Edge Rendering Architecture for multiuser XR Experiences and E2E Performance Assessment

TL;DR

The paper addresses the high QoS demands of XR by proposing an edge-rendering architecture that offloads rendering to a nearby edge server and streams per-user views via WebRTC. It introduces a practical QoS measurement framework using QR-code timestamps for video and synchronized audio tones for audio, all aligned with NTP clock synchronization, to quantify M2P, M2E, inter-device, and intra-media latencies. The authors implement and evaluate the system with Unity Render Streaming on a GPU-enabled edge, testing Ethernet, Wi-Fi, and 5G-SA across up to five concurrent users, demonstrating favorable latency and synchronization on wired and 5G networks but notable degradation on Wi-Fi. This work provides both a realizable XR edge-rendering solution and a portable QoS assessment methodology that can guide deployment and optimization of multiuser XR services in MEC environments.

Abstract

Holographic communications are gaining ground among emerging eXtended-Reality (XR) applications due to their potential to revolutionize human communication. However, these technologies are characterized by higher requirements in terms of Quality of Service (QoS), such as high transmission data rates, very low latency, and high computation capacity, challenging current achievable capabilities. In this context, computation offloading techniques are being investigated, where resource-intensive computational tasks, like rendering XR experiences, are shifted from user devices to a separate processor, specifically an Edge Computing instance. This paper introduces an Edge Rendering architecture for multiuser XR experiences, implements it on top of widely employed XR and Web technologies, and proposes a method based on image and audio processing to evaluate its performance in terms of end-to-end media streaming latency, inter-device, and intra-media synchronization when employing different access networks.

Figures (5)

• Figure 1: General architecture of the system.
• Figure 2: Testbed setup.
• Figure 3: M2P and M2E latency.
• Figure 4: Inter-device asynchrony.
• Figure 5: Intra-media asynchrony.