Research And Implementation Of Indoor Localization Algorithm Based On WiFi

Location based services play an important role in people's daily lives,and has shown great development prospects in the fields of target positioning,disaster prevention,emergency rescue,commercial advertising,road assistance and so on.At present,WiFi indoor positioning technology based on location fingerprint has become a research hot spot in the field of indoor positioning because of its advantages such as low deployment cost,strong universality and wide coverage.However,in the complex indoor environment,the time-varying nature of WiFi signals and the complex non-linear mapping between WiFi fingerprints and physical locations make the design of high-precision,low-complexity,and low-cost WiFi fingerprint positioning solutions still face many challenges.In this thesis,aiming at the low accuracy of WiFi indoor positioning and the decrease in accuracy caused by environmental changes in indoor positioning,the corresponding improved algorithm is proposed,and a set of indoor positioning system based on machine learning is developed on this basis.The main work of the thesis is as follows:Aiming at the problem of low positioning accuracy of the existing algorithm,introduce a new integrated learning based on extreme gradient boosting(XGBoost)positioning algorithm,establish its positioning algorithm model,and through the experimental test,the performance of the algorithm is verified from the three aspects of the algorithm's efficiency,accuracy and stability.In order to solve the problem of positioning accuracy decrease caused by environmental changes,a positioning error compensation algorithm based on XGBoost is proposed.The algorithm only needs to collect a small number of RSS samples in the area of environmental changes,and calculates the positioning coordinates and errors by using the XGBoost positioning algorithm.Then use polynomial regression to take positioning coordinates as input and positioning error as output to build an error model.In online positioning,first use the XGBoost positioning model to calculate the rough estimated position,input the rough estimated position into the trained error model to obtain the positioning error,and compensate the rough estimated position according to the positioning error to obtain a more accurate positioning.In addition,actual scene experiments are designed.The experimental results show that the proposed positioning error compensation algorithm based on XGBoost can significantly improve the impact of positioning accuracy caused by environmental changes.At the same time,compared to updating all the location fingerprint database,the workload of RSS sample collection is greatly reduced.From the perspective of practical application,integrating the work done in indoor positioning technology,design and implement a WiFi indoor positioning system based on machine learning,which is a three-tier structure based on the C/S framework,including Android client,Tomcat server and MySQL database.It can realize the establishment of location fingerprint database,the positioning of different machine learning algorithms and the navigation of the best path.In addition,an experimental scene is set up to test the function of the system.The test results show the practicability of the positioning system,and the performance analysis of different algorithms can be completed effectively with the help of the system.