Evolutionary Theory in Computer Animation

By Deane Barker on July 24, 2003

Emerging Technology: Darwin in a Box: Here a good piece on artificial intelligence. It details an attempt to give an animated stick figure a set of “rules” (gravity works like this, you have joints here, your center of gravity must stay above a certain height, etc.), and then trying to get the figure to learn to walk across the screen under the “weight” of those rules. It took some evolution:

The algorithm generated 100 animated characters, each with a randomly assembled neural network controlling its muscles. Then the algorithm let them all try walking. Predictably enough, the first generation was almost completely inept. But a few figures were slightly better than the rest — they took one hesitant step before crumbling to the ground. By the standards of the fitness function, they became the winners of round one. The software made 20 copies of their neural networks, introduced subtle mutations in each of them, added 80 new participants with randomly wired networks, and started the next generation walking.

Interestingly, they had to include the rule about the center of gravity staying above a certain height because some figures learned to crawl and tumble across the screen, rather than walk.



Comments are closed. If you have something you really want to say, tweet @gadgetopia.