A robust pedestrian dead-reckoning positioning based on pedestrian behavior and sensor validity

In this paper we describe a novel sensor data fusion method that estimates the location of a pedestrian. The key information to estimate position is the distance, the direction and the altitude. Several estimation methods exist to determine direction and altitude. They have strength and limitations depending on urban conditions. We introduce two types of Reliability Indexes (RI): pedestrian behavior and sensor validity, to selectively integrate reliable sensor data. We have adopted a particle filter to integrate the results of several sensor data. Our method adopted RI to choose reliable sensor data and update sensor offset in a particle filter. We have conducted preliminary experiments to validate our method. The preliminary experiment results showed our method can remove the noise effect of sensor data almost always. We also evaluated the proposed method in an urban area. Our experimental results showed that proposed method achieves higher accuracy and more robust than using either of the other sensor data is isolation.