Skill-based vehicle guidance by use of artificial neural networks

This article addresses the use of artificial neural networks for vehicle guidance in local maneuvers. As a typical example, the docking or parking of a semi-autonomous robot is studied. In the proposed approach a feed-forward artificial neural network trained by the back-propagation learning algorithm acts as a nonlinear state-space controller reproducing and generalizing the steering actions of a skilled human driver. The results are experimentally validated with the robot vehicle MACROBE.