ROS_CANOpen motor synchronization problem

asked 2019-01-12 01:16:17 -0500

jxy
11 ●2 ●3 ●4

updated 2019-01-15 09:07:12 -0500

I am using ROS_CANOpen to control a differential drive robot, and using the "profiled position" drive mode. I used schunk_lwa4p as a template: https://github.com/ipa320/schunk_robo...

When I publish my target position to the topic /arm/joint_group_position_controller/command, sometimes my motors do not start at the same time. Checking my logs, I found that sometimes one motor will begin moving one sync interval later than the other. This is very noticeable because I am having to run at a very slow sync interval of 200ms as a result of this ROS_CANOpen issue: https://github.com/ipa320/schunk_robo... and the seemingly related https://github.com/ros-industrial/ros... (which increasing tx queue length helps but doesn't resolve).

It may be notable that I could not reproduce this problem using "profile velocity" mode.

Thank you for reading

Comments

Cross-posted: ros-industrial/ros_canopen#315.

gvdhoorn ( 2019-01-12 02:25:21 -0500 )edit

Correct. Is that ok?

jxy ( 2019-01-12 07:07:07 -0500 )edit

cross-posting is typically frowned upon, as it leads to split discussions, with duplication of effort as a result.

gvdhoorn ( 2019-01-12 07:58:46 -0500 )edit

Sorry about that. Should I delete one of them? I thought if there was a fix on my end it would get addressed here and if there was a fix needed on the ROS_CANOpen end it would get addressed there.

jxy ( 2019-01-12 14:04:02 -0500 )edit

add a comment

answered 2019-01-16 04:50:25 -0500

Mathias Lüdtke

1596 ●2 ●26 ●26

to control a differential drive robot, and using the "profiled position" drive mode

I don't think that profiled position mode is suitable for controlling a diffential drive. It is is setpoint-based, so the timing might vary between different modules.

I used schunk_lwa4p as a template

While you can reuse some parts of it, I don't think that it is helpful to use it as template. It is meant for schunk hardware with highly specific settings.

This is very noticeable because I am having to run at a very slow sync interval of 200ms

You should really try reduce this, I would target at least 20ms.

Checking my logs, I found that sometimes one motor will begin moving one sync interval later than the other.

The motion should get commanded in the same sync slot. How do you determine the start of the motion?

as a result of this ROS_CANOpen issue: https://github.com/ipa320/schunk_robo...

That's a non-issue, and quite specific the Schunk LWA hardware.

edit flag offensive delete link

Comments

"You should really try reduce this, I would target at least 20ms."

If I set the sync interval lower than 200ms the startup/init sequence fails. The console prints "CAN not ready" and I don't know how to fix. ~10% of the time the init sequence succeeds and after that everything runs properly.

jxy ( 2019-01-16 09:56:31 -0500 )edit

"CAN not ready" should be rare. You can try see whats happening in the candump.. Please make sure that your PDO mappings are correct: http://wiki.ros.org/canopen_402

Mathias Lüdtke ( 2019-01-16 14:09:11 -0500 )edit

PDO mappings look good. I noticed that the "CAN not ready" message correlates in time with the SYNC message going out from the master while the 2nd node on the bus is being initialized. Could this be the root cause? Is it possible to delay SYNCs until after all nodes have initialized?

jxy ( 2019-02-05 17:49:09 -0500 )edit

Implementing the fix discussed here solved my issue: https://github.com/ros-industrial/ros...

jxy ( 2019-02-18 10:33:54 -0500 )edit

add a comment

ROS_CANOpen motor synchronization problem

Comments

1 Answer

Comments

Question Tools

Stats

Related questions

ROS_CANOpen motor synchronization problem edit

Comments

1 Answer

Comments

Question Tools

Stats

Related questions

ROS_CANOpen motor synchronization problem