Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Ten Ways in which Virtual Reality Differs from Video Streaming

Gustavo de Veciana, Sonia Fahmy, George Kesidis, Voicu Popescu

TL;DR

This paper argues that networked virtual reality (VR) fundamentally differs from stored 2D video and therefore demands new system and network designs. It presents a taxonomy of ten key differences spanning four areas: application characteristics, rendering/adaptation, prefetching/caching, and transport, and discusses how edge/cloud support, adaptive LoD strategies, and tailored transport mechanisms are essential. The work highlights approaches such as near-far LoD, visibility-based adaptation, predictive caching, multicast edge delivery, and out-of-order reliable transport concepts, and identifies open problems in QoE measurement, benchmarking, and reproducible VR workloads. Together, these insights illuminate pathways to scalable, low-latency VR that can support future metaverse-like applications and multi-user interactions.

Abstract

Virtual Reality (VR) applications have a number of unique characteristics that set them apart from traditional video streaming. These characteristics have major implications on the design of VR rendering, adaptation, prefetching, caching, and transport mechanisms. This paper contrasts VR to video streaming, stored 2D video streaming in particular, and discusses how to rethink system and network support for VR.

Ten Ways in which Virtual Reality Differs from Video Streaming

TL;DR

This paper argues that networked virtual reality (VR) fundamentally differs from stored 2D video and therefore demands new system and network designs. It presents a taxonomy of ten key differences spanning four areas: application characteristics, rendering/adaptation, prefetching/caching, and transport, and discusses how edge/cloud support, adaptive LoD strategies, and tailored transport mechanisms are essential. The work highlights approaches such as near-far LoD, visibility-based adaptation, predictive caching, multicast edge delivery, and out-of-order reliable transport concepts, and identifies open problems in QoE measurement, benchmarking, and reproducible VR workloads. Together, these insights illuminate pathways to scalable, low-latency VR that can support future metaverse-like applications and multi-user interactions.

Abstract

Virtual Reality (VR) applications have a number of unique characteristics that set them apart from traditional video streaming. These characteristics have major implications on the design of VR rendering, adaptation, prefetching, caching, and transport mechanisms. This paper contrasts VR to video streaming, stored 2D video streaming in particular, and discusses how to rethink system and network support for VR.

Paper Structure

This paper contains 9 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Application Characteristics
  3. Interaction, Content Dynamics, and Resource Needs
  4. QoE Requirement
  5. Rendering and Adaptation
  6. Prefetching and Caching
  7. Transport
  8. Open Research Problems
  9. Conclusions

Figures (3)

  • Figure 1: Manhattan urban virtual environment. The triangle mesh representation (top), shown here in wireframe, has 3.7 million triangles and 2 million vertices. The point-based representation (bottom), shown here with 2 $\times$ 2 pixel splats, has 20 million points and a density of 0.1 points/m$^2$.
  • Figure 2: The discontinuity between the near and far regions (left) is eliminated with an intermediate region morphed to connect the two regions (middle), yielding frames comparable to ground truth frames rendered from the geometry of the entire virtual environment (right) popescu2022ismar.
  • Figure 3: Progressive refinement of point cloud rendering of "Tikal" with an increased level of detail in the right image (screen shots obtained from https://doi.org/10.26301/708h-ss96 from openheritage3d.org).