RIS-Aided Sensing System With Localization Function: Fundamental and Practical Design

Sensing-assisted technology will be an essential instrument in sixth-generation (6G) wireless networks to effectively enhance the reliability of communication. In this article, not only staying at the theoretical level, a low-cost reconfigurable intelligent surface (RIS)-aided sensing system prototype is built, and an atomic norm minimization-direction of arrival estimation network (ANM-DOAnet) method is proposed to realize the effective fusion of sensing and localization. First, instead of the high hardware cost phased array, a cost-effective and low-complexity structure of RIS is designed, and a sensing system with the aid of the designed RIS is built in non-line-of-sight (non-LoS) scenario. Then, a practical signal model received by antennas is formulated by considering the phase shifts of RIS. Different from the traditional array for direction of arrival (DOA) estimation, the existing atomic norm minimization (ANM) methods cannot estimate the angle directly due to the phase reflection errors in the RIS. Thus, an ANM-DOAnet method is proposed by using the ANM to obtain the spatial spectra and the DOAnet to estimate the DOA via classification. To reduce the workloads, a generative adversarial network (GAN) is adopted to generate a large dataset. The experimental results show that the proposed ANM-DOAnet can localize the source effectively in the RIS-aided sensing system.