Design And Implementation Of UWB Three-dimensional Accurate Indoor Positioning Algorithm Based On Machine Learning

With the rapid development of the intelligent manufacturing,the research and application of industrial indoor positioning technology have developed rapidly.In industrial production,the indoor positioning technology can be used to locate and track goods,personnel,dangerous goods,etc.,thereby helping to simplify management,improve production efficiency,and reduce production risk factors.At present,most indoor positioning systems have the problem of floating positioning coordinates during positioning,and the problem of positioning failure caused by abnormal positioning data.The occurrence of these problems will greatly reduce the positioning accuracy and stability of the positioning system.At the same time,most of the current indoor positioning algorithms are positioned in a two-dimensional coordinate system,and there is a problem that they cannot truly reflect the height coordinates of the tag nodes.Therefore,in this paper,the research and testing of indoor positioning algorithms are carried out in a three-dimensional coordinate system.Firstly,for the problem of floating positioning coordinates of the tag nodes,after analyzing and processing the collected positioning data,it is found that the positioning data collected each time when the tag nodes are stationary is different,and there are certain errors,which leads to position coordinates to float.From the perspective of the tag nodes motion state classification,the motion state is judged by using a machine learning classification algorithm,which the acceleration and the difference in distance are specified as the model feature for model training.The resulting classification algorithm model can judge the motion state well.According to the determined motion state,it is further determined whether the position coordinates of the current tag node need to be recalculated,thereby solving the problem of floating position coordinates,and improving the stability and accuracy of the indoor positioning system.Secondly,to solve the problem that the indoor positioning system has no solution or misinterpretation due to abnormal positioning data,after preprocessing the collected positioning data,the analysis found that it belongs to time series data.From the perspective of prediction processing of positioning data,through the use of time series prediction models and machine learning prediction algorithms,the prediction processing of abnormal positioning data is performed to avoid the failure of positioning.By dividing the collected positioning data into data sequences,training the prediction model,and analyzing the prediction effects of the prediction model,it is found that the positioning data prediction process can solve the problem of no solution or misinterpretation,and further improve the stability and accuracy of the indoor positioning system.Finally,according to the solutions of the above two problems,the integrated processing of the 3D indoor positioning algorithm is performed.By using the UWB-based asynchronous TDOA positioning model,the clock synchronization problem faced by the indoor positioning system is successfully solved,and the system complexity is reduced to a certain extent.Then,by using the UWB wireless communication method based on TDMA,each tag node in the system is successfully allocated time slots to allocate communication resources,thereby avoiding the problem of data collision,and realizing real-time positioning of multi-label nodes.In this paper,the three-dimensional indoor positioning test system actually constructed is tested.The test results verify that the three-dimensional indoor positioning algorithm has high positioning accuracy and stability.