[ODE] Pushing through collisions

[Andrew Aye]

Cases like that are impossible to correctly fix in the general case without
using further information.  A penetration system that used the objects
velocities could for instance potentially "weight" the axis seperation tests
to have it tend to give axis that are as close to the negative of the
direction of movement as possible.  Havent really looked into it much -
finshing off some cylinder code myself -)  Without context and with high
speeds, or "high" rotations you'll always be able to find incorrect
seperation determinations ( ie your radial response was correctly the
smaller seperation ).  I would be wary of hacking in fixes to try to correct
for this - as crystal-ball methodolgies tend to break down at the worst
times -)

 - Andrew Aye

-----Original Message-----
From: Nguyen Binh [mailto:ngbinh at glassegg.com]
Sent: Thursday, April 29, 2004 4:24 AM
To: Alen Ladavac
Cc: ode
Subject: Re[4]: [ODE] Pushing through collisions


Hi Alen,

AL> 5) The cylinder-sphere combination in our collider. Nguyen, I belive
that we
AL> have found a case where contact normal is NAN if sphere center is on the
AL> cylinder axis. I'll send you a fix when we have time to do it.

    Possibly not! The small sphere will be stuck in the center of
    large cylinder when its center is deep inside: i.e sphere's center
    is between cylinder high-base and low-base. In this case, your
    code assumes that it's a side penetration, but it's actually a
    very deep bottom penetration.

    For the case sphere center is on the cylinder axis, I didn't
    notice it. I had successful made a large stable stack of cylinders and
    spheres. And I'm sure all sphere center is in cylinders axis.

    Nguyen Binh.

    I'm very very interested in another iterative solvers but
    I'll finish porting Croteam's cylinder-vs anything to ODE before
    reading more about it. Before, I had tried but without any
    success :( . So please keep your iterative solver progress
    updated. :)

    

_______________________________________________
ODE mailing list
ODE at q12.org
http://q12.org/mailman/listinfo/ode