Deep Reinforcement Learning-Based Optimization for End-to-End Network Slicing With Control- and User-Plane Separation

Control- and user-plane separation (CUPS) network slicing are two key technologies to support increasing traffic diverse wireless services. However, the benefit of CUPS in decoupling coverage data service functions has not been fully utilized facilitate slicing. In this paper, we present a novel CUPS-based end-to-end (CUPS-E2E) scheme. First, base stations (BS) classified into control BSs (CBS) that provide plane (CP) (TBS) deliver user (UP) traffic. Next, upon CBSs TBSs being virtualized, define four typical (E2E) slices: one for CP coverage, high-throughput services, computation-intensive other delay-sensitive The utilities E2E slices defined based on their throughput, computing capability delay requirements, respectively. Then, deep deterministic policy gradient (DDPG)-based algorithm is employed maximize long-term sum-utility by jointly optimizing allocation communication resources activation virtual TBSs, while meeting requirements all users. Simulation results show our proposed CUPS-E2E scheme conjunction with DDPG-based maximization can wide-coverage massive access as well UP high-throughput, services simultaneously, outperforms existing schemes terms sum-utility, percentage, throughput delay.