Thomas de Groot <tho### [at] degrootorg> wrote:
>  From what I see, the end result of the physics animation is much better
> (with objects) than what Poser Physics does (its ragdoll animation is
> good). Somehow I remain with objects not touching each other and
> floating in space within a bowl (See image; Poser render). The objects
> fall very rapidly into position and then remain thus for the rest of the
> animation, even after about a thousand frames.

It looks like the collision shapes just aren't corresponding to the actual
geometry closely enough. Is a hull or mesh collision shape available in Poser
Physics?
-------------------------------------------------
www.McGregorFineArt.com

Post a reply to this message

Thomas de Groot <tho### [at] degrootorg> wrote:
> Somehow I remain with objects not touching each other and
> floating in space within a bowl (See image; Poser render).

Nuts! Is there any sort of padding setting for colliding objects? I mean, do the
objects try to keep a distance from other objects to avoid errors?

Another thing that occurs to me (although I haven't used Poser since around
1999) is that you are using displacement maps for your objects. Perhaps Poser is
using the /unmodified/ meshes for collisions, not the modified versions. That
would cause some spacing issues for sure.

> The objects fall very rapidly into position and then remain
> thus for the rest of the animation, even after about a thousand frames.

I'm guessing the objects are "sleeping" too soon after settling. Is there a "no
sleeping" or similar option you can disable?

Sam

Post a reply to this message

"Robert McGregor" <rob### [at] mcgregorfineartcom> wrote:
> "Samuel Benge" <stb### [at] hotmailcom> wrote:
> > Attached is another test, this time showing a ball blasting through a wall of
> > cubes. For the motion blur effect, I saved two states of the sim at different
> > time steps and interpolated the transformations in POV over a series of frames.
> > The images were then averaged together.
>
> That motion blur effect is outstanding! How many frames did you average to
> achieve this shot?

Forty, though it looks like fewer because I jittered the camera's position at
each frame for antialiasing/focal blur. All frames together finished rendering
in about 4 minutes 45 seconds using three 2.4GHz cores, which isn't too bad at
all. Radiosity and area_light settings were low, everything got a scaled-down
bumps normal, and no AA was used.

On a related subject, I noticed that POV-Ray's actual focal blur uses the same
set of samples for each frame. I'm guessing this is due to the Halton sequence
used, but is very annoying when trying to have very low focal blur settings with
the intention of averaging frames together. If each frame used a different
sample set, you could get away with a blur_samples value of 1, which as far as I
can tell, adds nothing to the render time. It would be nice to have the option
of using random focal blur samples for each frame :)

Sam

Post a reply to this message

To Robert:
I think that is the case indeed. Poser Physics uses Ball, Box, and 
Capsule (a cylinder with rounded ends), and Ragdoll as collision 
objects. While I couple each mesh to one of the collision objects, it 
seems that the mesh has to be exactly conform, so no initial transform 
to it should be done, or that is what I guess presently.

To Sam:
No, there is no padding setting.
My meshes are /pure/ meshes; however see my answer to Robert.
There is no sleeping option available.

I want to do the same with only collision objects and see what happens 
but I believe that there is a serious constrain on Poser Physics... :-(

Thomas

Post a reply to this message

On 18-9-2012 9:19, Thomas de Groot wrote:
>
> I want to do the same with only collision objects and see what happens
> but I believe that there is a serious constrain on Poser Physics... :-(
>

It seems to be as I thought. Simulating with only collision objects 
works well. This means that mesh objects need to be as close in shape to 
the collision objects as possible.

Thomas

Post a reply to this message

Attachments:
Download 'poserphysicstest_02.png' (178 KB)

Preview of image 'poserphysicstest_02.png'

Thomas de Groot <tho### [at] degrootorg> wrote:
> It seems to be as I thought. Simulating with only collision objects
> works well. This means that mesh objects need to be as close in shape to
> the collision objects as possible.

Maybe you could use combined collision shapes like Sam did with the nails? After
seeing that little trick I dug into Bullet a bit more and found how to do it
directly. While I was poking around I also found that the ragdoll is really just
a simple collection of capsules and joint constraints:

class RagDoll
{
   enum
   {
      BODYPART_PELVIS = 0,
      BODYPART_SPINE,
      BODYPART_HEAD,

      BODYPART_LEFT_UPPER_LEG,
      BODYPART_LEFT_LOWER_LEG,

      BODYPART_RIGHT_UPPER_LEG,
      BODYPART_RIGHT_LOWER_LEG,

      BODYPART_LEFT_UPPER_ARM,
      BODYPART_LEFT_LOWER_ARM,

      BODYPART_RIGHT_UPPER_ARM,
      BODYPART_RIGHT_LOWER_ARM,

      BODYPART_COUNT
   };

   enum
   {
      JOINT_PELVIS_SPINE = 0,
      JOINT_SPINE_HEAD,

      JOINT_LEFT_HIP,
      JOINT_LEFT_KNEE,

      JOINT_RIGHT_HIP,
      JOINT_RIGHT_KNEE,

      JOINT_LEFT_SHOULDER,
      JOINT_LEFT_ELBOW,

      JOINT_RIGHT_SHOULDER,
      JOINT_RIGHT_ELBOW,

      JOINT_COUNT
   };

   ...
   ...

   m_shapes[BODYPART_PELVIS] = new btCapsuleShape(
      btScalar(0.15), btScalar(0.20));
   m_shapes[BODYPART_SPINE] = new btCapsuleShape(
      btScalar(0.15), btScalar(0.28));
   m_shapes[BODYPART_HEAD] = new btCapsuleShape(
      btScalar(0.10), btScalar(0.05));
   m_shapes[BODYPART_LEFT_UPPER_LEG] = new btCapsuleShape(
      btScalar(0.07), btScalar(0.45));
   m_shapes[BODYPART_LEFT_LOWER_LEG] = new btCapsuleShape(
      btScalar(0.05), btScalar(0.37));
   m_shapes[BODYPART_RIGHT_UPPER_LEG] = new btCapsuleShape(
      btScalar(0.07), btScalar(0.45));
   m_shapes[BODYPART_RIGHT_LOWER_LEG] = new btCapsuleShape(
      btScalar(0.05), btScalar(0.37));
   m_shapes[BODYPART_LEFT_UPPER_ARM] = new btCapsuleShape(
      btScalar(0.05), btScalar(0.33));
   m_shapes[BODYPART_LEFT_LOWER_ARM] = new btCapsuleShape(
      btScalar(0.04), btScalar(0.25));
   m_shapes[BODYPART_RIGHT_UPPER_ARM] = new btCapsuleShape(
      btScalar(0.05), btScalar(0.33));
   m_shapes[BODYPART_RIGHT_LOWER_ARM] = new btCapsuleShape(
      btScalar(0.04), btScalar(0.25));

   ...
   ...
}

Another thought, maybe try making the collision shapes smaller for the floating
meshes?

-------------------------------------------------
www.McGregorFineArt.com

Post a reply to this message

On 18-9-2012 13:59, Robert McGregor wrote:
> Maybe you could use combined collision shapes like Sam did with the nails? After
> seeing that little trick I dug into Bullet a bit more and found how to do it
> directly. While I was poking around I also found that the ragdoll is really just
> a simple collection of capsules and joint constraints:
>
> Another thought, maybe try making the collision shapes smaller for the floating
> meshes?
>

The problem is that in Poser Physics you have no control on these. You 
can set one of the pre-defined collision objects to (e.g.) a mesh, but 
you have no control over those collision objects. This seems to be a 
serious limitation on the Poser Physics plugin for Poser.

Thomas

Post a reply to this message

On 18/09/2012 8:19 AM, Thomas de Groot wrote:
> I want to do the same with only collision objects and see what happens
> but I believe that there is a serious constrain on Poser Physics... :-(

Too true. I believe that it is unsupported now.

-- 
Regards
     Stephen

Post a reply to this message

On 18-9-2012 21:30, Stephen wrote:
> On 18/09/2012 8:19 AM, Thomas de Groot wrote:
>> I want to do the same with only collision objects and see what happens
>> but I believe that there is a serious constrain on Poser Physics... :-(
>
> Too true. I believe that it is unsupported now.
>

However, there is a Poser Physics 2012, for Poser Pro 2012.

Thomas

Post a reply to this message

On 19/09/2012 7:44 AM, Thomas de Groot wrote:
> On 18-9-2012 21:30, Stephen wrote:
>> On 18/09/2012 8:19 AM, Thomas de Groot wrote:
>>> I want to do the same with only collision objects and see what happens
>>> but I believe that there is a serious constrain on Poser Physics... :-(
>>
>> Too true. I believe that it is unsupported now.
>>
>
> However, there is a Poser Physics 2012, for Poser Pro 2012.
>
I don't think it is worth upgrading from Pro 2010 just for that, do you.

I do intend to get to grips with the latest version of Blender though.

Sometime


-- 
Regards
     Stephen

Post a reply to this message