基于机器学习的近海广域无线通信资源管理方法研究

高效、高质量的近海广域覆盖无线通信系统是促进海洋经济发展的重要保障，也是未来无线通信技术发展的重要方向之一。近海广域覆盖无线通信场景具有信道不确定，用户分布呈总体稀疏、局部密集的特点。直接应用现有蜂窝架构的陆地基站无线通信网络，无法满足快速发展的近海通信业务需求。本项目拟基于机器学习，通过设计配置波束、功率、时隙等资源，实现近海通信系统的物理层资源优化管理，从而提高近海通信服务能力。针对近海无线信道不确定的特点，采用区域网格划分模型，研究基于强化学习的单网格通信资源管理方法；探索多网格在时域资源上的协同分集，设计基于公平性虚拟队列的通信资源管理方法；基于船只移动与历史决策，设计面向网格需求变化的通信资源管理迁移方法。通过项目研究，旨在提出适应近海通信业务需求的广域无线通信资源管理方法，为保障近海宽带无线网的通信质量，推进近海无线通信技术的应用发展提供参考。

Efficient and high-quality short-sea coverage wireless communications is an important guarantee for promoting the development of the marine economy, which is also envisioned as one of the most important trending of future wireless communication systems. The channels of the short-sea wide-area coverage wireless communications are uncertain, and the user distribution is generally sparse and locally dense. The direct application of the conventional terrestrial cellular networks cannot meet the rapidly growing demand for the short-sea communications. This project will conduct research on the physical layer resource management by optimizing the beamformers, power, and timeslots using machine learning technologies. To deal with the uncertain channel, the system model based on gridding is adopted, and the single-grid resource management is studied by using reinforcement learning. Meanwhile, the cooperation and diversity of time resources for the multi-girds are studied, and a virtual queue with fairness constraint-based resource management is to be designed. Further, a transfer management is studied according to the vessels’ trajectory and the history policies. The goal of the project is to develop a resource management method that meets the requirement of the short-sea communications, and provides a reference for ensuring the communication quality and promoting the application and development of short-sea wireless communication technologies.