[ODE] ODE collision qustion

Jon Watte hplus-ode at mindcontrol.org
Thu Jun 10 14:44:39 MST 2004


> Can not there be an universal algorithm as for example "Beziere curved 
> mesh" colision. You could describe box, mesh,sphere,capped cylinder with 
> it. And maybe convertion function for each primitive..So you reduce 
> number of algorithms from N*N/2 to N+1.

That wouldn't be very efficient. I believe the state of the art in 
collision is to require N*(N-1)/2 algorithms. Most libraries you pay 
money for (even the very large ones) seem to do this. The trick is to 
keep N small.

An alternative is to define subsets, where you can say that "I guarantee 
that X, Y and Z always work together" and add colliders only for those 
subsets.

Cheers,

			/ h+




More information about the ODE mailing list