An evaluation of Lidar-based 2D SLAM techniques with an exploration mode

Abstract

SLAM is a fundamental problem in robotic field and there have been many techniques on it. It is necessary to give an insight on weakness and strength of these techniques specific to the intended final application. This paper presents a study of three most common laser-based 2D SLAM techniques: Gmapping, KartoSLAM and Cartographer. Each technique was applied to construct maps combined with autonomous exploration. All the approaches have been evaluated and compared in terms of inaccuracy of constructed maps against the ground truth. In order to draw conclusions on the performance of the tested techniques, a metrics of average distance to the nearest neighbor (ADNN) was applied. Moreover, the computational load of each technique is examined.