A Study on Deep Learning Based Resource Allocation for Future Cellular Networks

随着5G的传输速率和网络容量的显著提高，甚至预计2030年的6 G，越来越多的终端将连接到互联网以服务于用户。而新兴的网络业务对时延、安全性、丢包率等指标都有用户感知或内容感知的需求，网络资源的复杂性和动态性要求网络资源分配方法比传统的数学模型更加先进，以满足日益严格和多样化的业务需求。人工智能（AI）方法，特别是最先进的深度学习技术，被认为是唯一的方法，因为它可以有效和灵活地解决极其复杂的问题。基于人工智能的网络管理已被视为6 G的网络范式。本项目的目的是研究未来蜂窝网络中基于人工智能的资源分配，在项目的第一年，对现有的研究进行了广泛的调查和研究。并取得了一些成熟的成果，发表在IEEE高影响因子期刊/学报上。人工智能技术已被证明是有效的智能分配网络资源的蜂窝网络。

With the significantly improved transmission rate and network capacity for 5G and even the expected 6G in 2030, an increasing number of end terminals will be connected to the Internet to serve the users. And the emerging network services are expected to have user-aware or content-aware requirements for various metrics including delay, security, packet loss rate, and so on. To meet the stringent and diversified service requirements, the network resource needs to be allocated by more advanced methods than conventional mathematical models due to the extreme complexity and frequent dynamics. The Artificial Intelligence (AI) methods, especially the state-of-the-art deep learning technologies, have been regarded as the only method since it can tackle the extremely complex problems efficiently and flexibly. And AI-based network management has been regarded as the network paradigm toward 6G. The purpose of this project is to study the AI-based resource allocation for future cellular networks.In the first year of this project, the existing research has been widely surveyed and studied. And some mature results have been obtained and published on and submitted to the high impact factor IEEE journals/transactions. The AI techniques have been demonstrated to be effective for intelligently allocate the network resource for cellular networks.