Codebook-Based Channel Estimation and Beamforming for RISs With Sensing in a Small Subset of Unit Cells

This paper discusses reconfigurable intelligent surfaces (RISs) featuring a limited number of sensing unit cells (UCs) alongside the remaining reflecting UCs. Such configurations provide multiple notable advantages, particularly in facilitating the rapid deployment of RISs into existing wireless systems. We consider a codebook-based approach for both channel estimation and beamforming in these RIS systems. A key challenge arises from performing channel estimation with only a few sensing UCs, while beamforming requires collaboration across all UCs. To address this, we develop a novel enhancement of the codebook-based approach, which leverages the least squares (LS) method and builds upon techniques recently introduced for linear antenna arrays, extending and optimizing them for planar arrays of sensing UCs. Numerical results compare various codebook-based methods, including both basic and proposed LS-enhanced approaches, applied to a planar RIS UC array. The evaluation focuses on channel estimation accuracy and achievable rate, considering different ratios of the number of sensing UCs to the total number of UCs. Furthermore, we assess the impact of RIS UC operation with and without amplification during reflection. Additionally, we compare the performance of these methods in both single-path line-of-sight and multi-path fading channels. Finally, we provide a quantitative comparison of training overhead and computational complexity, demonstrating the efficiency of the proposed LS-enhanced method.