@article{article_1634390, title={Performance analysis of different SLAM packages in autonomous robots using structural similarity index measure}, journal={International Journal of Automotive Engineering and Technologies}, volume={14}, pages={181–189}, year={2025}, DOI={10.18245/ijaet.1634390}, author={Erhan, Koray and Sildir, Mehmet Zeki}, keywords={Mobil Robot, Robot İşletim Sistemi (ROS), SLAM, Kalman Filtresi, Partikül Filtresi, SSIM}, abstract={In this study, a performance analysis is conducted on robot localization and mapping in the context of differential-drive autonomous mobile robots, a field commonly referred to as Simultaneous Localization and Mapping (SLAM). For a mobile robot to move from a starting point to a target location and complete assigned tasks effectively, it requires a reliable environmental representation namely, a map. The creation of such a map is managed through various sensor inputs and algorithmic processes, predominantly implemented within Robot Operating Systems (ROS). With the transition from ROS1 to the more advanced ROS2 framework, SLAM software packages have also evolved. This study evaluates and compares the performance of three SLAM packages Gmapping, Cartographer, and SLAM Toolbox within the ROS2 environment using the Create-3 mobile robot platform, which is specifically developed for ROS2 applications. The experiments are conducted in three different indoor environments with varying characteristics, including a narrow, obstacle-dense room; a corridor; and a wide room with sparse features. All trials are performed using a consistent speed profile and sensor setup. Performance is assessed based on the Structural Similarity Index Measure (SSIM) between the generated maps and manually created reference maps, as well as mapping speed. The results indicate that SLAM Toolbox consistently outperformed the other packages, achieving the highest SSIM values (particularly 0.72 in the wide-room scenario) and delivering more accurate maps in complex environments. Cartographer SLAM performed well in corridor-like spaces but exhibited decreased accuracy in larger, obstacle-rich environments. Gmapping provided relatively stable results but lagged behind in terms of fine-grained mapping precision. These findings demonstrate that SLAM Toolbox is better suited for applications requiring high-accuracy 2D mapping in ROS2-based systems. The study offers a comprehensive comparison framework and concrete performance data, guiding researchers and developers in selecting the most suitable SLAM approach based on environmental conditions and system requirements.}, number={3}, publisher={Murat CİNİVİZ}