Robust URLLC Beamforming in SWIPT via Data-Augmented Model-Driven Deep Learning

This letter studies robust beamforming for ultra-reliable low-latency communications (URLLC) downlink transmission with imperfect channel state information (CSI) in a simultaneous wireless information and power transfer (SWIPT) system. By modeling CSI estimation errors via statistical distributions, we maximize the minimum rate quantile among all users through the joint optimization of the beamforming matrix and the power-splitting factor under outage probability constraints. We propose a data augmentation and model-driven (DA-MD) deep neural network (DNN) algorithm: data augmentation addresses imperfect CSI via massive error sampling, while the model-driven beamforming recovery structure reduces complexity. Our results validate the superiority of the proposed approach for URLLC.