We present a graph representation that fuses information from the robot
sensory system and an emergency-map into one. Emergency-maps are a support
extensively used by firemen in rescue mission. Enabling the robot to use such
prior maps instead of starting SLAM from scratch will aid planning and
navigation for robots in new environments. However, those maps can be outdated,
information might be missing, and the scales of all rooms are typically not
consistent. To be able to use them during robot navigation while correcting the
map's errors is an important problem for firemen.
We use corners as a common landmark, between the emergency and the robot-map,
to create a graph associating information from both types of maps. The graph is
optimized, using a combination of robust kernels, fusing information from the
emergency and the robot-map into one map even when faced with scale
inaccuracies and inexact start poses.
Experiments on an office environment show that, we can handle up to 71% of
wrong correspondences and still get the expected results. By incorporating the
emergency-map's information in the robot-map, the robot can navigate and
explore while taking into account places it hasn't yet seen. We also show that
the emergency-map is enhanced by adding information not represented such as
closed doors or walls not represented.