| 摘要: |
| 为了减少毫米波大规模多输入多输出（Multiple-Input Multiple-Output，MIMO）系统在数据传输过程中所使用的射频链数目，解决高硬件成本和高功耗的问题，可以将透镜天线阵列作为一个实际有效的解决方案部署在毫米波MIMO系统的基站端。利用透镜天线阵列具有基于方向的能量聚焦特性，对透镜阵列进行天线选择，并为用户分配合适的天线进行数据传输，能够在维持一定的系统性能的同时减少射频链的使用。结合机器学习在数据处理及分类方面的优势，提出了一种基于全连接神经网络（Fully Connected Neural Network，FNN）的天线选择算法。仿真结果表明，所提方案能够达到和穷搜法基本一致的系统容量，且比逻辑回归、k-近邻算法的准确率分别高了7%、2%。 |
| 关键词: 毫米波通信 天线选择 透镜天线阵列 机器学习 全连接神经网络（FNN） |
| DOI： |
|
| 基金项目:国家自然科学基金青年科学基金项目(61701433)；云南省科技厅面上项目(2018FB099)；云南大学研究生科研创新项目(20200309) |
|
| FNN antenna selection algorithm for millimeter wave lens array |
| QIAN Rongrong,LU Songqin,REN Wenping,LU Yuyuan |
| (School of Information Science and Engineering,Yunnan University,Kunming 650500,China) |
| Abstract: |
| In order to reduce the number of RF links used in the data transmission process of millimeter wave(mmWave) large-scale multiple-input multiple-output(MIMO) system and solve the problems of high hardware cost and high power consumption,lens antenna array can be deployed in the base station of mmWave MIMO system as a practical and effective solution.By using the energy focusing characteristic of lens antenna array based on direction,the antenna selection of lens array and the allocation of appropriate antenna for data transmission can maintain a certain system performance and reduce the use of RF link.According to the advantages of machine learning in data processing and classification,an antenna selection algorithm based on fully connected neural network(FNN) is proposed.Simulation results show that the proposed scheme can achieve the same system capacity as the finite search method,and compared with logistic regression and k-nearest neighbor algorithm,the accuracy is 7% and 2% higher respectively. |
| Key words: mmWave communication antenna selection lens antenna array machine learning fully connected neural network(FNN) |
