Convolutional neural network and Kalman filter-based accurate CSI prediction for hybrid beamforming under a minimized blockage effect in millimeter-wave 5G network

Millimetre-wave (mmWave) communication is subjected to different types of blockages. Nonline of sight (NLoS) is the key problem that is caused by blockage. This study aimed ayt predicting the occurrence of self-blockage, dynamic blockage and static blockage and performs hybrid beamforming. Self-blockage is addressed by selecting and alternating the optimal access point by using an enhanced artificial flora algorithm. The fitness value estimation for optimal selection is based on the signal-to-noise ratio, distance and elevation angle. A drone base station (BS) is used to overcome dynamic blockages by inferring the drone altitude from free space path loss, transmit power and number of ground users. Static blockage is avoided by selecting the nearest ground BS based on the measurement of the SNR value. For optimum matching, bipartite matching theory is used that matches UEs to more than one ground BS. It avoids the delay of UEs. Once the user is connected and starts to receive signals, hybrid beamforming of regularised channel diagonalisation with the Convolutional Neural Network (CNN) and Kalman filter (KF) is applied to establish a communication beam. In hybrid beamforming the Channel State Information (CSI) is computed accurately by considering the frequency band, location, temperature, humidity, and weather which is performed by hybrid CNN with KF algorithm. Accurate CSI prediction ensures the formation of perfect beams from BSs and serves numerous users participating in the 5th Generation (5G) environment. The proposed system is implemented on a Matlab tool, and its performance is evaluated in terms of latency, blockage probability, number of users served, and spectral efficiency.