Reconfigurable intelligent surfaces (RISs) have become one of the key technologies in 6G wireless communications. By configuring the reflection beamforming codebooks, RIS focuses signals on target receivers. In this paper, we investigate the codebook configuration for 1-bit RIS-aided systems. We propose a novel learning-based method built upon the advanced methodology of implicit neural representations. The proposed model learns a continuous and differentiable coordinate-to-codebook representation from samplings. Our method only requires the information of the user's coordinate and avoids the assumption of channel models. Moreover, we propose an encoding-decoding strategy to reduce the dimension of codebooks, and thus improve the learning efficiency of the proposed method. Experimental results on simulation and measured data demonstrated the remarkable advantages of the proposed method.

翻译：可重构智能表面（RIS）已成为6G无线通信的关键技术之一。通过配置反射波束赋形码本，RIS可将信号聚焦至目标接收机。本文研究面向1比特RIS辅助系统的码本配置问题，提出一种基于隐式神经表征先进方法的新型学习方法。该模型可从采样中学习连续可微的坐标-码本表征。所提方法仅需用户坐标信息，无需依赖信道模型假设。此外，我们提出编码-解码策略以降低码本维度，从而提升所提方法的学习效率。仿真数据与实测数据的实验结果表明，该方法具有显著优势。