Research On Indoor Location Algorithm Based On Wi-Fi/Bluetooth Fusion

Location-based services have a wide range of applications in large shopping malls,airports and other indoor environments.As we continue to expand indoor activity areas for office and living,rapid and accurate positioning of users in large indoor Spaces has a high value for location-based services.Traditional indoor positioning research uses a single signal source for positioning.With the continuous improvement of positioning requirements,the integration of multiple signal sources for positioning is increasingly becoming a trend.In this dissertation,the positioning algorithm based on Wi-Fi/ Bluetooth fusion is studied.After the positioning results are obtained by combining Wi-Fi/ Bluetooth signals at the data layer and Bluetooth separately,the positioning results of Wi-Fi/Bluetooth are integrated at the decision level.The main research contents of this dissertation are as follows:Firstly,aiming at the problem that the positioning method using multi-mode signal source is not high enough in large indoor environment,a deep learning model semi-AE-BLSTM based on information bottleneck theory is proposed for indoor positioning.semi-AE-BLSTM is composed of an encoder based on an encoder network and a predictor based on a bidirectional long term and short term memory network.The data layer feature fusion of Wi-Fi/ Bluetooth signals is carried out through the autoencoder network.The bidirectional long term and short term memory network predictor trains the feature of the middle layer vector of the autoencoder network and completes the position coordinate prediction.Comparative experiments show that the proposed semi-AE-BLSTM model can effectively improve the positioning accuracy compared with other traditional KNN algorithms.Secondly,aiming at the problem of instability caused by large error of traditional ranging method of Bluetooth,a quadrilateral centroid location algorithm based on quadratic weighting is proposed.The ranging method and range loss model of Bluetooth signal are analyzed,the traditional centroid positioning method is improved to four-sided centroid positioning,and its class is introduced in practical application scenarios.Considering that AP nodes with different distances have different influences on the measured coordinates,the positioning coordinates obtained by the four-sided centroid positioning algorithm are weighted twice.The experimental results show that compared with the least square method,the quadrilateral centroid location algorithm based on quadratic weight has better positioning performance.Finally,in view of the respective shortcomings of Wi-Fi and Bluetooth positioning technologies,the positioning results of Wi-Fi and Bluetooth are integrated at the decision-making level,combining the advantages and advantages of the two,so as to obtain higher accuracy and more stable positioning results.According to the positioning requirements,the fusion algorithm of Wi-Fi/ Bluetooth location results based on improved Bayes is adopted.Experimental analysis shows that this algorithm can obtain more accurate and stable positioning effect.