| With the rapid development of mobile internet,location based service(LBS)in large public indoor places,such as shopoing malls,offices,and airports,have become increasingly popular.Therefore,the study of indoor localization technology has become particularly urgent.Bluetooth Low Energy(BLE)based indoor localization has attracted increasing interests for its low cost,low power consumption,and ubiquitous availability in mobile devices.In this paper,the current situation of researches on indoor localization is studied,and the advantages and disadvantages of various indoor localization technologies are analyzed from two aspects:measuring techniques and algorithms.A 3D bluetooth indoor localization method based on autoencoder is proposed.The method combines a deep learning model,and fingerprinting,and extends the 2D fingerprinting into 3D fingerprinting to realize high-precision 3D indoor localization.The denoising autoencoder is introduced into this method to cope with unpredictable BLE beacon lost and to further improve the localization accuracy and stability.A deep learning model,called autoencoder,is adopted to extract robust fingerprint patterns from RSSI measurements to replace the traditional fingerprint.A fingerprint databased is then constructed with reference locations in 3D space,rather than traditional 2D plane.In localization phase,the posteriori probabilities that the target locates at each reference location are calculated from distance between RSSI measurements and that reconstructed from each autoencoder.Finally,experiments are designed to compare with the traditional fingerprinting method in terms of three aspects:horizontal localization accuracy,vertical localization accuracy and localization accuracy in case of beacon lost.The results show that our method can not only achieve the accuracies of 1.09m and 0.34m in horizontal and vertical localization,but also retain high accuracy in case of beacon lost,which proves the effectiveness,accuracy and stability of our method. |
PDF Full Text Request