A Deep-Reinforcement-Learning-Based Approach to Dynamic eMBB/URLLC Multiplexing in 5G NR

Authors:
Publisher website
Google Scholar
Abstract
This paper investigates dynamic multiplexing of enhanced Mobile BroadBand (eMBB) and Ultra-Reliable and Low Latency Communications (URLLC) on the same channel in 5G NR. Due to significant difference in transmission time scale, URLLC employs a preemptive puncturing technique to multiplex its traffic onto eMBB traffic for transmission. The optimization problem to solve is to minimize the adverse impact of such preemptive puncturing on eMBB users. We present DEMUX – a model-free deep reinforcement learning (DRL) based solution to this problem. The essence of DEMUX is to use deep function approximators (neural networks) to learn an optimal algorithm for determining preemption solution in each eMBB transmission time interval (TTI). Our novel contributions in the design of DEMUX include the first use of DRL method with large and continuous action domain for resource scheduling in NR, mechanism to ensure fast and stable learning convergence by exploiting intrinsic properties of the problem, and mechanism to obtain feasible preemption solution from unconstrained output of a neural network while minimizing loss of information. Experiment results show that DEMUX significantly outperforms state-of-the-art algorithms proposed in the 3GPP standards body and the literature.
Funding Information
  • NSF (1642873)

This publication has 15 references indexed in Scilit: