Kester Maddock has been working on enhancing the
ODE support in blender. Here are win/linux binaries
of his work thus far:
http://bebop.cns.ualberta.ca/~cwant/kester/
tuhopuu2-gameengine-test.tar.gz are demo files.
Please provide constructive critisism and feedback.
Cheers,
Chris
Kester's ODE tuhopuu2 demo
Moderators: jesterKing, stiv
Nope thats not it. There is a problem with the polyheder implementation.
Some more things I noticed:
On some polyheders, their collision boundaries seem quite a bit bigger than the actual polygons.
On the test with the stacked boxes, they dont stand up very well, jiggle around a lot, and with the same setup, presents different outcomes in the simulation. I thought one of the things with ODE was that with same input you get very close to the same output, but that obviously isn't the case here.
Not sure how to fix these, but I thought there was some component in ODE to prevent some of the jittering and maybe make it easier for stacked boxes to stand up, don't remember what its called but we should be using it. I'll test some more and look at the source and see if I find anything else.
Some more things I noticed:
On some polyheders, their collision boundaries seem quite a bit bigger than the actual polygons.
On the test with the stacked boxes, they dont stand up very well, jiggle around a lot, and with the same setup, presents different outcomes in the simulation. I thought one of the things with ODE was that with same input you get very close to the same output, but that obviously isn't the case here.
Not sure how to fix these, but I thought there was some component in ODE to prevent some of the jittering and maybe make it easier for stacked boxes to stand up, don't remember what its called but we should be using it. I'll test some more and look at the source and see if I find anything else.
You should get similar outcomes for the simulation - I am using the StepFast solver with one iteration - O(n) - instead of the accurate solver - O(n^3). The docs suggested that 20 iterations was 'a lot' and the simulation was good enough with one, so I went with that.
The stacked boxes do tend to jitter about a bit. I'm not sure what to do about that. I saw stacked boxes for the tokamak physics sdk and thought it would be cool if something collided with them.
Tests:
Game Engine Polytest.blend -- doesn't work: Polyhedra don't collide with boxes. The ODE tests seem to confirm this.
Game Engine Polytest1.blend -- doesn't work: that .blend might be corrupt. The shute doesn't load with a bound type.
Game Engine Test3.blend -- works when I flip the terrain normals.
Game Engine Test4.blend -- works when I flip the prism normals.
Game Engine Test8.blend -- the 'catapult' is the only polyhedra - very non-deterministic.
With the size of collision boundaries, it has to take into account the object's scale vector and the size option in realtime buttons. Size is basically being used as a radius for spheres and cylinders.
If you're on Linux, you might like to try my compile (Redhat 9, pentium3):
http://mysite.reblended.com/tuhopuu2-cv ... 03.tar.bz2
The source patches:
http://mysite.reblended.com/tuhopuu2-ga ... tch.tar.gz
Hos apparently had a mission getting it compiled (thanks Hos). Any more info? You will need all the gameengine prerequisites working - OpenAL & ODE spring to mind. You also need an ODE from ODE cvs. Also, the source patch didn't apply clean.
Check the mailing list for more info.
Kester
The stacked boxes do tend to jitter about a bit. I'm not sure what to do about that. I saw stacked boxes for the tokamak physics sdk and thought it would be cool if something collided with them.
Tests:
Game Engine Polytest.blend -- doesn't work: Polyhedra don't collide with boxes. The ODE tests seem to confirm this.
Game Engine Polytest1.blend -- doesn't work: that .blend might be corrupt. The shute doesn't load with a bound type.
Game Engine Test3.blend -- works when I flip the terrain normals.
Game Engine Test4.blend -- works when I flip the prism normals.
Game Engine Test8.blend -- the 'catapult' is the only polyhedra - very non-deterministic.
With the size of collision boundaries, it has to take into account the object's scale vector and the size option in realtime buttons. Size is basically being used as a radius for spheres and cylinders.
If you're on Linux, you might like to try my compile (Redhat 9, pentium3):
http://mysite.reblended.com/tuhopuu2-cv ... 03.tar.bz2
The source patches:
http://mysite.reblended.com/tuhopuu2-ga ... tch.tar.gz
Hos apparently had a mission getting it compiled (thanks Hos). Any more info? You will need all the gameengine prerequisites working - OpenAL & ODE spring to mind. You also need an ODE from ODE cvs. Also, the source patch didn't apply clean.
Check the mailing list for more info.Kester
They should implement sound, replace the sound with one you have on your computer (C:\Windows\Media...)
That sound isn't there because I didn't pack it because I don't have rights to distribute it (it's from a commercial game.)
It should be set up to play when there is a collision - ie it's a test for the Collision sensor.
That sound isn't there because I didn't pack it because I don't have rights to distribute it (it's from a commercial game.)
It should be set up to play when there is a collision - ie it's a test for the Collision sensor.
Kester Hos, and anyone else working or going to work on the game engine in any form or way, thank ya big time... 
I wish you knew how important this was to me, If I knew how to program I'd be right there with ya..
PS: Most of the tests worked, I had to re calculate a few normals but most worked, there is manly a jitter problem and collison sometimes bouncey calculates before it touches..
I just tried phase's earlier collection of ode blends with this:
http://www.tncci.com/odeblends.zip
They all work, but it's interesting to note that the cola cans
fall over immediately with this new version.
Chris
P.S. Kester: are you on the tuhopuu-devel mailing list?
http://www.tncci.com/odeblends.zip
They all work, but it's interesting to note that the cola cans
fall over immediately with this new version.
Chris
P.S. Kester: are you on the tuhopuu-devel mailing list?
Hi Hanzo,hanzo wrote:hos I want to be on the mailing list! how do I get on it?....
The sign up page for the developer list is here:
http://www.blender.org/mailman/listinfo/tuhopuu-devel
The cvs mailing list is here (sends a mail everytime somebody
commits new code).
http://www.blender.org/mailman/listinfo/tuhopuu-cvs
The cooresponding lists for the official project are here:
http://www.blender.org/mailman/listinfo/bf-committers
http://www.blender.org/mailman/listinfo/bf-blender-cvs
Please note that the development lists are meant primary
for development so it's probably a bad plan to inundate those
lists with lots of feature requests. Another list that is intended for
the discussion for potential features is the 'functionality board'
list:
http://www.blender.org/mailman/listinfo/bf-funboard
Chris