PaperlayerView exact revision on arXiv

Source-linked AI summary

RIS-Aided Cell-Free Massive MIMO Systems for 6G: Fundamentals, System Design, and Applications

Enyu Shi, Jiayi Zhang, Hongyang Du, Bo Ai, Chau Yuen, Dusit Niyato, Khaled B. Letaief, Xuemin Shen

arXiv:2310.00263v3cs.ITeess.SP

TL;DR

Rising 6G demands motivate integrating CF mMIMO and RIS, but the combined systems introduce distinctive design and implementation challenges. This paper surveys their models, algorithms, practical constraints, and interactions with emerging technologies, synthesizing current findings and future research directions.

  • Problem

    6G requires high spectral and energy efficiency, ultra-low latency, and ultra-high reliability, motivating integrated CF mMIMO and RIS systems while their research remains fragmented and sparse.

  • Method

    The paper provides a comprehensive survey covering system models, channel estimation, joint beamforming, resource allocation, practical challenges, and emerging 6G technologies.

  • Results

    The survey consolidates reported analyses and solutions for RIS-aided CF mMIMO operation, resource allocation, hardware impairments, electromagnetic interference, fronthaul limits, deployment, and related technologies.

  • Takeaways & Limitations

    RIS-aided CF mMIMO is presented as a research direction combining CF mMIMO and RIS capabilities while requiring new multi-layer processing, estimation, and practical-system designs.

Abstract

from arXiv · show

An introduction of intelligent interconnectivity for people and things has posed higher demands and more challenges for sixth-generation (6G) networks, such as high spectral efficiency and energy efficiency, ultra-low latency, and ultra-high reliability. Cell-free (CF) massive multiple-input multiple-output (mMIMO) and reconfigurable intelligent surface (RIS), also called intelligent reflecting surface (IRS), are two promising technologies for coping with these unprecedented demands. Given their distinct capabilities, integrating the two technologies to further enhance wireless network performances has received great research and development attention. In this paper, we provide a comprehensive survey of research on RIS-aided CF mMIMO wireless communication systems. We first introduce system models focusing on system architecture and application scenarios, channel models, and communication protocols. Subsequently, we summarize the relevant studies on system operation and resource allocation, providing in-depth analyses and discussions. Following this, we present practical challenges faced by RIS-aided CF mMIMO systems, particularly those introduced by RIS, such as hardware impairments and electromagnetic interference. We summarize corresponding analyses and solutions to further facilitate the implementation of RIS-aided CF mMIMO systems. Furthermore, we explore an interplay between RIS-aided CF mMIMO and other emerging 6G technologies, such as next-generation multiple-access (NGMA), simultaneous wireless information and power transfer (SWIPT), and millimeter wave (mmWave). Finally, we outline several research directions for future RIS-aided CF mMIMO systems.

I. INTRODUCTION

6G requirements motivate combining user-centric cell-free massive MIMO with RIS, which can reshape propagation using low-power, low-cost surfaces. This survey consolidates the fragmented literature across models, operation, practical constraints, emerging technologies, and future directions.

  • Motivation: CF mMIMO uses geographically distributed APs connected to a CPU to jointly serve all UEs on identical time-frequency resources.This user-centric architecture differs from conventional centralized, BS-centric mMIMO.
  • Motivation: RIS alters radio waves without complex digital signal processing or active power amplifiers and can assist users with poor channel conditions.Its low-power, low-cost fabrication supports flexible deployment, but its passive nature makes integration with other technologies important.
  • Motivation: RIS-aided CF mMIMO combines a CPU, distributed APs, and distributed RISs without cell boundaries to pursue ubiquitous high-capacity coverage.The integration introduces changes to system architecture, protocols, signal processing, channel models, and channel estimation rather than merely combining two technologies.
  • Key contributions: The survey presents the first comprehensive review of RIS-aided CF mMIMO across different aspects and identifies open problems and future directions.Its scope includes emerging technologies such as NOMA, SWIPT, mmWave/THz, and UAVs.
  • Key contributions: The survey addresses system architecture, application scenarios, channel models, communication protocols, system operation, resource allocation, and multi-layer signal processing.It also compares RIS-aided CF mMIMO with standalone RIS and CF mMIMO systems.
  • Key contributions: The paper reviews practical considerations including hardware impairments, electromagnetic interference, limited fronthaul capacity, and RIS deployment.It emphasizes analyses and proposed solutions intended to facilitate implementation.

C. Organization of the Survey

The survey is organized from foundational system principles to operation, practical performance considerations, emerging-technology integration, and future research directions.

  • Section II presents the RIS-aided CF mMIMO system’s architecture, application scenarios, channel model, and communication protocol.
  • Section III surveys system operation and resource allocation, including channel estimation, joint beamforming, multi-stage transmission, signal processing, and resource allocation.
  • Sections IV–VI address practical system performance, integration with emerging 6G technologies, and future research directions.

II. RIS-AIDED CF MMIMO SYSTEM MODEL

The system model section introduces RIS-aided CF mMIMO architecture and application scenarios, compares channel models, and discusses operation under FDD and TDD protocols.

  • The section introduces the system architecture and major application scenarios of RIS-aided CF mMIMO systems.
  • Application scenarios span data-demand cases—mMTC, high-mobility, XL-MIMO, mmWave, and Metaverse—and energy-demand cases such as physical-layer wireless energy transfer.
  • Channel models among APs, UEs, and RISs are compared and analyzed under FDD and TDD communication protocols.

A. System Architecture and Application Scenarios

RIS-aided CF mMIMO combines distributed APs and RISs connected to a CPU, supporting shared service for users across data- and energy-demand scenarios. Its channel model includes direct and RIS-cascaded links.

  • System Architecture: The architecture comprises L APs with M antennas, T RISs with N elements, and K randomly distributed UEs with single or multiple antennas.
  • System Architecture: A CPU connects all APs and RISs through fronthaul links, exchanging power-control coefficients and payload data while enabling spatial multiplexing for all users.
  • Application Scenarios: RIS-aided CF mMIMO can support stable information and energy transmission even when the direct path is obstructed.
  • Channel Model: The model distinguishes direct AP–UE links from cascaded AP–RIS–UE links formed by two channels through each RIS.
  • Channel Model: The RIS phase-shift matrix is diagonal, with each element phase ϕ_nt constrained to [−π, π].
  • Channel Model: Cascaded links are commonly modeled as two independent channels, enabling separate channel design and closer correspondence to real-world conditions.

C. Communication Protocol

TDD dominates current RIS-aided CF mMIMO channel-estimation research because it is simpler, while FDD introduces nonreciprocity, feedback overhead, and beam-misalignment concerns. The section surveys estimation techniques and broader system-operation methods.

  • Communication Protocol: TDD uses the same frequency band for uplink and downlink, with coherence blocks divided into uplink training, uplink data, and downlink transmission phases.
  • Communication Protocol: FDD uses separate uplink and downlink frequency bands, so channel nonreciprocity requires downlink channel estimates for downlink precoding.
  • Communication Protocol: Additional downlink training in FDD improves channel acquisition but consumes time-frequency resources, reduces communication rates, and increases CSI exchange load with many APs.
  • Channel Estimation: Pilot-based TDD estimation has UEs transmit τ_p-length pilots to APs, using orthogonal or non-orthogonal sequences according to UE count and channel coherence time.
  • Channel Estimation: Separation channel estimation divides the aggregated channel into direct and cascaded components, estimating them in separate sub-phases by switching RIS elements or RISs.
  • Channel Estimation: Two-timescale compressive-sensing estimation reduces NMSE and approaches Oracle LS, while GST lowers NMSE relative to standard ST and regular pilot schemes.

B. Joint Beamforming Design

Joint beamforming design coordinates RIS phase shifts with AP transmission strategies under objectives including security, sum rate, fairness, and energy efficiency. The reviewed methods address imperfect CSI, distributed APs, multiple users, and practical RIS constraints.

  • Motivation: RIS phase shifts must be jointly designed with multiple APs to avoid reflected interference and support collaborative service.Poor beamforming can turn reflected beams into interference, degrading expected-signal detection and user performance.
  • Sum rate maximization: Distributed APs and inaccurate real-time channel information make sum-rate optimization more constrained and complex.The reviewed approaches include distributed ADMM variants with small signaling overhead for large-scale systems.
  • User fairness: Fairness-oriented designs maximize the minimum user rate, often combining statistical RIS CSI with instantaneous AP CSI.Representative methods use genetic algorithms, SDR, ZF, integer linear programming, or successive refinement under practical channel and phase-shift settings.
  • EE and EEF maximization: Well-designed active and passive RIS beamforming improves worst-user energy efficiency by 13.8% and 50%, respectively.With random phase shifts, active RIS still achieves a 40% improvement in worst-user energy efficiency.
  • Information security: Information-security designs jointly optimize RIS phases and AP power coefficients to reduce information leakage while maintaining legitimate-user spectral efficiency.One study uses statistical CSI, ZF precoding, and SDP for RIS phase-shift design under spoofing attacks.

2) Approaches for Joint Beamforming:

Joint beamforming methods mainly comprise traditional optimization and machine learning. Traditional methods handle non-convex formulations through techniques with differing complexity, convergence, scalability, and optimality trade-offs, while machine learning improves data-driven optimization but has interpretability and optimality limitations.

  • Traditional optimization: Traditional joint-beamforming methods mainly use alternating optimization for multi-variable non-convex problems.Fixing passive beamforming makes active beamforming a conventional optimization problem.
  • Semi-definite programming: SDP relaxes unit-modulus constraints into convex problems but may produce suboptimal or infeasible rank-one solutions.It supports complex constraints and can converge quickly, while rank-one reconstruction remains a practical limitation.
  • Iterative algorithm: Iterative algorithms trade complexity for performance and support large-scale non-convex problems, but may converge slowly to initialization-sensitive local optima.Their solutions can depend strongly on initialization.
  • Sub-gradient: Sub-gradient methods reduce computational complexity for nonsmooth problems but converge slowly and depend strongly on initialization.They avoid global and higher-order derivative computations.
  • Manifold optimization: Manifold optimization offers fast convergence and high accuracy for some structured problems, but has high computational complexity and limited universality.Its applicability is restricted to particular geometric problem structures.
  • Machine learning: Machine learning extracts features and patterns from complex data, improving optimization accuracy and stability, but lacks interpretability and cannot guarantee global optimality.Reliable results often require large amounts of labeled data.

C. Multi-stage Transmission Procedure and Multi-layer Signal Processing

RIS-aided CF mMIMO adds a RIS layer to the CF architecture, creating a multi-layer system with four transmission stages and hierarchical signal processing. This structure can reduce inter-user interference and resource costs, but requires strict protocols and standard guidance.

  • System architecture: The RIS-aided CF mMIMO architecture adds a cascading RIS link between distributed APs and UEs, forming a 3.5-layer structure.Its multi-layer processing is associated with reduced inter-user interference and enhanced system performance.
  • Wireless energy transfer: Stage 1 transfers wireless energy from APs to RISs for energy harvesting, leveraging RIS’s low power consumption.The CPU issues energy-control commands after receiving fronthaul information and signal detection.
  • Uplink pilot transmission: Stage 2 uses UE pilot transmission and AP-local channel estimation, with RIS phase shifts fixed to preserve estimation accuracy.APs forward required channel-estimation information to the CPU for later processing.
  • Uplink data transmission: Stage 3 combines direct and RIS-cascaded uplink signals locally at APs before CPU decoding.MR/GMR or MMSE combining and central decoding or LSFD can be used.
  • Downlink data transmission: Stage 4 sends CPU-precoded downlink data through APs and RISs while APs control RIS phase adaptation.Dynamic RIS control may require revised frame structures and dedicated control slots.
  • Signal processing: Hierarchical CPU, AP, and RIS processing can reduce resource costs from multi-read information interaction but requires strict protocols and standards.The multi-stage procedure follows from the system’s distributed structure.

D. Resource Allocation

Resource allocation in RIS-aided CF mMIMO must coordinate time-frequency-power assignments with AP-RIS-UE associations in a user-centric, cell-free network. Existing studies address selection complexity through structured optimization and grouping strategies, but large-scale resource management remains incompletely explored.

  • Joint AP-RIS-UE selection extends conventional UE-AP association because RIS changes channel propagation and introduces additional access-design decisions.
  • UE-AP selection: Full AP-UE access can become computationally expensive with large-scale RIS, motivating partially connected selection schemes that reduce resource consumption while improving performance.A relaxed linear approximation addresses the resulting BIQP formulation.
  • UE-RIS selection: UE-RIS association strongly affects network performance; assigning RIS to poorly served nearby users achieved a 5-times improvement for the worst user at a 100 m RIS-user distance.
  • RIS-AP selection: Heuristic RIS-AP distance selection assigns each RIS to its closest AP, although obtaining accurate RIS distance information can be challenging in practice.
  • Future resource management should investigate channel-quality, service-quality, and deep-learning-based grouping and dynamic allocation strategies.

A. Hardware Impairments

RIS-aided CF mMIMO performance is constrained by coupled hardware impairments, including spatial correlation, phase-shift quantization, and low-resolution converters. Studies identify practical operating points, while emphasizing that joint impairment effects remain insufficiently understood.

  • Spatial correlation: Spatial correlation in RIS elements and AP antennas jointly degrades system performance by disrupting beamforming and reducing achievable spectral-efficiency gains.With half-wavelength RIS-element spacing, the degradation is minimal.
  • Phase-shift errors: 3-bit RIS phase quantization performs close to continuous phase shifting, while 5-bit quantization is nearly indistinguishable from continuous phase shifting.
  • ADCs/DACs: 5-bit ADCs/DACs provide results sufficiently close to ideal converters for achievable ergodic secrecy rate.
  • All hardware impairments cause performance degradation, with ADCs/DACs especially severe because they distort signals from both direct and RIS-cascaded links.
  • Current research mainly studies hardware impairments individually, leaving their coexistence and mutual effects as an important open direction.

B. Electromagnetic Interference

RIS reflects electromagnetic interference as well as useful signals, complicating performance characterization in CF networks. EMI-aware analysis shows that increasing AP antennas can become less cost-effective, while fronthaul and deployment constraints remain practical challenges.

  • RIS reflection of EMI introduces beamforming inaccuracies, so EMI modeling is needed for accurate RIS-aided CF mMIMO performance characterization.
  • EMI significantly degrades uplink spectral efficiency, and the performance gap grows with the number of AP antennas because RIS-reflected EMI power increases.
  • Under EMI, continuously increasing AP antennas is not cost-effective for achieving further system-performance improvements.
  • Fronthaul capacity: Limited fronthaul capacity constrains AP-CPU data and control exchange and also affects RIS control, channel-information acquisition, and beamforming coordination.
  • RIS deployment: In a two-UE, two-AP scenario, placing RIS 10 m from a UE cluster improves sum rate 2.3 times relative to placement 60 m away.Multi-RIS, multi-antenna settings make deployment and interference management more challenging than simplified single-RIS scenarios.
  • Practical implementation requires balancing performance and cost because hardware, capacity, and deployment limitations can degrade system performance.

V. RIS-AIDED CF MMIMO WITH OTHER ENABLING TECHNOLOGIES TOWARDS 6G

RIS-aided CF mMIMO is being combined with NOMA and SWIPT to address spectral-efficiency, connectivity, reliability, and IoT energy-sustainability goals. These integrations introduce coupled clustering, interference, beamforming, power-control, and protocol-design challenges.

  • A. NOMA: Power-domain NOMA enables non-orthogonal parallel transmission, with successive interference cancellation separating users’ signals.
  • A. NOMA: Multi-agent DDPG-based joint RIS-AP beamforming can significantly enhance performance in the considered RIS-aided CF mMIMO system.
  • A. NOMA: NOMA-based RIS-aided CF mMIMO requires effective UE clustering, RIS-UE/AP matching, beamforming, and power control to manage inter-user interference.
  • B. SWIPT: SWIPT supports IoT energy sustainability, while wireless energy transfer can power passive RISs and remove the need for a dedicated RIS power supply.
  • B. SWIPT: A joint beamforming design under UE energy-harvesting thresholds can maximize sum rate, but equal AP power transmission may increase interference as AP number or transmit power grows.
  • B. SWIPT: Large RIS-aided CF mMIMO networks need control protocols for energy and data transmission, with coupled beamforming and power-splitting decisions complicating design.

C. mmWave and THz

The survey discusses mmWave-related RIS-aided CF mMIMO designs, including joint AP–RIS beamforming and partially connected architectures, while identifying channel-estimation and deployment challenges. It also reviews UAV integration schemes, their measured gains, network topologies, and open coordination problems.

  • mmWave: mmWave offers large bandwidth, high data rates, directional beamforming, and interference mitigation, while RIS can further improve beamforming and coverage.
  • mmWave: Joint AP–RIS beamforming can maximize weighted sum rate under perfect CSI using Saleh–Valenzuela channels with 1 LoS and 3 NLoS paths.
  • Challenges: Accurate estimation and tracking are difficult for sparse, directional mmWave channels, while RIS deployment requires careful location planning, alignment, and cost optimization.
  • UAV: 50% higher average system rate is achieved by combining UAV and RIS, with effectiveness depending on UAV height and RIS element count.With 60 RIS elements, performance improves by 75% at 16 meters compared with 100 meters.
  • UAV: UAVs can replace RISs, APs, or CPUs, with deployments organized as conventional star or mesh networks.
  • UAV: Star networks route UAV communication through the CPU, whereas mesh networks provide direct multihop links with greater flexibility and reliability.
  • Challenges: UAV-enhanced systems still require trajectory optimization, partial-information interaction design, and joint beamforming among multiple UAVs and RISs.

A. Semantic Communications

The survey identifies semantic communications as a future direction for RIS-aided CF mMIMO, using semantic information to tailor channel processing, RIS configuration, scheduling, and resource allocation. It also discusses ISAC and SAGIN integration, emphasizing distributed sensing, propagation control, expanded coverage, and coordination challenges.

  • A. Semantic Communications: Semantic communications can use semantic information for effective resource allocation in RIS-aided CF mMIMO systems.
  • A. Semantic Communications: Semantic information such as content type and priority can guide RIS phase-shift optimization and beamforming during channel estimation and signal processing.
  • A. Semantic Communications: Context about UE location, movement, and data usage can support dynamic RIS configuration during transmission.
  • A. Semantic Communications: Semantic scheduling could allocate more resources to edge users transmitting high-priority or delay-sensitive data to ensure QoS.
  • ISAC: ISAC combines sensing and communication on shared time-frequency resources, with distributed CF APs collaboratively supporting both functions.
  • ISAC: RIS can support ISAC by controlling reflected-signal direction and power while mitigating interference between sensing and communication signals.
  • SAGIN: SAGIN combines space, aerial, and terrestrial networks to provide seamless coverage, high data rates, and reliable communication services.
  • SAGIN: CF AP distribution can extend terrestrial coverage in SAGIN, but joint trajectory planning, mobility management, adaptive beamforming, synchronization, and interference coordination remain necessary.

D. Extremely Large-scale MIMO and Near Field Communications

The survey considers integrating XL-MIMO and near-field communications with RIS-aided CF mMIMO to enhance wireless-system capabilities, while highlighting increased channel, interference, configuration, and processing complexity. It also connects the framework to secure communication and broader 6G objectives.

  • D. Extremely Large-scale MIMO and Near Field Communications: XL-MIMO uses massive antenna arrays with potential gains in spectral efficiency, data rates, and energy efficiency when integrated with RIS-aided CF mMIMO.
  • D. Extremely Large-scale MIMO and Near Field Communications: Increasing XL-MIMO antenna numbers intensifies near-field channel complexity, which RIS integration further complicates despite improving signal coverage.
  • D. Extremely Large-scale MIMO and Near Field Communications: RIS-aided CF XL-MIMO must address interference management and multipath fading alongside XL-MIMO signal-focusing capabilities.
  • D. Extremely Large-scale MIMO and Near Field Communications: Scenario adaptability becomes more complex because RIS-aided CF XL-MIMO involves high signal-processing complexity and an ultra-large-dimensional RIS–XL-MIMO channel matrix.
  • Secure RIS-aided CF mMIMO: RIS security research addresses active eavesdropping through joint AP power-control and RIS phase-shift optimization.
  • Secure RIS-aided CF mMIMO: Security and energy efficiency can be jointly designed using distributed active beamforming, artificial noise, and RIS passive beamforming.
  • Secure RIS-aided CF mMIMO: Artificial-noise-aided secure power control exploits CF spatial diversity for legitimate users while countering passive eavesdroppers.
  • Conclusions: The survey reviews system models, channel estimation, joint beamforming, resource allocation, transmission procedures, practical challenges, technology integration, applications, and future directions.
Loading 2310.00263v3…