Adaptive Range-Based Nonlinear Filters for Wireless Indoor Positioning System Using Dynamic Gaussian Model

It is hard to obtain a general error model for the range-based wireless indoor positioning system due to the complicated hybrid line-of-sight/non-line-of-sight (LOS/NLOS) environment. The performance of the conventional Gaussian-based nonlinear filters is degraded in the indoor scenario. In this paper, we employ a dynamic Gaussian model (DGM) to describe the indoor ranging error. A general Gaussian approximated model is constructed first to fit the potential distribution. The instantaneous LOS or NLOS error at a typical time is considered as the drift from this general distribution dynamically. The relationship between the instantaneous error of the DGM and the estimation accuracy of nonlinear filters is analyzed. Based on our analysis, we propose a measurement adaptation method to further reduce the error according to the DGM. Then, the nonlinear filters based on the Gaussian model, which are simple and accurate, can be applied. A biased extended Kalman filter (EKF) and an adaptive Gaussian particle filter (PF) integrated with the measurement adaptation method are designed. The real indoor experiment demonstrates that the estimation accuracy of our algorithms is greatly improved without imposing complexity and that our algorithms are suitable for the dynamic indoor environment.