ReDBeam: Real-time MU-MIMO Beamforming with Limited CSI Data Samples

MU-MIMO beamforming is a key technology for 5G/NextG networks. In practice, MU-MIMO beamforming requires Channel State Information (CSI) and is prone to uncertainty. Furthermore, a beamforming solution must be derived within a millisecond (ms) to be useful for real-time (RT) 5G applications. We present ReDBeam-a RT data-driven beamforming solution for MU-MIMO using limited CSI data samples. The main contribution of ReDBeam is a parallel algorithm and an optimized GPU implementation. ReDBeam minimizes the base station (BS)'s power consumption while offering a probabilistic guarantee of users' data rates. It is purposefully designed to take advantage of the vast parallel processing capability in commercial off-the-shelf GPUs. Through extensive experiments, we show that ReDBeam can meet the 1 ms RT requirement and is orders of magnitude faster than other state-of-the-art algorithms for the same problem.