Joint Resource Allocation and Adaptive Task Scheduling in 5G Using Hybrid Deep Reinforcement Learning

Abstract

In this paper, we design a hybrid deep reinforcement learning strategy that takes into account the problems related to combined resource allocation and task scheduling in dynamic 5G systems. Using a combination of DDQN and Prioritized Experience Replay (DDQN-PER), we create an effective approach that manages both eMBB and URLLC efficiently, guaranteeing high rates and low latencies. Specifically, we model the problem of scheduling using a Markov Decision Process to make the method adaptive under varying traffic and channel conditions. Our simulation results show the superiority of the proposed solution, yielding an optimal accuracy of 97%, reducing the latencies by up to 35%, and improving spectral efficiency. The developed solution tackles the problem of heterogeneity in 5G services, as the proposed approach deals with eMBB, URLLC, and mMTC simultaneously in a unified network environment. With the help of the hybrid DRL technique, the scheme is capable of learning from changes and achieving fairness in traffic distribution.