Accurate channel estimation and hybrid beamforming using Artificial Intelligence for massive MIMO 5G systems

Abstract

In a large-scale massive Multi User-Multiple Input Multiple Output (MU-MIMO) environment channel estimation and beamforming is a breathtaking task for enhancing the array gain without utilizing the many Radio Frequency (RF) chains. Somehow, several state-of-the-art works perform channel estimation and Hybrid Beamforming (HB) using Artificial Intelligence (AI) Algorithms but the massive computation intricacy and power consumption hindered the performance of the existing system. By considering the existing issues, we designed a DL-based Hybrid Beamformer for the MIMO environment with 5G communication technology (DLHB-MIMO 5G). The proposed HB design centers on three progressive processes such as accurate channel estimation, hybrid beamformer design, and hybrid beamforming. In the accurate channel estimation phase, the noise and interference-free channels are estimated using the Improved Extreme Learning Machine-Adaptive Orthogonal Matching Pursuit (IELM-AOMP) algorithm based on channel parameters and user feedback. In the HB design stage, the shortcoming of prior DL models is resolved by adopting Transfer Learning Lite Convolutional Neural Network (TL-LiteCNN) for designing a hybrid beamformer. Beforehand, we select the appropriate antenna numbers using Stackelberg Game Theory (StGT) using adequate parameters. In the hybrid beamforming stage, the problem of less Spectral Efficiency (SE) during low SNR conditions is fixed by adopting the Improved Proximal Policy Optimization (IPPO) algorithm with several beamforming parameters to generate highly resourceful hybrid beams. The realization of the proposed research is carried out using the MATLAB R2020a simulation tool and the performance of the proposed work is compared with the major state-of-the-art works in terms of useful performance metrics. The comparative results show that the proposed work beats the existing works.