Sign in to follow this  

Swept Bounding Circles - first collision

This topic is 4401 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Hi, I finished a sweep and prune function for my collision code, but I'm stuck at one point. My code returns a list of all the overlapping bounding circles -eg. 1-3, 2-1, 3-4, 2-4 But these are using swept circles and each circle can only hit one other circle. -eg. if 1-3, and 1-4, overlap, I must find whether 1 hits 3 first or whether 1 hits 4 first. Does anyone have a link to some article that discusses this point? When I get lots of objects moving, the interdependencies get really confusing. Thanks a lot -Cuppo

Share this post


Link to post
Share on other sites
This thread contains a somewhat documented algorithm for finding the time of first collision between two spheres. It assumes linear movement between the start position and the end position.

Share this post


Link to post
Share on other sites
Oops, forgot to mention what I actually had trouble on.

Thanks for the code ToohrVyk, I'll take some time to crawl through that. (Objective Caml !!??)

My real problem though, is in the broad-phase detection.

such as:

If the bounding boxes of :
1 overlaps 2 and
1 overlaps 3

say 1 hits 2 before 1 hits 3. You may think aha, therefore 1 collides with 2 and not with 3.

However: what if something else (say 4) hits 2 before 1 hits 2? Then the actual collisions are:
4 hits 2,
and 1 hits 3

I'm looking for a fast way to prune all these checks.

Has anyone ever read the "Kinetic Sweep and Prune" paper? Does it talk about what I'm looking for? I can't understand it right now, but if it does, I'll try and read it again.

Thanks
-Cuppo

Share this post


Link to post
Share on other sites
Sign in to follow this