Mobile RIS-Channel Estimation with Kalman Filter and Learning

This paper presents a novel approach aimed at minimizing the Mean Squared Error (MSE) in estimating channels between mobile users and Reconfigurable Intelligent Surfaces (RIS) within a wireless communication system. Our system model comprises a mobile user, an RIS equipped with multiple elements, and a base station (BS) having several antennas. We employ a Kalman Filter (KF) to reduce the MSE of channel estimation, leveraging pilot signals transmitted by the user and the capability to adjust the phases of RIS elements. To enhance the KF's effectiveness, we integrate learning techniques that adaptively provide the KF parameters, especially beneficial in scenarios where reliable system model knowledge is not available. Our results indicate that the KF, employing a distance-dependent state space model, achieves a lower MSE compared to currently known methods. Furthermore, the integration of machine learning with the KF substantially enhances performance in instances of initial parameter inaccuracies. The paper concludes with numerical comparisons and comprehensive discussions on the improvements facilitated by our approach.