A Bayesian Density Model Based Radio Signal Fingerprinting Positioning Method for Enhanced Usability

Zheng Li,Xiaoguang Niu,Ruizhi Chen,Leilei Li,Jingbin Liu,Fan Yang,Zemin Wang

doi:10.3390/s18114063

Abstract

Indoor navigation and location-based services increasingly show promising marketing prospects. Indoor positioning based on Wi-Fi radio signal has been studied for more than a decade because Wi-Fi, a signal of opportunity without extra cost, is extensively deployed for internet connections. Bayesian fingerprinting positioning, a classical Wi-Fi-based indoor positioning method, consists of two phases: radio map learning and position inference. Thus far, the application of Bayesian fingerprinting positioning is limited due to its poor usability; radio map learning requires an adequate number of received signal strength indication (RSSI) observables at each reference point, long-term fieldwork, and high development and maintenance costs. In this paper, based on a statistical analysis of actual RSSI observables, a Weibull–Bayesian density model is proposed to represent the probability density of Wi-Fi RSSI observables. The Weibull model, which is parameterized with three parameters that can be calculated with fewer samples, can calculate the probability density with a higher accuracy than the traditional histogram method. Furthermore, the parameterized Weibull model can simplify the radio map by storing only three parameters that can restore the whole probability density, i.e., it is not necessary to store the probability distribution based on traditionally separated RSSI bins. Bayesian positioning inference is performed in the positioning phase using probability density rather than the traditional probability distribution of predefined RSSI bins. The proposed method was implemented on an Android smartphone, and the performance was evaluated in different indoor environments. Results revealed that the proposed method enhanced the usability of Wi-Fi Bayesian fingerprinting positioning by requiring fewer RSSI observables and improved the positioning accuracy by 19–32% in different building environments compared with the classic histogram-based method, even when more samples were used.

Highlights

Because of their social and commercial value, indoor location-based services (ILBS), which are predicted to be worth US$10 billion by 2020 and US$58 billion by 2023 [1,2], have attracted substantial attention in recent years
In this paper, we introduce the Weibull signal model to approximate the received signal strength indication (RSSI) probability distribution of all access points (APs) received at each fingerprinting point
The average error and root-mean-square error in the positioning map to first model the probability distribution of the RSSI measurements between an AP Am and a results of the three algorithms were compared in different actual scenarios

Summary

Introduction

Because of their social and commercial value, indoor location-based services (ILBS), which are predicted to be worth US$10 billion by 2020 and US$58 billion by 2023 [1,2], have attracted substantial attention in recent years. Smartphones, which are equipped with a variety of sensors that can be used for indoor positioning, are the most preferred platforms for such services. Techniques that are used to collect data with various sensors in a smartphone include wireless communication technologies (Wi-Fi [3], BLE [4], RFID [5,6]); optical and vision [7], and magnetic [8] among others. Among the growing number of techniques, Wi-Fi-based indoor positioning has become a research hotspot [9] because Wi-Fi access points (APs) are widely deployed throughout indoor environments, such as offices and airports. Wi-Fi signals can be used for positioning signals of opportunity, thereby requiring no extra cost. Wi-Fi-based indoor positioning applied in smartphone have many favorable features, such as low deployment costs, required accuracy, tolerable uncertainty, and fewer necessary computational resources [10]

Methods

Results

Conclusion