| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
The gmapping system misses glass surfaces since it makes use of laser data. Unfortunately, we have an environment that has quite a few glass walls/doors. Has anyone run into this issue? If yes, how do you folks suggest we handle it? I can always modify the map manually in an image editor. Is there a better way?
There's a related question here, which links to some answers. The problem is that if these glass walls are actually invisible in the laser scan, the local trajectory planner can still try to go through them, even if you manually paint the walls into the global map. So you'll need something like layered costmaps which contain the "no go zones".
Or you could do what we did with our (small) glass door, which is stick a couple sheets of paper onto it at a certain height... :-)
Or try additionally using an ultrasound ranger for local obstacle avoidance, if your robot happens to have one (not tried, but should work).
Thanks for your answer :). We thought of covering up glass surfaces already. The problem is that our environment is an entire floor of the building, and covering up all the glass surfaces will be tedious. I was hoping I can somehow detect glass and augment the maps.
(I've observed that the laser does seem to detect glass when incidence is near normal.)
Traditionally people have just drawn in a line in an image editor. You probably want to make the line thin as possible so that it has the least effect possible on other components, like AMCL, or, possibly give an unedited map to AMCL for localization.
Asked: 2013-08-11 15:23:42 -0500
Seen: 597 times
Last updated: Aug 12 '13
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
