Robust Probabilistic Positioning based on High-Level Sensor-Fusion and Map Knowledge Juergen Bohn and Harald Vogt Abstract Location information as a piece of context is considered to be of particular interest to ubiquitous computing applications. In this work, we present a probabilistic positioning service that employs an available ubiquitous computing infrastructure for the localization of mobile devices. The service draws on various heterogeneous sensors that serve as sources of location information. Data from these sources are transformed independently of each other into an abstract representation of location estimates. By means of a grid-based probabilistic fusion process, these estimates are then combined into a single position value. The quality of the result increases with the number of available sensors. Also, since the fusion procedure is not depending on the type and number of available sensors, failures of single sensors or sensor types can be tolerated. Our goal is to provide a robust positioning service to support location-aware applications that need to be resilient against (temporary) disruptions of parts of the infrastructure. Though the quality of position information may degrade in the presence of sensor failures, the overall availability of the service is maintained as long as some sensor input can still be obtained. Furthermore, the positioning algorithm is designed to be scalable and to support an autonomous and decentralized operation. Thanks to its modularity, the positioning service also facilitates seamless transition between indoor and outdoor positioning.