The wall histogram method

Occupancy maps are a standard method for probabilistic spatial representations of the environment of a mobile robot. They are easy to build and maintain and suffice for many navigation tasks. However, for solving more complex tasks the need to integrate metric occupancy maps with topological or relational spatial information arises. An important step in the automatic extraction of symbolic representations from sub-symbolic data such as occupancy maps is the segmentation of occupancy maps into areas with similar occupancy. Formerly suggested methods usually lead to unintuitive segmentation, which are difficult to match with symbolic concepts like rooms or hallways. We present the wall histogram method which makes use of the concepts and algorithms developed in computer vision, to yield more intuitive occupancy map segmentation.