Research On Multi-source Fusion Positioning Method Based On UWB/PDR/map Information

With the increasing popularity of location services,the demand for indoor location services is growing.In the field of indoor positioning research,a variety of indoor positioning technologies have been proposed,including UWB positioning,inertial navigation positioning,and pseudolite positioning technology.Each indoor positioning technology has its advantages and disadvantages.However,the indoor environment is complex,and it is difficult for a single positioning technology to achieve full coverage and high-precision positioning.Therefore,the fusion of multiple positioning technologies has become the main solution.This paper studies UWB positioning technology,studies pedestrian dead reckoning technology based on MEMS-IMU,and proposes a multi-source information fusion positioning model based on UWB/PDR/map model.The main research contents are:Aiming at the problem of ranging error in UWB indoor positioning,this paper analyzes the error source of the TW-TOF ranging method,and establishes a ranging error correction model by polynomial fitting.After the fitting and verification of the ranging value,the error of the original ranging is reduced by 72.06%,which weakens the influence of the systematic error.Aiming at the problem of non-line-of-sight error,this paper proposes an improved robust Kalman filter algorithm,which adjusts the size of the gain matrix according to the size of the prediction residual vector to suppress the influence of outliers on the positioning results.Through comparative analysis,it has the advantage of higher positioning accuracy.Aiming at the problem of few-station positioning when the base station is abnormal,a fewstation positioning method based on LSTM trajectory prediction is proposed.This method trains the LSTM network through historical trajectories,and predicts future trajectories when the positioning base station signal is lost.The experimental verification shows that after the abnormal shutdown of the positioning base station,although there will be an "inertial" error phenomenon at the corner,the overall positioning trajectory is basically accurate,and the predicted positioning error is within 0.2m,which can meet the positioning requirements in the case of temporary sudden base station failure.The positioning accuracy is better.For pedestrian gait detection,the pedestrian gait cycle is obtained by analyzing the acceleration data in time domain and frequency domain.Through the analysis of each parameter,a multi-parameter constraint gait detection model is established.The experimental verification shows that the detection error is about 1%,and it has high detection accuracy and stability.In terms of step size estimation,linear and nonlinear models were fitted by walking experiments,and the estimated error of the Weinberg model was about 2.63 m.Aiming at the cumulative error of heading estimation,a heading correction method based on graphical model is proposed,which improves the accuracy of heading angle estimation.In summary,the optimization of the three technologies improves the PDR positioning accuracy and provides a guarantee for the fusion positioning.Aiming at the problem of multi-source information fusion,a multi-source information fusion positioning model based on UWB/PDR/vector map is proposed.The model integrates UWB positioning information,PDR estimation results(gait,step length,heading angle,etc.)information,and vector map information.And designed the indoor environment closed path experiment,verified the fusion positioning algorithm,and compared and analyzed the results of UWB positioning and PDR positioning.The experimental results show that the particle filter positioning model based on UWB/PDR/vector map and other multi-source information fusion has higher positioning accuracy,better positioning stability,smoother trajectory and more in line with the real trajectory.The positioning accuracy of the fusion positioning method is about 7% higher than that of UWB positioning,and 52.6% higher than that of PDR positioning.Moreover,the vector map information can also provide effective constraints on the PDR heading angle,step size estimation,etc.,and effectively suppress the cumulative error generated during the operation of the MEMS-IMU,so that the final positioning trajectory is more in line with the actual walking trajectory.