A Navigation-Field-Based Semi-Autonomous Nonholonomic Vehicle-Parking Assistant

We present a closed-loop approach to solve the semi-autonomous nonholonomic vehicle-parking assistant problem. The method allows foraprioriunknown and potentially high vehicle speeds and dynamic environment variations. The computationally low-cost, memory-efficient algorithm respects constraints on the steering angle and its first two derivatives, allowing for comfortable, safe, and smooth steering motions. A parking situation is captured by an arrangement of precalculated vector fields of feasible nonholonomic paths, by which the vehicle is controlled toward a desired goal configuration inside a parking bay. Using standard onboard environment perception sensors, the system adapts to changes in the environment and avoids collisions whenever possible, without aborting a parking procedure.