Meeting ROS-I specs #46
Conversation
|
Ok, probably I won't be able to tackle all the list above in this PR... but let's see where it can get. |
carlosjoserg
force-pushed
the
ros-i-specs
branch
from
f5c544d to
fbb829a
Compare
|
Rebased, merged, etc... waiting for travis... |
…sual. The lwr4plus/collision folder is created, but recall that in this package the collision geometry is formed of basic primitives within the urdf model.
… the new joint names.
… controllers now are updated at a fixed sampling rate defined in the script being used in the robot
Most likely, I will discard the different pre-configured modes I had in mind, and leave the kuka ones: joint position = 10 cartesian impedance = 20 joint impedance = 30
Not working yet
carlosjoserg
force-pushed
the
ros-i-specs
branch
from
ffe95b8 to
28aeb14
Compare
|
Finally, this is ready to merge... This PR contains a lot of changes, so I'll leave it open one more day in case someone has something to say, and I'll merge tomorrow morning otherwise. |
|
And I'm going to add the pending items from the list above as issues. |
| //the controller manager | ||
| controller_manager::ControllerManager manager(&lwr_robot, lwr_nh); | ||
|
|
||
| // run as fast as possible | ||
| while( !g_quit ) | ||
| { | ||
| // get the time / period | ||
| if (!clock_gettime(CLOCK_REALTIME, &ts)) | ||
| if (!clock_gettime(CLOCK_MONOTONIC, &ts)) |
There was a problem hiding this comment.
Using CLOCK_MONOTONIC to set the current time in the call to update has caused some issues in other packages before (e.g. scheduling trajectories far into the future by the joint trajectory controller, publishing messages with timestamps incompatible with other ROS nodes). See conversation from this comment.
Both the ur_modern_driver and the ros_control_boilerplate have settled on using the monotonic clock to compute the elapsed time, but setting the current time using ros::Time::now().
There was a problem hiding this comment.
I see... just reading all discussions, jumping from issue to issue in a wikipedia navigation style.
I've never experienced any issue here though, but good to have this in the background just in case.
Thanks for the review ;)
| @@ -99,7 +99,7 @@ int main( int argc, char** argv ) | |||
| while( !g_quit ) | |||
| { | |||
| // get the time / period | |||
| if (!clock_gettime(CLOCK_REALTIME, &ts)) | |||
| if (!clock_gettime(CLOCK_MONOTONIC, &ts)) | |||
This PR contains changes according to ROS-I specs:
ToDo:
Note: Once we have all interfaces in
masterthen we can do this in a different PRNote: I don't think we will find better geometries... the current ones do not actually match perfectly the real ones, but I think we can live with that.
Note: This has to be worked in a different PR that tackles the Stanford FRIL iface.