Bayesian Near-Field Multiuser Tracking via Non-stationary Reconfigurable Intelligent Surface

Reconfigurable intelligent surface (RIS) has attracted enormous interest in future 6G systems. Due to its low cost, it can be made sufficiently large in scale for localization enhancement. When deploying a large RIS, users are often operating in the near field of the RIS and the channels exhibit spatial non-stationarity (SNS), posing new challenges for near-field user localization and tracking. In this paper, we develop a bi-level array partitioning strategy to flexibly characterize the spatial non-stationarity effect in user localization and tracking problem. Considering user mobility and the group sparsity of the channels between users and the RIS, we introduce Markov processes and establish a probabilistic transition model to characterize the user tracking process. Subsequently, we propose a novel Bayesian user tracking algorithm, termed bi-level array partitioning for Bayesian near-field tracking (BLAP-BNT), which jointly estimates users’ positions and their visible region (VR). The simulation results show the effectiveness of our algorithm.