[ODE] Convex Volume Geom

[Erwin Coumans]

The weaker performance of Bullet+ODE is not due to GJK but because of the 
brute-force penetration depth algorithm. There is a switch in the code that 
allows you to enable Solid EPA penetration depth. This makes the penetration 
case much faster. Unfortunately I haven't had the time to write EPA yet.

Hopefully I will have time for this in 2006, and provide you with some 
comparisons. Erwin




----- Original Message ----- 
From: "Jaroslav Sinecky" <jsinecky at tiscali.cz>
To: <ode at erwincoumans.com>; "ode" <ODE at q12.org>
Sent: Thursday, December 22, 2005 7:43 AM
Subject: RE: [ODE] Convex Volume Geom


> Hi Erwin,
> did you (or someone else) tested the performance of your GJK based 
> collision
> detection in comparision to ODE geoms?
>
> I tried your ODE+Bullet because I wanted to use plane-capped cylinders but
> realized that collisions like cylinder-box or cylinder-sphere (or I think
> any other) were MUCH slower then ODE's specific geoms code.
> I didn't investigate much about this so it could also be some problem on 
> my
> side. I just went to use dCylinder2 contrib and I'm quite happy now.
>
> Jaroslav
>
>
>> -----Original Message-----
>> From: ode-bounces at q12.org [mailto:ode-bounces at q12.org]On Behalf
>> Of ode at erwincoumans.com
>> Sent: Wednesday, December 21, 2005 11:31 PM
>> To: ode
>> Subject: Re: [ODE] Convex Volume Geom
>>
>>
>>
>> Rodrigo Hernandez writes:
>> >
>> > No, I want to stay clear from GJK (any implementation), due to its
>> > numerical instability,
>> > perhaps that's an option for the original poster though.
>>
>> For games GJK is a good solution in my experience. Did you know that even
>> commercial engines succesfully use GJK for their convex collision
>> detection?
>>
>> Erwin
>>
>>
>> Rodrigo Hernandez writes:
>>
>> >
>> > No, I want to stay clear from GJK (any implementation), due to its
>> > numerical instability,
>> > perhaps that's an option for the original poster though.
>> >
>>
>>
>> _______________________________________________
>> ODE mailing list
>> ODE at q12.org
>> http://q12.org/mailman/listinfo/ode
>
>