Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Traffic Priority-Aware 5G NR-U/Wi-Fi Coexistence with Deep Reinforcement Learning

Mohammad Reza Fasihi, Brian L. Mark

TL;DR

This work addresses the coexistence of 5G NR-U and Wi-Fi by first benchmarking leading collision-resolution schemes and then extending the best-performing baseline (gCR-LBT) to support traffic-priority-aware channel access. It introduces two approaches: a Dynamic Transmission Skipping method that protects high-priority traffic by forcing lower-priority transmissions to skip, and a Multi-Objective Deep Q-Network (MO-DQN) that jointly optimizes high-priority latency and network fairness using a scalarized objective. The MO-DQN framework enables two network agents (NR-U gNB and Wi-Fi AP) to adapt contention windows and cross-network parameters to maintain low PC1 delay while achieving higher Jain’s fairness between NR-U and Wi-Fi. Simulation results validate that the proposed methods can reduce high-priority latency and improve fairness, with the α parameter enabling a tunable trade-off. Ongoing work targets multi-channel scenarios and joint PHY-layer parameter tuning to mitigate cross-channel leakage effects.

Abstract

Coexistence of 5G new radio unlicensed (NR-U) and Wi-Fi is highly prone to the collisions among NR-U gNBs (5G base stations) and Wi-Fi APs (access points). To improve performance and fairness for both networks, various collision resolution mechanisms have been proposed to replace the simple listen-before-talk (LBT) scheme used in the current 5G standard. We address two gaps in the literature: first, the lack of a comprehensive performance comparison among the proposed collision resolution mechanisms and second, the impact of multiple traffic priority classes. Through extensive simulations, we compare the performance of several recently proposed collision resolution mechanisms for NR-U/Wi-Fi coexistence. We extend one of these mechanisms to handle multiple traffic priorities. We then develop a traffic-aware multi-objective deep reinforcement learning algorithm for the scenario of coexistence of high-priority traffic gNB user equipment (UE) with multiple lower-priority traffic UEs and Wi-Fi stations. The objective is to ensure low latency for high-priority gNB traffic while increasing the airtime fairness among the NR-U and Wi-Fi networks. Our simulation results show that the proposed algorithm lowers the channel access delay of high-priority traffic while improving the fairness among both networks.

Traffic Priority-Aware 5G NR-U/Wi-Fi Coexistence with Deep Reinforcement Learning

TL;DR

This work addresses the coexistence of 5G NR-U and Wi-Fi by first benchmarking leading collision-resolution schemes and then extending the best-performing baseline (gCR-LBT) to support traffic-priority-aware channel access. It introduces two approaches: a Dynamic Transmission Skipping method that protects high-priority traffic by forcing lower-priority transmissions to skip, and a Multi-Objective Deep Q-Network (MO-DQN) that jointly optimizes high-priority latency and network fairness using a scalarized objective. The MO-DQN framework enables two network agents (NR-U gNB and Wi-Fi AP) to adapt contention windows and cross-network parameters to maintain low PC1 delay while achieving higher Jain’s fairness between NR-U and Wi-Fi. Simulation results validate that the proposed methods can reduce high-priority latency and improve fairness, with the α parameter enabling a tunable trade-off. Ongoing work targets multi-channel scenarios and joint PHY-layer parameter tuning to mitigate cross-channel leakage effects.

Abstract

Coexistence of 5G new radio unlicensed (NR-U) and Wi-Fi is highly prone to the collisions among NR-U gNBs (5G base stations) and Wi-Fi APs (access points). To improve performance and fairness for both networks, various collision resolution mechanisms have been proposed to replace the simple listen-before-talk (LBT) scheme used in the current 5G standard. We address two gaps in the literature: first, the lack of a comprehensive performance comparison among the proposed collision resolution mechanisms and second, the impact of multiple traffic priority classes. Through extensive simulations, we compare the performance of several recently proposed collision resolution mechanisms for NR-U/Wi-Fi coexistence. We extend one of these mechanisms to handle multiple traffic priorities. We then develop a traffic-aware multi-objective deep reinforcement learning algorithm for the scenario of coexistence of high-priority traffic gNB user equipment (UE) with multiple lower-priority traffic UEs and Wi-Fi stations. The objective is to ensure low latency for high-priority gNB traffic while increasing the airtime fairness among the NR-U and Wi-Fi networks. Our simulation results show that the proposed algorithm lowers the channel access delay of high-priority traffic while improving the fairness among both networks.

Paper Structure

This paper contains 7 sections, 2 equations, 9 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Evaluation of Collision Resolution Schemes
  3. Traffic Priority-Aware Channel Access
  4. Dynamic Transmission Skipping Method
  5. Multi-Objective DQN Method
  6. Simulation Results
  7. Conclusion

Figures (9)

  • Figure 1: Intra-gNBs collisions probability for NR-U.
  • Figure 2: Channel efficiency of NR-U.
  • Figure 3: Channel access delay of NR-U.
  • Figure 4: Jain's fairness index between Wi-Fi and NR-U.
  • Figure 5: Example of gNB UE running gCR-LBT Loginov:2021 with a) no skipping, b) skipping to the next slot boundary, and c) skipping to the next transmission opportunity.
  • ...and 4 more figures