Prediction, Communication, and Computing Duration Optimization for VR Video Streaming

TL;DR

This work addresses proactive VR video streaming by jointly optimizing the observation window for tile prediction and the durations for computing and transmitting predicted tiles within a fixed proactive horizon $T_{ ext{ps}}$. It derives a closed-form, globally optimal solution through problem decomposition into P2 and P3, and reveals a critical boundary $T_{ ext{cc}}^{\max}$ that separates resource-limited and prediction-limited regimes. The approach is validated with three predictors (LR, CB, GRU) on a real dataset, showing QoE gains from the joint optimization over non-optimized baselines and providing practical insights into system design. The results quantify when to invest in better predictors versus more capable computing/communication resources, enabling informed MEC-assisted VR streaming strategies.

Abstract

Proactive tile-based video streaming can avoid motion-to-photon latency of wireless virtual reality (VR) by computing and delivering the predicted tiles to be requested before playback. All existing works either focus on designing predictors or allocating computing and communications resources. Yet to avoid the latency, the successively executed prediction, communication, and computing tasks should be accomplished within a predetermined time. Moreover, the quality of experience (QoE) of proactive VR streaming depends on the worst performance of the three tasks. In this paper, we jointly optimize the duration of the observation window for predicting tiles and the durations for computing and transmitting the predicted tiles, aimed at balancing the performance for three tasks to maximize the QoE given arbitrary predictor and configured resources. We obtain the closed-form optimal solution by decomposing the formulated problem equivalently into two subproblems. With the optimized durations, we find a resource-limited region where the QoE increases rapidly with configured resources, and a prediction-limited region where the QoE can be improved more efficiently with a better predictor. Simulation results using three existing predictors and a real dataset validate the analysis and demonstrate the gain from the joint optimization over non-optimized counterparts.

Figures (7)

• Figure 1: Proactively streaming the ($l$+1)th segment of a VR video. $t^b$ is the start time for streaming the ($l$+1)th segment, and $t^e$ is the start time of playback of the ($l$+1)th segment.
• Figure 2: The relation of $\frac{1}{T_{\mathrm{cc}}^{\max}}$ with $C_{\mathrm{com}}$ and $C_{\mathrm{cpt}}$, and the optimal durations v.s. $\frac{1}{T_{\mathrm{cc}}^{\max}}$.
• Figure 3: Optimal durations v.s. $\frac{1}{T_{\mathrm{cc}}^{\max}}$ and $T_{\mathrm{ps}}$.
• Figure 4: Prediction performance of three existing predictors and corresponding fitting functions.
• Figure 5: Impact of predictors, prediction performance and CC task completion rate on QoE.