Research On WiFi Indoor Positioning Technology Based On Fingerprint Optimization And GRNN

With the development of the Internet of Things industry and the maturity of wireless communication technology,services based on location information have gradually attracted attention.Under outdoor conditions,Global Navigation Satellite Systems(GNSS)combined with some ground-based or space-based systems can almost meet the user's location information requirements with any accuracy,but under indoor conditions,due to the occlusion of buildings and indoor objects,It is difficult for users to receive real-time and effective satellite signals,and satellite positioning technology cannot meet the needs of users for indoor location information.Therefore,researching an accurate and effective indoor positioning technology has become a hot topic in the current positioning research field.Due to the wide coverage,low cost,easy expansion and simple deployment of WiFi,WiFi positioning technology has become the main technology for indoor positioning.However,the indoor environment is complex.During the indoor propagation of WiFi signals,due to the influence of people walking,multipath effects,electromagnetic interference and other factors,it exhibits a high degree of uncertainty,which seriously affects the positioning effect of WiFi positioning technology.There are still huge challenges in practical applications.This paper is based on the theory of location fingerprint positioning.Aiming at the problem of the uncertainty of WiFi signal in the positioning process,focusing on reducing the impact of this uncertainty on all aspects of WiFi positioning,applying some theoretical results in the field of deep learning,the WiFi location fingerprint The problems of positioning have been studied,and corresponding algorithm improvements have been made.The research content mainly includes the following aspects:(1)Analysis of WiFi signal characteristics.Taking into account the complexity of the indoor environment,WiFi signals will be affected by a variety of uncontrollable factors during the indoor propagation process.This article uses statistics and analysis of the fingerprint data collected in the experimental area to conclude: First,due to multipath effects,signals Influenced by factors such as interference and people moving,the Received Signal Strength(RSS)of the WiFi signal presents a random fluctuation in time.Second,the RSS distribution of the WiFi signal does not completely obey the normal distribution.In the actual situation,the phenomenon of bimodal distribution will appear randomly;third,increasing the RSS dimension of the WiFi signal can improve the distinguishability of different locations in the RSS signal space to a certain extent,but it will also increase the calculation of the algorithm accordingly.the amount.(2)Research on optimization algorithm of fingerprint database.Taking into account the timevarying characteristics of the RSS of the WiFi signal and the distinguishing characteristics of the RSS at different locations,the optimization method of the traditional fingerprint library is improved for the purpose of reducing the influence caused by the uncertainty of the RSS.First,in view of the GRNN WiFi AP redundancy problem in WiFi location fingerprint positioning,an AP selection algorithm based on mutual information is proposed.The algorithm is used to remove redundant APs in the sample library and select the best reference point for each sample in the fingerprint library.AP collection;then,for the search efficiency of the best sample in the fingerprint matching algorithm,the algorithm of Dividing the Fingerprint database based on the Optimal Aps(DFOA)is proposed.The DFOA algorithm is introduced into the K-Nearest Neighbor(KNN)to build the DFOA-KNN indoor positioning model.It is found through experiments that the DFOA algorithm can effectively reduce the computational complexity of the indoor positioning algorithm and significantly improve the algorithm's performance Positioning efficiency and positioning accuracy.(3)Research on indoor positioning algorithm based on multilayer neural network.Taking into account the dependence of the Generalized Regression Neural Network(GRNN)positioning prediction model on the model parameters,the GRNN positioning algorithm is improved for the purpose of reducing the influence of human factors and improving the positioning effect.First of all,to solve the problems of particle swarm optimization and poor convergence accuracy,the Linear Dynamic Change factor Particle Swarm Optimization algorithm(LDCPSO)is proposed;Integrating the respective performance advantages of the LDCPSO algorithm,the Cukoo Search(CS)algorithm and the K-means algorithm,a hybrid optimization algorithm of LDCPSO and CS based on multiple clusters is proposed;finally,the algorithm is used to optimize GRNN Network,find the best model parameters of GRNN,establish an indoor positioning prediction model,and get through experiments.Compared with the positioning model built by traditional positioning algorithms,the positioning model built in this article has better positioning accuracy and positioning efficiency.