Deep optimized hybrid beamforming intelligent reflecting surface assisted UM-MIMO THz communication for 6G broad band connectivity

Abstract

For 6G communications, the Ultra Massive Multiple Input Multiple Output (UM-MIMO) systems with Intelligent Reflecting Surface (IRS) assistance are capable since they can efficiently get beyond the limitations of restricted blockage and coverage. However, in the far field, a robust THz channel sparsity is unfavorable to spatial multiplexing, whereas excessive UM-MIMO and IRS dimensions extend the near field region. To address these issues, a hybrid beamforming IRS assisted UM-MIMO THz system with Deep Siamese Capsule Network is designed with the cascaded channel. The near and far field codebook-based beamforming is developed to model the proposed communication channel. The channel estimation is done based on the deep siamese capsule adaptive beluga whale neural network. The simulation results of the bit error rate, Normalized Mean Square Error (NMSE), spectral efficiency, sum rate, data rate, normalized channel gain, beamforming gain, and array gain loss shows that the proposed system achieves reliable performances compared with existing techniques. The suggested approach also demonstrates the outstanding adaptability to various network configurations and good scalability. The method provides a better channel estimation accuracy and less complexity which shows an NMSE of − 11.2 dB at an SNR of 10 dB.