A Study on the Application of ORBSLAM3 on a Mobile Differential Drive Robot

Abstract

A mobile indoor robot has been developed to navigate its environment using Visual Simultaneous Localisation and Mapping (VSLAM) technology. This technology allows the robot to perceive and interpret its surroundings in 3D space, providing a more stable and accurate map of the environment. The robot is equipped with ORBSLAM3, a VSLAM algorithm, which enables autonomous navigation using maps generated by MLMapping. Experiments were conducted to study the accuracy of VSLAM localisation and navigation. The results showed that the estimated trajectory had a root-mean-square error of 0.13 meters compared to the lidar-based localisation. The map created using VSLAM had an average distance to the nearest neighbour (ADNN) of 0.14 meters when compared to the map generated by lidar-based SLAM.