An Enhanced Density-Based Clustering Algorithm for the Autonomous Indoor Localization

Indoor localization applications are expected to become increasingly popular on smart phones. Meanwhile, the development of such applications on smart phones has brought in a new set of potential issues (e.g., high time complexity) while processing large datasets. The study in this paper provides an enhanced density-based cluster learning algorithm for the autonomous indoor localization algorithm DCCLA (Density-based Clustering Combined Localization Algorithm). In the enhanced algorithm, the density-based clustering process is optimized by "skipping unnecessary density checks" and "grouping similar points". We conducted a theoretical analysis of the time complexity of the original and enhanced algorithm. More specifically, the run times of the original algorithm and the enhanced algorithm are compared on a PC (personal computer) and a smart phone, identifying the more efficient density-based clustering algorithm that allows the system to enable autonomous Wi-Fi fingerprint learning from large Wi-Fi datasets. The results show significant improvements of run time on both a PC and a smart phone.