A comprehensive survey of artificial intelligence advances in Reconfigurable Intelligent Surfaces-assisted wireless networks

초록

Existing surveys on Reconfigurable Intelligent Surfaces (RISs) either focus on hardware fundamentals or provide preliminary discussions of Artificial Intelligence (AI) integration without systematic analysis. This survey bridges that gap by delivering a structured taxonomy of AI techniques applied to RIS-assisted wireless networks. We first examine RIS variants from passive and active designs to Beyond Diagonal RIS (BD-RIS), analyzing their architectures and operational constraints. Subsequently, we review AI methodologies including supervised learning, deep learning, reinforcement learning, deep reinforcement learning, federated learning, graph learning, transfer learning, and meta-learning, emphasizing their suitability for specific RIS challenges. The core contribution is a five-domain analysis covering channel estimation, sum rate maximization, energy efficiency, security enhancement, and performance evaluation, supported by comprehensive summary tables detailing system models, channel state information, algorithms, and optimization strategies. Following that, we synthesize lessons learned, identify open challenges, and outline future research directions, underscoring AI's transformative role in advancing RIS toward sixth generation networks.

키워드