Research on Map Building by Mobile Robots

Map building is the basis of navigation, path planning and obstacle avoiding for mobile robots. So it has been an active research area in robotics in the last two decades. In this paper, the map building problem is surveyed in the following three aspects: sensor techniques used to acquire environment information, paradigms for mapping used to represent environments, and methods for information processing. We introduce internal, range finder and vision sensors as the main ones employed in information acquiring. Then, we compare three fundamental paradigms of mapping including occupancy grid, feature and topological maps, and especially present new examples of hybrid maps. Finally, as localization is a key problem in mapping, we divide information processing methods into two types, which are the probabilistic methods and scan matching methods according to their different considerations on localization problem. Representative methods are also presented, along with open research problems in map building.