Real-time Symbolic Indoor Map Matching

Abstract

Recently, many indoor positioning methods have been developed such as Wi-Fi positioning, Geo-magnetic sensor positioning, Ultra-Wide Band (UWB) positioning, and Pedestrian Dead Reckoning (PDR). These indoor positioning methods provide (x, y, z) coordinates in R3 or (x, y) in R2 space. However the context of a user in indoor space is mainly determined by the indoor cell, such as room, where the user stays rather than (x, y, z) coordinates. The determination of the cell from (x, y, z) coordinates collected by indoor positioning is therefore a fundamental requirement of indoor spatial applications. We call this function symbolic indoor map matching (SIMM), like map matching that is to determine the road segment from GPS in road network space for navigation services. A direct indoor map matching is to discover the cell containing the point (x, y, z) collected from indoor positioning. However it results in a low accuracy due to the errors contained in the indoor positioning methods. Considering the average error 1-5 meters of current indoor positioning methods, it is not sufficiently accurate to determine the cell of user. In order to improve the accuracy, we propose several symbolic indoor map matching methods based on the Hidden Markov Model (HMM) for probabilistic properties of moving objects and additional properties of indoor space. Since our indoor map matching methods assume a certain level of errors from indoor positioning methods, they yield significant improvements of accuracy particularly when the accuracy level of indoor positioning is low. We analyze the accuracy and performance of the proposed methods in comparison with the direct indoor map matching by experiments.