Source-linked AI summary
Structured Massive Access for Scalable Cell-Free Massive MIMO Systems
Shuaifei Chen, Jiayi Zhang, Emil Björnson, Jing Zhang, Bo Ai
TL;DR
B5G networks need scalable ways to serve more users at higher data rates with stringent QoS, while scalable massive access remains unresolved in cell-free massive MIMO. The paper proposes a structured massive-access framework combining access, decoding, pilot assignment, power control, and closed-form SE analysis, and reports improved pilot schemes, scalable decoding, and a fairness–SE trade-off.
Problem
Scalable massive access implementation remains a vital unresolved practical issue in cell-free massive MIMO systems serving increasingly many users.
Method
The paper develops a structured massive-access framework with scalable initial access, P-LSFD, two pilot assignment schemes, fractional power control, and closed-form MR-combining SE expressions.
Results
User-Group and IB-KM pilot schemes offer 22.4% and 5.9% improvement in 95%-likely SE, while P-LSFD’s reduction versus optimal LSFD is marginal at 1.7% when K = 60.
Takeaways & Limitations
The framework provides a feasible solution for structured massive access, with scalable decoding and controllable fairness–average-SE adjustment.
Takeaways & Limitations
The study assumes error-free fronthaul connections and does not focus on fronthaul provisioning, including limited fronthaul capacity.
Abstract
from arXiv · showhide
How to meet the demand for increasing number of users, higher data rates, and stringent quality-of-service (QoS) in the beyond fifth-generation (B5G) networks? Cell-free massive multiple-input multiple-output (MIMO) is considered as a promising solution, in which many wireless access points cooperate to jointly serve the users by exploiting coherent signal processing. However, there are still many unsolved practical issues in cell-free massive MIMO systems, whereof scalable massive access implementation is one of the most vital. In this paper, we propose a new framework for structured massive access in cell-free massive MIMO systems, which comprises one initial access algorithm, a partial large-scale fading decoding (P-LSFD) strategy, two pilot assignment schemes, and one fractional power control policy. New closed-form spectral efficiency (SE) expressions with maximum ratio (MR) combining are derived. The simulation results show that our proposed framework provides high SE when using local partial minimum mean-square error (LP-MMSE) and MR combining. Specifically, the proposed initial access algorithm and pilot assignment schemes outperform their corresponding benchmarks, P-LSFD achieves scalability with a negligible performance loss compared to the conventional optimal large-scale fading decoding (LSFD), and scalable fractional power control provides a controllable trade-off between user fairness and the average SE.
I. INTRODUCTION
Cell-free massive MIMO is presented as a promising B5G technology for serving more users with higher data rates and stringent QoS, but scalable massive access remains challenging. The paper addresses this gap with a structured access framework covering access, decoding, pilot assignment, and power control.
- Motivation: Cell-free massive MIMO uses distributed APs connected to a CPU to jointly serve UEs through coherent transmission and reception.Its user-centric design lets subsets of APs serve individual UEs.
- Motivation: Macro-diversity can improve coverage probability, while user-centric AP cooperation manages interference and accommodates more UEs than cellular networks.Cellular networks are limited by inter-cell interference and pilot shortage.
- Challenge: Limited pilot resources force reuse among UEs, causing pilot contamination that degrades channel estimation and interference rejection.Proper pilot assignment is therefore critical, especially when K is roughly the same as L.
- Challenge: Practical implementation requires each AP’s signal-processing complexity and resource requirements to remain finite as K →∞.Earlier joint initial-access, pilot-assignment, and power-control methods were not designed for massive access with L ≈K.
- Objective: The paper designs a structured massive-access framework including initial access, data decoding, pilot assignment, and power control while accounting for imperfect CSI, SE, user density, and fairness.The framework targets scalable cell-free massive MIMO networks.
- Related work: Prior scalable pilot assignment was evaluated mainly for L ≫K, while several alternative methods were not scalable or did not optimize pilot assignment.The paper focuses on suppressing pilot contamination in structured access.
B. Main Contributions
The paper proposes a scalable structured massive-access uplink framework for cell-free massive MIMO, combining access selection, decoding, pilot assignment, power control, and closed-form SE analysis. Its components are designed to support many UEs while addressing interference, scalability, fairness, and imperfect CSI.
- Decoding: The framework proposes scalable P-LSFD for multi-antenna APs with roughly the same performance as the optimal alternative.P-LSFD is paired with a closed-form SE expression using MR combining.
- Initial access: A competitive-mechanism algorithm enables many UEs to access the network and select appropriate serving APs.The user-centric AP selection is represented through UE–AP association subsets.
- Pilot assignment: User-Group and IB-KM pilot assignment schemes partition UEs to suppress mutual interference from pilot sharing.Their design targets structured massive access.
- Power control: A scalable fractional power-control policy provides a fairness–average-SE trade-off adjustable through a parameter.The paper evaluates this structured access framework numerically.
- Analysis: Two novel closed-form SE expressions are derived with MR combining for fixed pilot assignments and random pilot switching.The system model includes K single-antenna UEs, L APs, and user-centric AP selection.
A. Pilot Transmission and Channel Estimation
The section develops uplink transmission and decoding for cell-free massive MIMO, including achievable SE, LSFD weighting, and a scalable partial-LSFD alternative.
- The CPU combines locally processed AP estimates through LSFD to decode each UE's data signal.
- The achievable SE for UE k is log2 (1 + SINRk), with SINRk determined by the uplink combining and interference terms.
- Optimal LSFD maximizes SINRk through a generalized Rayleigh quotient, but its fronthaul load and computational complexity grow rapidly with network size.
- P-LSFD restricts weighting to UEs sharing APs with UE k, limiting the relevant interferer set independently of the total UE count.
- The proposed P-LSFD is scalable because its complexity does not grow with K, while LP-MMSE expectations require Monte-Carlo computation rather than closed-form evaluation.
III. INITIAL ACCESS AND AP SELECTION
The initial-access procedure assigns serving APs through capacity-aware competition while ensuring each UE receives at least one AP under the one-UE-per-pilot assumption.
- Each AP serves at most one UE per pilot and uses all its N antennas for those UEs.
- The competitive mechanism guarantees every UE at least one serving AP while preventing weak-channel UEs from being abandoned.
- The algorithm lets UEs select APs using large-scale fading, with APs admitting additional UEs until their pilot-specific capacity reaches τp.
- When an AP is full, it compares competing UEs and prioritizes those with stronger channel conditions.
- After competitions and fallback assignments, the procedure repeats until all UEs finish AP selection and constructs the association matrix A.
IV. PILOT ASSIGNMENT
The pilot-assignment section motivates structured assignments as a way to suppress pilot contamination in massive access and introduces two corresponding schemes.
- Proper pilot assignment improves system performance by suppressing pilot contamination, particularly in massive access scenarios.
- The section derives a novel closed-form SE expression for random pilot switching.
- It identifies drawbacks of random pilot assignment and proposes two pilot assignment schemes dedicated to suppressing pilot contamination.
A. Random Pilot Assignment and Random Pilot Switching
Random pilot assignment and switching determine pilot-sharing relationships, enabling closed-form MR-based SE analysis while exposing strong pilot-contamination cases.
- With random pilot assignment, each UE receives one of τp orthogonal pilots and uses that pilot in all blocks.
- Random pilot switching changes the pilot-sharing UEs across blocks to average over pilot contamination.
- A binary variable χik indicates whether UE i shares a pilot with UE k, simplifying the statistics used in the derivation.
- The analysis derives a closed-form SINR expression for MR combining with random pilot switching and obtains a corresponding SE corollary.
- The closed-form SE is treated as a worst case because random pilot switching can expose all UEs to strong pilot contamination.
- Structured pilot assignment avoids occasional pilot sharing between nearby UEs, which otherwise creates strong mutual interference.
B. Interference-Based K-Means Pilot Assignment Scheme
The Interference-Based K-Means (IB-KM) pilot assignment separates UEs into clusters using interference-related distance information, then pairs UEs across clusters to reuse pilots while limiting pilot contamination. It improves on centroid-only geographic clustering by considering UE-level interference relationships.
- Interference rationale: UEs with disjoint serving-AP subsets generate less inter-user interference when sharing a pilot than UEs with common serving APs.The example gives Dis12 = 9150 > Dis13 = 315 because UE 1 and UE 3 share AP 3, whereas UE 1 and UE 2 do not.
- Interference metric: Disik measures the difference between the service quality received by UEs i and k from their corresponding APs.Smaller Disik indicates stronger potential inter-user interference when the UEs share a pilot.
- Cluster construction: IB-KM separates K UEs into ⌈K/τp⌉ disjoint clusters whose centroids have maximized minimum Dis values.Each cluster contains at most τp UEs selected by their smallest Dis values to the corresponding centroid.
- Pilot reuse: After forming clusters, IB-KM assigns mutually orthogonal pilots within one cluster and matches UEs in other clusters to reuse those pilots.When multiple UEs compete for a match, the one with the largest Dis value is selected, and the remaining UEs seek other matches.
- Scope and motivation: Unlike centroid-centric K-means assignment, IB-KM still risks pairing cluster-edge UEs served by similar AP subsets, motivating direct UE-level separation.The paper therefore introduces User-Group pilot assignment as a user-centric alternative.
C. User-Group Pilot Assignment Scheme
User-Group pilot assignment constructs serving, interference, and grouping relationships from AP-UE associations, then forms disjoint UE groups with minimal common serving APs. A threshold δ controls the number of groups, allowing adjustment to the pilot budget.
- Design principle: User-Group assigns mutually orthogonal pilots to UEs served by fewest common APs, reducing pilot contamination among pilot-sharing users.Pilot contamination occurs when pilot-sharing UEs communicate with the same AP; fewer common serving APs reduce the contamination caused by sharing a pilot.
- Relationship matrices: Matrix S retains the strongest AP-UE serving relationships, with threshold δ determining how many relationships remain and how many groups result.The retained relationships are selected from sorted large-scale fading coefficients.
- Relationship matrices: Matrix T records UE interference relationships, where zero entries identify UE pairs served by few common APs.The relevant serving sets satisfy M′k ∩ M′i = ∅ for zero-valued entries.
- Grouping procedure: Matrix G organizes candidate group members from the zero entries of T, and the grouping procedure forms disjoint UE groups under explicit membership constraints.The paper illustrates this construction with five UEs and nine APs.
- Pilot-budget control: The grouping procedure produces M disjoint groups for a given δ, and bisection adjusts δ until M = τp.As δ decreases, retained serving relationships shrink, increasing the chance that UEs have no common serving APs.
D. Online Complexity Analysis
The online complexity differs substantially between IB-KM and User-Group assignment. IB-KM uses centroid selection and cross-cluster matching, whereas User-Group constructs three UE/AP relationship matrices.
- IB-KM complexity: IB-KM operates by assigning K UEs to ⌈K/τp⌉ centroids after centroid locations are determined.Centroid locations can be computed offline from AP geography, so that step can be omitted from online complexity.
- IB-KM complexity: Online IB-KM complexity is driven by UE-to-centroid selection and matching one UE from each other cluster to share each pilot.The resulting complexity is described after accounting for these two online operations.
- User-Group complexity: User-Group pilot assignment has complexity O(KL + K^2L + K/2).This includes computation involving matrices S, T, and G.
- Scaling comparison: Under L ≈ K ≫ τp, IB-KM has a more attractive complexity scaling than User-Group.The considered massive-access setting assumes the number of pilots is far smaller than the numbers of APs and UEs.
V. SCALABLE FRACTIONAL POWER CONTROL
The proposed scalable fractional power control policy locally minimizes large-scale SIR variance while compressing received-power differences. Its parameter θ controls the fairness–average-SIR trade-off.
- Policy objective: The proposed fractional power control policy is scalable and locally minimizes the variance of the large-scale SIR.The local-average desired signal power uses only large-scale fading coefficients from APs selected by each UE.
- Power-control parameter: The data transmission power pk is controlled by θ ∈ [0, 1], which determines how strongly received-power ranges are compressed.The policy is derived for the local-average desired signal.
- Trade-off: Smaller θ favors average SIR, whereas larger θ promotes user fairness.Thus θ provides a controllable trade-off between these objectives.
VI. NUMERICAL EVALUATION
The numerical evaluation tests the structured massive-access components under defined cell-free massive MIMO settings. The proposed initial access, pilot assignment, P-LSFD, and power-control designs improve performance, preserve scalability, and expose fairness–SE trade-offs.
- Initial access: The proposed initial access algorithm outperforms the benchmark across all four pilot assignment schemes for K = 40 and K = 60 UEs.Its competition mechanism allows each UE to be served by as many APs as possible under the scalability condition, although the advantage becomes less prominent at higher UE density.
- Power control: Fractional power control trades average SE against fairness: smaller θ improves average SE, whereas larger θ promotes fairness among UEs.Average SE also decreases as UE density increases, while the SE range is insensitive to the number of UEs.
- Overall comparison: User-Group and IB-KM outperform GB-KM and random assignment in both UE fairness and average SE.The analytical results from Lemma 2 closely match the simulations, and LP-MMSE achieves higher SE than MR combining.
VII. CONCLUSION
The conclusion presents a structured massive-access framework for scalable cell-free massive MIMO that addresses access, decoding, pilot assignment, and power control. Simulations show improved SE and a scalable P-LSFD trade-off, while the framework’s scope remains focused on SE, user density, and fairness.
- Framework and motivation: The framework targets structured massive access in scalable cell-free massive MIMO, where pilot contamination from pilot sharing is a central issue.It combines a scalable initial access algorithm with User-Group and IB-KM pilot assignment schemes.
- Analysis: The framework evaluates SE with LP-MMSE and MR combining while accounting for UE density and fairness, and derives two closed-form MR-combining SE expressions.The analysis focuses on the uplink, with similar results expected in the downlink due to channel reciprocity.
- Initial access: The proposed initial access algorithm enables each UE to be served by as many APs as possible under the scalability condition.This design is presented as part of a feasible solution for structured massive access.
- Scalable decoding: P-LSFD sacrifices only 1.7% in 95%-likely SE at K = 60 relative to optimal LSFD, providing an acceptable scalability trade-off.The performance reduction grows with user density.
- Power control: Scalable fractional power control provides a trade-off between fairness among users and average SE.The framework therefore offers controllable performance along these two objectives.
- Scope: The study focuses on SE under user-density and fairness considerations, leaving energy efficiency, hardware impairment, and limited fronthaul for future analysis.The proposed geometry-based schemes may also extend to multi-antenna UEs, but exact details remain future work.