Path planning for autonomous vehicles by trajectory smoothing using motion primitives

We present a novel planning strategy which is applicable to high performance unmanned aerial vehicles. The proposed approach takes as input a 3-D sequence of way-points connected by straight flight trim conditions, and "smooths" it in an optimal way with the goal of making it compatible with the vehicle dynamics. The smoothing step is achieved by selecting appropriate sequences of alternating trims and maneuvers from within a precomputed library of motion primitives. The resulting extremal trajectory is compatible with the vehicle and therefore trackable with small errors; furthermore, it is guaranteed to stay within the flight envelope boundary, alleviating the need for flight envelope protection systems. Yet it can be computed in real-time using closed-form expressions, all nonlinearities due to the vehicle model being confined to the stored library of motion primitives. The new method is demonstrated for the aggressive maneuvering of a helicopter.