The pursuit of energy-efficient solutions in the context of Reconfigurable Intelligent Surface (RIS)-assisted wireless networks has become imperative and transformative. This paper investigates the integration of RIS into an orthogonal frequency division multiple access (OFDMA) framework for multiuser downlink communication systems. We address the challenge of jointly optimizing RIS reflection coefficients alongside OFDMA frequency and power allocations, with the aim of maximizing energy efficiency. This optimization is subject to specific quality-of-service (QoS) requirements for each user equipment (UE) and a constraint on transmission power and RIS phase shift matrix. To address this complex optimization problem, we propose a practical and low-complexity approach that derives a computationally efficient and numerically tractable lower bound on energy efficiency. Our approach strikes a balance between performance enhancement and complexity. The numerical results highlight the effectiveness of our approach, showing a substantial increase in energy efficiency compared to scenarios with random RIS integration and without RIS.

Abstract— Even though, reconfigurable intelligent surfaces (RISs) are adopted in various scenarios to enable the implementation of a smart radio environment, there are still challenging issues for its real-time operation due to the need for a costly full dimensional channel estimation with offline exhaustive search or online exhaustive beamtraining. The application of the deep learning (DL) tools is favored to enable feasible solutions. In this work, we propose two low training overhead and energy efficient adversarial bandit-based schemes with outstanding performance gains compared to reference DL based reflection beamforming methods. The resulting deep learning models are also discussed using state of-the art model quality prediction trends.