Design Of Bluetooth Indoor Positioning System Based On Neural Network

Due to the complex indoor environment and the non-linearity of Bluetooth signal transmission model,the classical Bluetooth fingerprint positioning suffers from low accuracy,poor robustness,and insufficient environmental adaptability,which makes it difficult to meet the demand for high-precision and highly stable indoor positioning.To improve Bluetooth fingerprint positioning performance,the thesis combines Radial Basis Function(RBF)neural network with fingerprint localization principle to build Bluetooth indoor localization system,combines hybrid filtering algorithm and Beetle Antennae Search(BAS)algorithm to optimize the system performance and improve the accuracy,stability and environment of indoor localization.The main contributions of this study are summarized as follows.1.Based on the analysis of the principle of Bluetooth fingerprint indoor positioning and the factors affecting the positioning performance,the overall scheme of indoor positioning based on RBF network is designed.Based on the Received Signal Strength(RSS)values and hybrid filtering,a fingerprint library is established to improve the accuracy of fingerprint data.The RBF neural network is used for the determination of position coordinates in the localization phase,and the optimization of the RBF network by the clustering algorithm and the BAS algorithm is used to improve the accuracy and stability.2.To address the problem of RSS data fluctuation,the accuracy of fingerprint database establishment is improved by signal filtering.By performing median,mean,Gaussian and Kalman filtering on the collected RSS data,and conducting comparison experiments to propose a hybrid filtering algorithm that combines the advantages of multiple filtering algorithms to improve the stability of signal strength.On this basis,the signal strength difference method is used to weaken the problem of RSS data differences caused by different models of equipment to ensure positioning accuracy and stability.3.The principle of RBF neural network is studied,and a positioning model based on RBF neural network is established to improve the accuracy and environmental adaptability of Bluetooth indoor positioning.Meanwhile,to further optimize the performance,the clustering algorithm and BAS algorithm are used to optimize the performance of the RBF network.The improved K-Means clustering algorithm is used to determine the sample centers,and the initial parameters of the RBF neural network are optimally set based on BAS,so as to obtain further improvement of localization accuracy and stability.4.The experimental system is established based on i Beacon system and thesis algorithm,including i Beacon anchor node,Android-based client,server side and positioning database and other parts.The Android-based client can provide functions such as Bluetooth signal detection,Bluetooth signal scanning,RSS data collection and positioning result display.Based on the experimental system,localization experiments and data analysis were conducted in two indoor environments: small space with large changes in environmental factors and large space with small changes in environmental factors.Theoretical research and experimental results show that the organic combination of RBF neural network and fingerprint algorithm to improve the accuracy of Bluetooth indoor positioning,and the average positioning error is 1.06 m and 0.98 m in both experimental environments,and the confidence probability of positioning error less than 1m is 62% in all case,which can meet the needs of different positioning scenarios and has good environmental adaptability.Combining the optimization algorithm of hybrid filtering,clustering and BAS fusion,the average positioning error is further reduced to 0.75 m and 0.68 m,and the confidence probability of positioning error less than 1m is further improved to 78% and 79%,and the positioning stability is also effectively improved.