PaperlayerView exact revision on arXiv

Source-linked AI summary

5G Wireless Network Slicing for eMBB, URLLC, and mMTC: A Communication-Theoretic View

Petar Popovski, Kasper F. Trillingsgaard, Osvaldo Simeone, Giuseppe Durisi

arXiv:1804.05057v2cs.NIcs.IT

TL;DR

The paper addresses how heterogeneous eMBB, mMTC, and URLLC services can share RAN resources beyond conventional orthogonal slicing. It develops a communication-theoretic model for H-NOMA and finds that non-orthogonal slicing can provide significant performance-trade-off gains in some regimes.

  • Problem

    The paper studies how to enable coexistence of eMBB, mMTC, and URLLC services with heterogeneous requirements within the same RAN architecture.

  • Method

    The study develops a communication-theoretic model accounting for the services' differences in reliability, latency, device population, and traffic dynamics.

  • Results

    H-NOMA is advantageous over H-OMA in some regimes, and H-NOMA with SIC is always beneficial for the eMBB-URLLC case studied.

  • Takeaways & Limitations

    Non-orthogonal RAN slicing can exploit reliability diversity across services to improve performance trade-offs while supporting heterogeneous guarantees.

Abstract

from arXiv · show

The grand objective of 5G wireless technology is to support three generic services with vastly heterogeneous requirements: enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). Service heterogeneity can be accommodated by network slicing, through which each service is allocated resources to provide performance guarantees and isolation from the other services. Slicing of the Radio Access Network (RAN) is typically done by means of orthogonal resource allocation among the services. This work studies the potential advantages of allowing for non-orthogonal sharing of RAN resources in uplink communications from a set of eMBB, mMTC and URLLC devices to a common base station. The approach is referred to as Heterogeneous Non-Orthogonal Multiple Access (H-NOMA), in contrast to the conventional NOMA techniques that involve users with homogeneous requirements and hence can be investigated through a standard multiple access channel. The study devises a communication-theoretic model that accounts for the heterogeneous requirements and characteristics of the three services. The concept of reliability diversity is introduced as a design principle that leverages the different reliability requirements across the services in order to ensure performance guarantees with non-orthogonal RAN slicing. This study reveals that H-NOMA can lead, in some regimes, to significant gains in terms of performance trade-offs among the three generic services as compared to orthogonal slicing.

I. INTRODUCTION

5G network slicing must accommodate eMBB, mMTC, and URLLC traffic with heterogeneous rate, activity, latency, and reliability requirements. The paper studies H-NOMA as a non-orthogonal alternative to orthogonal RAN slicing and identifies performance trade-offs across services.

  • Orthogonal slicing: Conventional RAN slicing assigns orthogonal time-frequency resources to the three services to provide isolation and performance guarantees.This approach can leave resources reserved for intermittent URLLC or mMTC traffic unused.
  • Heterogeneous non-orthogonal access: H-NOMA shares RAN resources among heterogeneous services, allowing eMBB users to access frequency resources otherwise assigned only to mMTC or URLLC traffic.The approach differs from conventional NOMA, which shares resources among users with homogeneous requirements.
  • Performance trade-offs: H-NOMA can improve resource use and eMBB spectral efficiency by exploiting intermittent traffic, but interference and reliability-resource allocation create non-trivial trade-offs.Under H-OMA, eMBB rate is unaffected by URLLC arrival rate, whereas H-NOMA experiences interference; URLLC reliability can be improved by reallocating frequency resources.
  • Model and evaluation: The paper develops a communication-theoretic model to capture performance trade-offs and design insights for H-OMA and H-NOMA.The model considers uplink transmissions from eMBB, mMTC, and URLLC devices to a common base station.

B. Further Related Works

The paper reviews prior work on heterogeneous-service slicing and then develops a tractable communication-theoretic framework for H-NOMA. It analyzes H-OMA and H-NOMA trade-offs, emphasizing reliability diversity and service isolation through non-orthogonal slicing.

  • Further Related Works: Prior studies address heterogeneous-service slicing through orthogonal allocation, downlink URLLC-eMBB multiplexing, feedback, inter-cell interference, and grant-free access.
  • Main Contributions: The paper proposes a tractable model that captures heterogeneous service features, including differences in arrival processes and traffic dynamics.
  • Main Contributions: The analysis first evaluates orthogonal slicing for all three services, then studies H-NOMA in focused eMBB-URLLC and eMBB-mMTC cases.
  • Main Contributions: The study focuses on two service pairs; H-NOMA between URLLC and mMTC may be problematic because mMTC creates random interference patterns affecting URLLC reliability guarantees.
  • Main Contributions: Reliability diversity leverages differences in reliability requirements and definitions to provide service isolation under non-orthogonal slicing.
  • Main Contributions: When properly exploited, non-orthogonal slicing can yield important performance-trade-off gains among the three generic services in some regimes.
  • Organization: The paper reports numerical results illustrating service trade-offs for both H-OMA and H-NOMA schemes.

II. SYSTEM MODEL

The system model represents uplink coexistence of eMBB, URLLC, and mMTC devices over shared time-frequency resources. It specifies traffic, channel, reliability, CSI, and asymptotic-analysis assumptions for evaluating heterogeneous slicing.

  • System Model: The model studies efficient uplink sharing of radio resources by eMBB, URLLC, and mMTC devices communicating with a common base station.
  • Resource Structure: Each resource contains n symbols divided into S minislots, and Fig. 4 illustrates the resulting time-frequency grid.
  • Transmission Model: eMBB occupies one radio resource, while URLLC transmits within minislots and across FU frequency channels because of latency constraints.
  • Traffic Model: mMTC devices access a specified resource through random access, with active-device count AM ∼ Poisson(λM).
  • Channel Model: The model assumes independent Rayleigh fading across resources, normalizes average transmission power and noise power, and represents received power through SNR.
  • Analysis Assumptions: The analysis assumes sufficiently large minislots for asymptotic information-theoretic treatment, leaving finite-blocklength effects for future work.
  • CSI Assumptions: eMBB has perfect CSI, whereas URLLC and mMTC devices have no CSI because of latency and protocol constraints.
  • Reliability Model: Service error probabilities must satisfy reliability requirements, whose differing levels and definitions motivate reliability diversity.

A. Signal Model

The signal model represents shared radio resources across eMBB, URLLC, and mMTC traffic, while the eMBB baseline uses channel-aware truncated power inversion to satisfy an outage target under average power constraints.

  • eMBB model: eMBB users transmit on a scheduled frequency channel only when the instantaneous channel gain exceeds a minimum threshold.
  • Signal model: The received vector is indexed by minislot and frequency channel and includes service-specific transmitted signals and Gaussian noise.
  • Resource model: H-OMA allocates resources exclusively to one service, whereas H-NOMA permits resource sharing and mutual interference.
  • eMBB model: The optimal eMBB power-control strategy is truncated power inversion: transmit power is inversely proportional to channel gain above the threshold and zero otherwise.
  • Design criteria: The analysis uses outage probability and average power as design constraints, while noting that average transmission-rate maximization is an alternative criterion.
  • eMBB baseline: The resulting eMBB outage rate, denoted rB^orth, does not depend on frequency under the section’s assumptions.

C. URLLC

The URLLC model prioritizes immediate decoding and reliability without transmitter-side CSI, while the mMTC model evaluates the fraction of active devices in outage and supports sequential SIC decoding.

  • URLLC: URLLC transmissions occupy FU frequency resources per minislot, and their count across S minislots follows SU ∼ Bin(S, aU).
  • URLLC: Each URLLC message must be decoded in its receiving minislot, so URLLC devices cannot code across minislots.
  • URLLC: URLLC devices lack channel-state information and therefore cannot adapt transmission power or rate.
  • URLLC: The URLLC performance metric is the maximum rate satisfying the reliability condition P(EU) = ϵU.
  • mMTC: For mMTC, reliability is measured by the fraction of active devices in outage, and the objective is the maximum supported arrival rate λM.
  • mMTC: mMTC SIC orders devices by channel gain, decodes them sequentially when log2(1 + σ[m′]) ≥ rM, and subtracts successfully decoded signals.

III. SLICING FOR EMBB AND URLLC

The eMBB–URLLC study compares orthogonal slicing with H-NOMA under heterogeneous latency and reliability requirements, evaluating achievable rate pairs while testing SIC and puncturing-based decoding.

  • H-OMA: Under H-OMA, FU resources are reserved for URLLC and the remaining FB = F − FU resources are assigned to eMBB users.
  • Performance objective: The system performance is characterized by achievable rate pairs (rB, rU) at reliability levels (ϵB, ϵU).
  • H-NOMA with SIC: H-NOMA shares all F frequency channels, with eMBB decoded first and canceled before URLLC decoding; URLLC is therefore decoded while treating eMBB as noise.
  • H-NOMA with SIC: Because eMBB tolerates greater latency, URLLC can be decoded and canceled first, enabling SIC for subsequent eMBB decoding.
  • Reliability diversity: Reliability diversity exploits the different reliability requirements, with eMBB performance expected to approach ideal orthogonal operation under the described SIC decoder.
  • Puncturing: The analysis also considers puncturing, in which minislots containing URLLC transmissions are treated as erasures and corrected using an outer eMBB erasure code.
  • Rate-region analysis: Achievable rate pairs are obtained by fixing URLLC reliability and optimizing eMBB design variables subject to joint reliability constraints.

1) SIC decoder:

The SIC decoder analysis bounds eMBB reliability under URLLC interference and shows that the required eMBB activation probability remains close to its orthogonal benchmark when URLLC reliability is very high.

  • SIC decoder: Under H-NOMA, eMBB decoding is affected by whether URLLC transmissions are present and successfully decoded or canceled.
  • Reliability analysis: The eMBB error probability is bounded by separating cases with and without URLLC interference and applying the reliability constraint Pr(EB) ≤ ϵB.
  • Reliability analysis: URLLC interference can cause an eMBB error even when eMBB SNR exceeds its threshold, so aB exceeds 1 − ϵB under non-orthogonal slicing.
  • Reliability analysis: Because ϵU is very small, the eMBB activation probability aB remains close to 1 − ϵB, indicating typically minimal URLLC interference impact.
  • Rate region: The resulting eMBB and URLLC rates exhibit a non-trivial interdependence, and maximum eMBB rate is found through constrained numerical maximization.

2) Puncturing and erasure decoder:

The paper analyzes puncturing-based H-NOMA alongside SIC and compares both with orthogonal slicing for eMBB–URLLC coexistence. The preferred scheme depends on relative SNRs and whether the objective emphasizes URLLC rates or eMBB sum-rates.

  • Puncturing and erasure decoder: The eMBB reliability requirement is decomposed according to whether the erasure code corrects URLLC-induced erasures.When the number of erasures is correctable, outage is determined by the instantaneous SNR; otherwise decoding fails.
  • Numerical comparison: The simulations evaluate rate regions for H-OMA and H-NOMA with SIC and puncturing under two opposite relative-SNR settings.The cases use ΓU = 20 dB, ΓB = 10 dB and ΓU = 10 dB, ΓB = 20 dB, respectively.
  • Numerical comparison: When ΓU > ΓB, H-NOMA with SIC dominates the rate region achievable by orthogonal slicing.The base station can decode and cancel URLLC transmissions by leveraging reliability diversity.
  • Numerical comparison: When ΓU < ΓB, orthogonal slicing can achieve rate pairs unavailable to H-NOMA with SIC, especially at large URLLC rates.The latency constraint prevents decoding and canceling eMBB transmissions before URLLC decoding.
  • Numerical comparison: H-NOMA offers significant gains when large eMBB sum-rates are desired because eMBB users can occupy more spectral resources without significant URLLC-interference effects.The reported lower bound captures the shape of the region from more accurate Monte Carlo simulations and is easier to evaluate.

A. Orthogonal Slicing: H-OMA for eMBB and mMTC

Orthogonal slicing partitions resources between eMBB and mMTC through time sharing, scaling each service’s achievable performance by its allocated fraction. The supported region is characterized by the eMBB rate and mMTC arrival rate under reliability and transmission-rate requirements.

  • Orthogonal resource allocation: H-OMA allocates fractions α and 1 − α of a frequency resource to eMBB and mMTC devices, respectively.The two services use the resource in a time-sharing manner.
  • Orthogonal resource allocation: Orthogonal slicing scales both the achievable eMBB rate and mMTC transmission rate according to each service’s allocated time fraction.The supported region consists of pairs (rB, λM) for given mMTC rate rM and error probability ϵM.
  • SIC decoding: Reliability diversity motivates decoding strong mMTC devices before eMBB because mMTC error probability measures the fraction of incorrectly detected active users.Some high-gain mMTC devices can create substantial interference and may therefore be decoded and canceled first.
  • SIC decoding: The non-orthogonal SIC procedure decodes either eMBB or the next mMTC device in decreasing channel-gain order, canceling each successfully decoded transmission.The procedure terminates when no further transmissions can be reliably decoded.
  • Performance characterization: The SIC analysis characterizes mMTC and eMBB error probabilities from the expected numbers of decoded mMTC and eMBB devices.Computing the maximum supported mMTC arrival rate as a function of eMBB rate requires Monte Carlo simulation.

C. Numerical Illustration

Numerical results reveal three operating regimes for the eMBB–mMTC trade-off and show that non-orthogonal slicing is beneficial only in selected rate and reliability ranges. Orthogonal slicing becomes preferable when eMBB interference dominates.

  • Three operating regimes: The supported mMTC arrival rate is nearly constant at very small eMBB rates because eMBB can usually be decoded and canceled first.Small increases in rB therefore have little effect on mMTC performance in this regime.
  • Three operating regimes: At intermediate eMBB rates, eMBB is decoded after some strong mMTC signals are canceled, reducing mMTC performance through interference.The SIC process may stop after eMBB detection, with subsequent mMTC decoding failing because other mMTC signals remain undecoded.
  • Three operating regimes: At large eMBB rates, mutual interference causes the supported mMTC arrival rate to decay to zero.The reported upper and lower bounds agree with Monte Carlo results in the third regime.
  • Parameter effects: Increasing ΓB from 20 dB to 30 dB shifts the first-to-second regime transition to larger rB but does not change the second-to-third transition rate.The shift reflects earlier eMBB decoding for a larger set of eMBB rates.
  • Parameter effects: Tighter eMBB reliability constraints reduce supported mMTC arrival rates in the second and third regimes through increased eMBB interference.Higher eMBB transmission power required for reliability impairs mMTC decoding.
  • Orthogonal versus non-orthogonal slicing: Non-orthogonal slicing is beneficial across the second and third regimes for moderate eMBB reliability requirements, whereas orthogonal slicing is superior at very large eMBB rates.The latter regime is limited by interference caused by eMBB users.

V. DISCUSSION, CONCLUSIONS AND FUTURE WORK

The paper develops a communication-theoretic model for orthogonal and non-orthogonal uplink slicing among heterogeneous 5G services. Its analysis shows that reliability diversity guides effective H-NOMA designs and that H-NOMA is advantageous in specific operating regimes.

  • Model and slicing paradigms: The model captures differences among eMBB, mMTC, and URLLC in reliability, latency, and supported-device population.It studies uplink slicing over shared multiple-access resources while considering heterogeneous performance requirements and service isolation.
  • Reliability diversity: Reliability diversity is required to guide effective non-orthogonal slicing solutions across the studied service pairings.The design principle leverages heterogeneous reliability requirements rather than only their numerical reliability levels.
  • eMBB-URLLC coexistence: For eMBB-URLLC coexistence, SIC should decode URLLC before eMBB because URLLC reliability and latency requirements are stricter.URLLC decoding therefore cannot depend on decoding the less stringent eMBB transmission first.
  • eMBB-mMTC coexistence: For eMBB-mMTC coexistence, regimes exist where decoding eMBB after one or more mMTC devices enables non-orthogonal slicing gains.The ordering reflects the high probability of many active mMTC devices and the need to account for the fraction of correctly decoded transmissions.
  • Numerical conclusions: H-NOMA is advantageous over H-OMA in some regimes, including very large eMBB rates for eMBB-URLLC and sufficiently small eMBB rates for eMBB-mMTC.For eMBB-mMTC, the eMBB rate must be small enough not to hamper mMTC decoding; H-OMA is advantageous in other regimes.
  • Future work and scope: The analysis relies on simplifying assumptions and identifies extensions involving more general models, alternative decoding, multiple mMTC channels, frequency hopping, and non-Poisson arrivals.The current model assumes a Poisson arrival process, while burstier arrivals may change the achievable gain from non-orthogonal slicing.

APPENDIX A

Appendix A derives bounds on supported mMTC arrival rates and eMBB error probabilities for non-orthogonal slicing. The bounds combine analytical inequalities, Monte Carlo evaluation, and alternative decoding schemes.

  • Numerical evaluation: The expectation in the bound is evaluated by Monte Carlo simulation, with t ≥ 0 selected to maximize the resulting lower bound.The derivation uses Markov inequality and an i.i.d. fading assumption in preceding steps.
  • Upper-bound construction: The appendix derives an upper bound on the non-orthogonal mMTC arrival rate by bounding eMBB error probability under residual mMTC interference.The construction assumes correctly decoded mMTC devices are cancelled before eMBB decoding while unsuccessfully decoded devices remain as interference.
  • Alternative decoding: For an alternative lower bound, the decoder is forced to decode eMBB first and then the mMTC devices.The maximal arrival rate supported by this modified decoder is a lower bound on the non-orthogonal arrival rate.
  • Orthogonal reference: The appendix defines the orthogonal mMTC error probability without eMBB and uses total probability to account for eMBB presence.It then compares the resulting admissible arrival-rate conditions with the non-orthogonal bounds.
  • Combined bounds: The appendix combines the bound from (41) with another upper bound to tighten the constraint when the eMBB rate is small.The resulting arrival-rate bound is formed using the smallest relevant admissible value.
Loading 1804.05057v2…