Closed aPonza closed 2 years ago
@aPonza, could you check, whether https://github.com/ros-planning/moveit_tutorials/pull/387 fixes this issue? If so, please close this issue.
It doesn't seem to, the output is mainly the same. Just to double check, I had to clone moveit, moveit_resources and moveit_tutorials, stash my changes in panda_moveit_config (which wasn't modified) and rebuild the workspace to get this fix in, correct? I don't understand how a fix in a tutorial would affect the issue I'm seeing with the floating/fixed joint to be honest, but I'll have time to look a bit more into it in some days' time hopefully.
https://github.com/ros-planning/moveit_tutorials/pull/387 addressed an issue in the motion_planning_api_tutorial
, which directly uses the C++ API of MoveIt, circumventing the ROS interface provided via the move_group
node.
Please provide exact commands to reproduce your issue. You didn't even mention, which of the tutorials you tried.
I'm not trying tutorials, I'm using my own nodes and launch file of a pick and place process. It's exactly the same roslaunch command with floating vs fixed. I'll make it reproducible, you're right. It'll take some time, I'm working on another thing at the moment.
@rhaschke Maybe we can close this issue since kinetic is EOL?
The virtual joint should be type fixed
. This doesn't require a virtual joint broadcaster, which is not provided by the real robot.
tl;dr 6ae03fac5cf19f830f60e10a5bf669e0148a8e1f doesn't fix for me the virtual joint issue that was fixed for #22, and it's not the case of just propagating the publisher to another launch file.
The issue
With the same exact process running, checking out panda_moveit_config 0.7.1 runs fine but the current kinetic-devel gives erratic behaviour and this diff in the output (I tried to keep it as short as possible:
Basically what I see is kinetic-devel giving many more collision states and joint constraint violations without any physical differences between the two plannings.
What I tried
Skimming the commits that happened between the two tags (convenience link to comparison), I can see:
kinematics_solver_attempts
: should be irrelevantUnable to update multi-DOF joint 'virtual_joint': TF has no common time between '/world' and 'panda_link0'
(which is incidentally different from this).Seems to be this this problem but I tried to apply the fix to my own higher level launch file or directly to panda_control_moveit_rviz, and it doesn't solve the issue, whereas like the comment says if I change
floating
tofixed
it does work.EDIT1: The addition of the virtual_joint_publisher removes these two messages:
but produces no further change in behaviour:
Insufficient states in sampleable goal region
,Orientation constraint violated for link 'x'
,More than 80% of the sampled goal states fail to satisfy the constraints imposed on the goal sampler.
, etc as before.EDIT2: