Viewing Distance-Aware Volumetric Video Caching and Rendering for XR Services

In this paper, we propose a novel volumetric video caching and rendering approach tailored for interactive extended reality (XR) services, aiming to enhance users' quality of experience (QoE). Specifically, an XR user's QoE is jointly determined by the actual quality of the video delivered to the user's XR device and the distance between the user's viewpoint and the location of virtual objects within the videos. Thus, we adopt a customized QoE model and develop an adaptive caching and rendering approach for XR where the users' viewing distances vary over time. In our proposed approach, point clouds representing virtual objects with various densities can be dynamically cached at the edge server, and then the most appropriate cached point cloud will be selected for rendering to each user. A long-term optimization problem is formulated to maximize the accumulated QoE of XR users over time. Solving the optimization problem is very challenging due to the unavailability of future information and the NP-hardness. Therefore, we first use a regularization technique to decouple the optimization problem into a series of one-shot optimization subproblems, whose relaxed optimal solutions can be efficiently obtained by using convex tools. Then, we design a dependent rounding algorithm to recover relaxed solutions to integral solutions without violating network resource constraints. Simulation results demonstrate that our proposed approach outperforms two benchmark algorithms in terms of the accumulated QoE.