Practical RIS-Assisted RF Energy Harvesting via Hierarchical Beam Search and Tracking

Reconfigurable Intelligent Surfaces (RIS) offer a promising way to mitigate the severe path loss of far-field RF energy harvesting (RFEH). However, many existing RIS beamforming techniques rely on channel state information (CSI) or RF-domain feedback, which are unavailable on low-power harvesters that expose only noisy capacitor-voltage readings. This paper presents a practical RIS-assisted RFEH system that operates with commodity transmitters using only harvester-side DC feedback. We develop a geometry-guided hierarchical control framework in which the first stage performs a coarse sector search around the nominal direction inferred from deployment geometry, and the second stage performs local bandit-based tracking over a finer-grained codebook. We implement the system using a custom 16$\times$16 2-bit RIS at 2.4~GHz, an Arduino controller, and a Powercast harvester. Experiments with a commodity Wi-Fi router and a USRP-based OFDM transmitter show that our design increases average harvested DC power by up to 8.13$\times$ and extends the maximum regulated operating distance by 1.8$\times$ over the no-RIS baseline, while outperforming Geometry-only and three competitive CSI-free blind baselines.