PaperlayerView exact revision on arXiv

Source-linked AI summary

NOMA for Next-generation Massive IoT: Performance Potential and Technology Directions

Yifei Yuan, Sen Wang, Yongpeng Wu, H. Vincent Poor, Zhiguo Ding, Xiaohu You, Lajos Hanzo

arXiv:2104.04911v1cs.ITeess.SPmath.IT

TL;DR

Massive IoT requires support for many connections, while its small packets and uncoordinated transmissions make NOMA performance bounds a relevant research question. The paper analyzes short-blocklength and multi-user-diversity bounds and discusses technology directions for 6G. It reports a widening NOMA advantage over OMA as load grows, alongside a boundary in what each analysis captures.

Abstract

from arXiv · show

Broader applications of the Internet of Things (IoT) are expected in the forthcoming 6G system, although massive IoT is already a key scenario in 5G, predominantly relying on physical layer solutions inherited from 4G LTE and primarily using orthogonal multiple access (OMA). In 6G IoT, supporting a massive number of connections will be required for diverse services of the vertical sectors, prompting fundamental studies on how to improve the spectral efficiency of the system. One of the key enabling technologies is non-orthogonal multiple access (NOMA). This paper consists of two parts. In the first part, finite block length theory and the diversity order of multi-user systems will be used to show the significant potential of NOMA compared to traditional OMA. The supremacy of NOMA over OMA is particularly pronounced for asynchronous contention-based systems relying on imperfect link adaptation, which are commonly assumed for massive IoT systems. To approach these performance bounds, in the second part of the paper, several promising technology directions are proposed for 6G massive IoT, including linear spreading, joint spreading & modulation, multi-user channel coding in the context of various techniques for practical uncoordinated transmissions, cell-free operations, etc., from the perspective of NOMA.

I. INTRODUCTION

The introduction traces how 5G IoT inherited 4G-based access and why NOMA, considered but not standardized for 5G mMTC, remains relevant to 6G. The paper outlines massive-IoT performance goals and NOMA-related technology directions.

I. INTRODUCTION

  • 6G is envisioned to require 10 million connections per square kilometer, ten times the 5G mMTC KPI.The introduction says this target should spur practical enabling solutions.Source: §I · p. 1
  • NB-IoT and eMTC, originally designed for 4G, were carried into the 5G network architecture to meet IMT-2020 requirements.The introduction traces IoT standardization from a simplified LTE specification in 2011.Source: §I · p. 1
  • NOMA with grant-free transmissions was considered for 5G mMTC because it can improve connection density and reduce system overhead, but no NOMA-based mMTC solution was specified.The introduction links this outcome to NB-IoT’s effect on standardization and the postponement of advanced features such as NOMA to Rel-16.Source: §I · p. 1
  • The paper outlines massive-IoT performance goals and NOMA technology directions for 6G, while noting that relay-connected devices may have different channel conditions.The paper directs readers to for relay-related channels.Source: §I · p. 1

II. PERFORMANCE POTENTIAL

Massive IoT traffic consists of small, infrequent packets, making extensive link adaptation overhead burdensome and Shannon’s long-block assumption unsuitable. The section distinguishes grant-free access from autonomous contention and frames performance in terms of supported users and outage.

II. PERFORMANCE POTENTIAL

  • Massive IoT packets are typically 40–320 bits, so Shannon’s asymptotic capacity formula is not suitable for evaluating them.The section contrasts these small packets with long-block assumptions used for data-rate evaluation.Source: §II · p. 2
  • Link adaptation can require hundreds of control bits before a small payload, motivating grant-free or autonomous transmission that sends the packet immediately.The described setup includes random-access exchanges, CSI estimation, feedback, scheduling grants, and connection setup.Source: §II · p. 2
  • Grant-free access pre-allocates signatures to avoid collisions, whereas autonomous access is contention-based and users may choose the same signature.Pre-allocation may suit stationary devices; contention-based access is more general when pre-allocation and closed-loop power control are impractical.Source: §II · p. 2
  • For grant-free or autonomous transmissions, performance is evaluated by the number of users supported at a target per-user rate and outage probability.An outage occurs when the instantaneous fading channel supports less than the target data rate; signatures may also serve as temporary IDs and aid channel estimation.Source: §II · p. 2

A. Performance bounds of a multiple access channel having short block length

The section uses short-blocklength bounds for independently Rayleigh-faded multiple-access channels to assess NOMA against OMA. The required total Eb/No is similar at low load, but the gap widens sharply in the cliff region.

A. Performance bounds of a multiple access channel having short block length

  • Finite-blocklength rate bounds account for channel dispersion, block length, and error probability, with shorter blocks or lower error probabilities reducing the maximum achievable rate.The bounds are described as tighter than many previous estimates for block lengths as short as 100 bits.Source: A. Performance bounds of a multiple… · p. 2
  • The analysis uses independently Rayleigh-faded users because massive IoT can rely only on open-loop power control, leaving received SNR subject to fast fading.Figure 2 assumes an information block length of 100 bits and compares perfect receiver CSI (CSIR) with no CSI.Source: A. Performance bounds of a multiple… · p. 3, Figure 2 · p. 3
  • At system load μ = 0.16 (16 bps/Hz), the total Eb/No performance gap between NOMA and OMA exceeds 20 dB.At low system load or spectral efficiency, their required Eb/No values are very similar; through the cliff region, NOMA’s required total Eb/No remains almost constant as load increases.Source: A. Performance bounds of a multiple… · p. 3, Figure 2 · p. 3
  • Beyond the cliff region, the required Eb/No for NOMA increases at a pace similar to OMA.Source: A. Performance bounds of a multiple… · p. 3

B. Upper bound of multi-user diversity gain

The section estimates NOMA’s multi-user diversity gain using individual outage probability derived from joint outage events. It reports that the gain grows with the number of participating users and sum rate, then saturates at an example user count.

B. Upper bound of multi-user diversity gain

  • NOMA’s multi-user diversity gain increases as more users participate, then saturates at 4 users for a 3 bps/Hz sum rate and 8 users for 9 bps/Hz.The gain also increases as the sum rate rises to 9 bps/Hz.Source: B. Upper bound of multi-user diversity… · p. 3, Figure 3 · p. 4
  • The analysis defines individual outage probability by summing joint outage probabilities across user subsets containing that user, assuming joint detection and decoding.A joint outage occurs when a subset’s sum rate falls below its target; the resulting analysis is an upper bound for uplink grant-free NOMA.Source: B. Upper bound of multi-user diversity… · p. 3

C. Synergy of information-theoretic study and multi-user diversity analysis

The information-theoretic and multi-user diversity analyses show consistent NOMA gains over OMA as the number of superimposed users increases. Each approach has a different limitation, motivating bounds that combine short-blocklength and multi-user-diversity effects.

C. Synergy of information-theoretic study and multi-user diversity analysis

  • Both analyses show that NOMA’s performance gain over OMA increases as the number of superimposed users rises, across a similar 3–16 bps/Hz spectral-efficiency range.Source: C. Synergy of information-theoretic… · p. 4
  • Information-theoretic bounds capture finite blocklengths but require asymptotic treatment, whereas multi-user diversity bounds cover finite user counts and receiver types but omit blocklength effects.Source: C. Synergy of information-theoretic… · p. 4
  • Jointly considering short blocklengths and multi-user diversity is expected to yield more industrially relevant bounds and guidance for practical massive-IoT NOMA design.Source: C. Synergy of information-theoretic… · p. 4

III. TECHNOLOGY DIRECTIONS

The section frames NOMA’s potential spectral-efficiency advantage over OMA as an open problem for grant-free massive IoT. It proposes physical-layer signal formats that approach performance bounds while addressing uncoordinated transmission and inter-cell interference.

III. TECHNOLOGY DIRECTIONS

  • For grant-free or autonomous massive-IoT transmissions, NOMA can potentially offer much higher spectral efficiency than OMA, but its uplink performance remains incompletely understood.Source: §III · p. 4
  • The primary signal-format goal is to approach Section II’s system-level bounds with reasonable receiver complexity in uncoordinated transmissions.Source: §III · p. 4
  • Spreading maps short user information sequences to longer modulated-symbol sequences, using linear spreading, joint modulation and spreading, or multi-user channel coding.Source: §III · p. 4
  • Practical NOMA design must address inter-cell interference and the challenges of uncoordinated transmissions through signal, waveform, receiver, and frame designs.Source: §III · p. 4

A. Signal spreading based schemes

Signal-spreading schemes include symbol-level linear spreading and joint modulation and spreading. Their receiver complexity, performance trade-offs, and scalability differ.

A. Signal spreading based schemes

  • Signal-spreading schemes comprise symbol-level linear spreading and joint modulation and spreading; linear spreading applies a sequence to each traditionally modulated symbol and needs large sequence pools to limit collisions.Source: A. Signal spreading based schemes · p. 4
  • MMSE de-spreading with hard interference cancellation offers relatively low complexity and compatibility with typical 4G/5G base stations, but its outage-performance gap to joint detection grows with user count or per-user spectral efficiency.Source: A. Signal spreading based schemes · pp. 4–5
  • Joint modulation and spreading maps encoded bits directly to spread symbols, can exploit multidimensional diversity and shaping gains, and may use sparse codes to reduce receiver complexity.Source: A. Signal spreading based schemes · p. 5
  • Joint modulation and spreading requires a high-complexity maximum-likelihood receiver, and codebooks optimized for specific spreading lengths and user counts create scalability challenges for massive IoT.Source: A. Signal spreading based schemes · p. 5

B. Multi-user channel coding

The section argues that short packets and fading motivate multi-user channel codes and receivers that exploit multi-user diversity. It outlines code-design options and iterative detection and decoding.

B. Multi-user channel coding

  • Multi-user channel codes are proposed because massive-IoT packets are short and legacy single-user codes target large blocklengths; with advanced receivers, they can exploit multi-user diversity in fading channels.Source: B. Multi-user channel coding · p. 5
  • Practical multi-user code designs include repetition with LDPC and spatially coupled LDPC, adding design freedom to improve robustness in complex uncoordinated multi-user channels.Source: B. Multi-user channel coding · p. 5
  • Candidate designs can use structured codes such as Polar or other block codes, traditional design analyses, or AI methods such as genetic algorithms for short-blocklength LDPC codes.Source: B. Multi-user channel coding · p. 5
  • The MAC receiver may iteratively exchange extrinsic information between detector and decoder, using soft bits and potentially joint decoding across users while targeting low complexity.Source: B. Multi-user channel coding · p. 5

C. Cell-free architectures for massive IoT

Cell-free access is especially important for NOMA-assisted massive IoT, where inter-cell interference poses a system-level challenge. Cell-free operation may mitigate interference and balance load, but requires new random-access and control-plane designs.

C. Cell-free architectures for massive IoT

  • NOMA faces a system-level challenge because inter-cell interference is more limiting, and typical cell-edge SNRs rarely reach the stated 20 dB needed to support massive IoT with NOMA.A spreading-code-domain MMSE detector can suppress inter-cell interference to some extent, but seamless integration of multiuser channel coding with interference suppression remains an open challenge.Source: C. Cell-free architectures for massive… · p. 5
  • Cell-free access is particularly important for NOMA-assisted massive IoT.The passage identifies inter-cell interference as a key capacity limit that becomes more restrictive in NOMA systems.Source: C. Cell-free architectures for massive… · p. 5
  • Cell-free access lets a group of base stations support a group of devices, unlike cell-based access, where each base station serves devices within its cell boundary.The described cell-free arrangement offers load balancing and interference mitigation.Source: C. Cell-free architectures for massive… · p. 5
  • Massive-IoT cell-free access requires new random-access procedures, with multiple base stations performing preamble detection and related control-plane processing for a user.This differs from traditional single-cell random access and scheduled, RRC-connected uplink CoMP, with implications for next-generation control-plane design.Source: C. Cell-free architectures for massive… · pp. 5–6

D. Handling practical issues in uncoordinated

Practical uncoordinated transmissions must address reference-signal collisions, sparse device activity, timing and frequency asynchrony, and limits on base-station antenna counts. The paper discusses alternative channel-estimation and frame-design approaches alongside these constraints.

transmission

  • With hundreds of simultaneous users, limited orthogonal reference signals make collisions likely; non-orthogonal references or data-only transmission are alternatives, while iterative interference cancellation can mitigate the issue.Data-only transmission estimates channels from received-payload constellation geometry and can be accurate for BPSK or QPSK, but increases receiver complexity through blind decoding or detection.Source: transmission · p. 6
  • Sparse activity in massive IoT motivates compressed sensing for joint device-activity detection and channel estimation; AMP can extend to temporally or spatially correlated channels.The cited work describes the AMP extension.Source: transmission · p. 6
  • Without closed-loop timing adjustment, propagation delays make precise time/frequency synchronization difficult, motivating tolerant waveforms and frame structures such as slotted Aloha.Narrower sub-carrier spacing in low-mobility scenarios can increase cyclic-prefix duration and tolerance to timing errors or differences.Source: transmission · p. 6
  • Although large receive-antenna counts and sufficient received power per degree-of-freedom can let SIC achieve very low error probability, wide-coverage massive IoT may limit base-station antennas because carrier frequencies may not be high.The relatively large antenna size at those frequencies constrains the number of antennas at the base station.Source: transmission · p. 6

IV. Conclusions

The paper analyzes NOMA performance bounds and reports gains over OMA, then discusses technology directions and practical considerations for massive IoT. It also identifies cell-free NOMA as having high performance potential.

IV. Conclusions

  • NOMA shows promising gains over OMA for short-blocklength coding and multiuser diversity in Rayleigh-faded multiple-access channels.The paper analyzes performance bounds for these settings.Source: §IV · p. 6
  • Proposed directions include linear spreading, joint modulation and spreading, multiuser channel coding, and cell-free operation for massive IoT.Multiuser channel coding may potentially approach the performance bounds of uncoordinated MAC systems; cell-free NOMA is described as having high performance potential.Source: §IV · p. 6
  • The paper discusses practical concerns for robust NOMA-aided massive access, including channel estimation, reference-signal design, signature collision, asynchronization, receiver algorithms, and MIMO deployment.Figure 5 compares cell-based and cell-free access for massive IoT.Source: §IV · p. 6, Figure 5 · p. 6
Loading 2104.04911v1…