Optimizing low-power, anti-interference routing with IRS through spatiotemporal reinforcement and federated learning

Abstract This paper introduces the SAC-IRS algorithm, a novel approach that combines Intelligent Reflecting Surfaces (IRS) and Reinforcement Learning (RL) to optimize network performance, specifically targeting packet delivery rate and energy efficiency. The proposed algorithm dynamically adjusts IR...

Full description

Saved in:
Bibliographic Details
Main Authors: Chengfu Fan, Shoujiu Xiong, Fei Zhou
Format: Article
Language:English
Published: Springer 2025-07-01
Series:Discover Applied Sciences
Subjects:
Online Access:https://doi.org/10.1007/s42452-025-07552-7
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract This paper introduces the SAC-IRS algorithm, a novel approach that combines Intelligent Reflecting Surfaces (IRS) and Reinforcement Learning (RL) to optimize network performance, specifically targeting packet delivery rate and energy efficiency. The proposed algorithm dynamically adjusts IRS phase shifts at the physical layer, leveraging cross-layer optimization to coordinate the physical, link, and network layers. Unlike traditional approaches, SAC-IRS incorporates real-time channel state information (CSI) feedback, enabling the system to adapt to varying network conditions and improve signal strength while reducing energy consumption. Experimental results demonstrate that SAC-IRS achieves a 90% packet delivery rate and lowers energy consumption to 0.25 mJ/bit, outperforming existing methods such as AODV, DeepIR, and IRS-Q-Routing. These findings highlight the algorithm’s ability to balance signal optimization with energy efficiency. Additionally, the paper discusses the challenges of deploying IRS hardware in real-world settings, addressing issues such as scalability in large networks, handling node mobility, and adapting to dynamic topology changes. The proposed protocol’s real-world applicability is considered, and future directions are outlined for overcoming scalability and adaptation challenges in practical deployments.
ISSN:3004-9261