Revision history [back]

Thank you for updating your question with content of your parameter files and for posting the video.

After investigating your parameters, I think that your local costmap configuration causes the high computation time. Your local costmap consist of a static map (static map: true) which is not recommended. In that case the width and height parameters are ignored, leading to a local costmap representing your whole environment. The teb_local_planner treats each occupied costmap cell as single point obstacle (if no costmap conversion plugin is activated, which is the default). With that in mind, it is not surprising that the performance is poor, since each obstacle causes a huge amount of distance calculations during optimization.

My current local costmap configuration contains the following parameters:

static_map: false
rolling_window: true
width: 5.5
height: 5.5
resolution: 0.1

I hope this resolves your issue with huge computation times. If you still have problems with your performance afterwards, there are other parameters that influence the computational resources (number of optimization iterations, area/distance behind the robot that should be checked for obstacles, temporal trajectory resolution, ...). Please let me know in that case. Some hints are also given in the tutorials.

PS: A local costmap resolution of 0.025 is not feasible for the planner on most computers, but it is also not necessary. I am running the planner with a resolution of 0.1 (similar to your 3rd example in the video).

PPS: You might try out the new version 0.3 of the planner. Until it is available in the official repos, you can checkout the master branch of the github repository. Parameters and Tutorials are already updated to be compliant with version 0.3 (some minor parameter changes and a new tutorial).