Digital twin‐enabled deep reinforcement learning for joint scheduling of ultra‐reliable low latency communication and enhanced mobile broad band: A reliability‐guaranteed approach

Abstract

In the coexistence of ultra‐reliable low latency communication (URLLC) and enhanced mobile broad band (eMBB) in 5G networks, the arriving URLLC traffic with strict latency requirements will be scheduled by puncturing ongoing eMBB transmissions, negatively impacting eMBB data rate. In this article, we add reliability measurement for eMBB users with high data rate requirements. The scheduling problem is formulated as an optimization problem with the goal of maximizing the data rate of eMBB users while meeting the requirements of URLLC latency and eMBB data rate. To jointly optimize resource allocation policy and puncturing policy, an algorithm based on deep reinforcement learning (DRL) is introduced. Considering that the utilization of DRL in the resource scheduling problem is limited due to the nonstationary communication environment, a digital twin‐enabled DRL architecture is presented to fine‐tune the DRL model according to feedback from the real‐world network. The DRL agent can explore in the digital twin model, avoiding the loss of quality of service caused by exploration in the real‐world environment. According to simulation results, the approach proposed in this article can improve the reliability of eMBB users with data rate requirements.