Machine-Learning-Based Design of RIS-Enabled Wireless Systems

Reconfigurable Intelligent Surface (RIS) is considered an energy-efficient solution for future wireless communication networks due to its fast and cost-effective configuration. Deploying RISs in the existing communication environment can effectively mitigate signal-blocking issues between the Base Station (BS) and users, providing a practical and efficient means to improve service coverage. However, it’s important to note that using a large number of reflective elements in RISs to achieve better beamforming performance can lead to larger channel dimensions, increasing the overhead and complexity of solving the related optimization problems. Machine Learning (ML) proves to be a valuable approach for maximizing the potential benefits of RIS-aided communication systems, especially as the computational complexity of operating and deploying RISs escalates with an increasing number of interactions between users and infrastructure. In this chapter, we present a comprehensive survey of relevant research in which ML techniques have been applied in RIS-aided communication networks to enhance various design and functional aspects, including channel estimation, beamforming design, resource management, security, and detection. ML can contribute to optimizing these elements, making RIS technology even more efficient and effective in wireless communication networks.