基于CPU-GPU的软件无线电基带并行算法与通信系统架构研究

随着无线通信网络的发展，通信协议日益繁多，迫切需要支持多协议多模式通信的软件无线电系统。近年来，基于通用硬件的软件无线电系统由于其成本低、通用性强、编程难度低等优势受到学术界和产业界的重视，而CPU-GPU异构结构成为其中的研究热点。本项目以提高软件无线电系统传输性能和灵活性为目标，基于移动CPU-GPU异构平台，开展软件无线电基带处理并行算法和通信系统架构研究，旨在解决软件无线电系统成本高、通用性差、编程难度高的问题，推动软件无线电走向通用化。首先，基于CPU-GPU异构结构特点和基带算法特征，构建基于CPU-GPU的基带算法并行映射模型；然后以该模型为基础，设计基于CPU-GPU的参数化5G基带并行算法；最后，研究基于CPU-GPU的多模式MIMO软件无线电原型系统，验证模型和算法的有效性。

With the development of wireless communication networks, communication protocols are becoming increasingly numerous, and there is an urgent need to support multi-protocol multi-mode software defined radio (SDR) system. In recent years, the SDR system based on general hardware has been paid much attention by academia and industry because of its low cost, versatility and low programming difficulty. In this research area, CPU-GPU heterogeneous structure has become one of the research hotspots. In order to improve the transmission performance and flexibility of SDR system and to solve the problem of high cost, low versatility and high programming complexity, this research focuses on the parallel baseband algorithms and communication system architecture based on mobile CPU-GPU heterogeneous platform, which can promote the versatility of the SDR system. Firstly, based on the characteristics of baseband algorithms, a parallel mapping model of baseband algorithms based on CPU-GPU is constructed. Then, derived from the model, CPU-GPU based parametric 5G parallel baseband algorithms are designed. At last, we propose multi-mode MIMO SDR prototype system based on mobile CPU-GPU heterogeneous architecture, which can verify the validity of the presented model and algorithms.