As the progress of electronics and information processing technology continues, indoor localization has become a research hotspot in wireless sensor networks (WSN). The adverse non-line of sight (NLOS) propagation usually causes large measurement errors in complex indoor environments. It could decrease the localization accuracy seriously. A traditional grey model considers the motion characteristics but does not take the NLOS propagation into account. A robust interacting multiple model (R-IMM) could effectively mitigate NLOS errors but the clipping point is hard to choose. In order to easily cope with NLOS errors, we present a novel filter framework: mixture Gaussian fitting-based grey Kalman filter structure (MGF-GKFS). Firstly, grey Kalman filter (GKF) is proposed to pre-process the measured distance, which can mitigate the process noise and alleviate NLOS errors. Secondly, we calculate the residual which is the difference between the filtered distance of GKF and the measured distance. Thirdly, a soft decision method based on mixture Gaussian fitting (MGF) is proposed to identify the propagation condition through residual value and give the degree of membership. Fourthly, weak NLOS noise is further processed by unscented Kalman filter (UKF). The filtered results of GKF and UKF are weighted using the degree of membership. Finally, a maximum likelihood (ML) algorithm is applied to get the coordinate of the target. MGF-GKFS is not supported by any of the priori knowledge. Full-scale simulations and an experiment are conducted to compare the localization accuracy and robustness with the state-of-the-art algorithms, including robust interacting multiple model (R-IMM), unscented Kalman filter (UKF) and interacting multiple model (IMM). The results show that MGF-GKFS could achieve significant improvement compared to R-IMM, UKF and IMM algorithms.
A Mobile Localization Method in Smart Indoor Environment Using Polynomial Fitting for Wireless Sensor Network
