Research on the Target Tracking Algorithm for Wireless Sensor Network Based on Improved Particle Filter
Liang Peng -,Xinhan Huang -,Li He -
DOI: https://doi.org/10.4156/jcit.vol7.issue11.2
2012-01-01
Journal of Convergence Information Technology
Abstract:Wireless Sensor Network (WSN) is a new technology integrated with communications, embedded technology, and network functions; having capability of real-time collecting, communicating, and sensing, for which dynamic target locating and tracking is its critical task, as well as a basic element in its commercial and military application such as battlefield information accessing and traffic monitoring. Sensor target tracking is the technology of target location estimate based on measurements from corresponding sensors. An inaccurate estimated location may very easily lead to target tracking failure which reduces the reliability and stability of the target tracking [1]. Experts and scholars have done a lot of researches to solve problems exited in wireless sensor target tracking, and many new tracking algorithms have been proposed, among which tracking algorithms such as target contrail-fitting, linear prediction, Kalman filter, unscented Kalman filter(UKF) and particle filter[2] are most common applied. Sikdar, with his partners, has proposed target location tracking algorithm based on linear sensor which is proved better than conventional contrail fitting algorithm in its effects. Zhang Xiaoping , with his partners, has proposed WSN target tracking algorithm through modeling for quadratic polynomial motions which has improved target tracking accuracy[3]. However, there is a problem in common: all the algorithms above assume the tracking target contrail-fittings being in linear motion, while the actual contrail fitting is of uncertainty and variability, and the estimated accuracy can hardly be improved after being to a certain level. In this condition, the tracking error of the above algorithms can be greater [4]. Kalman filter algorithm can be applied for weakly nonlinear models, but for strongly nonlinear models, Unscented Kalman Filter algorithm can be better. Kalman filter work well in cases based on model-linear and Gaussian noise, thus its estimated accuracy is not that ideal [5, 6]. Particle filter algorithm is based on optimal regression Bayesian filtering algorithm in Monte Carlo simulation, which can accurately estimate the any distribution state with only a small amount of mobile anchor nodes. PF algorithm has distinguished advantages in processing non-linear, non-Gaussian cases [7]. However, in PF application process after several iterations, diversified degradation of the particles will occur, which is likely to lead a target tracking loss, thus further affect WSN target tracking accuracy [8]. In order to solve the diversified degradation of the particles in Particle Filter algorithm, this paper Research on the Target Tracking Algorithm for Wireless Sensor Network Based on Improved Particle Filter Liang Peng,Xinhan Huang, Li He