Question about TreeCollision

[aqnuep]

Hi,

Again a question just for the sake of clarification.

As far as I know TreeCollision is a static collision so any body which has a TreeCollision type collision will be implicitly static. At least this is how it was in 1.53 as I remember.

Has this thing changed? Are TreeCollision primitives dynamic in Newton 2.0? If not, is there a way to create triangle soup based concave dynamic collisions without using compound collisions?

I don't ask these questions because I'm not satisfied with compound collisions, just want to be sure that currently there is valid information on the Wiki about the topic.

Thanks in advance!

[Julio Jerez]

you are right in all questions.
and not there is no way to make dynamics collision trees


In the pass I studied many alternatives to make dynamics collision tress but I get to the conclusion that it is a futile effort.
The main obstacle is the there is no way to assign volume information to a collision tree, therefore the physics of a dynamics collision tree will be unpredictable.

It is more profound that that than that, if you have a very large body the rigid body abstraction break down. When you have a large body the Nerwton/Euler breaks down and the reason for the is the a force applied at one point in the body is not felt instantaneously across all other points of the body.

Bodies are elastic and lot of energy is absolved in compression and deformations. With need an elasticity model of compatibility,
We need to modify the Lagrange equation to accepts reactions forces are equal but opposite plus some a net internal non conservative loses.
This premise should lead to a model of joints that can allows for a better formulation of the motion of large bodies with elastic joints.
we can cheat using spring stuff but it is something I need to investigate more. Whatever the solution is it will use Compound shapes in the future.

This is one for the reason for NewtonMesh, to assist in the creation of these kind of effects.
Say for example you want to make a space ship, or even a deformable model.
All this stuff transcends the simplicity of collision tree and the separation between Graphics and Physics.
Collision Trees are a dinosaur of the 90’s.

The attention will be focus more and more in Compound collision and NewtonMeshes.

[JernejL]

The info on wiki is still valid, while the trimeshes are still not moved by newton, various forum users did find workarounds to make them dynamic, some methods have decent amout of success.

There's even a wiki article on that manual method:
http://www.newtondynamics.com/wiki/inde ... TriMesh%3F

The info there is based on: viewtopic.php?p=35089

[aqnuep]

Thanks Julio for the very thorough answer.

The info on wiki is still valid, while the trimeshes are still not moved by newton, various forum users did find workarounds to make them dynamic, some methods have decent amout of success.

Yes, I know, I just wanted to be sure that it's not outdated.

Thanks for both of you for the info, for me compound collisions are perfect, just I wanted to know the details.

[JernejL]

The viewtopic.php?p=35089#p35089 could be useful to you, if it works as acceptible workaround in your case.

[aqnuep]

The http://newtondynamics.com/forum/viewtop ... 089#p35089 could be useful to you, if it works as acceptible workaround in your case.

Thanks, as I mentioned for me compound collision is okay, just I'm working on the Wiki, that's why I ask these many stupid questions

Btw maybe I'll write a tutorial about the topic you've linked...