A scalable indoor localization algorithm based on distance fitting and fingerprint mapping in Wi-Fi environments
Hao Zhang,Kai Liu,Feiyu Jin,Liang Feng,Victor Lee,Joseph Ng
DOI: https://doi.org/10.1007/s00521-018-3961-8
2019-01-01
Neural Computing and Applications
Abstract:With ever-increasing demands on location-based services in indoor environments, indoor localization technologies have attracted considerable attention in both industrial and academic communities. In this work, we propose a scalable indoor localization algorithm (SILA) consisting of two components, namely an annulus-based localization (ABL) component and a local search-based localization (LSL) component, with the objectives of enhancing localization accuracy and reducing online computational overhead. First, the ABL component is developed based on distance fitting using received signal strength indicator (RSSI) of Wi-Fi-based devices. In particular, a distance-RSSI fitting model is proposed based on multinomial function fitting, which is adopted to estimate the distance between the Wi-Fi access point (AP) and the mobile device. On this basis, an annulus construction scheme is proposed to confine the online searching space for possible locations of the mobile device. In addition, based on the observation of signal attenuation characteristics in different physical environments, we design a subarea division scheme, which not only enables the system to choose proper distance-RSSI fitting functions in different areas, but also reduces the overhead of distance fitting. Second, the LSL component is developed based on fingerprint mapping using RSSIs collected at APs. In particular, an RSSI distribution probability model is derived to better map the signal features of an online point (OP) with that of reference points (RPs). Then, an online localization algorithm is proposed, which selects a set of candidate RPs based on Bayes theorem and estimates the final location of an OP using K-nearest-neighbor (KNN) method. Finally, we implement the system prototype and compare the performance of SILA with two representative solutions in the literature. An extensive performance evaluation is conducted in real-world environments, and the results conclusively demonstrate the superiority of SILA in terms of both localization accuracy and system scalability.
computer science, artificial intelligence