Archon & OpenDriveLab at RSS 2026

How can human-centric data become a principled substrate for whole-body intelligence? Drawing on sustained work in embodied intelligence, autonomous systems, and human-centric robot learning, the speaker will examine how behavior traces and demonstrations can be structured for cross-embodiment transfer. The discussion will foreground evaluation beyond isolated demos, toward robot capabilities that generalize across bodies, tasks, and environments.

What should a robot world model predict, abstract, and control to support long-horizon decision making? Drawing on a research trajectory in embodied AI systems and model-based autonomy, the speaker will connect controllable prediction with planning, imagination-based policy improvement, and evaluation. The talk will frame world models as operational interfaces between perception, memory, and physical consequence.

What does robust execution require when whole-body policies encounter contact, latency, and task drift in the physical world? Drawing on deployment-oriented research in VLA systems and long-horizon robot control, the speaker will examine execution-time feedback, recovery, and progress estimation. The discussion will position robustness not as a post-hoc correction, but as a design principle for reliable humanoid autonomy.