Digital Twin-Aided Learning to Enable Robust Beamforming: Limited Feedback Meets Deep Generative Models

In massive multiple-input multiple-output (MIMO) systems, robust beamforming is a key technology that alleviates multi-user interference under channel estimation errors. Recently, stochastic weighted minimum mean square error (WMMSE) algorithm was proposed to achieve robust risk-neutral beamforming, assuming the knowledge of channel probability distribution. However, this information is difficult to acquire especially in the fifth generation new radio (5G NR) frequency division duplex (FDD) systems with limited feedback. To tackle this challenge, different from previous works relying on Gaussian assumptions, we propose a digital twin-aided approach that directly learns the channel generative process from the limited feedback information via deep generative models. Numerical results are presented to show the excellent performance of the proposed algorithm in terms of both channel generation and robust beamforming.