O PEN T EACH: A Versatile Teleoperation System for Robotic Manipulation Aadhithya Iyer 1 Zhuoran Peng 1 Yinlong Dai 1 Irmak G¨ uzey 1 Siddhant Haldar 1 Soumith Chintala 2 Lerrel Pinto 1 1 New York University 2 Meta https://open-teach.github.io/ Open Teach Fig. 1: We present OPEN TEACH, a unified robot teleoperation framework that supports multiple arms and hands, allows mobile manipulation, is calibration-free, and works across both simulation and real-world environments. OPEN TEACH uses a VR headset for teleoperation and offers low latency and high-frequency visual feedback. This high-frequency operation allows human users to correct for robot errors in real time, facilitating the execution of intricate, long-horizon tasks. From making a sandwich and ironing cloth to placing items in a basket and lifting it and approaching a cabinet and opening it,OPEN TEACH delivers a comprehensive, user-friendly teleoperation experience for a wide range of applications. OPEN TEACH is fully open-source. arXiv:2403.07870v1 [cs.RO] 12 Mar 2024