Indoor localization algorithm based on behavior-driven predictive learning in crowdsourced Wi-Fi environments

Abstract

The indoor localization algorithm based on the behavior-driven predictive learning (BDPLA) executes machine-learning predictions by computing the shortest path from a starting location to a destination. The proposed algorithm selects a set of reference points (RPs) to predict the shortest path using all available RPs from the crowdsourced Wi-Fi environment. In addition, the proposed algorithm utilizes the collected received signal strength indicator (RSSI) values to determine the error distance. Using principal component analysis (PCA), the existing crowdsourced RSSI data can be calibrated to help decrease the inconsistent RSSI values among all received signals by reconstructing the values. The average error distance of 3.68 m achieved better results compared with the traditional fingerprint map with an average result of 6.96 m.