Joint Channel Estimation and Hybrid Beamforming via Deep-Unfolding

In this work, we propose an end-to-end deep-unfolding neural network (NN) based joint channel estimation and hybrid beamforming (JCEHB) algorithm to maximize the sum rate in massive multiple-input multiple-output (MIMO) systems. Specifically, the recursive least-squares (RLS) and stochastic successive convex approximation (SSCA) algorithms are unfolded for channel estimation and hybrid beamforming, respectively. We consider a mixed-timescale scheme, where analog beamforming matrices are designed based on the channel state information (CSI) statistics once in each frame, while the digital beamforming matrices are designed at each time slot based on the equivalent CSI matrices. Simulation results show that the proposed algorithm can significantly outperform conventional algorithms.