Automatic collecting of indoor localization fingerprints: An crowd-based approach

For typical indoor positioning systems employing a training/positioning model based on Wi-Fi fingerprints, significant training costs extremely restrict this kind of indoor localization system to be widely deployed and implemented with real location based applications. In this paper, we present a crowd-based approach to solve this problem, which automatically collects and constructs fingerprints database for anonymous buildings through common crowd customers with their smart-phones. However, such a crowd-based approach also introduces an accuracy degradation problem as crowd customers are not professional trained and equipped. So in this approach we employ fixed and hint landmarks to do error resetting. In our practical system, common corridor crossing points will serve as fixed landmarks and cross points between different crowd paths serve as hint landmarks. Machine-learning techniques are utilized for short range approximation around fixed landmarks and fuzzy logic decision technology is applied for searching hint landmarks in crowd traces space. We test this crowd-based automatic collecting approach on a dataset of about 5.09km walking in four corridors and the rooms besides. Experimental results indicate that this automatic collecting approach successfully construct indoor fingerprint radio map with rather high accuracy.