A complete sensor-based system to navigate through a cluttered environment

This article deals with the autonomous navigation problem of a mobile robot in a cluttered environment. We propose to have a different perspective than the traditional way of splitting the problem into two categories: the map-based ones and the mapless ones. Here we divide navigation systems into six processes: perception, modeling, localization, planning, action and decision. Then we present how those processes are organized into an architecture to perform a navigation. It is shown that this framework embraces any navigation system proposed in the literature and how it allows to create new combination of processes. We then detail our solution to the problem which mainly consists in coupling sensor-based controllers with a topological map. Moreover we present the used tools that we have developed over the last years as well as the ones from the literature. Finally we present experimentation results of a long-range navigation based on the proposed approach where a robot drives through an environement despite of occlusions and possible collisions due to obstacles.