A real-time indoor localization approach integrated with a Geographic Information System (GIS)

Nowadays, accurate maps from mostly anywhere in the world can be obtained for free, with the exception of indoor spaces. However, the evidence seems to suggest that in the next few years indoor maps will be more and more available for anyone. Thus, profiting from the idea of easily obtainable indoor maps, we present a novel approach for real-time mobile robot localization that focuses on spatial reasoning at a high abstraction level. In order to manage and query existing indoor spatial models, we rely on the power of Geographic Information Systems (GIS) and spatial databases. Moreover, to extract symbolic information from the environment, we have developed a door detection system that fuses 2D laser and vision data. We have integrated these two ideas into an extended Kalman filter localization framework. Our proposal has been implemented and tested through autonomous navigation missions in real-world scenarios. Extensive experimental results are provided, which show robustness and accuracy concerning both door detection and localization.