Accurate and Efficient Indoor Location by Dynamic Warping in Sequence-Type Radio-map

Abstract

An efficient way to overcome the calibration challenge and RSS dynamics in radio-map-based indoor localization is to collect radio signal strength (RSS) along indoor paths and conduct localization by sequence matching. But such sequence-based indoor localization suffers problems including indoor path combinational explosion, random RSS miss-of-detection during user movement, and user moving speed disparity in online and offline phases. To address these problems, this paper proposes an undirected graph model, called WarpMap to efficiently calibrate and store the sequence-type radio-map. It reduces RSS sequence signature storage complexity from O(2N) to O(N) where N is the number of path crosses. An efficient on-line candidate path extraction algorithm is developed in it to find a set of the most possible candidate paths for matching with the on-line collected RSS sequence. Then, to determine the user's exact location, a sub-sequence dynamic time warping (SDTW) algorithm is proposed, which matches the online collected RSS sequence with the sequential RSS signatures of the candidate paths. We show the SDTW algorithm is highly efficient and adaptive, which localizes user without backtracking of warping path. Extensive experiments in office environments verified the efficiency and accuracy of WarpMap, which can be calibrated within thirty minutes by one person for 1100m2 area and provides overall nearly 20% accuracy improvements than the state-of-the-art of radio-map method.