Autonomous mobile robotics : Localisation and navigation with a region based map

Usual environmental representations either tend to complexify the world or to restrain a robot’s movements in order to locate it correctly. Overly localised perceptions are often to blame. They focus on details such as distance to a wall. For large scale navigation, this type of information is not relevant. This article introduces a concise representation system which allows a behaviour independent localisation. The localisation unit is the region. A region is a large portion of space exhibiting certain topological properties: a room, a section of a corridor, or a dead−end. Consequently, alleviated treatment simplifies localisation. The robot builds, through autonomous exploration, a graph in which a node represents a region and a vertex the border between two regions. Localisation is based on global environment perception. Sensor (sonars in this case) classification is only slightly affected by robot position and orientation within the region. We also present our results in simulated and actual experiments.