Research On Beamforming Based On Deep Learning Millimeter Wave Channel Estimation

As people's quality of life improves,mobile data transmission services are increasingly in demand.The use of millimeter-wave multi-antenna transmission technology in 5G will substantially increase the rate and reduce the delay.In this thesis,beamforming is investigated based on channel estimation for millimeter-wave systems.This thesis designs a Deep Learning Coordinated Beamforming(DLCBF)solution based on deep learning to address the problem of frequent switching of base stations during communication between users and base stations,which will lead to large delay overhead in channel transmission between users and base stations.The solution estimates the channel response by deep learning,coordinates multiple base stations to serve a mobile user at the same time,and trains the learning model by receiving the user uplink guide sequence and pre-coded codebook.The model constructs the link between uplink guide channel response characteristics and downlink beam vector by deep learning model,and when the user's guide frequency is received again,the base station predicts the best downlink beam vector based on the training model,which reduces the huge training overhead required to select the best beamforming vector in millimeter-wave large-scale multi-antenna systems,thus effectively reducing the latency.Simulation results show that the proposed scheme has higher spectral efficiency compared with the conventional millimeter wave beamforming scheme.The conventional user localization technique has complicated steps and the system in which the user's location information is not estimated with high accuracy.In this thesis,we address this problem by designing a beam selection-based localization algorithm in conjunction with DLCBF.The scheme considers the channel parameters such as line-of-sight,non-line-of-sight and multipath to estimate the channel information,and analyzes the connection between the channel response information and the location information by selecting the optimal beam between the user and the base station to estimate the user location information.Simulation analysis shows that the scheme has improved in localization accuracy and also has better performance improvement in communication transmission delay,compared with the conventional scheme,this scheme has better performance improvement in transmission signal-to-noise ratio,time delay and localization accuracy.