Proximal Policy Optimization for Cross-Layer Joint Design of MISO Beamforming and RIS Phase Configuration

Video traffic has become the dominant form of data in modern wireless networks, making high-quality transmission and efficient resource allocation increasingly critical. Reconfigurable Intelligent Surfaces (RIS) are considered a key technology in development of upcoming wireless systems. such as B5G and 6G, offering new opportunities to enhance wireless communication performance. Although prior works have applied Deep Reinforcement Learning (DRL) to optimize power control and resource allocation in RIS-assisted systems, most rely on structurally complex algorithms whose training stability can be sensitive to hyperparameters, such as Soft Actor-Critic (SAC), and the consideration of cross-layer design remains largely underexplored. This paper focuses on replacing SAC with the more stable and lightweight Proximal Policy Optimization (PPO), while also incorporating application-layer video parameters into the state representation to support learning and improve overall system performance, Therefore, we propose a cross-layer resource allocation method based on the PPO algorithm for RIS-assisted communication systems. As demonstrated in Section IV, the proposed scheme achieves a Peak Signal-to-Noise Ratio (PSNR) improvement of approximately 0.5 to 1 dB under various scenarios, while significantly reducing the total number of model parameters by approximately 86% to 89% compared to the baseline scheme. Furthermore, the proposed scheme is extended to the Mobile Edge Computing (MEC) offloading framework. As demonstrated in Section V-E, the proposed method achieves an average PSNR improvement of 1 to 3 dB compared to the baseline scheme.