Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Hi schultza,

the extracted obstacles are published as Marker type from the teb_local_planner on topic teb_markers. Just add the marker display to your rviz configuration.

Note, the conversion algorithms are still experimental. I implemented them "quickly" in spare-time. So if you observe some issues or if you have some suggestions or better default parameter sets, please let me know :-). Also I need to perform some benchmarks in the future between the different algorithms.

According to your console output, the converter is indeed loaded properly. You can tune some of the parameters using rqt_reconfigure. By the way, better start with a convex hull or line algorithm since calculating concave hulls takes much more time and is often not required.

1 Hz is indeed slow. I recently added a FAQ section to the tutorials with some hints about parameters and their influences on computation time: Link

Hi schultza,

the extracted obstacles are published as Marker type from the teb_local_planner on topic teb_markers. Just add the marker display to your rviz configuration.

Note, the conversion algorithms are still experimental. I implemented them "quickly" in spare-time. So if you observe some issues or if you have some suggestions or better default parameter sets, please let me know :-). Also I need to perform some benchmarks in the future between the different algorithms.

According to your console output, the converter is indeed loaded properly. You can tune some of the parameters using rqt_reconfigure. By the way, better start with a convex hull or line algorithm since calculating concave hulls takes much more time and is often not required.

1 Hz is indeed slow. I recently added a FAQ section to the tutorials with some hints about parameters and their influences on computation time: Link

EDIT #1:

After trying to reproduce your parameter setting I recognized, that the planner wasn't loading the correct parameter namespace. Instead of namespace /.../TebLocalPlannerROS/costmap_converter/CostmapTo* it tried to access /.../TebLocalPlannerROS/CostmapTo*. Consequently, your parameters (even in rqt_reconfigure) did not changed anything. I fixed that in the source code and it is included in the new release I am going to push later today).

Regarding your rviz output: Each occupied cell (lethal obstacle, yellow cell) is transformed into a single point. No conversions are performed at all. The reason is that you are using a low costmap resolution (large resolution value >0.5?), but I configured the default values for resolution of 0.1 meters. Note, the default value of parameter cluster_max_distance is too low. Maybe I can change the default parameters such that they scale better with the resolution of the costmap in the future. However, in my opinion your resolution is way too high. The costmap raytrace algorithm marks a lot of free space in your map as obstacles. This could also make navigating through narrow doors nearly impossible. Additionally, after some tests with the converter and a resolution of 0.5, I am not really satisfied with the results of the current implementation.

I am really interested in your computer specs, if the planner runs with just 1 Hz for you?

Hi schultza,

the extracted obstacles are published as Marker type from the teb_local_planner on topic teb_markers. Just add the marker display to your rviz configuration.

Note, the conversion algorithms are still experimental. I implemented them "quickly" in spare-time. So if you observe some issues or if you have some suggestions or better default parameter sets, please let me know :-). Also I need to perform some benchmarks in the future between the different algorithms.

According to your console output, the converter is indeed loaded properly. You can tune some of the parameters using rqt_reconfigure. By the way, better start with a convex hull or line algorithm since calculating concave hulls takes much more time and is often not required.

1 Hz is indeed slow. I recently added a FAQ section to the tutorials with some hints about parameters and their influences on computation time: Link

EDIT #1:

After trying to reproduce your parameter setting I recognized, that the planner wasn't loading the correct parameter namespace. Instead of namespace /.../TebLocalPlannerROS/costmap_converter/CostmapTo* it tried to access /.../TebLocalPlannerROS/CostmapTo*. Consequently, your parameters (even in rqt_reconfigure) did not changed anything. I fixed that in the source code and it is included in the new release I am going to push later today).

Regarding your rviz output: Each occupied cell (lethal obstacle, yellow cell) is transformed into a single point. No conversions are performed at all. The reason is that you are using a low costmap resolution (large resolution value >0.5?), but I configured the default values for resolution of 0.1 meters. Note, the default value of parameter cluster_max_distance is too low. low. Try to change all parameters with rqt_reconfigure first. Maybe I can change the default parameters such that they scale better with the resolution of the costmap in the future. However, in my opinion your resolution is way too high. The costmap raytrace algorithm marks a lot of free space in your map as obstacles. This could also make navigating through narrow doors nearly impossible. Additionally, after some tests with the converter and a resolution of 0.5, I am not really satisfied with the results of the current implementation.

I am really interested in your computer specs, if the planner runs with just 1 Hz for you?