← All topics

Guide · Data and hardware · Topic 23 of 32

Dexterous hands

Multi-fingered hands in depth: real hardware and its specifications, how a hand connects to the action head, retargeting, and using touch.

14 minute readBuilds on: Embodiments, Senses beyond vision

Why hands are a separate problem

Most robot learning results use a parallel gripper: two fingers and one number, open or closed. It picks things up and puts them down. It cannot turn a key, use scissors, or shift an object in its grasp without setting it down. Those need several fingers, and a hand changes every part of the pipeline at once: the hardware, how data is collected, what the policy has to sense, and how many numbers it outputs at every step.

This note goes through each of those in turn, with real hardware and its published specifications.

What a robot hand is made of

A Shadow Dexterous Hand holding a light bulb between thumb and fingers, with a bundle of tendons running down into the forearm
Shadow Dexterous Hand. Tendon-driven. The cables running into the forearm are the tendons.
A Schunk SVH five-finger hand with fingers spread, photographed against a window
Schunk SVH. Nine drives move twenty joints, so several joints move together.
An Allegro four-finger hand fitted with soft tactile sensor pads on every finger segment
Allegro Hand fitted with uSkin touch sensors on each finger segment.

Four design decisions account for most of the differences between hands.

  • How many joints are driven. A fully actuated hand has a motor for every joint and can place each finger segment independently. An under-actuated hand has fewer motors than joints. One motor curls a whole finger, and the segments wrap around whatever they meet. That makes a secure grasp easy and precise finger placement impossible.
  • Where the motors are. Motors inside the fingers and palm make a self-contained hand that is thick and heavy at the fingertips. Motors in the forearm, pulling tendons, make slim fingers and a hand that needs a forearm, with cables that stretch and wear.
  • How force reaches the joint. Direct drive and low gearing let the motor feel what the finger touches, and let the finger give way on impact. High gearing and screw-type linear actuators are strong and compact, and are stiff, slow to respond and blind to contact.
  • The thumb. Most useful grasps need a thumb that can swing across to oppose the fingers. How many of the thumb’s joints are driven is the first thing to check on a specification sheet.

The fundamentals note explains why stiffness and compliance matter so much once a robot touches something.

Real options

Bars of independently driven joints and total moving joints for eight end effectors: parallel gripper 1; Inspire RH56 6 driven, 12 joints; Unitree Dex3-1 7; Schunk SVH 9 driven, 20 joints; Allegro Hand V4 16; LEAP Hand 16; Unitree Dex5-1 16 driven, 20 joints; Shadow Dexterous Hand 20 driven, 24 joints.
How much there is to control. Solid is what the policy can command. The outline is what moves.
HandFingersDriven / jointsDriveWeightTouch sensingNotes
Inspire RH5656 / 12Linear actuators and linkagesabout 0.54 kgOn some versionsCompact and widely fitted to humanoids
PSYONIC Ability Hand56Motors in the fingersabout 0.49 kgFingertip sensorsDesigned as a prosthesis; sold in a robotics edition
Unitree Dex3-137 / 7Motors in the handabout 0.71 kg33 sensors on the tactile versionThumb with 3 joints, two fingers with 2 each
Schunk SVH59 / 20Motors in the hand, coupled joints1.3 kgNone as standardIndustrial product, the size of a human hand
Allegro Hand V4416 / 16A motor at every jointabout 1.1 kgAdd-on skins availablePosition and current control; a long-standing research standard
LEAP Hand416 / 16Off-the-shelf servos at every jointnot confirmedNone as standardOpen design; about $2,000 in parts and four hours to assemble
Unitree Dex5-1516 / 20Motors in the handabout 1.0 kg94 force sensors on the tactile versionLifts about 3.5 kg palm-down
Shadow Dexterous Hand520 / 24Tendons from motors in the forearm4.3 kg with forearmTactile fingertips on thumb and indexThe closest to a human hand’s range of motion

These figures come from the makers’ pages, the LEAP Hand paper and published listings, checked in November 2026. Versions change and resellers disagree, so confirm against the current data sheet before buying. Humanoid makers also build their own hands. Tesla’s latest is reported to have 22 degrees of freedom with the actuators in the forearm, a figure that comes from press coverage and not from a published data sheet.

Weight against independently driven joints for six hands. Inspire RH56, Unitree Dex3-1, Unitree Dex5-1, Allegro Hand V4 and Schunk SVH lie between 0.5 and 1.3 kilograms. The Shadow Dexterous Hand, with 20 driven joints, weighs 4.3 kilograms including its forearm. A shaded band marks the weight of a human hand, around half a kilogram.
More driven joints cost weight. Only the six-motor hands approach the weight of a human hand, and weight at the end of an arm comes straight out of what the arm can carry.

Choosing one

  • Start from the task. Picking, placing and holding tools need a secure grasp, and six motors is enough. Turning, re-gripping and using fingers one at a time need independent joints.
  • Count the hours between repairs. A hand that fails every few hours cannot collect a dataset. Ask who can repair it, how long that takes, and what a finger costs.
  • Check the control interface. Some hands accept only position targets. Others also offer current or torque control, which is what allows gentle and force-aware grasping.
  • Check the weight against the arm. A 1 kg hand on an arm rated for 3 kg leaves 2 kg for the object.
  • Look for a simulation model. A hand with an accurate, maintained simulation model can be used for reinforcement learning and for testing. One without cannot.
  • Prefer what others use. Shared hardware means shared code, datasets and retargeting tools.

Two pieces of hand mechanics

Tendon coupling. In a tendon-driven or under-actuated hand, the joint angles qq and the tendon or motor displacements ll are related by a coupling matrix, and the same matrix relates tensions to joint torques:

Δl=P Δqτ=P⊤f\Delta l = P\, \Delta q \qquad\qquad \tau = P^{\top} f

If PP is square and invertible the hand is fully actuated and every joint can be placed independently. If it has fewer rows than columns, several joints share a motor. Their combined motion is then determined only once the finger meets something, which is why under-actuated fingers wrap around objects and why the policy needs to see measured joint angles.

When a grasp holds. A fingertip pressing on a surface can resist sliding only up to a limit set by friction. With normal force fnf_n, tangential force ftf_t and friction coefficient μ\mu:

∣ft∣≤μ fn| f_t | \le \mu\, f_n

The set of forces satisfying this is a cone around the surface normal, the friction cone, with half-angle arctan⁡μ\arctan \mu. A grasp is in force closure when the fingers, each pushing within its cone, can together resist a force or twist in any direction. For two fingers that means the line between the contacts must lie inside both cones. On a slippery object the cones are narrow and the fingers must oppose each other almost exactly. On a grippy one there is room for error. This is the geometry a learned grasp has to satisfy, whether or not the policy represents it explicitly, and it is why soft, high-friction fingertips make learning easier.

Connecting a hand to the action head

For the policy, a hand is a longer action. How that action is written decides how much data the policy needs and how well anything transfers.

Five ways to write one step of action. Arm and gripper: 8 numbers. Arm and hand as joint targets: 23. Arm and under-actuated hand, one number per motor: 13. Arm and a few blended hand shapes: 10. Arm and a 3D target for each of four fingertips: 19.
Five ways to write one step for a seven-joint arm with a four-finger, 16-joint hand, or with a six-motor hand in the third row.
RepresentationWhat the policy outputsStrengthWeakness
Joint targetsAn angle for every driven jointNothing is lost; the hand can do anything it is capable ofThe largest action, tied to one hand
Motor commandsOne number per motor of an under-actuated handSmall, and the mechanism handles the wrappingNo control of individual segments
Hand shapesA few numbers that blend learned whole-hand posturesVery small; easy to learn from little dataFine finger motion cannot be expressed
Fingertip targetsA 3D point for each fingertip, relative to the wristMeans the same on different hands; matches how contact is describedNeeds a solver; the solver’s choices are outside the policy

The hand-shapes idea rests on a finding about people. When human grasps are recorded and analysed, two or three underlying patterns account for most of the variation in hand posture. Hands mostly move in coordinated shapes, not joint by joint, so a handful of numbers reaches most everyday grasps. It does not reach the rest, which is exactly the fine in-hand motion that dexterous hands are bought for.

Action values to produce per second at 30 steps a second. One arm with a gripper: 8 numbers per step, 240 per second. One arm with a 16-joint hand: 23 and 690. Two arms with two 16-joint hands: 46 and 1,380. Two arms, two 20-joint hands, head and waist: 60 and 1,800.
How the action grows. A two-handed humanoid produces seven or eight times as many numbers per second as an arm with a gripper.

That growth has four practical consequences.

  • The kind of action head matters more. A head that writes actions as tokens pays for every number, so its output gets long and slow unless whole chunks are compressed into a few tokens. A flow-matching or diffusion head produces the whole vector in each refinement step, so a wider action costs it almost nothing in time. This is one reason those heads dominate work with hands.
  • Normalise each joint separately. Finger joints have very different ranges. The basics note shows what happens when values on different scales share one loss.
  • Arm and hand must be one action. Reaching and pre-shaping the fingers happen together. Predicting them in one chunk keeps them coordinated. Separate policies for arm and hand drift apart.
  • Pretraining on grippers helps only partly. Large robot datasets are gripper data. They teach a policy where to reach and what the task is. The finger motion has to come from hand data.

Under-actuated hands add one more wrinkle. The policy commands the motor, and the finger’s final shape depends on what it touches. The body state the policy sees should therefore include the measured joint positions where the hand reports them, not just the last command.

Retargeting: a human hand is not a robot hand

Dexterous data almost always starts with a person’s hand. The person’s fingers and the robot’s differ in length, joint count and thumb placement, so the motion has to be translated. That translation is called retargeting.

Three panels of a finger touching a contact point. A human finger with three joints reaches the point. A two-joint robot finger given the same joint angles ends 2.7 centimetres away. The same robot finger, solved to match the fingertip position, reaches the point with different joint angles.
Computed for a three-joint human finger and a two-joint robot finger of different lengths. Copying joint angles misses the contact point by 2.7 cm. Matching the fingertip reaches it with a different pose.

Copying joint angles keeps the hand’s shape and loses the contact. Matching fingertip positions keeps the contact, which is what manipulation depends on, and changes the shape. Most systems match fingertips, with extra terms to keep the thumb opposed to the fingers and to avoid the fingers colliding. No mapping is exact, and the leftover error is one reason hand policies need more data than gripper policies.

Collecting data for hands

MethodHow it worksStrengthLimit
TeleoperationA person’s hand is tracked by a glove or a headset and the robot hand mirrors itReal robot actions and observationsSlow; the operator feels nothing; needs the robot
Portable motion captureThe person wears tracking gloves and a camera and works with their own hands. DexCap is an example: it tracks wrist and fingers and records the scene in 3D.Fast, natural, collected anywhereThe data is of a human hand; it has to be retargeted, and the camera sees a hand, not a robot
Wearable exoskeletonThe person wears a frame that limits their hand to motions the robot hand can make. DexUMI is an example: it also replaces the human hand in the video with the robot hand.Closes both the motion gap and the visual gap; the person feels the contactOne exoskeleton design for each robot hand
Reinforcement learning in simulationThe policy learns by trial and error on a simulated handNo human data; has produced skilled in-hand rotationContact is hard to simulate; transfer needs care

The trend is away from teleoperating the robot and toward capturing people directly, because a person using their own hands works much faster and feels what they are doing.

Using touch with a hand

Fingers cover the object they hold, so the camera loses sight of it exactly when it matters. Touch fills that gap, and a hand has room for a lot of it.

Sensor typeHow it worksReportsTrade-off
Camera-basedA small camera watches a soft gel skin from insideA detailed image of the contact: shape, texture, shearThick, so mostly fingertips; too slow for vibration
Magnetic skinMagnets in soft rubber move over magnetic field sensorsForce in three directions at each point, at high ratesDisturbed by nearby magnets and motors; needs calibration
Pressure arraysA grid of cells whose resistance or capacitance changes when pressedWhere contact is and roughly how hardLittle about sideways force; coarse
Motor currentThe effort each finger motor is makingWhole-finger contact and load, with no added sensorOnly on hands with low gearing; no location

The photo of the Allegro hand above shows the second kind: thin skins on every finger segment, each reporting force in three directions.

Two loops. In the slow loop, cameras and fingertip sensors feed a learned policy, which outputs a chunk of finger targets 10 to 30 times a second. In the fast loop, the fingertip sensors feed a slip or overload check that adjusts grip force hundreds of times a second, correcting what the policy asked for before it reaches the motors.
Touch is used twice. The policy reads it as tokens at its own rate. A simple reflex underneath reads it far faster and corrects the grip between the policy’s decisions.
  • As policy input. Each sensor patch is encoded into one or a few tokens, usually one group per finger, and joins the camera and body-state tokens. The policy then learns things a camera cannot supply: whether the object is seated, which way it is tilting in the grasp, when a peg has touched down.
  • As a reflex. A slip must be answered within tens of milliseconds, faster than a policy that plans in chunks. A small rule beneath the policy, tighten when vibration or sideways force rises, loosen when force exceeds a limit, handles this at the rate of the sensor. The policy sets the intent and the reflex keeps the grip.
  • During data collection. Feeding touch back to the operator, through vibration in a glove or directly through an exoskeleton, produces gentler and more consistent demonstrations.
  • As a success signal. Contact on the expected fingers at the expected time is a cheap check that a grasp happened.

Two cautions carry over from the note on senses, which shows what these signals look like and how they are turned into tokens. A policy trained with cameras and touch tends to lean on the cameras, so drop the camera input at random during training and then test with the touch input removed. And sensors of the same model differ from one another and wear, so calibrate each one and expect to replace skins.

How hand policies are trained

  • Imitation with a generative action head. The mainstream route: demonstrations from teleoperation or wearables, and a flow-matching or diffusion head that outputs arm and hand together in chunks.
  • Reinforcement learning in simulation. The route behind the best in-hand rotation results. Thousands of simulated hands practise with randomised physics, and the policy is then run on the real hand.
  • Pretraining on human video. The only data source large enough to cover the variety of things hands do. It supplies what to do and where to touch, and robot data supplies how.

These are combined in practice: video for breadth, demonstrations for the task, and practice on the robot to make it reliable. The notes on reinforcement learning and robot data cover each part.

Open questions

  • Is a five-finger hand needed? Many useful tasks need two or three fingers. A human-like hand fits human tools and human video. It is not clear how much of the benefit comes from which.
  • How should the action be written? Joint angles, a handful of hand shapes, or fingertip positions. Smaller choices learn faster and can do less, and no one representation has won.
  • Can human video carry the load? It is the only source large enough. How much robot data is needed on top is unsettled.
  • How much does touch add? Results are strong on specific contact-rich tasks and mixed in general, partly because there is no large shared touch dataset.
  • How is success measured? There are few shared benchmarks, and results on different hands are hard to compare.

Common mistakes

  • Choosing the most capable hand. A simpler hand that survives a day of data collection teaches more.
  • Copying joint angles. Check where the fingertips end up, not whether the pose looks right.
  • Treating the hand as a separate policy. Arm and fingers move together and should be predicted together.
  • Adding touch and assuming it is used. Test the policy with the touch input removed. If nothing changes, it was ignored.

Terms

Degrees of freedom
The number of independent ways a mechanism can move.
Under-actuated
Having fewer motors than joints, so some joints move together.
Tendon drive
Moving a joint with a cable pulled by a motor located elsewhere, usually in the forearm.
Retargeting
Converting a human hand’s motion into the motion of a robot hand.
Hand synergy
A coordinated whole-hand posture; a few of them, blended, reproduce most grasps.
In-hand manipulation
Moving an object within the hand without putting it down.
Friction cone
The set of contact forces a fingertip can apply without slipping.
Force closure
A grasp that can resist a force or twist in any direction.

Further reading