Digital Twin-in-the-Loop Bayesian Tracking and Joint Beamforming for RIS-aided System

Abstract

Digital twins (DTs) are emerging as a promising tool for closed-loop control of next-generation wireless networks, but their integration with reconfigurable intelligent surfaces (RISs) remains challenging. We propose a multi-rate DT-in-the-loop framework for multiuser tracking and joint beamforming in RIS-assisted downlink systems. A DT-assisted Unscented Kalman Filter (DT-UKF) uses ray-traced channel features as propagation-aware measurement predictions, while explicitly accounting for both pilot-estimation noise and DT mismatch through a regularized covariance update. We further construct a differentiable surrogate of the ray-traced RIS-assisted channel to enable joint optimization of RIS phases and BS precoders. Numerical results show that the proposed DT-UKF improves tracking accuracy over analytical EKF and baseline DT-UKF methods, while joint RIS/precoder optimization improves spectral efficiency over fixed RIS and gradient-based baselines. Results further demonstrate the effects of DT mismatch and measurement-update frequency on closed-loop tracking performance.