OpenArm Releases Open-Source Dual-Arm Robot and Evaluation Platform
Key point
OpenArm has released an open-source dual-arm robot hardware stack featuring 7 degrees of freedom, along with tools for data learning and policy evaluation.
Details
Tokyo-based startup Enactic has released OpenArm, an open-source dual-arm robot project based on research into the autonomous stack of its commercial humanoid, Ena. The project aims to enhance the reproducibility of SOTA (state-of-the-art) results by providing an integrated ecosystem ranging from CAD designs to data formats and evaluation services.
Hardware and Control Specifications
OpenArm supports 7 degrees of freedom arms, adopting Damiao motors to provide 1kHz CAN-FD communication and bidirectional force feedback. The rated payload is 4.1kg, with a peak of 6.0kg, while the effective payload considering end-effector weight is approximately 2.6kg. The leader arm (KER) for teleoperation implements 1:1 motion without motors and is ready for use within 30 seconds of being worn.
Software and Learning Ecosystem
The software is built on dora-rs nodes and supports ROS2, MuJoCo, and Isaac Lab simulation environments. Data can be converted to Hugging Face's LeRobot format, ensuring compatibility with existing ecosystems. Enactic has released a cube grasping dataset collected via VR teleoperation (30 episodes, 5,405 frames) and an ACT policy model, enabling training and validation in simulation environments without physical hardware.
Evaluation Service and Licensing
OpenArm Online offers a policy submission and evaluation service based on Docker images. Currently, only simulation tasks are supported, with limited integration for evaluation on actual equipment (Cell). The software is licensed under Apache-2.0, while the hardware CAD is licensed under CERN-OHL-S-2.0, mandating the disclosure of modified designs.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.