[ODE] RE: gjk convex collision preview for ODE

Wed Sep 7 08:14:46 MST 2005

Bullet collision detection already has support for arbitrary triangle
meshes, so it is just a matter of integrating into ODE. Some 'midphase'
optimization are still missing (tree to do early culling, for example a
k-dop tree).

Another option is to mix the gjk-convex with the trimesh collider  for
general convex objects. Then you get the Opcode 'midphase' optimizations. So
I would say it is not 'technically difficult', more laborous work.

Erwin

----- Original Message ----- 
From: "Jaroslav Sinecky" <jsinecky at tiscali.cz>
To: <phpbb at erwincoumans.com>
Cc: <ODE at q12.org>
Sent: Wednesday, September 07, 2005 7:58 AM
Subject: [ODE] RE: gjk convex collision preview for ODE

> How about collision between convex solids and general trimesh (including
> concave trimesh i.e. terrain) ? Is it possible/difficult to implement? I
> don't know much about GJK and neighter about how actual ODE trimesh
> colliders work, so sorry if it's silly question.
>
> Jaroslav
>
>
> > -----Original Message-----
> > From: Megan Fox [mailto:shalinor at gmail.com]
> > Sent: Tuesday, September 06, 2005 10:56 PM
> > To: phpbb at erwincoumans.com
> > Cc: Rodrigo Hernandez; Jaroslav Sinecky
> > Subject: Re: gjk convex collision preview for ODE
> >
> >
> > That's great news - granted, you can "kinda" fake  convex solids, but
> > it's just really irritating to make something that can roll AND settle
> > on a side without them.
> >
> > Boulders look kinda silly no matter if I use boxes or spheres :(
> > Never mind all those cylinder objects that never stop rolling.
> >
> > On 9/6/05, phpbb at erwincoumans.com <phpbb at erwincoumans.com> wrote:
> > >
> > > Hi Megan,
> > >
> > > This is just a 2 day integration job, and the problems are
> > related to ode
> > > integration, not to the gjk implementation. If I get the time, and
> > > motivation there will be fast convex collision detection. Bunny
> > was a very
> > > bad example to show convex by the way :)
> > >
> > > Cheers,
> > > Erwin
> > >
> > >
> > >
> > > Megan Fox writes:
> > >
> > > > Are the issues (like the bunny penetrating the ground) all solvable
> > > > issues, or is GJK always going to have these sorts of issues?  Or
are
> > > > they more a problem with proper ODE integration?
> > > >
> > > > That is... at some point in the future, can we expect convex solids
to
> > > > be as stable as boxes and spheres, without a severe performence
> > > > penalty?
> > > >
> > > >> Hi,
> > > >>
> > > >> Thanks!
> > > >>
> > > >> Indeed, the cylinders gain a bit of energy due to
> > penetration depth recovery
> > > >> in the ODE solver. It needs a little bit of damping, then it
> > should be fine.
> > > >> You can press 'n' as well, for a convex hull bunny. Bad
> > example obviously,
> > > >> to add 400+ vertices to a convex hull object. The graphics
> > drawing of the
> > > >> bunny seems to be very slow too. Also notice that there is
> > interaction
> > > >> between capped cylinder, box, sphere and gjk cylinder/convex
> > hull too. In
> > > >> gjk you can just take the convex-hull of two spheres for the capped
> > > >> cylinder.
> > > >>
> > > >>
> > > >> The modified ODE source code and Bullet source code can be
> > downloaded here,
> > > >> including visual studio 6 and 8 projectfiles:
> > > >> http://207.234.235.180/ftp/pub/test/physics
> > > >>
> > > >> See a file called BulletWithOdeSampleIntegration.zip
> > > >>
> > > >> Open \Bullet\Extras\ode-0.5\VC6\Samples\MakeAllTests.dsw (or
> > .sln for VC8)
> > > >>
> > > >> Thanks for testing !
> > > >> Erwin
> > > >>
> > > >>
> > > >> Megan Fox writes:
> > > >>
> > > >> > Yep, works fine now - looks great.  The cylinders are excellent,
> > > >> > though it does seem as if they tend to gain velocity when
> > rolling on
> > > >> > their side (they sort of accelerate off into the distance
> > and fall off
> > > >> > the ground plane)
> > > >> >
> > > >> > On 9/6/05, phpbb at erwincoumans.com <phpbb at erwincoumans.com> wrote:
> > > >> >>
> > > >> >> Hi,
> > > >> >>
> > > >> >> this was a difficult to spot problem, related to Visual
> > Studio 2005.
> > > >> >> Apparently when it converts the vc6 files, it enables a
> > 'manifest', and it
> > > >> >> requires to add the file Microsoft.VC80.CRT.manifest and
> > some dll's.
> > > >> >>
> > > >> >> I disabled this 'manifest' setting (whatever it means)
> > and recompiled again.
> > > >> >> I think this should work.
> > > >> >>
> > > >> >>
> > http://207.234.235.180/ftp/pub/test/index.php?dir=physics/&file=bu
> > lletgjk_od
> > > >> >> e.zip
> > > >> >>
> > > >> >> Can you let me know ?
> > > >> >>
> > > >> >> Usage: press 'l' for bullet gjk cylinder, 'f' for gjk
> > box, 'b' for ode box,
> > > >> >> c and s. 't' for contact points (regular ode demo).
> > > >> >>
> > > >> >> Thanks,
> > > >> >> Erwin
> > > >> >>
> > > >> >> Megan Fox writes:
> > > >> >>
> > > >> >> > Still no dice, at least for me.  Still complains of improper
> > > >> >> > application configuration (an error I've never seen,
> > oddly enough -
> > > >> >> > I'm used to illegal operations and such)
> > > >> >> >
> > > >> >> > On 9/6/05, Erwin Coumans <erwin at erwincoumans.com> wrote:
> > > >> >> >>
> > > >> >> >> Hi,
> > > >> >> >>
> > > >> >> >> thanks for testing. I added one missing dll, msvcr80.dll.
> > > >> >> >>
> > > >> >> >> Could you check if it works, and if not, mail me which
> > file is missing ? I'm
> > > >> >> >> using visual studio 8, this might explain the problem.
> > > >> >> >>
> > > >> >> >> If it works, please mail at the ode forum ;)
> > > >> >> >>
> > > >> >> >> Thanks,
> > > >> >> >> Erwin
> > > >> >> >>
> > > >> >> >> ----- Original Message -----
> > > >> >> >> From: "Megan Fox" <shalinor at gmail.com>
> > > >> >> >> To: "Erwin Coumans" <erwin at erwincoumans.com>
> > > >> >> >> Cc: <ODE at q12.org>
> > > >> >> >> Sent: Tuesday, September 06, 2005 1:32 AM
> > > >> >> >> Subject: Re: [ODE] gjk convex collision preview for ODE
> > > >> >> >>
> > > >> >> >>
> > > >> >> >> This is wonderful news... but when I attempt to run
> > the exe, I get a
> > > >> >> >> "this application is configured improperly" error.
> > Alas.  Any ideas?
> > > >> >> >>
> > > >> >> >> ... I remain extremely excited at the prospect of
> > convex collision, though.
> > > >> >> >>
> > > >> >> >> On 9/5/05, Erwin Coumans <erwin at erwincoumans.com> wrote:
> > > >> >> >> >
> > > >> >> >> > Hi,
> > > >> >> >> >
> > > >> >> >> > There is a small preview of Bullet GJK Convex
> > Collision Detection
> > > >> >> >> integrated
> > > >> >> >> > with Open Dynamics Engine (ODE). Press l for a
> > bullet gjk cylinder, f for
> > > >> >> >> a
> > > >> >> >> > bullet gjk box, b, s and c for ode colliders. Other
> > convex primitives like
> > > >> >> >> > cone, convex mesh etc. are a matter of adding
> > graphics drawstuff for ODE.
> > > >> >> >> > Disclaimer: it doesn't work perfect ;)
> > > >> >> >> >
> > > >> >> >> >
> > > >> >> >>
> > http://207.234.235.180/ftp/pub/test/index.php?dir=physics/&file=bu
> lletgjk_ode.zip
> > >> >> >> >
> > >> >> >> > Sources are checked into CVS at
> > >> >> >> > http://sourceforge.net/projects/bullet
> > >> >> >> > Actual ODE demo source will be added soon.
> > >> >> >> >
> > >> >> >> > Erwin
> > >> >> >> > http://www.continuousphysics.com/Bullet
> > >> >> >> >
> > >> >> >> > _______________________________________________
> > >> >> >> > ODE mailing list
> > >> >> >> > ODE at q12.org
> > >> >> >> > http://q12.org/mailman/listinfo/ode
> > >> >> >> >
> > >> >> >> >
> > >> >> >> >
> > >> >> >>
> > >> >> >>
> > >> >> >> --
> > >> >> >> -Megan Fox
> > >> >> >> Lead Developer, Elium Project
> > >> >> >> http://www.elium.tk
> > >> >> >>
> > >> >> >>
> > >> >> >>
> > >> >> >
> > >> >> >
> > >> >> > --
> > >> >> > -Megan Fox
> > >> >> > Lead Developer, Elium Project
> > >> >> > http://www.elium.tk
> > >> >>
> > >> >>
> > >> >>
> > >> >
> > >> >
> > >> > --
> > >> > -Megan Fox
> > >> > Lead Developer, Elium Project
> > >> > http://www.elium.tk
> > >>
> > >>
> > >>
> > >
> > >
> > > --
> > > -Megan Fox
> > > Lead Developer, Elium Project
> > > http://www.elium.tk
> >
> >
> >
>
>
> --
> -Megan Fox
> Lead Developer, Elium Project
> http://www.elium.tk
>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>