enactic/openarm

What it solves

OpenArm provides an affordable, open-source humanoid arm designed for physical AI research. It addresses the need for a standardized, human-scale robotic arm that is safe for human-robot interaction and capable of handling contact-rich tasks in real-world environments.

How it works

OpenArm is a 7-degree-of-freedom (7DOF) humanoid arm. It utilizes a high backdrivability and compliance system to ensure safety and flexibility. The project includes a comprehensive ecosystem of software and hardware:

• Hardware: CAD data (STL, STEP, Fusion 360) for building the arm.
• Control: A CAN control library for low-level motor communication and ROS2 integration packages.
• Simulation: Support for Isaac Lab and MuJoCo for training and testing in virtual environments.
• Teleoperation: Packages for unilateral and bilateral control.
• Data: Tools and a Python API for recording and managing datasets for imitation learning.
• Environment: The OpenArm Cell, a standardized environment with unified lighting and camera placement to ensure research reproducibility.

Who it’s for

Researchers and developers working in physical AI, imitation learning, and humanoid robotics who need a flexible, open-source platform for data collection and deployment.

Highlights

• Open-Source Hardware: Full CAD data provided for DIY or assembled purchase.
• Human-Scale Proportions: Designed for human-like movement and movement in human environments.
• Safe Interaction: High backdrivability and compliance for safe human-robot interaction.
• Standardized Evaluation: The OpenArm Cell provides consistent conditions for global research reproducibility.
• Affordable Bimanual System: A complete bimanual system is available for $6,500 USD.