UAV Obstacle Mapping for Multi-UGV Exploration and Mapping

Although autonomous Unmanned Ground Vehicles (UGVs) can accomplish a vast variety of tasks, they do have one weakness; their limited field of view. Teams of UGVs can be used to navigate and collect data in high-risk environments that are dangerous to humans, but can be rendered useless if they get trapped by an unseen obstacle. This can occur with UGVs that use a 2D LiDAR for their navigation and cannot sense objects or holes below the height of their sensor. To mitigate this issue, an Unmanned Aerial Vehicle (UAV) is proposed to be added to the multi-robot system to use its high vantage point and communicate the locations of previously unseen obstacles. In this work, an algorithm is developed to process pointcloud data created from a UAV depth camera and isolate obstacles. These obstacles are added to the UGVs’ costmaps so that they can be included in path-planning and subsequently safely avoided. Experimental results show that the proposed method can be used to detect obstacles independent of their shapes, colour, or unique markings and can accurately add their locations and geometries to the costmap.