[ODE] TR: Cylinder / Sphere

[LR]

Hi Bram,

I've made more test with the cylinders,

1] It's possible the contact distance problem was a visual effect ... I'm
not sure but it's possible.
But, The force of the contact is really too heavy. As you suggested, i
watched the _DEPTH_RECOVER_METHOD_ : its set to 0

2] I made some tests:


1) Cylinder Fall on a box:

   ---
   | |
   | |
   | |
   | |
   ---
----------
-        -
----------

Contactpoints : 10 for the cylinder (!!!) , 1 for the capsule
Visual Effect : OK, cylinder may takes 4 or 5 time step to stabilise
(ts=0.01s), 1 or 2 for the capsule ...

2) Cylinder Fall on a sphere:

   ---
   | |
   | |
   | |
   | |
   ---
    O
  O   O
  O   O
    O

Contactpoints : 1 for the cylinder, 1 for the capsule
Visual Effect : OK,

2) Cylinder Fall on a sphere (there is a problem ...):

|--------------|
|              |
|--------------|

    O
  O   O
  O   O
    O

Contactpoints : 1 for the cylinder, BUT the CONTACTPOINT IS INSIDE THE
SPHERE (NEAR SPHERE CENTER )
Visual Effect : The cylinder bounce with a hard force ...



2) Sphere Falls on a cylinder:


    O
  O   O
  O   O
    O
|--------------|
|              |
|--------------|

Contactpoints : 1 , ALL IS OK 
Visual Effect : ALL IS OK 

It seems that the speed of the cylinder impacts the location of the contact
point.

I've tested with the capsule and all is OK. An idea ?

Laurent.



-----Message d'origine-----
De : Bram Stolk [mailto:bram at sara.nl] 
Envoyé : jeudi 23 mars 2006 23:18
À : LR
Cc : ode at q12.org
Objet : Re: [ODE] TR: Cylinder / Sphere

LR wrote:
>  
> Hi,
>  
> I have a little problem with the Cylinder:
>  
> My test is a cylinder that falls on a phere :
> Problem 1: It seems the contact detection start finding some contact 
> when the cylinder bounding sphere tutch sphere bounding sphere (it's 
> that i believe)

You see false positives on the intersection test?

Is it with end-caps near the sphere, or with the cyl body near the sphere?

Do you have a test program?

> Problem 2 : the cylinder is projected with a force witch seems to be 
> 2* too large

Hmm... there is a mechanism of 'depth exaggeration' in
collision_cylinder_sphere.cpp, but that is disabled by default.
Did you redefine _DEPTH_RECOVER_METHOD_ ?

> If i change the cyclinder for a capsule, there is no problem if i 
> change the sphere for a box (i keep cylinder) , there is no problem 
> too
>  
> some one have an idea ?
>  
> Laurent.
>  
>  
>  
> 
> 
> ----------------------------------------------------------------------
> --
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode