PDA

View Full Version : Bullet and mdd



marchermitte
12-12-2014, 08:36 AM
Hi there!
Is there a trick to get a bullet simulation to collide with MDD animated files ?
I can't get It to work

Thank you!

imshadi
02-19-2017, 07:00 PM
Hi there!
Is there a trick to get a bullet simulation to collide with MDD animated files ?
I can't get It to work

Thank you!

1. Add your MDD collision object to Bullet as a DEFORMING object.
2. Set it's shape retention to 100%
3. Set Shape Lock to <Translation and Rotation>

It worked for me.

marchermitte
01-09-2019, 06:05 AM
done that. With a very basic test scene (a cube animated with MDD, a pool filled with dynamic spheres) I get inconsistent results, like the test mesh (the mdd driven one) does deform like chewing gum, sticks to the hard body spheres. It doesn't matter if I use the filter mesh or not, linear and angular stiffness to 100, lock trans and rot + shape retention to 100 (even tried with keep volume) Noting works correctly. I don't get It, why can't we just have what the traditional LW hard body and cloth fx used to have, an fx_collision function that scan the mesh and use It's vertex position to collide with other dynamic meshes? What are those inefficient workaround?
I can post the scene and videos.

mummyman
01-09-2019, 08:21 AM
Is the displacement MDD file added through nodes? That's how mine was set up in an older scene I did that used MDD files and a bullet sphere bouncing on them. Just a thought. Mine worked either way, though nodes or the MDD reader.

marchermitte
01-09-2019, 12:34 PM
Nope. Trying that right now

marchermitte
01-09-2019, 01:23 PM
Same thing. look at the video and at my settings... It does kind of work, because , as you can see, the sphere are pushed by the cube (not when the cube moves too fast) BUT the cube keep deforming with the sphere interpenetrating It.

143845


https://youtu.be/z9RqMhfK5L0

jwiede
01-09-2019, 06:07 PM
Is the cube only supposed to be deforming due to the MDD? If so, I'm not sure setting it as a "deforming" body is the right setting in Bullet, as that seems to tell Bullet it should be deforming based on collisions as well, which doesn't seem to be what you're seeking.

I believe there's a setting that tells Bullet the object can undergo external deformation-type animation (effectively tells Bullet to continually update it's shape), but where Bullet still treats it as a "rigid body" for its own purposes. Maybe try "kinematic" for the cube? Just a guess, I'd need to experiment a bit, and can't really do so at the moment.

erikals
01-09-2019, 07:23 PM
might want to try a 6 sided Box instead, or a lowres proxy object.

marchermitte
01-10-2019, 01:23 AM
Done that in the beginning, same problem , that's why I subdivided the cube (and collision detection with the sphere seems to be better now)
Jwiede, The cube isn't supposed to be deformed. Using deforming body on a mdd animated mesh, with lock shape and shape retention at 100% is normally the way to get the mdd file to interact with bullet meshes. But The MDD isn't supposed to deform with the 2 options (shape retention and lock)

marchermitte
01-10-2019, 08:35 AM
just made another test, It doesn't work properly, bullet meshes do react to the mdd geometry but interpenetrate it or stay inside It after going trough. Not sure It was that lame before 2018....
Just for the story, the old dynamic system hard fx, can detect any deformed geometry which had applied the advanced collision to It. How's bullet a progress ? Might be better at simulating but ... how does a tool fit in a production workflow?

jwiede
01-10-2019, 11:48 AM
just made another test, It doesn't work properly, bullet meshes do react to the mdd geometry but interpenetrate it or stay inside It after going trough. Not sure It was that lame before 2018....

What's the "Collision Margin" set to in Bullet for cube and sphere(s)? What scale are the objects?

Also, what's your "Dynamics Framerate" setting in Bullet->World?

marchermitte
01-11-2019, 10:52 AM
Last test, a 2 m hight mesh, deformed with MDD, collsion: Tried different values up to 100mm
Spheres: Collision 25mm (they're smaller than the mdd object)
Framerate is 180, the deforming object solver iteration, I tried several values up to 30
Same bad results.
I tried this same exact scene with the of hard fx + advanced collision on the MDD mesh. It works almost perfectly. (and actualy It calculates FASTER!)

jwiede
01-11-2019, 01:44 PM
Last test, a 2 m hight mesh, deformed with MDD, collsion: Tried different values up to 100mm
Spheres: Collision 25mm (they're smaller than the mdd object)
Framerate is 180, the deforming object solver iteration, I tried several values up to 30
Same bad results.
I tried this same exact scene with the of hard fx + advanced collision on the MDD mesh. It works almost perfectly. (and actualy It calculates FASTER!)

Yeah, see this thread (https://forums.newtek.com/showthread.php?158892-Can-bullet-rigid-bodies-collide-with-bullet-deforming-bodies), apparently Carm3D was told it's a known Bullet issue in LW2018.

marchermitte
01-11-2019, 02:13 PM
ok, cool. Thank you for the link!

jwiede
01-11-2019, 02:45 PM
This does kind of highlight why it would be highly beneficial for Newtek to maintain some kind of publicly-viewable and maintained "Known Issues in <current LW version>" site (since bugs reports are now basically private to the reporter and Newtek). These "known issue reports", esp. when involving significant issues in features/systems of LW, affect many more than just the person who reported the issue, and currently there's no way for users to even differentiate whether an issue is "already known" or not.

Expecting "every customer should report every issue they encounter" is wildly unrealistic, customers' time is (at least) as scarce as Newtek employees' time. If Newtek wishes to crowdsource how to prioritize among such known issues, there are much better, UserVoice-type systems they could deploy. Users could easily/quickly vote on or otherwise indicate (depends on system) which issues they consider most important to them, with full analytics, etc. Trying to use bug report quantities as an issue priority indicator just isn't a reliable approach, esp. given the "barriers to entry" that currently exist.

erikals
01-11-2019, 03:12 PM
This does kind of highlight why it would be highly beneficial for Newtek to maintain some kind of publicly-viewable and maintained "Known Issues in <current LW version>" site (since bugs reports are now basically private to the reporter and Newtek). These "known issue reports", esp. when involving significant issues in features/systems of LW, affect many more than just the person who reported the issue, and currently there's no way for users to even differentiate whether an issue is "already known" or not.

Expecting "every customer should report every issue they encounter" is wildly unrealistic, customers' time is (at least) as scarce as Newtek employees' time. If Newtek wishes to crowdsource how to prioritize among such known issues, there are much better, UserVoice-type systems they could deploy. Users could easily/quickly vote on or otherwise indicate (depends on system) which issues they consider most important to them, with full analytics, etc. Trying to use bug report quantities as an issue priority indicator just isn't a reliable approach, esp. given the "barriers to entry" that currently exist.

Yup!

marchermitte
01-12-2019, 06:27 AM
You're absolutely right. I was a beta tester for another software and the bug report system was pretty good, It worked like a dedicated web site where you can search by bug, by author, by developer, resolved / open, etc.
For instance, I report a bug (say MDD not colliding correctly with bullet), if someone else encounter the same problem, he can confirm the bug, adding his name and description + scene files and/or video. We can then follow up and see if/when It's resolved.
Since I started using 2018, I'm encountering many problems, most of them being bugs , hub or stability related. (Lw 11.6 is still my favorite version so far features and stability wise.)
I remember when Lightwave was the very first and best Sub-D modeler, what happens next, other packages (such as XSI) take the technologies and make It better (Catmul Clarck sub-D with Ngon for instance)
Lightwave was also the 1rst software (correct me if I'm wrong) to have the morphs embedded into one single mesh, opposed to blend shapes in Maya for instance, where you need one separate object as a target to morph with.
It seems that Newtek really shines at finding new technologies but lack of the ability of developing them further and making them evolve with new versions.
Look at SpeedEdit... video Toaster...

jwiede
01-22-2019, 07:30 PM
Marchermitte, have you checked whether the Bullet-v-MDD problem you were encountering is fixed in LW2019.0.0? Might be worth checking, even if just with trial license.