Hybrid Puncturing and Superposition Scheme for Multiplexing uRLLC and eMBB Services Based on Deep Reinforcement Learning

Abstract

In 5G network, enhanced mobile broadband (eMBB) service requires high data rate while ultra reliable low latency communication (uRLLC) service requires low latency and high reliability. Multiplexing two heterogeneous services with quite different Quality of Service(Qos) requirements on the same wireless radio resource is a challenging wireless resource allocation problem. In this paper, a hybrid superposition and puncturing scheme is adopted to realize the coexistence of eMBB and uRLLC. The user pairing and the radio resource allocation problem is formulated as a non-convex problem aims at minimizing the rate loss of eMBB caused by uRLLC. We propose Proximal Policy Optimization(PPO), a state-of-the-art deep reinforcement learning (DRL) algorithm, to solve this non-convex problem. Simulation results show that our proposed scheme can minimize the loss of eMBB users while strictly ensuring the latency and reliability requirements of uRLLC users, when compared to other baseline schemes.