check the conservation of angular momentum

[yzd0014]

I'm not sure if mojoco keeps track the total angular momentum of all bodies with respect to a reference point just like energy. I'm not sure if subtree_angmom will give what I want.

[yuvaltassa]

Yes, subtree_angmom is what you want. Note that this computation is optional. You need to add the appropriate sensor to your model if you want this to computed automatically.

[v-r-a]

Hello,
I'm a student and MuJoCo user. Here are a few details as I understand:

I'm not sure if mojoco keeps track the total angular momentum of all bodies with respect to a reference point just like energy. I'm not sure if subtree_angmom will give what I want.

You will have to calculate the angular momentum of each body (about its own COM ($h_g$) to start with) yourself. It is not readily available anywhere. But this is the place where it is calculated in the subtree_angmom code.

The quantity readily available is the angular momentum of a chain of bodies (aka subtree) about the subtree's centre of mass. If you access subtree_angmom, you will get the angular momentum of the whole robot about its COM.
If you access subtree_angmom +3*bodyid you will get the angmom of the rest of the robot start from bodyid about the rest of the robot's COM.

(Just reiterating that) there is no difference in the angmom value if your bodyid is 0 or 1 if there is only one tree/robot in the model.

As @yuvaltassa has rightly mentioned, $h_p = h_g + (r_g - r_p) \times (m v_g)$ for a body, where $h_p$ is the angmom about point P fixed w.r.t. the world frame.

[yuvaltassa]

I'm not sure if COM is computed for this body and all its children or only all its children.

The former: the body and all its children. Because the world has no momentum itself, if there is only one subtree in the model, it doesn't matter if you choose the world body or the root body. If you choose a leaf body you will get something else (basically the angmom of the leaf body about its own com). Makes sense?

[yzd0014]

@v-r-a Thank you very much for the info you provided!

[yzd0014]

@yuvaltassa Yes, it makes sense. By reading the reply from @v-r-a, I think I have wrong understanding previously. Please let me know if this is correct. Previously I thought every three entries in subtree_angmom stores angular momentum of a rigid body wrt com, but it actually stores angular momentum of a rigid and all its children?

[yzd0014]

I was able to verify the conservation of angular momentum by adding the extra term as you suggested. Thank you for your help! @v-r-a @yuvaltassa