3D Loose-Coupled Fusion of Inertial Sensors for Pedestrian Localization

This work focuses on the development of a position tracking algorithm for pedestrians using, simultaneously, two inertial sensors. Our goal is to extend, from 2D to 3D, a loosely-coupled system that combines a foot-mounted inertial navigation system (INS) and a pocket-mounted INS. In order to estimate the vertical displacement of the pedestrian's walked trajectory, we use the pocket INS to detected when the user is walking the stairs. Upon stairs detection, the vertical displacement is estimated through the position estimated by the foot INS. The vertical displacement is then integrated in the filter of the loosely coupled system. The results show an improvement of 48% regarding the average height error of the foot INS. In fact, the estimated height drifts less than the foot INS estimated height when the user walks on a horizontal surface. Nevertheless, the proposed approach does not outperform the pocket INS. In conclusion, the 3D loose coupling has a reduced height drift in horizontal surfaces, while it estimates vertical displacements without any height model.