[ODE] Swept collisions?

[nlin at nlin.net]

On Wed, Oct 01, 2003 at 08:37:36PM +0000, Roel van Dijk wrote:
> If you could come up with a way to do swept collisions you would also have to 
> change the inferface. Now you use the dWorldStep function to 'advance' the 
> simulation by some time interval. And collisions can only occur at those 
> intervals and that causes tiny sphere's to fly through thin walls at high 
> speed. But if you would do swept collisions a collision could occur any time 
> between the intervals. How would you handle that? 

Currently the collision callback is called if two objects currently
intersect. How about extending this callback logic such that a collision
callback is called also if the objects' swept volumes intersect, plus
the time of the first collision? This way, the user could decide what
(if anything) to do with the "missed" collision. I think this could also be
done without breaking the current collision infrastructure, since we're just
talking about adding new callbacks, not changing existing ones.

-Norman