Statistical-CSI-Based RIS User Clustering: From Interference-Free TDMA to Simultaneous Multiuser Transmission

Reconfigurable intelligent surfaces (RISs) can improve wireless coverage and spectral efficiency, but practical deployments are limited by channel estimation overhead and by the number of RIS reconfigurations within a transmission frame. Statistical channel state information (S-CSI) is a low-overhead alternative for RIS phase-shift design and user clustering. Previous work showed that S-CSI-based minimum-distance (minD) clustering performs close to perfect CSI (P-CSI)-based methods in interference-free time division multiple access (TDMA) operation. This paper extends that framework to simultaneous multiuser transmission, where users assigned to the same cluster are served in the same time slot under a common RIS configuration. The analysis is reformulated from signal-to-noise ratio (SNR) to signal-to-interference-plus-noise ratio (SINR), accounting for multiuser interference and spatial correlation. Numerical results show that, for the considered setup, the prelog gain obtained by reducing the number of time slots does not compensate for the interference penalty. Thus, the results identify an operational boundary for the S-CSI/minD framework: it is effective for sharing RIS configurations among temporally multiplexed users, but serving one user per slot remains the most effective strategy when multiuser interference is significant.