Deep Reinforcement Learining for Video Quality in RIS-Enabled Mobile Edge Computing

Reconfigurable Intelligent Surface (RIS) is a highly promising technology that can enhance the performance of Mobile Edge Computing (MEC) networks. Various Deep Reinforcement Learning (DRL)-based methods have been proposed for power and phase-shift matrix allocation in RIS-assisted communication systems, applicable to environments with hardware defects and imperfect Channel State Information (CSI). However, existing methods primarily focus on resource allocation at the physical layer and fail to fully address cross-layer issues related to video transmission. Therefore, this paper proposes a cross-layer resource allocation method for RIS-assisted communication systems in video transmission, changing the reward function from maximize computation rate to jointly optimize computation rate and average PSNR Ratio (PSNR)