Gives a complete description of the general methodology of SLAM algorithms (simultaneous construction of maps and definition of its own locations on it), based on 2D-scanning of real rooms. Existing approaches can be classified within the framework of the presented model, and it is also possible to compare them. The following algorithms were considered: Google Cartographer, GMapping, tinySLAM. According to their assessment, the Cartographer and GMapping are more accurate than tinySLAM, and the Cartographer is the most reliable of the algorithms. The problem of quantitative and qualitative estimation of the results of the algorithm of simultaneous construction of maps and determination on it is considered. The quantitative estimation assumes presence of a true trajectory of movement of the robot during record of the test data. A qualitative assessment can be performed only by a person: it is necessary to assess the consistency of the map, the presence of self-intersections, discontinuities of straight lines, and so on.

Authors: Ar. Yu. Filatov, An. Yu. Filatov, A. T. Huletski, D. A. Kartashov, K. V. Krinkin

Direction: Informatics and Computer Technologies

Keywords: 2D laser SLAM, Cartographer, gmapping, tinySLAM, groundtruth

View full article