Sum-Rate Maximization for Double RISs Assisted Downlink RSMA System in Overloaded Networks

As a critical evaluation metric for the sixth generation (6G) communication systems, the sum-rate has attracted increasing attention in recent research. In this letter, we consider the double reconfigurable intelligent surfaces (RISs) assisted rate-splitting multiple access (RSMA) downlink multiple-input single-output (MISO) system under overloaded network scenarios. We aim to maximize the system sum-rate by optimizing the common rate allocation, RIS phase shifts, and base station (BS) beamforming. Specifically, a deep learning algorithm based on graph neural network (GNN) is employed to solve this problem. Through simulation experiments, the proposed novel GNN model is proven to be more lightweight and well-adapted to overloaded network scenarios. Compared to the successive convex approximation (SCA) and traditional deep learning (DL) algorithm, the GNN based RSMA demonstrates higher computational efficiency, robustness against imperfect channel state information (CSI), and effective generalization capabilities across different user sets.