Using 2D LiDAR and RGB Camera for Human Agnostic Mapping

Human agnostic mapping has great potential to enhance automation in industrial settings and smart cities. Despite established research in mapping using light detection and ranging (LiDAR), mapping in the presence of undesirable objects, such as pedestrians, remains an under-explored field. This project presents a novel method to create a more precise map of an environment even with the presence of humans. This method was achieved by utilising a combination of camera based object detection and LiDAR point clouds. A Simultaneous Localisation and Mapping (SLAM) algorithm was chosen, in this case Hector SLAM, to create maps of the environment using sensor fusion. A comparison was done between the newly created map with existing floorplans to validate the proposal method. It was found that the newly produced map enhanced accuracy and robustness in the presence of humans. From the testing conducted in this project, the filtering mechanism worked in an accurate manner except for cases where the detection was at the edge of the frame. This paper demonstrates usability and accuracy of the proposed human-agnostic mapping to catalyse intelligence in next-generation autonomous navigation systems.