search

moveit / moveit_ros

THIS REPO HAS MOVED TO https://github.com/ros-planning/moveit
69 stars 118 forks source link

No planning scene updates in Rviz after #716. #736

Closed dornhege closed 7 years ago

dornhege commented 7 years ago

I run a robot in simulation, start move_group and Rviz. 

roslaunch pr2_gazebo pr2_empty_world.launch 
roslaunch pr2_moveit_config move_group.launch 
rosrun rviz rviz 
roslaunch pr2_teleop_general pr2_teleop_general_keyboard.launch

The planning scene initially appears, but when moving the robot head via teleop is never updated. The robot state doesn't change and the Octomap isn't updated. This seems quite basic, can someone confirm this is an issue?

If I go back to before #716 everything works as expected. I haven't followed, what exactly is happening there, but as it deals with updates I suspect, but cannot confirm that this is the cause. The newPlanningSceneCallback seems to never be called, although rostopic hz shows messages on the monitored_planning_scene topic.

I've tried with indigo-devel and also merging in #728 in its current state.

v4hn commented 7 years ago

do you have a simple way by which I could reproduce this bug?

dornhege commented 7 years ago

The most objective thing I have found besides "looking" is to put a printout in planning_scene_monitor.cpp newPlanningSceneCallback. Then just run moveit and the RViz plugin. This printout should (as far as I understand everything) always come, when things are working correctly. To be safe put rostopic hz on the topic set in rviz (i.e. /move_group/monitored_planning_scene). Compare indigo-devel and ros-planing/moveit_ros@ae51d07d5b1b848d87acb8c7dacb078b17d9c5ca.

dornhege commented 7 years ago

@v4hn : The comment in #728 about the spinner might be the reason for this. The newPlanningSceneCallback is called via the spinner. If that is in the "wrong" thread and something doesn't work as expected, messages might never be dispatched.

dornhege commented 7 years ago

Digging through RViz warnings I found this:

[1470311817.138041964, 2197.807000000] ros.roscpp: AsyncSpinnerImpl: Attempt to start() an AsyncSpinner failed because another AsyncSpinner is already running. Note that the other AsyncSpinner might not be using the same callback queue as this AsyncSpinner, in which case no callbacks in your callback queue will be serviced.

v4hn commented 7 years ago

On Thu, Aug 04, 2016 at 05:10:29AM -0700, Christian Dornhege wrote:

Digging through RViz warnings I found this:

[1470311817.138041964, 2197.807000000] ros.roscpp: AsyncSpinnerImpl: Attempt to start() an AsyncSpinner failed because another AsyncSpinner is already running. Note that the other AsyncSpinner might not be using the same callback queue as this AsyncSpinner, in which case no callbacks in your callback queue will be serviced.

I suppose this means seals the revert question then. I'll have a look at it this evening.

dornhege commented 7 years ago

Fixed with #742 .