Reserch On Terminal Indoor Location Positioning Technology With WiFi And Inertial Sensor

With the increasing demand for location services in the fields of personal location navigation,emergency rescue,industry location and so on,the demand for mobile terminal indoor location technology is becoming increasingly obvious.Therefore,indoor positioning technology based on mobile terminals is becoming a research hotspot at present.Because WiFi has the advantages of extensive indoor coverage and deployment,and intelligent terminals are generally integrated with various inertial sensors,the fusion positioning technology based on WiFi and inertial sensors has become the mainstream technology of terminal indoor positioning.This paper mainly focuses on the fusion positioning technology and algorithm based on WiFi and inertial sensors to carry out the following research.Firstly,in order to solve the problems of large amount of on-line matching calculation and low positioning accuracy caused by the fixed number of Euclidean distance and neighboring points in traditional WKNN location fingerprint matching positioning algorithm when indoor WiFi positioning system processes large area,a WiFi positioning algorithm based on k-means spatial partition and EWKNN matching is proposed to perform partition fast index and accurate positioning estimation.Firstly,the algorithm divides the location area according to the k-means algorithm,which solves the problem of large computation caused by the matching process between the point to be measured and the fingerprint point,and realizes the fast index location of the partition.Then,in order to solve the problem that the traditional WKNN algorithm uses weighted positioning estimation with fixed number of neighbor points,which will lead to the increase of positioning error caused by the participation of neighbor points with low similarity to the point to be measured,an adaptive EWKNN algorithm is proposed to adaptively select the number of neighbor point fingerprints,eliminate the error caused by remote neighbor points,and improve the positioning accuracy and robustness.The experimental comparison results show that the WiFi positioning algorithm based on k-means spatial partition and ewknen matching achieves a positioning accuracy of 1.65 m under the condition of 3d Bm Gaussian noise,which is 8.63% better than the traditional WKNN algorithm,and the positioning matching computation is reduced by more than 50%.Secondly,in order to reduce the influence of accumulated errors in PDR positioning,the algorithms of step frequency detection,step size estimation and heading angle estimation are improved.Aiming at the difficulty of the step frequency detection algorithm based on fixed threshold to adapt to the pace change and pedestrian difference,a step frequency detection method based on dynamic threshold constraint peak-valley value discrimination is proposed.Experimental results show that the step frequency detection accuracy of the algorithm reaches 96% on average,and it can adapt to the gait changes of pedestrians.In order to improve the accuracy of step size estimation,a Kalman filter adaptive nonlinear step size estimation algorithm is proposed.Experimental data show that compared with the traditional nonlinear step size estimation method,it has higher accuracy and universality.In order to make the heading angle estimation closer to the actual trajectory,the improved heading angle estimation method is to filter the mean of heading angle and judge the turning criterion.The experimental results show that the improved heading angle estimation method is closer to the actual trajectory and has less fluctuation.Finally,in order to solve the problems of large fluctuation of WiFi positioning results and accumulated errors in PDR positioning in single system positioning,a WiFi-PDR fusion positioning algorithm based on unscented Kalman filter(UKF)is proposed.Using PDR positioning to predict the next position;The current WiFi positioning result and PDR step size,step frequency and heading angle estimation are used as measurement information to update the PDR prediction value.The experimental comparative analysis shows that the algorithm improves the problem of long-term accumulated errors in PDR positioning,and its positioning accuracy reaches 1.52 m,which is 33% higher than WiFi positioning alone.It has good robustness when heading changes,and can be applied to high-precision indoor positioning systems in complex environments.