Study On The Robustness Of UWB/PDR Positioning Algorithm Based On Graph Optimization

The demand of realizing high precision indoor positioning promotes the vigorous development of indoor positioning technology.Inertial navigation technology and UWB positioning method are the main indoor positioning technologies nowadays.The UWB/PDR positioning method based on graph optimization fuses the data of two sensors for positioning.High positioning accuracy can be achieved under the ideal condition,but there are inevitable errors in the fusion process.One is the error in the process of estimating orientation,the other is the error resulting from the refraction and reflection of UWB signal in NLOS environment.The above errors will reduce the positioning accuracy.There are two main ways to improve the robustness of the fusion positioning algorithm based on graph optimization: One is to reduce the error in the front end,the other is to add the robust algorithms including robust kernel function at the back end to reduce the influence of errors in optimization.Aiming at the reasons causing errors in the process of UWB/PDR fusing positioning,two studies are carried out to improve the robustness of UWB/PDR positioning algorithm based on graph optimization.Firstly,in order to reduce the influence of orientation error effectively,the improved Madgwick orientation estimation method and PDR dynamic confidence method are proposed in this thesis.The improved Madgwick orientation estimation method uses low-pass filtering to process the sensor data to reduce the impact of errors on positioning results.This method uses Dogleg algorithm to replace the gradient descent method to update the position calculated jointly by the accelerometer and magnetometer.PDR dynamic confidence method is proposed for the PDR error which cannot be eliminated in the front end,considering with the UWB errors.The method restrains the error by setting the threshold and adjusting the confidence of the PDR and UWB constraint.Experiments show that both methods can improve the robustness of the fusion positioning algorithm and achieve high positioning accuracy.In order to solve the problem that UWB constraint greatly affected by NLOS noise,IDCS robust kernel function and IWelsh robust kernel function are proposed in this thesis.Firstly,it is found that the basic robust kernel function is weak in suppressing the error under large NLOS noise.DCS function and Welsh function are reconstructed with UWB sensor characteristics.The experiments show that IDCS algorithm and IWelsh algorithm can improve the robustness of UWB/PDR algorithm based on graph optimization.It can be concluded that the robust kernel function hasbetter precision and robustness than the basic robust kernel function.This thesis has 51 maps,8 tables and 80 references.