Data-Driven Control and Optimization Models for Large-Scale Reconfigurable Metasurface-Enhanced Wireless Systems
Year: 2025
Type: article
Abstract: The newly developed reconfigurable intelligent metasurfaces (RISs) have become a game changer in the next generation wireless networks since they are able to control the electromagnetic propagation programmatically. Nevertheless, the real time optimization of large scale RIS-enhanced a wireless system is still a challenging issue because of non-convex optimization goals, high dimensionality control space and a time-varying environment. This paper proposes a data-based control and optimization system, which is a deep reinforcement learning (DRL)-based one, to reconfigurable metasurface-enhanced wireless systems at large scale. The formulation of the RIS control problem as a Markov decision process, in which the adaptation phase-shift policies are learned directly and online through observation of the environment and without explicit channel modeling or iterative convex programming. To make the proposed framework scalable and be able to operate in real-time, it will ensure that real-time and scalability are integrated through metasurface clustering and sharing parameters, allowing high-level element counts to be controlled efficiently in RIS deployments. Numerous simulations show that the suggested DRL-based method performs willingly superior to the traditional alternating optimization and fixed RIS configuration schedules in spectral-productivity, fast convergence, and ability to adjust the dynamic channel alterations. The stability tests also ensure stability in conditions where channels are estimated poorly and users move about. The findings underscore the opportunities of DRL-guided data-centric optimization as an effective and scalable method to smart operation of large-scale metasurface-based wireless infrastructures in 6G networks in the future.
Author: Robbi Rahim
Institution: Sekolah Tinggi Ilmu Manajemen Sukma
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7220334165