Approximating Reference Trajectories for Autonomous Vehicles Using Motion Primitives

In this paper we describe an approach for validating and adapting a reference trajectory of an autonomous vehicle. It is assumed that the generation of the reference trajectory guarantees the approximation to be collision free if it stays within a tolerance area. We employ a precomputed set of motion primitives as used in state-of-the-art path planning libraries. The generated path is smooth and feasible. Two applications are presented, which have motivated the development of the approach.