Reactive Obstacle Avoidance for Multicopter UAVs via Evaluation of Depth Maps

Reacting to unforeseen obstacles is a major issue in the field of autonomous navigation. In the context of Unmanned Aerial Vehicles, an “obstacle” is any object that stands between the UAV and its desired position (waypoint). Therefore, obstacle detection can be reduced to the problem of assessing the visibility of the waypoint from the point of view of the drone. In this work, data acquired from an onboard depth camera are used to describe the visibility of the target waypoint in a qualitative framework, and to plan a new route when obstacles are detected.