Research On UWB Non-Line-of-Sight Propagation Identification And Positioning Algorithm

Due to the continuous development of technology and economy,people’s life concepts and behavior habits have also changed a lot.Indoor positioning services play a very important role in many scenarios such as automatic parking,intelligent medical care and intelligent storage.Ultra-wideband(UWB)indoor positioning technology has been widely paid attention by academia and industry due to its advantages of low power consumption,high temporal resolution,strong anti-interference capability,and high positioning accuracy.However,in the complex indoor environment,non-line-of-sight(NLOS)propagation can seriously affect the UWB positioning accuracy.To address this problem,a series of research works on NLOS identification and UWB positioning algorithms are carried out in this thesis,and the main contents and contributions are as follows:(1)A NLOS identification method based on One-Dimensional Wavelet Packet Analysis(ODWPA)and Convolutional Neural Network(CNN)is proposed.The ODWPA technique is used to transform different classes of Channel Impulse Respond(CIR)into the corresponding color coefficient profiles for constructing image datasets.Two CNN models with different structures are created to achieve 100% and near-100%accuracy in identifying test signals in the target sceneries,respectively,outperforming several classical CNN models while being less computationally intensive.Experimental results in several scenarios show that the proposed method has the highest average identification accuracy of about 95%,which is significantly higher than several other NLOS identification methods,and has the best identification robustness.(2)In order to reduce the UWB ranging errors,two different error models were developed for the LOS/NLOS environment for different sources of ranging errors and large differences between the LOS and NLOS conditions.A series of ranging experimental results and performance comparison analysis show that the polynomial function error model is more suitable for ranging error compensation in LOS scenarios,while the exponential function error model is more suitable for ranging error compensation in NLOS scenarios.(3)In order to further improve the UWB positioning accuracy,a method is proposed to classify the range values corresponding to different classes of UWB signals,and to compensate the error of LOS/NLOS range values using two range models in order to improve the range accuracy.Based on this,the target position solution in static and dynamic scenes is implemented by the joint Chan algorithm and Kalman filtering algorithm,respectively.The experimental results show that,compared with other algorithms,the proposed algorithm has the highest positioning accuracy of 12.6 cm on average in static scenario,which is about 32.8% better.In dynamic scenario,the average error of the proposed algorithm is about 10.7 cm and 8.9 cm in X-axis and Y-axis,respectively,and the positioning accuracy is much higher than other algorithms,with better smoothness and closer to the real trajectory of the tag.The thesis has 45 figures,17 tables,and 85 references.