Research On Indoor Joint Positioning Technology In Complex Environment

With the growing safety requirements of modern social production activities,in the complex indoor environment such as hydropower stations,the demand for high-precision positioning such as personnel positioning and object positioning is increasing.In the line-of-sight environment,Ultra-wideband(UWB)positioning technology can achieve centimeter-level accuracy.However,in complex indoor non-line-of-sight(NLOS)environment,occlusion of obstacles can lead to large UWB ranging error and even signal transmission interruption.As an autonomous navigation technology,inertial navigation technology is not affected by environment.But as the inertial navigation system running time increases,the positioning error accumulates.This thesis combines ultra-wideband and inertial navigation technology,focuses on indoor positioning problem in complex non-line-of-sight environment,and completes the following researches:Firstly,based on the mastery of UWB positioning technology and inertial navigation technology,the UWB and inertial navigation joint positioning algorithm based on nonlinear kalman filter algorithm is studied,including the joint localization algorithm based on extended kalman filter(EKF)and the joint localization algorithm based on unscented kalman filter(UKF).By systematically modeling the joint positioning system,the state equations and observation equations of the system are obtained and introduced into the EKF algorithm and the UKF algorithm respectively.Then,the two joint positioning models are designed and compared.The simulation results show that compared with UWB time of arrival(TOA)location algorithm,the root mean square error of EKF and UKF-based joint positioning algorithm both have a certain degree of decline.In complex non-line-of-sight environment,using ultra-wideband and inertial navigation data to estimate the location of tags can be classified as a time series prediction problem.Therefore,a joint positioning algorithm based on long short term memory(LSTM)network is proposed in this thesis,and then its overall architecture design,data preprocessing method,network structure design and model training method are studied.The simulation results show that compared with UKF-based joint positioning algorithm,both in the diamond path and the circular path,the positioning performances of LSTM-based joint positioning algorithm significantly improve.Finally,an indoor joint positioning system is designed and implemented in this thesis.For the function verification of the joint positioning system and joint positioning algorithm,the system function testing experiment and dynamic joint positioning experiment are designed and conducted in this thesis.The results of dynamic joint positioning experiments show that compared with the TOA localization algorithm,in the non-line-of-sight environment with four obstacles,the average distance error of LSTM-based joint localization algorithm drops significantly.