We propose a practical framework for designing a physically consistent reconfigurable intelligent surface (RIS) to overcome the inefficiency of the conventional phase gradient approach. For a section of Cape Town and across three different coverage enhancement scenarios, we optimize the amplitude of the RIS reradiation modes using Sionna ray tracing and a gradient-based learning technique. We then determine the required RIS surface/sheet impedance given the desired amplitudes for the reradiation modes, design the corresponding unitcells, and validate the performance through full-wave numerical simulations using CST Microwave Studio. We further validate our approach by fabricating a RIS using the parallel plate waveguide technique and conducting experimental measurements that align with our theoretical predictions.

翻译：本文提出了一种实用的物理一致可重构智能表面设计框架，以克服传统相位梯度方法的低效问题。针对开普敦某区域及三种不同的覆盖增强场景，我们利用Sionna射线追踪和基于梯度的学习技术，优化了RIS再辐射模式的幅度。随后，根据再辐射模式的期望幅度确定了所需的RIS表面/片阻抗，设计了相应的单元结构，并通过CST Microwave Studio的全波数值仿真验证了性能。我们进一步采用平行板波导技术制作了RIS原型，并进行了实验测量，结果与理论预测相符，从而验证了所提方法的有效性。