Active RIS-Assisted Sup-Degree of Freedom Interference Suppression for a Large Antenna Array: A Deep-Learning Approach With Location Awareness

Radio frequency interference (RFI) generated by satellites orbiting the Earth has become a significant threat to astronomical observations. Active reconfigurable intelligent surfaces (RISs) have the capacity to intelligently control the wireless propagation environment and compensate for double-path fading attenuation. Thus, this article exploits an active RIS to mitigate RFI by forming an additional path to reflect RFI. Specifically, the model provides a new solution to sup-degree of freedom (DoF) interference suppression, which means that the amount of interferences surpasses the array DoF. We intend to simultaneously design the receive beamforming of large antenna array and reflection coefficients of active RIS to maximize the received signal-to-interference-plus-noise ratio (SINR). Moreover, the system model can be correlated with a non-Euclidean graph structure due to its graph-like nature. To achieve this objective, a deep-learning (DL) approach, named the location awareness graph ordering attention (LAGOAT) network, is presented to map the locations of RFI and array into the receive beamforming and reflection coefficients. The simulations not only demonstrate that our proposed system can improve significantly the received SINR but also show that proposed LAGOAT network can achieve superior performance compared with other DL approaches, namely, graph neural network (GNN), graph attention network (GAT), and the network without GOAT.