A Deep Learning Scheme for Integrated Active and Passive Beamforming in Reconfigurable Intelligent Surface Aided Wireless MISO Networks

Abstract

The fifth-generation wireless networks deployment has stared since 2020. We consider a Reconfigurable Intelligent Surface (RIS) aided multi-user multiple-input single-output (MISO) downlink system in this work. The RIS elements phase shift and beamforming matrices are optimized together to achieve maximum sum rate. The iterative optimization algorithms are adopted in most of the prior works to get suboptimal solutions, which are computationally complex. In this work, a deep learning based approach is proposed to decrease computational complexity for integrated active and passive beamforming with adequate performance. We propose an unsupervised two-stage neural network that can be trained and implemented online for real-time prediction.