GNN Based Beamforming Design for RIS Assisted Near-Field ISAC Systems

In this paper, we investigate a very large reconfigurable intelligent surface (RIS) assisted near-field integrated sensing and communication (ISAC) system, where a multi-antenna base station (BS) simultaneously sends unicast data to multiple single-antenna communication users (CUs) and senses multiple targets. The BS, CUs and targets are all assumed to be located in the near-field region of the RIS. We aim to maximize the achievable sum rate of all CUs, subject to the sensing beampattern gain constraint for each target and the transmit power constraint for the BS. Different from the existing works, we adopt a graph neural network (GNN) based approach to optimize the BS’s transmit beamforming vectors and the RIS’s reflecting coefficients via learning the useful channel state information (CSI) in the ISAC system. Numerical results show that our proposed design scheme achieves a better performance than the existing benchmarks.