Site-Specific Design Optimization of Reconfigurable Intelligent Surface Codebooks

Multipath propagation is ubiquitous in wireless communications and plays a critical role in channels enabled by reconfigurable intelligent surfaces (RISs). To date, most methods of RIS codebook design, i.e., the selection of each unit cell state, neglect multipath propagation and rely on idealized models that consider only the Tx–RIS–Rx link. In this letter, we introduce a numerical, site-specific (SS) design paradigm in which RIS codebooks are tailored to the realistic multipath conditions of the deployment site. The framework combines ray tracing with a convolutional neural network to map RIS codebooks directly to signal strengths at designated Rxs, accounting for multipath propagation and practical factors, such as cell-to-cell coupling. Using genetic algorithm optimization, we show that SS codebooks consistently achieve higher received signal strengths than conventional, site-agnostic designs. In addition, our developed framework supports comprehensive predeployment assessment of link budgets and design parameters, such as the number of unit cells.