Robust indoor localization with wi-fi and IMU sensor fusion

Abstract

Indoor localization remains a necessity thing in Global Positioning System (GPS) -denied places such as hospitals, warehouse and airports where the indoor positioning is demanded. Even if there are many indoor localization applications and systems, the companies and the developers do not publish the source code and its implementation which leaves a huge gap to reproduce or improve it. Modern smartphones cannot easily identify the current location of the person in indoor environments due to the limitations in Android operating system (OS), which can scan the Wi-Fi once in 30 seconds. Because of the lack of data, the quality of standard Wi-Fi fingerprints becomes worse. This study addresses the limitation of current operating systems by fusing pedestrian dead reckoning which is based on Inertial Measurement Unit (IMU) sensors, and Wi-Fi Received Signal Strength Indicator (RSSI) in order to overcome those 30-second gaps. The system combines Madgwick filter for estimating the heading, Weinberg formula to calculate the step length and Zero Velocity Update (ZUPT) to calibrate the sensors in order to detect the stops in real time. The experiment results on some complex trajectories showed that the Pedestrian Dead Reckoning (PDR) algorithm drifts a lot and the mean error of trajectories were more than 7.8 meters. However, the fusion algorithm showed stable results, where the mean error was 4.8 - 6.1 meters under throttling simulations. The results say that a fusion algorithm can provide reliable localization without any hardware modifications.