8
$\begingroup$

Although simulation models like Boids give good results for bird flocks or fish shoals on a small scale, simulating every single member in real time becomes unrealistic for huge numbers. Is there a way I can model a flock in the distance where only the density of birds is visible? I'd like to have that flowing, changing density gradient with a much smaller number of variables to process.

I've tried using a much smaller population and displaying each boid as a blurred area with Gaussian density so that as they overlap the density rises and falls through their interaction. This is reasonably cheap but it never leads to sharp changes in density, either spatially or temporally, which makes it look too uniform.

Is there any other way of getting away with a much smaller number of individuals? Or is the only way to get realistic results to prerender?

$\endgroup$
  • $\begingroup$ It would be great if you had reference footage for the effect you want. Say, something like this? youtube.com/watch?v=XH-groCeKbE $\endgroup$ – John Calsbeek Aug 22 '15 at 15:20
  • $\begingroup$ @JohnCalsbeek yes that would make it easier to get across what I want. In the video you linked to the individual birds are discernable (just). I'm looking to render a flock a little more distant so that individuals are not visible, but the variations in density are still consistent and realistic. $\endgroup$ – trichoplax Aug 22 '15 at 15:24
4
$\begingroup$

One way that might work is to have boids but render each boid as a group of birds. This way the simulation is still simple but the rendering makes it seem complex. You might even allow a little drift per bird in the boid to make it look so obviously rigid.

$\endgroup$
4
$\begingroup$

I'm not sure if this counts, but you can use fluid dynamics to render large crowds (of birds, people, etc). With SPH (smoothed particle hydrodynamics) to model the "fluid", you aren't really describing the motion of each bird per-se, since you can sample the crowd of birds with a representative set, and then draw birds around your samples. Take a look at this paper that compares the different methods:

http://gamma.cs.unc.edu/lookahead/golas-2013-hybridcrowd.pdf

$\endgroup$
  • 1
    $\begingroup$ This is actually a pretty good method. But is it less expensive? $\endgroup$ – joojaa Aug 22 '15 at 12:28
  • $\begingroup$ It's as expensive as you want your simulation to be. If it's too expensive, use fewer samples. $\endgroup$ – Mokosha Aug 22 '15 at 16:02
  • $\begingroup$ link don't work $\endgroup$ – Drakonoved Mar 4 '18 at 8:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.