Indoor Localization based on Time-of-Flight Fingerprinting

Abstract

The emerging of a positioning/localization system allows user to track their position in real-time by using a knowledge of radio frequency signal (RF). Global Positioning System (GPS) is the most famous outdoor localization system with leveraging RF signal from satellites. Even though it has been widely used in many applications, it is not be able to localize in indoor environment such as building because the signal from satellite cannot penetrate inside and if it does, the signal power is not strong enough to maintain the connection. Many researchers have found a way to solve this issue by deliberately makes use of off-the-shelf wireless infrastructures such as Wi-Fi, and named these applications as Indoor Positioning System (IPS). IPS has been currently researched in a way that it should provide higher accuracy, centimeter to a few meter-scale, than outdoor system and more reliable. Ranging-based localization technique that is mainly used in GPS, turns out to be inapplicable in IPS because this technique relies on signal propagation time to estimate users location which is hardly to accurately measure in indoor environment due to strong multipath and hardware limitation of wireless devices, which is not optimized for specific purpose. Therefore, the feasible method goes to RSS-based fingerprinting where location is estimated by matching Received Signal Strength (RSS) profile with a database that contains the known locations with its profiles. However RSS profile is not robust and time-varying, causing large error in some scenario. This work presented a new way of fingerprinting–based localization by using Time-of-Flight (ToF) as a reference information instead of RSS. Theoretically speaking, ToF is better than RSS in term of reliability and accuracy because they are less time-varying and more robust than RSS. Moreover, measuring ToF does not require synchronization between access point and user device, thus it provides this system less complexity. Although, it is not a famous measurement when ToF is used in ranging-based localization where hardware limitation reduces accuracy significantly to 10 meters error, our approach has shown that it provides precisely localization when combining with fingerprinting technique. We have experimented various estimation method in order to find a better solution than basic fingerprinting algorithm where accuracy is limited by a gap between fingerprinting. We found that by using a knowledge of neighbor fingerprinting, it can achieve sub-optimal resolution than the basic algorithm. We evaluated the performance, and the result has shown that our work outperformed previous works.