Matching Prediction to Communication and Computing in Proactive VR Video Streaming
share
Proactive tile-based video streaming can avoid motion-to-photon latency of wireless virtual reality (VR) by computing and delivering the predicted tiles in a segment to be requested before playback. However, all existing works either focus on the tile prediction or focus on tile delivering and computing, overlooking the important fact that prediction, communication, and computing have to share the same duration. Moreover, the quality of experience (QoE) of proactive tilebased streaming depends on the worst performance of prediction, communication, and computing, thus it is vital to match the prediction performance to the communication and computing capability. In this paper, utilizing optimizing duration allocation for prediction, communication, and computing, we match the prediction performance to the communication and computing capability to maximize the QoE of watching a VR video. We find the global optimal solution with closed-form expression by decomposing the original problem equivalently into two matching subproblems. From the optimal solution, we find two regions where tile prediction and communication and computing capability respectively play the dominant role, and reveal the tradeoff between the prediction performance and the communication and computing capability. Simulation results using two existing tile prediction methods with a real dataset demonstrate the gain of the optimized duration over the non-optimized duration in prediction, communication, and computing.
READ FULL TEXT