Dynamic indoor navigation with bayesian filters

Abstract

Indoor space navigation has always been an issue without GPS localization. Especially for complicated cases such as emergency evacuation and dynamic navigation, there is no existing efficient solution to the best of our knowledge. Localization in indoor spaces has to rely on sensing devices (e.g., Radio Frequency Identification(RFID) readers, WiFi routers, bluetooth beacons) rather than GPS, and indoor floor plans are more complicated than road networks. Consequently, existing spatial outdoor query techniques are not suitable for this new challenge. However, raw data generated by sensing devices suffers from false negatives and errors. As a result, filtering methods are necessary for accurate localization. We propose a novel dynamic shortest path navigation strategy to enable efficient navigation for emergency evacuation in indoor spaces. This strategy achieves optimal time efficiency by: 1) using a Bayesian inference based concurrent model, which integrates dynamic shortest path searching into the filtering process, thus achieving an efficient and accurate search for any time-sensitive situation; 2) storing alternative parent nodes along the shortest path search for a fast, dynamic search. We use both particle filters and the Kalman filter to study which one is more suitable for dynamic environments. In general, we develop an innovative, dynamic shortest path navigation solution based on Bayesian inference localization.