David J. Watkins

David J. Watkins

Research Lead | Foundation models and robot manipulation at scale | Advisor @ Tola Capital | Built teams of 40 | Ph.D. Columbia

Bridging Handheld and Teleoperated Supervision for Contact-Rich Manipulation via State-Gated Experts

Published in Submitted to ICRA 2027, 2026

Recommended citation: Surendran, V., Peri, N., & Watkins, D. (2026). Bridging Handheld and Teleoperated Supervision for Contact-Rich Manipulation via State-Gated Experts. Submitted to the IEEE International Conference on Robotics and Automation (ICRA). https://nperi-rai.github.io/bridge-project/

Abstract

BRIDGE combines handheld and teleoperated data collection for robotic manipulation using a mixture-of-experts approach that routes between specialist task-phase heads conditioned on the current robot state. This state-gating improves success rates by up to 36.7% over handheld-only baselines on contact-rich tasks.

Project page, code, and datasets: nperi-rai.github.io/bridge-project. Work with Vidullan Surendran and Neehar Peri (RAI Institute, Carnegie Mellon University).