[ODE] contact points

Rodrigo Hernandez kwizatz at aeongames.com
Fri Apr 13 08:47:05 MST 2007

Personally I am not a fan of GJK, I studied it and even implemented it 
once, and found it to be too unstable and too cpu intensive
to be of much use, but apparently Erwin managed to come up with a good 
(at least better than mine) implementation of it in Bullet.

So I say, stick with SAT, since Bullet would provide GJK anyway.


Lewis Foster wrote:
> Jaroslav, in that case then something is wrong. I'll have to try and 
> find whats causing the problems. Thanks.
> Rodrigo, Yea i'm working on the code you wrote. I see what you mean 
> about the box example, something more needs to be done with the planes 
> of each convex that are most likely in contact. At the minute contact 
> points can only be on vertices so something needs to be done there. 
> I've done edge-edge collision but again as contact points can only be 
> present on vertices it's not going to work very well.
> I'm tempted to look more closely at GJK, perhaps combined with SAT but 
> it depends on time really. I suppose i should see where this technique 
> can go before trying another.
> Thanks for the help everyone,
> Lewis
> ------------------------------------------------------------------------
> _______________________________________________
> ODE mailing list
> ODE at ode.org
> http://mooshika.org/mailman/listinfo/ode

More information about the ODE mailing list