Research Of Error Mitigation In UWB Positioning System Based On Machine Learning

With the widespread popularity of Io T devices and smart devices,location-aware services have played an increasingly important role in life.In recent years,the huge application and commercial potential of indoor location-based services has stimulated the development of location-based services related technologies and industries towards indoor location-based services.Among the existing network technologies,ultra wideband(UWB)indoor positioning technology stands out among many wireless positioning technologies due to its advantages such as low power consumption,good anti-multipath effect,high security,low system complexity and high positioning accuracy.At present,based on UWB's higher transmission rate and better real-time performance,combined with the time difference of arrival(TDOA)positioning algorithm,it can roughly meet the requirements of wireless positioning.However,when this positioning method is blocked by obstacles such as walls or furniture during the propagation of wireless signals,it will produce multipath effects and non-line-of-sight errors,which will eventually result in low positioning accuracy and poor positioning realtime performance.At the same time,the existing methods for identifying and suppressing non-line-of-sight errors are difficult to implement in a complex indoor environment.This thesis focuses on the indoor positioning technology in the non-line-of-sight environment,and combines the non-line-of-sight error suppression algorithm with the traditional positioning algorithm.The main research contents and results include:Firstly,this thesis studies UWB point-to-point positioning under wooden and iron obstacles,analyzes the non-line-of-sight error of different propagation distances under different obstacles,and can quantify the impact of non-line-of-sight propagation.Comprehensive consideration of the distribution of non-line-of-sight errors under different obstacles provides a theoretical basis for the recognition of non-line-of-sight errors.Secondly,this thesis studies the error suppression algorithm based on machine learning,and uses a sliding window to dynamically calculate the variance and mean of multiple sets of TDOA measurement values in a short period of time as the supplementary input of the regression model.Under static positioning,it can significantly improve nonuniformity.Based on the unsupervised learning non-line-of-sight error recognition algorithm,the above features are used as the input of the recognition algorithm,and they are divided into three categories: within-line-of-sight,low non-line-of-sight error and high non-line-of-sight error.Positioning strategy to achieve more precise positioning.Design simulation experiments verify the effectiveness of the algorithm,and verify the effectiveness and stability of the proposed algorithm based on the existing UWB actual measurement positioning data.