WiFi indoor positioning based on regularized online sequence extreme learning machine

ABSTRACT WiFi positioning based on fingerprint has received widespread attention and practical applications. However, the fingerprints are susceptible to environmental changes, such as shadowing, multipath, temperature, humidity and obstacles. Due to the instability of received signal strength (RSS), it brings plenty of difficult for WiFi positioning with high accuracy. In this paper, a regularised online sequence extreme learning machine with forgetting parameters (FP-ELM) is adopted to solve the issue accordingly. Forgetting factor and regular factor are adopted in FP-ELM to cope with the time-varying nature of RSS and overcome the issue of irreversible matrix in OS-ELM. The fast running speed of the online sequence extreme learning machine (OS-ELM) is also maintained in FP-ELM. Extensive experiments are carried out in simulation and real experimental areas to explore the characteristics of FP-ELM. Moreover, the positioning results of FP-ELM are compared with the conventional algorithms (OS-ELM and KNN). The simulation and experimental results show when the regular factor is set properly, the positioning result based on FP-ELM algorithm is better than conventional algorithms Figures 1.