Qualitative distances and qualitative image descriptions for representing indoor scenes in robotics

Patterns of qualitative concepts are extracted from robot sensors in order to describe the shapes, colours, spatial orientations and topology situations of natural landmarks in the robot environment and also the distance to them. Those qualitative patterns are obtained at a low level sensor data processing and without using training on datasets or learning techniques. A qualitative distance integration approach is parametrized and applied to detect glass windows and mirrors. Corners and columns are detected by the laser sensor and described qualitatively as relevant landmarks. Images taken by the robot camera are described qualitatively for completing the description of the objects located in the robot environment.Experimentation carried out shows that the integration of the information provided enhances the robot perception.

► Patterns of fuzzy distances from sonar and laser sensors identify any obstacle.
► Qualitative shapes, colours, topologies, and orientations describe objects in images.
► Indoor natural landmarks (i.e. corners) are detected and described qualitatively.
► Approach tested on a Pioneer 2dx robot but extendable to other robots.
► Successful results obtained in navigation inside a building with glass windows.