OPEN TEACH: A Versatile Teleoperation System for Robotic Manipulation

Open-sourced, user-friendly tools form the bedrock of scientific advancementacross disciplines. The widespread adoption of data-driven learning has led toremarkable progress in multi-fingered dexterity, bimanual manipulation, andapplications ranging from logistics to home robotics. However, existing datacollection platforms are often proprietary, costly, or tailored to specificrobotic morphologies. We present OPEN TEACH, a new teleoperation systemleveraging VR headsets to immerse users in mixed reality for intuitive robotcontrol. Built on the affordable Meta Quest 3, which costs 500, OPEN TEACHenables real-time control of various robots, including multi-fingered hands andbimanual arms, through an easy-to-use app. Using natural hand gestures andmovements, users can manipulate robots at up to 90Hz with smooth visualfeedback and interface widgets offering closeup environment views. Wedemonstrate the versatility of OPEN TEACH across 38 tasks on different robots.A comprehensive user study indicates significant improvement in teleoperationcapability over the AnyTeleop framework. Further experiments exhibit that thecollected data is compatible with policy learning on 10 dexterous andcontact-rich manipulation tasks. Currently supporting Franka, xArm, Jaco, andAllegro platforms, OPEN TEACH is fully open-sourced to promote broaderadoption. Videos are available at https://open-teach.github.io/.

Further reading#

• Access Paper in arXiv.org