Performance evaluation of DHRR-RIS based HP design using machine learning algorithms

Abstract

Reconfigurable Intelligent Surfaces (RIS) have emerged as a promising technology for improving the reliability of massive MIMO communication networks. However, conventional RIS suffer from poor Spectral Efficiency (SE) and high energy consumption, leading to complex Hybrid Precoding (HP) designs. To address these issues, we propose a new low-complexity HP model, named Dynamic Hybrid Relay Reflecting RIS based Hybrid Precoding (DHRR-RIS-HP). Our approach combines active and passive elements to cancel out the downsides of both conventional designs. We first design a DHRR-RIS and optimize the pilot and Channel State Information (CSI) estimation using an adaptive threshold method and Adaptive Back Propagation Neural Network (ABPNN) algorithm, respectively, to reduce the Bit Error Rate (BER) and energy consumption. To optimize the data stream, we cluster them into private and public streams using Enhanced Fuzzy C-Means (EFCM) algorithm, and schedule them based on priority and emergency level. To maximize the sum rate and SE, we perform digital precoder optimization at the Base Station (BS) side using Deep Deterministic Policy Gradient (DDPG) algorithm and analog precoder optimization at the DHRR-RIS using Fire Hawk Optimization (FHO) algorithm. We implement our proposed work using MATLAB R2020a and compare it with existing works using several validation metrics. Our results show that our proposed work outperforms existing works in terms of SE, Weighted Sum Rate (WSR), and BER.

Keywords

• reconfigurable intelligent surfaces (ris)
• dynamic hybrid relay reflecting (dhrr)-ris
• multi user multiple input multiple output (mu-mimo)
• hybrid precoder
• machine learning and deep learning algorithms
• channel state estimation