Many robotic capabilities require spatial information about the robot, the interaction partners, or of objects and places of interest. This information is stored in maps. Even if a-priori information about the robots environment is available, mapping is needed to deal with dynamic and changing environments.
The maps generated by robots are usually used to enable the robot to perform certain tasks like for example autonomous navigation using path planning. Maps also assist an operator of a remotely teleoperated robot in locating the robot in the environment. They do this by providing information of features of interest like corners, hallways, rooms, objects, voids, landmarks, etc. Robotic mapping is interesting for many areas of applications in various scenarios. This could be the classical ser- vice robotics in industry or office environments, robots participating in the normal traffic, or mining, wilderness or search and rescue robotics and of course military robotics.
The quality of the mapping algorithms and the generated maps has to be ensured. In order to be able to specify and test the performance of mapping systems, the result - the maps - have to be analyzed and evaluated in a systematic, repeatable and reproducible way. The focus of this line of research is thus to develop and test algorithms to assess the quality of maps generated by mobile robots. The evaluation of generated maps is very important to be able to make assessments about the performance of mapping algorithms. A topology representation is extracted from the 2D grid maps and compared to the topology of a ground truth map.
Map evaluation was already the topic of my PhD thesis. We are currently working on extending and generalizing the TopologyGraph algorithms. Future work includes work on 3D map evaluation (see Image). Also see this dataset page regarding the maps for the paper on 3D map evaluation at the IEEE International Symposium on Safety, Security, Rescue Robotics 2015.
3D Ground Truth map of the 2013 RoboCup Rescue arena in Eindhoven.