Multidimensional In- and Outdoor Pedestrian Tracking using OpenStreetMap Data

This paper presents a sensory and algorithmic solution to track and find pedestrians, particularly persons with dementia in combined outdoor and multi-level indoor environments. The shown solution uses on body MEMS sensors and doesn't require any pre- installed infrastructure or specific measurements. It combines even rudimentary map data with a relative trajectory obtained from a foot sensor, absolute positions from a GNSS sensor and height change events triggered by a barometer in a probabilistic framework. The particle filter based implementation can adjust to varying level of detail in the available map data. Also with very basic building information from OpenStreetMap the achieved accuracy suffices to easily find persons in an unknown multilevel building. By providing high resolution map data the accuracy of the system can be improved, but it already performs sufficiently well with low detailed map data to be considered helpful in the investigated dementia use case and potential similar tracking problems.