Server-Sided Automatic Map Transformation in RoboEarth

RoboEarth aims at providing a distributed cloud-based web platform from robots for robots that is publicly accessible and enables robots to autonomously share knowledge among each other and to generate new knowledge from previously stored data. As a result robots don’t have to gain the same knowledge over and over again, but can build upon it right from the start. Currently, shareable data are abstract task descriptions, object models and environment maps. In this paper we describe RoboEarth’s approach to automatically and transparently generate 2D maps for localization and navigation, which are extracted from shared 3D maps and suited for a specific robot configuration. The parameters of the map generation process get inferred from a robot’s semantic self-description. Using RoboEarth for knowledge generation enables simple platforms with low computational power to execute complex tasks in complex environments. Furthermore the approach effectively simplifies the time consuming process of generating new maps every time a new robot platform with different specifications is used.