Wideband Channel Estimation with Imperfect Hardware for Reconfigurable Intelligent Surfaces

Abstract

Channel estimation is central part of reconfigurable intelligent surface (RIS) aided communication for active and passive beamforming. The primary challenge behind channel estimate is large dimensionality stemming from not only massive number of RIS elements but also many antennas in base station. Large dimensionality leads to excessive usage of pilot tones for OFDM signals implying high overhead and decreased throughput. To alleviate the high usage of pilots for channel estimation, in this study we propose to have a low complexity transmitter at the RIS simplifed with an aggressive clipping policy and a robust channel estimator against clipping. For robust channel estimation, a generative machine learning model is adapted to exploit prior information to compensate the information loss due to clipping. The simulation results clearly indicate that the proposed estimator has quite a large resiliency for clipped transmitted signals as compared to linear channel estimators.