Evaluation of Modern Laser Based Indoor SLAM Algorithms

One of the key issues that prevents creation of a truly autonomous mobile robot is the simultaneous localization and mapping (SLAM) problem. A solution is supposed to estimate a robot pose and to build a map of an unknown environment simultaneously. Despite existence of different algorithms that try to solve the problem, the universal one has not been proposed yet [1]. A laser rangefinder is a widespread sensor for mobile platforms and it was decided to evaluate actual 2D laser scan based SLAM algorithms on real world indoor environments. The following algorithms were considered: Google Cartographer [2], GMapping [3], tinySLAM [4]. According to their evaluation, Cartographer and GMapping are more accurate than tinySLAM and Cartographer is the most robust of the algorithms.