Robot hand · hardcore

What it does.

A hand you can bolt to a robot arm or stand on a desk and drive from Python. Built to show grip, gentleness and speed in one short demo.

GripsPower grip, pinch, point, open, and rock, paper and scissors poses. Fingers wrap around odd shapes on their own.
Feels contactIt reads force from motor current, slows down just before it touches, then holds a set force. That's how the egg survives.
Holds without cookingOnce it has a grip, it backs the current off until the gears hold the load, so long holds stay cool.
Talks to your codeUSB or CAN-FD, a simple command protocol and a Python library. A rock-paper-scissors demo ships with it.
Mounts on an armAn adapter plate fits standard cobot flanges (ISO 9409-1-50-4-M6, UR3e/UR5e class).

5 / 6fingers / motors

~490 ghand + forearm (CAD estimate)

200 mmforearm, 68 x 56 mm section

≤ 0.7 sopen to closed (target)

≥ 5 Npinch (target; 6.5 N design point)

1 kgbottle hold (target)

1 kHzposition and current loops

~5.4 h+agent time so far; routing in progress

~$596parts per hand, estimate incl. spares

Targets and design numbers. Nothing here has been measured on hardware yet.

The trade-offs.

Hands are one of the hardest packaging problems in robotics. Motors, tendons, sensing, drivers and firmware all fight for the same space. Here's how we settled each fight.

Small geared motors vs custom brushless

Brushless motors give finer force control, but need more electronics and firmware, and more room.

Call: small geared motors with built-in encoders. They fit the weight budget with about 300 g to spare, and gear friction lets the hand hold a bottle at almost no current. Brushless stays on the list for v2.

Gear ratio vs speed vs pinch force

The limit isn't the motor, it's what the gearbox can take. Ratio and spool size trade closing speed against pinch force.

Call: one ratio for all five finger motors that meets both the 0.7 s close and the 5 N pinch, and a slower, stiffer one for thumb rotation, which needs to hold still more than move fast.

Motors in the palm vs the forearm

Motors in the palm make a self-contained hand. In the forearm, they keep weight and heat away from what you're holding.

Call: forearm. About 100 g less ahead of the wrist, simpler wiring, and seven tendons through the wrist in low-friction tubes.

Board heat vs forearm space vs long grips

The board runs cool. The motors are the real heat problem during a long hold.

Call: fix it in firmware. After gripping, current ramps down until the encoder sees a tiny slip, then holds just above that. A per-motor heat model limits hard pinches to safe bursts.

Finger joints vs shapes it can grab

A finger needs to wrap a bottle and also stay stiff for a fingertip pinch. With every joint free, it wraps, but the tip buckles in a pinch.

Call: the base joint is free and the last two are linked, so the finger wraps a bottle by itself and stays stiff for a pinch.

Where it stands.

Designed. First build in progress.

Design: Printable CAD for every part with fits and an assembly order, the forearm board with fab files, firmware, a Python library, a parts list checked against live stock, and a log of design mistakes we caught and fixed.
Checks done: No part collisions in five poses. Schematic checks clean; 10 board connections are still being routed by hand before fab. Host library and protocol tests pass.
Not yet known: Real grip force, motor temperature in a long hold, tendon stretch over time, and gearbox life at full pinch. These need the real thing.
Next: Nothing is built yet. The last few board connections are being finished by hand. Then boards and printed parts get ordered, and the first hand gets assembled and tested. We'll post the results, good or bad.

A five-finger robot hand for your robot.