btConeShape orientation

nomad · Post by **nomad** » Thu Nov 16, 2006 5:03 pm

The constructor for btConeShape is documented as btConeShape::btConeShape ( btScalar radius, btScalar height) .
But when you try to match a btConeShape to a mesh, where the mesh is a cone standing on its base (i.e. height along y-axis), the collision shape does not match the geometry.
This is simply because the btConeShape is oriented along the z-axis initially, so the height param is really its length. Using cones in the CCDPhysicsDemo shows this, if you disable the sim and swing the camera round to the back of the stack.
This is not much of a problem really, it is just a matter of passing the correct dimension as the cone's length, and orienting the mesh correctly.
But perhaps it would be better to change the 'height' param to 'length', to avoid confusion?

Jack · Post by **Jack** » Fri Nov 17, 2006 12:23 pm

I have a troubles with ConeMesh too. How does it orientated? Where is zero point?

nomad · Post by **nomad** » Fri Nov 17, 2006 3:02 pm

btConeShape is aligned like this:
cone.jpg
(The 3 arrows showing the x,y,z axes 'meet' at the world origin)

Cone aligned to z-axis with center at (0,0,0). This is how the btConeShape is constructed. If your mesh is also like this, you should not have any problems. Just use the mesh length (from base to apex) in the btConeShape constructor, i.e. something like

Code: Select all

m_CollisionShape = new btConeShape btConeShape(radius, length).

Works for me anyway.

Apart from this, I have had no problems with cones.

Jack · Post by **Jack** » Fri Nov 17, 2006 4:46 pm

nomad wrote:btConeShape is aligned like this:
cone.jpg
(The 3 arrows showing the x,y,z axes 'meet' at the world origin)

Cone aligned to z-axis with center at (0,0,0). This is how the btConeShape is constructed. If your mesh is also like this, you should not have any problems. Just use the mesh length (from base to apex) in the btConeShape constructor, i.e. something like
Code: Select all
m_CollisionShape = new btConeShape btConeShape(radius, length).
Works for me anyway. Apart from this, I have had no problems with cones.

I've made a mesh like on your picture, but it is not working...

nomad · Post by **nomad** » Fri Nov 17, 2006 6:09 pm

Jack wrote:I've made a mesh like on your picture, but it is not working...

How exactly is it not working? Is the cone colliding with anything?

Jack · Post by **Jack** » Fri Nov 17, 2006 6:50 pm

nomad wrote:
Jack wrote:I've made a mesh like on your picture, but it is not working...
How exactly is it not working? Is the cone colliding with anything?

It is colliding with static plane, but it does not fit the mesh. It looks like apex is in (0,0,0). But I can not find correct mesh...Or maybe it does not collide correctly...

nomad · Post by **nomad** » Fri Nov 17, 2006 7:17 pm

If you are creating the mesh in a graphics program (like 3ds Max) then perhaps the mesh coordinates are not centered about the origin? I had a problem like this with Max, which exports a transformation matrix with the mesh.

Jack · Post by **Jack** » Fri Nov 17, 2006 7:33 pm

nomad wrote:If you are creating the mesh in a graphics program (like 3ds Max) then perhaps the mesh coordinates are not centered about the origin? I had a problem like this with Max, which exports a transformation matrix with the mesh.

FrameTransformMatrix is Identity. I am using this exporter:
http://www.andytather.co.uk/Panda/direc ... loads.aspx
I like it more then one included in DirectX SDK.

nomad · Post by **nomad** » Fri Nov 17, 2006 8:07 pm

Still sounds like an export issue, but maybe your problem is different to mine. If you can let me have your MAX file (I'm using MAX 7) then I could export it and see how your cone works in my game engine (I use .3ds and .X formats). Or you could try my cone: cone.max

Jack · Post by **Jack** » Fri Nov 17, 2006 8:34 pm

I made it. I corrected btConeShape.cpp to:

Code: Select all

#ifdef WIN32
//static int coneindices[3] = {1,2,0};
[b]static int coneindices[3] = {2,1,0};[/b]
#else
static int coneindices[3] = {2,1,0};
#endif

Now it is working. In fact cone is vertical (Y) aligned. Its bottom is at zero level....

Jack · Post by **Jack** » Fri Nov 17, 2006 8:52 pm

It also looks like scaling is not working for cone shape..

Erwin Coumans · Post by **Erwin Coumans** » Sat Nov 18, 2006 3:51 am

Jack wrote:It also looks like scaling is not working for cone shape..

Some fixes have been made to the btConeShape. Unfortunately no scaling yet. You can apply scaling yourself, directly on the height/radius.

Some updates have been made in the Bullet User Manual:
http://www.continuousphysics.com/ftp/pu ... Manual.pdf

Thanks for the feedback!
Erwin

Jack · Post by **Jack** » Sat Nov 18, 2006 10:09 am

Erwin Coumans wrote:
Jack wrote:It also looks like scaling is not working for cone shape..
Some fixes have been made to the btConeShape. Unfortunately no scaling yet. You can apply scaling yourself, directly on the height/radius.

Some updates have been made in the Bullet User Manual:
http://www.continuousphysics.com/ftp/pu ... Manual.pdf

Thanks for the feedback!
Erwin

Is it hard to add scaling? Objects are scaled dynamically during game play...It would be very great to have scaling for cone.

nomad · Post by **nomad** » Sat Nov 18, 2006 10:42 am

Jack wrote:I made it. I corrected btConeShape.cpp to:
Code: Select all
#ifdef WIN32
//static int coneindices[3] = {1,2,0};
[b]static int coneindices[3] = {2,1,0};[/b]
#else
static int coneindices[3] = {2,1,0};
#endif
Now it is working. In fact cone is vertical (Y) aligned. Its bottom is at zero level....

Yep. Cone is now y-aligned, with center at (0,0,0). Much easier to work with.

Erwin Coumans · Post by **Erwin Coumans** » Sat Nov 18, 2006 4:33 pm

Jack wrote:Is it hard to add scaling? Objects are scaled dynamically during game play...It would be very great to have scaling for cone.

It is easy to add one scale-factor for all directions, is that enough for a start? (Re)scaling with different scale factors for x,y and z is more work.

Thanks,
Erwin