IRB2400 URDF Seems Wrong

[jbeck28]

Hello!

I've been working on setting up an IRB2400 in ROS2, and along the way I've found something that seems very wrong, and I'm not sure how to proceed with it.

Basically, the way the 2400 URDF is set up is there are lever joints which mimic J3... however, this isn't how an actual 2400 works. When jogging J2 in joint mode on a 2400, the pivot point which is J3 in the URDF would actually take on the value that I think is J2 - joint_lever_b... This can be seen by jogging J2 with a physical IRB2400 and seeing very clearly that the angle at J3 in the URDF changes.

I'm going to work on correcting this but I'm not at the time of writing sure how I'm going to do this.

Is this something anyone else has bumped into?

Thanks,
Josh

[gavanderhoorn]

I've been working on setting up an IRB2400 in ROS2, and along the way I've found something that seems very wrong

First: it's always appreciated when people report things they believe are incorrect with any of the models here. Without feedback, we can't improve them.

Having written that though: this support package is very old (10+ years), and used quite a bit, so if there was something fundamentally wrong with it, I would have expected more users reporting it and/or asking about it. That doesn't mean I'm dismissing your observation though.

(edit: you don't mention it, but I'm assuming you're referring to irb2400_12_155_macro.xacro, as that's the only IRB 2400 variant which (attempts to) model the parallel structure)

Basically, the way the 2400 URDF is set up is there are lever joints which mimic J3... however, this isn't how an actual 2400 works.

That could be, but something to remember: the xacro macro doesn't model how an actual IRB 2400 works.

It models just enough of the kinematics to be able to visualise the state of a real robot and to use that for motion planning with something like MoveIt.

KDL -- the default IK library in many ROS applications -- doesn't / didn't support parallel kinematics, so none of the extra stuff related to the parallel structure and extra mimic joints contribute to the real kinematic chain.

What's left is a regular, as you would expect, serial chain of 6 joints with 5 connecting links. The default driver (ie: abb_driver) then takes care of publishing joint angles that have been corrected for the J2-J3 coupling you mention. That's what the J23_coupled parameter configures (here).

I don't know which driver you are using, but you might want to verify it also takes care of this coupling if needed.

When jogging J2 in joint mode on a 2400, the pivot point which is J3 in the URDF would actually take on the value that I think is J2 - joint_lever_b... This can be seen by jogging J2 with a physical IRB2400 and seeing very clearly that the angle at J3 in the URDF changes.

I'm going to work on correcting this but I'm not at the time of writing sure how I'm going to do this.

My suggestion would be to first make sure there is something actually wrong with the model.

If you could show some screenshots of rviz showing an incorrect pose, or a diagram with some poses you feel are incorrect and show what the correct pose should be, that would help (me).

[jbeck28]

[gavanderhoorn]

It is now a discussion, but it'd be better to keep things on-topic. Especially something like exporting/importing calibration data into URDFs/XACROs, as it is something which is actually a cross-cutting concern (ie: something that happens to be interesting for/with multiple OEMs) and therefore deserves its own thread.

Please start a new thread and remove your related problem here.