MIMO requires joint processing because pilots are coupled across transmit and receive antennas.
  • Single-User System with Single IRS: High-mobility IRS systems require protocols that track direct and cascaded CSI efficiently, while doubly-selective channels remain challenging.Existing protocols cover SISO, MISO, and MIMO high-mobility settings, but the time-varying multipath case needs further study.
  • Multi-User System with Single IRS: Applying single-user estimation successively to multiple users makes training time scale with the number of users K and can become unaffordable when K is large.Multi-user designs must jointly tailor pilot sequences, IRS reflection patterns, and estimation algorithms to improve training efficiency.
  • Multi-User System with Single IRS: Exploiting the common IRS-BS channel lets reference-user and prior-knowledge methods estimate other users’ cascaded channels with reduced training overhead.Some designs first estimate one typical user or the quasi-static common IRS-BS channel, then use it to estimate user-specific channels in real time.
  • Multi-User System with Single IRS: Fully-passive multi-user downlink estimation still requires users to feed back direct and cascaded CSI to a central processor, creating high CSI-feedback overhead.Semi-passive IRS alternatives estimate the common IRS-BS and user-IRS channels in parallel at dedicated sensing devices.

3) Double/Multi-IRS System:

Double- and multi-IRS systems introduce inter-IRS reflections and multiple reflection orders, coupling the effective links and increasing estimation complexity. Broadband operation adds multipath convolution and frequency-flat reflection constraints, leaving several extensions open.

  • Double/Multi-IRS System: Inter-IRS reflection creates higher-order passive beamforming gains but introduces double-reflection, single-reflection, and direct links into the effective channel.The effective channel therefore contains more coupled components than single-IRS systems.
  • Double/Multi-IRS System: Coupled links with different reflection orders entail more channel coefficients, making single-IRS channel-estimation techniques inapplicable to double- and multi-IRS systems.Dedicated estimation methods are consequently required for these architectures.
  • Double/Multi-IRS System: Semi-passive and fully-passive double-IRS schemes have been studied for single-user SISO systems under assumptions such as LoS inter-IRS channels or blocked direct and single-reflection links.These assumptions simplify estimation but restrict the considered propagation settings.
  • Double/Multi-IRS System: ON/OFF and always-ON training reflection schemes were proposed to acquire cascaded CSI for single- and double-reflection links in double-IRS multi-user MISO systems.The always-ON design addresses error propagation and reflection-power loss associated with ON/OFF control.
  • Broadband System with Single IRS: Broadband IRS channels are convolutions of multi-tap user-IRS and IRS-BS channels, requiring substantially more cascaded coefficients than narrowband channels.Frequency-flat IRS reflection also affects every OFDM subcarrier identically, limiting per-subcarrier design flexibility.
  • Double/Multi-IRS System: Existing double- and multi-IRS estimation studies mainly address narrowband flat-fading channels, leaving broadband frequency-selective extensions open.The survey identifies this direction as requiring further research.

C. Signal Processing Methods for IRS Channel Estimation

IRS channel estimation uses classical LS/LMMSE methods alongside training-pattern, grouping, and protocol designs to reduce complexity and overhead. These methods trade estimation requirements, implementation simplicity, reflection power, and passive-beamforming performance.

  • Classical Channel Estimation: LS/LMMSE estimation is widely used because of low implementation complexity, but it generally requires at least as many observations as unknown channel parameters.LMMSE additionally exploits second-order channel and noise statistics to minimize overall MSE.
  • Classical Channel Estimation: Unique estimation requires a full-column-rank observation matrix, which in the stated model requires T ≥ MuN pilot symbols.The observation matrix depends on the IRS training reflection pattern and pilot design.
  • Classical Channel Estimation: ON/OFF training estimates the direct channel with all IRS elements off and cascaded channels by activating elements sequentially.Although simple, it suffers reflection-power loss and direct-channel interference that degrade estimation accuracy.
  • Classical Channel Estimation: Element-grouping estimates one equivalent aggregated channel per subsurface, reducing training overhead by a factor of B.Adjusting subsurface size B provides a trade-off between training overhead, design complexity, and passive beamforming performance without assuming a specific channel model.
  • Classical Channel Estimation: Reference-user protocols exploit lower-dimensional scaled representations of other users’ cascaded channels to improve multi-user training efficiency.These protocols have been proposed for both narrowband and broadband systems.
  • Classical Channel Estimation: Hierarchical training progressively resolves cascaded CSI and successively refines passive beamforming for dynamic IRS environments.The design targets reduced training delay before data transmission.

2) Compressed Sensing:

At mmWave and THz frequencies, limited scattering paths make IRS-associated channels sparse and low-rank, enabling compressed-sensing-based channel estimation. Existing methods exploit angular sparsity and shared sparsity structures to reduce estimation complexity and training overhead.

  • 2) Compressed Sensing:: Limited scattering paths at mmWave and THz frequencies produce strong spatial/angular sparsity and low rank in IRS-associated channels.Severe path loss and occasional blockage restrict the number of scattering paths between the IRS and BS/user.
  • 2) Compressed Sensing:: Geometric channel models represent IRS-BS and user-IRS channels with over-complete array-response dictionaries and sparse path-gain vectors.The sparse vectors contain dG and dk spatial paths, respectively, with dG ≪ LBLR and dk ≪ LRLu.
  • 2) Compressed Sensing:: Cascaded channel estimation can be formulated as a sparse signal recovery problem and solved using compressed-sensing approaches.The cascaded channel inherits strong sparsity from the IRS-associated channels when represented appropriately.
  • 2) Compressed Sensing:: A common IRS-BS channel gives cascaded channel matrices for all users a shared row-column-block sparsity structure.This shared structure is exploited in compressed-sensing-based estimation methods.
  • 2) Compressed Sensing:: OMP provides a low-complexity approach by selecting the best-matching projections of received measurements in the beamspace or angular domain.OMP has been applied to cascaded channel estimation, alongside other compressed-sensing algorithms.

3) Matrix Factorization/Decomposition:

Matrix factorization and decomposition methods address the bilinear nature of fully passive IRS channel estimation by separating cascaded observations into component channels. However, the recovered IRS-BS and user-IRS channels have a scaling ambiguity that generally does not prevent passive beamforming design.

  • 3) Matrix Factorization/Decomposition:: Fully passive IRS channel estimation is a bilinear problem because the cascaded user-IRS-BS channel is the product of two channel matrices.Compared with conventional linear channel estimation, the bilinear problem is generally harder because of the high-dimensional cascaded channel.
  • 3) Matrix Factorization/Decomposition:: Matrix factorization or decomposition seeks to resolve the IRS-BS and user-IRS channel matrices from the cascaded channel.The approach directly targets the component channels underlying the received cascaded observations.
  • 3) Matrix Factorization/Decomposition:: The factorization has a scaling ambiguity: for any invertible diagonal matrix Λ, G and Hk can be transformed without changing the cascaded channel.Consequently, the two component channels cannot generally be uniquely resolved separately from the received signal model.
  • 3) Matrix Factorization/Decomposition:: Parallel factor tensor modeling unfolds the 3D cascaded MIMO channel into 2D user-IRS and IRS-BS channels for efficient estimation.The same methods and results can be adapted to downlink systems by swapping the roles of the multi-antenna BS and multiple users.
  • 3) Matrix Factorization/Decomposition:: Signal-processing methods for IRS channel estimation differ in applicability and complexity, with LS/LMMSE broadly applicable and matrix factorization among methods requiring higher complexity.The appropriate method depends strongly on the underlying IRS and signal models.

III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI

Practical IRS passive beamforming must account for imperfect, statistical, hybrid, or unavailable instantaneous CSI because extra IRS-associated channels increase estimation burden. The section organizes robust CSI-based design, beam training, and deep-learning approaches around these CSI scenarios.

  • III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI: Perfect CSI is difficult to obtain because of channel aging, limited training or feedback, noise, interference, and extra IRS-associated channels.These factors complicate joint optimization of IRS reflection and BS active beamforming.
  • III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI: The section covers imperfect CSI, statistical or hybrid CSI, beam training and channel tracking, and no-explicit-CSI methods.Its organization is summarized in Fig. 9 and includes deep-learning-based reflection design.
  • III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI: Deterministic CSI-error models bound error norms, whereas stochastic models treat CSI errors as random variables with specified distributions and variances.For fully passive IRS, the cascaded channel is modeled as L = L̃ + EL; for individually estimated channels, each channel is expressed as an estimate plus error.
  • III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI: Deterministic robust beamforming targets worst-case utility over all CSI within an error bound, while stochastic design targets non-outage performance for random CSI errors.Both formulations must handle infinitely many possible CSI realizations in worst-case or probabilistic form.
  • III. IRS PASSIVE BEAMFORMING DESIGN UNDER PRACTICAL CSI: Larger deterministic error bounds or stochastic error variances generally reduce worst-case or non-outage performance, respectively.The bound and variance quantify CSI uncertainty in the two error models.

1) Deterministic Model:

Deterministic-model robust IRS beamforming has been widely studied using tractable optimization transformations across diverse communication scenarios. By contrast, stochastic-model design is less developed because its probabilistic constraints are more difficult, while practical systems also face substantial estimation overhead.

  • 1) Deterministic Model:: Deep reinforcement learning can jointly optimize IRS phase shifts and reflection amplitudes while minimizing BS transmit power under worst-case reflected and harvested signal-power requirements.The approach was proposed for a single-user MISO system aided by an energy-harvesting IRS.
  • 1) Deterministic Model:: The S-procedure transforms deterministic worst-case objectives or constraints into tractable linear matrix inequalities, typically combined with AO and SDR.These techniques address robust design and IRS unit-modulus constraints.
  • 1) Deterministic Model:: Deterministic robust beamforming has been studied for broadcast, MIMO-OFDMA THz, secure multi-IRS, NOMA, UAV-ground, and secrecy-energy-efficiency systems.The reviewed works jointly optimize active and passive beamforming with additional variables such as trajectories or jamming signals in some settings.
  • 1) Deterministic Model:: Despite CSI errors, robust passive beamforming can dramatically improve system performance over traditional systems without IRS and non-robust designs.This conclusion summarizes the deterministic-model literature across assorted scenarios.
  • 1) Deterministic Model:: Stochastic-model robust beamforming is less studied because probabilistic constraints are more challenging, motivating convex approximations and new optimization techniques.Existing approaches include constrained stochastic SCA, sphere bounding, and Bernstein-type inequalities, often with AO and SDR.
  • 1) Deterministic Model:: Statistical CSI reduces training time by using slowly varying channel statistics, while hybrid CSI estimates only selected channels in real time.The reduction in overhead comes at the cost of real-time performance when instantaneous CSI is unavailable.

1) Statistical CSI:

Statistical and hybrid CSI support long-term or two-timescale IRS beamforming designs that trade channel-information overhead against optimization complexity. Existing studies span single-user, multiuser, interference, security, and resource-allocation settings, while broader system configurations remain underexplored.

  • Statistical CSI: Statistical CSI enables long-term IRS beamforming by optimizing tractable bounds, deterministic equivalents, or statistical objectives such as coverage probability and ergodic capacity.Studies cover correlated Rayleigh, Rician, double-scattering, and second-moment-based channel models.
  • Statistical CSI: Some statistical-CSI studies report that IRS assistance may not outperform conventional systems in ergodic-rate performance under specific conditions or setups.
  • Open issues: Existing statistical-CSI studies have mostly considered basic single-user systems, leaving more general system setups for future exploration.
  • Hybrid CSI: Hybrid CSI designs combine instantaneous and statistical channel information, including joint optimization for multiple-access, broadcast, massive-MIMO, and multicell systems.Two-timescale designs commonly use short-term active beamforming with long-term IRS passive beamforming or resource allocation.
  • Hybrid CSI: Compared with statistical- or instantaneous-CSI-only designs, hybrid CSI flexibly balances performance and overhead but generally requires approximations and higher optimization complexity.In two-timescale beamforming, short-term active and long-term passive beamforming can be coupled.

2) Deep-learning Based Reflection Design:

Reflection design without explicit CSI uses side information, deep learning, random reflections, or quantized practical controls instead of full channel knowledge. These approaches reduce estimation requirements or address hardware constraints, but introduce trade-offs in performance, overhead, complexity, and unresolved design limits.

  • Deep-learning Based Reflection Design: Separating IRS channel estimation from passive beamforming can cause error propagation, fail to align estimation error with beamforming performance, and require high computational complexity.
  • Deep-learning Based Reflection Design: Deep learning can directly map training data, user locations, or received pilots to IRS passive beamforming without explicit CSI.Location-based design constructs a fingerprint database of optimal beamforming at prescribed positions, while pilot-based design learns channel and beamforming features jointly.
  • Other approaches: Random IRS reflections avoid CSI and heavy channel-estimation overhead but sacrifice full passive beamforming gain.In one multicast study, random reflections achieved lower outage probability than a CSI-based scheme requiring large estimation overhead.
  • Other approaches: Without explicit CSI, received SNRs, training signals, and user location provide side information for inferring IRS beam directions.
  • Open issues: Efficiently combining available side information and characterizing performance limits for no-explicit-CSI reflection designs remain open problems.
  • Practical reflection controls: Discrete finite phase-shift and amplitude levels reduce implementation cost but make channel estimation and passive beamforming combinatorial.The number of levels is controlled by bits through Kβ = 2^bβ and Kθ = 2^bθ.

B. Reflection with Phase-dependent Amplitude

Practical IRS reflection design must account for phase-dependent amplitude, mutual coupling, hardware impairments, and alternative active or relaying architectures. These effects alter optimization, channel estimation, achievable rates, coverage, and deployment complexity.

  • Phase-dependent amplitude: Practical IRS elements can have reflection amplitudes that vary nonlinearly with phase shift, invalidating independent amplitude-phase control assumptions.The amplitude typically reaches a minimum at zero phase shift and then increases toward an asymptotic value.
  • Phase-dependent amplitude: Element-wise BCD methods optimize phase shifts while accounting for their effect on phase-dependent amplitudes in single- and multi-user MIMO-OFDM systems.
  • Mutual coupling: Mutual coupling between closely spaced reflecting elements creates coupled reflection coefficients and complicates element-level channel estimation and beamforming.Element grouping estimates subsurface-level channels, reducing training overhead while partially circumventing coupling effects.
  • Hardware impairments: Transceiver and IRS impairments include phase noise, RF distortions, analog imperfections, quantization errors, and low-resolution ADC effects.Existing channel-estimation methods model some distortions statistically, but their combined effects remain insufficiently characterized.
  • Hardware impairments: IRS phase noise can degrade achievable rate more severely as the number of reflecting elements increases, although phase-shift errors may preserve square SNR and linear diversity scaling orders.
  • New architectures: Active, relaying, and passive IRS architectures present different trade-offs in energy and spectral efficiency, complexity, coverage, and serving range.Relaying IRS architectures seek complementary benefits by using the IRS controller to actively relay information.

3) Intelligent Refracting/Transmitting Surface (IRS/ITS):

IRS/ITS architectures and applications extend beyond reflection to refraction, wireless power transfer, reflection modulation, UAV communications, physical-layer security, cognitive radio, and RF sensing. Their practical deployment requires addressing channel acquisition, mobility, reciprocity, interference, and coverage constraints.

  • Intelligent Refracting/Transmitting Surface (IRS/ITS): Refraction-type metasurfaces can serve transmitters and receivers on opposite sides, but incur signal penetration loss and raise channel-reciprocity questions.
  • Applications: IRS-assisted wireless power transfer can establish line-of-sight paths and exploit large apertures, but accurate IRS-associated CSI remains crucial, especially in frequency-selective channels.
  • UAV communications: High UAV mobility makes UAV–IRS channels more dynamic and demands tracking within short channel coherence times; high-altitude UAVs can also cause pilot contamination.
  • Physical-layer security: IRS-based physical-layer security requires robust passive beamforming under statistical cascaded-CSI errors, with a trade-off between secrecy performance and CSI accuracy.
  • Cognitive radio: Cognitive-radio applications require CSI associated with primary users, which is challenging when dedicated feedback channels are limited or unavailable.
  • Survey scope: The survey organizes practical IRS research around channel estimation, reflection design, hardware constraints, emerging architectures, and application-specific design challenges.
Loading 2110.01292v3…