Abstract

Reconfigurable intelligent surfaces (RISs) have been introduced as a remedy for mitigating blockages in millimeter wave (mmWave) and terahertz (THz) communications networks. However, perfect or nearly perfect channel state information (CSI) is fundamental in order to achieve their full potential. Tra-ditionally, an RIS is fully passive without any baseband processing capabilities, which poses great challenges for CSI acquisition. Thus, we focus on the hybrid RIS architecture, where a small portion of RIS elements are active and able to processing the received pilot signals for estimating the corresponding channel. The channel estimation (CE) is done by resorting to off-the-grid compressive sensing technique, i.e., atomic norm minimization, for extracting channel parameters through two stages. Simulation results show that the proposed scheme outperforms the passive RIS CE under the same training overhead.