Research On Indoor High Precision Positioning Algorithm Based On Improved XGBoost

With the rise of intelligent devices and the continuous development of wireless network technology,location-based services(LBS)and applications are becoming increasingly popular.Lbs has been using global navigation satellite for a long time System(GNSS)can navigate and determine the location information of intelligent devices in the outdoor environment,but GNSS can provide accurate location information in the outdoor environment,but it can not achieve the same effect in the indoor environment,because the GNSS signal is too weak in the indoor environment,it is difficult to penetrate the building walls and other obstacles,and it can not provide reliable indoor location-based services Therefore,more and more researchers have carried out research on indoor positioning technology,among which WiFi based positioning technology has been paid more attention due to the popularity of WiFi hardware.In this paper,WiFi indoor location technology based on fingerprint is studied.Firstly,the traditional K-nearest neighbor fingerprint location algorithm is improved by calculating the weighted distance using distance and attribute features.Then,based on the improved algorithm FWKNN(Feature Weighted K-nearest Neighbors),e Xtreme Gradient Boosting is applied to indoor positioning model training to effectively improve the positioning accuracy.The main work is as follows:(1)In order to improve the positioning accuracy of WiFi fingerprint location method based on received signal strength,KNN algorithm is improved.Firstly,the weights of the main feature RSS with different distances are assigned to calculate the weighted distance.Then,the weighted distance is used to match the offline fingerprint database to determine the unknown position.Finally,the positioning accuracy is further improved with the accurate positioning results of Kalman filter.The algorithm unites the benifits of distance weighting and attribute weighting,which can calculate the distance of each point more accurately and select the appropriate nearest neighbor.(2)In order to reduce the computational complexity of the location algorithm of XGBoost,while reducing the influence of noise and outliers and improve data reliability,this paper uses the Density-Based Spatial Clustering of Applications with Noise method with noise based on the weighted distance in FWKNN to remove the influence of interference points,effectively reduce the amount of data involved in the operation and improve the positioning effect of XGBoost.Firstly,the weighted distance proposed in FWKNN method is used to select some nearest neighbor fingerprint points;then,DBSCAN clustering method is used to further screen out the core points;finally,the XGBoost model of machine learning method is used to locate the core points.The experimental results show that the positioning accuracy of the improved xgboost positioning method based on FWKNN is improved,effectively reduces the amount of data involved in the operation,improves the positioning accuracy,and reduces the energy consumption of intelligent devices.