VR 全景视频时空切片传输的流量评估与建模

To characterize the traffic load of virtual reality (VR) panoramic video under the spatio-temporal slicing transmission mode, firstly, based on a real-word VR video viewing dataset, the relationship between the traffic load and the spatio-temporal slicing strategy was evaluated via simulations. Secondly, on the basis of the simulation results, a statistical traffic model was established, describing the interaction between the traffic load and the number of spatial tiles as well as the duration of temporal chunks. Finally, based on the established model, a hierarchical spatial-temporal slicing strategy optimization method was proposed, where the outer layer found the optimal number of spatial tiles by exhaustive searching while the inner layer could directly compute the optimal duration of temporal chunks. Research and simulation results show that the traffic load approximately follows a normal distribution, where the mean and variance are functions of the spatio-temporal slicing strategy. By establishing the traffic model for each type of videos, the optimized slicing strategy based on the established model performs close to the optimal strategy obtained by searching by simulations.