Neural and fuzzy approaches to vision-based parking control

Abstract This paper presents an approach to the acquisition and transfer of an experienced driver's skills to an automatic parking controller. The controller processes input information from a video sensor and generates the corresponding steering commands. Two neural control architectures are considered. In the direct neural control architecture, the controller is a single artificial neural network. In the fuzzy hybrid control architecture the controller is configured as a combination of an artificial neural network and a fuzzy network. Both control architectures have been experimentally validated with a mobile robot.