Comparison of Visual SLAM Algorithms ORB-SLAM2, RTAB-Map and SPTAM in Internal and External Environments with ROS

• Kesse Jonatas de Jesus UFSM
• Henry Julio Kobs UFSM
• Anselmo Rafael Cukla UFSM
• Marco Antonio de Souza Leite Cuadros IFES
• Daniel Fernando Tello Gamarra UFSM

Resumo

This work compares three visual Simultaneous Localization and Mapping (vSLAM) algorithms: RTAB-Map, ORB-SLAM2 and SPTAM. Simulations were carried out in an indoor and an outdoor environment on gazebo using ROS (Robot Operating System). It was used a robot differential drive with RGB-D and stereo cameras in both scenarios. The efficiency of vSLAM methods is shown. As a result of the experiments, the S-PTAM showed a better performance in indoor and outdoor environment. The measures for the trajectory distance in the ORB-SLAM2 had more accuracy in an indoor environment and the RTAB-Map had more accuracy for measures of the trajectory distance in an outdoor environment with a stereo camera. The code of the project can be found at (https://github.com/Jhonan01/jhonan.git).

Palavras-chave: Visualization, Simultaneous localization and mapping, Heuristic algorithms, Operating systems, Robot vision systems, Pose estimation, Cameras, Visual SLAM, ORB SLAM, RTAB SLAM, SPTAM, ROS