[ros-users] Detecting Glass Walls

[Bill Smart]

On Fri, 2010-04-02 at 12:50 -0400, Eric Perko wrote: 
> Thanks Radu,
> 
> I haven't gotten into a serious literature search yet, but thats my
> next stop. Figured I'd ping this list first and see if any of the ROS
> users out there had already solved this or done some experimenting
> with any methods and found them to be good/bad.

At the risk of offering the hackiest solution, you could think about
putting a strip of tape at the level of the laser plane, along the
glass.  The simple solution is to use opaque tape, which would look
ugly, with high probability, but you might get away with if you get the
colour and size right (and make it look like a feature of the decor).
Alternatively, is there a tape, or glass treatment, that is transparent
to visible light, but reflective in the IR range?

Our experience with a SICK hitting glass has been much the same as
yours.  Most of the beams go through, but some reflect sporadically.
Might there be a way to notice this, and combine it with other
information (from the non-transparent posts) to infer the glass.

-- Bill