DexhandsIndustry hub

DEXTEROUS MANIPULATION

The world of dexterous hands.

Compare hardware. Discover applications. Share what works. An open reference and community for dexterous manipulation.

Dexterous Hands · Course contents
Learn / Dexterity Foundations
6 min reading + activities · Practical · no coding required

Choose a hand for the task, not the spec sheet

Turn “we need a dexterous hand” into requirements that a supplier or research team can actually evaluate.

Dexterity starts with purposeful finger coordination. A pinch is one grasp; moving an object within the hand is a further capability.Concept illustration · not a product specification or test result.
What you will learn

Translate a manipulation task into a comparable, evidence-based shortlist.

Describe the object and the action

A hand that lifts a pen may not be able to turn it into a writing grip. A hand that holds an egg may not control contact while moving it. Start with an object set, starting conditions, intended motion and end state. Note which actions belong to the arm and which must happen inside the hand.

Separate essential capability from preference

Write down the fingers and contacts needed for the task. Consider thumb opposition, fingertip reach, passive compliance and the space around the object. Then check sensing and control: what observation would tell you that contact, slip or the desired pose has occurred? An attractive DOF count is not a substitute for that answer.

Check the complete setup

Compare the delivered assembly, including external actuators or controllers where required. Ask about mount geometry, power, communication, supported operating systems, SDK examples and calibration. A simulation model and a hardware driver solve different problems. Identify which exact hand revision each resource supports.

Make an acceptance task before asking for a quote

For an illustrative pen task, ask for a repeatable pickup and repositioning sequence across agreed starting poses. Specify what counts as a drop, a reset or operator help, and whether the trial is teleoperated or autonomous. Agree the attempt count and acceptable outcome with the supplier; there is no universal threshold suitable for every task.

Keep the shortlist explainable

Choose two or three plausible hands and mark each requirement as supported by evidence, needing a test, or currently unsupported. Compare source dates and test conditions. Use the same task brief when requesting information so the responses answer the same questions.

Worked example: from object to requirement

Illustrative selection questions, not measured product performance.
TaskCapability to investigateEvidence to request
Turn a pen in the palmContact changes and fingertip reachFull attempts, start poses and control mode
Move a fragile objectContact feedback and compliant behaviorSensor logs plus the damage criterion
Place a connectorPose accuracy and arm–hand coordinationTolerance, fixture and repeatability setup
Collect demonstrationsAction mapping and synchronized observationsAn anonymized episode and its schema

Try it yourself

Your practical task

  1. Choose a real object and write the desired starting and ending states.
  2. List three essential capabilities and one integration constraint.
  3. Compare two hand cards and mark evidence, unknowns and required tests.
  4. Write the same acceptance task for both candidates.

What to produce: A one-page task brief and a two-hand evidence comparison.

Check your understanding

A hand holds a pen in a video. What remains unproven for a pen-reorientation task?

Original sources & next steps

NIST: grasping and manipulation measurement ↗MuJoCo Menagerie: model-specific documentation ↗

Original Dexhands teaching material. Lesson and linked references reviewed 2026-09-24. Research links are not endorsements or evidence of hardware compatibility.

Start a discussion

Published posts are public. First contributions are reviewed. Only share material you have permission to disclose.

Report a post