A Comparative Study of 2D SLAM Algorithms for ROS-Based Mapping

Simultaneous Localization and Mapping (SLAM) is a fundamental capability for autonomous systems, enabling them to perceive and navigate their surroundings. Numerous SLAM algorithms have been developed and implemented within the Robot Operating System (ROS) framework to facilitate real-time mapping and localization. This work presents a comparative review of three widely used 2D SLAM algorithms in ROS: GMapping, KartoSLAM, and SLAM Toolbox. The study evaluates these algorithms based on their mapping quality and computational load through a series of experiments on diverse datasets. Each algorithm's performance is analyzed in terms of map quality and adaptability to different operational conditions. Additionally, the CPU usage of each algorithm was monitored during the experiments to assess their computational demands. The results indicate that KartoSLAM provides the most consistent mapping accuracy but has a higher initial CPU load due to intensive graph initialization. GMapping exhibits reliable boundary representations but experiences increased CPU usage over time due to its particle filter-based approach. SLAM Toolbox demonstrates robust loop closure but suffers from occasional misalignment issues and higher final CPU usage during optimization. The findings highlight the trade-offs between accuracy and computational efficiency, providing insights into the practical deployment of these algorithms on resource-constrained robotic platforms. This review aims to assist researchers and developers in selecting the most suitable SLAM algorithm for their specific robotic applications, particularly in indoor navigation and autonomous mobile systems.