Fingerprintbuilding and positioning based on wireless sensor networks for underground.

Abstract

This paper proposes an algorithm of fingerprintconstructing and positioning based on wireless sensor networks. The positioning area of underground is divided into several sub-areas according to the neighborhood principle. The reliability mechanism based on the calibration node is established in each sub-area, and the availability of the reference pointfingerprintis later determined by the reliability mechanism of the sub-area. When the fingerprintdata of the sub-region reference pointchanges too much, the neighborhood mapping fingerprintmodel trained by the back propagaption neural network is used to construct the fingerprint. The principle of the neighborhood mapping model is to train the neighborhood relationship between each reference pointand its adjacent calibration node in the offline phase to form a network model structure. Then, we use this model to build the fingerprintof reference pointin the online stage. In the real-time positioning stage, we use the positioning model based on the adaptive network-based fuzzy inference system. The average positioning error of our proposed algorithm is 3.03m, when there is a seven day interval between the training dataset and testing dataset, which confirms that the proposed algorithm can be better adapted to the changing environment and to achieve better positioning results.