Channel Estimation and Prediction in Wireless Communications Assisted by Semi-Passive RIS

When the line-of-sight between the base station and mobile users is unavailable, reconfigurable intelligent surfaces (RIS) can be exploited to ensure connectivity and improve data transmission performance. The objective of this paper is to estimate and predict timevarying user-RIS channels with low pilot overhead using a small number of sparsely distributed active RIS elements. Structured covariance matrix interpolation is performed to fully utilize the array aperture from the sparse semi-passive RIS. Channel variation over time due to user movement and environmental factors can make it difficult to utilize all time slots for channel estimation and data transmission. To address this challenge, we propose a model based on long short-term memory (LSTM) networks for channel estimation and prediction to reduce the required training pilot signals and increase the transmission data rate using parallel computation. Simulation results verify the capability of the proposed approach to enhance data transmission in wireless networks and demonstrate its effectiveness compared to other machine learning models.