On 13/09/12 20:45, Samuel Benge wrote:
>> Searching for such a program on the net I found Koppi's Playground
>
> Would I really need to build Koppi's Playground myself? Seems like a
> lot of work to go though. Tangling with Blender is looking more and
> more like an easier way to go when I consider trying to download and
> install the tool chain necessary to get KBPP up and running %(

   No, you are already doing more complex things than Koppi's playground
allows... so it' would not be of much use for you. At this moment,
Koppi's playground is very limited.

> Nice updated pencils scene, BTW. Those "no_image" erasers are a kick
> :) Are you using geometries other than boxes now? It would appear
> that the erasers' rounded edges are behaving as such, although some
> pencils seem to be balancing on their edges, so I'm not sure.

   No, just boxes... even for the pencils (that's why you see them
floating a little bit at some places), as cylinders were behaving a
bit erratically instead resting on the floor. The rounded erasers look
right just because I tried the simulation several times until I found a
result without obvious flaws.

--
Jaime

Post a reply to this message

On 14/09/12 01:19, Robert McGregor wrote:
> Jaime Vives Piqueres <jai### [at] ignoranciaorg> wrote:
>> It would be nice to have a tool specifically made to interface the
>> Bullet Physics library with POV-Ray, to have a more simpler
>> process (run the simulation, export and render)...
>
> I have been thinking about (and working on) this quite a lot lately
> :)

   I've plenty of ideas for the interface, if you really are going to
start such a tool...

> fyi, I noticed some mirroring of the "no_image" text on some erasers
> (literally mirrored text) that may need to be fixed...

   Yeah... I was lazy enough to just use planar mapping without realizing
it would give and inverted image at the other side: time for some
uv_mapping.

--
Jaime

Post a reply to this message

On 13/09/12 23:02, Robert McGregor wrote:
> Okay, here is Gille Tran's shopping cart filled to overflowing with primitives
> via a custom program I wrote using the Bullet library. Still pondering easier
> integration with POV-Ray...

   Just... :O


--
Jaime

Post a reply to this message

"Robert McGregor" <rob### [at] mcgregorfineartcom> wrote:
> Okay, here is Gille Tran's shopping cart filled to overflowing with primitives
> via a custom program I wrote using the Bullet library. Still pondering easier
> integration with POV-Ray...
>

Well that's excessively cool, Robert. Congrats!

What kind of geometry evaluations did you use for the object collisions? Convex
meshes, triangle (concave) meshes, primitive mathematical representations? Are
they colliding with a low-poly mesh representation of the cart?

More than any of those things, I'd like to know how stable everything was at
runtime. When working in Blender, solid bodies tend to jump around a lot when
confined in any space, sometimes never stabilizing. This mainly happens when
performing mesh-mesh collisions.

BTW, those front wheels seem off. More often than not you'll see them pointed
the other way... maybe the force of all those dropped objects pushed the cart
back? :)

Sam

Post a reply to this message

On 14-9-2012 22:29, Samuel Benge wrote:

> Well that's excessively cool, Robert. Congrats!

I join that!

Jaime, Sam, Robert, you are presenting very interesting work here.

>
> [...]
>
> More than any of those things, I'd like to know how stable everything was at
> runtime. When working in Blender, solid bodies tend to jump around a lot when
> confined in any space, sometimes never stabilizing. This mainly happens when
> performing mesh-mesh collisions.

I am not far enough with Blender, but I have the same experience using 
Poser Physics. Confined to a container, objects seem indeed to bounce 
around more, even when the bounce parameter is set to zero.

Thomas

Post a reply to this message

On 14/09/12 22:29, Samuel Benge wrote:
> More than any of those things, I'd like to know how stable everything
> was at runtime. When working in Blender, solid bodies tend to jump
> around a lot when confined in any space, sometimes never stabilizing.
> This mainly happens when performing mesh-mesh collisions.

   I find this problem too when using Koppi's Playground, specially
cylinders are unusable as they start to roll as if they were spheres,
instead of resting on the floor. With boxes, when they are long and
thing, I also have stability problems: there is always some that will
start shaking, sometimes spreading the problem to near boxes. Seems the
bullet physics simulation is more suited for animations of the
simulation, rather than for stills of the simulation result...

--
Jaime

Post a reply to this message

On 15-9-2012 9:39, Thomas de Groot wrote:
> I am not far enough with Blender, but I have the same experience using
> Poser Physics. Confined to a container, objects seem indeed to bounce
> around more, even when the bounce parameter is set to zero.

One issue (at least with Poser Physics) is when objects "touch" at the 
start; maybe even when they are "too" close. The result is unpredictable.

As in Poser Physics collision is calculated from box, sphere or capsule, 
all other objects need to have a shape fairly close to one of those to 
get believable results.

Thomas

Post a reply to this message

On 15/09/12 10:32, Thomas de Groot wrote:
> One issue (at least with Poser Physics) is when objects "touch" at
> the start; maybe even when they are "too" close. The result is
> unpredictable.
>

   This is happening also to me on Koppi's playground... reading about
the Bullet Physics Library, seems this is on purpose: if the objects are
overlapping, the simulation attaches some initial forces to them so they
move apart (and then they start colliding before you wanted).

--
Jaime

Post a reply to this message

On 15-9-2012 11:56, Jaime Vives Piqueres wrote:
>
>    This is happening also to me on Koppi's playground... reading about
> the Bullet Physics Library, seems this is on purpose: if the objects are
> overlapping, the simulation attaches some initial forces to them so they
> move apart (and then they start colliding before you wanted).

It is understandable as one does not want to have overlapping objects, 
but the force moving them apart seems too strong to me, and cannot be 
controlled by hand unfortunately.

With random objects, the difficulty is to have them as close in shape as 
possible to the basic collision objects. That puts some constrains upon 
the modelling...

Thomas

Post a reply to this message

"Samuel Benge" <stb### [at] hotmailcom> wrote:
> Well that's excessively cool, Robert. Congrats!

Thanks Sam :)

> What kind of geometry evaluations did you use for the object collisions? Convex
> meshes, triangle (concave) meshes, primitive mathematical representations? Are
> they colliding with a low-poly mesh representation of the cart?

For the shopping cart I made a simple cart-shaped stand-in box, with the top
removed, that fits inside the actual cart geometry (10 triangles). For the
collisions themselves I used:

[Dynamic objects]
Primitives:
   btBoxShape
   btSphereShape
   btConeShape
   btConvexHullShape

[Kinematic objects]
Cart box: btTriangleMesh
Ground: btStaticPlaneShape

After the sim is stable (objects have stopped moving) the transformation data
for each object (via btTransform::getBasis() for rotation and
btTransform::getOrigin() for translation) is written to a POV-Ray include file
as a set of multidimensional arrays, something like this:

// for each collision object
btCollisionObject* obj = m_dynamicsWorld->getCollisionObjectArray()[i];
btRigidBody* body = btRigidBody::upcast(obj);

btMatrix3x3 rot = body->getWorldTransform().getBasis();
btVector3 trans = body->getWorldTransform().getOrigin();

// write to file...

> More than any of those things, I'd like to know how stable everything was at
> runtime. When working in Blender, solid bodies tend to jump around a lot when
> confined in any space, sometimes never stabilizing. This mainly happens when
> performing mesh-mesh collisions.

Working with Bullet is extremely stable. While a sim is running it's pretty much
real-time, like a video game.

I haven't delved into the soft body capabilities yet (although they are
amazing); I just wanted to get my feet wet with the rigid body stuff first.

> BTW, those front wheels seem off. More often than not you'll see them pointed
> the other way... maybe the force of all those dropped objects pushed the cart
> back? :)

Hmm, I hadn't noticed that, it must've rolled backwards a bit indeed :)
-------------------------------------------------
www.McGregorFineArt.com

Post a reply to this message