Reconfigurable Intelligent Surface Assisted Millimeter-Wave MISO Systems:Channel Estimation

Millimeter-wave(mm Wave)has propagation loss,so it will be very challenging to achieve full coverage in specific application scenarios.With the research of the next generation mobile communication(B5G/6G)technology,the Reconfigurable Intelligent Surface(RIS),as a relay method that can reflect incident signals without loss,can effectively improve the coverage of the base station and the communication system capacity.The basis of reliable communication requires an accurate and effective channel estimation method to obtain large-scale Channel State Information(CSI),but the lack of signal processing functions of RIS brings challenges to channel estimation,and there are large training overhead,complex calculations and difficult to adapt to the complex time-varying environment of mobile communications of traditional channel estimation algorithms.In this regard,this article has carried out the following analysis and research:The channel estimation method is studied in the RIS assisted millimeter-wave multiple input single output(MISO)system.Among them,RIS is used to assist the data transmission between the user and the base station in the mm Wave MISO system.The time division duplex(TDD)protocol is used to estimate the uplink,and the method based on deep learning is discussed.Firstly,through MATLAB simulation,the Saleh-Valenzudel model is used to construct the Base Station-RIS,RIS-User channel model,and the cascade channel is obtained through deformation,and then the received signal is obtained.After analyzing the performance of traditional channel estimation algorithms(based on the least square method,based on the minimum mean square error method and compressed sensing method)in simulation,the channel estimation based on the deep learning method is explored.Secondly,in the channel estimation research based on deep learning,the neural network framework constructed by deep learning is analyzed,and a suitable network model is found to train the cascaded channel information.A Deep Neural Network(DNN)+ Lambda structure network model is proposed,including an input layer,a Flatten layer,5 fully connected layers,and two Lambda layers.The received signal is used as a data set and input into the network for training.Specifically,the simulation is performed through PYTHON,the Tensor Flow framework in deep learning is used,and the parallel calculation of the GPU and the CUDA calculation library is used to complete the network construction and simulation.Finally,the simulation results show that compared with traditional channel estimation algorithms,the method proposed in this paper has the following advantages:(1)Through the selection of a reasonable network structure,the deep learning based RIS assisted mm Wave MISO system used in this article improves the accuracy of channel estimation and can improve the transmission performance of the communication system;(2)By simulating the pilot frequency overhead,the method based on deep learning is robust to the changes of the pilot frequency and can reduce the pilot frequency overhead;(3)Since the data takes into account the complex environment of communication,such as effective multipath number and other factors,the algorithm has strong environmental adaptability and can track channel changes;(4)The training of the network is carried out in an offline state,and real-time channel estimation is realized online,thereby reducing computational complexity,saving estimation time,and improving the time efficiency of estimation.