[ODE] - Re: Once more the trimesh vs sphere collision problem - Bayesian Filter detected spam

[Jon Watte (ODE)]

Johan Öberg wrote:

> So I tried to just keep the contact with the deepest penetration and it 
> seems to work! Right now it's just a hack which just checks contact 
> depth and doesn't care about if the contacts are conincident. I guess 
> I'll have to try to sort them somehow, to tell what contact points to 
> throw out and which ones to keep.
>  
> How would I go about doing this? Would using a dot-product between two 
> normals to decide if they are coincident be a good idea, or is there a 


Coincidence has to do with position, not normal. Thus, you want to merge 
two contacts with different normals, but the same (or close) contact 
points. The best way to merge contacts is to treat the contact points as 
"vertices" and do "vertex welding" with some large epsilon (say, 0.05 
meters or so).

Cheers,

			/ h+