PaperlayerView exact revision on arXiv

Source-linked AI summary

Next Generation Advanced Transceiver Technologies for 6G and Beyond

Changsheng You, Yunlong Cai, Yuanwei Liu, Marco Di Renzo, Tolga M. Duman, Aylin Yener, A. Lee Swindlehurst

arXiv:2403.16458v3cs.ITeess.SP

TL;DR

The paper examines advanced transceiver design across near-field communications, RISs, HMIMO, and semantic-aware systems. It synthesizes recent progress, design challenges, and future research directions, including the need for tailored CSI estimation under near-field and higher-frequency conditions.

  • Problem

    EM information theory highlights limits of conventional channel abstractions, while semantic-aware systems still require principled characterization of semantic information for transmission-rate analysis.

  • Method

    The paper provides a comprehensive overview of near-field channel models, RISs, flexible antennas, HMIMO, semantic-aware transceivers, and other promising NGAT technologies.

  • Results

    The survey identifies beamforming, beam training, channel estimation, mutual coupling, EM information theory, and semantic information characterization as central NGAT design issues and research directions.

  • Takeaways & Limitations

    Future NGAT research must address physical-layer coupling and semantic-aware design challenges across emerging transceiver technologies.

  • Takeaways & Limitations

    Near-field communications and higher frequency bands make CSI estimation more challenging, increasing overhead and complexity and motivating tailored estimation schemes.

Abstract

from arXiv · show

To accommodate new applications such as extended reality, fully autonomous vehicular networks and the metaverse, next generation wireless networks are going to be subject to much more stringent performance requirements than the fifth-generation (5G) in terms of data rates, reliability, latency, and connectivity. It is thus necessary to develop next generation advanced transceiver (NGAT) technologies for efficient signal transmission and reception. In this tutorial, we explore the evolution of NGAT from three different perspectives. Specifically, we first provide an overview of new-field NGAT technology, which shifts from conventional far-field channel models to new near-field channel models. Then, three new-form NGAT technologies and their design challenges are presented, including reconfigurable intelligent surfaces, flexible antennas, and holographic multi-input multi-output (MIMO) systems. Subsequently, we discuss recent advances in semantic-aware NGAT technologies, which can utilize new metrics for advanced transceiver designs. Finally, we point out other promising transceiver technologies for future research.

I. INTRODUCTION

6G applications require substantially stronger performance than 5G, motivating NGAT designs across near-field channels, new-form antennas, and semantic-aware communications. The paper surveys these technologies, their design challenges, and open research directions.

  • 6G Requirements: 6G targets up to 1 Tbps peak data rate, 100 bits/s/Hz spectral efficiency, 0.1∼1 ms latency, and centimeter-level positioning accuracy.It also targets network reliability of 1-10−5∼1-10−7 and up to 100 times higher energy efficiency than 5G.
  • 6G Use Cases: Emerging 6G scenarios extend 5G eMBB, mMTC, and URLLC toward immersive, massive, and HRLLC communications for XR, metaverse, transportation, and other applications.Additional use cases include ubiquitous connectivity, ISAC, and integrated AI-oriented communications and computing support.
  • Motivation: Existing 5G technologies such as massive MIMO and mmWave communications cannot solely satisfy 6G requirements, motivating NGAT for efficient signal transmission and reception.NGAT spans transceiver architecture, hardware, modulation and channel coding, and waveforms.
  • New-Form NGAT: RISs sculpt radio propagation with low-cost metamaterial elements, FAs dynamically adjust antenna positions, and HMIMO uses densely spaced continuous apertures for electromagnetic-field control.These technologies respectively address virtual LoS links, spatial diversity and multiplexing, and flexible electromagnetic-field control.
  • Semantic-Aware NGAT: Semantic communication incorporates semantics into wireless-system design to handle massive multimodal data for integrated physical and cyber worlds.The paper treats semantic-aware NGAT as a distinct direction alongside near-field and new-form technologies.
  • Paper Scope: The paper organizes NGAT into near-field communications, new-form technologies based on RISs, FAs, and HMIMO, and semantic-aware transceiver designs.It introduces hardware architectures, system models, design challenges, current research, and open problems for these technologies.

II. NEW FIELD: NEAR-FIELD TRANSCEIVERS

Near-field XL-MIMO replaces far-field planar-wave assumptions with spherical-wave channel models and enables beam focusing by location. The section surveys field regions, channel models, and resulting transceiver-design issues.

  • Near-field Region: XL-MIMO fields comprise reactive near-field, radiative near-field, and far-field regions, distinguished by propagation behavior and distance from the array.The discussion mainly focuses on the radiative near-field because the reactive region is very small in practice.
  • Near-field Region: The classical Rayleigh distance is rRayl = 2D^2/λ; for a 0.5 m aperture at 30 GHz, it is about 50 m.Direction-dependent, effective, uniform-power, and Björnson distances provide additional criteria based on phase, coherence, amplitude, or array gain.
  • Near-field Applications: Near-field spherical wavefronts enable beam focusing on specific locations or regions rather than steering energy only toward angles.This can serve users at the same angle but different distances with small or negligible inter-user interference and improve wireless-power-transfer efficiency.
  • Channel Models: Near-field channel models account for spherical wavefronts and spatial non-stationarity, using deterministic LoS or multipath formulations at high frequencies.The deterministic models use near-field array responses for LoS and multipath channels with multiple transmit and receive antennas.
  • Channel Models: The USW model assumes negligible amplitude variation but nonlinear phase variation, whereas the NUSW model includes non-negligible amplitude variation and nonlinear phase variation.The NUSW model should be used for asymptotic SNR analysis as the antenna count grows large.

2) Stochastic Channel Model for Spherical Wavefronts:

The stochastic near-field channel model incorporates spherical wavefronts and distance-dependent path loss through covariance-based formulations. These models support correlation and performance analysis in rich-scattering environments.

  • Stochastic channel modeling: Stochastic near-field channels in low-frequency rich-scattering environments can be characterized using channel covariance matrices.This facilitates analysis using SINR and ergodic capacity metrics.
  • Stochastic channel modeling: The near-field stochastic channel combines spatial correlation with antenna-dependent large-scale path loss under the non-uniform spherical-wave model.The path-loss vector depends on each antenna's link distance to the user.
  • Stochastic channel modeling: Near-field spatial correlation depends on the power location spectrum, whose scatterer distribution is determined jointly by scatterer angles and distances.The covariance expression accounts for distances from scatterers to the array reference and individual antennas.

3) Near-field Spatial Non-stationarity:

Near-field spatial non-stationarity arises from spherical-wave amplitude variation and visibility regions, requiring VR-aware channel models and new transceiver designs. The section also highlights beamforming, beam-split, and hardware-cost challenges for XL-MIMO.

  • 3) Near-field Spatial Non-stationarity:: Near-field spatial non-stationarity results from spherical-wave amplitude differences across antennas and visibility regions exposing array segments to distinct propagation environments.VR information is especially relevant in environments containing obstacles or scatterers.
  • 3) Near-field Spatial Non-stationarity:: VR-aware deterministic channel models use binary masks to indicate whether each antenna sees a user along each propagation path.The mask vector assigns 1 to visible antennas and 0 to invisible antennas.
  • 3) Near-field Spatial Non-stationarity:: VR-aware stochastic models modify the channel covariance matrix to represent antenna visibility, including diagonal visibility matrices and VR-dependent spatial correlation.These formulations extend covariance-based channel modeling to spatially non-stationary channels.
  • 3) Near-field Spatial Non-stationarity:: Complex environments may require a two-tier VR model because scatterers can be only partially visible to the XL-array and the user.The two visibility relationships may differ from one another.
  • 3) Near-field Spatial Non-stationarity:: Near-field XL-MIMO design challenges include beamforming, beam training, channel estimation, and hybrid beamforming that balances performance against hardware and energy cost.The section frames these as central transceiver-design problems under near-field channel models.
  • 3) Near-field Spatial Non-stationarity:: Phase-shifter-based wide-band beamforming can create beam-split because frequency-flat phases focus different frequencies at different angles and distances, degrading SNR.True time delays are identified as a hardware-based way to generate frequency-dependent beams.

2) Near-field Beam Training:

Near-field beam training must search jointly over angle and distance because spherical wavefronts spread energy across conventional codewords. Proposed approaches trade exhaustive-search accuracy for lower overhead through hierarchical, multi-beam, learning-based, and wide-band methods.

  • 2) Near-field Beam Training:: Near-field beam training requires angle-and-distance codebooks because DFT angle codebooks suffer energy spread across multiple codewords.Cartesian, polar, and slope-domain codebooks are among the proposed alternatives.
  • 2) Near-field Beam Training:: Exhaustive search over polar-domain codewords provides a straightforward narrow-band training method but incurs very high overhead.The overhead reduces the time available for data transmission.
  • 2) Near-field Beam Training:: Hierarchical near-field codebooks reduce training overhead by progressively refining angle and distance coverage across codebook layers.Two-stage designs can first estimate a coarse angle with part of the array and then resolve finer angle and distance jointly.
  • 2) Near-field Beam Training:: Multi-beam and deep-learning methods further reduce search demands by generating multiple beams or learning the mapping from pilot powers to optimal beam angle and distance.Sparse-array grating lobes support simultaneous multi-beam generation, while neural networks predict beam parameters.
  • 2) Near-field Beam Training:: Wide-band beam training must address frequency-dependent defocusing, with TTD-controlled beam-split mitigation and rainbow training covering sampled angles and ranges.Rainbow training separates angle sweeping from range determination.
  • 2) Near-field Beam Training:: Near-field channel estimation is harder because spherical wavefronts invalidate conventional angle-domain sparsity and require more channel parameters per pilot.Wide-band and hybrid-field settings introduce additional estimation challenges.
  • 2) Near-field Beam Training:: Existing near-field work commonly assumes all scatterers lie entirely in either the near- or far-field, leaving hybrid-field beam-split, spatial non-stationarity, and interference management insufficiently addressed.The paper identifies these as open research challenges when near- and far-field paths coexist.

2) DL for Near-field Communications:

Near-field and RIS-based transceiver designs address nonplanar propagation, hardware constraints, channel-estimation burdens, and practical impairments. RISs can extend coverage and provide strong scaling gains, but their benefits depend on accurate CSI and realistic hardware and channel models.

  • Near-field communications: Deep-learning methods target near-field beam training and channel estimation, but more comprehensive studies remain necessary.Near-field spherical wavefronts create nonlinear phase variations across arrays, complicating transceiver design.
  • RIS technologies: RISs reshape propagation using controllable reflecting or refracting elements, while nearly-passive implementations use low-power controllers instead of RF chains.Their hardware may use mechanical actuation, functional materials, or electronic devices such as PIN diodes and MEMS switches.
  • RIS technologies: Nearly-passive RIS performance has been demonstrated from sub-6 GHz through mmWave and THz bands, with the effective cascaded channel represented by gH diag(ν) Q.The cascaded channel combines the BS–RIS and RIS–user links.
  • RIS performance: RISs can establish virtual LoS links and achieve received-signal-power scaling proportional to M^2 under far-field assumptions, although near-field modeling predicts saturation for extremely large surfaces.The corresponding transmit-power reduction is 1/M^2 without compromising received SNR under the stated model.
  • RIS design challenges: RIS design must address difficult CSI acquisition, discrete phase and amplitude control, coupled reflection amplitude and phase, hardware impairments, and channel-estimation errors.Nearly-passive RISs lack RF chains for pilot transmission, while practical phase control creates combinatorial optimization problems.

2) Active RISs:

Active, STAR, and related RIS designs extend the capabilities of nearly-passive surfaces through amplification or simultaneous transmission and reflection. These gains introduce amplification noise, self-interference, additional CSI requirements, and more complex optimization problems.

  • Active RISs: Active RISs mitigate product-distance path loss through amplification, but self-interference can degrade performance because they operate in full-duplex mode.Active RIS deployment also faces practical hurdles related to size, placement, and amplification hardware.
  • Active RISs: Active RIS hardware uses cascaded amplifying and phase-shifting circuits or load-modulation circuits that convert DC power into amplified reflected RF signals.A prototype with 64 elements operating at 3.5 GHz is described in the supplied passage.
  • Active RISs: Active RIS elements introduce non-negligible amplification noise, requiring joint control of reflected signal power and correlated noise.This trade-off produces non-convex reflection-coefficient constraints and harder optimization problems.
  • Active RIS performance: O(M) received-SNR scaling occurs for active RISs under a total active-element power constraint, while nearly-passive RISs provide O(M^2) scaling; active RISs can nevertheless achieve higher rates for moderate M or sufficiently large amplification power.The comparison reflects amplification gain versus the limited beamforming gain of nearly-passive RISs at moderate element counts.
  • STAR-RISs: STAR-RISs extend coverage to both sides of a surface by jointly controlling reflected and transmitted signals through energy-splitting, mode-switching, or time-switching protocols.Their transmission and reflection coefficients must satisfy energy conservation, and CSI is needed for both links.
  • STAR-RISs: STAR-RIS designs can achieve significant coverage extension and full diversity order, but require protocol-dependent joint training and CSI acquisition for reflection and transmission links.The additional links make channel estimation schemes for nearly-passive RISs insufficient without modification.

4) Future Directions:

Future transceiver research must move beyond isolated, idealized RIS elements and fixed-position antennas. Promising directions include interconnected RIS architectures and flexible antennas that adapt antenna locations or shapes to changing wireless environments.

  • RIS directions: Existing RIS studies largely focus on single-RIS or single-reflection systems and static or quasi-static channels, leaving multi-reflection and high-mobility scenarios insufficiently investigated.These limitations define important scope boundaries for current RIS results.
  • RIS directions: Beyond-diagonal RISs exploit controllable inter-element impedances, replacing diagonal reflection matrices with non-diagonal models that provide additional propagation-control degrees of freedom.The added variables also increase the complexity of beamforming and channel-estimation optimization.
  • Flexible antennas: Flexible antennas dynamically adjust some or all antenna locations within a designated space, including designs with near-continuous movement among closely spaced ports.They are also called fluid antennas or movable antennas.
  • Flexible antennas: Fluid antennas can change shape to adapt to wireless conditions, whereas movable antennas use positioning mechanisms such as mechanical slides or rotation shafts.The two architectures represent distinct hardware approaches to antenna flexibility.
  • Flexible antennas: Movable antennas operate within regions spanning several to tens of wavelengths, where spatial correlation makes small-scale fading a dominant determinant of performance.This arrangement can require shorter connecting cables than distributed antenna systems.

2) Performance Gains:

Flexible antennas can improve diversity and multiplexing by changing antenna positions or ports, while HMIMO offers highly flexible electromagnetic-field control through densely spaced arrays and surfaces.

  • Flexible antennas: Flexible antennas provide diversity and multiplexing gains compared with conventional fixed-position arrays.
  • Flexible antennas: Moving antennas to better channel positions can reduce outage probability and exploit spatial diversity, especially as the number of channel paths increases.
  • Flexible antennas: Flexible antennas reconfigure channel matrices through antenna-position or port selection, enabling higher MIMO capacity.
  • Flexible antennas: Accurate CSI is essential for flexible-antenna gains, but exhaustive position measurement is impractical for large transmit or receive regions.
  • Flexible antennas: [214] achieves 50% training-overhead reduction compared with benchmark approaches.
  • HMIMO: HMIMO uses ultra-thin, spatially continuous arrays or surfaces to manipulate electromagnetic-wave amplitudes and phases with high flexibility.
  • HMIMO: HMIMO implementations include leaky-wave antenna and tightly coupled antenna-array architectures, with feeds, substrates, and antenna elements forming the surface.
  • HMIMO: HMIMO channel models account for spatial correlation, mutual coupling, angular scattering, and nearly continuous apertures using array-response and Green’s-function approaches.

3) Transceiver Design Challenges:

HMIMO design must address mutual coupling, electromagnetic-domain information processing, spatial sampling, and high-dimensional channel estimation while preserving the gains of dense apertures.

  • Mutual Coupling: Dense HMIMO packing strengthens spatial correlation and mutual coupling, reducing channel rank and antenna radiation efficiency.
  • Mutual Coupling: Mutual coupling can also enable super-directivity, with array gain scaling with the square of the antenna number in the described setting.
  • EM Information Theory: HMIMO motivates electromagnetic information theory because signal processing can occur in the electromagnetic domain rather than only digitally.
  • EM Information Theory: Transmit-current patterns or basis functions are central to HMIMO performance, with WDM and PDM proposed for pattern-based multiplexing.
  • Spatial Sampling: Spatial sampling schemes seek to preserve maximum electromagnetic mutual information using the minimum number of samples.
  • Channel Estimation: HMIMO channel estimation is difficult because nearly innumerable antennas produce high-dimensional covariance matrices and prohibitively costly matrix inversions.
  • Channel Estimation: Subspace, compressed-sensing, parametric, DFT-based, and unsupervised estimators reduce HMIMO estimation complexity or overhead while targeting MMSE-level accuracy.

4) Future Directions:

Future HMIMO research must verify hardware architectures in practice and characterize distortions, coupling, spacing, and other implementation factors comprehensively.

  • Future Directions: The practical feasibility of HMIMO hardware architectures such as leaky-wave antennas remains to be verified under hardware imperfections.
  • Future Directions: Reverse effects of HMIMO-radiated fields on reference waves may distort holographic imaging and degrade holographic communications.
  • Future Directions: Future studies should comprehensively characterize antenna spacing, mutual coupling, and other important HMIMO implementation factors.

IV. NEW METRIC: SEMANTIC-AWARE TRANSCEIVERS

Semantic-aware transceivers shift communication from bit-level transmission toward semantic-level processing for diverse data and tasks. Their designs combine semantic and channel processing, with performance assessed through task- or reconstruction-oriented metrics.

  • Motivation and Framework: Semantic communication addresses the growth of diverse data modalities by incorporating semantics into communication-system design.
  • Motivation and Framework: Semantic and channel encoders and decoders may be designed jointly or separately, with trainable parameters assigned to the corresponding modules.
  • Motivation and Framework: A typical semantic-communication pipeline maps source data to symbols, transmits them through an impaired physical channel, and reconstructs the source at the receiver.
  • Design Directions: Semantic coding performance is evaluated using metrics such as reconstruction distortion, classification accuracy, PSNR, and BLEU.
  • Source Recovery: Source-recovery systems compress source data directly into channel symbols while incorporating redundancy for noise resistance.
  • Source Recovery: Deep-learning-based semantic transmission has advanced across vision, text, audio, and point-cloud modalities.

2) Intelligent Task Execution:

Intelligent task-oriented semantic communication extends source recovery toward classification and other application tasks, while raising questions about semantic characterization, multitask operation, and security across system preparation.

  • Intelligent Task Execution: Task-oriented semantic communication targets specific applications such as classification rather than only recovering transmitted signals.Its motivation is to reduce the bandwidth needed for reconstruction-focused communication.
  • Semantic Information Characterization: Information bottleneck methods trade off task performance and transmission rate by maximizing task-relevant mutual information while minimizing the rate.These methods have been applied to classification and extended to distributed scenarios.
  • Multitask Execution: Existing semantic communication systems often specialize in one task and one modality, whereas practical transceivers may need source recovery and intelligent tasks simultaneously.Multitask systems therefore seek unified handling of diverse data and objectives.
  • Threats During Deployment: Deployment threats include semantic information leakage and attacks against semantic models, including eavesdropping, adversarial manipulation, and inference of sensitive properties.These risks affect both transmitted semantic content and the models used to process it.
  • Threats Before Deployment: Semantic communication faces security threats during implementation, data preparation, pretraining, and collaborative learning.Reported risks include framework exploitation, data poisoning, compromised public models, and attacks by malicious participants.

2) Threats and Countermeasures in Deployment Phase:

Deployment-phase semantic communication threats target either model behavior or private information, while practical designs combine adaptive transmission, multi-antenna techniques, and RIS-assisted channels to improve operation.

  • Deployment Threats: Deployment threats primarily include adversarial attacks against trained models and privacy leakage from semantic information.These threats arise after the semantic transceiver has been trained and deployed.
  • Adversarial Attacks: Adversarial samples can be injected during encoding or transmission, causing semantic errors, decoding failures, or misinterpretations.Unlike conventional jamming, the attack can target semantic correctness rather than only legitimate reception.
  • Privacy Leakage: Semantic communication can leak private information because channel symbols remain directly correlated with source data and semantic representations.Inference attacks may reveal properties such as demographics, medical conditions, economic status, or preferences.
  • Adaptive Transmission: Rate adaptation adjusts transmission power, data rate, or coding rate using receiver channel-state information and feedback.These controls are intended to improve reliability and spectral efficiency.
  • Multi-Antenna Transmission: Multi-antenna semantic communication addresses the limitations of studies based mainly on SISO transmission over AWGN channels.MIMO designs incorporate channel conditions and feature importance into transmission decisions.
  • RIS-Aided Transmission: RIS-assisted semantic communication can improve channel quality and spectral and energy efficiency while outperforming point-to-point systems in reported BLEU scores.The cited RIS-aided system also showed greater robustness to channel-estimation errors in Rayleigh channels.

5) Digital Modulation:

Digital semantic communication converts source information into discrete representations or bit sequences for modulation, while current research addresses adaptability, resource allocation, large-model integration, coding efficiency, and hardware realization.

  • Digital Modulation: Digital semantic communication research commonly maps source information to discrete constellation points or to bit sequences followed by standard modulation.These are the two principal design categories described for digital semantic transmission.
  • Trainable Discretization: Differentiable soft quantization, sigmoid approximations, VAE-based learning, and VQ-VAE techniques enable trainable mappings to discrete symbols or latent representations.The methods support end-to-end optimization despite discrete modulation choices.
  • Retransmission Design: End-to-end semantic networks may require retraining when channel conditions change and may not dynamically adjust coding length to those conditions.HARQ-based designs address these limitations by adapting code rates using feedback.
  • Resource Allocation: Semantic communication evaluates performance with task-specific metrics and resource constraints, motivating semantic metrics and resource-allocation strategies.These metrics differ from conventional communication evaluations.
  • Large AI Model-Empowered Design: Large AI models can create low-dimensional signal representations, reconstruct multimodal data, and allocate transmission emphasis to important words, image patches, or video frames.Their training and inference costs remain important research considerations.
  • Advanced Coding Techniques: Advanced source coding methods use VAE, learned entropy modeling, and hyper-priors, while joint black-box source-channel coding may lack specifically designed error protection.Combining advanced source and channel coding is identified as a potential improvement.
  • Hardware Realization: Practical semantic communication requires hardware realization alongside algorithm design, with challenges involving performance, energy, real-time operation, security, and customization.Dedicated hardware platforms have demonstrated performance enhancements over conventional systems.

4) Semantic Information Theory:

Semantic-aware NGAT must address immature semantic information theory, cross-layer protocol adaptation, security, and practical NOMA challenges. The section also identifies semantic and NGAT-specific metrics and opportunities for improving connectivity.

  • Semantic information theory: Semantic information theory remains immature because semantic ambiguity varies across tasks, motivating suitable performance metrics and semantic similarity measures.The framework is intended to guide semantic communication system design and support further theoretical development.
  • Protocol and security challenges: Semantic communication requires joint optimization across protocol layers rather than conventional layer separation, while preserving semantic-information security.The text highlights encryption and protocol design as additional security requirements.
  • NGAT-enabled NOMA: Near-field beam focusing and new antenna types create NOMA opportunities, but enhanced degrees of freedom complicate user clustering and SIC operation.RIS-aided NOMA has been studied, while FA/HMIMO-aided NOMA remains an open research avenue.
  • Semantic-aware NOMA: Semantic communication can improve NOMA resource efficiency because it is generally more robust and requires fewer radio resources than conventional communication.An opportunistic semantic-and-bit communication strategy was proposed for uplink NOMA.
  • NOMA design challenges: NOMA faces practical challenges from CSI-estimation overhead, SIC error propagation, and the complexity of clustering users under near-field and new-form antenna configurations.The text proposes tailored CSI estimation, imperfect-CSI designs, and semantic communication for mitigating SIC error propagation.

B. Localization and Sensing

NGAT supports localization and sensing by exploiting near-field propagation, RIS-assisted paths, and hardware-aware transceiver models. These approaches improve target-motion characterization and account for electromagnetic coupling and circuit-level effects, while introducing communication–sensing trade-offs and signal-loss limits.

  • Near-field L&S: Near-field localization and sensing estimates both radial and transverse target velocities, enabling more detailed motion profiles for trajectory prediction and tracking.Far-field Doppler estimates are limited to radial motion, whereas near-field Doppler frequencies contain both velocity components.
  • RIS-assisted L&S: RISs can create virtual line-of-sight paths around obstructions and provide additional observations, but passive RIS and STAR-RIS sensing can suffer cumulative echo-path loss.Active sensors or amplifiers are proposed to increase received SNR or echo-signal amplitude.
  • Integrated sensing and communications: ISAC transceivers must balance communication and sensing performance because both functions compete for shared resources, while also requiring new waveforms.The section frames efficient multifunction transceiver design as a central ISAC issue.
  • Hardware-aware transceiver design: Circuit and electromagnetic theory connect processed digital signals with generated electromagnetic fields, enabling hardware-aware analysis of communication systems.The approach models antennas as wave transformers and multi-antenna systems as multi-port networks.
  • Circuit theory based design: The multi-port model captures mutual coupling, impedance matching, structural losses, and intrinsic and extrinsic noise in transceiver and RIS-assisted systems.Transmit–receive mutual impedances and receive-to-transmit back-scattering impedance are included in the model.
  • Section scope: The paper surveys near-field channel and transceiver designs, RISs, flexible antennas, holographic MIMO, semantic-aware NGAT, and other future transceiver technologies.The overview emphasizes beamforming, beam training, channel estimation, and open challenges across these technologies.
Loading 2403.16458v3…