Inertial pocket navigation system for pedestrians

Abstract

There is nowadays a high demand of pedestrian navigation systems, which are integrated in safety-of-life services such as disaster management for rescue personnel or locationbased services such as guidance in hospitals, airports or shopping malls. In this work, indoor and urban environments constitute the targeted scenarios and the navigation is performed with inertial and magnetic sensors due to their wide availability, light-weight and infrastructureless nature. Investigations are carried out that improve or cover specific gaps of pedestrian navigation areas to offer versatile pedestrian navigation systems for a wide range of applications. First, the use of magnetic field measurements to compensate the systematic errors of inertial sensors and their effect on the estimated orientation of the sensor has been comprehensibly analyzed. Reference measurements with known error values have been used combined with different magnetic field distributions and the results have been endorsed with real measurements of medium-cost sensors. It is concluded that the use of magnetic measurements is beneficial to estimate the systematic errors, yielding to bounded orientation estimation errors. However, the targeted scenarios commonly present perturbed magnetic fields and the error estimation becomes prohibitively slow. Second, several algorithms have been proposed in this work that outperform the accuracy of the horizontal displacement of the pedestrian with respect to the state of the art. Additionally, an innovative vertical displacement estimation algorithm has been proposed and tested in real-world scenarios. This algorithm makes it possible for the first time to solve unaided 3D inertial positioning for non-foot-mounted sensors. Lastly, a novel drift estimation algorithm capable of preventing positioning errors caused by heading errors is proposed. The computation of the drift is based on landmarks automatically detected using solely inertial measurements. Landmarks defining the building or city layout have been chosen to be stairs and corners. By re-visiting these landmarks it is possible to observe the accumulated drift, which is fed back to the orientation estimation algorithm in order to bound the heading error. The proposed algorithm has been extensively tested with reference and real measurements of medium-cost sensors. Two types of corrections, online and offline, are presented to adapt the pedestrian navigation system to the particular application.