Fix sjtc to correctly aborts on violation of constraints #810
Conversation
|
@destogl as discussed this contains our updates on the SJTC. That could be a base for re-addressing ros-controls/ros2_controllers#301. |
fmauch
force-pushed
the
update_sjtc_rolling
branch
from
60a0acd
to
39b5d60
Compare
fmauch
force-pushed
the
update_sjtc_rolling
branch
from
39b5d60
to
2932698
Compare
| if (normalize_joint_error_[index]) { | ||
| // if desired, the shortest_angular_distance is calculated, i.e., the error is | ||
| // normalized between -pi<error<pi | ||
| error.positions[index] = angles::shortest_angular_distance(current.positions[index], desired.positions[index]); |
There was a problem hiding this comment.
When should this be actual. I think we always would like to catch the absolute angle. More or less all the UR joints have a range of +-2pi if the physically and UR3e's wrist3 joint can rotation infinty of resolutions
There was a problem hiding this comment.
That particular line is (as most of the update function) from the upstream JTC. This is used only if
- closed loop control is activated.
angle_wraparoundis set totruein the controller configuration.
Both isn't true inside our controller configuration and also has to be explicitly activated when desired. I would vote for sticking as close to the upstream code as possible because
a) we want to make our lives easy to adapt to future changes
b) we want to switch to the upstream JTC as soon as the scaling feature got merged there.
you mean preempt? Yes, we could add that, as well. |
Yes, to check that the about function also works. |
|
|
Merging since the necessary changes in ros2_controllers have just been released. |
urrsk
changed the title
This aims at fixing both #800 and #804.
In a first step I'd like to add a test reproducing the error reported in #800 and add a fix afterwards which I assume to be the same fix as #804 requests.