The vision subsystem of the autonomous vehicle ROBERT

The vision subsystem of an autonomous navigating vehicle is presented. This subsystem is composed of two parts: a data processing module which finds and extracts the image of a "marker" from the background; and a recognizer of the marker itself. Both the modules are based on neural network architectures, the first on a simple but efficient one, the second on a standard backpropagation. Percentages of correct classification are given.