0
$\begingroup$

Rosanswers logo

It sounds horrid, and that is exactly what it looks like :p

I have a visual robot model, now only consisting of a base and a head. Both are separate links and they have their separate mesh. The head is attached through continuous joints in the neck. If I now drive with the robot, the base will move seemingly smooth, however the head will detach from its link occasionally and drag behind a bit. See the picture for more information.

If I change the joints to fixed joints, this does not happen. If I decrease the refresh rate of the joint publisher node, it gets worse, if I increase the rate it gets better, but you still notice some detachments on occasion. Is there anything I can do about this?

image description

Best regards, Hans


Originally posted by Hansg91 on ROS Answers with karma: 1909 on 2014-04-17

Post score: 0

$\endgroup$

1 Answer 1

0
$\begingroup$

Rosanswers logo

This typically happens if you connect to a robot via a WiFi connection and the available bandwidth is maxed out, leading to tf transform information getting dropped or arriving with delay (see here for an example of that). Another reason could be wrong time synchronization between involved machines.


Originally posted by Stefan Kohlbrecher with karma: 24361 on 2014-04-17

This answer was ACCEPTED on the original site

Post score: 0


Original comments

Comment by Hansg91 on 2014-04-17:
Hmm quite strange, I'm just running this locally, using robot_state_publisher and joint_state_publisher. I expect those to have their usage of time stamps correct. What I find most weird is that you define links relative to other links, how can child links then move independently?

Comment by Hansg91 on 2014-04-17:
To add to my comment, I'd expect that if I only update the base_link for instance, the other links are automatically updated as well, since they have indirectly changed as well. This apparently does not happen in ros?

Comment by Hansg91 on 2014-04-18:
I just tried with the tutorial urdf file, changing the joint to a continuous joint and running joint_state_publisher, same thing happens. In addition, the same thing also happens to the tf axes, so it is not just the visual model.

Comment by Stefan Kohlbrecher on 2014-04-18:
You can think of the robot model as using tf but additionally drawing the robot geometry, so if something is wrong with transforms, it will always affect both. Your system is probably not maxed out CPU wise when you do your tests?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.