Resource Allocation of URLLC and eMBB Mixed Traffic in 5G Networks: A Deep Learning Approach

Ultra-reliable low-latency communication (URLLC) has been considered as a major use case for the fifth generation (SG) wireless networks. Therefore, the Third Generation Partnership Project (3GPP) has targeted the support of URLLC in the new radio (NR) air interface by introducing several technologies such as short transmission time intervals (sTTIs) and puncturing scheduling. However, scheduling URLLC without impacting the quality-of-service (QoS) of enhanced mobile broadband (eMBB) traffic is a challenging task. In this paper, we optimize the resource allocation and scheduling process of URLLC puncturing eMBB transmissions by considering the QoS of eMBB and the transmission errors associated with the finite blocklength coding of the URLLC traffic. In addition, we propose a deep supervised learning approach to predict the optimized resource allocation in a computationally efficient manner to be practically used in real-time operation. The numerical results show that by adjusting the model parameters, we can increase the accuracy of the low-complexity predictions for an efficient scheduling scheme with superior performance as compared to other techniques.