| 摘要: |
| 为解决软件化雷达系统实时处理大规模数据的问题，提出了一种分层级的分布式并行计算方法，并设计了一种低延时大规模数据处理能力的软件化雷达系统。该系统采用三层并行计算方法，利用ZeroMQ技术实现了任务级的分布式计算，多线程技术实现了线程级的多核并行计算，Arrayfire平台实现了数据级的图形处理单元(Graphic Processing Unit,GPU)并行计算。系统具有部署灵活、扩展性强、并发性高、延时低的特征。实验结果表明，数据率为1.92 Gb/s时，系统能在30 ms内实时显示空中目标，完全满足实际工程应用环境需求。 |
| 关键词: 软件化雷达 分布式并行计算 多线程 |
| DOI： |
|
| 基金项目: |
|
| A software radar system with hierarchical parallel computing |
| ZHAO Bang,LI Wuxu,ZHAO Haoran |
| (Sichuan Jiuzhou Air Traffic Control Technology Co.,Ltd.,Mianyang 621000,China) |
| Abstract: |
| In order to solve the problem of real-time processing of large-scale data in software radar system,a hierarchical distributed parallel computing method is proposed and a software radar system with low-latency large-scale data processing capabilities is designed.The system adopts a three-layer parallel computing method.ZeroMQ technology is used to implement distributed computing based on task level.Multi-threading technology is used to realize parallel computing based on thread level.The Arrayfire platform is used to implement parallel computing based on graphic processing unit(GPU) and data level.The system has the characteristics of flexible deployment,strong scalability,high concurrency,and low latency.Experimental results show that when the data rate is 1.92 Gb/s,the system can display air targets in real time within 30 milliseconds,which fully meets the actual engineering application environment requirements. |
| Key words: software radar distributed parallel computing mult-threading |
《电讯技术》编辑部 