Benchmarking Unmanned Rotorcraft Trajectories based on Dynamic Fit

This paper presents a performance assessment of a method to generate trajectories for a small unmanned rotorcraft within an obstacle field using a standardized set of benchmark scenarios. The dynamics of the rotorcraft are discretized into a set of motion primitives. For the generation of these primitives a formulation of a path-tracking optimal control problem is used. This optimal control problem has been previously shown to achieve trajectories that can be flown very accurately in flight tests. The generated primitives are used to convert the original task within the obstacle field into a discretized planning problem using sequential decision making that is solved using an optimal graph search algorithm (e.g., Dijkstra). This algorithm selects from and combines the available motion primitives to achieve the transition between a start and a goal waypoint in minimum time. We present results of this planning approach and include a scaling analysis for simple artificial as well as more complex urban environments.