Map Matching with WiFi-RSSI GRU Indoor Room Switching Classifier

In recent years, indoor location services have gradually become a new research direction in addition to outdoor location services. Traditionally, WiFi-based indoor location technologies are divided into two categories: (i) using WiFi channel propagation model based on RSSI values for ranging, whose localization error can reach 10~20 m; (ii) using RSSI to establish fingerprint library for fingerprint matching and machine learning algorithms for matching, whose localization error can reach 3~20 m. Existing WiFi-based location technologies are usually applied to position coordinate solving, but the accuracy is not high enough. However, using WiFi-RSSI for indoor room switching pattern recognition will effectively improve the accuracy of map matching. In this paper, we use the gate recurrent unit (GRU) algorithm to determine the user's indoor room switching model from the WiFi RSSI timing data set of the mobile terminal in the indoor room switching scenario, which can effectively correct the traditional Hidden Markov Model (HMM) indoor map matching algorithm and significantly improve the accuracy and stability of the indoor map matching algorithm. This algorithm improves the matching accuracy by up to 25.1% compared with the traditional HMM indoor map matching algorithm under the limited accuracy of the original solution coordinates provided by the indoor positioning system.