Blind beamforming for reconfigurable intelligent surfaces via quadratic modeling

Recently, blind beamforming has become a promising method to configure reconfigurable intelligent surfaces (RISs) without channel state information (CSI) or geometric models, making it readily usable on commodity hardware.In this paper, we introduce a blind beamforming algorithm, so-called Blind Optimal RIS Beamforming with Sensing (BORN), that relies solely on received signal strength (RSS).Unlike prior approaches built on majority-voting heuristics, BORN leverages the inherent quadratic structure of the received signal-to-noise ratio (SNR).The method has two stages: a sensing stage, which estimates a quadratic model from RSS measurements, followed by an optimization stage, which computes the RIS configuration using the learned model.Extensive simulations and real-world experiments show that BORN achieves near-optimal performance and outperforms state-of-the-art blind beamforming schemes.