User Clustering and Phase-Shift Design for RIS-Aided Systems Based on Statistical CSI

Reconfigurable Intelligent Surfaces (RISs) are a key technology to improve cell coverage and energy efficiency in the upcoming sixth-generation (6G) systems. Most papers in the field assume the existence of a wired link between the base station (BS) and RIS, and impose no constraint on the information exchange between them, as well as on the frequency with which the RIS can tune the phase-shifts of its elements. On the other hand, it has been shown recently that a more appealing scenario is to consider a limited number of RIS reconfigurations per time period, due to the imperfection of the BS-RIS communication link and/or energy consumption entailed by each reconfiguration. A time division multiple access is thus an interesting approach to serve the users, making it necessary to cluster them in groups to be served under the same RIS configuration. In this paper, we propose a user clustering and phase-shift design method based solely on the users’ statistical channel state information (CSI). Compared with the approaches based on instantaneous CSI available in the literature, our proposed method has several advantages: it requires much lower training overhead, alleviates the need for active RIS elements, the obtained solution is valid for a much larger time basis, and for the entire frequency bandwidth available. Furthermore, extensive numerical results show that the performance loss regarding the approaches based on instantaneous CSI is very little for typical scenarios expected for 6G systems.