Channel Access Optimization in Unlicensed Spectrum for Downlink URLLC: Centralized and Federated DRL Approaches

Yan Liu, Hui Zhou, Yansha Deng*, Arumugam Nallanathan

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)


The sixth-generation (6G) communication research is currently in the early stage, where ultra-reliable low-latency communication (URLLC) is still an important service as in the fifth-generation (5G). Since 6G networks are expected to provide even higher levels of massive connectivity, high spectrum efficiency, high reliability, and low latency than 5G communication, it would confront much more severe spectrum scarcity problems, which make the new radio in unlicensed spectrum (NR-U) technology attractive. However, how to achieve URLLC requirements in NR-U networks is extremely challenging due to interference and collisions among multiple radio access technologies (e.g., WiFi). Therefore, it is urgent to design efficient spectrum-sharing algorithms to support URLLC in emerging 6G networks. In this paper, we develop novel centralized deep reinforcement learning (CDRL) and federated DRL (FDRL) frameworks, respectively, to optimize the downlink URLLC transmission in NR-U and WiFi coexistence systems through dynamically adjusting energy detection (ED) thresholds. Our results show that both CDRL and FDRL approaches have improved the reliability of the NR-U system significantly, but the CDRL framework has sacrificed the reliability of the WiFi system. To guarantee the reliability of the WiFi system while improving the NR-U system, we take fairness into account by redesigning the reward of CDRL.

Original languageEnglish
Pages (from-to)2208-2222
Number of pages15
JournalIEEE Journal on Selected Areas in Communications
Issue number7
Publication statusPublished - 1 Jul 2023


  • channel access
  • deep reinforcement learning
  • NR-U
  • WiFi


Dive into the research topics of 'Channel Access Optimization in Unlicensed Spectrum for Downlink URLLC: Centralized and Federated DRL Approaches'. Together they form a unique fingerprint.

Cite this