Deep Learning-Based CSI Feedback for Beamforming in Single- and Multi-Cell Massive MIMO Systems

The potentials of massive multiple-input multiple-output (MIMO) are all based on the available instantaneous channel state information (CSI) at base station (BS). Therefore, user in frequency-division duplexing (FDD) systems has to keep feeding back CSI BS, thereby occupying large uplink transmission resources. Recently, deep learning (DL) achieved great success feedback. However, existing works just focus improving feedback accuracy and ignore effects following modules, e.g., beamforming (BF). In this paper, we propose a DL-based framework for BF design, called CsiFBnet. key idea CsiFBnet is maximize performance gain rather than accuracy. We apply it two representative scenarios: single- multi-cell systems. CsiFBnet-s single-cell system autoencoder architecture, where encoder compresses decoder BS generates vector. CsiFBnet-m feed kinds CSI: desired interfering CSI. entire neural networks trained by an unsupervised strategy. Simulation results show improvement complexity reduction compared with conventional methods.