Multibeam Beamforming Design in RIS-aided SISO Communications Subject to Reradiation Masks – Optimization and Machine Learning Methods

Reconfigurable intelligent surfaces (RISs) are considered as one of the most promising technologies to reduce the power consumption and improve the spectral efficiency in future wireless systems. In this paper, we investigate the design of a two-user RIS-aided single-input single-output communication system and propose a beamforming design for RIS optimization, which accounts for specified reradiation masks at the design stage. First, we formulate a max-min optimization problem that aims to maximize the minimum of the normalized powers received at the two users, and we solve it by using semidefinite programming. Then, we devise an efficient implementation of the proposed algorithm based on neural networks. The obtained simulation results show that the proposed schemes are capable of shaping the desired dual-beam radiation pattern towards the directions of interest, while fulfilling the imposed reradiation masks towards other directions. The implementation based on neural networks is shown to reduce the execution time.