[ros-users] Packages for 2D slam based on floor images

Hi!

I am building a robot that should localize itself on a 2-dimensional
plane by taking straight pictures of the floor and doing SLAM on this data.

Because I don't want to reinvent any wheels, I decided to use ROS and
did the tutorials and read a lot of stuff. Looking through the
repositories I found a lot of packages doing SLAM on laser scan data, 3D
point clouds or stereo cameras.

But does anybody know some packages that do this task in 2-dimensional
space? There are a lot of videos, in which you can see robots building
up the groundplan of buildings. But it is always done by detecting walls
or other obstacles in the environment beside the robot.

I could not find any robots, which are interested in what lies directly
under them. I will have some more detailled surface than carpet, though.

If nobody has done something similar, it would be nice if someone with
an better overview could point out to me, what parts of existing stacks
I maybe could reuse and what I have to write myself.


Probably I want to have a simulation for that robot. I played with
gazebo and stage a bit, but I can't really decide which one is the right
for me. The bot has a simple mechanical model and I primarily need the
feature of getting different sections from a surface texture while moving.

I would appreciate any hints or tips.


Cheers,
David