PaperlayerView exact revision on arXiv

Source-linked AI summary

Ultra-Reliable Low-Latency Communications in Autonomous Vehicular Networks

Xiaohu Ge

arXiv:1903.01863v1cs.NI

TL;DR

Autonomous vehicular networks require ultra-reliable low-latency communication, while reliability and latency interact. The paper proposes a Euclidean-norm joint function and an expanded network-slicing solution; simulations report improvements in both reliability and latency utility with slicing.

  • Problem

    Reliability and latency interact in 5G autonomous vehicular networks, motivating their joint study to meet URLLC requirements.

    Source: §I · p. 1, §III · p. 3

  • Method

    The paper proposes a Euclidean-norm function for jointly evaluating reliability and latency and a slicing solution extending resource slicing to service and function slicing.

    Source: §I · p. 2

  • Results

    With network slicing, both reliability and latency utility-function values improve compared with networks without slicing.

    Source: §V · p. 9

  • Takeaways & Limitations

    The proposed network-slicing solution is presented as a way to implement URLLC in 5G autonomous vehicular networks.

    Source: §VI · p. 10

Abstract

from arXiv · show

Autonomous vehicles are expected to emerge as a main trend in vehicle development over the next decade. To support autonomous vehicles, ultra-reliable low-latency communications (URLLC) is required between autonomous vehicles and infrastructure networks, e.g., fifth generation (5G) cellular networks. Hence, reliability and latency must be jointly investigated in 5G autonomous vehicular networks. Utilizing the Euclidean norm theory, we first propose a reliability and latency joint function to evaluate the joint impact of reliability and latency in 5G autonomous vehicular networks. The interactions between reliability and latency are illustrated via Monto-Carlo (MC) simulations of 5G autonomous vehicular networks. To improve both the reliability and latency performance and implement URLLC, a new network slicing solution that extends from resource slicing to service and function slicing is presented for 5G autonomous vehicular networks. The simulation results indicate that the proposed network slicing solution can improve both the reliability and latency performance and ensure URLLC in 5G autonomous vehicular networks.

I. INTRODUCTION

Autonomous-vehicle URLLC motivates jointly evaluating reliability and latency and developing a slicing solution; the paper reports that slicing improves both performance measures.

I. INTRODUCTION

  • The paper addresses reliability–latency interactions in 5G autonomous vehicular networks, where optimizing either measure can affect the other.Source: §I · p. 1, §III · p. 3
  • The proposed network slicing extends resource slicing to service and function slicing to improve both reliability and latency.Source: §I · p. 2
  • The proposed slicing solution improves both reliability and latency, and its optimization depends on vehicle density.Source: §I · p. 2

II. SYSTEM MODEL

  • The system model uses Manhattan Poisson line process roads, uniformly distributed RSUs, and independently Poisson-distributed infrastructure providers forming Poisson–Voronoi cells.Source: §II · p. 2

A. Reliability and Latency Joint Function

B. Reliability Model

The reliability model estimates successful message delivery over vehicle–RSU wireless links, including links that use multiple hops.

B. Reliability Model

  • The model defines network reliability through successful transmission probabilities over vehicle–RSU links, including multi-hop paths.Source: B. Reliability Model · p. 4
  • Wireless links use 75 GHz millimeter-wave transmission, with vehicle–RSU distance modeled uniformly over the covered road length.Source: B. Reliability Model · p. 4
  • Beamforming is used and interference is ignored; a one-hop message succeeds when received SNR meets threshold θ.Source: B. Reliability Model · p. 4

C. Latency Model

The latency model combines wireless propagation and RSU message handling; propagation latency rises with vehicle density and falls with RSU density.

C. Latency Model

  • Total message latency is divided into wireless propagation latency and handling latency in RSU queues.Source: C. Latency Model · p. 4
  • For multi-hop links, propagation latency includes transmission latency and processing at relaying vehicles; RSU handling is modeled as a GI/M/1/∞ queue.Source: C. Latency Model · pp. 4–5
  • At fixed RSU density, propagation latency increases with vehicle density; at fixed vehicle density, it decreases as RSU density increases.Source: Figure 2 · p. 5

D. Performance Analysis of Reliability and Latency

Monte Carlo results show a vehicle-density trade-off between reliability and latency utility, while higher RSU density improves both. The joint function peaks around a vehicle density of 0.19 veh/m, then declines.

D. Performance Analysis of Reliability and Latency

IV. NETWORK SLICING SOLUTION

The section identifies a reliability–latency conflict and argues that resource slicing alone is difficult to use to meet URLLC requirements. It extends slicing to service and function slicing alongside resource slicing.

IV. NETWORK SLICING SOLUTION

  • The reliability–latency conflict means both performance dimensions must improve to implement URLLC in 5G autonomous vehicular networks.Source: §IV · p. 7
  • Conventional network slicing, which slices only network resources, is described as insufficient on its own to satisfy vehicular URLLC requirements.Source: §IV · p. 7
  • The proposed solution extends slicing from resources to services and functions to improve reliability and latency.Source: §IV · p. 7

A. Service Slicing

The solution normalizes vehicular services into service slices, then combines function and resource slicing in a layered architecture. Its algorithm allocates local and pooled RSU resource blocks according to service latency requirements.

A. Service Slicing

  • Vehicular services are normalized into state-report, event-driven, and entertainment-application slices, with differing arrival, handling-time, and latency requirements.State-report slices carry vehicle state and need quick access; event-driven slices include emergency broadcasts and road updates and require low latency and high reliability.Source: A. Service Slicing · p. 7

B. Function Slicing and Resource Slicing

  • The architecture dynamically assembles function modules and combines InP monitoring and allocation with shared-resource and RSU interface, virtualization, and resource layers.The RSU interface separates service-slice resource requirements, virtualization allocates local blocks, and the resource layer manages blocks at each RSU.Source: B. Function Slicing and Resource Slicing · p. 7, Figure 8 · p. 8

C. Network Slicing Algorithm

  • The a_kg matrix represents inter-RSU resource sharing, while a_max limits the share of local RSU blocks used for other RSUs.Adjacent RSUs can supply blocks to reduce a service slice’s handling latency.Source: C. Network Slicing Algorithm · p. 8
  • For each service slice, the InP compares local handling latency with its requirement, releasing surplus blocks below the target and scheduling pooled blocks when latency exceeds it.The algorithm updates resource allocation until handling latency meets the slice requirement.Source: C. Network Slicing Algorithm · pp. 8–9

V. SIMULATION RESULTS AND DISCUSSION

Monte Carlo simulations compare reliability, latency, and their joint function with and without network slicing. Slicing improves both separate utility functions, while the joint-function comparison varies with vehicle density and the upper bound ratio.

V. SIMULATION RESULTS AND DISCUSSION

  • The simulations classify messages as SRSS, EDSS, and EASS, with slot times of 25, 50, and 100 milliseconds and a configured message-count ratio of 8:1:1.The slot times correspond to SRSS, EDSS, and EASS, respectively.Source: §V · p. 9
  • Network slicing improves both reliability and latency utility-function values compared with no slicing.The separate utilities still move in opposite directions as vehicle density increases.Source: §V · p. 9, Figure 9 · p. 9
  • At a fixed upper bound ratio, network slicing yields higher reliability-and-latency joint-function values than no slicing.Fig. 10 examines this comparison across vehicle density and different upper bound ratios.Source: §V · p. 9
  • With aMax_kg = 0.9 below 0.12 veh/m, excess resource-block scheduling by adjacent RSUs reduces the local and network joint-function values.The passage attributes the reduction to resource blocks scheduled by adjacent RSUs.Source: §V · p. 9

VI. CONCLUSION

The conclusion presents a reliability-and-latency joint function and a network-slicing algorithm for 5G autonomous vehicular networks. Simulations indicate that slicing improves both metrics, while further optimization is proposed for service, resource, and function slicing.

VI. CONCLUSION

  • The reliability-and-latency joint function evaluates their combined impact, and simulations find a maximum value with respect to vehicle density.The joint function is based on Euclidean norm theory, and the interactions are analyzed using Monte Carlo simulations.Source: §VI · pp. 9–10
  • The proposed network-slicing algorithm improves both reliability and latency in 5G autonomous vehicular networks, and the results indicate slicing can support URLLC.The proposed architecture extends network slicing from resources to services and functions.Source: §VI · p. 10
  • The proposed vehicular-network architecture extends slicing from network resources to service and function slicing.The conclusion also describes a network-slicing algorithm for implementing URLLC.Source: §VI · p. 10
  • Future work should optimize service slicing for message types and counts, resource slicing for service-slicing results, and function slicing to support both.Source: §VI · p. 10

APPENDIX A

The appendix derives message-arrival interval distributions for vehicles and an RSU. It also states a conditional survival relation for a vehicle's message interval.

APPENDIX A

  • For the vehicle message interval T0, the conditional survival probability after time s equals the survival probability over an additional duration t.The passage gives this relation as an exponential expression.Source: Appendix A · p. 10
  • For n vehicles in an RSU's coverage, the appendix derives the probability that the RSU message-arrival interval is at most t.The derivation refers to messages arriving in the RSU coverage area within the time slot (0, t].Source: Appendix A · p. 10
  • The appendix then derives the cumulative distribution function and probability density function of the RSU message-arrival interval.Source: Appendix A · p. 10

APPENDIX B

Appendix B models RSU message handling with a GI/M/1/∞ queue and derives queue probabilities and message waiting and dwelling-time distributions. It specifies FIFO service and distinguishes messages that wait from those handled immediately.

APPENDIX B

  • The RSU message-handling process is modeled as a GI/M/1/∞ queue.Source: Appendix B · p. 10
  • For c = 1, the queue-length probability is p_j = (1 − δ) · δ^j for j = 0, 1, 2, … .The appendix identifies p_j as the probability of j messages in the RSU queue.Source: Appendix B · pp. 10–11
  • Under FIFO service, a message does not wait when fewer than c messages are queued; otherwise it waits for services to complete for j − c + 1 messages.The waiting-time distribution is discussed for cμ < 1.Source: Appendix B · p. 11
  • The message dwelling-time distribution combines waiting time and service time, which the appendix treats as independent.Source: Appendix B · p. 11

APPENDIX C

Appendix C models typical InP coverage-cell areas with a Gamma distribution and derives the expected RSU count under uniform RSU placement along roads.

APPENDIX C

  • A typical InP cell area follows a Gamma distribution when InP density is λInp.Source: Appendix C · p. 11
  • The Gamma model uses a as its shape parameter and bλInp as its inverse scale, with simulation defaults a = 3.61 and b = 3.57.Source: Appendix C · p. 11
  • With ρroad denoting urban road density, expected NRSU is derived assuming RSUs are uniformly distributed along the road.Source: Appendix C · p. 11

Related papers

Loading 1903.01863v1…