Recently, the installation of 5G networks offers a variety of real-time, high-performance and human-oriented customized services. However, the current laying 5G structure is unable to meet all of the growing communication needs by these new emerging services. In this paper, we propose a DQL (Deep Q-learning Network) based intelligent resource management method for 5G architecture, to improve the quality of service (QoS) under limited communication resources. In the environment of network function virtualization (NFV), we aim at improving the efficient usage of spectrum resources. In this two-step solution, our first goal is to guarantee the maximum communication quality with the smallest number of infrastructures. Then, a DQL-based wireless resource allocation algorithm is designed to realize the elaborate operation. Unlike previous studies, our system can provide the allocation policy in a more subdivided way and finally maximize the usage of bandwidth resources. The simulation also shows that our proposed MSIO improves 3.12% in the performance of the maximum coverage importance problem and the ARODQ algorithm improves 4.05% than other standard solutions.

中文翻译：

---

5G 中的细粒度管理：基于 DQL 的 C-RAN 网络功能虚拟化智能资源分配

近期，5G网络的安装提供了多种实时、高性能、人性化的定制化服务。然而，目前正在铺设的 5G 架构无法满足这些新兴业务不断增长的通信需求。在本文中，我们为 5G 架构提出了一种基于 DQL（深度 Q 学习网络）的智能资源管理方法，以在有限的通信资源下提高服务质量（QoS）。在网络功能虚拟化（NFV）环境中，我们的目标是提高频谱资源的使用效率。在这个两步式解决方案中，我们的首要目标是以最少的基础设施数量保证最高的通信质量。然后设计了一种基于DQL的无线资源分配算法来实现精细化操作。与以往的研究不同，我们的系统可以以更细分的方式提供分配策略，最终最大限度地利用带宽资源。仿真还表明，我们提出的 MSIO 在最大覆盖重要性问题的性能上提高了 3.12%，ARODQ 算法比其他标准解决方案提高了 4.05%。