Point Cloud Streaming with Latency-Driven Implicit Adaptation using MoQ

TL;DR

This work tackles the challenge of real-time point cloud streaming for VR/AR by introducing latency-driven implicit adaptation using Multiple Description Coding and Media Over QUIC. The system leverages delivery timeouts and per-Track QUIC priorities to trade latency for quality on a per-client basis, enabling ultra-low-latency teleconferencing as well as higher-quality broadcasts for relaxed latency targets. Empirical results show substantial throughput gains and PCQM improvements when allowing greater latency, and a content-aware PoC demonstrates how prioritizing salient points can further enhance perceived quality. The approach offers a practical, transport-layer solution that reduces the need for explicit bitrate estimation while supporting diverse XR applications and realistic DASH baselines.

Abstract

Point clouds are a promising video representation for virtual and augmented reality. Their high-bitrate, however, has so far limited the practicality of live streaming systems. In this work, we leverage the delivery timeout feature within the Media Over QUIC protocol to perform implicit server-side adaptation based on an application's latency target. Through experimentation with several publisher and network configurations, we demonstrate that our system unlocks a unique trade-off on a per-client basis: applications with lower latency requirements will receive lower-quality video, while applications with more relaxed latency requirements will receive higher-quality video.

Figures (5)

• Figure 1: Overall system diagram with an example from our testing. A client that subscribes with a higher latency target will generally receive more representations than a client with a lower latency target, resulting in a higher-quality frame reconstruction. For a given frame, each partition is sent across a unique QUIC stream.
• Figure 2: Stall rate across all test configurations. A frame is counted as stalling if no representation is received.
• Figure 3: Average throughput comparisons between DASH and MoQ for each encoder and packaging configuration.
• Figure 4: Time series comparison between MoQ and DASH on the ricardo9 sequence loop2016microsoft, showing the number of received representations at various bandwidth limitations. The configuration with 10 encoders and 30 FPG was used. For readability, we show only two MoQ traces, to illustrate the improvement in throughput as the delivery timeout increases.
• Figure 5: Impact of content-aware sampling on the andrew9 sequence loop2016microsoft, with higher transport priority given to points detected in the head region. Configuration: 10 Tracks, 30 FPG, 600 Mbps bandwidth, 100 ms timeout.