并行处理技术在通信系统中的应用研究

发布时间：2018-04-18 10:26

本文选题：基带信号处理 + 并行计算　；参考：《北京交通大学》2014年硕士论文

【摘要】：摘要：随着移动通信技术的突飞猛进,基于专门硬件实现的通信设备因其功能单一,出现了通信标准和系统之间不兼容以及升级困难等问题。软件无线电技术采用软件实现无线通信算法,可以灵活的将不同的通信系统在一个通用的硬件平台上实现,但是传统上的软件无线电平台是利用FPGA或DSP等可编程器件进行高速信号处理,其开发难度大,周期长。近年来,基于通用处理器(GPP)的软件无线电平台由于开发难度低并且易于升级等特点为软件无线电的发展提供新方向。但是,目前GPP并不是专为无线信号处理设计,因此基于GPP的软件无线电开发很难满足实时性需求。为此,本文采用计算机领域中高性能计算技术,基于多核处理器和图形处理器的并行处理技术来加速通信系统中典型基带信号的处理速度。这样可以从根本上突破计算瓶颈,使基于GPP的软件无线电平台开发继续向前迈进了一步。本文依托国家科技重大专项三,基于TD-LTE的高速铁路宽带通信的关键技术研究与应用验证,展开了对LTE系统中典型基带算法进行并行处理研究,并设计实现基于多核平台和GPU平台的并行通信算法。论文的主要工作可以归纳为以下几个方面：第一,针对MATLAB仿真平台仿真速度慢的问题,本文提出基于IT++软件平台进行通信系统的建模、分析和设计。并成功搭建基于IT++软件开发平台的LTE下行链路仿真平台,分析查找LTE系统中耗时基带算法。第二,以4核心CPU为硬件平台、Microsoft Visual Studio2010为软件环境,使用C语言搭建基于OpenMP共享存储模型和基于Win32API的多核并行处理技术的并行计算开发平台。第三,在搭建的并行计算平台上面分别设计实现多核并行FFT算法,多核并行Vitebi译码算法,并对性能指标加速比进行测试分析,最终加速比接近4。第四,以Gtx660为硬件开发平台,搭建基于CUDA的GPU并行计算平台,并以Fourier矩阵乘法的方法设计特定需求下的多线程IDFT算法,最终加速增益在187倍左右。
[Abstract]:Absrtact: with the rapid development of mobile communication technology, the communication equipment based on special hardware has many problems, such as the single function, the incompatibility between communication standards and systems, and the difficulty of upgrading.Software radio technology uses software to implement wireless communication algorithms, which can flexibly implement different communication systems on a common hardware platform.However, the traditional software radio platform uses FPGA or DSP to process high speed signal, which is difficult to develop and has a long period.In recent years, the software radio platform based on GPP provides a new direction for the development of software radio because of its low difficulty and easy upgrade.However, at present, GPP is not specially designed for wireless signal processing, so the development of software radio based on GPP is difficult to meet the requirement of real time.Therefore, in this paper, the high performance computing technology in the field of computer and parallel processing technology based on multi-core processor and graphics processor are adopted to accelerate the processing speed of typical baseband signal in communication system.In this way, the computing bottleneck can be fundamentally broken, and the development of software radio platform based on GPP is further advanced.Based on the research and application verification of the key technology of high-speed railway broadband communication based on TD-LTE, this paper carries out parallel processing of typical baseband algorithm in LTE system.The parallel communication algorithm based on multi-core platform and GPU platform is designed and implemented.The main work of this paper can be summarized as follows:First, aiming at the slow simulation speed of MATLAB simulation platform, this paper proposes a communication system modeling, analysis and design based on IT software platform.The LTE downlink simulation platform based on IT software development platform is successfully built, and the time-consuming baseband algorithm in the LTE system is analyzed.Secondly, using 4 core CPU as hardware platform and Microsoft Visual Studio2010 as software environment, C language is used to build parallel computing development platform based on OpenMP shared storage model and multi-core parallel processing technology based on Win32API.Thirdly, the multi-core parallel FFT algorithm and the multi-core parallel Vitebi decoding algorithm are designed and implemented on the parallel computing platform, and the speedup of the performance index is tested and analyzed, and the final speedup is close to 4.Fourthly, using Gtx660 as hardware development platform, the GPU parallel computing platform based on CUDA is built, and the multithreaded IDFT algorithm is designed by Fourier matrix multiplication method. Finally, the acceleration gain is about 187 times.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN929.5

【参考文献】