面向6G空口信号处理的新一代GPU基带研究

Research on the Next-generation GPU Baseband for 6G Air Interface Signal Processing

在6G网络中,随着新兴业务和场景的不断涌现,以及人工智能内生需求的日益增加,6G无线空口的灵活性成为当前研究的重点。这种灵活性要求软件算法与硬件平台的协同发展,然而现有的研究多集中在软件算法层面,对硬件平台的探索相对不足。面向6G空口信号处理,研究了基于GPU硬件平台的空口基带实现的可行性。首先基于GPU的特点及基带信号处理的要求,分析了GPU作为基带处理单元的可行性。随后选取吞吐量和算术强度作为GPU基带的关键性能指标进行分析,并基于Sionna平台对吞吐量部分进行了初步仿真验证。最后,总结了研究工作并对未来6G无线空口技术的发展进行了展望。

In 6G networks,with the continuous emergence of new services and scenarios,as well as the increasing demand for AI-native capabilities,the flexibility of the 6G radio air interface has become a key research focus.This flexibility requires the coordinated development of software algorithms and hardware platforms.However,current research has predominantly concentrated on the software algorithm layer,with relatively insufficient exploration of hardware platforms.Aiming at 6G air interface signal processing,this paper investigates the feasibility of an air interface baseband implementation based on a GPU hardware platform.First,the feasibility of using GPUs as baseband processing units is analyzed based on the characteristics of GPUs and the requirements of baseband signal processing.Subsequently,throughput and arithmetic intensity are selected as the key performance indicators for the GPU-based baseband,and preliminary simulations are conducted on the Sionna platform to validate throughput performance.Finally,the research work in this paper is summarized and the development of future 6G wireless air interface technology is prospected.