[ODE] How does everyone resolve world collision?

[Mohsin Hasan]

I have been able to load a 12.5 million polygon plant model as trimesh with
ODE and it works rock solid without any considerable loss in performance. We
started out loading 1 big giant trimesh for the whole data. Performance drop
was so minimal that we haven't changed anything yet. Our camera is basically
a vertically placed capped cylinder and it works almost flawlessly, even
climbs the stairs without much issues. 

Mohsin

-----Original Message-----
From: ode-bounces at q12.org [mailto:ode-bounces at q12.org] On Behalf Of Adam D.
Moss
Sent: Saturday, June 25, 2005 3:40 PM
To: ode at q12.org
Subject: Re: [ODE] How does everyone resolve world collision?

Vast wrote:
> I know I use one _HUGE_ trimesh to represent my 3d world, but what about 
> everyone else? What are other possibilities, and what did people do to 
> resolve the collision detection with the world before trimesh collision 
> was checked in into ODE?

I've been using ODE without trimeshes for a few years.  What I do
is place (in the modelling program, in separate layer) a big static
hashspaced soup of boxes and spheres around the place to approximate
the world geometry mesh for collision.

While it's straightforward to get reasonable results this way,
it's also extremely boring, time-consuming and somewhat imperfect
in various ways, especially as worlds get more complex.

I'm still taking this approach for moveables, but am currently
moving towards using a runtime-simplified world mesh for collision
with static geometry.  I expect that this world mesh will be modelled
as submeshes for each of the collision material types, mostly out of
toolchain considerations.  Apart from that, I expect that I will use
the one-giant-collision-trimesh approach as far as possible, as OPCODE's
internal representation is very good, though I'm a little worried that
this will produce a much-too-sloppy AABB on the ODE side, causing too
many redundant collision callbacks and probes.  Still too early for me to
tell.

--adam
-- 
Adam D. Moss   -   adam at gimp.org
_______________________________________________
ODE mailing list
ODE at q12.org
http://q12.org/mailman/listinfo/ode