Research On UWB Indoor Localization Method Based On Deep Learning

As a new indoor positioning technology,UWB positioning has a higher accuracy than traditional indoor positioning technology and can achieve centimeter-level positioning.It is of practical significance to carry out research related to UWB positioning technology,which can play a significant role in production and life.This paper studies the UWB high precision positioning algorithm and develops the positioning software which can be used in practice.The main research contents and innovative achievements include:(1)Aiming at the problem that TDOA method and TO A method in classical positioning schemes are greatly interfered by non-line-of-sight errors,a machine learning-based LOS and non-line-of-sight signal(NLOS)filter is proposed.In the traditional UWB location algorithm,when there is an obstacle between the label and the base station,the signal will be scattered and rebounded,which seriously affects the measurement accuracy.In this paper,convolutional neural network and support vector machine are used to filter NLOS signals to improve the stability of localization algorithm.(2)Using the inertial navigation to enhance the positioning effect of UWB positioning system in complex environment.Based on Kalman filter,this paper realizes UWB-INS fusion positioning,and uses INS to guide and correct the possible problem of positioning coordinate point drift in UWB positioning system.The experimental results show that the fusion localization method proposed in this paper can achieve good localization effect.(3)Constructing the hardware and software platform of UWB positioning system.The upper computer software of UWB positioning system is developed in C++11 environment.The software includes several module functions such as positioning and solving module,data playback module,map display module,data export module,log module and so on,and integrates the innovative algorithm in this paper.The four base station positioning system is deployed in the actual scene,and the field test shows that the positioning and tracking function can be realized and the positioning effect is good.The positioning error of moving object is less than 10cm,which proves the effectiveness and portability of the proposed method.