Tensor Based Channel Estimation for Multi-RIS Assisted OFDM System

We consider the problem of channel estimation for multi-reconfigurable intelligence durface (RIS) assisted orthogo-nal frequency division multiplexing (OFDM) systems. To detect the higher-rank signals reflected by multiple RISs, we divide the total RISs into several groups and propose a training protocol for filtering the signal components belonging to different groups. By concatenating the training signals on different subcarriers, we construct a low-rank third-order tensor model and develop a canonical polyadic decomposition (CPD) method to estimate the channel parameters. Simulation results indicate that the proposed tensor-based method effectively improves the estimation accuracy. The RIS grouping strategy trades off the training overhead and algorithm performance.