turtlebot stucks in free space while navigation
While testing turtlebot navigation with stage simulation I found interesting behavior. Robot gets stuck in empty space and starts to rotate to clear costmap. Could anybody explain why robot gets stuck in empty space?
[ WARN] [1400749002.714297597, 125.900000000]: Clearing costmap to unstuck robot.
[ WARN] [1400749013.113075404, 136.300000000]: Clearing costmap to unstuck robot.
move_base configuration and video
You can test with turtlebot_stage package.
Installation
> wstool init . https://raw.githubusercontent.com/turtlebot/turtlebot/hydro-devel/turtlebot.rosinstall > wstool merge https://raw.github.com/robotics-in-concert/rocon/hydro-devel/rocon_app_platform.rosinstall > catkin_make
Start
roslaunch turtlebotstage turtlebotin_stage.launch
Asked by jihoonl on 2014-05-21 23:10:19 UTC
Comments
Seems like a local planner problem.
Asked by David Lu on 2014-05-22 14:24:36 UTC
Yeah.. do you have any suggestion to track down the problem? I see this behavior with real robot and would like to solve it.
Asked by jihoonl on 2014-05-22 15:31:04 UTC
I retagged with eband, so maybe someone with that expertise can help you.
Asked by David Lu on 2014-05-22 16:10:55 UTC
It uses trajectory local planner which is given by default navigation stack. It is just switchable to eband but it used trajectory local planner while I test.
Asked by jihoonl on 2014-05-22 17:52:56 UTC
Can you please edit your question to include the precise launch file you are using to get the described behavior?
Asked by David Lu on 2014-05-27 09:34:13 UTC
The question is updated. Let me know if you need more information.
Asked by jihoonl on 2014-05-28 16:27:51 UTC
I also use my turtlebot to navigate with eband_local_planner. Yesterday, I removed a base_local_planner_params.yaml. I need to have both base_local_planner_params.yaml and eband_planner_params.yaml?
Asked by Ken_in_JAPAN on 2014-05-29 06:28:27 UTC
I watched your video. The movement you point out is interesting. As a number of particle is getting a few around the turtlebot, turtlebot can localize. but I think a turtlebot's rotation means turtlebot's localization. That's turtlebot lost the position of itself. Am I wrong?
Asked by Ken_in_JAPAN on 2014-05-29 06:45:03 UTC
To answer first question, no. It depends on what planner you uses. You can choose either elastic or default planner with is_elastic_planner arg.
Asked by jihoonl on 2014-05-29 19:56:51 UTC
For 2nd question, that is what I would like to know. Turtlebot might have lost its position like you said. It might be another reason.. I would like identify what the problem is and find a solution because I have seen similar behaviors with real robots.
Asked by jihoonl on 2014-05-29 20:00:04 UTC
mmm... I have a guess on what's going on: local planner gets pissed-off of going in an opposite direction of the goal, and chooses turn toward the goal instead of keep following the path. After a while of this happening (8 seconds), the oscillation detector starts a clear costmaps rec. behavior
Asked by jorge on 2014-05-31 09:27:10 UTC
What happens gets pretty clear by showing the /move_base/TrajectoryPlannerROS/cost_cloud on RViz. However... I have no solution : ( Someone with more expertise in base_local_planner could do. Btw, eband works because is based in a fully different approach.
Asked by jorge on 2014-05-31 09:30:43 UTC