A Prediction-Based Traffic Scheduling Framework for Multimedia Services in Dynamic Networks
Zhe Zhang, Xin Wei, Zhicai Zhang · IEEE Communications Letters · 2024
How to efficiently transmit multimedia content over dynamic networks is extremely challenging in current networks. Motivated by the advancements in using Artificial Intelligence (AI) models to predict video frames, we propose a novel traffic scheduling framework that can predict upcoming video frames at the edge server rather than transmitting them from the source node to save bandwidth. The proposed framework consists of a video prediction component (predicts upcoming video frames based on previous video frames) and a Deep Reinforcement Learning (DRL)-based triggering component (determines if the video prediction component needs to be triggered based on the current network conditions and user demands). Simulation results show the effectiveness of the proposed framework.