Have you ever watched a large flock of starlings? It’s pretty amazing, isn’t it. It’s been known for quite awhile that you can reproduce this sort of complex behavior by giving each bird a simple set of rules which only depend on its immediate neighbors. The high level patterns emerge from these simple rules. Brian Hayes has an interesting article about this in the upcoming issue of American Scientist. It’s about a research project called STARFLAG. They set out to determine whether these local rules are used in real starling flocks. They did this by tracking the 3D motion of individual birds using a pair of video cameras. It’s a really nice piece of work. Their results are that the boids algorithm only required some minor tweaks to fit the measured data.
BTW, the starling demos I posted earlier (link1, link2) don’t use this algorithm. They’re producing a somewhat similar looking behavior using a single attractive force and inertia. If you’re interested in how this sort of thing works, you should go check out the applets on Craig Reynold’s website.