Multi Sensor Pedestrian Navigation System for Indoor and Outdoor Environments

This paper presents a real-time multi sensor localization system for pedestrians. The system can be separated into a foot-mounted multi sensor platform for robust and precise localization in indoor and outdoor environments and a hand-held device with the same sensor platform and an additional RGB-D camera for mapping purposes. The foot-mounted module is based on an accurate relative inertial positioning system with high precise Zero-velocity-UPdaTes (ZUPTs) motivated on medical research. GNSS, barometric air pressure and OpenStreetMap information is fused to obtain accurate absolute position and orientation estimations. This system can cope with very challenging situations including multiple outdoor-indoor transitions, elevators and escalators. For the hand-held device a barometric altimeter supported Visual Inertial Simultaneous Localization and Mapping (VISLAM) approach is developed. Separate measurements for translational and rotational movement are generated from RGB-D images. The translation measurements are estimated by a visual inertial odometry approach which enables loop closures. The rotation is calculated by a novel visual inertial compass, which robustly integrates the inertial sensor information. Tests show that both modules generate high accurate trajectories with small errors.