Base station caching and multicast are two promising methods to support mass content delivery in future wireless network environments. However, existing scheduling designs do not take full advantage of the two methods. This article focuses on employing multicast scheduling and caching in a network architecture which involves both macro cell base stations (MBS) and small cell base stations (SBS) in order to achieve joint optimization of average delay and power consumption. We describe this co-optimization problem as the Multicast-Aware Caching Scheduling Problem (MACSP). This article proposes a novel pending request queue model, which aims to solve the problem of long waiting time for non-popular content, and transform this collaborative multicast-cache scheduling problem into a Markov Decision Process that can be solved using reinforcement learning methods. For actual deployment, the paper further introduces a Distributed Policy Gradient algorithm (DPG) with similar performance and lower complexity. The simulation-based testing results demonstrate that our model and algorithm have better performance and lower energy consumption than existing state-of-the-art approaches.

中文翻译：

---

联合优化组播调度和缓存，以提高无线异构网络的性能和节能

基站缓存和多播是在未来的无线网络环境中支持海量内容交付的两种有前途的方法。但是，现有的调度设计不能充分利用这两种方法。本文重点介绍在涉及宏小区基站（MBS）和小型小区基站（SBS）的网络体系结构中采用多播调度和缓存，以实现平均延迟和功耗的联合优化。我们将此共同优化问题描述为“多播感知缓存调度问题”（MACSP）。本文提出了一种新颖的待处理请求队列模型，旨在解决非热门内容的等待时间长的问题，并将此协作式多播缓存计划问题转化为可以使用强化学习方法解决的马尔可夫决策过程。对于实际部署，本文还介绍了性能相似且复杂度较低的分布式策略梯度算法（DPG）。基于仿真的测试结果表明，与现有的最新方法相比，我们的模型和算法具有更好的性能和更低的能耗。