PaperlayerView exact revision on arXiv

Source-linked AI summary

Towards Massive Connectivity Support for Scalable mMTC Communications in 5G networks

Carsten Bockelmann, Nuno K. Pratas, Gerhard Wunder, Stephan Saur, Monica Navarro, David Gregoratti, Guillaume Vivier, Elisabeth de Carvalho, Yalei Ji, Cedomir Stefanovic, Petar Popovski, Qi Wang, Malte Schellmann, Evangelos Kosmatos, Panagiotis Demestichas, Miruna Raceala-Motoc, Peter Jung, Slawomir Stanczak, Armin Dekorsy

arXiv:1804.01701v1cs.IT

TL;DR

mMTC requires access protocols that can handle massive numbers of sporadically active devices sending small payloads, while conventional reservation procedures have limited collision handling. The paper surveys one-stage and two-stage PHY/MAC approaches and evaluates them in a common framework using throughput and latency. Its comparison finds substantial gains from combining collision resolution, multi-user detection, coded or signature-based access, non-orthogonal transmission, and latency-oriented access designs, with trade-offs between throughput and latency.

  • Problem

    mMTC needs efficient access for massive numbers of devices contending sporadically with small payloads, motivating alternatives to conventional access procedures.

  • Method

    The paper surveys PHY and MAC access techniques and evaluates their protocol performance in a common framework using throughput and access latency.

  • Results

    The evaluated solutions show substantial gains, while each provides a trade-off between protocol throughput and access latency.

  • Takeaways & Limitations

    The results identify techniques that can guide robust massive-access protocol design for 5G NR.

Abstract

from arXiv · show

The fifth generation of cellular communication systems is foreseen to enable a multitude of new applications and use cases with very different requirements. A new 5G multiservice air interface needs to enhance broadband performance as well as provide new levels of reliability, latency and supported number of users. In this paper we focus on the massive Machine Type Communications (mMTC) service within a multi-service air interface. Specifically, we present an overview of different physical and medium access techniques to address the problem of a massive number of access attempts in mMTC and discuss the protocol performance of these solutions in a common evaluation framework.

I. INTRODUCTION

The paper addresses massive cellular access for mMTC, where sporadic small-payload transmissions from many devices conflict with conventional access procedures. It surveys challenges and presents approaches evaluated for throughput, latency, signaling, power, and multi-service integration.

  • Motivation: mMTC targets sporadic small-payload access by a massive, randomly active population of low-rate, low-power Machine-Type Devices.Applications include smart-grid, monitoring, asset, and health-management systems.
  • Access challenge: Current access reservation procedures assume relatively few devices and moderate-to-high data rates, conflicting with mMTC requirements.The paper argues that mMTC may require a more radical cellular-access redesign.
  • Contribution: The paper summarizes PHY and MAC solutions and evaluates access-protocol throughput and latency within a common framework.The scope also includes outlooks on RRC and waveforms, while focusing on PHY and MAC enhancements.
  • Control signaling: Control signaling can dominate transmission overhead when an idle device sends one small packet, motivating low-overhead MAC, PHY, and higher-layer procedures.The paper also considers transmitting small packets over the control plane.
  • Access challenge: LTE access capacity is severely degraded because PHY and MAC layers lack a specific collision-resolution procedure for massive contention.The paper identifies novel MAC and PHY approaches as necessary for supporting massive device populations.
  • Multi-service integration: Flexible frame definitions, robust waveforms, and flexible control channels are presented as requirements for coexistence among 5G services with different demands.The paper focuses on MAC and PHY enhancements for control signaling and throughput, using novel waveforms as enabling technology.

III. MMTC STATE OF THE ART

The surveyed state of the art includes non-3GPP LPWANs and 3GPP IoT solutions targeting coverage, cost, power efficiency, and low-rate connectivity. However, existing LPWAN approaches are not designed for massive simultaneous activity.

  • Non-3GPP LPWANs: LoRa uses a star-of-stars LPWAN architecture, chirp-based spread spectrum, and frequency-and-time ALOHA for 0.3 kbps to 50 kbps operation.Vendor-claimed coverage is 10–15 km rurally and 3–5 km in urban areas.
  • Non-3GPP LPWANs: Sigfox provides infrequent bidirectional LPWAN communication using ultra-narrow-band modulation and frequency-and-time ALOHA.Vendor-claimed coverage is 30–50 km rurally and 3–10 km in urban areas.
  • Non-3GPP LPWANs: IEEE 802.11ah extends low-power, long-range 802.11 operation below 1 GHz and restricts contention through device-ID-based access windows.A single access point can serve an area of up to 1 km.
  • State-of-the-art limitation: Presented LPWANs use relatively simple physical-layer processing and cannot cope with massive numbers of simultaneously active devices.This limitation distinguishes their existing capabilities from the massive-access target of mMTC.
  • 3GPP IoT systems: 3GPP IoT standardization introduced Cat-M1, NB-IoT, and EC-GSM solutions addressing cost, power consumption, coverage, and device complexity.These solutions exploit cellular infrastructure but support different device and coverage targets.

1) Cat-M1:

Cat-M1 and NB-IoT represent cellular IoT access options with reduced device capabilities, while the paper evaluates mMTC within a flexible multi-service NR-like air interface. Its comparison uses a single-cell baseline and focuses on PHY and MAC building blocks.

  • Cat-M1: Cat-M1 reduces supported downlink and uplink bandwidth from 20 MHz to 1.4 MHz while retaining LTE-compatible network capabilities.The design targets simpler devices through changes at both device and network-infrastructure levels.
  • Cat-M1: Cat-M1 retains the LTE preamble structure and access procedure, with a simplified option that removes security overhead.Its stated focus is increased coverage with LTE-like functionality.
  • NB-IoT: NB-IoT is a clean-slate access network for massive numbers of low-throughput, delay-tolerant, ultralow-cost devices.Its technical features include 180 kHz bandwidth, 23 dBm maximum transmit power, and 20 dB additional link budget over legacy GPRS.
  • NB-IoT: NB-IoT offers full, medium, and light access protocols, ranging from LTE-like procedures to preamble-followed-by-data transmission.Its main focus is extreme coverage, with a supported user count similar to LTE-M.
  • Evaluation assumptions: The paper models mMTC as part of a multi-service NR air interface using a multi-carrier system, suitable waveform, flexible numerologies, and allocated resource-grid portions.The mMTC service may use all or part of the available resource grid.
  • Evaluation assumptions: The evaluation uses a single-cell scenario and generic OFDM assumptions to compare baseline PHY/MAC concepts through common access-protocol performance measures.The paper classifies building blocks into PHY, MAC, RRC, and waveforms, focusing primarily on the first two.

1) Physical Layer:

The paper distinguishes multi-stage, two-stage, and one-stage access protocols, then focuses on two-stage and one-stage designs whose trade-offs depend on feedback, synchronization, collision handling, and PHY capabilities.

  • Access protocol types: Multi-stage access separates access, connection establishment, and data phases, whereas two-stage and one-stage protocols combine later operations to reduce staging.LTE connection establishment is given as a multi-stage example.
  • Two-stage access: Two-stage access separates access notification from data delivery, allowing intermediate feedback for power control, timing alignment, and resource allocation.The paper assumes initial authentication and security can be reused from a previous full connection-establishment session.
  • One-stage access: One-stage access performs access notification and data delivery in a single transaction using one or several consecutive packets or one transmission.The paper focuses its analysis on one-stage and two-stage protocols.
  • Signaling reduction: Connectionless transmission of small packets from registered and authenticated devices can reduce the number of required signaling messages.Each packet must include source and destination addresses together with the payload.
  • Synchronization and overhead: New 5G waveforms can relax tight uplink synchronization for small mMTC packets, enabling compressed or omitted broadcast signaling.For one-stage protocols, the random access response can be completely omitted; two-stage variants can reduce the overhead.
  • Key performance metrics: Protocol throughput counts served devices per TTI, while access latency measures TTIs from packet arrival to successful reception; both KPIs are required for fair comparison.Throughput alone can be increased through aggregated access opportunities and longer back-off times.

V. MAC PROTOCOL PROCEDURES

The paper compares one-stage and two-stage access protocols, SUD and MUD, and signature-based access for massive mMTC connectivity. Results show a latency–throughput trade-off, while signatures add throughput and authentication functionality but can generate false positives.

  • Approaches: Three MAC approaches are evaluated: one-stage versus two-stage access, signature-based access with integrated authentication, and non-orthogonal access with time-alignment-free transmission.The approaches use idealized physical-layer models to compare massive-access protocol performance.
  • One-Stage vs Two-Stages Access: Two-stage access achieves higher throughput than one-stage access because intermediate feedback assigns data resources and reduces collisions.The comparison is reported for both single-user and multi-user detection.
  • One-Stage vs Two-Stages Access: Multi-user detection improves both access protocols by allowing multiple colliding packets to be decoded and effectively increasing the number of resources.The idealized model decodes at most two superimposed packets when their preambles differ.
  • One-Stage vs Two-Stages Access: One-stage access provides lower latency at very low traffic loads, while two-stage access benefits from lower collision and retransmission rates around λ = 25.Multi-user detection further reduces latency, and larger preamble sets can also improve access latency.
  • Signature-Based Access: Signature-based access represents each device by a multi-preamble sequence that identifies the device and its resource requirements, reducing protocol exchanges and enabling authentication.Signatures are transmitted synchronously over frames composed of multiple PRACHs under orthogonality and simultaneous-detection requirements.
  • Signature-Based Access: Signature construction can produce false positives even with perfect preamble detection, so signature length and active-preamble count must control this effect.A false positive occurs when the base station detects a signature belonging to an inactive device.
  • Signature-Based Access: The signature scheme has a lower-bound throughput capped at 38 packets per TTI at high loads, while increasing arrival rate reduces its access-latency bounds through shorter signatures.The 38-packets-per-TTI ceiling corresponds to the maximum available PRBs per TTI.

C. Non-Orthogonal Access with Time Alignment Free Transmission (NOTAFT)

NOTAFT combines pulse-shaped OFDM and spatial processing to support grant-free, time-alignment-free short-packet transmission. Its one-shot procedure increases throughput at high arrival rates and reduces access latency and signaling overhead.

  • Physical-layer design: P-OFDM uses tailored transmit and receive pulse shapes to increase resilience against timing offsets, allowing asynchronous transmission without timing adjustment.The adopted orthogonalized Gaussian pulse spreads four symbol periods and permits omission of timing alignment during random access.
  • Physical-layer design: NOTAFT couples P-OFDM with SDMA and multiple base-station antennas to enable non-orthogonal grant-free access with collision resolution through MIMO detection.Each spatial layer carries an orthogonal DMRS for identifying the layer.
  • Access procedure: The procedure has each UE randomly select a resource block and transmit its short payload and identifier in one shot, after downlink synchronization and configuration acquisition.The base station then decodes the received signal; retransmission behavior is part of the subsequent procedure.
  • Performance: NOTAFT offers significantly higher protocol throughput, especially at relatively high arrival rates, because it provides many random access opportunities while allocating all PRBs to data.No resource is reserved for the random access procedure.
  • Performance: Removing timing adjustment yields lower access latency than the baseline approach and supports reduced end-to-end latency and signaling overhead for short packets.The single-shot design is also presented as potentially extending device battery life through improved sleep/wake operation.

VI. PHY AND MAC INTEGRATED SCHEMES

The paper integrates physical-layer processing with MAC protocols to compare four approaches for massive mMTC access in a common evaluation framework. The schemes target sporadic traffic through sparsity exploitation, coded random access, reduced control overhead, or dense-network forwarding.

  • Evaluation framework: The evaluation includes physical-layer transmission simulations with coding and modulation, and usually channel estimation, rather than a pure MAC-level comparison.This common framework is used to gauge access protocol performance across the presented concepts.
  • Integrated schemes: CSMUD exploits sporadic-activity sparsity for multi-user detection in each random access slot of a slotted ALOHA setting.It is the first of four integrated PHY/MAC approaches described.
  • Integrated schemes: CRAPLNC extends CSMUD to frame-based coded random access using network-coding ideas, while CCRA combines coded random access with an underlay control channel to reduce control overhead.SCF addresses very dense networks by forwarding messages through many mini base stations to a full base station.

A. Compressive Sensing Multi-User Detection (CSMUD)

CSMUD uses sparse user activity and pilot observations to jointly detect active devices and estimate channels in a one-stage random-access protocol. Its throughput scales nearly linearly at moderate load, while latency remains low across working points.

  • Signal model: Each active user transmits unique pilot symbols and spread data symbols over the bandwidth using pseudo-noise spreading sequences, enabling CSMUD-based activity and channel estimation.The pilots and data occupy a 10 MHz, 1 ms slot in the described setup.
  • Signal model: CSMUD exploits sporadic activity by modeling inactive-user channels as zero, making the channel vector strictly group-sparse for joint activity detection and channel estimation.The group size equals the channel dimension Nh.
  • Signal model: The received pilot signal is modeled with a preamble matrix, superimposed user pilots, and additive white Gaussian noise, with a Toeplitz structure describing pilot-channel convolution.The formulation commonly starts with a one-tap Rayleigh channel and extends to group-sparse multi-tap channels.
  • Evaluation: The evaluation combines CSMUD with a one-stage random-backoff protocol in which failed transmissions repeat up to four times, using full physical-layer processing.The implementation uses GOMP for activity and channel estimation, BPSK, convolutional coding, least-squares equalization, and BCJR decoding.
  • Results: At 10 dB, throughput scales nearly linearly with arrival rate up to 16 before interference and retransmissions reduce successful detection and decoding.The later throughput increase beyond λ = 32 is attributed to GOMP detection behavior and is ultimately limited by an overloaded CDMA system.
  • Results: CSMUD access latency is very low across working points and lower overall than the compared signature-based and frame-focused PLNC schemes.At 10 dB, a single transmission is sufficient most of the time; lower SNR and higher arrival rates increase repetitions.

B. Coded Random Access with Physical Layer Network Coding (CRAPLNC)

CRAPLNC combines coded random access, physical-layer network coding, and optional extended-Galois-field precoding to resolve collisions across a frame. Simulations report throughput gains over slotted ALOHA at moderate and high loads, including without precoding.

  • Protocol design: CRAPLNC uses frame-slotted ALOHA in which users transmit redundancy packets in randomly selected slots, with configurable redundancy distributions.The illustrative configuration uses R = 2 for all users, while optimized distributions are also possible.
  • Motivation: The CRAPLNC design targets reduced signaling while improving collision resolution for short-packet massive access.Its coded-access structure extends earlier schemes and emphasizes short packets with minimum coordination for packet synchronization.
  • Physical-layer processing: Each message may be linearly precoded through symbol-wise multiplication in an extended Galois field before channel encoding and modulation.The precoding coefficients are generated randomly, and the same channel code and modulation are assumed among users.
  • Receiver processing: A preamble supports user detection, channel estimation, and identification of precoding coefficients, while the receiver decodes messages or linear combinations slot by slot.Decoded combinations populate a frame matrix whose full rank can resolve collisions without requiring a singleton packet.
  • Receiver processing: The scheme relies on CSMUD for colliding-user detection and channel estimation within a slot, assuming channels with no delay spread for the simple CSMUD form.This PHY component can also be applied to the data-transmission stage of two-stage protocols.
  • Results: Relevant throughput gains over slotted ALOHA appear at moderate and high loads for very short binary LDPC codes, even with no precoding.The reported simulations use codeword length 164 coded symbols under block-fading channels.

C. Compressive Sensing Coded Random Access (CCRA)

CCRA combines coded random access with compressive-sensing detection to jointly recover user activity, channels, and data for massive mMTC access. Its design separates overloaded control from data resources and uses replica cancellation to improve throughput while reducing signaling overhead.

  • CCRA concept: CCRA combines Coded ALOHA with compressive sensing for sparse joint activity, channel, and data detection.The scheme targets reduced resources for activity signaling, channel estimation, and data detection, while supporting successive interference cancellation.
  • Signal model: The received-signal model represents each active user through a signature, data sequence, channel impulse response, and additive Gaussian noise.Control and data are transmitted through the channel convolution, with measurements taken on selected rows or subcarriers.
  • Resource structure: CCRA places all user preambles in a common overloaded control channel and data in its complementary resources, avoiding interference between them for orthonormal measurements.The complementary bandwidth is divided into frequency patterns addressed by the preambles.
  • Detection: The receiver exploits block-column sparsity with HiHTP, selecting large entries within user blocks and then the largest blocks by ℓ2 norm.The model assumes bounded channel support, sparse user activity, and sparse channel impulse responses.
  • Random access: Replica-based decoding identifies users in non-idle slots, learns replica locations, cancels interference, and iteratively decodes additional users.Users randomly select slots, transmit replicas containing data and pointers, and thereby enable successive interference cancellation.
  • Performance: 13% pilot-to-data ratio reduces overhead, while three replicas significantly improve throughput over traditional slotted ALOHA; replica-detection BER at 15 dB SNR is below 10−1.Traditional slotted ALOHA reaches only 40% normalized throughput, or 20 users/TTI, in the cited comparison.

D. Slotted Compute and Forward (SCF)

SCF extends compute-and-forward to random access by using network densification, physical-layer network coding, and multicarrier transmission. Mini base stations decode message combinations and a macro base station reconstructs messages when the coefficient matrix is full rank, yielding low latency under the evaluated conditions.

  • SCF concept: SCF combines compute-and-forward relaying, network densification, physical-layer network coding, and OFDM to exploit collisions.Simultaneously transmitted messages can be decoded as linear combinations rather than discarded as collisions.
  • Physical layer: Mini base stations rescale received signals and use integer-forcing to decode noisy linear combinations with finite-field coefficients.The effective noise is minimized through the scaling factors and integer coefficients.
  • Message recovery: The macro base station recovers original messages by inverting the coefficient matrix when the decoded equations have full rank over Fp.Devices repeat messages across frequencies to reduce rank deficiency probability.
  • Evaluation: The end-to-end evaluation includes coding, modulation, resource allocation, wireless transmission, mini-base-station processing, forwarding, and macro-base-station aggregation.The setup assumes prior channel estimation and device identification, single antennas, equal device rates, and LDPC coding at R=1/4.
  • Metrics: Throughput is the mean number of successfully transmitted messages for arrival rate λ, and it does not improve significantly above 20 dB SNR in the considered setup.The latency evaluation adds random backoff and repeats transmissions after failure until success or the retransmission limit.
  • Results: Users average one or two transmissions, keeping SCF access latency very low.This conclusion is reported for the evaluated SCF access procedure.

E. Massive MIMO

The Massive MIMO solution addresses uplink mMTC by jointly handling pilot-based channel acquisition and data transmission. Pilot hopping randomizes pilot contamination across slots, while large antenna arrays and pilot resources determine multiplexing and sum-rate performance.

  • Motivation: Massive MIMO uses many antennas to create spatial degrees of freedom for serving large numbers of machine-type devices.The approach supports dense spatial multiplexing under favorable propagation conditions.
  • CSI and throughput: The solution addresses both CSI acquisition through pilot training and uplink data communication, while pilot duration and pilot availability limit performance.Its main metric is an approximate uplink sum rate depending on the number of BS antennas M and pilot sequences τp.
  • Frame operation: Frames contain multiple slots in which active devices transmit a pilot followed by a codeword part, with independent block-fading channel realizations.The example uses four active devices and two mutually orthogonal pilot sequences.
  • Pilot hopping: Pilot hopping randomizes pilot-contamination effects over multiple slots so contamination-induced interference is averaged out.Each device follows a unique pseudorandom hopping pattern known to the base station for identification and combining.
  • Design guidance: One third of the transmission slot should be devoted to training, and the average number of active devices should be on the order of √Mτu.For K=400, M=100, 200, 400, τu=300 and τp=100 were selected as a near-optimal ratio for the reported study.

VII. DISCUSSION AND COMPARISON

The comparison finds a throughput–latency trade-off across massive-access solutions and reports substantial gains from combining physical-layer, medium-access, and protocol-design techniques. The surveyed designs generally shift complexity toward the base-station receiver while keeping device transmitters simple, with one stated exception.

  • Performance comparison: The evaluated solutions each provide a trade-off between throughput and latency.Figures 23 and 24 report protocol throughput and access latency versus arrival rate for K=400 and M=100, 200, 400.
  • Cross-layer techniques: Very significant gains can be achieved through coordinated massive-access techniques across the physical, medium-access, and protocol layers.The comparison identifies compressive sensing, multi-user decoding, redesigned preambles, spatial layers, retransmission coding, backoff, one-stage protocols, and low-overhead synchronization.
  • Physical layer: Physical-layer improvements include compressive-sensing multi-user detection, multi-user decoding, redesigned access preambles, and multiple spatial layers.These techniques are associated with CSMUD, CRAPLNC, OSTSAP, and NOTAFT in the comparison.
  • MAC and protocol design: Medium-access improvements include coding over retransmissions and back-off schemes, while protocol improvements include one-stage access and low-overhead synchronization.The listed schemes include SBA, CRAPLNC, OSTSAP, CSMUD, and NOTAFT.
  • Complexity: Most complexity is placed at the base-station receiver, while device transmitter complexity does not increase except for the CRPLNC scheme.This is stated as a cross-scheme design characteristic in the comparison.

VIII. CONCLUSIONS

The paper surveys access protocols and physical- and medium-access-layer techniques for supporting massive numbers of devices contending for mMTC network access. It evaluates these approaches in a common framework to inform robust protocol design and 3GPP NR recommendations.

  • The paper considers one-stage and two-stage random-access schemes alongside multi-user detection, collision resolution, interference harnessing, and non-orthogonal access techniques.
  • Medium-access techniques include coded random access, signature-based access, one/two-stage random access, and fast uplink access focused on latency reduction.
  • A common evaluation framework compares individual techniques to identify higher-performing approaches and support robust massive-access protocol design for 3GPP NR.
Loading 1804.01701v1…