fix behavior for multi-turn revolute joints

[codebot]

Previously, the behavior of revolute joints was unexpected when
the joint limits extended beyond [-pi, pi] due to unwrapping
happening in a few places in the angles package, which was called
from the joint position controller. This PR checks to see if the
range of motion extends beyond [-pi, pi] and, if so, goes directly
to the (clamped) target, rather than trying to unwrap the current
or target angles to find the shortest path.

[graiola]

Hello,

Could you show me a situation where the behavior is unexpected?
I am testing the code with and without the pr, but I don't see any difference.
Is the issue related to a different version of angles package? (I am testing with kinetic)

Thanks

[graiola]

ping @codebot

[IanTheEngineer]

@codebot @graiola I've experienced the same issue within the default_robot_hw_sim interface from gazebo_ros_control:
https://github.com/ros-simulation/gazebo_ros_pkgs/blob/kinetic-devel/gazebo_ros_control/src/default_robot_hw_sim.cpp#L338-L342
Any measured angle from Gazebo that is outside of -pi to pi is forced to be within that range. This is a problem for Sawyer, as the seventh joint has a range -4.15 to 4.15 rad. I can handle fixing Sawyer, but we should address this for the larger community.

Rather than fixing it here, shouldn't we just modify the angles shortest_angular_distance_with_limits to handle larger joint limits properly?

[bmagyar]

We love test driven development here. (I also like Sawyer ;)) Could you please add a test that replicates this? An alternative version of rrbot could work.

…

On Tue, Apr 24, 2018, 16:53 Ian McMahon ***@***.***> wrote: @codebot <https://github.com/codebot> @graiola <https://github.com/graiola> I've experienced the same issue within the default_robot_hw_sim interface from gazebo_ros_control: https://github.com/ros-simulation/gazebo_ros_pkgs/blob/kinetic-devel/gazebo_ros_control/src/default_robot_hw_sim.cpp#L338-L342 Any measured angle from Gazebo that is outside of -pi to pi is forced to be within that range. This is a problem for Sawyer, as the seventh joint has a range -4.15 to 4.15 rad. I can handle fixing Sawyer, but we should address this for the larger community. Rather than fixing it here, shouldn't we just modify the angles shortest_angular_distance_with_limits <https://github.com/ros/angles/blob/master/angles/include/angles/angles.h#L215> to handle larger joint limits properly? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#264 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADXH4asEUDK3vgJ15idkqB7I_9jJj9sTks5tr0qGgaJpZM4M2QTy> .

[IanTheEngineer]

@bmagyar hmm, on further investigation, I'm experiencing the joint limits being reported directly from Gazebo7 as being truncated to a reduced range. That doesn't mean that the shortest_angular_distance_with_limits is correct, but rather that this whole thing requires more investigation. I'll add more color (and possibly unit tests) to this as I grok the underlying issue better.

[bmagyar]

I think we addressed this in #477 and #466 , please reopen if not.