2nd International Conference on Computer Science and Engineering, UBMK 2017, Antalya, Türkiye, 5 - 08 Ekim 2017, ss.99-103
The fact that 6-DoF (Degree of Freedom) mobile robots become widespread, makes 2 dimensional Simultaneous Localization and Mapping (SLAM) systems insufficient in order to provide autonomous abilities like exploration and navigation. Thus, 3 dimensional SLAM has become a research subject, and various systems have been developed for the last decade. This paper compares performances of two state-of-the-art 3D SLAM systems: RTAB-Map (Real-Time Appearance-Based Mapping) and Kintinuous. Although both system uses RGB-D sensor for input data and similar aspects in gathering information from the input, their different approaches in implementation details affects performance results. Both systems were tested on several datasets. Experimental results show that RTAB-Map has more success on handling 3-Dimensional SLAM problems so that trajectory of RGB-D sensor can be estimated with remarkably less error than Kintinuous. The reason can stem from the fact that Kintinuous is more dependent to depth data which is less reliable than RGB data while RTAB-Map uses RGB data more effectively.