Fast approximate path coordinate motion primitives for autonomous driving
Matthew Sheckells, Timothy M. Caldwell, Marin Kobilarov · 2017
In autonomous driving, it is often useful to plan trajectories in a curvilinear coordinate frame with respect to a given reference curve, such as a path produced by a high-level route planner. In this domain, standard planning methods rely on expensive coordinate transformations or on solving computationally intensive boundary value problems for computing motion primitives between states. This work develops efficient, approximate path coordinate motion primitives appropriate for fast planning in autonomous driving scenarios. We gain a 1000x speed-up in primitive computation time relative to standard approaches at the loss of some precision with respect to the position along the reference line, which we statistically quantify. Motion primitive properties like path length, acceleration, and the reference line offset are exactly preserved.