Previous Abstract Return to Session C2 Next Abstract

Session C2: Signals of Opportunity-based Navigation Systems

Algorithm Analysis of WKNN and Bayes Estimation in WiFi Fingerprint Localization Method
Wei Gao, Kedong Wang, School of Astronautics, Beihang University, China
Location: Atrium Ballroom
Alternate Number 3

With the wide spread of Internet and the acceleration of 5G network construction, location information has become an indispensable basic information in the Internet society. However, due to the limitations of the indoor application of satellite navigation system, how to provide accurate and reliable location information for indoor users has been widely and deeply studied, which is a hot issue in the field of navigation. The localization principle used in this paper is the RSSI fingerprint localization method based on WiFi. On the basis of building indoor simulation environment model, RSSI offline fingerprint dataset and online data are constructed based on signal propagation model, which can be used in the simulation experiment of RSSI localization algorithm. Based on the simulation results, the localization accuracy, calculation and stability of the localization algorithm are compared and analyzed. The simulation results show that the localization accuracy and stability of WKNN and Bayes are quite excellent. In the field test, however, for the fact that the AP dimension is high but the failure dimension accounts for a large proportion in the actual data, it is necessary to select the dimension according to the signal strength; for the data problem that the sample point density of the actual dataset and the area of the test area are small, the K-value adaptive value method of WKNN algorithm and Bayes algorithm is proposed; for the problem of missing or weak signals, the calculation method of Bayes probability is adjusted. The RSSI fingerprint data in an indoor environment is measured by the hardware platform of ARM. The experimental results show that the localization accuracy and stability of the two improved algorithms in the experimental scenario are better than original algorithms.



Previous Abstract Return to Session C2 Next Abstract