Elevator and Escalator Classification for Precise Indoor Localization

This paper presents a new finite state machine based elevator and escalator classification for pedestrians with foot mounted MEMS sensors. Generally, stance phase detectors are used to reduce the position error drift by applying Zero-Velocity-Updates (ZUPTs) in the navigation filter. However, if the person rides an elevator or escalator ZUPTs should not be processed during the stance phases of the foot. A novel elevator and escalator classifier separating the motion of the moving platform in different sub-states is introduced. Unique is the consideration of foot movement, so that the person can walk during the ride with an elevator or escalator. In addition, powerful motion constraints differing for the various platforms are presented which enable accurate positioning performance. Horizontal ZUPTs and delta ZUPTs are introduced for elevator rides and absolute velocity measurements based on an escalator velocity model are used to limit the position error growth during escalator movement. We evaluated the presented algorithms with recorded sensor data and achieved an average return position error of 0.72% in the vertical axis and an average error of 0.58% in the horizontal plane of the total walked distance of 1162.09m.