Maintaining Demonstration Quality in a 100-Robot Teleoperation Pipeline
Published in RSS 2026 Workshop: It's the Demos, 2026
Recommended citation: Kapoor, K., Khan, S., McCalmon, J., Michel, J., Mohammed, A., Nikiforova, K., Oppenheimer, A., Szabo, V., & Watkins, D. (2026). Maintaining Demonstration Quality in a 100-Robot Teleoperation Pipeline. Workshop on "It's the Demos: A Deep Look at the Role of Demonstration Quality in Imitation-Based Robot Manipulation" at Robotics: Science and Systems (RSS), Sydney, Australia. /files/2026_rss_df1_demo_quality.pdf
Abstract
At fleet scale, maintaining demonstration quality and consistency becomes a first-order engineering problem. We study these problems in Data Factory 1 (DF1), a 100-robot fleet of bimanual manipulators that supports 1,000 hours of teleoperated demonstrations per day from a large, internationally distributed operator pool. We treat demonstration quality as an explicit objective at each stage of the pipeline: reducing teleoperation latency with PTeleop (a proprioceptive VR interface that improves teleoperation speed by 35.8% with over 10% fewer mistakes), monitoring specification adherence while preserving behavioral diversity, normalizing operator speed variation during preprocessing, grading demonstrations for reward-weighted training and data filtering, and using model rollouts with teleoperator interventions to surface and correct failures at fleet scale. Together these stages provide a blueprint for producing high-quality real robot data at scale.
Accepted to the It’s the Demos workshop at RSS 2026. Workshop poster.