PaperlayerView exact revision on arXiv

Source-linked AI summary

An Overview on Resource Allocation Techniques for Multi-User MIMO Systems

Eduardo Castañeda, Adão Silva, Atílio Gameiro, Marios Kountouris

arXiv:1611.04645v1cs.IT

TL;DR

MU-MIMO resource allocation is a complex joint optimization problem shaped by user channels, channel knowledge, interference mitigation, and system parameters. This survey reviews analytical, heuristic, feedback, precoding, scheduling, and power-allocation methodologies, highlighting practical trade-offs and supported design guidelines.

  • Problem

    MU-MIMO resource allocation jointly optimizes users, antennas, signaling, rates, power, and transmission techniques, while optimal scheduling and precoding remain computationally difficult.

  • Method

    The survey synthesizes analytical results, asymptotic scaling laws, heuristic algorithms, feedback strategies, precoding architectures, and resource-allocation methods for downlink MU-MIMO systems.

  • Results

    The surveyed literature provides fundamental performance limits, characterizes optimal and suboptimal operating points, and develops practical schemes balancing complexity and performance.

  • Takeaways & Limitations

    Resource allocation should balance conflicting objectives and system constraints, using practical suboptimal methods when exact optimization is infeasible.

  • Takeaways & Limitations

    Imperfect CSI can leave residual interference that makes massive-MIMO sum rate saturate at high SNR unless estimation and calibration accuracy improve.

Abstract

from arXiv · show

Remarkable research activities and major advances have been occurred over the past decade in multiuser multiple-input multiple-output (MU-MIMO) systems. Several transmission technologies and precoding techniques have been developed in order to exploit the spatial dimension so that simultaneous transmission of independent data streams reuse the same radio resources. The achievable performance of such techniques heavily depends on the channel characteristics of the selected users, the amount of channel knowledge, and how efficiently interference is mitigated. In systems where the total number of receivers is larger than the number of total transmit antennas, user selection becomes a key approach to benefit from multiuser diversity and achieve full multiplexing gain. The overall performance of MU-MIMO systems is a complex joint multi-objective optimization problem since many variables and parameters have to be optimized, including the number of users, the number of antennas, spatial signaling, rate and power allocation, and transmission technique. The objective of this literature survey is to provide a comprehensive overview of the various methodologies used to approach the aforementioned joint optimization task in the downlink of MU-MIMO communication systems.

I. INTRODUCTION

MU-MIMO resource allocation combines user scheduling, signaling, rate allocation, and power control to exploit spatial multiplexing while meeting QoS and network objectives. This survey classifies the interacting techniques, metrics, constraints, channel-information conditions, and deployment scenarios that shape downlink design.

  • MU-MIMO rationale: MU-MIMO transmits independent data streams simultaneously by exploiting spatial degrees of freedom, increasing throughput and multiplexing gains.Independent user channels provide multiuser diversity, while multiple antennas provide parallel spatial channels and reduce resource wastage.
  • Design challenges: Downlink scheduling is difficult because receiver locations and channels are random, joint detection is unavailable, and user selection depends on QoS, signaling, rate, and power decisions.Efficient allocation across heterogeneous resources requires balancing optimization quality against computational feasibility.
  • Survey scope: The survey classifies precoding, scheduling, spatial-compatibility metrics, optimization criteria, and constraints across channel-information levels, antenna settings, and single- or multi-transmitter scenarios.It also considers objective functions, transmitter coordination, and power-allocation techniques.
  • Survey approach: The survey emphasizes interactions among resource-allocation components and practical design guidance rather than detailed theory or exhaustive coverage.Mathematical formulations are largely left to the cited references.
  • Resource allocation foundations: Resource allocation assigns power, bandwidth, antennas, codes, or time slots among active users in the downlink broadcast channel.The downlink transmitter combines signals for co-scheduled users under total transmit-power constraints.
  • Performance and operating conditions: MU-MIMO mode selection depends on CSIT accuracy, allowed interference, target rates, user count, SNR, and the achievable capacity of each mode.With sufficient CSIT, MU-MIMO can provide diversity, directivity, interference mitigation, and multiplexing gains.

C. The need of User Scheduling

User scheduling selects a subset of users and assigns resources to optimize a performance metric under power and QoS constraints. The problem is combinatorial, globally coupled, and strongly dependent on channel knowledge and system assumptions.

  • User scheduling assigns radio resources to a selected user group so that a global performance metric is optimized subject to power and QoS constraints.
  • The scheduling objective can depend on each user’s channel, allocated power, data queue, encoding order, and QoS requirements.
  • The formulation is generally globally coupled and combinatorial because co-selected users change utilities, while binary scheduling variables and encoding order create discrete choices.
  • Exhaustive search can solve the scheduling problem optimally, but its computational complexity is prohibitively high.
  • Scheduling and resource allocation rely on channel knowledge, whose acquisition may use reciprocity or feedback and may be instantaneous, statistical, full, or partial.
  • Pilot contamination can arise when bandwidth and antenna constraints prevent orthogonal pilots across users, especially under universal frequency reuse.

III. MU-MIMO CHANNEL AND SYSTEM MODELS

MU-MIMO system models distinguish single- and multiple-transmitter scenarios and characterize channels through fading, covariance, correlation, and large-scale effects. In single-transmitter systems, inter-user interference is the main performance-limiting factor addressed by precoding.

  • The surveyed scenarios are classified into single-transmitter and multiple-transmitter systems, with techniques depending on transmitters, antennas, users, SNR, and CSIT accuracy.
  • The channel model includes joint transmission across users, fading, covariance determined by propagation conditions, and possible large-scale fading and Doppler effects.
  • Spatially uncorrelated Rayleigh fading assumes Σ_k = γI and corresponds physically to rich scattering with sufficient antenna spacing.
  • Correlated MU-MIMO channels can be modeled using jointly correlated or simplified Kronecker approaches, but the conventional Kronecker model may produce misleading capacity estimates in realistic scattering environments.
  • For single-transmitter MU-MIMO, inter-user interference is the principal performance-limiting factor and is addressed through precoding.
  • User distributions affect fading statistics and effective multiuser channel variances, while satellite channels can offer marginal MIMO gains because strong line-of-sight limits MIMO processing.

B. Scenarios with multiple transmitters

Multiple-transmitter MU-MIMO systems coordinate scheduling, beamforming, resource allocation, and interference mitigation across cells or transmitters. Cooperation can reduce coordination burdens compared with joint transmission, but depends on CSI, signaling, synchronization, and backhaul constraints.

  • B. Scenarios with multiple transmitters: Deploying multiple transmitters across an area supports communication for heterogeneous terminals, with resource allocation based on CSI and interference knowledge.
  • B. Scenarios with multiple transmitters: Clusters of base stations can coordinate resource allocation, scheduling, and inter-cell interference mitigation through cooperative approaches such as CoMP and network MIMO.
  • B. Scenarios with multiple transmitters: When user data is shared, coordinated base stations can jointly design coding, decoding, and other parameters using global data and CSI.
  • B. Scenarios with multiple transmitters: Coordinated scheduling with coordinated beamforming shares CSI without data sharing or signal-level synchronization, reducing information-exchange requirements relative to joint transmission.
  • B. Scenarios with multiple transmitters: CSI acquisition, control signaling, and coordinated scheduling remain challenging because feedback bandwidth and backhaul capacity are finite.
  • C. Commercial Deployments: LTE-Advanced and IEEE 802.11ac support MU-MIMO, with LTE allowing dynamic MU- or SU-MIMO switching and 802.11ac supporting up to eight streams and four users.
  • C. Commercial Deployments: At high SNR, out-of-cluster interference fundamentally limits capacity, although coordinated scheduling and user clustering can improve spectral efficiency and mitigate interference.
  • A. Non-linear Precoding with full CSIT: Precoding steers, scales, rotates, or projects independent signals before transmission to modify their power and spatial properties for a specified goal.

B. Linear Precoding with full CSIT

Linear precoding spans user-level and symbol-level designs, with practical schemes balancing interference mitigation, performance, and computational complexity. Full-CSIT precoder design is often difficult, while limited-feedback and random-beam approaches introduce feedback and quantization trade-offs.

  • Linear precoding: Linear precoding decouples input data into spatial beams and allocates power using channel state information at the transmitter.General constrained designs use optimization or heuristics such as SLNR maximization.
  • Linear precoding: Optimal precoder selection is NP-hard for many metrics, motivating suboptimal schemes that provide spatial multiplexing with lower computational complexity.Most linear precoders require M ≥ N; when M < N, interference can limit performance at moderate and high SNR.
  • User-level precoders: MRT maximizes intended-user signal power, while ZFBF suppresses inter-user interference and BD extends zero forcing to MIMO channels.Regularized channel inversion and regularized BD incorporate noise variance to improve low-SNR performance.
  • Symbol-level precoding: Symbol-level precoding addresses simultaneous transmitted symbols to different users, including constructive-interference zero forcing for MISO systems.This differs from user-level precoding, which transmits independent codewords intended for different users simultaneously.
  • Limited feedback: Limited-feedback systems use predefined codebooks or random beams to obtain channel-direction information and support spatial multiplexing with reduced transmitter knowledge.Quantization errors can create a high-SNR sum-rate ceiling, while larger codebooks improve CSIT at the cost of exponentially increasing feedback and receiver memory requirements.
  • Limited feedback: Random beamforming can asymptotically match full-CSIT downlink performance as K →∞, but it does not achieve full multiplexing gain at high SNR.Its performance is particularly discussed for K ≫ M with moderate transmitter antenna count M.

E. Precoding in LTE-Advanced

LTE-Advanced precoding combines standardized CSI feedback with dynamically updated linear precoders and practical multi-antenna architectures. The preferred approach depends on active-user density, channel information, antenna resources, and hardware constraints.

  • CSI acquisition: LTE CSI feedback includes RI for recommended stream count, PMI for codebook precoder selection, and CQI for channel quality.The codebook size should be optimized with the active-user count to balance multiuser diversity and multiplexing gains.
  • LTE precoding: With few active users, quantized-channel linear precoding outperforms codebook-based RBF or PU2RC, whereas the reverse holds when K ≫ M.Supported linear schemes include SLNR and ZFBF, whose precoders can be recalculated after CSI updates.
  • Practical implementations: 802.11n introduced sounding frames for channel estimation, while steering-matrix calculation remains implementation- and vendor-specific rather than defined by 802.11ac.Practical steering matrices can also be computed using ZFBF or MMSE.
  • Design criteria: Precoding objectives depend on CSIT accuracy, transmitter count, hardware characteristics, and the system objective, including spectral efficiency, error rates, fairness, and QoS.The survey classifies schemes and optimization criteria across single- and multi-transmitter scenarios.
  • Massive MIMO: Massive MIMO uses more transmit antennas than served users, but low-cost hardware can introduce impairments that affect performance.The cited discussion notes that excess antennas can average out hardware impairments, while TDD systems face pilot-contamination and calibration concerns.
  • Hybrid precoding: Hybrid mmWave architectures limit co-scheduled users per resource by M_RF and make sum-rate optimization non-convex under power and amplitude constraints.Hybrid precoding reduces hardware complexity but remains challenging to implement reliably and cost-effectively.

A. Null Space Projection

Null space projection measures the portion of a user’s channel that lies outside the interference subspace, directly linking channel geometry to zero-forcing gain. It therefore supports compatibility-aware grouping and scheduling, especially under ZF precoding.

  • Null space construction: For user k, the aggregated interference channels define V_k, while V_k^⊥ is their orthogonal-complement null space.Suppressing inter-user interference requires M > max_k rank(H̃_k).
  • Channel decomposition: The channel H_k decomposes into projections onto V_k and V_k^⊥, representing interference-subspace and zero-forcing directions.The V_k^⊥ component is the spatial direction free of inter-user interference.
  • Scheduling metric: The squared magnitude of the V_k^⊥ component is the null space projection and directly computes the effective channel gain obtained by ZF precoding.Large NSP components indicate semi-orthogonal, spatially compatible users.
  • Scheduling metric: For ZF-based schemes, maximizing the product of effective channel gains also maximizes sum-rate in the high-SNR regime.Some works instead maximize the sum of effective gains, with similar performance at high SNR and for K ≫ M.
  • Computation: NSP can be computed using SVD, projection matrices, Gram-Schmidt or QR decomposition, partial-correlation products, and determinant ratios.These methods support grouping and scheduling in general MU-MIMO settings.
  • Spatial clustering: Spatial clustering groups channels around orthonormal bases or codebook words, with hyperslab thresholds controlling correlation and target ε-orthogonality.The clustering parameter can be adjusted according to competing-user count, SNR, fading, and precoding scheme.

C. Compatibility between Subspaces

Subspace compatibility can be assessed through geometric and matrix-based metrics, but heterogeneous MU-MIMO requires care because spatial separation alone may miss channel-gain degradation. Resource allocation must ultimately satisfy both physical-layer and QoS-related constraints.

  • Heterogeneous channels: Heterogeneous MU-MIMO channels have different subspace dimensions, so angular or subspace-domain compatibility metrics must account for unequal dimensions.Standard normalized inner-product metrics cannot be directly applied across channels of different dimensions.
  • Geometric metrics: Orthogonality defect measures energy degradation from column correlation, while determinants represent the volume spanned by channel vectors.Condition numbers quantify eigenvalue spread and can measure spatial distance between MIMO channels.
  • Metric limitations: Principal angles and chordal distance do not fully measure compatibility in heterogeneous MU-MIMO because they may neglect useful correlation information.Metrics based only on spatial separation or eigenvalue dispersion can also miss effective channel-gain degradation.
  • Metric limitations: As active-user count grows, the set maximizing a spatial compatibility metric may differ from the capacity-maximizing set in users, cardinality, or both.This divergence limits the use of compatibility metrics as standalone capacity surrogates.
  • Grouping information: User grouping should consider both inter-user spatial correlation and each multi-antenna user’s internal eigenvector structure.Transmit-antenna correlation primarily affects precoding performance, while receive-antenna correlation has marginal or no impact according to the cited studies.
  • Resource allocation: Resource-allocation criteria are classified as physical-layer objectives using channel information alone or cross-layer objectives that also incorporate upper-layer QoS requirements.Feasibility requires satisfying individual and global constraints through precoding, power control, or both.

A. Weighted Sum Rate (WSR) Maximization

WSR-based resource allocation maximizes weighted user utilities under power and cardinality constraints. The survey relates α-fair and weighted-sum formulations to throughput-fairness tradeoffs and QoS requirements.

  • System optimization maximizes summed individual utility functions subject to power constraints and a limit on the scheduled-user set.
  • The α-fair utility family varies user priorities through α, spanning fairness, sum-rate, and minimum-rate objectives.α = 1 yields maximum fairness, α = 0 generates sum-rate utility, and α →∞ defines the minimum-rate function.
  • Weighted sum rate assigns non-negative, time-varying weights that prioritize users and can create different fairness levels.Setting all weights to one produces the sum-rate criterion.
  • Sum rate measures the maximum error-free information received by co-scheduled users, but disregards fairness.It is commonly used to assess resource-allocation effectiveness and simplify scheduling rules.
  • QoS-aware WSR incorporates network- or user-based requirements, with weights establishing service priorities and reducing allocation flexibility.QoS may be expressed through individual SINR or rate targets, or network stability over time.

1) Target SINR and Error Rates:

MU-MIMO optimization under SINR, error-rate, and queue constraints must balance individual service requirements with system-level objectives. The survey describes feasibility mechanisms and the effects of channel knowledge, interference, and competing objectives.

  • Target SINR and Error Rates: SINR requirements can enforce target error and peak-rate performance through monotonic functions such as BER or Shannon capacity.Different global-versus-individual tradeoffs arise under SINR constraints, including WSR and max-min weighted-SINR objectives.
  • Target SINR and Error Rates: For a preselected compatible user set, power allocation and precoder design solve the optimization, while scheduling is needed when constraints are infeasible.Infeasibility can require admission control or user-removal procedures to create a feasible set.
  • Target SINR and Error Rates: Transmission errors arise from noisy channels and inaccurate CSIT, including quantization, estimation errors, delay, and feedback distortion.Under partial CSIT, practical rate adaptation lies between worst-case outage and ideal mutual-information rates.
  • Target SINR and Error Rates: Queue-stability optimization requires bounded average queue lengths while exploiting available CSIT for throughput maximization.Queue-based WSR is presented as a criterion for combining these objectives.
  • Target SINR and Error Rates: System performance under QoS constraints depends on CSIT accuracy and statistics, estimation resources, spatial compatibility, SNR, service priority, and simultaneous streams.
  • Target SINR and Error Rates: Multi-objective optimization addresses conflicting goals because improving one objective can degrade others, and generally no single operating point optimizes all objectives.The survey describes solution-space sampling as a way to select an operating point satisfying a predefined tradeoff.

A. Aggregated Utility Based Selection

Aggregated utility-based selection decomposes MU-MIMO allocation into user grouping and inner precoding-power optimization. Because the grouping problem is combinatorial, practical methods use greedy construction, spatial metrics, and refinement.

  • Aggregated Utility Based Selection: The WSR formulation separates an outer user-selection problem from an inner rate optimization over precoding weights, powers, and encoding order.Under specific system settings and constraints, these subproblems can be decoupled.
  • Aggregated Utility Based Selection: The inner problem maximizes WSR for a fixed user set, requiring joint admission control, precoding design, and power control.A solution exists when the associated multiuser channel has spatial dimensions compatible with the precoding scheme.
  • Aggregated Utility Based Selection: Greedy opportunistic algorithms build the user set sequentially, selecting each new user by a locally maximizing utility or spatial-compatibility metric.Direct approaches use WSR, whereas indirect approaches use spatial compatibility.
  • Aggregated Utility Based Selection: Greedy incremental selection may retain redundant users because it cannot identify the optimal cardinality through non-iterative cumulative decisions.Delete and swap operations are used to address this redundancy.
  • Aggregated Utility Based Selection: Spatial-compatibility grouping is an NP-complete problem, and exhaustive search can be computationally prohibitive as the number of users grows.The grouping metric can be computed without precoders or powers, reducing the cost of the grouping stage.
  • Aggregated Utility Based Selection: Maximizing spatial compatibility alone does not guarantee WSR maximization in heterogeneous systems, and compatibility-metric reliability is sensitive to CSIT accuracy.Scheduling designs depend on the system model, antenna dimensions, and available channel information.

B. Metaheuristic Algorithms

Metaheuristic methods offer approximate solutions for complex, non-convex MU-MIMO allocation problems with mixed variables. Their practical use is constrained by convergence guarantees and rapidly changing wireless conditions.

  • Metaheuristic Algorithms: Stochastic optimization methods seek close-to-optimal solutions for complex, non-convex mixed problems as alternatives to classical programming.They do not generally provide mathematical convergence proofs or analytical performance results.
  • Metaheuristic Algorithms: Genetic algorithms support multi-objective optimization with continuous-discrete variables and poorly behaved non-convex solution spaces.They use populations of candidate points rather than a single starting point, reducing susceptibility to local minima.
  • Metaheuristic Algorithms: Particle swarm optimization searches through particles guided by both individual utility and swarm-level intelligence.
  • Metaheuristic Algorithms: GA and PSO iteratively sample candidate user sets, retain the best objective values, generate improved populations or swarms, and stop at a convergence criterion.Sample and elite-set sizes can limit computational complexity and speed convergence.
  • Metaheuristic Algorithms: Metaheuristics have limited practical MU-MIMO application because channel conditions, rate demands, and competing-user populations can change rapidly.

C. Classical Optimization

Classical optimization approaches formulate MU-MIMO resource allocation as convex, combinatorial, semidefinite, mixed-integer, or distributed problems involving scheduling, precoding, power, and coordination. These methods provide structured solutions but often face infeasibility, large search spaces, non-convexity, or difficult joint optimization.

  • Classical formulations jointly optimize variables such as user sets, precoders, and powers for objectives including weighted sum rate, power minimization, and fairness.
  • Convex reformulations can decouple channel assignment and signal-space selection, enabling per-subcarrier optimization of rates, powers, capacity, and consumption.
  • Semidefinite relaxation and branch-and-bound address joint precoding, power allocation, and user selection, while allowing |K| ≥ M can greatly expand complexity.
  • Admission control uses iterative user removal with SOCP-based precoder and power optimization to improve feasibility under target-SINR constraints.
  • User scheduling can be modeled through minimum set cover or sum assignment, but evaluating costs for every feasible subset increases computational requirements.
  • Distributed and coordinated multi-cell methods decouple or iteratively update user selection, coordination, precoding, and power variables under limited information exchange.

E. Scheduling for Massive MIMO

Massive MIMO changes scheduling priorities because excess antennas and channel hardening reduce the value of detailed user separability, while imperfect CSI and hardware constraints preserve important interference and complexity limits. Results indicate that selecting the number of active users and adapting to system conditions can matter more than selecting users solely by spatial compatibility.

  • Massive MIMO characteristics: Massive MIMO uses hundreds of base-station antennas to serve fewer users, providing additional spatial degrees of freedom but increasing hardware complexity and RF-chain constraints.
  • Massive MIMO characteristics: As M →∞, channel hardening makes user separability less important, so scheduling can rely more heavily on channel magnitudes.
  • Scheduling methods: JSDM groups users by channel-covariance subspaces, while conventional SUS scheduling can reach complexity O(M^3K) in large-antenna regimes.
  • Evaluation: For M = 64, direct selection is strongest near K = 32 at low SNR, whereas performance is optimized with fewer active users than antennas when K is larger.
  • Evaluation: Max-RSS selection may be efficient when M ≫|K|, while direct and indirect selection have a small gap for M ≤32 with rich multiuser diversity.
  • Practical limitations: Imperfect CSI can create interference-limited saturation, making the optimal number of served users more important than spatial compatibility alone.
  • Practical limitations: Channel-estimation overhead scales with M, making TDD more suitable for massive MIMO while FDD implementation remains an open problem.
  • Quantized channels: With quantized channels, high-SNR fixed-codebook systems become interference limited, whereas low-SNR scheduling should prioritize CQI over CDI.

B. Scheduling using RBF

RBF scheduling uses a fixed precoder codebook and channel-quality feedback to assign users to beams, then selects a beam-user subset according to a utility objective. Its benefits depend on user diversity, quantization, SNR, and QoS constraints.

  • Codebook and feedback: RBF defines precoders through a codebook and estimates a performance metric incorporating effective channel gain and quantization-induced interference.The approach differs from methods whose precoders are computed after user selection.
  • Codebook and feedback: Users estimate channels on all spatial beams, compute a CQI per beam, and report their largest CQI together with the associated CDI.The reported information supports beam-user matching at the transmitter.
  • Complexity and diversity: In the large-user regime, MUDiv scales as log(log(KN)), enabling spatial or threshold-based preselection to reduce feedback and scheduling complexity.Threshold schemes feed back only when a user’s CQI exceeds γth.
  • User selection: The transmitter selects users by assigning beams to users or by searching for a subset that maximizes a utility function such as WSR.Selection can treat receive antennas separately, assign at most one beam per user, or allocate multiple beams per user.
  • Complexity and diversity: For i.i.d. channels, each selected user cannot achieve maximum SINR on more than two beams per antenna when K > M, reducing scheduling complexity.Analytical results also state that M must scale proportionally to log(K) to avoid user starvation.
  • Limitations and extensions: Finite codebook granularity prevents full spatial separation, making inter-user interference unavoidable and particularly harmful at high SNR.When K ≈ M, RBF cannot exploit multiuser diversity effectively and performs poorly; the optimal active-beam count depends on users and SNR.

C. Two-stage Feedback Scheduling

Two-stage feedback scheduling partitions feedback between coarse channel-quality information and refined channel-direction information. The optimal partition depends on user diversity, SNR, system dimensions, and precoding, while delayed or limited feedback constrains achievable gains.

  • Two-stage design: The total feedback Bt is divided into B1 CQI bits and B2 CDI bits, separating user ranking from refined channel information for precoding.The two phases are implemented as training or pilot communication under partial CSIT.
  • CQI feedback: In the first phase, users send coarse quantized performance metrics, with B1 ∈ {1, 2, 3, 4} sometimes sufficient to attain unquantized-CQI MUDiv gains.In the large-user regime, small B1 can achieve asymptotically optimal system performance.
  • Partitioning feedback: When K ≈ M, larger B1 is needed to identify strong users, whereas high SNR favors B1 < B2 because CDI quantization errors limit performance.The CQI allocation should also match the granularity of available modulation and coding schemes.
  • CDI feedback: In the second phase, users report CDI using B2 bits, providing more accurate spatial directions or effective channel gains at the transmitter.When total feedback bandwidth is constrained, B2(k) can be assigned heterogeneously across users.
  • Adaptive feedback: Grouping users by location or long-term channel gain improves feedback-bandwidth assignment and simplifies user grouping.This is especially relevant when users have heterogeneous average long-term channel gains.
  • Practical constraints: Training and data transmission must fit within channel coherence time; otherwise, increasing B2 alone is insufficient to achieve MIMO gains.For multi-basis codebooks, large B2 may provide little capacity gain when performance is highly sensitive to MUDiv.

E. Implementation in WLAN Scenarios

WLAN and broader practical MU-MIMO implementations balance user grouping, stream assignment, computational processing, feedback overhead, and precoding constraints. These trade-offs make system design dependent on CSI availability, architecture, objectives, and channel conditions.

  • WLAN implementation: In IEEE 802.11ac, MU-MIMO identifies a transmission user set and assigns a Group ID with downlink signaling; Group ID selection resembles a two-dimensional coloring problem.A heuristic determines user positions according to occurrence probabilities.
  • WLAN implementation: Users receiving multiple streams must use the same MCS, so stream allocation requires careful design; BD with geometric-mean decomposition can enforce equal MCS allocation.The cited scheme highlights the importance of choosing the number of streams per selected user.
  • Complexity: Scheduling complexity includes implementation complexity from signaling overhead and computational complexity from transmitter or centralized-unit processing time.Quantized channel information is assumed to reduce implementation complexity.
  • CSI and codebooks: With partial CSI, computational complexity depends on codebook resolution, deployed antennas, and active users; higher resolution improves peak rates and reduces interference but bottlenecks FDD uplink feedback.Dynamic channel-based codebooks can speed allocation, reduce pairing complexity, and mitigate interference.
  • CSI and scheduling: Practical scheduling must trade signaling overhead against accurate channel-quality estimation, while CSI availability also determines SU-MIMO or MU-MIMO transmission mode and scheduling decisions.LTE-Advanced scheduling algorithms are proprietary and implementation-specific rather than standardized.
  • Open design issues: Hybrid-precoding MU-MIMO performance depends on transmitter RF-chain count and CSI accuracy, with architecture and precoding selected according to objective function and user sparsity.Joint optimization of feedback-bit and stream allocation remains a stated research direction.

A. Summary and Future Directions

The survey identifies practical and analytical challenges for MU-MIMO resource management, including energy efficiency, CSI acquisition, realistic power-control constraints, and scalable allocation. It also organizes asymptotic results by system parameters and antenna configurations while highlighting open problems in emerging deployments.

  • Open challenges: CSI acquisition and affordable-complexity algorithms remain central challenges for grouping users and allocating carriers, time slots, antennas, and transmitters.The required CSI is difficult to obtain in practical systems.
  • Open challenges: 4G MU-MIMO performance is limited because terminals lack interference estimation and typically provide only SU-MIMO CSI feedback.The reviewed works often assume feedback and capabilities that are not available in current LTE systems.
  • Open challenges: Discrete LTE downlink power control differs from the continuous power control assumed by most reviewed works.LTE uses a user-specific data-to-pilot-power offset parameter for downlink power control.
  • Future directions: Future MU-MIMO designs require novel precoding and stronger use of the multiuser dimension to deliver spectral efficiency, energy efficiency, and user satisfaction.These requirements are stated in the context of considered MIMO deployments.
  • Future directions: Massive MIMO research still needs acceptable-overhead FDD operation, pilot-contamination mitigation, efficient wideband hybrid precoding, and fast mmWave beam adaptation.The passage also identifies low-resolution and cost-effective signal processing as an open direction.
  • Analytical synthesis: Asymptotic analyses study performance limits across high- and low-SNR, large-user, large-antenna, and large-codebook regimes, with results depending on system parameters.Table VII organizes reviewed results by MISO or MIMO configuration and by the parameters analyzed.
Loading 1611.04645v1…