An Incremental Hybrid Approach to Indoor Modeling

Most of mobile robots basic functions are highly dependent on a model of their environment. Proper modeling is crucial for tasks such as local navigation, localization or route planning. This paper describes a novel solution for building models of indoor environments. We use a 3D Laser on a mobile robot to scan the surroundings and obtain sensing information. While the robot explores the environment, perceived points are clustered forming models of rooms. Room modeling is solved using a new variation of the Hough transform. The result of the modeling process is a topological graph that represents the rooms (nodes) and their connecting doors (edges). Each node in this representation contains the metric parametrization of the room model. Using this basic metric information, robots do not need to maintain in parallel a metric map of the environment. Instead, this metric map can be totally or partially built from the topological representation whenever it is necessary. To test the approach we have carried out a modeling experiment of a real environment, obtaining promising results.