Toward Localization in Silence: Novel Fingerprinting Techniques for WLAN-based Indoor Localization

靜默室內定位:基於無線局域網的新穎指紋定位技術

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date31 Aug 2018

Abstract

Indoor localization has attracted increasing research attention for years as it provides high commercial value for indoor location-based services (LBS), such as location-based advertisement and indoor navigation. Among various techniques proposed for indoor localization, fingerprinting technique using Wi-Fi signal from the infrastructure (i.e., access point) of wireless local area networks (WLAN) has attracted continuous attention in both academic and industrial communities. The fingerprinting technique does not require line-of-sight (LoS) measurement of access points (APs) while achieving fine-grained localization accuracy in a complex indoor environment.

In previous fingerprinting techniques proposed for WLAN-based indoor localization, fingerprints are constructed by taking measurements of received signal strength (RSS) of packets sent by APs at a fixed transmission power. In order to improve performance of fingerprinting technique, we consider using path loss rather than RSS to better distinguish signal quality at different locations. In particular, we propose an innovative formulation, namely Multiple Power Path Loss (MPPL) fingerprint, based on path losses obtained from the packets sent by APs at multiple power levels. By analyzing signal propagation in an indoor environment, MPPL fingerprint reveals its subtle advantage of using multiple transmission power. In addition, a location estimation algorithm is designed to exploit the performance characteristics of the the MPPL fingerprint. Extensive experimental results illustrate the usability and superiority of our proposed method.

However, typical RSS-based fingerprinting techniques cannot achieve localization in silence, which is a desirable requirement for indoor localization system, as an application/software has to be installed on the target (i.e., the WiFi client) for uploading RSS information to the localization system. In this work, we propose data rate (DR) fingerprinting to replace RSS by data rate to form fingerprints. DR fingerprinting can achieve localization in silence because data rate information is readily available to be observed by APs. However, some intrinsic features of data rate, including low-resolution and serious fluctuation, significantly impair the performance of DR fingerprinting. To maintain a high localization accuracy of DR fingerprinting, we leverage multiple transmission power levels, propose a time window mechanism with different fingerprint formulations, and design a new matching algorithm namely dynamic nearest neighbors (DNN). We conduct extensive experiments in a real-world testbed to study the performance of DR fingerprinting. Experimental results illustrate the feasibility and effectiveness of the proposed DR fingerprinting technique.

However, while a target is in deep-silence (i.e., not associated with APs), there is no data frame transmitted from AP to the target for obtaining data rate information, rendering DR fingerprinting useless. Hence, we propose packet delivery ratio (PDR) fingerprinting to provide localization in deep-silence. In PDR fingerprinting, we leverage RTS/CTS (control frames) to trigger frequent interaction between APs and the target, thereby obtaining sufficient packet delivery ratio information. A channel filter (CF) algorithm is designed to figure out the time slot for capturing RTS/CTS frames in the correct channel. Furthermore, we fuse the packet delivery ratio information derived from multiple data rates and multiple transmission power levels to build high-dimensional PDR (HD-PDR) fingerprints, for achieving higher localization accuracy. The experimental results illustrate that PDR fingerprinting provides competitive performance while achieving localization in deep-silence.