Research And Implementation Of Indoor Positioning System Based On LSTM And TCN

With the improvement of people's living standards,indoor positioning service is now becoming more and more important and this technology is widely used in various life services.In recent years,many researchers have joined the research of indoor positioning technology and proposed a variety of interrelated technologies.Fingerprint localization algorithm has become the mainstream method of indoor location technology.It is worth mentioning that the Bluetooth Low Energy(BLE)technology proposed in 2012 and the Apple Beacon technology proposed in 2013 have enabled the Bluetooth-based fingerprint localization algorithm to be used in many fields.The key step of the fingerprint localization algorithm is to find out the'locatio n fingerprint"corresponding to the tocalization area through the data in database,which corresponds to the process of learning the features from the data by deep learning algorithms.With the rapid development of artificial intelligence(AI)and deep learning technologies,more and more indoor positioning researchers apply the experiences of deep learning technologies into the fingerprint localization algorithm Experiments have shown that the fingerprint localization algorithm based on deep learning technologies has significantly improved the positioning accuracy comparing with the traditional fingerprint localization algorithms.Based on the introduction of some deep learning and neural networks technologies,this paper proposes two fingerprint localization algorithms based on deep learning technologies,which are the Long Short-Term Memory Network Based on Attention Mechanism and Temporal Convolutional Network.These two methods are different from some of the previously proposed fingerprint localization algorithms based on deep learning technologies.Combining the wireless signals according to the reception time to form a wireless signal sequence,and time domain continuity between each wireless signal in the sequence.Therefore,based on the spatial information of wireless signal in the positioning process,considering the time information of the wireless signal in the positioning process,using the Long Short Term Memory Network and the Temporal Convolutional Network extracting the features ofthe signal sequence and then perform the final position calculation.In order to verify the effectiveness of the algorithms,firstly,the actual positioning scene is established in the underground garage of the Innovation Building in the campus of Shandong University,the Bluetooth Beacon is arranged in the scene,and the data collection work is performed.These signals are then collated to create a fingerprint database,followed by the design of the neural network structure,using the data in the database to train and test the network.Finally,comparing the proposed algorithm with existing fingerprint localization algorithms based on deep learning technologies,it can be seen that the proposed method significantly improved the positioning accuracy.In order to visualization of the effect and the integrity of the system,for the case of multi-drivers,the Android client software and the server host software for the application scenario of the undergrournd garage is designed,realizing the real-time navigation and positioning,storage,reverse car search,real-time personnel monitoring,personnel track record saving,hotspot map displaying and other functions.In addition,in order to verify the effectiveness of the systen,on-site tests were conducted and some intermediate key processes were recorded.The experiments proved that the proposed system has practical application value.