AI EncyclopediaRobots

Simulation

시뮬레이션

Building a space inside a computer that imitates reality, so you can try things there before trying them for real.

Overview

Practicing with a real robot takes a long time, and a robot that falls can break. So people build a virtual space inside a computer that even imitates gravity and friction, and have many virtual robots practice walking or picking up objects there all at once. Practice that would take years in reality can be finished in much less time.

Self-driving cars also face dangerous situations first on virtual roads. The problem is that virtual and real are not the same. A robot that did well in the virtual world often slips on a real floor. Narrowing this gap so that what was learned in the virtual world carries over to reality is called sim-to-real.

An everyday example

It is like a pilot practicing many times in a flight simulator before flying a real plane. If you crash, you just start over.

Why it matters

It is a big reason robots have clearly become better at walking and running in just a few years. It is also why robot companies team up with chip companies that make virtual training tools.

See it on LastAI

These concept articles were written by AI and checked again for accuracy before publishing. The English edition is translated from the Korean articles. They leave out things that change quickly, such as prices and rankings. Tell us if something is wrong. Report an errorLast updated

You