Space-air-ground integrated network (SAGIN) is a new type of wireless network mode. The effective management of SAGIN resources is a prerequisite for high-reliability communication. However, the storage capacity of space-air network segment is extremely limited. The air servers also do not have sufficient storage resources to centrally accommodate the information uploaded by each edge server. So the problem of how to coordinate the storage resources of SAGIN has arisen. This paper proposes a SAGIN storage resource management algorithm based on distributed deep reinforcement learning (DRL). The resource management process is modeled as a Markov decision model. In each edge physical domain, we extract the network attributes represented by storage resources for the agent to build a training environment, so as to realize the distributed training. In addition, we propose a SAGIN resource management framework based on distributed DRL. Simulation results show that the agent has an ideal training effect. Compared with other algorithms, the resource allocation revenue and user request acceptance rate of the proposed algorithm are increased by about 18.15\% and 8.35\% respectively. Besides, the proposed algorithm has good flexibility in dealing with the changes of resource conditions.
翻译:空-空-地综合网络(SAGIN)是一种新型的无线网络模式。对SAGIN资源的有效管理是高可靠性通信的一个先决条件。然而,空间-空网络部分的存储能力极为有限。空气服务器也没有足够的存储资源来集中接收每个边缘服务器上传的信息。因此产生了如何协调SAGIN储存资源的问题。本文件提议了一个基于分布式强化学习的SAGIN储存资源管理算法(DRL) 。资源管理程序以Markov决定模型为模型。在每一个边缘物理领域,我们提取以存储资源为代表的网络属性,以便代理建立培训环境,从而实现分布式培训。此外,我们还提议了一个基于分布式DRL的SAGIN资源管理框架。模拟结果表明,该代理具有理想的培训效果。与其他算法相比,拟议的算法的资源分配收入和用户要求接受率分别增加了大约18.15 ⁇ 和8.35 ⁇ 。此外,拟议的算法在应对资源条件的变化方面具有良好的灵活性。