Home | Tutorials | Wiki | Issues
Ask Your Question
1

Inertia again: significant figures

asked 2013-10-06 08:21:45 -0500

Arn-O gravatar image

updated 2013-10-06 14:06:04 -0500

Hi.

I'm facing really bad bugs which seem to be related to inertia. I'm trying to understand the issue with Gazebo and found that the inertia tensors are [0] for the smaller parts:

image description

This is consistent with my bugs, since with the large arm parts, it works more or less, but with the gripper, that becomes crazy. The small parts of the arm and the gripper have 1e-4 values for the inertia tensor.

Here is a short video:

http://www.youtube.com/watch?v=3T2s96...

The joint 3 is moved to 1.0 rad, the next links should stay in line (at 0 rad), but they don't. If you stay long enough (several tenth of second), they finally go to there stable position.

So I wonder how many figures are considered in Gazebo. What are the best practice for simulating this kind of robot? (we could imagine to use a simple kinematic simulation for this part of the body, as they have a low contribution to the overall behavior)

Any inputs will be helpful.

This question is related to:

http://answers.gazebosim.org/question...

edit retag flag offensive close merge delete

Comments

Can you please describe the bugs.

nkoenig gravatar imagenkoenig ( 2013-10-06 10:56:24 -0500 )edit

YouTube video included in the text. It looks like the upper part of the arm has a very huge inertia.

Arn-O gravatar imageArn-O ( 2013-10-06 14:07:01 -0500 )edit

1 Answer

Sort by ยป oldest newest most voted
0

answered 2013-10-09 15:06:21 -0500

Arn-O gravatar image

The main bug was due to the inertia of the fingers:

  <inertial>
    <mass value="${finger_right_link_mass}"/>
    <inertia ixx="0.00001" ixy="0.000" ixz="0.000" iyy="0.00001" iyz="0.00001" izz="0.00001"/>
  </inertial>

I changed the values to 0.0001 (x10) and the behavior is far more realistic. So i guess that there is some kind of limit there.

edit flag offensive delete link more

Comments

If inertia values are too small, especially when connected to links that have a much large inertia, then simulation will have unexpected results. I think you found a good solution to your problem.

nkoenig gravatar imagenkoenig ( 2013-10-09 23:15:38 -0500 )edit

Thx Nate. Is there a rule for "too small"? Is it possible to cancel this effect, since the overall contribution of the finger inertia is probably negligible?

Arn-O gravatar imageArn-O ( 2013-10-11 12:45:29 -0500 )edit
2

A good rule is to make connected links have less than one order of magnitude difference in their intertias.

nkoenig gravatar imagenkoenig ( 2013-10-11 16:48:44 -0500 )edit
Login/Signup to Answer

Question Tools

Stats

Asked: 2013-10-06 08:21:45 -0500

Seen: 501 times

Last updated: Oct 09 '13