Flash ALOHA project authors (project page by John Yao and collaborators) · Fast robot imitation learning

Flash ALOHA: Learning Fast and Precise Manipulation

Flash ALOHA is a system for learning fast, precise bimanual robot manipulation from human demonstrations. It collects near-human-speed demonstrations through kinesthetic teaching on a backdrivable platform. The learning pipeline adds a lightweight smoothing algorithm to address tracking instabilities from action chunking, and masks out the operator visible in the camera views. With 50 demonstrations per task, it completes ball tossing and catching, folding shorts, battery insertion and table bussing at over 80% success while close to human speed. The page also reports that a visual-inertial tracking device loses fidelity at high speed. No licence or code availability is stated.

Open the source ↗

Similar Agents & Automation releases

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.