GNN-Based Channel Estimation for Intelligent Reflecting Surface Aided Multiuser Systems Relying on User Locations

For intelligent reflecting surface (IRS)-assisted wireless systems, it is practically challenging to acquire accurate channel state information due to the massive number of IRS elements without the capacity of processing signals. In this letter, we propose a graph neural network (GNN) based channel estimation method. Specifically, the proposed GNN is used to learn the mapping from the received signals to cascaded channels. To further improve the channel estimation performance, the location information of the users is incorporated into the neural network. Then, we develop a GNN based channel estimation method by incorporating the graph topology of the IRS-aided system into the GNN architecture. Numerical results show that the proposed GNN maintains good performance and outperforms existing deep learning based schemes under low pilot overhead.