动态传输环境下基于在线学习的高效混合波束赋形技术研究

数据驱动的混合波束赋形因其极高的设计效率成为了当前的研究热点，但其现有相关研究工作都是基于静态传输环境的理想假设。而在真实的无线传输中，因散射体或用户的位移等因素，传输环境是动态变化的。但因动态传输环境下的信道特征的统计分布具有时变特性，普遍采用固定设计模型的现有方法无法直接应用到该环境中。不同于静态传输环境下的数据驱动方法，动态传输环境下的设计模型不仅要保证设计效率，还要及时地适应不断变化的传输环境以保证传输性能。为此，申请人选取三个典型动态传输环境场景（时变的传播路径角度分布、路径增益的统计分布和传播路径数目），以传输任务产生的实时信道数据流作为训练数据，结合每个动态场景中传输环境变化的特点，尝试以在线学习的研究思路赋予混合波束赋形设计模型适应传输环境各类变化的能力。该项目力争突破现有数据驱动的混合波束赋形方法的局限性，为有效解决动态传输环境下的混合波束赋形问题提供新的方法和研究范式。

Due to the super-high design efficiency, the data-driven hybrid beamforming becomes a research hotspot. But the existing works are almost based on the ideal assumption of the static transmission environment. However, in the real wireless transmission, due to movements of scattering bodies or users, the transmission environment is dynamic. Since the statistical distribution of channel characteristics is time-varying in the dynamic transmission environment, the existing works, which are almost based on the static design model, cannot work directly in the environment. Different to the data-driven methods in the static transmission environment, the design model in the dynamic transmission environment not only has to guarantee the design efficiency, but also has to adapt to the varying transmission environment timely for guaranteeing the transmission performance. Therefore, the applicant picks up three typical cases of the dynamic transmission environment, namely, time-varying angle distribution of propagation paths, time-varying distribution of path gain and time-varying number of propagation paths, by taking the real-time channel data flow generated by the real transmission tasks as the training data, combining the characteristics of each case of the varying transmission environment, and tries to give the design model of data-driven hybrid beamforming the ability of adapting to variant changes of the transmission environment, via the idea of on-line learning. This project tries to break through the limitations of the existing data-driven hybrid beamforming methods, and provides the effective solutions and new research paradigm for hybrid beamforming under the dynamic transmission environment.