Moving sensor nodes' tracking by WT-UKF in Wireless Sensor Networks

Abstract

Tracking moving sensor node is a key point in Wireless Sensor Networks (WSN). However, it's difficult to achieve high tracking accuracy quickly because many problems should be solved. After analyzing those fundamental problems, Unscented Kalman Filter (UKF) is established to estimate the moving sensor node's localization and environment argument simultaneously. To adapt for the case of colored measurement noise, the UKF model is improved to form a time-lag UKF model with Gauss Noise based on the first-order Auto Regressive (AR) model of the colored noise. To decrease the influence of dynamic variance of measurement noise in the time-lag UKF model, Wavelet Transformation (WT) is used to estimate the standard deviation of the measurement noise on-line fast, and the new model is named WT-UKF. In simulations, results show that WT-UKF is always better than UKF when the number of the measurement values analyzed by WT in one interval is set properly.