GELLO

Berkeley's open-source low-cost leader-follower teleoperation device, using a scaled-down replica arm to control the real robot.

GELLO is a teleoperation framework proposed in 2023 by Pieter Abbeel's group at UC Berkeley. The approach is to 3D-print a scaled-down replica arm with the same joint structure as the target robot arm, fitting each joint with an off-the-shelf servo that reads its angle; the operator moves the replica arm, and the real robot follows along joint angle by joint angle — a form of leader-follower teleoperation. Because the two arms share the same kinematics, joint angles map directly across, unlike a VR controller, which has to convert a hand's pose into a robot-arm pose. Parts cost under $300, and all the hardware and software designs are open-sourced. The paper builds versions for the Franka, UR5, and xArm arms; user studies show it collects demonstrations more stably and faster than a VR controller or a 3D mouse, and it also supports bimanual and contact-rich tasks. The community has since contributed designs for other arms, including the FR3 and YAM, and it's commonly used to collect imitation-learning data.

ExampleAn operator holds the GELLO replica arm on the desk and performs a pick-and-place motion; the real Franka arm follows along in sync, and the camera footage and joint angles are recorded as one training demonstration.

Also called

GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators, GELLO Teleoperation Arm

Related

Leader-Follower Teleoperation · Teleoperation · ALOHA · Demonstration Data · SpaceMouse Teleoperation · VR Teleoperation

Sources

GELLO: A General, Low-Cost, and Intuitive Teleoperation Framework for Robot Manipulators (arXiv)
GELLO 项目主页 (Chinese)
gello_mechanical (GitHub)

As of

2024-07