Improving Signal Coverage in Millimeter-Wave Massive MIMO via Efficient Predefined-Time Adaptive Neural Network–Based Beam Training

This paper proposes an advanced deep learning framework for efficient beam training in millimeter wave (mmWave) massive multiple-input multiple-output (MIMO) systems. To overcome the limitations of conventional beam training approaches such as high overhead, slow adaptation to dynamic environments, and poor scalability, an Improving Signal Coverage in Millimeter Wave Massive MIMO via Efficient Predefined Time Adaptive Neural Network based Beam Training (ISC-MMIMO-EPTANN-BT) model is proposed. The proposed model used deep neural network (DNN) to learn complicated nonlinearities in channel power leakage (CPL) and used an efficient predefined time adaptive neural network (EPTANN) to provide real-time responsiveness and temporal synchronism in beam training. The parameters of the model are also optimized using fire hawk optimization algorithm (FHOA) to get better convergence speed and signal coverage. The proposed technique is executed in MATLAB. The proposed approach attains better performance under successful rate by significantly less beam training overhead and also increases signal coverage based on simulation results. The proposed ISC-MMIMO-EPTANN-BT method attains 26.15%, 21.08%, and 33.75% higher successful rates and 16.32%, 28.94%, and 20.24% lower normalized mean square error compared with existing methods such as deep learning for beam training in millimeter wave massive MIMO schemes (BT-MMIMO-DNN), deep learning for combined feedback and channel prediction in large-scale MIMO systems (CNN-JCS-MMIMO), and triple-refined hybrid-field beam training in mmWave extremely large-scale MIMO (TR-FBT-MIMO), respectively. The ISC-MMIMO-EPTANN-BT technique reduced beam training overhead, enhanced signal coverage, and identified a promising candidate for successful beam training in mmWave massive MIMO schemes.