Dynamic Network Optimization with Deep Learning and Deployment of Intelligent Reflecting Surfaces

The proliferation of 5G technologies has presented challenges such as signal attenuation and high energy consumption. This study proposes a novel approach to enhance wireless networks by integrating Intelligent Reflecting Surfaces (IRS) and Artificial Intelligence (AI) to optimize signal propagation and predict the Received Signal Strength Indicator (RSSI). A new framework combines K-means clustering for IRS deployment with a Temporal Convolutional Network-Gated Recurrent Unit (TCN-GRU) model for signal prediction. Using data from 15,000 Wi-Fi access points in Herefordshire, UK, the results demonstrate significant improvements in signal strength, with the proposed model achieving a Mean Squared Error (MSE) of 37% lower than the existing models, indicating higher prediction accuracy. In addition to enhancing signal quality, the framework achieves throughput increases of 21.26% with single-group IRS and 64.91% with multi-group IRS. These results validate the proposed the potential of the framework to improve performance, making it suitable for next-generation wireless networks.

I. Introduction

The rapid adoption of 5G technology has led to an explosive increase in wireless devices such as IoT devices and smart terminals, bringing innovation to modern communication environments and expanding applicability across industries [1]. High-frequency signals enable greater bandwidth but are more susceptible to attenuation due to their reduced penetration capability, particularly in denseurban environments [2]. These issues reduce the coverage range of 5G networks and significantly impact communication stability and overall network performance. Moreover, in 5G networks, power consumption of various associated devices needs attention [3].