Robots — Winning at Chess but Unable to Pick Up a Cup

Difficulty is inverted

We think of arithmetic as hard and walking as easy, because school made us learn the first and nobody taught us the second.

For a machine it is the other way round. Calculation has clear rules that can simply be written down. Walking and gripping have rules written nowhere.

Walking is falling on purpose

Standing on two feet is precarious: small soles, tall body. Even standing still you are tipping slightly all the time and correcting without noticing.

Walking goes further. You fall forward deliberately and put a foot out before you land, then do it again with the other foot.

The orbit article described a satellite as falling continuously and missing continuously. Walking is close: falling continuously and getting a foot down each time.

The last push comes from the big toe

Look closely at one step and there is a distinct final moment: just before the foot leaves the ground, the toes press down and send the body forward. The big toe does that.

It can only do it because it lies in line with the others. A chimpanzee's is different — set off to the side like our thumb. Far better for gripping a branch, and useless for pushing the ground away behind you. That is why a chimpanzee walking upright rolls from side to side.

One foot holds twenty-six bones and more than thirty joints, arched in the middle so that it compresses on landing and springs back on release — storing a little force and returning it.

Knowing that, a robot foot looks poor. Most are a single flat plate. With no joints there is no final push from the toes, so the whole sole is lifted and set down as one. That is why robot walking rolls, runs slow and spends so much energy — and why some current work adds a joint at the toe.

And gripping

Picking up a cup means judging force. Too little and it drops; too much and it breaks. We hold an egg and a hammer with the same hand and adjust without thinking.

And it needs friction. A slippery cup must be gripped much harder, and how slippery it is can only be found by touching. No calculation gives it in advance.

A little further in

So you measure and correct, continuously

Calculating it all in advance does not work, because the world does not cooperate: the floor tilts a little, the cup is heavier than expected, the hand slips.

So the method is feedback: measure the present state, see how far it is from the target, correct by that much, and repeat very fast. A walking robot is doing this hundreds of times a second.

Why factory robots manage

The arm in a car plant is faster and more accurate than any person. Same machinery — why does it succeed?

Because its world is fixed. The part always arrives in the same place facing the same way, so little feedback is needed. Following instructions suffices.

You cannot make what you cannot measure

The institute's other axis is precision machining — cutting and finishing to very fine tolerances.

Here it meets the article on units. Holding a part to a fraction of a hair's width requires measuring to that fineness first. Measurement accuracy is the ceiling on manufacturing accuracy — which is why keeping standards sits underneath industry.

What you would study to do this

One robot gathers several trainings.

1. Someone who designs the joints and frame — mechanical engineering.
2. Someone who keeps it from falling over — control engineering.
3. Someone who builds its eyes and skin — electronics and sensors.
4. Someone who decides what it should attend to — computing and artificial intelligence.
5. Someone who finds a body both light and strong — materials engineering.
6. Someone who machines the drawing into real parts — the shop floor.

That last line matters as always: without someone cutting the parts, a robot exists only on screen. And this field has an unusually large share of knowledge that lives in the hands. Why a joint rattles is often answered faster by feel than by calculation.

The work is done at the Korea Institute of Machinery and Materials in Yuseong-gu, Daejeon.

The question that remainsIf the easiest-feeling tasks are the most practised ones — what is the hardest thing you do without thinking about it at all?