PDA

View Full Version : Genetic Algorithms for animation


jHromika
07-17-2003, 05:51 PM
The process starts with random walkers, none of which can walk properly. The best ones (those that make at least one step without falling over) are allowed to produce offspring, which are again selected according to how far they walk. This selection is repeated over a number of generations. At the end of process the biped can walk without falling over.

http://www.naturalmotion.com/pages/technology_hiw.htm

The link is to a related web site, but the original article I read was in Discover Vol. 24 No.8 (August 2003). This algorithm is incorporated into an animation package called Endorphin. Interesting read, although I'm really not sure what to think of it :shrug:

Thalaxis
07-17-2003, 07:48 PM
Originally posted by jHromika
http://www.naturalmotion.com/pages/technology_hiw.htm

The link is to a related web site, but the original article I read was in Discover Vol. 24 No.8 (August 2003). This algorithm is incorporated into an animation package called Endorphin. Interesting read, although I'm really not sure what to think of it :shrug:

I remember studying genetic algorithms in school (we even had a
talk by John Holland, who came up with them, explaining his first
experiments with GA's), in which some researchers used GA's to
get a sigma-shaped entity to learn to jump. (They used a
Connection Machine, by Thinking Machines, subsequently acquired
by Sun.) What it came up with was clever, even though it was in
two-D space, and only jumped vertically. The GA learned how to
squat down, extend itself, and then pull its lower segments up,
maximizing the height of its center of gravity.

Neat stuff... but it's even cooler to see that someone's putting it
to good use :)

CGTalk Moderation
01-15-2006, 04:00 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.