Enabling autonomous asbestos removal: Towards automatic unbiased evaluation of estimated maps

S. C. Eddine,Tobias Haschke,Sascha Weil,B. Corves
DOI: https://doi.org/10.1109/ICINFA.2017.8078892
2017-07-01
Abstract:Autonomous asbestos removal requires a highly reliable and precise map of the robot's environment. These maps are usually obtained through Simultaneous Localization and Mapping (SLAM) yielding maps that vary in accuracy and exhibited noise. Given the high number of available SLAM algorithms on the one hand and the map's significance for the asbestos removal process on the other hand, a comparative evaluation of these estimated maps is required. In this paper, we present a method to assess their accuracy as well as their noise by comparing them against a given reference map. We measure accuracy by means of Global Map Deviation (GMD) and Local Map Deviation (LMD), whereas the noise is measured through Mean Noise Distances (MND). Our method allows for an automatic, highly-precise overall evaluation of different maps without subjective bias introduced by the individual operator. Experimental results clearly demonstrate the efficiency and validity of our approach.
What problem does this paper attempt to address?