Distributed WebRTC-Based Forwarding for Scalable Volumetric Video Streaming

Abstract

As immersive media becomes more accessible, virtual counterparts to real-world experiences such as concerts and conferences have emerged, enabled by volumetric streaming pipelines for virtual reality (VR). User representation is central to these systems, with point clouds widely adopted for realistic avatars due to their balance between quality and performance. However, most systems struggle to scale to larger user counts. While recent work has proposed more scalable architectures, these typically focus on computational optimizations and fail to scale to high user counts due to network bottlenecks. To address this gap, we present an open-source, modular, distributed WebRTC-based volumetric streaming pipeline that employs multiple selective forwarding units (SFUs) to improve latency, throughput, and quality compared to a centralized SFU. Results show that, with 64 users, the distributed setup receives 147% more points with comparable transport latency, and reduces peak latency at lower user counts. Furthermore, leveraging quality adaptation enables stable latency across scales, achieving approximately 25 ms.