2D Collison Response (Guendelman vs. Catto)

Post Reply
crashlander
Posts: 41
Joined: Sat Apr 08, 2006 11:20 am

2D Collison Response (Guendelman vs. Catto)

Post by crashlander »

I've implemented a 2D rigid body system that supports convex polygon collision detection.

Its collision detection uses a Distance Grid similar to that discussed in the 'Nonconvex Rigid Bodies with Stacking' paper by Guendelman

For the collision response, I've implemented Erin Cattos Impulse based system described in his Box2D demo.

Everything seems to work pretty good. I'm curious though the differences and trade offs between using the Box2D impulse system vs. using the system described in the Guendelman paper where the response is broken in to Collision Response and Resting Contact.

Just thought if there was anyone familiar with both techniques that could do a little quick compare and contrast on the techniques. I'm trying to decide which is better from both a stabilty stand-point as well as performance stand point.

Thanks.
Dirk Gregorius
Posts: 861
Joined: Sun Jul 03, 2005 4:06 pm
Location: Kirkland, WA

Post by Dirk Gregorius »

Did you extend the Guendelman paper to deal also with joints? From my experience this was not trivial and it didn't work to well for all scenarios. See the work of R. Weinstein and how it integrates into Guendelman contact processing. The Guendelman approach has also much worse performance than Erin's work. On the other side you have maybe a little less penetration and nicer collision. It really depends on what you want to do...
crashlander
Posts: 41
Joined: Sat Apr 08, 2006 11:20 am

Post by crashlander »

No, I have not implemented joints (from either papers) yet. I'm mostly interested in using my engine for games, so if Erin's collision response technique is more performant, then I'll probably just stick with it.

I will most likely use the Joint system from Box2D as well.
Post Reply