Intelligent Reflecting Surface-Assisted Interference Mitigation With Deep Reinforcement Learning for Radio Astronomy
Junhui Peng, Hailin Cao, Zahid Ali, Xiaodong Wu, J. H. Fan
Abstract
Radio frequency interference (RFI) is a significant threat to astronomical observations. Thus, this letter exploits the intelligent reflecting surfaces (IRSs) to mitigate RFI by adjusting the reflection coefficients of IRSs. Aiming to synthesize a spatial quiet zone in the control area of a radio telescope, an optimization problem for joint multiple reflected beamforming at IRSs is formulated. As the interference behavior and direction are dynamic, an IRS relative position encoding attention deep deterministic policy gradient (RPEA-DDPG) learning algorithm is proposed to jointly optimize the reflected beamforming of IRSs without the knowledge of the interference model. Simulation results demonstrate that the proposed technique can effectively establish an open electromagnetic field quiet zone to prevent RFI from entering the receiver.