High-precision WiFi indoor localization algorithm based on CSI-XGBoost

Considering the influence of complex indoor environment and multi-path effects on the WiFi fingerprint positioning performance, this paper extracts channel state information (CSI) from the Intel 5300 wireless network card and utilizes the modified CSI amplitude and phase information as fingerprint features. A high-precision fingerprint database was built using the extreme gradient boosting (XGBoost) algorithm to achieve indoor positioning at a decimeter level. Experiments in the actual indoor environment have been conducted to evaluate the effects of sampling interval, line of sight (LOS) and non line of sight (NLOS), missing values, and data dimensions on the localization performance of the proposed method. The results of real indoor experiment show that the proposed CSI-XGBoost method is less affected by NLOS and robust to complex indoor environments. In addition, this method can handle high-dimensional sparse data well and solve the mismatching problem of CSI fingerprinting. Moreover, this method is insensitive to missing data, with localization accuracy of better than 90%. © 2018, Editorial Board of JBUAA. All right reserved.