大规模MIMO系统中的低复杂度检测算法研究

发布时间：2018-04-28 20:14

本文选题：大规模MIMO + 低复杂度　；参考：《西安电子科技大学》2015年硕士论文

【摘要】：大规模多输入多输出(MIMO,multiple-input multiple-output)系统通过在基站端配置数十甚至上百根天线,可以显著提高系统的频谱效率和能量效率,增强系统的可靠性,被认为是未来无线通信,尤其是第五代(5G,the 5th generation)移动通信的关键技术之一。但是,天线的大规模增加也带来了运算复杂度的大幅度增加。因此,如何设计低复杂度高性能的检测算法是大规模MIMO系统上行链路需要解决的关键问题之一。在大规模MIMO系统中,对于基站天线数远大于用户数的场景,传统MIMO系统中的线性检测算法就可以取得理想的检测性能,但是会涉及到复杂的矩阵求逆运算。而对于用户数较多,甚至与基站天线数相当的场景,需要进一步研究能取得理想性能的低复杂度检测算法。本文围绕上述大规模MIMO系统中检测算法存在的问题,对现有大规模MIMO系统中的典型算法进行了研究,并针对大规模MIMO系统中的两种不同场景分别提出了改进算法。主要的研究内容及研究成果为:1.对低复杂度大规模MIMO检测算法进行研究。首先给出了大规模MIMO系统的模型,并简要介绍了传统MIMO系统中的经典线性检测算法,包括匹配滤波(MF,matched filter)、迫零(ZF,zero forcing)和最小均方误差(MMSE,minimum mean squared error)算法。然后介绍了三种近年来从神经网络和图像处理领域引入的大规模检测算法,分别为似然上升搜索算法(LAS,likelihood ascent search)、马尔科夫链-蒙特卡罗(MCMC,markov chain monte carlo)算法和置信传播(BP,belief propagation)算法,并对它们的性能进行了仿真研究。2.针对基站天线数远大于用户数的场景,提出了一种避免矩阵求逆的MMSE改进算法,称之为松弛迭代检测(RID,relaxation iteration detection)算法。该改进算法利用松弛迭代方法和大规模MIMO系统中的信道特性,使用迭代的方式避免了传统MMSE算法中的矩阵求逆运算,并给出了一种优化的初始解,加快了RID算法的收敛速度,从而进一步降低了算法的计算复杂度。另外,考虑到信道非理想的情况,结合RID算法的特点,指出需要进行信道估计时应采用的信道估计方法,并分析了存在大尺度衰落情况的RID算法及信道估计误差对RID算法的影响。3.针对用户数较多,甚至与基站天线数相当的场景,为了进一步降低消息传递检测(MPD,message passing detection)算法的计算复杂度,提出了一种MPD改进(IMPD,improved MPD)算法。所提的IMPD算法分别从初始化方式、迭代更新方式及收敛条件三个方面对MPD算法进行了优化,进一步加快了MPD算法的收敛速度,从而实现了计算复杂度的降低。仿真结果验证了所提IMPD算法在降低计算复杂度的同时可以实现非常接近于MPD算法的检测性能。
[Abstract]:Large scale Mimo multiple-input multiple-output (Mimo) multiple-input multiple-output) systems can significantly improve the spectral efficiency and energy efficiency of the system by configuring dozens or even hundreds of antennas at the base station, and enhance the reliability of the system. It is considered to be the future wireless communication. In particular, the fifth generation 5 G 5th generation is one of the key technologies of mobile communication. However, the large-scale increase of antenna also brings a large increase in computational complexity. Therefore, how to design a detection algorithm with low complexity and high performance is one of the key problems that need to be solved in the uplink of large-scale MIMO systems. In large-scale MIMO system, the linear detection algorithm in traditional MIMO system can obtain ideal detection performance for the scene where the number of base station antennas is far larger than the number of users, but it will involve complex matrix inversion operation. For the scenario with large number of users or even equivalent to the number of base station antennas, it is necessary to further study the low complexity detection algorithm which can achieve ideal performance. This paper focuses on the problems of detection algorithms in large-scale MIMO systems, studies the typical algorithms in existing large-scale MIMO systems, and proposes two improved algorithms for two different scenarios in large-scale MIMO systems. The main research contents and results are as follows: 1. A large scale MIMO detection algorithm with low complexity is studied. Firstly, the model of large-scale MIMO system is given, and the classical linear detection algorithm in traditional MIMO system is introduced briefly, including matching filtering mf matched filter, zero forcing and minimum mean square error (MMSE) minimum mean squared error) algorithm. Then, three large scale detection algorithms introduced from neural networks and image processing in recent years are introduced. They are likelihood ascent search algorithm, Markov chain Monte Carlo MCMCmarkov chain monte carlo) algorithm and confidence Propagation algorithm. And their performance is simulated. 2. 2. For the scenario where the number of base station antennas is far larger than the number of users, an improved MMSE algorithm is proposed to avoid matrix inversion, which is called relaxation iteration detection algorithm. The improved algorithm uses relaxation iterative method and channel characteristics in large-scale MIMO system to avoid matrix inversion in traditional MMSE algorithm, and gives an optimized initial solution, which speeds up the convergence of RID algorithm. Thus, the computational complexity of the algorithm is further reduced. In addition, considering the non-ideal channel situation and combining the characteristics of RID algorithm, the paper points out the channel estimation method that should be used when channel estimation is needed. The RID algorithm with large scale fading and the influence of channel estimation error on RID algorithm are analyzed. In order to further reduce the computational complexity of message passing detection algorithm, an improved MPD improved passing algorithm is proposed for scenarios with large number of users, even equivalent to the number of base station antennas. The proposed IMPD algorithm optimizes the MPD algorithm from three aspects: initialization, iterative updating and convergence conditions, which further speeds up the convergence of the MPD algorithm and reduces the computational complexity. The simulation results show that the proposed IMPD algorithm can reduce the computational complexity and achieve very similar detection performance to that of the MPD algorithm.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN919.3

【相似文献】