GELLO Open Source Robotic Arm Framework Affordable Intuitive Remote Control for Franka UR5

Imitative learning based on human teachings is a powerful framework that can be used to teach robots new skills. However, the performance of learning strategies is limited by the quality, scale, and diversity of demonstration data. The open-source project GELL0 aims to lower the threshold for collecting large-scale, high-quality human demonstration data by building a low-cost, intuitive robotic teleoperation system. GELLO is a general-purpose framework for building low-cost controllers for target robotic arms that are consistent with their kinematic structure, implemented with 3D-printed parts and off-the-shelf motors. The system is easy to build and intuitive to operate. Through extensive user research, GELLO has been tested to collect demonstration data more reliably and efficiently than teleoperation devices (such as VR controllers and 3D space markers) commonly used in the field of imitation learning. We also demonstrated GELL0's ability to perform complex two-handed and high-touch tasks.

To popularize GELL0, we have designed and built corresponding systems for three commonly used robotic arms (Franka, UR5), and all software and hardware are open-source.