PaperlayerView exact revision on arXiv

Source-linked AI summary

User-centric Cell-free Massive MIMO Networks: A Survey of Opportunities, Challenges and Solutions

Hussein A. Ammar, Raviraj Adve, Shahram Shahbazpanahi, Gary Boudreau, Kothapalli Venkata Srinivas

arXiv:2104.14589v2cs.ITcs.NIeess.SPeess.SY

TL;DR

Dense transmitter deployments make interference management difficult for future mobile networks. This survey examines user-centric cell-free massive MIMO, synthesizes proposed solutions to deployment challenges, and reports that the architecture can improve spectral efficiency while alleviating cell-edge and coverage problems.

  • Problem

    Dense deployments needed for future mobile-network QoS make interference management difficult, motivating coordinated and cooperative architectures.

  • Method

    The paper surveys user-centric cell-free massive MIMO architecture, deployment challenges, proposed solutions, and management technologies across fronthaul, channel estimation, resource allocation, delay, and scalability.

  • Results

    Cell-free massive MIMO has shown large median and 95%-likely spectral-efficiency improvements over traditional networks, including five-fold and ten-fold gains over small-cell schemes under different shadow-fading conditions.

  • Takeaways & Limitations

    User-centric serving clusters can alleviate cell-edge users and non-uniform coverage while providing enhanced signal strength, connectivity, and macro-diversity-based reliability.

  • Takeaways & Limitations

    TDD is more appropriate than FDD for this network, while short coherence times constrain pilot training and data transmission within one coherence time.

Abstract

from arXiv · show

Densification of network base stations is indispensable to achieve the stringent Quality of Service (QoS) requirements of future mobile networks. However, with a dense deployment of transmitters, interference management becomes an arduous task. To solve this issue, exploring radically new network architectures with intelligent coordination and cooperation capabilities is crucial. This survey paper investigates the emerging user-centric cell-free massive Multiple-input multiple-output (MIMO) network architecture that sets a foundation for future mobile networks. Such networks use a dense deployment of distributed units (DUs) to serve users; the crucial difference from the traditional cellular paradigm is that a specific serving cluster of DUs is defined for each user. This framework provides macro diversity, power efficiency, interference management, and robust connectivity. Most importantly, the user-centric approach eliminates cell edges, thus contributing to uniform coverage and performance for users across the network area. We present here a guide to the key challenges facing the deployment of this network scheme and contemplate the solutions being proposed for the main bottlenecks facing cell-free communications. Specifically, we survey the literature targeting the fronthaul, then we scan the details of the channel estimation required, resource allocation, delay, and scalability issues. Furthermore, we highlight some technologies that can provide a management platform for this scheme such as distributed software-defined network (SDN). Our article serves as a check point that delineates the current status and indicates future directions for this area in a comprehensive manner.

I. INTRODUCTION

User-centric cell-free massive MIMO assigns each user a cooperative cluster of distributed transmitters, eliminating conventional cell edges. The architecture offers cooperation, coordination, energy efficiency, reliability, and reported spectral-efficiency gains, while requiring scalable clustering and fronthaul solutions.

  • Architecture: User-centric cell-free networks define a separate serving cluster of neighboring transmitters for each user, eliminating conventional cell edges.Clusters can be constructed using serving distance, network performance, or optimized two-stage procedures.
  • Architecture: Distributed massive MIMO deploys many access points serving few users instead of co-locating antennas, addressing non-uniform cellular coverage.Traditional co-located massive MIMO particularly suffers at cell edges and in shadowed areas.
  • Key benefits: The architecture combines cooperation, interference coordination, transmitter proximity, and macro diversity to improve connectivity, interference suppression, energy efficiency, and reliability.Distinct path loss and shadowing across serving transmitters provide macro diversity.
  • Key benefits: Five-fold and ten-fold spectral-efficiency improvements over a small-cell scheme are reported with uncorrelated and correlated shadow fading, respectively.The cited comparison concerns median and 95%-likely spectral efficiency under different scenarios.
  • Deployment challenges: Serving every user with every transmitter is impractical because distant links consume power and bandwidth while contributing little useful signal, and resource-allocation signaling is not scalable.User-centric clustering limits the number of serving transmitters to address these constraints.
  • Deployment challenges: Cell-centric clustering retains weak cell-edge signals, scales inter-cluster interference with intra-cluster signals, and produces higher delay spread than user-centric clustering.The surveyed standard remains cell-centric, but separates control and user planes, leaving user-centric deployment theoretically feasible.

A. Scope, Contributions and Related Work

This survey comprehensively reviews user-centric cell-free MIMO, emphasizing deployment challenges, proposed solutions, comparative context, and open research directions. It distinguishes the architecture through user-specific serving clusters and examines centralized and multiple-CU deployment considerations.

  • Scope and contributions: The survey compiles recent literature on user-centric cell-free MIMO and identifies proposed solutions for fronthaul, CSI estimation, clustering, resource allocation, delay, and scalability challenges.
  • Open issues: The literature reveals gaps in scalable distributed resource allocation, delay analysis for uRLLC, mobility effects on cluster re-formation, and protocols for multi-CU operation.
  • Related work: Unlike related surveys centered on cellular access architectures, this review focuses on user-centric cell-free systems with distinct serving-cluster, cell-boundary, and CSI assumptions.
  • Survey organization: The paper surveys challenge areas selected for their performance impact and categorizes topics from the literature under general challenge titles.
  • System architecture: User-centric clustering limits each user's serving transmitters, reducing DU load and signaling while distinguishing the scheme from approaches without explicitly defined user clusters.
  • System architecture: Multiple-CU deployments offer greater scalability and lower delay than centralized-CU designs through hierarchical organization, local signal collection, and relatively low fronthaul traffic.

A. Transmission Mode

User-centric cell-free networks support coherent and non-coherent transmission modes, with different synchronization, receiver, rate, and fronthaul requirements. Fronthaul management can use SDN to coordinate CUs and DUs while addressing clustering and cooperation challenges.

  • Transmission modes: Joint transmission shares user data among cooperating DUs, which simultaneously transmit it to the user.
  • Transmission modes: Dynamic point selection transmits from selected DUs, while coordinated scheduling or beamforming transmits from one DU using coordinated decisions.The latter reduces network-resource demand because user data needs to be available at only one transmitter.
  • Transmission modes: Coherent transmission uses the same data symbol across DUs and requires phase synchronization, although cyclic prefixes can relax this requirement.User-centric clustering can reduce signal delay spread by centering the user within its serving cluster.
  • Transmission modes: Non-coherent transmission independently precodes different data streams, avoids strict phase synchronization, and generally provides lower rates with a more complex SIC receiver.Its smaller serving clusters may make SIC complexity affordable at the user.
  • Transmission modes: Transmission modes impose different data-rate expressions, fronthaul capacity requirements, and usage models, with coherent joint transmission dominating the literature.
  • Fronthaul coordination: Fronthaul carries CSI, beamforming, power-control, scheduling, mobility, and user-data information; JT is the most demanding because data must reach every serving DU.
  • Fronthaul coordination: Multiple CUs require DUs to form virtual cells controlled by individual CUs, creating an open challenge for CU assignment and cooperation.
  • SDN management: SDN provides application, control, and infrastructure layers, while OpenFlow operates on the southbound interface between controllers and network nodes.

A. Minimizing Fronthaul Communication

Fronthaul communication can be reduced through compression, limited CSI exchange, partial centralization, alternative processing splits, and compute-and-forward. These choices trade fronthaul usage against performance, DU complexity, deployment cost, and coordination.

  • Compression and signaling: Source coding and data compression reduce fronthaul signaling by sending quantized data or channel estimates to the CU.
  • Compression and signaling: Quantized CSI-and-signal transmission provides slightly higher uplink rate than quantized weighted-signal transmission, with the gap shrinking as antennas per DU increase.
  • Compression and signaling: More than 7 quantization bits yields performance close to perfect fronthaul under the studied network configuration.
  • Compression and signaling: Bussgang decomposition models quantization as a power-dependent linear function plus distortion noise, with step size chosen to maximize output SDNR.
  • Compression and signaling: Quantize-and-estimate schemes send quantized received signals and pilots to the CU, while optimized quantization levels can depend on signal statistics and use Lloyd-Max algorithms.
  • Compression and signaling: Compute-and-forward uses structured lattice physical-layer network coding to reduce fronthaul traffic and increase system throughput.
  • Compression and signaling: Three compress-and-forward strategies—compress-forward-estimate, estimate-compress-forward, and estimate-multiply-compress-forward—require progressively more DU processing power.The schemes are studied under limited-capacity fronthaul and hardware impairments using rate-distortion theory for fronthaul allocation.
  • Centralization trade-offs: Reduced CSI sharing lowers fronthaul signaling but requires more DU baseband processing and can reduce performance through less coordinated operation.

B. Wired/Wireless fronthaul

The survey examines wired and wireless fronthaul designs for dense user-centric deployments, emphasizing scalability, flexibility, processing, and coordination trade-offs.

  • Fronthaul architectures: Serial fronthaul lets DUs relay data between one another, but wide deployment requires an efficient wired bus network to remain scalable.Radio stripes integrate antennas, processing, and wired CU connections inside adhesive tapes.
  • Deployment challenges: Dense DU and fronthaul deployment is logistically challenging because environments require different placement strategies, flexible links, and substantial maintenance.The illustrated environments include urban, crowded, suburban, and industrial areas; the figure is not intended to represent DU density.
  • Wireless fronthaul: Wireless fronthaul offers flexible DU placement and can support wired infrastructure, but requires careful resource tuning and may need advanced interference cancellation.It may be infeasible when many users share channel resources through beamforming.
  • Wireless technologies: Millimeter-wave fronthaul and IAB can provide wireless connectivity, but millimeter-wave fronthaul optimization remains an open research topic.IAB reuses part of the access spectrum for fronthaul or backhaul and can provide one- or two-hop wireless connections.
  • Performance and trade-offs: The user-centric approach achieves a huge sum-SE advantage over cell-free massive MIMO in the reported FSO-fronthaul study, attributed to lower interference.The FSO setup uses access-point-to-aggregation links followed by fiber to a single CU; hardware-impairment modeling outperforms clipping modeling.
  • Management and compression: Distributed SDN can manage fronthaul traffic and dynamically assign DUs to CUs, while compression choices trade DU processing load against fronthaul communication requirements.The surveyed compression scenarios are compress-forward-estimate, estimate-compress-forward, and estimate-multiply-compress-forward.

A. Favorable Propagation and Channel Hardening

The section reviews favorable propagation and channel hardening as properties relevant to CSI acquisition and signal detection, while showing that their validity depends strongly on network configuration.

  • Favorable propagation: Favorable propagation means sufficiently large serving-channel vectors can be treated as orthogonal, thereby canceling inter-user interference.The interpretation follows the large-antenna limit for independently fading user channels.
  • Channel hardening: Channel hardening occurs when effective channel gain approaches its mean, making mean-gain-based signal detection effective.The underlying channel model separates small-scale fading from large-scale fading represented by D_b.
  • Validity and limitations: The survey cautions that channel hardening may be limited to special distributed scenarios, although 5–10 antennas per DU can substantially improve it.Lower path-loss exponents and line-of-sight components can also increase hardening under the reported analysis.
  • Validity and limitations: The literature often assumes channel hardening and favorable propagation, but robust models should evaluate the corresponding conditions for each network configuration first.The survey identifies their applicability in user-centric cell-free networks as unresolved, particularly for highly distributed deployments.
  • Simulation findings: In the reported simulation, co-located antennas harden channels most, followed by distributed antennas with correlated and then independent shadowing.Lower path-loss exponents improve hardening for distributed deployments, but other configurations may yield different profiles.

B. Pilot Assignment

Pilot assignment (PA) manages pilot contamination by separating co-pilot users, but must balance estimation quality against overhead, computation, and limited coherence time.

  • A well-structured PA policy can reduce pilot-sequence length by 3–3.75 while achieving negligible contamination.Negligible contamination is defined as a 3% decrease in average SE.
  • Rate-based PA can impose substantial overhead and computation, making repeated pilot reassignment and channel re-estimation impractical within limited channel coherence time.
  • Location-based clustering assigns orthogonal pilots within user groups, with further optimization possible inside each group.The groups can be formed using HAC based on user locations.
  • Downlink PA can assign orthogonal pilots to users with higher pilot utility while allowing others to decode using statistical CSI.The utility depends on Doppler spread, channel-hardening degree, and prioritization weights.
  • Non-orthogonal PA is relevant to crowd scenarios with many users and intermittent access, including IoT and cell-free networks underlaid with D2D communication.
  • Covariance-aided Bayesian estimation and pilot-power control are proposed to improve CSI acquisition, while semi-blind estimation remains an open direction.

C. High-mobility Users

High mobility creates channel-aging and serving-cluster-update problems that can increase signaling, delay, and effective-throughput losses; the required countermeasures remain incompletely studied.

  • High-mobility users remain an open research topic because rapidly changing channels create CSI mismatch between estimation and data transmission.This mismatch is termed channel aging.
  • Time-varying channel models can incorporate channel aging through temporal autocorrelation linked to propagation geometry, velocity, frequency, and antenna characteristics.Jakes’ and autoregressive models are identified as prominent examples.
  • Frequent serving-cluster updates for mobile users generate signaling overhead and delay that affect effective throughput.Conventional LTE handover latency is about 45–50 ms, while cell-free networks may require longer durations.
  • Mobility-dependent handover schemes and trace-based or synthetic mobility models are candidate tools for studying and improving mobility management.Suggested evaluation metrics include handoff rate, sojourn time, and handoff probability.
  • Under channel aging, cell-free massive MIMO outperforms small-cell networks in both static and mobile scenarios in the cited study.Fractional power control is also used to improve system performance.
  • Channel prediction and synchronization-error mitigation require further investigation for user-centric clustering, particularly when massive-MIMO channel properties may not apply.
  • FDD studies report more than 60% lower uplink feedback overhead than conventional CSI feedback, while antenna calibration remains insufficiently studied.Angle reciprocity may be less effective at millimeter-wave frequencies because uplink and downlink carriers can be far apart.

V. FORMATION OF SERVING CLUSTER

Serving clusters make cell-free deployment practical by limiting each user’s cooperating transmitters, reducing resource and signaling burdens while introducing clustering-performance trade-offs and allocation challenges.

  • V. FORMATION OF SERVING CLUSTER: Serving clusters address DU capacity, processing, fronthaul, and CSI-overhead constraints, while excluding transmitters whose path loss makes their contribution negligible.
  • A. Utility-based Clustering: Clustering metrics trade complexity against performance optimality, with the appropriate balance depending on network conditions and service type.
  • A. Utility-based Clustering: Adding DUs does not necessarily improve users already receiving sufficient signal strength and can instead introduce additional costs.
  • A. Utility-based Clustering: Clustering strategies range from static to semi-dynamic and dynamic, creating different trade-offs between overhead and interference cancellation capability.
  • A. Utility-based Clustering: User-centric clustering can use large-scale-fading thresholds followed by scheduling or weighted-sum-rate optimization, but the dynamic process is more complex than cell-centric clustering.
  • B. Lessons Learned: Explicit clusters limit users served per DU, computational complexity, CSI estimation, DU and fronthaul load, and path-loss-induced inefficiency.
  • B. Lessons Learned: Clusters may be formed from offline or online parameters, and utility-based frameworks can prioritize different metrics through static, semi-dynamic, or dynamic operation.
  • VI. RESOURCE ALLOCATION: Resource-allocation studies target QoS objectives including spectral efficiency, energy efficiency, and transmit-power minimization under constraints.Common decision variables include beamformers, user scheduling, and allocated power.

B. Energy Efficiency (EE)

Energy-efficiency research in cell-free MIMO optimizes power, cooperation, and deployment choices under QoS and hardware constraints, while distributed deployment creates distinctive accounting and management issues.

  • Cell-free massive MIMO outperforms its collocated counterpart in energy efficiency under a QoS constraint.
  • Energy efficiency is defined as data rate divided by power consumption, with radiated and circuit power commonly included.
  • Area-based EE can optimize pilot reuse and DU density; a suitable pilot reuse factor can lower interference and increase EE per unit area up to a specific value.
  • EE per unit area is needed for cell-free networks because users receive joint transmission from many DUs, unlike conventional single-BS service.
  • Cell-free and millimeter-wave integration is considered promising because cooperation can alleviate millimeter-wave connectivity problems.
  • EE optimization can jointly control power, beamforming, transmit covariance, active DUs, antenna activation, and DU–user association.
  • Distributed deployment changes EE trade-offs through shorter service distances and lower BS transmission-power budgets, while the number of DUs also affects EE.

C. Distributed Approaches

Distributed resource allocation reduces per-node complexity and supports scalability, while different decomposition and game-theoretic methods address optimization trade-offs. Reported studies also examine federated learning, scheduling, and distributed beamforming, with performance depending on coordination architecture.

  • Distributed resource allocation: Distributed resource allocation reduces computational complexity per node and better accommodates growing system size and fronthaul load.More network resources become available as the system grows, helping compensate for increased complexity.
  • Game-theoretic approaches: Game-theoretic methods support distributed utility maximization but may require costly feedback, converge away from sum-rate optima, and require simplified formulations.Auction and matching approaches are among the related distributed tools.
  • Decomposition methods: Decomposition methods split optimization into master and derivative problems, with optimality available when the original problem is convex and separable.For uplink max-min SINR, related work separates receiver-filter design from power allocation.
  • Energy-efficiency optimization: A three-stage scheme maximizes rate, then energy efficiency, and finally minimizes power under the rate constraint when earlier solutions fail the requirement.Fractional programming and bisection are used in the successive subproblems.
  • Distributed learning: Federated learning over cell-free massive MIMO reduces training time by up to 55% over baseline approaches, although communication-overhead latency is not studied.Training proceeds through iterative sharing and aggregation with the central unit.
  • Distributed scheduling and beamforming: CU-distributed scheduling and beamforming achieve 90% of centralized performance and provide 1.3- to 1.8-fold higher network data rate than DU-distributed schemes.The corresponding DU-distributed schemes achieve 72% of centralized performance.

D. Lessons Learned

The surveyed lessons emphasize practical trade-offs among beamforming, energy efficiency, connectivity, and scalability in dense cell-free deployments. Performance and deployment choices depend on interference suppression, fronthaul constraints, access-point density, and coordination architecture.

  • Beamforming: Conjugate beamforming improves tractability and scalability but does not explicitly suppress interference; LSFD, local MMSE, and weighted MMSE are alternatives.These alternatives may be preferable in practice, particularly without channel hardening.
  • Propagation and density: Increasing access-point density from 128 to 1024 per km2 raises the probability of line-of-sight communication from 40% to 95%.
  • Power control: Power-control algorithms balance fairness, latency, and throughput, but the latency–data-rate trade-off remains unaddressed.
  • Energy efficiency: Energy efficiency can be improved through fractional programming, Dinkelbach optimization, transmit-power adaptation, energy harvesting, and sleep modes.LSFD is promising for uplink operation but has been studied only with a single central unit.
  • Millimeter-wave integration: Combining user-centric cell-free networks with millimeter-wave communication may address millimeter-wave connectivity limitations while supporting dense transmitter deployment.Millimeter-wave short range and large bandwidth are presented as compatible with dense cell-free deployments.
  • Scalability: Distributed allocation scales better than centralized allocation, but game and auction methods can introduce feedback requirements, delay, and non-optimal equilibria.

VII. LATENCY AND SYNCHRONIZATION

Latency and synchronization are important deployment challenges because distributed serving and multiple central units can delay delivery and distort coherent transmission. The surveyed solutions include queue-aware optimization, edge computing, punctured scheduling, coding techniques, and timing protocols, but several applicability gaps remain.

  • Latency: Data-delivery latency is rarely studied in cell-free massive MIMO, motivating delay-aware transmission control and scheduling.
  • Latency: Lyapunov optimization jointly controls uplink power and scheduling while managing queue admission to reduce average network delay.
  • Latency-aware technologies: Latency-critical services motivate MEC, short-packet communication, HARQ-IR, finite-blocklength coding, and punctured scheduling.Punctured scheduling trades uRLLC latency against eMBB rate loss.
  • Synchronization: Serving users with multiple DUs can increase signal-delivery delay, especially when DUs are controlled by different CUs.
  • Synchronization: Distributed DU timing differences increase signal delay spread and inter-DU carrier-frequency offset, while data-synchronization errors can worsen the power delay profile.
  • Synchronization: Over-the-air synchronization and calibration achieved sufficient accuracy for satisfactory performance, but the system was designed for cell-centric clustering or small cells.
  • Synchronization: Coherent transmission is more affected by synchronization errors than non-coherent transmission, requiring relative time and phase synchronization among DUs.Quasi-synchronous systems, cyclic prefixes, and IEEE 1588v2 PTP are proposed mitigation directions.

C. Lessons Learned

The lessons learned frame latency and scalability as distinct but connected design concerns. Queue-aware methods, uncertainty-aware decision processes, synchronization techniques, distributed allocation, and careful cluster sizing are proposed, while scalability itself remains inconsistently defined.

  • Latency: uRLLC prioritizes latency sensitivity over data rate, making delay-aware solutions important for cell-free communication.
  • Latency: Proposed delay-reduction strategies include MEC, punctured scheduling, short packets, HARQ-IR, finite-blocklength coding, and limiting serving-cluster size.
  • Latency: Queueing theory, MDPs, dynamic programming, and Lyapunov optimization provide tools for studying latency in coordinated networks.
  • Latency: POMDPs can model uncertainty in punctured-scheduling states and derive strategies that optimize the relevant parameters.
  • Synchronization: Distributed DU time-of-arrival differences increase delay spread and carrier-frequency offset, making synchronization important for coherent transmission.
  • Scalability: Scalability lacks a precise rigorous definition, although one formulation requires finite per-DU complexity as the number of served users approaches infinity.
  • Scalability: A scalable LSFD and clustering-based pilot assignment were proposed, but a single CU may become a fronthaul and processing bottleneck.
  • Scalability: A distributed resource-allocation system can scale better than a centralized one because growing system size brings additional resources to handle complexity.

B. Intelligent Reflecting Surfaces

Intelligent reflecting surfaces (IRSs) are presented as low-cost, passive structures that can improve channel quality and energy efficiency in user-centric cell-free networks. The surveyed literature highlights joint DU beamforming and IRS reflection optimization, while machine learning remains underexplored.

  • Intelligent Reflecting Surfaces: IRSs use controllable passive phase shifts to coherently reflect signals and enhance coverage, unlike active relays with complex signal processing.Their reflected channels can be combined with DU-user channels, and the phase-shift matrix can be optimized.
  • Intelligent Reflecting Surfaces: Joint optimization of DU transmit beamformers and IRS reflection coefficients is studied under limited-capacity fronthaul constraints.The coupled optimization problem is described as difficult to solve directly.
  • Intelligent Reflecting Surfaces: More than 2-fold gain in the reported metric is obtained by varying IRS transmit power, density, and size in the surveyed study.The supplied passage reports the gain but does not identify the metric in its excerpt.
  • Intelligent Reflecting Surfaces: Machine learning applications in user-centric cell-free networks remain limited, leaving performance gains for beamforming, allocation, association, and mobility management to be quantified.Existing examples include deep convolutional neural networks for uplink power control and reinforcement learning for antenna selection.
  • Intelligent Reflecting Surfaces: IRS deployment can enhance energy efficiency at lower cost and power consumption than denser DU deployments.The survey identifies low deployment cost and low operating power as central motivations for IRS-aided cell-free networks.

IX. POSSIBLE FUTURE RESEARCH DIRECTIONS

The survey identifies open directions spanning architecture, fronthaul, channel estimation, mobility, multi-CU processing, and full-duplex operation. These directions address limited investigation, scalability, pilot overhead, changing channels, and coordination constraints.

  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: Multiple-CU architectures remain less investigated than single-CU designs, especially regarding fronthaul effects and adapting LSFD.The survey calls for comparisons under the same network assumptions.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: Distributed SDN could dynamically associate DUs with CUs and migrate DUs between neighboring CUs according to fronthaul capacity.The survey connects programmable management with the difficulty of managing dense DU deployments.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: Non-uniform fronthaul quantizers, including Lloyd-Max designs and companders, are proposed to reduce distortion when signal levels are skewed.The passage contrasts these approaches with uniform quantizers and relates them to rate-distortion analysis.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: Wireless fronthaul, millimeter-wave fronthaul, and integrated access and backhaul require further performance analysis, including system-level IAB simulations.Wireless fronthaul capacity varies with parameters such as DU locations, while IAB can provide spectrum where dedicated spectrum is unavailable.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: FDD channel estimation needs fewer pilots, while downlink estimation of individual DU channels becomes problematic as the serving-cluster size grows.The survey also identifies semi-blind CSI estimation as an open cell-free topic and cites gains reported in massive MIMO studies.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: Distributed power-allocation optimization is needed because centralized optimization may be infeasible in actual deployments.The survey also identifies CSI-sharing limits as important for scalable beamforming.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: High mobility remains insufficiently investigated because channel aging, pilot orthogonality loss, frequent cluster redefinition, and handoff complications affect performance.The survey suggests robust pilot designs, predictive stable handoffs, and redefined handoff metrics as research directions.
  • IX. POSSIBLE FUTURE RESEARCH DIRECTIONS: The benefits and operating conditions of in-band full-duplex communication in cell-free networks remain open research questions.The survey presents full-duplex operation as a less-visited topic requiring scenario- and parameter-specific investigation.

D. Delay, Serving Cluster, Other Areas

The survey emphasizes delay, mobility, management, IRS, and scalability as unresolved aspects of user-centric cell-free networks. It presents distributed control, IRS integration, and learning-based optimization as promising directions while retaining important deployment constraints.

  • D. Delay, Serving Cluster, Other Areas: Delay-related studies are rare, although queue-state information and multi-CU coordination are important for resource allocation and synchronization.The survey specifically proposes studying CU density and transmission synchronization for high-mobility users.
  • D. Delay, Serving Cluster, Other Areas: Cell-free communication may support uRLLC reliability, but the data-rate–delay tradeoff and multiple-CU MEC implementations require investigation.Punctured scheduling is also identified as an unstudied topic for user-centric cell-free networks.
  • D. Delay, Serving Cluster, Other Areas: Distributed SDN and SON platforms are proposed to manage serving-cluster construction and scalable network operation in dense DU deployments.SDN can support dynamic CU-DU association, but complete network models and solutions remain necessary.
  • D. Delay, Serving Cluster, Other Areas: IRSs are expected to enhance energy efficiency, with frequency-selective channels identified as a future direction.The survey frames IRS deployment as a novel topic in cell-free networks.
  • D. Delay, Serving Cluster, Other Areas: Machine learning is proposed for real-time decisions after offline training, addressing computational complexity in optimization frameworks.The survey notes that its treatment of machine learning is limited and directs readers to dedicated surveys.
  • D. Delay, Serving Cluster, Other Areas: TDD is considered more appropriate than FDD because channel reciprocity can reduce pilot requirements, although short coherence times still challenge both modes.FDD downlink pilots depend on the number of DUs, while TDD also suffers when coherence times are short.
  • D. Delay, Serving Cluster, Other Areas: Millimeter-wave and cell-free systems are described as complementary: cell-free improves connectivity, while millimeter wave offers short range and large bandwidth for dense deployment.The survey associates their combination with a wireless Gbit/s experience.
  • D. Delay, Serving Cluster, Other Areas: High-mobility performance and latency remain rarely investigated, particularly for channel aging, pilot design, cluster formation, synchronization, and uRLLC.These topics are identified as priorities for future work.
Loading 2104.14589v2…