Neurons that know how to drive

ROBIN is a template-based visual autonomous road-following system based on a radial basis function architecture. ROBIN is trained by a human operator to drive on a specific road-type. These training instantiations can be saved in a repertoire of road types and activated at a later time. By using a diverse sensor set such as a color camera, FLIR and low light camera, ROBIN can function in both night and day without the requirement for artificial lighting. ROBIN has successfully driven on a diverse set of road types ranging from worn tracks in the grass to well maintained highways. Image understanding is the primary processing modality of ROBIN in the generation of a steering and speed response, but it can also exploit other modalities such as a priori road geometry or embedded roadway magnets to monitor inappropriate vision-based control responses. ROBIN has been thoroughly tested in the course of multiple automation programs.