Research On Indoor Localization Technology Based On WiFi

With the increasing demand for indoor location service,more and more researchers are investing in the research field of indoor localization technology.Among them,due to the widespread deployment of WiFi networks and the popularity of smart mobile terminals,WiFi-based indoor localization technology has become a research hotspot.This paper studies the indoor localization technology based on WiFi,analyzes the research status at home and abroad,compares the advantages and disadvantages of different indoor localization methods,and finally carries out research work based on fingerprinting-based indoor localization.This article studies localization algorithms and the reconstruction and update of fingerprint databases.In order to solve the problem that the WiFi localization method only based on the received signal strength indication has low localization accuracy and the cost of the establishment and update of the offline fingerprint database is high,corresponding improvement schemes are proposed.This article uses Convolutional Neural Networks(CNN)to improve the positioning performance of traditional RSSI fingerprinting algorithms.The received signal strength at each location is used as a training input for CNN and the position estimation is output.Then,from the point of view of error correction,a localization error correction algorithm based on random forest regression is proposed.After the algorithm executes the CNN localization algorithm to obtain the localization coordinates and calculates the localization errors of the horizontal and vertical coordinates.The received signal strength vector and the vertical and horizontal coordinates localization error are used as training sets to train the random forest regression model,aiming to find the nonlinear mapping relationship between the received signal strength and the localization error.In the online localization step,using trained random forest regression model predict the localization errors of the horizontal and vertical coordinates of the localization point,and then the localization coordinate from CNN is corrected.Aiming at the problem of high maintenance and update cost in the offline fingerprint database,in this paper,combining with the Bayesian compressed sensing theory and the similarity of reference point fingerprints,an offline fingerprint database construction algorithm is proposed.The algorithm makes full use of similarity between reference points.By measuring the change in the signal strength of a small number of reference points in a new environment,it can reconstruct the change value of signal from a AP due to the change of the environment in the entire localization area from a certain value.The fingerprinting of the original offline fingerprint database is corrected by the change value,and then the fingerprinting data in the new environment can be obtained.Finally,this paper implements the reference points in the real environment.The actual environment data was collected to verify and analyze the two proposed algorithms.Compared with DNN and SVM,the CNN algorithm proposed in this paper improves the localization accuracy by 7.49% and 22.03%,respectively.The error correction algorithm based on CNN+RF can further improve the localization accuracy.Compared with CNN and CNN+SVR,the localization accuracy is improved by 21.63% and 9.11%,respectively.The similarity-based Bayesian compressive sensing algorithm can reconstruct the fingerprint database well,reduce the workload of the offline stage,and improve the localization algorithm's localization accuracy under the same conditions.