Hi,

I am testing move_base with sensor data from our stereo vision sensor.  
Thus, I have a PointCloud as observation source for the costmap. I  
want to use the parameters ~<name>/max_obstacle_height (=0.7) and  
~<name>/<source_name>/min_obstacle_height (=0.05) to mark the  
appropriate obstacles in the costmap - assuming that the costmap  
considers values smaller than min_obstacle_height as ground (= free  
space).

This seems to work for a static environment generally. But as soon as  
e.g. somebody walks through the picture, his trace is marked as  
obstacle cells. After the person leaves the picture I would expect the  
costmap to clear these cells, because the corresponding "z"-values  
from the point cloud are less than min_obstacle_height. Instead the  
cost map obviously ignores values smaller than min_obstacle_height.

Is there any way to use the values smaller than min_obstacle_height to  
clear the former obstacle cells?

Regards,
Sabrina