Multi-Resolution Elevation Mapping and Safe Landing Site Detection for an Autonomous UAV

In this paper, we propose a resource-efficient approach to provide an autonomous UAV with suitable, hazardfree landing sites by building and analysing a local elevation map. We introduce a perception system that allows to build a detailed map of the UAV’s local environment in the form of a two-dimensional, robot-centric elevation map with multiple scales that is probabilistic. Through continuous depth map fusion based on the concept of dynamic Level of Detail, we are able to build the map incrementally. Using depth maps from a Structure-from-Motion approach, the system requires only the input of a single downward-looking camera and an inertial measurement unit. In addition, we present an approach for an efficient landing site detection that exploits the features of the underlying multi-resolution map.