0
$\begingroup$

Rosanswers logo

Hi,

I am seeing my urdf model, and all tf frames specified by robot_state_publisher, lag behind the actual robot motion (which the transform /odom->/base_footprint appears to track perfectly). The lag is around 5 seconds.

I have also tried with a simplified .urdf model that uses only fixed joints (and thus removes /joint_state publications as a potential cause). No help.

I have set the ~publish_frequency param of robot_state_publisher to 10 hz, and I have a static_transform_publisher publishing at 10 hz to link the top-level urdf frame (/chassis) to the /base_footprint frame published by the rest of the localization system.

Any bright ideas? Thanks!

PS - using diamondback and Ubuntu 10.10


Originally posted by Nick Armstrong-Crews on ROS Answers with karma: 481 on 2011-08-26

Post score: 0


Original comments

Comment by AHornung on 2011-08-26:
Is all running on the same machine, or are parts running on a laptop attached to a robot? Are all involved clocks in sync (ntp)?

Comment by David Lu on 2011-08-26:
Is this over a wireless connection?

Comment by Nick Armstrong-Crews on 2011-08-26:
Thanks for pointer, but... Nope, I actually think my problem might be a direct result of the fix described there; all elements of my robot model move together atomically, and all with the same 5-second lag. I suppose I should double-check tf pub rates with tf_monitor...

Comment by Eric Perko on 2011-08-26:
Is your behavior similar to that described here: http://answers.ros.org/question/110/has-there-been-progress-for-urdf-transform-jitter ?

$\endgroup$

1 Answer 1

0
$\begingroup$

Rosanswers logo

Yup, good suggestions! Indeed, I am operating over wifi, and today I don't see the problem (after machines have power cycled and been connected to the internet).

Diagnosis: must have been a time sync issue.

Thanks!


Originally posted by Nick Armstrong-Crews with karma: 481 on 2011-08-26

This answer was ACCEPTED on the original site

Post score: 2


Original comments

Comment by Nick Armstrong-Crews on 2011-08-26:
I also want to note that it didn't matter on which machine I ran robot_state_publisher; as long as at least one node is listening to /tf topic on each machine, all those transforms get published and sent over the wire and timestamps compared. So this stymied my initial debug efforts. Watch out!

$\endgroup$

Your Answer

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