Research And Fusion Implementation Of Indoor Location Method Based On Wi-Fi And Inertial Technology

Based on the indoor location service,personnel positioning,security rescue,accurate advertising and others have great application value.Mature GPS location services,because satellite signals are attenuated when they penetrate buildings,they can't provide reliable indoor location services.In order to solve the problem of indoor location services,researchers have proposed a variety of different indoor positioning technologies.This paper studies the existing indoor positioning technology at domestic and foreign,and analyzes the reasons why indoor positioning services cannot be popularized.The key research work is carried out around Wi-Fi fingerprint positioning and inertial technology positioning.The focus includes Wi-Fi fingerprint positioning based on Gradient Boosting Decision Tree(GBDT),the positioning error correction model based on Random Forest(RF),Pedestrian Dead Reckoning(PDR)based on Android mobile phone inertial sensor positioning and the implementation of the fusion positioning of the two positioning methods,the main research focuses are as follows:In this paper,the GBDT algorithm is used to improve the positioning performance of the fingerprint matching positioning model.The data collected in the offline phase is used as the training input of the GBDT algorithm to establish an online positioning model.Afterwards,aiming at the problem of accuracy reduction of positioning model caused by indoor environment change,the positioning error correction model based on RF is proposed.The model uses the GBDT positioning model to predict the position coordinates in the current environment,and calculates the error between the predicted and real coordinates in the x,y direction.Using the error and received signal strength as RF training inputs to establish a nonlinear regression model of signal strength and error.In the online positioning stage,the trained RF model is used to predict the positioning error,and then the positioning coordinates predicted by the GBDT positioning model are corrected to obtain the final positioning result.Aiming at the traditional peak method,when the gait changes,the detection accuracy is reduced.An improved gait detection model is proposed.The peaks and valleys of the acceleration data are constrained by setting time thresholds and amplitude thresholds.After that,the step size estimation and heading angle detection were analyzed,and the PDR positioning based on Android mobile phone was verified.Finally,for the problem of poor continuity of Wi-Fi positioning trajectory,obvious bunching and jumping phenomenon and PDR positioning can't eliminate the cumulative error,two models of Wi-Fi and PDR fusion positioning are proposed.One is the fusion positioning model based on extended Kalman filter.Second,the fusion positioning model based on conditional decision.Experimental data shows that the fusion positioning model can effectively improve the problems existing when the two are used alone.