Paper Overview
- Field: ML
- Authors: Nicolas Kosanovic, Jordan Dowdy, Jean Chagas Vaz
- arXiv: 2507.18393
Summary
Full-sized humanoid robot capabilities have grown exponentially in recent years, aiming towards general-purpose deployment in human environments. A popular control method used by manufacturers utilizes Virtual Reality for upper-body teleoperation and Reinforcement Learning for lower-body balance and locomotion control. As a result, a single remote operator can see, manipulate, and navigate about a real, distant physical environment.
However, this powerful control stack is often relegated to expensive full-sized robots, many of which are inaccessible to the research community. Miniature humanoids are more prevalent, but employ less biomimicry in their design (e.g., fewer sensors, degrees of freedom, etc.) and lack similar developments.
This paper describes a compliant full-body telepresence control stack developed from scratch for a miniature humanoid. Experiments on ROBOTIS OP3 hardware demonstrate walking speeds of up to 0.45 m/s, independent of arm motion. Tele-loco-manipulation is demonstrated through cube relocation experiments with expert human operators: on average, the teleoperation system moved two distinct 40-gram cubes within 10 minutes, over a total walking distance of 5 meters.
Overall, the developed system demonstrates the potential of miniature humanoid tele-loco-manipulation.
--- *Auto-collected on 2026-07-24*