Optimization of XGBoost indoor positioning method with HF and fluctuation parameter assistance
Liu Gaohui, Ling Fengzhi
School of Automation and Information Engineering, Xi′an University of Technology
Abstract： Aiming at the problem that the received signal strength measurement data in complex indoor environments contain noise which makes them show fluctuation and leads to low positioning accuracy, an optimized extreme gradient boosting (XGBoost) indoor positioning method based on hybrid filtering (HF) and fluctuation parameters assistance is proposed. Firstly, the HF method is used to optimize the data subset, reduce the influence of noise, and obtain the initial database; in addition, considering that the fluctuation can′t be completely eliminated, the fluctuation parameter that can reflect the degree of data change is introduced; secondly, for the performance of the XGBoost algorithm is susceptible to the influence of the initial parameter, the particle swarm (PSO) algorithm is used for optimization of the parameter, and the fluctuation parameter and optimized data are used as inputs to train the algorithm to generate the localization model; finally, the target point information is input into the model to complete the position estimation, and the point data is saved into the database to complete the update. The experimental results show that compared with the traditional algorithms, the algorithm in this paper has a good localization effect, and the localization accuracy is improved by 9.2%, 14.1%, and 18.45% in the range of 1 m, 2 m, and 3 m, respectively.
