Efficient or bad: tree collision compared to primitive

[Henry00]

Hi everyone it's me again from 00Laboratories, I was wondering about something..

Currently once someone makes a model, for example a wheel or a barrel, I would use a primitive shape NewtonCreateCylinder.
But since the representation of the cylinder ( newtonCollisionForEachPolygonDo ) is just a triangle based cylinder, would it be efficient to use a treecollision, or does Newton have problems when I build a cylinder myself at simulating the physics?
This decreases alot of work when making the physical world, and allows advanced "decoration" to have presize physics as well instead of a cylinder.

Thanks! ( and @ JoeJ, we are getting closer with lighting http://henry00.ftpaccess.cc/henry00/projects/Wind16_light2.png )

[Julio Jerez]

there is a limitation with teh collison system and that si that each collison shape must be convex, therefore you cna no make a cylinder usign collision tree since the algorithm will not have a way of knowing what is outside and inside the shapen.

[Henry00]

Yeah I can apply that rule thanks for the post, but somehow treecollision aint falling, and the primitives are..

That is if I understand convex correct, no inside holes? just plain outside right

yeah now I get it, you mean literally triangles in the world, and not a solid shape to check with that has an inside!
Would there be a different way to achieve this or should I just give up on the idea of making things more perfect..

[Julio Jerez]

Tree collision are by definitions static objects.
The engine conve tree collision to and array of flar conve polygon.

a Flat conve polygon is a special case of a conve solid with zere volume.
The engine uses that condition to determine if a solid volume shape with volume collide with a shape plane.
The functions of a solid-shape-flatPolygon collsion test are special, and the know the the first faces has a volume.

you can use a compound collision,
there is an add on to the engine that can decomposes a solid into an agregate fo convex solid, but the integration is still incompleted.
the first time I tryed I found some Bugs, they were fixed but I have not re integrated the latest version.

[Henry00]

Thanks Julio, even though it's strange that your english doesn't match your programming skills, I guess I have to wait until newton increases in functionality!
Good luck, I will be waiting!

[jiandingzhe]

Tree collision are by definitions static objects.
The engine conve tree collision to and array of flar conve polygon.

a Flat conve polygon is a special case of a conve solid with zere volume.
The engine uses that condition to determine if a solid volume shape with volume collide with a shape plane.
The functions of a solid-shape-flatPolygon collsion test are special, and the know the the first faces has a volume.

you can use a compound collision,
there is an add on to the engine that can decomposes a solid into an agregate fo convex solid, but the integration is still incompleted.
the first time I tryed I found some Bugs, they were fixed but I have not re integrated the latest version.

I guess some modelling software might have the function of separating arbitrary shapes into convex shapes. It's just a guess...

[JernejL]

Ofcourse, but you are still dealing with 1-dimensional flat polygon shapes, and they are very complex to handle efficiently in real-time, you can separate your meshes into compound shapes made of convex hulls as a offline model preparation step, that works very well in newton.