Ruka-v2: Tendon Driven Open-Source Dexterous Hand with Wrist and Abduction for Robot Learning
The development of dexterous robotic hands has faced significant challenges due to the lack of accessible and efficient hardware. In a groundbreaking move last year, researchers released Ruka, a fully open-sourced, tendon-driven humanoid hand designed to mimic human dexterity. With 11 degrees of freedom (2 per finger and 3 at the thumb), Ruka can be built for under $1,300, marking a significant milestone in the field.
Ruka was one of the first fully open-sourced humanoid hands, introducing a novel data-driven approach to finger control that incorporates tendon dynamics into its control system. Despite its innovative design, Ruka lacked critical features necessary for closely imitating human motions: wrist mobility and finger abduction/adduction. Acknowledging these limitations, the creators of Ruka have now unveiled Ruka-v2, which is set to revolutionize the capabilities of robotic hands further.
Introducing Ruka-v2
Ruka-v2 boasts an advanced tendon-driven design that includes a decoupled 2-degrees of freedom (2-DOF) parallel wrist and the ability to perform finger abduction and adduction. The enhancements in Ruka-v2 address the limitations of its predecessor and provide a more versatile tool for robotic manipulation.
- Parallel Wrist: The new wrist design allows for smooth and independent flexion/extension and radial/ulnar deviation. This feature significantly improves the hand’s ability to navigate and manipulate objects in confined spaces, such as cabinets.
- Finger Abduction: This enhancement permits the hand to grasp thin objects, perform in-hand rotations, and even execute tasks like calligraphy. The precision and versatility offered by abduction can facilitate a range of applications previously unattainable with Ruka.
Performance Evaluation
The researchers conducted a series of user studies to evaluate Ruka-v2 against its predecessor, Ruka. The findings were promising, demonstrating significant improvements in both efficiency and effectiveness:
- A remarkable 51.3% reduction in task completion time.
- A notable 21.2% increase in success rates across various teleoperated tasks.
These results underscore Ruka-v2’s enhanced capabilities and its potential to facilitate more complex robotic tasks. The hand’s design and functionality allow it to be utilized in diverse applications, from bimanual to single-arm teleoperation.
Applications in Robot Learning
Ruka-v2 is not just a mechanical marvel; it is also a powerful tool for robot learning. The hand can be employed in a wide array of tasks, including:
- Bimanual and single-arm teleoperation across 13 dexterous tasks.
- Autonomous policy learning on 3 complex tasks.
The flexibility and adaptability of Ruka-v2 make it an invaluable asset for researchers and developers in the field of robotics.
Access and Resources
All 3D print files, assembly instructions, controller software, and demonstration videos for Ruka-v2 are publicly available on the project’s website: https://ruka-hand-v2.github.io/. This commitment to open-source development encourages collaboration and innovation within the robotics community.
In conclusion, Ruka-v2 represents a significant step forward in creating accessible and versatile robotic hands, paving the way toward achieving human-level dexterity in robotic applications.