大规模MIMO系统中训练序列优化算法的研究
发布时间：2018-08-20 09:39
【摘要】：5G是面向2020年以后移动通信需求而发展的新一代移动通信系统,其核心技术包括高能效的无线传输技术以及高密度无线网络技术,其中重点表现为大规模MIMO技术。大规模MIMO系统具有超高的频谱利用率和能量效率等优点,自被提出之后就获得了广泛而深入的研究,是当前通信领域研究的重点内容。然而,大规模MIMO系统实现超高性能增益的前提是系统可以正确地获得信道状态信息,随着发射天线数增多,大规模MIMO系统无法承担信道空间相关信息的反馈开销,此时训练序列的计算发生在用户设备端而非基站端,基站由于不具备信道长期统计信息而需要不断接收反馈回来的训练信号。本文正是针对大规模MIMO系统中基站不具备信道长期统计信息这一问题,研究FDD大规模MIMO系统中下行链路的训练算法优化问题。对于传统训练算法在大规模MIMO系统存在着性能上的限制。本文研究了开环训练算法对于大规模MIMO系统的适用性问题,在推导了开环单射情况下使得信道估计均方误差最小的训练序列结构后,针对优化后的训练序列,本文通过理论推导和实际仿真证明了开环单射训练方法存在的天花板效应,大规模MIMO系统中训练序列固定时系统的归一化接收信噪比将受到限制。使用开环有记忆的训练方法可以缓解开环单射训练所存在的性能限制,使用卡尔曼滤波来预测信道的变化信息可以提升大规模MIMO系统中信道估计的准确度。为了解决大规模MIMO系统中基站端无法完成训练序列计算的问题,本文重点研究了大规模MIMO系统中的闭环有记忆训练算法。研究了使得信道估计性能最好的训练序列结构设计方法,通过寻找高性能的训练信号集以及使得MSE最小的全反馈训练信号,使得信道估计的性能得到提升。仿真结果表明闭环有记忆训练方法可以在只增加几比特反馈开销的前提下使得信道估计更准确,当用户端反馈全部的训练信号结构时,系统的下行链路可以进一步获得更高的信道估计性能。本文创新性地提出了引入功率分配策略的闭环有记忆训练算法,基站可以在不具备信道长期统计信息的前提下完成更加准确的下行链路信道估计。通过计算机仿真评估了新算法的信道估计均方误差性能指标。提出的算法可以在系统性能与链路开销取得较好的折中,适用于大规模MIMO系统中的训练设计问题。
[Abstract]:5G is a new generation of mobile communication system, which is developed to meet the requirement of mobile communication after 2020. Its core technologies include high energy efficiency wireless transmission technology and high density wireless network technology, in which the emphasis is on large-scale MIMO technology. Large-scale MIMO system has the advantages of ultra-high spectral efficiency and energy efficiency, and has been widely and deeply studied since it was proposed, which is the focus of the research in the field of communication. However, the premise of realizing super-high performance gain in large-scale MIMO systems is that the channel state information can be obtained correctly by the system. With the increase of the number of transmitting antennas, the large-scale MIMO system cannot afford the feedback overhead of the space-related information of the channel. In this case, the calculation of the training sequence takes place at the end of the user equipment rather than the base station, and the base station needs to receive the feedback training signal because of the lack of long-term statistical information of the channel. Aiming at the problem that the base station does not have the long-term statistical information of the channel in the large-scale MIMO system, this paper studies the optimization of the downlink training algorithm in the FDD large-scale MIMO system. There is a limitation on the performance of traditional training algorithms in large scale MIMO systems. In this paper, the applicability of open-loop training algorithm to large-scale MIMO systems is studied. After the structure of training sequence with minimum mean square error of channel estimation is derived in the case of open-loop monojection, the optimized training sequence is proposed. In this paper, the ceiling effect of open-loop single-shot training method is proved by theoretical derivation and practical simulation, and the normalized reception SNR of large-scale MIMO system is limited when the training sequence is fixed. The performance limitation of open-loop single-shot training can be alleviated by using open-loop memory training method. Using Kalman filter to predict channel variation information can improve the accuracy of channel estimation in large-scale MIMO systems. In order to solve the problem that the base station can not complete the training sequence calculation in the large-scale MIMO system, this paper focuses on the closed-loop memory training algorithm in the large-scale MIMO system. The training sequence structure design method which makes the channel estimation performance the best is studied. The performance of channel estimation is improved by searching for the high performance training signal set and making the MSE minimum full feedback training signal. The simulation results show that the closed-loop memory training method can make the channel estimation more accurate under the condition of adding only a few bits of feedback overhead. The downlink of the system can further achieve higher channel estimation performance. In this paper a novel closed-loop memory training algorithm with power allocation strategy is proposed. The base station can complete more accurate downlink channel estimation without long-term statistical information of the channel. The performance index of channel estimation mean square error of the new algorithm is evaluated by computer simulation. The proposed algorithm can achieve a good compromise between system performance and link overhead, and is suitable for training design in large-scale MIMO systems.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.3
[Abstract]:5G is a new generation of mobile communication system, which is developed to meet the requirement of mobile communication after 2020. Its core technologies include high energy efficiency wireless transmission technology and high density wireless network technology, in which the emphasis is on large-scale MIMO technology. Large-scale MIMO system has the advantages of ultra-high spectral efficiency and energy efficiency, and has been widely and deeply studied since it was proposed, which is the focus of the research in the field of communication. However, the premise of realizing super-high performance gain in large-scale MIMO systems is that the channel state information can be obtained correctly by the system. With the increase of the number of transmitting antennas, the large-scale MIMO system cannot afford the feedback overhead of the space-related information of the channel. In this case, the calculation of the training sequence takes place at the end of the user equipment rather than the base station, and the base station needs to receive the feedback training signal because of the lack of long-term statistical information of the channel. Aiming at the problem that the base station does not have the long-term statistical information of the channel in the large-scale MIMO system, this paper studies the optimization of the downlink training algorithm in the FDD large-scale MIMO system. There is a limitation on the performance of traditional training algorithms in large scale MIMO systems. In this paper, the applicability of open-loop training algorithm to large-scale MIMO systems is studied. After the structure of training sequence with minimum mean square error of channel estimation is derived in the case of open-loop monojection, the optimized training sequence is proposed. In this paper, the ceiling effect of open-loop single-shot training method is proved by theoretical derivation and practical simulation, and the normalized reception SNR of large-scale MIMO system is limited when the training sequence is fixed. The performance limitation of open-loop single-shot training can be alleviated by using open-loop memory training method. Using Kalman filter to predict channel variation information can improve the accuracy of channel estimation in large-scale MIMO systems. In order to solve the problem that the base station can not complete the training sequence calculation in the large-scale MIMO system, this paper focuses on the closed-loop memory training algorithm in the large-scale MIMO system. The training sequence structure design method which makes the channel estimation performance the best is studied. The performance of channel estimation is improved by searching for the high performance training signal set and making the MSE minimum full feedback training signal. The simulation results show that the closed-loop memory training method can make the channel estimation more accurate under the condition of adding only a few bits of feedback overhead. The downlink of the system can further achieve higher channel estimation performance. In this paper a novel closed-loop memory training algorithm with power allocation strategy is proposed. The base station can complete more accurate downlink channel estimation without long-term statistical information of the channel. The performance index of channel estimation mean square error of the new algorithm is evaluated by computer simulation. The proposed algorithm can achieve a good compromise between system performance and link overhead, and is suitable for training design in large-scale MIMO systems.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.3
【相似文献】
相关期刊论文 前10条
1 齐祥明;许道峰;杨绿溪;;基于叠加训练序列快速信道辨识方法[J];电路与系统学报;2007年02期
2 陈佑红;张科峰;;基于短训练序列的符号同步改进算法及其实现[J];现代电子技术;2012年11期
3 林韵,肖自美;矢量量化器的典型训练序列方法[J];中山大学学报(自然科学版);1993年04期
4 杜茂森;吉华芳;;几种训练序列相关性能分析[J];信息安全与通信保密;2006年02期
5 罗仁泽;高，
本文编号：2193172
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2193172.html
