A common goal in modern autonomous mobile indoor and outdoor robotics is to achieve a level of semantic understanding of the environment. While researchers have successfully managed to create accurate metric environment representations in the past, true understanding of the environment remains a very challenging task. In order to move towards a semantic representation of areas, rooms, objects, persons etc. in the robot’s vicinity, the robot must first of all be able to robustly extract and classify the corresponding portions from its sensor data.

In this line of research we focus only on data obtained from a 3D laser range finder setup. This is in parallel to the research area “Cognitive understanding of the Environment”, that focuses on the fusion of vision and range data. While visual sensors provide a very convenient way of gathering information about the environment at high rates, the information from laser range finders is more accurate and contains all necessary information for classification on a level of geometric primitives.

The raw data consists of a 3D point cloud enhanced by the laser remission values (i.e. the amount of light reflected by objects). This data is enhanced by artificially generated features and a number of clustering techniques is employed to extract meaningful portions from the data. The obtained clusters can be further processed to obtain geometric primitives, which can in turn serve as input to a learning architecture.