大规模MIMO中基于低复杂度雅克比预编码算法及实现  

LOW COMPLEXITY JACOBI BASED PRECODING FOR LARGE-SCALE MIMO

作　　者：王雷[1] 

机构地区：[1]南京理工大学电子工程与光电技术学院,江苏南京210000

出　　处：《计算机应用与软件》2017年第10期294-297,共4页Computer Applications and Software

摘　　要：迫零线性预编码可以获得接近最优的系统容量,不同于传统MIMO系统,大规模MIMO将会配置成百根天线,随着天线数量增加,使得迫零线性预编码矩阵求逆计算复杂,不利于在应用中实现。为了减小线性预编码计算复杂度,提出基于低复杂度的雅克比迭代算法,该算法通过线性迭代,避免了矩阵求逆运算,减少了计算量。为了更进一步的减少计算时间,提出基于统一计算架构的异构多核并行算法,该方法利用GPU具有多核多线程结构特点,实现了异构多核并行计算。仿真结果表明,基于低复杂度雅克比预编码算法可以达到迫零预编码算法性能,同时与传统的线性预编码相比,该算法的计算量更少、时间更短。Linear precoding techniques,such as zero forcing,can achieve the near optimal capacity. In contrast to traditional multiple input multiple output（ MIMO）, large scale MIMO installed hundreds of antennas, with the increasing numbers of antennas,zero forcing precoding involve the matrix inversion of large size with high computational complexity,which may cause difficulty for the realization of the application. To reduce the computational complexity of linear precoding,we propose a low complexity iterative algorithm based on the degree of Jacobi. This method realized through the linear iteration meanwhile avoided the matrix inversion. To further reduce the computation time,we propose a heterogeneous multicore parallel algorithm based on CUDA（ compute unified device architecture）,which leverage the Graphics Processing Unit（ GPU） characteristic of multicores and multithreads to realize heterogeneous multicore architecture parallel computation. Experimental result demonstrates this proposed algorithm not only can achieve the same performance of zero forcing precoding,but also has less computation time and shorter time than traditional linear precoding.

关 键 词：预编码 雅克比 统一计算架构 迫零 异构多核 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]