ORCA Dexterity announces three new open source robotic hands
Announcement video from ORCA Dexterity (the ETH Zurich Soft Robotics Lab spinout behind the ORCA hand, arXiv:2504.04259) unveiling three open-source, 3D-printable tendon-driven anthropomorphic hands for manipulation research: orcahand lite (9-DoF adaptive, from 3,500), and orcahand touch (17-DoF + 351 taxels × 5 digits with 3D force vectors, from $6,100). All three share the same open firmware (orca_core), are field-repairable, and mount to the ISO 9409-1 standard. Video itself has no transcript; details reconstructed from the ORCA project page and third-party coverage.
Key claims
Section titled “Key claims”- Three-tier product line from a single tendon-driven platform: 9-DoF
lite(from 3,500), and 17-DoFtouchwith per-finger tactile arrays (from 17,900 [video description]. - The
touchvariant integrates 351 taxels per hand, each outputting a full 3D force vector (shear + slip + normal) at 0.1 N sensitivity and 1 mm spatial resolution [description]. - All three hands share firmware (
orca_coreon GitHub), mount to ISO 9409-1, and are designed to be field-repairable and 3D-printable end-to-end [description]. - Underlying platform is the ETH Zurich ORCA v1: 17-DoF tendon-driven anthropomorphic hand, integrated tactile sensors, one-person assembly in <8 hours, material cost below 2,000 CHF ([arXiv:2504.04259 abstract]).
Method
Section titled “Method”The ORCA design is tendon-driven with popping joints, auto-calibration, and integrated tensioning. The touch tier extends this with a whole-hand taxel array (351 taxels × 3D force per digit), delivered as an open hardware package: 3D-print files, BoM, firmware, and control code released together. The lite 9-DoF variant swaps the finger kinematics for an adaptive underactuated design to hit the $1,500 price floor while keeping the same software interface and mounting standard as the higher tiers.
Results
Section titled “Results”Video is an announcement only — no benchmark numbers are shown. The underlying ORCA v1 hand has separately been evaluated on a 7 h 17 min continuous pick-and-place run under an imitation-learned policy [srl.ethz.ch/orcahand.html] and across teleoperation / imitation-learning / zero-shot sim-to-real RL tasks in the arxiv paper [arXiv:2504.04259 §I]. The new three-tier lineup extends the same platform to a lower price floor ($1,500 lite) and a richer contact-sensing tier (351-taxel touch).
Why it’s interesting
Section titled “Why it’s interesting”Directly extends the sensor-placement-dominates-type finding from Tactile Genesis: Exploring Tactile Sensors at Scale for Learning Dexterous Tasks into commercially procurable hardware: the touch tier commits to whole-hand tactile coverage rather than fingertip-only, at a price point that undercuts commercial dexterous hands by an order of magnitude. Complements the open-hardware direction of Aero Hand Open: low-cost, open-source, lightweight, anthropomorphic dexterous hand (Aero Hand Open at $314, no tactile) and Solving Dexterity: A Full-Stack Approach (mimic hand M1 + wearable U1) (mimic M1 with motor-current implicit force sensing) on the same open-anthropomorphic-hand axis — three distinct bets on which sensing modality is load-bearing (whole-hand taxel arrays vs no tactile vs joint-torque implicit force). Puts a hardware substrate under the Universal Manipulation Exoskeleton: Learning Compliant Whole-body Policies with Real-time Torque Feedback and Getting a Grip on Robotic Data Collection — Co-Designed Koala Gripper for Force-Mediated Manipulation “force-in-the-collection-loop” recipes.
See also
Section titled “See also”- Tactile Genesis: Exploring Tactile Sensors at Scale for Learning Dexterous Tasks — the simulator-driven case for whole-hand tactile coverage the
touchtier operationalizes - Aero Hand Open: low-cost, open-source, lightweight, anthropomorphic dexterous hand — sibling open-anthropomorphic hand at the sparse-sensing end of the axis
- Solving Dexterity: A Full-Stack Approach (mimic hand M1 + wearable U1) — mimic M1’s motor-current implicit-force alternative to per-taxel arrays
- T-Rex: Tactile-Reactive Dexterous Manipulation — policy-side answer that the
touchtier’s taxel stream is well-matched to - Tactile sensing for manipulation — concept page
- Human-to-Robot Retargeting — anthropomorphic form-factor rationale (learning from human-hand video)