Home | Tutorials | Wiki | Issues
Ask Your Question
0

Simple model always becomes unstable

asked 2019-08-13 05:13:56 -0600

Sietse gravatar image

updated 2019-09-02 08:59:54 -0600

kumpakri gravatar image

Hi list,

I have a model with 5 revolute joints, nothing special. It is a kind of shoulder and elbow construct. When started it seems stable, but after a few seconds it becomes violantly unstable. This all with default settings. How can it be that this does not work with gazebo, what am I doing wrong?

Tried several versions of gazebo, 9 and 10, on Ubuntu and Debian (from source), but the behavior always is the same.

The model can be found on my github page. There is the boot.urdf that I created and the boot.sdf generated from it.

Please see instable.mp4 for the instabilities.

Update: This model is just a simplified version of my intended model, which can be found here. I created the simplified model to solve the instabilities and crashes of gazebo first.

The inertia values of all 1 kg links a calculated using the values in <visual> using the standard formulas. The values are all realistic physical values, so I assume they should work with gazebo. If I however change the first link from 1 to 3 kg, with correct inertia values, it becomes stable.

But that is no solution for my much more complex model. So my strategy is to first get the simplified model to work.

Thanks in advance, Sietse

PS. This is actually a question I asked before, but is still unresolved. I hope that now maybe someone else can help me in some way!

edit retag flag offensive close merge delete

Comments

How did you arrive to the inertia values? Maybe they are too small. Can you try bigger values?

kumpakri gravatar imagekumpakri ( 2019-08-13 08:04:38 -0600 )edit

Thanks for the reply. Please see the update.

Sietse gravatar imageSietse ( 2019-08-14 03:57:18 -0600 )edit

If you use controllers for the joints, inspect whether the PID values make sense. In case the adjustment of the inertials solved the instability, then adjusting PID values should do the same. It's usually one of those two that need to be fixed.

kumpakri gravatar imagekumpakri ( 2019-08-14 05:34:55 -0600 )edit

Note that there are no controllers here. I just start gazebo and the model in itself becomes unstable.

Sietse gravatar imageSietse ( 2019-08-14 05:54:53 -0600 )edit

1 Answer

Sort by ยป oldest newest most voted
0

answered 2019-08-27 12:51:03 -0600

updated 2019-08-27 12:51:37 -0600

Your model is stable if I increase the number of solver iterations from 50 to 100, although I'm not sure why it's unstable in the first place.

Side note: you don't have any collisions, just FYI in case that's a mistake. I think that should be unrelated to this issue though.

image description

edit flag offensive delete link more

Comments

Thanks! This works, but my complete model, with 31 revolute joints, still crashes in a similar fashion, even with iterations set to 200. Sometimes it takes 30 seconds or so, but in the end it always crashes violantly. With some added friction in the joints the model almost comes to rest, but then suddenly explodes just as violenty, and gazebo also has to be restarted. So trying to make the simple model work with default iterations still seems to be the best strategy. But how?

Sietse gravatar imageSietse ( 2019-08-28 03:48:35 -0600 )edit
Login/Signup to Answer

Question Tools

1 follower

Stats

Asked: 2019-08-13 05:13:56 -0600

Seen: 92 times

Last updated: Aug 27