基于无监督DNN和Sub-6GHz的毫米波功率控制算法  

Millimeter wave power control algorithm based on unsupervised DNN and Sub-6GHz

作　　者：孙长印 毛亚宁[1] 江帆 王军选 SUN Changyin;MAO Yaning;JIANG Fan;WANG Junxuan(School of Communications and Information Engineering,Xi’an University of Posts and Telecommunications,Xi’an 710121,China;Key Laboratory of Information Communication Network and Security,Xi’an University of Posts and Telecommunications,Xi’an,710121,China)

机构地区：[1]西安邮电大学通信与信息工程学院,陕西西安710121 [2]西安邮电大学信息通信网络与安全重点实验室,陕西西安710121

出　　处：《西安邮电大学学报》2023年第2期29-38,共10页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金项目(61801382,61871321,62071377);陕西省自然科学基金项目(2019JZ-06);陕西省重点产业链项目(2020ZDLGY02-06,2019ZDLGY07-06)。

摘　　要：针对毫米波(millimeter Wave,mmW)系统中有监督深度神经网络(Deep Neural Networks,DNN)功率控制算法的性能受限及mmW信道信息测量质量不佳的问题,提出一种基于无监督DNN(Unsupervised DNN,UDNN)和Sub-6GHz频段的mmW功率控制算法。将最大化系统和速率设置为UDNN的损失函数,改进有监督学习的性能上限并利用集成学习进一步提升算法性能,最终实现输入Sub-6GHz频段信道信息即可得到mmW频段最优功率分配。为了验证所提算法的可行性,将所提算法与加权最小均方误差算法、最大功率控制算法、随机功率控制算法和二进制穷举搜索算法的系统和速率进行对比。验证结果表明,所提算法的和速率性能分别是其他4种算法的1.121倍、2.322倍、1.843倍和1.022倍,且采用Sub-6GHz预测mmW功率控制可全程逼近使用mmW信道预测功率控制的性能。For the performance limitation of the power control algorithm under the supervised Deep Neural Networks(DNN),and the poor measurement quality of the millimeter wave(mmW)system,a predictive mmW power control algorithm is proposed based on the unsupervised DNN(UDNN)and the Sub-6GHz frequency band.In order to break through the performance constrains of the supervised learning,and to further improve the algorithm performance of the ensemble learning,the maximum sum rate is selected as the UDNN loss function,to realize the mmW optimal power allocation according to the input Sub-6GHz channel information.To verify the effectiveness of the proposed algorithm,its system and rate are compared with the weighted minimum mean square error algorithm,the maximum power control algorithm,the random power control algorithm,and the binary exhaustive search algorithm.The verification results show that the average system sum rate under the proposed algorithm could be increased by 1.121,2.322,1.843 and 1.022 times separately compared with the other four algorithms.It is also shown that the predictive mmW power control by the Sub-6GHz can approximate the performance of the mmW channel predictive power control in the whole procedure.

关 键 词：毫米波通信 无监督DNN 深度神经网络 功率分配 系统和速率 集成学习 

分 类 号：TN928[电子电信—通信与信息系统]