[ODE] contact joints / attaching them

Nate W coding at natew.com
Sat Oct 11 14:29:29 MST 2003


On Sat, 11 Oct 2003, gl wrote:

> So let's say you run collision detection on a fast moving object twice
> during each physics step, as was suggested.  This means that when a
> collision is detected (the good part), it will penetrate less than all other
> objects in the simulation, given the same parameters - and thus the results
> will be a little different.  At least that's how it seems to me.  It may not
> make much difference - but then again it may.

I see what you're getting at now.  I still think it won't be a problem
though...

If I recall correctly, the forces applied to the object on the next
timestep are a function of penetration depth and the duration of the
previous timestep - ODE applies enough force to completely resolve the
penetration (perhaps scaled by ERP?) during the next timestep, with the
assumption that the next timestep will be the same duration as the
previous timestep.

A problem scenario might be a fast-moving bullet passing through a thin
wall... the tunneling is detected and the bullet is stepped with a shorter
time-step to make it penetrate...  If at the next dynamics step the world
world (bullet and everything) is stepped using the 'regular' timestep
interval, rather than the cleverly-shortened timestep interval, the
penetration should resolve properly.

At least, I think so.  :-)

-- 

Nate Waddoups
Redmond WA USA
http://www.natew.com/




More information about the ODE mailing list