new collision detection algorithm


New algorithm for collision detection cuts cost and processing time. Movie will be shown at Siggraph.




Can’t wait for it to implemented. :bounce:


This will improve physics in games and animations 10000 folds.


That chair sim is crazy. It’s almost comedy how much shit they threw at it…


Wow! This stuff looks ripe for code implementation. I’m not a coder but I connect with some of the best out there. I will see what they think of this. This is cool stuff.


Wow that’s very impressive.

Though collision is an active research project so it’s not like this comes out of the blue. Last year Stanford had a similar paper though with rigid objects:

Ron Fedkiw page

This thing is actually part of a SIGGRAPH papers session, so it’ll be interesting to see the video demonstrations.

Dynamics & Modeling SIGGRAPH 2004 Papers session


Oh man that is sooooo sweet. I’m still picking my jaw up off the floor. The next Havoc engine needs to implement this. mycodesoft wasn’t exajerating, this is going to change video game physics forever.


Awe inspiring animations.


Well I don’t think it’s exactly perfect for games just yet. On first glance, it’s fast, but it will increase the file size for a piece of geometry by quite an order of magnitude depending on how the data is stored. Storage is still a premium in console and PC games, so game developers will get nice collisions at the expense of pretty much the quality of all the other assets in their game.


thats amazing, lots of work has finally paid off, will be interesting when its implemented into games and such




There is also going to be a course about collision detection algorithms at SIGGRAPH. It most probably won’t cover this new stuff but it still should cover the state of the art. You’ll notice a few guys from animation and game studios, so it’s bing used out there:
Collision Detection and Proximity Queries


Couple of points to note…

  1. This only deals with deformable objects. And only makes it nearly as fast as rigid body deformations

  2. The memory overhead’s gotta be pretty high

  3. There’s a big difference between someone presenting something that “looks cool” at Siggraph, and it actually being useful in production…


This looks really great, I would like to see more examples of how it can be used for cloth, but I guess it’s going to take a while for it to be implemented in the standard cloth-solutions out there. If it’s really that fast then clothsolving should be superfast and animated hair could benefit too.



there is a small article about this in the newest issue of wired. apparently the whole thing (if i remember the article correctly) only took 16 hours to do all of the calculations! to me that is more impressive than the simple scale of the animation itself.


WOW! Wish blender could do that :stuck_out_tongue:


Anyone remember a program called “REAL 3D” back in the mid 90’s? It had a collision detection system in the modeler. As you built objects they would “drop” to the floor. If you turned the wheels on a car model the car would turn. I know most packages have these options in animation if you want to set them all up, but what about a global modeling option so you could “toss a chair into a scene and “mess it up” a bit. Kind Cool? What do Y’all think?

Cheers, THOM


Remember it? I did the training videos for that program back in the day. It still has some cool features that Maya still doesn’t today. Ahhh… memories…


Yup! Wired isn’t online yet, so I googled the name of the guy behind it all and got the official site. That’s where I originally saw it though. :slight_smile:


ACK! I want!


well I beleive u can achiveve the same thing in houdini.
but it requires some cleaver nodes and smart coding.
Its a great concept!