Model-Based Local Path Planning for UAVs

Autonomous aviation continuously becomes more and more important. Algorithms that enable this autonomy have developed quickly in the last years. This paper describes a concept for a reactive path planning algorithm. The aim is to develop a method for static obstacle avoidance of an unmanned aerial vehicle (UAV) by calculating collision-free paths within the field of view of a UAV's obstacle detection sensor. In contrast to other algorithms, this method considers the properties of the obstacle detection sensors, plans paths that the UAV is able to track, and is applied in three-dimensional space without access to an inner loop controller. In this work we represent the field of view of a UAV as a grid map and apply the wavefront algorithm as the local path planning algorithm. We reduce the configuration space of UAVs within the field of view by calculating an approximated worst-case reachable set based on a linearized reference model. We evaluate the method with approximated specifications for the unmanned helicopters ARTIS and Yamaha RMAX, and with specifications for the obstacle detection sensors LIDAR - and stereo camera. Experiments show that this method is able to generate collision-free paths in a region constricted by obstacles.