PaperlayerView exact revision on arXiv

Source-linked AI summary

Massive Access for 5G and Beyond

Xiaoming Chen, Derrick Wing Kwan Ng, Wei Yu, Erik G. Larsson, Naofal Al-Dhahir, Robert Schober

arXiv:2002.03491v2cs.ITeess.SP

TL;DR

Massive access for B5G cellular IoT must support many low-power devices with broad coverage despite short packets, shared resources, and challenging channels. This paper surveys theories, protocols, access techniques, coverage, energy, security, and related research directions. It concludes by synthesizing existing results and identifying open challenges across these aspects.

  • Problem

    Massive access requires information-theoretic and wireless-network designs beyond conventional finite-user, long-code multiple access because IoT involves massive transmitters and short packets.

  • Method

    The paper provides a comprehensive survey of massive access in B5G wireless networks across theory, protocols, techniques, coverage, energy, security, and future research directions.

  • Results

    The survey synthesizes massive random and short-packet access theories, grant-free protocols, orthogonal and non-orthogonal techniques, coverage methods, and associated design issues.

  • Takeaways & Limitations

    Massive access design must jointly address low-complexity short codes, large-dimensional processing, energy consumption, coverage, and analytical limitations of learning-based methods.

Abstract

from arXiv · show

Massive access, also known as massive connectivity or massive machine-type communication (mMTC), is one of the main use cases of the fifth-generation (5G) and beyond 5G (B5G) wireless networks. A typical application of massive access is the cellular Internet of Things (IoT). Different from conventional human-type communication, massive access aims at realizing efficient and reliable communications for a massive number of IoT devices. Hence, the main characteristics of massive access include low power, massive connectivity, and broad coverage, which require new concepts, theories, and paradigms for the design of next-generation cellular networks. This paper presents a comprehensive survey of aspects of massive access design for B5G wireless networks. Specifically, we provide a detailed review of massive access from the perspectives of theory, protocols, techniques, coverage, energy, and security. Furthermore, several future research directions and challenges are identified.

I. INTRODUCTION

Massive IoT growth exposes the limits of existing wireless access and motivates massive access designs for B5G cellular networks. The paper surveys this area comprehensively across multiple theoretical and practical perspectives.

  • Motivation: IoT device growth is projected from 8.4 billion connected devices in 2017 to more than 75.4 billion by 2025, with hundreds of billions and 10 million devices per km2 envisioned by 2030.These projections drive the need for scalable wireless access.
  • Motivation: Zigbee, Bluetooth, and WiFi provide short-range access for moderate device populations but cannot ultimately support reliable access for massive numbers of IoT devices.Cellular IoT reuses existing cellular infrastructure and is presented as a more beneficial and economical LPWAN option than LoRa for service providers.
  • Research Challenges: Massive access is difficult because conventional information theory targets few devices, while IoT commonly uses short packets to reduce latency and decoding complexity.The paper identifies a lack of information-theoretic concepts directly suited to massive access.
  • Research Challenges: Massive access differs from eMBB and URLLC because cellular IoT must address low power, massive connectivity, and broad coverage rather than primarily high data rates or ultra-reliable low latency.These differing objectives require access designs tailored to cellular IoT.
  • Contribution: Earlier surveys focused on individual perspectives, whereas this paper provides a comprehensive overview spanning theory, protocols, techniques, coverage, energy, security, and future directions.The survey is organized around the aspects illustrated in Fig. 1.

II. MASSIVE ACCESS IN CELLULAR IOT

Cellular IoT combines sporadic, small-payload, low-power, widely distributed devices with heterogeneous service requirements. Massive access must therefore overcome wireless-channel bottlenecks involving spectrum, coherence time, coverage, and CSI.

  • Wireless Access: Limited radio spectrum forces multiple devices to share bandwidth through multiple access techniques, whose performance also depends on channel conditions and device requirements.Access design must coordinate shared resources while accounting for fading, interference, noise, and QoS.
  • Service Characteristics: Cellular IoT devices exhibit sporadic traffic, small payloads, low power, ubiquitous distribution, limited capability, stringent latency constraints, and heterogeneous QoS requirements.These characteristics distinguish massive access from conventional cellular access.
  • Wireless-Channel Challenges: The last-mile wireless channel is a major bottleneck because limited spectrum and short coherence times constrain frame length, CSI acquisition, and massive-access performance.When CSI is unavailable, non-coherent transmission may be used but suffers performance degradation compared with ideal coherent transmission.
  • Wireless-Channel Challenges: Non-coherent transmission can address scenarios where full CSI is unavailable, but it incurs performance degradation compared to ideal coherent transmission.This trade-off becomes relevant for short-coherence-time or high-mobility applications such as smart traffic.

III. MASSIVE ACCESS THEORIES

Massive-access theory extends conventional multiple-access analysis to settings with many transmitters, random activity, and short packets. The reviewed results cover capacity scaling, activity-identification cost, and extensions to MIMO channels.

  • Classical MAC limitations: Conventional MAC theory becomes inadequate as the number of transmitters grows, because per-transmitter capacity approaches zero and long codes conflict with IoT requirements.Co-channel interference dominates with massive transmitter counts, while conventional capacity assumes infinitely long codes.
  • Massive-access requirements: Massive-access analysis must account for more than 10 million devices per km2, random device activity, and short-packet transmission.These requirements motivate information-theoretic guidelines distinct from conventional MAC results.
  • Many-access channel: The many-access channel models transmitter count and coding blocklength increasing together, using random coding and Feinstein threshold decoding under a maximum power constraint.When transmitter count grows sublinearly with blocklength, each transmitter can communicate with arbitrarily small error probability for sufficiently large blocklength.
  • Random activity: Random activity adds an activity-identification cost equal to the entropy H2(α) of the activity probability.The activity-identification stage detects active transmitters before data decoding; the cost is independent of the numbers of transmit antennas and receive antennas.
  • MIMO extensions: MIMO massive-access theory extends the channel model to multiple antennas at the base transceiver station and IoT devices, with capacity expressions involving channel matrices and active-transmitter sets.The individual capacity term describes a transmitter’s capacity when activity information is available at the receiver.

B. Massive Short-Packet Access

Massive short-packet access addresses finite-blocklength IoT communication, where low latency and bursty traffic make asymptotic error-free transmission difficult. Approximate finite-blocklength rates enable performance evaluation, but available capacity results remain limited, especially for fading channels.

  • Motivation: Short packets reduce latency for bursty cellular-IoT communications, but make it difficult to guarantee error-free reception at finite blocklength.Existing asymptotic results often assume coding blocklength grows with the number of transmitters and permit packet error rates approaching zero.
  • Finite-blocklength theory: The finite-blocklength rate R*(M, ϵ, P) generalizes established capacity results, recovering Shannon capacity as M →∞ and the diversity-multiplexing tradeoff as P increases.The exact massive short-packet capacity remains challenging to derive in closed form.
  • Rate approximation: A tight approximation based on the inverse Gaussian Q function and channel dispersion evaluates achievable rates after substituting the massive-access SINR.As blocklength increases, the approximation reduces to the Shannon capacity formula.
  • Open limitations: Available massive-access capacity results remain limited because most theoretical studies consider Gaussian channels, whereas fading and channel estimation change the relevant capacities.For fading channels, short-packet capacity essentially reduces to outage capacity.

IV. MASSIVE ACCESS PROTOCOLS

Massive-access protocols coordinate sporadically active IoT devices seeking cellular access. Grant-based random access is simple but vulnerable to collisions, latency, and signaling overhead as device numbers grow.

  • IV. MASSIVE ACCESS PROTOCOLS: Random access protocols coordinate data exchange between active IoT devices and the BTS under randomly varying device activity.Only active devices request access because IoT traffic is sporadic.
  • A. Grant-Based Random Access: Grant-based random access, adopted in 5G NB-IoT, requires active devices to obtain BTS permission before accessing the network.
  • A. Grant-Based Random Access: The grant procedure uses four transmissions: preamble selection, BTS authorization, connection request, and contention resolution with resource allocation.A preamble selected by only one active device can lead to a granted request; collisions are not granted.
  • A. Grant-Based Random Access: A finite pool of orthogonal preambles creates high collision probability and access failure when many IoT devices contend simultaneously.Short coherence time limits the number of available orthogonal preamble sequences.
  • A. Grant-Based Random Access: Four required transmissions increase signaling overhead, while collision-driven access failures can make average access latency intolerably high.

B. Grant-Free Random Access

Grant-free random access removes the grant stage by combining preamble-based activity detection with direct data transmission. Its central challenge is recovering massive, structured device-state signals under nonorthogonal preambles, noise, and computational constraints.

  • B. Grant-Free Random Access: Grant-free protocols let active devices transmit unique preambles and then data directly, reducing access latency and signaling overhead.The BTS must detect active devices from received preambles without granting access first.
  • B. Grant-Free Random Access: Nonorthogonal preambles and short packets cause severe co-channel interference, requiring sophisticated activity-detection algorithms at the BTS.
  • 1) CS Formulation:: Compressed sensing formulates massive device detection as sparse signal recovery from noisy preamble measurements.The received preamble is sparse because only sporadically active devices transmit.
  • 1) CS Formulation:: The sensing matrix contains device preamble sequences, while the device-state matrix represents activity indicators and channel responses.The CS formulation uses a zero-norm sparsity objective with a Frobenius-norm reconstruction constraint.
  • 1) CS Formulation:: Detection methods approximate the nonconvex zero-norm problem using optimization, greedy, and Bayesian approaches with different complexity-performance trade-offs.Greedy methods simplify implementation but can degrade under noise; Bayesian methods exploit prior channel information.
  • 1) CS Formulation:: High-dimensional device-state matrices can be simultaneously sparse and low-rank, offering another route to reduce detection complexity.Nonconvex l_p minimization may improve activity-detection accuracy but usually has high computational complexity.

2) Covariance Formulation:

Covariance-based detection formulates massive device detection as maximum-likelihood estimation when only device activity is required and the BTS has many antennas. The approach can also jointly recover activity and data, but remains limited by complexity and preamble length.

  • 2) Covariance Formulation:: When channel estimation is unnecessary and the BTS has many antennas, device activity can be formulated as a maximum-likelihood problem based on received-signal covariance.
  • 2) Covariance Formulation:: Covariance-based detection can jointly identify activity and data by assigning each device a unique sequence set whose transmitted sequence encodes data.Detecting the received sequence provides both activity and data information.
  • 2) Covariance Formulation:: Massive device detection remains open because recovery algorithms can be computationally expensive and preamble sequences may be too long for short-packet transmission.

C. Unsourced Random Access

Unsourced massive random access uses a shared codebook rather than device-specific preambles, placing device identity inside each message. The broader massive-access discussion contrasts conventional orthogonal allocation with approaches that add resources or improve utilization efficiency.

  • C. Unsourced Random Access: Unsourced massive random access gives all devices one shared codebook instead of assigning each device a unique preamble sequence.
  • C. Unsourced Random Access: Devices include their identities in the information messages, and the BTS decodes the active-message list up to permutations.
  • C. Unsourced Random Access: Conventional orthogonal multiple-access techniques allocate each time-frequency resource block to one device, making massive access difficult under limited spectrum.
  • C. Unsourced Random Access: B5G massive access can pursue additional wireless resources or improve resource-utilization efficiency.

A. Massive Orthogonal Access

Massive access requires new orthogonal-access designs that exploit spatial and frequency resources because conventional approaches face limited theory, feedback, energy, and bandwidth. Massive MIMO and higher-frequency operation offer capacity and spatial gains, but introduce substantial CSI, hardware, propagation, and interference-management challenges.

  • A. Massive Orthogonal Access: Massive access requires extra spatial and frequency-domain resources because strict latency constraints make time-domain resources scarce.Conventional orthogonal access over limited spectrum cannot satisfy massive-access QoS requirements.
  • 1) Massive MIMO: Massive MIMO can improve active-device detection and transmission performance while asymptotically suppressing co-channel interference through channel hardening.The cited work also reports significant spectral- and energy-efficiency improvements.
  • 1) Massive MIMO: Conventional FDD quantized feedback becomes prohibitive because each device requires many bits for its high-dimensional channel vector.Sparse-CSI compression can reduce feedback, but its applicability is limited to sparse channels and FDD systems.
  • 1) Massive MIMO: TDD avoids downlink CSI feedback by exploiting channel reciprocity, but pilot reuse causes pilot contamination and reduces CSI-estimation accuracy.Limited pilot-sequence length forces reuse across many devices, creating co-channel interference.
  • 1) Massive MIMO: Massive MIMO hardware can consume substantial energy because large antenna arrays require many RF chains and ADC modules.Hybrid precoding reduces RF-chain requirements but increases design complexity, while ADC cost remains a vital issue.
  • 2) Millimeter-Wave/Terahertz: mmW and THz bands provide additional spectrum for massive access, but severe propagation loss shortens transmission distance and complicates CSI acquisition and precoding.Sparse channels permit compressed-sensing and Bayesian CSI methods.
  • B. Massive Non-orthogonal Access: NOMA lets multiple devices share one time-frequency resource block, potentially admitting more devices than OMA at a given bandwidth and per-device spectral efficiency.Its massive-access operation is limited by severe co-channel interference and the need for interference management.

1) Power-Domain Non-Orthogonal Multiple Access:

The survey examines non-orthogonal access and coverage-enhancement mechanisms for supporting massive connectivity under constrained resources and weak links. These approaches increase spectrum sharing or coverage, but often require sophisticated processing, careful resource allocation, or additional infrastructure.

  • 1) Power-Domain Non-Orthogonal Multiple Access:: PD-NOMA uses power-weighted superposition coding and successive interference cancellation to separate devices sharing the same resource.Power allocation affects interference and fairness, with higher power intuitively assigned to devices with smaller channel gains.
  • 1) Power-Domain Non-Orthogonal Multiple Access:: PD-NOMA improves spectral efficiency but can impose prohibitive computational complexity and signal-processing delay when SIC is performed for all devices.Device clustering reduces complexity by restricting SIC to smaller groups, although frequency-domain clustering decreases spectral efficiency.
  • 2) Code-Domain Non-Orthogonal Multiple Access:: CD-NOMA schemes use sparse codes or codebooks to multiplex devices, enabling message-passing detection and spreading-based interference suppression.The survey discusses LDS-CDMA, LDS-OFDM, MUSA, and SCMA as code-domain approaches.
  • B. Massive Non-orthogonal Access: Both PD-NOMA and CD-NOMA exploit additional channel-sharing degrees of freedom, but high-dimensional channels can make transceiver complexity prohibitive.The survey identifies simple but effective transceiver design as an important future research topic.
  • VI. MASSIVE COVERAGE ENHANCEMENT: Cell-edge and indoor IoT signals are weak because devices transmit at low power, while existing coverage-enhancement schemes consume resources inefficiently.Rural deployment further limits the coverage of current cellular IoT.
  • A. Cell-Free Massive MIMO: Cell-free massive MIMO distributes access points across an area and connects them to a central processing unit, shortening access distances and broadening coverage.The distributed antenna system serves users through one or multiple access points rather than partitioning the area into cells.
  • A. Cell-Free Massive MIMO: Cell-free massive MIMO improves 95%-likely per-user throughput by nearly fivefold under uncorrelated shadow fading and ten-fold under correlated shadow fading versus a small-cell architecture.The passage presents it as a promising low-power strategy for outdoor coverage enhancement.
  • A. Cell-Free Massive MIMO: Optimal cell-free performance requires allocating wireless resources across access points because each access point has a different impact on the overall system.A max-min power-control scheme was proposed to provide equal throughput for all users.

B. Intelligent Reflecting Surface

Indoor and wide-area massive access requires coverage enhancement because propagation loss and missing rural cellular infrastructure limit connectivity. The paper reviews IRS-based indoor beamforming and multi-beam LEO satellite access, while noting deployment complexity and cost.

  • Indoor IoT deployment faces weak received signals because walls attenuate wireless propagation, making coverage enhancement crucial.
  • Indoor coverage enhancement: An intelligent wall uses sensors, a cognitive engine, and an active frequency-selective surface to control reflections and improve radio coverage.A reconfigurable reflect-array, also called an intelligent reflecting surface (IRS), is one implementation.
  • Indoor coverage enhancement: IRS phase-shift control enhances desired signals and cancels interference through spatial beamforming across many reflecting units.This can significantly improve received-signal quality for many indoor wireless devices.
  • Rural coverage enhancement: Rural IoT applications lack cellular coverage, while deploying new rural cellular networks is prohibitively expensive, motivating satellite communications.
  • Rural coverage enhancement: Multiple-beam LEO satellites can serve many devices simultaneously and provide low-latency, reliable access across large rural areas.Beamforming studies include robust power-minimizing designs under imperfect CSI and cooperative multicast transmission.
  • Deployment considerations: Combining coverage strategies such as cell-free massive MIMO and IRS can improve hotspot signal quality but increases implementation complexity and cost.

A. Wireless Energy Transfer

Massive access is constrained by limited IoT battery capacity, difficult battery replacement, and security risks from shared spectrum. The paper reviews wireless energy transfer, physical-layer security, and open challenges including long-distance charging, mobility, and secure CSI acquisition.

  • Wireless energy transfer: Most IoT devices are battery powered, and limited battery capacity forces very low transmit power, such as 23 dBm.Higher transmit power requires frequent battery replacement, creating human and environmental costs.
  • Wireless energy transfer: Energy beamforming uses spatial beamforming to address wireless energy transfer losses caused by path loss and channel fading.
  • Wireless energy transfer: Long-distance wireless energy transfer remains an open research problem because low transfer efficiency makes its effective distance too short for broad coverage.
  • Access security: Shared-spectrum massive access creates information-leakage risks, and grant-free random access can make key-distribution-based encryption inapplicable.Physical-layer security complements conventional encryption by exploiting fading, interference, and noise.
  • Access security: Multiple-antenna techniques improve secrecy by strengthening legitimate signals and reducing eavesdropper signals; with full CSI, linear precoding can make leakage asymptotically tend to zero.The paper notes that acquiring the eavesdroppers’ CSI remains a challenge.
  • Future challenges: Mobility causes fast time-varying fading that makes accurate CSI acquisition difficult and causes frequent handoffs.

B. Modulation and Coding

B5G massive access must support sporadic short-packet traffic, high-dimensional received signals, and stringent reliability and latency requirements. The paper identifies short FEC codes, big-data analytics, machine learning, and integrated sensing-computation-communication as important directions.

  • B. Modulation and Coding: Sporadic IoT traffic favors short packets, requiring short forward-error-correction codes for massive access.The paper identifies modulation and coding schemes as central to both efficiency and reliability.
  • B. Modulation and Coding: Massive access produces large data volumes and high-dimensional received signals, increasing the burden on B5G networks.Methods for big-data analytics and large-dimensional processing are therefore needed to improve efficiency.
  • B. Modulation and Coding: URLLC remains difficult because massive access causes severe co-channel interference while short packets have high decoding error rates.
  • Future research directions: Machine learning is being applied to resource allocation, signal processing, channel estimation, and transceiver design to reduce network design complexity while maintaining high performance.
  • Future research directions: Jointly designing sensing, computation, and communication can reduce resource consumption by reusing sensed-signal transmissions for over-the-air computation.
  • Survey scope: The survey covers massive-access theory, protocols, access techniques, coverage extension, energy, security, and future research challenges.
Loading 2002.03491v2…