An adaptive full-duplex deep reinforcement learning-based design for 5G-V2X mode 4 VANETs

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review

5 Citations (Scopus)

Abstract

This paper exploits full-duplex (FD) technology and deep reinforcement learning (DRL) algorithm jointly and adaptively to enhance the performance of 5G-V2X networks that operate based on the 5G-V2X Mode 4 standard. Specifically, we propose a novel physical- (PHY) and medium access control- (MAC) layer cross-layer design, in which collision detection capability is enabled during broadcasting without introducing redundant signalling. Besides, the resource reservation scheme, collision resolution mechanism and scheduling policy are also designed. As the proposed adaptive method is fully decentralised, vehicular users adapt to the unknown and fast-changing environment autonomously without any help from gNBs. Simulation results demonstrate the superiority of our proposed design over the standardised sensing-based semi-persistent scheduling (SB-SPS) protocol. Therefore, the proposed cross-layer design can be considered as a solution for future 5G-V2X VANETs.

Original languageEnglish
Title of host publication2021 IEEE Wireless Communications and Networking Conference, WCNC 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728195056
DOIs
Publication statusPublished - 2021
Event2021 IEEE Wireless Communications and Networking Conference, WCNC 2021 - Nanjing, China
Duration: 29 Mar 20211 Apr 2021

Publication series

NameIEEE Wireless Communications and Networking Conference, WCNC
Volume2021-March
ISSN (Print)1525-3511

Conference

Conference2021 IEEE Wireless Communications and Networking Conference, WCNC 2021
Country/TerritoryChina
CityNanjing
Period29/03/20211/04/2021

Keywords

  • 5G-V2X Mode 4 standard
  • Decentralised network
  • Deep reinforcement learning
  • Full duplex
  • VANETs

Fingerprint

Dive into the research topics of 'An adaptive full-duplex deep reinforcement learning-based design for 5G-V2X mode 4 VANETs'. Together they form a unique fingerprint.

Cite this