Integrated Deep Implicit CSI Feedback and Beamforming Design for FDD mmWave Massive MIMO Systems

Abstract

The heavy overhead of channel state information (CSI) feedback and the non-convexity of the hybrid beamforming (HYB) design pose great challenges for frequency-division-duplex (FDD) millimeter wave (mmWave) massive MIMO systems. In this letter, a deep HYB approach is proposed with limited and implicit CSI feedback. Specifically, we develop an autoencoder-based integrated deep implicit CSI feedback and beamforming neural network (CsiBFNet), which jointly conducts limited CSI feedback in the encoder(s) distributedly deployed on the user side and designs low-complexity beamformer in the decoder at the BS without explicit CSI reconstruction. The CsiBFNet aims to improve the achievable rate of the system rather than the CSI recovery accuracy. Simulations show that the proposed scheme outperforms the benchmark methods at the same feedback dimension and achieves 80% of the sum-rate of the existing state-of-the-art traditional HYB methods with perfect CSI even under severe compression rates (i.e., <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu =32$ </tex-math></inline-formula> ).