Research On WLAN Indoor Localization Based On Channel State Information

With the development of the Internet of things and the mobile Internet,the demand for indoor localization is on rapid growth in recent years.In outdoor environments,the satellite positioning system can be used to obtain the current positon.However it does not work well in indoor environment because the signals are affected by buildings.Because of the popularity of current indoor WLAN access,many indoor positioning systems do not need the installation of additional localization devices.At present,indoor localization based on WLAN uses Received Signal Strength Indicator.However,multipath effect leads to low precision of indoor localization based on Received Signal Strength Indicator.The physical layer Channel State Information(CSI)with high granularity can better adapt to the indoor multipath environment to improve the accuracy of indoor localization based on fingerprint.In this thesis,we conduct the indoor localization research based on CSI.The major contributions are outlined as follows:In this thesis,an indoor localization scheme based on CSI fingerprint is proposed.This scheme only uses 1 AP as the signal source,which includes data preprocessing,offline training and on-line localization.Data preprocessing will be done in the offline and online phases for dealing with abnormal data.Based on the distribution characteristics of CSI,the Kmeans algorithm is used to extract the fingerprint and improve the discrimination between different positions.In the online phase,the number of clusters is determined dynamically according to the distribution of CSI.Then,the clustering results are compared with the reference points,and some fingerprints are removed to further improve the accuracy of the fingerprint.Finally,by using 3 kinds of matching methods,we can make use of the fingerprint information reasonably and reduce the influence of the limitation on the fingerprint information.Based on the Kmeans clustering scheme,the Kmeans-MIMO scheme is proposed based on the independence of the CSI between the MIMO antennas.Based on the analysis of the position estimation value of the multiple antenna pairs,the position estimation value of the antenna pair which has a good fingerprint characteristic and is matched with the reference point is selected as the final position estimation value.This scheme can solve the problem that the number of data packets is limited in the localization stage because of the limitation of the localization time,and there is no obvious difference in the location of the fingerprint generated by the antenna.Through the selection of antenna pair,the localization accuracy is further improved.At last,the performance of the localization scheme is evaluated from several angles.Experiments show that,in the case of using a single AP Kmeans-MIMO,the average errors of the proposed scheme in two experimental scenarios are 1.14 meters and 1.01 meters.