SILVA, F. L.; http://lattes.cnpq.br/0497494110355822; SILVA, Felype de Lucena.
Resumen:
AMRs di er from AGVs due to their autonomous navigation capabilities and adaptability
to environments without xed tracks. Autonomous navigation requires precise perception,
localization, and mapping systems. In the context of LiDAR-based SLAM, the Iterative
Closest Point (ICP) stands out as one of the primary algorithms used for re ning localization
and point cloud alignment. This study proposes a SLAM and ICP-based solution for
navigation exclusively utilizing a 2D LiDAR sensor, without additional sensor support, under
computational constraints. The solution was experimentally implemented on an AGV
vehicle in a laboratory, enabling its reclassi cation from an AGV to an AMR. The main
contributions include: a comparative study of ICP methods (classical, Gauss-Newton pointto-
point, and point-to-plane) using real data; incremental implementation of ICP, allowing
continuous alignment of multiple scans; adaptation of the system for real-time operation;
and the development of an interactive interface to visualize the trajectory and generated
map. The proposed ICP technique demonstrated superior performance and was experimentally
validated, highlighting its potential for extension to other real-time SLAM applications
involving vehicles.