Source-linked AI summary
Extremely Large-Scale MIMO: Fundamentals, Challenges, Solutions, and Future Directions
Zhe Wang, Jiayi Zhang, Hongyang Du, Wei E. I. Sha, Bo Ai, Dusit Niyato, Mérouane Debbah
TL;DR
XL-MIMO research lacks a unified account of its diverse hardware schemes and practical near-field modeling, analysis, and processing requirements. The paper reviews these areas, discusses challenges and solutions, and develops tutorials and case studies for hybrid propagation modeling and practical EDoF computations. Numerical investigations examine EDoF for unparallel XL-MIMO surfaces and multiple user equipments.
Problem
XL-MIMO has diverse hardware schemes whose characteristics and relationships are not well discussed, while practical near-field performance analysis and scalable signal processing remain important research challenges.
Method
The paper comprehensively reviews XL-MIMO hardware, channel modeling, performance analysis, and signal processing, then proposes solutions, tutorials, and two practical case studies.
Results
The paper provides numerical investigations of EDoF performance for unparallel XL-MIMO surfaces and multiple user equipments, and reports that near-field LoS propagation can enhance DoF.
Takeaways & Limitations
The review and proposed solutions provide foundations for further XL-MIMO channel-modeling and performance-optimization research.
Abstract
from arXiv · showhide
Extremely large-scale multiple-input-multiple-output (XL-MIMO) is a promising technology to empower the next-generation communications. However, XL-MIMO, which is still in its early stage of research, has been designed with a variety of hardware and performance analysis schemes. To illustrate the differences and similarities among these schemes, we comprehensively review existing XL-MIMO hardware designs and characteristics in this article. Then, we thoroughly discuss the research status of XL-MIMO from "channel modeling", "performance analysis", and "signal processing". Several existing challenges are introduced and respective solutions are provided. We then propose two case studies for the hybrid propagation channel modeling and the effective degrees of freedom (EDoF) computations for practical scenarios. Using our proposed solutions, we perform numerical results to investigate the EDoF performance for the scenarios with unparallel XL-MIMO surfaces and multiple user equipment, respectively. Finally, we discuss several future research directions.
I. INTRODUCTION
XL-MIMO extends massive MIMO through extremely large antenna deployments, but its new near-field electromagnetic characteristics create unresolved modeling, analysis, and processing challenges. This article reviews the technology, organizes existing schemes, proposes solutions and practical case studies, and identifies future directions.
- XL-MIMO deploys an extremely large number of antennas in compact spaces, encompassing schemes such as holographic MIMO, LIS, ELAA, and CAP-MIMO.
- Near-field electromagnetic operation introduces new channel characteristics and makes XL-MIMO performance analysis and optimization difficult.
- Existing DoF studies mainly use primitive parallel, single-pair, or scalar-channel scenarios, motivating analysis in more practical XL-MIMO systems.
- Large arrays require near-field signal-processing methods, while broadband systems must address near-field beam squint and acceptable processing complexity.
- The article reviews hardware schemes and their relationships, covers channel modeling, performance analysis, and signal processing, and discusses challenges, solutions, case studies, and future directions.
II. FUNDAMENTAL THEORIES OF XL-MIMO
XL-MIMO hardware designs range from discrete linear and planar arrays to spatially continuous apertures. The schemes are related through limiting or special-case constructions, although CAP-based and discrete implementations differ practically.
- ULA-based XL-MIMO: ULA-based XL-MIMO uses extremely large linear arrays, potentially reaching 1024 antennas or more and changing electromagnetic characteristics beyond antenna-count growth.
- UPA-based XL-MIMO: UPA-based XL-MIMO uses thousands of patch or point antennas with typically sub-half-wavelength horizontal or vertical spacing.
- Relationships among schemes: CAP-based XL-MIMO differs from discrete-antenna schemes in practical design and signal-processing protocols, despite their discussed mathematical and physical relationships.
- CAP-based XL-MIMO: CAP-MIMO uses infinitely many infinitely small-spaced antennas, forming a spatially continuous electromagnetic surface.
- Relationships among schemes: The ULA can be treated as a special case of a UPA with one antenna column, while patch and point models are connected by reducing element size toward zero.
- Relationships among schemes: A discrete UPA becomes a continuous aperture when patch size equals spacing or point-antenna spacing becomes infinitesimal.
B. Channel Modeling
XL-MIMO channel modeling must account for near-field electromagnetic propagation, where vectorial spherical waves describe the receiver-side field more accurately than far-field simplifications.
- XL-MIMO channel modeling describes current distributions at the transmitter and induced electric fields at the receiver across LoS and NLoS propagation environments.
- The near-field receiver region requires electromagnetic modeling based on vectorial spherical waves and flexible transmitter current-density distributions.
- LoS and NLoS channel fundamentals are treated through free-space and scattered propagation descriptions, respectively.
1) LoS Propagation:
XL-MIMO’s large aperture makes LoS propagation predominant while requiring distance- and angle-varying channel responses across the array. The modeling framework therefore distinguishes spherical-wave and plane-wave regimes.
- LoS propagation: Large XL-MIMO arrays shorten transmission ranges, making LoS propagation predominant and motivating Green’s-function modeling of electromagnetic characteristics.
- LoS propagation: Dyadic Green’s functions solve Maxwell’s equations numerically and model full polarization through three transmit/receive-point polarizations.
- Complex channel response: Near-field distances and angles vary across the array, so complex channel responses require models that capture spatial amplitude and phase behavior.
- Propagation regions: The radiative near-field uses non-uniform spherical waves because both amplitude and phase vary over the receiver aperture.
- Propagation regions: The Fresnel region neglects amplitude variation but retains noticeable phase variation through a uniform spherical-wave model.
- Propagation regions: The far-field region uses uniform plane waves when link distance greatly exceeds array dimensions, with phase determined by incident angle.
2) NLoS Propagation:
XL-MIMO NLoS channels can be modeled either through Fourier plane-wave representations or superpositions of array response vectors, with near-field models incorporating richer spatial information than far-field plane-wave models.
- Fourier Plane-Wave Representation: Fourier plane-wave modeling supports arbitrary scattering conditions because spherical waves can be exactly decomposed into infinitely many plane waves.For tractability, channels may be represented as a finite superposition of steering vectors or discretized plane waves.
- Fourier Plane-Wave Representation: Fourier plane-wave representations couple incident and received fields through a scattering kernel integral operator.The channel response is expressed using source, receive, and angular responses, then approximated with discretized plane waves and random Fourier coefficients.
- Fourier Plane-Wave Representation: The source response maps transmitted excitation currents to propagation directions, while the receive response maps directions to induced currents.
- Fourier Plane-Wave Representation: The angular response maps each source propagation direction onto each receive propagation direction.
- Array Response Vector Representation: Array response vector models represent the NLoS channel as a superposition whose size depends on the number of path components associated with effective scatterers.Far-field vectors depend only on AoA/AoD, whereas near-field vectors reflect near-field channel characteristics.
C. Performance Analysis
XL-MIMO performance analysis uses DoF and EDoF within electromagnetic information theory, while recognizing that scalar and far-field approximations can be inaccurate for practical near-field configurations.
- DoF for the XL-MIMO: DoF is the rank of the channel matrix and cannot exceed the smaller of the transmitting and receiving antenna counts.
- DoF for the XL-MIMO: Near-field LoS propagation provides a large range of angles, enabling XL-MIMO to enhance DoF even without NLoS propagation.
- DoF for the XL-MIMO: Approximate XL-MIMO DoF scales with transmitter-receiver length products for linear arrays and area products for large surfaces, with distance-dependent reductions.These expressions are based on prolate spheroidal wave functions and differ by array geometry.
- DoF for the XL-MIMO: EDoF is the equivalent number of SISO systems, approximately computed from the dyadic-Green’s-function channel matrix using (tr(HH^H)/∥HH^H∥_F)^2.
- DoF for the XL-MIMO: Approximate DoF formulas can overestimate DoF when link distance is comparable to transmitter and receiver physical sizes.The cited expressions assume link distance is much larger than the physical sizes and rely on scalar Green’s functions.
- Electromagnetic Information Theory: Traditional capacity theory mismatches four-dimensional electromagnetic fields, motivating electromagnetic information theory to characterize fundamental capacity bounds.
D. Signal Processing
XL-MIMO signal processing must capture near-field electromagnetic characteristics while controlling complexity, spanning channel estimation and beamforming or precoding design.
- Overview: Channel estimation and beamforming/precoding design are treated as the main XL-MIMO signal-processing areas.Their distinct requirements arise from XL-MIMO’s physical size and electromagnetic characteristics.
- Channel estimation: The RS-LS estimator exploits the spatial-correlation eigenspace and array geometry to reduce implementation difficulty.It outperforms conventional LS and achieves performance similar to MMSE.
- Channel estimation: Polar-domain estimation jointly represents angle and distance to capture near-field channels, using a Fresnel-based transform and OMP algorithm for NLoS estimates.
- Beamforming/Precoding design: Pattern-division multiplexing designs electric current density distributions through iterative optimization to maximize sum capacity for multiple users.
- Beamforming/Precoding design: A low-computational ZF precoder replaces matrix inversion with Neumann series expansion in multi-UE UPA-based XL-MIMO.
- Overview: XL-MIMO processing schemes must simultaneously capture electromagnetic characteristics and maintain low computational complexity.
III. KEY CHALLENGES & SOLUTIONS
The paper identifies three XL-MIMO challenges: tractable hybrid-channel modeling, practical EDoF analysis, and processing with acceptable complexity, then discusses corresponding solutions and case studies.
- Challenge 3: Low-complexity processing: The third challenge is designing XL-MIMO processing schemes with acceptable complexity.
- Challenge 1: Hybrid-channel modeling: Hybrid propagation requires accurate and tractable modeling that includes both LoS and NLoS components.
- Challenge 2: Practical EDoF analysis: Existing DoF analyses rely on idealized parallel links, single transmitter-receiver pairs, and scalar Green’s-function models rather than practical configurations.
- Challenge 2: Practical EDoF analysis: The paper analyzes EDoF for unparallel transmitter-receiver surfaces and multiple-transmitter scenarios using UPA-based XL-MIMO with point antennas.
- Challenge 2: Practical EDoF analysis: Parallel transmitter and receiver achieve the highest EDoF, while increasing antenna count and decreasing spacing initially increase EDoF before saturation.
- Challenge 2: Practical EDoF analysis: For two UEs, EDoF is below the sum of the individual-UE EDoFs because coupling between UEs can strongly affect performance.
C. Signal Processing
XL-MIMO signal processing must address high complexity, latency, and power consumption caused by its large antenna arrays. The paper discusses distributed processing, antenna selection, low-complexity learning-based methods, and reduced-hardware architectures as potential solutions.
- C. Signal Processing: Large XL-MIMO arrays make it challenging to design signal-processing schemes with acceptable complexity.High processing complexity also increases signal-processing latency.
- C. Signal Processing: The practical EDoF studies consider unparallel XL-MIMO surfaces and XL-MIMO systems with multiple single-antenna UEs.Fig. 5 varies N for unparallel transmit and receive surfaces, while Fig. 6 varies d1 for one- or two-UE settings.
- C. Signal Processing: Distributed XL-MIMO uses local processing units, central processing units, and fronthaul links to flexibly support processing requirements.The topology and CPU assistance can reduce BS computation overhead.
- C. Signal Processing: Antenna selection and activation can serve particular UEs with only a partial group of antennas.Optimization can select antenna groups or switch particular antennas on or off for different requirements.
- C. Signal Processing: Deep learning, low-complexity processing, and semantic-based schemes are advocated to reduce processing complexity, latency, or transmitted data size.Examples include DL-based beam training and semantic extraction from source data.
- C. Signal Processing: Few-RF-chain, hybrid-beamforming, mixed-resolution-ADC, antenna-selection, modular, and sparse-processing architectures are potential approaches to reduce XL-MIMO power consumption.The few-RF-chain design connects each RF chain to a certain antenna module.
IV. FUTURE DIRECTIONS
Future XL-MIMO research includes distributed implementations and semantic communications. The paper highlights distributed optimization, hardware and synchronization issues, and semantic feature extraction with resource-aware beamforming.
- A. Distributed XL-MIMO Implementation: Distributed XL-MIMO remains an open area involving topology-dependent processing, BS and antenna selection, hardware cost, and synchronization.Low-resolution ADCs, hybrid beamforming, and over-the-air reciprocity calibration are identified as possible responses.
- A. Distributed XL-MIMO Implementation: Distributed learning is regarded as a possible solution for optimization across XL-MIMO base stations.The paper specifically points to distributed optimization at each XL-MIMO BS.
- B. Semantic Communications: Semantic communications can extract task-related information to improve network efficiency and reduce wireless data-transmission burden.The paper proposes joint semantic channel encoding for semantic extraction and XL-MIMO transmission.
- B. Semantic Communications: Semantic features can be weighted through attention-based scaling parameters that guide resource allocation and beamforming design.Finer beamforming can be used for relatively more important semantic information under limited resources.
C. Integrated Sensing and Communications
The paper identifies integrated sensing and communications and near-field wireless power transfer as future XL-MIMO directions. It also summarizes case studies that apply proposed methods to practical channel-modeling and EDoF scenarios.
- C. Integrated Sensing and Communications: ISAC can integrate sensing and communication to use wireless resources efficiently in XL-MIMO systems.Sensory data can assist beamforming, while XL-MIMO degrees of freedom can benefit localization and tracking.
- C. Integrated Sensing and Communications: The relationship between XL-MIMO degrees of freedom and localization or tracking is identified as a promising research direction.The paper specifically highlights exploiting XL-MIMO DoF for these sensing tasks.
- D. Wireless power transfer: Near-field wireless power transfer is proposed as an emerging XL-MIMO topic for addressing future wireless-network energy-supply issues.The paper points to near-field WPT protocols and the relationship between XL-MIMO DoF and WPT capabilities.
- V. CONCLUSIONS: The article reviews XL-MIMO fundamentals, challenges, solutions, and future directions across channel modeling, performance analysis, and signal processing.Its numerical studies investigate EDoF for unparallel XL-MIMO surfaces and multiple UEs.