Reference Points Density Invariant Map Matching for Wi-Fi Fingerprinting Positioning

Wi-Fi fingerprinting positioning is one of the most popular methods for indoor positioning. However, its positioning error is influenced by the abnormal large distance between the time-adjacent locations. To tackle the problem, the map matching method is used to improve the precision of the positioning by the constraint of the indoor environment. Hidden Markov model (HMM) is a popular model for map matching to treat locations as a sequence and refine them according to the connection of the indoor reference points and their neighbours. However, the traditional HMM-based method constructs the transition matrix just by the adjacent reference points in indoor environments. As the transition between the nonadjacent reference points is forbidden, the precision of the positioning is highly dependent on the interval size of the reference points. In this paper, we combine the long and short distances together to design a new transition matrix that considers both adjacent and nonadjacent transition. The probabilities of the transition matrix are assigned based on the distance between reference points. In the experiment, we evaluate our method in a complex indoor environment, which consists of various regions. The results indicate that our method is insensitive to the change of the moving speed and the interval size of the reference points, which affects the positioning error of the traditional methods significantly.