PaperlayerView exact revision on arXiv

Source-linked AI summary

Millimeter Wave Cellular Networks: A MAC Layer Perspective

Hossein Shokri-Ghadikolaei, Carlo Fischione, Gabor Fodor, Petar Popovski, Michele Zorzi

arXiv:1503.00697v4cs.ITmath.OC

TL;DR

Highly directional mmWave links create new MAC-layer challenges involving blockage, deafness, synchronization, access, mobility, and resource allocation. The paper presents design approaches and an integrated MAC-layer view, finding that fully-directional operation can substantially improve performance while leaving the operating regime for noise-limited systems as an open question.

  • Problem

    Highly directional mmWave channels introduce blockage and deafness, complicating link establishment and making conventional interference-limited MAC assumptions inapplicable.

  • Method

    The paper identifies key mmWave MAC-layer challenges and develops design approaches spanning beamforming, synchronization, random access, handover, scheduling, and dynamic cellular operation.

  • Results

    Fully-directional mode generally outperforms other modes; compared with omnidirectional operation in experiment 10, sum rate increases by factors of 113 and 75 in experiments 3 and 9.

  • Takeaways & Limitations

    Directionality improves link budget and reduces interference, while spatial resources must be allocated jointly with time and frequency resources in MAC design.

  • Takeaways & Limitations

    The conditions defining the noise-limited regime in terms of densities, transmission powers, beamwidths, and traffic remain an open topic for future study.

Abstract

from arXiv · show

The millimeter wave (mmWave) frequency band is seen as a key enabler of multi-gigabit wireless access in future cellular networks. In order to overcome the propagation challenges, mmWave systems use a large number of antenna elements both at the base station and at the user equipment, which lead to high directivity gains, fully-directional communications, and possible noise-limited operations. The fundamental differences between mmWave networks and traditional ones challenge the classical design constraints, objectives, and available degrees of freedom. This paper addresses the implications that highly directional communication has on the design of an efficient medium access control (MAC) layer. The paper discusses key MAC layer issues, such as synchronization, random access, handover, channelization, interference management, scheduling, and association. The paper provides an integrated view on MAC layer issues for cellular networks, identifies new challenges and tradeoffs, and provides novel insights and solution approaches.

I. INTRODUCTION

mmWave cellular networks create distinctive MAC-layer challenges because directional, blockage-sensitive links blur cell boundaries and complicate control, access, mobility, and resource management. The paper surveys these challenges and proposes approaches spanning synchronization, random access, control channels, and resource allocation.

  • Paper scope: The paper identifies principal MAC challenges and develops novel approaches for three aspects of mmWave cellular communications.Its scope includes synchronization, random access, handover, channelization, interference management, scheduling, and association.
  • Control plane: A directional control plane is necessary in mmWave bands, while an omnidirectional microwave control channel can significantly improve control-plane performance.The paper highlights options for realizing directional control and explains the benefit of cross-band control signaling.
  • Initial access and mobility: The proposed framework combines microwave and mmWave control channels through two-step synchronization and favors contention-based initial access as beams narrow.The paper also proposes a MAC signal to address prolonged random-access backoff and discusses mobility management and handover.
  • Resource allocation: Directional pencil beams expand cell-formation and resource-allocation options while simplifying interference management, but introduce tradeoffs among throughput, fairness, and connection robustness.The paper formulates long-term resource allocation and examines additional RF chains as a source of gains with limits.
  • Directed-channel challenges: mmWave’s sparse, directional channels introduce blockage and deafness, making conventional cell boundaries and MAC procedures less applicable.Obstacles can require alternative beam searches, while misaligned beams prevent link establishment but reduce interference.

B. Heterogeneity

mmWave cellular networks combine heterogeneous spectrum and deployments with directional beamforming, creating new control, channel, and resource-allocation requirements. Beamforming can compensate attenuation and suppress interference, but implementation choices trade flexibility and performance against hardware, power, and CSI overhead.

  • Heterogeneity: mmWave networks may combine high-frequency data links with lower-frequency control links and coexist across macrocell, microcell, femtocell, and picocell deployments.The resulting spectrum and deployment heterogeneity supports stand-alone and integrated network scenarios.
  • Beamforming: Beamforming compensates severe channel attenuation, reduces interference, and supports multiplexing or spatial diversity through multi-beam directional operation.Fully directional pencil beams reduce inter-cell interference in both downlink and uplink.
  • Noise-limited operation: Reduced multiuser interference does not necessarily make the network noise-limited because channel-establishment and maintenance overhead may instead limit performance.The qualification is particularly relevant when interpreting resource-allocation and interference-management results.
  • Digital beamforming: Fully digital beamforming offers high beam-shaping flexibility but requires one RF chain per antenna element, increasing hardware cost, complexity, and power consumption.Its channel-estimation complexity also scales at least linearly with the number of transmitter antenna elements.
  • CSI and implementation constraints: CSI acquisition can be constrained by limited UE power, weak uplink reference-signal beamforming gains, and the channel-reciprocity requirement in TDD systems.Reciprocity holds when duplexing time is much shorter than channel coherence time.
  • Hybrid beamforming: Hybrid digital-analog beamforming uses many antennas with fewer RF chains by applying digital precoding to an effective channel formed through analog weights and the physical channel.Analog beamforming supplies directivity and spatial-division gains, while digital beamforming can reduce intra-sector interference and enable multiplexing.

III. REALIZATION OF PHYSICAL CONTROL CHANNELS

Realizing a physical control channel (PHY-CC) in mmWave cellular networks requires balancing fallback reliability against directional efficiency, coverage, interference, and spatial-search overhead. The paper analyzes these options quantitatively and discusses their implications for control-plane design.

  • B. Available Options and Design Aspects: PHY-CC design involves fallback between microwave and mmWave bands, and directionality choices among omnidirectional, semi-directional, and fully-directional communication.Microwave channels offer larger coverage and stability, while mmWave channels enable a single transceiver but face attenuation and blockage.
  • B. Available Options and Design Aspects: Directional control channels increase coverage and reduce interference, but require spatial search that adds delay and alignment overhead.Fully-directional communication provides the greatest coverage and interference reduction at the expense of the highest spatial-search overhead.
  • B. Available Options and Design Aspects: With path-loss exponent 3, semi-directional communication with 16 dBi gain increases range roughly 3.5×, while 30 dBi combined gain enables a 10× coverage gain.The paper notes that one fully-directional BS can cover an area requiring up to 100 omnidirectional BSs.
  • B. Available Options and Design Aspects: The paper compares PHY-CC options using a Poisson-distributed BS network and a typical UE receiving strong signals from line-of-sight BSs.It also considers hierarchical combinations of options, including a novel two-step synchronization procedure.
  • B. Available Options and Design Aspects: At α = 3 and one LoS BS in a 250x250 m2 area, Options 1, 2, and 3 cover 63.6%, 99.9%, and 100% of the area, respectively.With LoS BS density 2×10^-6, Option 3 covers 99.8% while Option 2 covers 60%; the extra coverage requires more complicated alignment.

IV. INITIAL ACCESS AND MOBILITY MANAGEMENT

Initial access in mmWave cellular networks must handle directional synchronization, gain asymmetry, and mobility across short-range, blockage-sensitive cells. The section uses an illustrative network scenario to frame these MAC-layer design challenges.

  • 1) Synchronization and Cell Search:: Figure 5 illustrates UE1 performing initial access and UE2 requiring handover within a macrocell containing three microcells and an obstacle.The coverage boundaries are idealized to simplify the discussion.
  • 1) Synchronization and Cell Search:: Initial access consists of synchronization, system-information extraction, and random access before the UE can transmit and receive data.The process registers the UE as active and connects it to the data plane.
  • 1) Synchronization and Cell Search:: Unlike LTE’s omnidirectional synchronization, mmWave cell search must account for directional propagation and beamforming.LTE UEs know when and where synchronization signals are transmitted, whereas mmWave procedures require spatial considerations.
  • 1) Synchronization and Cell Search:: A 30 dBi combined directivity gain can make the data range at least 4× larger than the synchronization range even under severe attenuation.This gain asymmetry makes an omnidirectional mmWave PHY-CC unsuitable for initial cell search in the described setting.
  • 1) Synchronization and Cell Search:: Directional synchronization and cell search using semi-directional or fully-directional mmWave control channels can provide the needed spatial information, but spatial search may delay system-information acquisition.Microwave synchronization cannot provide sufficient spatial synchronization for the mmWave band because the propagation characteristics differ.

2) Extraction of System Information:

The paper proposes directional control and a two-step synchronization procedure to support system information extraction, random access, and handover in mmWave networks. Directionality reduces contention but creates deafness and spatial-search overhead, motivating tailored access and mobility mechanisms.

  • Random Access: Contention-based random access becomes more justifiable as beamwidths narrow because directionality reduces contention, although deafness can cause repeated backoffs.A proposed MAC signal addresses prolonged backoff caused by deafness during random access.
  • Synchronization: A two-step procedure first synchronizes entities in time and frequency through an omnidirectional microwave control channel, then supports directional cell search and access.The macrocell BS broadcasts periodic synchronization signals, after which entities extract directional information and proceed with system information and random access.
  • Synchronization: Semi-directional synchronization marginally outperforms fully-directional synchronization in average discovery epochs and epochs required to guarantee a target discovery probability.Fully-directional operation needs more epochs when its search space per epoch is smaller; increasing its search space narrows the difference.
  • Synchronization: For α = 3.5 and discovery probability 0.99, both options must search all Ns = 18 sectors at θ = 20° or all Ns = 6 sectors at θ = 60°.Under these conditions, semi-directionality has no advantage over full directionality, motivating selection of a suitable fully-directional beamwidth.
  • Mobility Management and Handover: Pencil-beam operation complicates mobility because obstacles and channel changes can trigger frequent handovers, each potentially adding spatial synchronization overhead.The paper suggests maintaining connections to several BSs and using beam tracking or fast switching to reduce reassociation delay.
  • Mobility Management and Handover: A reliable microwave-band control channel can facilitate handover negotiations, while more efficient directional control channels can perform periodic connection checks.This combines robust control signaling with directional operation for recurring link monitoring.

V. RESOURCE ALLOCATION AND INTERFERENCE MANAGEMENT

Directional mmWave MAC design expands resource allocation from time and frequency into space, enabling spatial reuse and higher throughput. The resulting gains require beam-aware grouping, coordination, and explicit management of throughput, fairness, robustness, and hardware constraints.

  • Resource Allocation: MmWave resource blocks extend LTE’s time-frequency definition into the time-frequency-space domain because directional beams add spatial resources.Digital beamforming would require precise CSI and substantial pilot complexity, motivating hybrid approaches.
  • Resource Allocation: UE groups can share one analog beamforming vector when their channel covariance matrices are similar, while spatial scheduling changes less frequently than time-frequency scheduling.Time-frequency decisions are recalculated each channel coherence time and bandwidth; spatial decisions follow meaningful covariance changes.
  • Scheduling: UEs belonging to multiple groups create coupled scheduling decisions that require cooperation between serving stations.A shared UE can link group schedules across a BS and relay station, complicating resource allocation.
  • Fully-directional Communications: Fully-directional operation increases degrees of freedom by allowing colocated multi-antenna UEs to use different analog beams without additional BS hardware complexity.Beam management can also separate colocated UEs into different effective groups rather than assigning them together.
  • Scheduling: 60, 120, 120, and 240 resource blocks are achieved by traditional, semi-directional, semi-directional-with-relay, and fully-directional scheduling, respectively.The 120-resource-block cases arise from spatial division at the BS, while 240 results from additional spatial division at multi-antenna UEs.
  • Resource Allocation: The paper formulates long-term resource allocation to improve the throughput-fairness tradeoff while guaranteeing a minimum QoS level.Additional RF chains can improve throughput, fairness, and minimum UE rate, but gains have limits.

C. Interference Management

Highly directional beams reduce interference and enable spatial reuse, but mmWave networks still require interference-aware association and resource allocation. Dynamic, user-centric cells are proposed to balance throughput, fairness, QoS, and blockage robustness under changing network conditions.

  • Interference Management: Pencil-beam operation mitigates intra-cell and intra-group interference through spatial orthogonality and suitable scheduling and beamforming.Inter-cell interference may also be substantially reduced by time-frequency-space scheduling with fully-directional communication.
  • Interference Management: Inter-cell interference coordination may be unnecessary in mmWave networks under rare-interference conditions, with on-demand management available when needed.Analog beamforming can further minimize inter-group interference, whereas omnidirectional control channels still require careful pilot and message design.
  • Dynamic Cell: Traditional RSRP/RSSI association can become inefficient in non-uniform directional deployments, motivating dynamic cells formed by beam-served UE groups.Dynamic cells need not contain colocated UEs and can be redefined to improve an objective function.
  • Dynamic Cell: Dynamic cell formation can meet UE QoS demands, improve the fairness-spectral-efficiency tradeoff, and place each UE in at least two groups for blockage robustness.The approach requires network state covering traffic demands, QoS, dynamic formations, and neighboring connectivity.
  • Dynamic Cell: A traffic-driven reassignment can move a UE between cells by changing analog beamforming vectors, even without mobility or blockage.The paper distinguishes this dynamic reconfiguration from reassociation after handover and notes that its benefit depends heavily on interference.
  • Performance: Compared with omnidirectional experiment 10, fully-directional experiments 3 and 9 increase sum rate by factors of 113 and 75, respectively.Minimum offered spectral efficiency increases by factors of 207 and 43 in the same comparisons; results average 10 random topologies.
  • Performance: Increasing RF chains by a factor of 4 improves fully-directional sum-rate performance by a factor of 3.2 while also improving the minimum achievable rate.The paper reports additional gains from hardware-enabled degrees of freedom, with limits as directionality is applied at BSs and UEs.

VI. CONCLUDING REMARKS

The paper concludes that mmWave cellular networks require MAC-layer redesign around directionality, spectrum heterogeneity, and new tradeoffs. It presents solution approaches for control channels, synchronization, scheduling, association, interference management, and future noise-limited operation.

  • Concluding remarks: mmWave networks offer improved area spectral and energy efficiencies but require fundamental MAC-layer changes because of high attenuation, blockage vulnerability, directionality, and limited interference.The paper examines synchronization, random access, handover, channelization, interference management, scheduling, and association.
  • Concluding remarks: A hierarchical control-plane architecture combines reliable omnidirectional microwave signaling with efficient directional mmWave signaling for initial access and synchronization.The proposed two-step procedure uses microwave coverage and reliability together with directional mmWave cell search.
  • Concluding remarks: Directional pencil beams create flexible cell and resource configurations while simplifying interference management, but introduce tradeoffs among throughput, fairness, and connection robustness.The paper frames these tradeoffs through long-term resource allocation and scheduling.
  • Concluding remarks: Proper fully-directional scheduling with limited RF chains can improve throughput and UE fairness over omnidirectional operation.Additional RF chains improve sum rate, minimum UE rate, and Jain’s fairness index, with gains saturating differently for semi-directional and fully-directional modes.
  • Concluding remarks: Future analysis should identify when mmWave networks are noise-limited using UE and BS densities, transmission powers, beamwidths, and UE traffic.The paper notes that noise-limited operation can facilitate concurrent transmissions and simplify MAC intelligence for spectrum sharing and interference avoidance.

APPENDIX A: SPATIAL SEARCH OVERHEAD

Appendix A models spatial-search overhead for directional synchronization using LoS base stations, sectorized antennas, and distance-dependent attenuation. It derives discovery probabilities and shows how beamwidth, BS density, and propagation assumptions shape discovery delay.

  • Model and assumptions: The analysis bounds spatial-search delay assuming nb LoS BSs can deliver synchronization pilots to a typical UE and that all BSs sequentially scan Ns sectors.Each BS transmits one directional pilot per epoch in a random sector order.
  • Discovery model: The probability model combines the distribution of discoverable LoS BSs with the probability that scanned sectors include the UE’s direction.The appendix derives joint and marginal discovery probabilities using a zero-truncated Poisson distribution and sector scanning.
  • Discovery metrics: The appendix gives closed-form metrics for average discovery epochs, discovery within l epochs, guaranteed discovery within Ns epochs, and the minimum epochs needed for probability µ.Discovery is guaranteed after scanning all Ns = ⌈2π/θ⌉ sectors.
  • Limitations: The conclusions depend on assumptions about antenna radiation patterns and channel models, including a sectored antenna approximation and distance-dependent attenuation with α > 2.Preliminary Nakagami-fading results are mentioned as applying the insights while changing the exact expressions.
  • Model and assumptions: The sectored antenna and distance-dependent path-loss models connect beamwidth and directivity to pilot reception range and LoS-BS density.The effective area is πdmax^2 for semi-directional communication and θdmax^2/2 for fully-directional communication.
  • Results: Given discoverability, semi-directional control channels require fewer discovery epochs on average than fully-directional control channels.Fully-directional reception covers a smaller effective sector area, yielding lower discoverable LoS-BS density under the stated model.

APPENDIX B: OPTIMAL CELL FORMATION

The appendix formulates optimal cell formation as a network-utility optimization over association, resource allocation, beam orientations, and beamwidths. It characterizes fairness, communication-mode relationships, and the limitation that an efficient solution method remains future work.

  • Optimization formulation: The model uses long-term SINR and rate allocation, with narrower beams generally increasing SINR through higher received power and lower interference.Fast fading is averaged out because association operates on a longer timescale than instantaneous channel fluctuations.
  • Optimization formulation: The optimization jointly selects BS–UE associations, resource shares, and BS and UE beam orientations and beamwidths to maximize network utility.The formulation assumes known network topology and uses binary association variables with resource-allocation variables.
  • Optimization formulation: Constraint (9c) enforces association with only one BS, while constraint (9d) enforces minimum UE QoS and restricts positive resource shares to associated UEs.The resulting solution provides a long-term association policy together with beam orientation and resource allocation.
  • Limitations and extensions: The paper focuses on fundamental limitations and leaves development of an efficient solution method for the cell-formation optimization problem to future work.Backup associations can be incorporated so a UE can use its backup connection until a new solution enables handover to the appropriate BS.
  • Fairness properties: Replacing yij with equal shares within each analog beam yields micro-level fairness, while logarithmic utility ensures macro-level proportional fairness.The equal allocation is expressed through the reciprocal of the number of associated UEs served by a BS.
  • Communication modes: The optimal utility is upper bounded in the omnidirectional mode by the semi-directional mode, and in the semi-directional mode by the fully-directional mode.These bounds follow because each less-directional mode has a feasible solution set contained within the more-directional mode's feasible set.
Loading 1503.00697v4…