Low-Latency Networks: The Nervous System for Robots | JB Kempf, Kyber | FUSE Toulouse 2026

Robots and autonomous vehicles still need humans to watch them, and sometimes to take control. Jean-Baptiste Kempf, creator of VLC and founder of Kyber, explains why video and networking built for people watching people break down for machines, and how an open-source stack built on QUIC delivers 8 ms glass-to-glass latency for video, sensors and controls. Topics: remote observation and control, latest data vs best data, TCP, UDP, WebRTC and QUIC, separating sensor, control and data streams, forward error correction, clock drift across sensors, multipath, and real-world builds from cloud gaming to drones and boats. Recorded at FUSE Toulouse, October 6, 2026, Orange Innovation Center, Balma, France. FUSE is a global technical conference series for localization and Physical AI engineers. Next: FUSE San Francisco, November 19, Foundry SF. Register: https://fuse.pointonenav.com/sf Chapters 00:00 Intro: VLC, open source and Kyber 01:52 Why remote control: projecting compute and skills 03:23 Latest data beats best data 04:19 8 ms glass to glass for video, sensors and control 06:19 TCP, UDP, WebRTC and why QUIC 09:17 One connection, three stream types 12:15 Demos: cloud gaming and “negative latency” 16:28 Video calls vs control loops 18:52 Reliability, clock drift and portability 25:46 Multipath, Kyber Hub and real-world builds 31:38 Q&A: localization, safety-critical use, remote AI, business model