Robot Platforms
Meet the fleet: Chloe the Unitree G1, Dexter the overbuilt Aloha Mini 2, and Rover the LeKiwi — all contributing to the Hugging Face LeRobot dataset collection initiative.
Every teleoperation session contributes to democratizing robotics AI through community-driven dataset collection. Our platform integrates with LeRobot's standardized data formats, helping create diverse, real-world datasets that advance robotic learning across different embodiments and behaviors.
The Social Robot Path
Social Teleop
Operate Chloe as a presence robot: gestures, greetings, and crowd interaction with a human in the loop.
Righting Robot Track
A dedicated recovery track lets Chloe attempt high-intensity choreography with automated, safe fall recovery.
Dance Performance
Routines built in the G1 Dance Studio (studio.ultimatebots.com) are validated against joint limits, then performed live on UFB.gg.
Dataset Contribution
Chloe generates full-body humanoid motion data: choreography sequences, balance recovery episodes, and human-robot interaction demonstrations, all formatted for LeRobot compatibility.
Data Standardization
All robot data is automatically converted to LeRobot Dataset v2.1 format, ensuring compatibility and contributing to the global robotics dataset.
Quality Assurance
Advanced filtering and validation ensure high-quality episodes, with AI-generated task descriptions and standardized camera views.
Real-world Variation
Multiple robot embodiments and diverse scenarios promote generalization across different behaviors and environments.