A robust indoor localization algorithm based on polynomial fitting and Gaussian mixed model

Abstract

Wireless sensor network (WSN) positioning has a good effect on indoor positioning, so it has received extensive attention in the field of positioning. Non-line-of sight (NLOS) is a primary challenge in indoor complex environment. In this paper, a robust localization algorithm based on Gaussian mixture model and fitting polynomial is proposed to solve the problem of NLOS error. Firstly, fitting polynomials are used to predict the measured values. The residuals of predicted and measured values are clustered by Gaussian mixture model (GMM). The LOS probability and NLOS probability are calculated according to the clustering centers. The measured values are filtered by Kalman filter (KF), variable parameter unscented Kalman filter (VPUKF) and variable parameter particle filter (VPPF) in turn. The distance value processed by KF and VPUKF and the distance value processed by KF, VPUKF and VPPF are combined according to probability. Finally, the maximum likelihood method is used to calculate the position coordinate estimation. Through simulation comparison, the proposed algorithm has better positioning accuracy than several comparison algorithms in this paper. And it shows strong robustness in strong NLOS environment.