Reconfigurable Intelligent Surface Assisted Interference Suppression With Impedance Regulated Deep Neural Network (IR-DNN)

Radio frequency interference (RFI) poses a substantial risk to microwave radiometry. This paper addresses a machine learning-driven reconfigurable intelligent surface (RIS) absorber for RFI cancellation. The proposed wideband absorber regulates the scattering parameters through a universal surface impedance management scheme, considering all possible physical and electromagnetic attributes. It is based on a low-cost Impedance Regulated Deep Neural Network (IR-DNN), yielding an impressive mean square error of only 0.004. The presented simulation results indicate that in the X-band, the IR-DNN-based absorber can reflect more than 80% of signals received through the main beam while blocking at least 80% of the interfering signals arriving from other directions. Notably, the proposed IR-DNN model can inversely predict design variables with higher than 97% accuracy.