Research On Indoor Combined Positioning Based On WiFi/Bluetooth/Barometer

With the popularity and rapid development of Io T devices and sensors,location-based products and application services have become an indispensable part of people’s lives,and the demand for indoor precise location services is becoming more and more urgent.The prevalence of WiFi and Bluetooth modules in mobile devices makes WiFi and Bluetooth positioning technologies have a broad application and development prospect,and it is important to deeply study WiFi and Bluetooth and the positioning technology of their fusion.The paper combines the advantages of WiFi’s perfect infrastructure,wide signal coverage and Bluetooth devices’ low power consumption and easy deployment,and proposes a WiFi/Bluetooth/barometer combination indoor localization method with the assistance of barometer.The main research elements of the paper are as follows:(1)To address the problems of low localization accuracy,poor stability and high computational complexity when the area to be localized is large in large indoor localization scenarios based on the fingerprint library localization method,an Improved Sparrow Search algorithm based on fusion clustering is proposed to optimize the fingerprint library indoor localization algorithm of the Kernel Extreme Learning Machine.The paper adopts fusion clustering to optimize the traditional K-Means algorithm,and the selection of K and clustering centers is first completed by hierarchical clustering combined with Silhouette Coefficients and Error Sum of Squares,and then the similarity calculation formula of physical coordinate weights is introduced to avoid inaccurate fingerprint point classification;compared with the poor localization accuracy caused by the random determination of the input layer and implied layer neuron weights of the traditional Extreme Learning Machine,the kernel function is used instead of random mapping matrix,and then the sparrow search algorithm with multi-strategy optimization is used to find the optimization of the kernel parameters to obtain the localization model of each partition.The experimental results show that the algorithm can achieve better localization compared with other traditional algorithms,the accuracy and stability of localization are significantly improved.(2)To address the problems that WiFi and Bluetooth fingerprint localization methods require a large number of labeled training samples and that single-mode localization accuracy and stability are difficult to meet the requirements of large-scale localization scenarios,a Multifeature Fusion Semi-supervised Extreme Learning Machine(MFSELM)based on WiFi and Bluetooth is proposed.The method utilizes the mutual correlation theory to fuse the WiFi and Bluetooth fingerprint feature vectors for feature extraction and obtains a mutual correlation feature with high stability.In addition,manifold constraints are constructed for the bluetooth,WiFi and mutual correlation features to enhance the learning ability of unlabeled samples.Test analysis shows that the proposed multi-feature fusion improves localization accuracy by about25% while improving stability compared to single features;the number of labeled samples required in the localization process can be reduced by about 90% when using the semisupervised learning method of constructing manifold constraints separately.(3)The paper designs and implements an indoor 3D positioning system based on smartphones as the mobile carrier by using barometric pressure sensors to achieve the determination of floors in the indoor environment based on the combination of WiFi and Bluetooth indoor accurate 2D positioning.In the offline phase,the positioning system is based on the cell phone APP to complete the collection of WiFi,Bluetooth and air pressure data in the positioning area and upload them to the server side.In the online phase,the above data are collected in real time using the mobile phone APP and communicated with the server side,and the ISSA-KELM and MFSELM algorithms proposed in the thesis are used to estimate the 2D location coordinates of the user,and then the difference in air pressure is used to calculate the location of the user’s floor to achieve indoor 3D positioning.