A practical implementation of a multilevel architecture for vision-based navigation

The paper summarizes the current work of the DIST-University of Genoa in the field of mobile robotic control. In particular the authors describe the implementation of a multilevel architecture for navigation, based on a stereo vision system, which allows obstacle avoidance, detection of free corridors and low level planning by means of few known landmarks such as color spots or vertical edges. Each level is, in principle, independent from the others. Appropriate motor actions are carried out combining the parallel behavior of the different levels.<
>