ORCA Hand

An ETH Zurich/ORCA Dexterity open-source dexterous hand family focused on low-cost reproducibility and tactile variants

ORCA Hand

One-line judgment

ORCA Hand is the candidate for starting dexterous learning experiments with a reproducible, lower-cost hand instead of buying a high-end commercial platform. For flower work, its value is strongest when you expect to modify fingertips, compliance, or tactile sensing.

Key Specs

ItemValueInterpretation
ORCA v1 DoF17 DoF, 16 finger + 1 wristEnables hand and wrist posture experiments
Material costUnder 2,000 CHFMuch lower entry cost than Shadow/Allegro class hands
Assembly targetUnder 8 hoursPlausible for lab-built deployments
Learning experimentReported 7h17m and about 2,000 grasp cyclesSmall-dataset policy experiments are plausible
2026 familyLite 9 DoF, Hand 17 DoF, Touch 17 DoFBudget and tactile needs can be separated
Touch sensingReported 351 taxels per handCandidate for stem/petal contact classification

Meaning for Flower Work

ORCA’s strongest point is not a finished commercial spec sheet; it is an open, modifiable hand platform. Flower manipulation will likely need fingertip shape changes, softer pads, and contact-sensing experiments, so a reproducible open design is a real advantage.

Best initial tasks:

  • Grasp a single stem or stem bundle
  • Hold folded wrapping paper in place
  • Pull ribbon or tape ends without slipping
  • Train a tactile classifier for “too much force”

Strengths

  • Papers and project material make the design intent and reproduction path relatively transparent.
  • The low-cost/open direction makes fingertip modification, added sensors, and part replacement easier to justify.
  • The Touch line is relevant for tactile datasets around stems, petals, and soft contact.

Risks

  • A build-it-yourself route introduces quality variation and maintenance work.
  • ORCA Dexterity’s 2026 product line needs direct confirmation for availability, lead time, SDK, and support.
  • Flower-shop realities such as moisture, debris, pollen, and pad wear cannot be inferred from paper specs alone.

Integration Checklist

  • Decide whether the project needs a purchasable product or a lab-built open-source hand.
  • Confirm CAD, BOM, firmware, control code, and license coverage.
  • For tactile variants, verify raw tactile-stream access and calibration tooling.
  • Measure fingertip-pad wear and replacement cost after repeated grasping.

References