setInertia Values In Python Script

[Brad Humphreys]

I am attempting to set the inertia tensor of a body in a Python Script. I can tell from the docs that I need to use a Mat33 for SimTK. Not sure how to do this though.

For example, to set the center of mass I just use:

cmVec = modeling.ArrayDouble.createVec3(1, 2, 3)
block.setMassCenter(cmVec)

What do I use for setInertia?

[Ayman Habib]

Hi Brad,

SimTK namespace classes are not exposed to scripting, instead we have utility methods that take modeling.ArrayDouble and converts them to SimTK::Inertia. E.g.

Code: Select all

pelvis = m.getBodySet().get('pelvis')
unitInertia = modeling.ArrayDouble(1.0, 6)
pelvis.setInertia(unitInertia)

A similar getInertia() method exists.

Hope this helps,
-Ayman

[Brad Humphreys]

Ayman,

Thanks much for the quick reply. Your code works, but it sets all of the inertia terms (the entire tensor) to the value of 1.0 (not physically practical). How do you set each of the values?

I tried your code but with:
unitInertia = modeling.ArrayDouble([10.0,10.0,10.0,0,0,0],1) #A typical/practical Inertia

I get an error that the first argument must be a double.

Thanks,

Brad

[Yannik Schröder]

There is probably an easier way to do it, but you could try to first create the ArrayDouble and then edit the values of the array via the set(index, value) method (e.g. unitInertia.set(3, 0.0)).

Regards,
Yannik

[Brad Humphreys]

Yannik, Ayman,

That works beautifully! So the code becomes:

Code: Select all

pelvis = m.getBodySet().get('pelvis')
unitInertia = modeling.ArrayDouble(1.0, 6)
unitInertia.set(0,10.0)  
unitInertia.set(1,10.0)
.....
pelvis.setInertia(unitInertia)

Thanks!

[Brad Humphreys]

On a related note, I understand that Inertia are with respect to the center of mass. But is the directionality (xx, yy, zz. xy, etc.) of the tensor with respect to the global space coordinate system of the model?

As I don't see a means to define a body coordinate system, I believe this is the only means to provide directionality for the inertia tensor. I did look through the forum and the location question is answered (the body CM) but not the direction.

Thanks!

Brad

[Michael Sherman]

Hi, Brad.

The inertia matrix is defined in the local frame of the body. The reason there is no way to specify a body's local frame is that the local frame defines the body. Every body begins as just an unadorned body frame B. Then you add geometry and mass properties in that frame. Later the multibody system positions the frame B with respect to inertial space, taking the geometry and mass properties along with it.

Regards,
Sherm

[Brad Humphreys]

Sherm,

Thanks for the response. Just to clarify a bit more. I am building the bodies "from scratch" using python scripting. When the unadorned B frame is placed, it would be aligned to the inertial space coordinate system, correct? So when I set the inertia values they end up initially aligned to the global inertial frame. Your point is then later as constraints/joints/motions drive the body positions, this frame of course changes. Am I interpreting that correct?

Regards,

Brad

[Michael Sherman]

No, you're thinking too hard about this Brad! A body frame doesn't have a placement in space until you have built a set of generalized coordinates for the multibody system containing it. The inertia properties you specify are simply aligned with the body frame -- there is no need to talk about the placement of the body in space at that point. You can read more about this in the Simbody theory manual, chapter 2 if you want: https://simtk.org/docman/view.php/47/23 ... Manual.pdf.

Most likely you are thinking about bodies whose motion is parameterized in Cartesian space, and then restricted by constraints to form joints, like in ADAMS for example. OpenSim is an internal coordinate code -- the generalized coordinates that define the joints are the coordinate system; there is no underlying Cartesian system.

Regards,
Sherm

[Brad Humphreys]

Sherm,

Thanks - the light bulb just finally lit. And this is for importing our ADAMS models into OpenSim. What is interesting, is even with the completely different "build-up" methodologies of OpenSim and Adams and my lack of understanding that till now, the models import nicely and I have it working for all of the base joint types in OpenSim. I was just concerned with the inertia values, but I may have to go back and think of model topology issues that could possibly be caused by the differences here.

Thanks,

Brad