A Markov decision model for source video rate allocation and scheduling policies in mobile networks

We consider a scenario with a mobile user who either upstreams a video clip to a portal in the Internet, where video consumers can watch it in real-time, or, conversely, downstreams a multi-view video from the Internet to be played on his/her mobile device. Thus, we design a delivery framework that takes into account the correlation of the video frames, due to the dynamics of the scene both in the temporal and spatial domains (multiviews), and the available time-varying wireless channel capacity, to predict the evolution of the structure of the video segment to be sent. We implement a Markov Decision Process (MDP) model to foresee the optimal video encoding and packet scheduling policies to be used at the video source to generate the video frames with the goal of minimizing the overall distortion of the video consumed at the end user side. By means of simulation, we discuss the insights gained into the impact of frame correlation and time-varying wireless channel on the video stream generation and transmission, and the possible benefits from extending the model to more complex video coding and scheduling mechanisms.