The ORCA Hand: Open-Source Dexterity at Your Fingertips
ORCA is a reliable, cost-effective robotic hand that combines human-level dexterity with open-source accessibility. With plug-and-play design, tactile sensors, and zero-shot learning support, ORCA accelerates manipulation research for all.
General-purpose robots should possess human-like dexterity and agility to perform tasks with the same versatility as us. A human-like form factor further enables the use of vast datasets of human-hand interactions. However, the primary bottleneck in dexterous manipulation lies not only in software but arguably even more in hardware. Robotic hands that approach human capabilities are often prohibitively expensive, bulky, or require enterprise-level maintenance, limiting their accessibility for broader research and practical applications. What if the research community could get started with reliable dexterous hands within a day? We present the open-source ORCA hand, a reliable and anthropomorphic 17-DoF tendon-driven robotic hand with integrated tactile sensors, fully assembled in less than eight hours and built for a material cost below 2,000 CHF. We showcase ORCA's key design features such as popping joints, auto-calibration, and tensioning systems that significantly reduce complexity while increasing reliability, accuracy, and robustness. We benchmark the ORCA hand across a variety of tasks, ranging from teleoperation and imitation learning to zero-shot sim-to-real reinforcement learning. Furthermore, we demonstrate its durability, withstanding more than 10,000 continuous operation cycles---equivalent to approximately 20 hours---without hardware failure, the only constraint being the duration of the experiment itself.
All design files, source code, and documentation are available at:
The ORCA v1 hand
The ORCA v1 hand is a tendon-driven, anthropomorphic robotic platform designed to closely replicate the structure and movement of a human hand. It features 17 degrees of freedom—16 in the fingers and 1 in the wrist—enabling a wide range of dexterous manipulation tasks. Its design balances functionality, reliability, and accessibility.
Open-Source:
All design files (STLs), control software, material lists, and step-by-step assembly instructions (including videos) are publicly available on our external page dashboard. The project is built to encourage replication, customization, and experimentation by researchers, educators, and hobbyists.
Reliable:
The ORCA hand incorporates several design choices focused on robustness and longevity. Joints are engineered to safely dislocate under excess load rather than break, which protects components and simplifies repairs. Additionally, auto-calibration, low-friction tendon routing through the center of rotation of each joint, and a modular layout contribute to consistent performance and long-term usability.
Cost-Effective:
The hand is centered around a 3D-printable design, with a total material cost under 2,000 CHF. All structural components can be printed using a standard 3D printer, and the remaining off-the-shelf parts are widely available and easy to source online, ensuring accessibility for a wide range of users.
Anthropomorphic:
Designed to match the proportions and joint layout of the human hand, the ORCA v1 includes an opposable thumb and an actuated wrist. It mirrors human joint configurations, including MCP, PIP, and ABD joints, enabling it to interact with everyday tools and objects originally designed for human use, and to perform a broad spectrum of manipulation tasks with human-like agility. This anthropomorphic structure also significantly simplifies teleoperation/retargeting and facilitates a more straightforward training on human hand data.
Together, these features make the ORCA v1 hand a practical and versatile platform for hands-on exploration in robotics and manipulation-focused learning and development.
A Fully Open Platform
The ORCA v1 is a fully open robotic hand, with all design files (STLs), core control code, and beginner-friendly documentation soon available. The platform will include essential resources such as step-by-step assembly instructions, repair guides, and visual references to support easy replication and integration into research workflows. Designed for accessibility, it relies on non-proprietary, widely available materials and includes a comprehensive, regularly updated Bill of Materials (ΒΟΜ), with direct sourcing links. Its low cost, simplicity, and modularity make it a practical tool for education and research labs, especially in the development and benchmarking of machine learning-based manipulation models. By offering a shared, accessible and standardized hardware platform, ORCA fosters reproducibility and accelerates collaboration across institutions, enabling the robotics community to collectively build datasets, compare policies, and push the boundaries of dexterous manipulation. Released under permissive MIT and Creative Commons licenses for non-commercial use, the project actively encourages open contribution and knowledge sharing.
7h+ Uninterrupted Imitation Learning
To highlight the reliability of the orca v1 hardware platform, we designed a continuous pick-and-place evaluation task. The robotic hand is required to pick up a cube from a table and place it on a sliding surface, which then causes the cube to fall back onto a random location on the table, thus resetting the experiment. We trained different policies via imitation learning and deployed the most successful policy for 7h 17min (which corresponds to approx. 2,000 grasping cycles) with no human intervention on the ORCA hand's hardware and minimal intervention in aiding in the pick-and-place task. Throughout the test, the policy maintained consistent perfor- mance, with no tendon slack or rupture, and the experiment was concluded not due to failure but because it sufficiently demonstrated the system's reliability, robustness and effectiveness in long-duration tasks. The orca hand's ability to reliably perform dexterous tasks without human intervention for multiple
hours could potentially accelerate real-world RL applications for dexterous manipulation research.
Zero-shot Sim-To-Real Reinforcement Learning
We demonstrate the use of reinforcement learning in simulation to train the orca hand for dexterous manipulation tasks in the real world. Using an IsaacGymEnvs wrapper, we train 4,096 simulated orca hand models in parallel with an advantage actor-critic architecture to learn in-hand ball reorientation. After just one hour of training with domain randomization, we successfully deploy a robust policy from simulation to the physical orca hand, enabling it to reliably reorient a tennis ball along a specified rotation axis.
Reliability Experiment
To evaluate the reliability and robustness of the ORCA hand in long-duration tasks, we conducted an experiment in which we actuate the hand joints continuously for 2.5 hours. We attach a plush animal to the palm of the hand and have it grasp it with all fingers every four seconds. Moreover, to test the durability of the wrist joint, we flex and extend the wrist to 40◦ every 16 seconds. The hand reliably performs the same grasping movement for all 2,250 grasping cycles without breaking, motor shut- down, or excessive tendon slack buildup. Additionally, the maximum current used by each motor remains roughly the same during 2.5 hours of uninterrupted joint movement, which further demonstrates the robustness of the hand, the high repeatability of joint movement and its capacity for long-duration operation. In particular, the experiment was not terminated due to failure, but was concluded after 2.5 hours because it was deemed a sufficient demonstration of the system's reliability.
Fast & Easy Assembly
The orca hand is designed to be easily assembled and repaired. Given all the necessary mechanical parts, it takes one person less than 8 hours to assemble one entire orca v1 hand with integrated sensors, tensioned tendons, and soft skin. A lot of time has been invested to develop a comprehensive, easy-to-follow assembly guide atour dashboard, which includes detailed instructions, images, and videos for the full assembly process. Specially designed "poppable" joints simplify both the assembly and repair processes, as joints can safely dislocate under load instead of breaking, reducing downtime and extending the hardware's lifespan. Because all parts are 3D-printed, they can be easily replaced in case of damage. This makes the orca hand a great product for any robotics research lab, education group, or hobbyist who want to create and build projects with a reliable, cost-effective open-source robotic hand.
Citation
If you use our work, please cite us using the following BibTeX entry:
@misc{christoph2025orcaopensourcereliablecosteffective,
title={ORCA: An Open-Source, Reliable, Cost-Effective, Anthropomorphic
Robotic Hand for Uninterrupted Dexterous Task Learning},
author={Clemens C. Christoph and Maximilian Eberlein and Filippos Katsimalis
and Arturo Roberti and Aristotelis Sympetheros and Michel R. Vogt and
Davide Liconti and Chenyu Yang and Barnabas Gavin Cangan and Ronan J. Hinchet and Robert K. Katzschmann},
year={2025},
eprint={2504.04259},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2504.04259}
}