Research On UWB/IMU Indoor Location Method Based On Machine Learning And Improved SRCKF

As people pay more and more attention to personal location service in production and life,indoor positioning technology has become one of the hot spots in the field of navigation and positioning.Satellite navigation technology has been widely used in the field of outdoor positioning,but it is not suitable for indoor positioning because the satellite signal is easily blocked by buildings.Ultra wide band(UWB)has the advantages of high transmission rate,low power consumption and strong penetration,but its signal propagation is easily affected by non line of sight(NLOS);The positioning method based on inertial measurement unit(IMU)is not affected by external information,all-weather,autonomous positioning and other advantages.However,due to the accumulation of errors,it is not suitable for long-time positioning.UWB / IMU integrated positioning has become one of the research hotspots in the field of indoor positioning.In this paper,the high precision,high reliability and sustainable positioning method of UWB / IMU combination is studied.Firstly,the UWB / IMU integrated positioning experimental data acquisition system is built.The UWB-mini3 s ultra wideband module is used to measure the distance between the mobile target and the base station,and the WT901 C inertial measurement unit is used to measure the acceleration and rotation angular velocity of the carrier,which provides data source for positioning algorithm experiment.Secondly,the principles of UWB positioning and IMU positioning are studied.In view of the characteristics of pedestrian dead reckoning(PDR)such as small calculation and simple implementation,the PDR positioning method based on IMU and UWB are combined.The performance of UWB single location and IMU single location is analyzed,which lays the foundation for UWB / IMU data fusion.Thirdly,the state space model of UWB / IMU integrated positioning system is deduced,and a data fusion algorithm based on robust adaptive square root culture Kalman filter(RASRCKF)is proposed.The improved Sage-Husa method is used to estimate the covariance matrix of measured noise,which can adapt to the change of statistical characteristics of measurement noise;The noise covariance of measurement is modified by the method of robust estimation to restrain the adverse effect of abnormal measurement value.Experimental results show that RA-SRCKF algorithm has higher accuracy and stability than SRCKF algorithm.Fourthly,a machine learning aided UWB / IMU integrated positioning scheme is proposed to solve the problem of continuous positioning in the case of UWB ranging failure.When the UWB ranging is effective,the machine learning model is trained by the results of acceleration,step size,heading,RA-SRCKF position estimation and PDR positioning;In the case of UWB ranging failure,the machine learning model is used to predict and compensate the error of IMU positioning system.BP neural network,random forest(RF)and support vector regression(SVR)are used to assist UWB/IMU combination positioning.The experimental results show that in the indoor environment with limited range of activities,BP neural network and RF assistance have no effect on the positioning accuracy,and the SVR assisted UWB/IMU combined positioning method can reduce the positioning error by49%.The research results of this paper provide a reference for high-precision and reliable indoor positioning algorithm,and lay a theoretical and practical foundation for further implementation of UWB / IMU indoor integrated positioning system.