Abstract
We propose a robust method for tracking nonlinear target with the fusion unscented Kalman filter (FUKF). We noticed that when some outliers exist in the measurements of the sensors, they cannot track the target accurately by using the standard Kalman filters. The robust statistics theory is used in this paper to solve this problem. The measurement noise variance which is at the time of the outlier is restructured through minimizing the designed cost function. Then, the standard fusion unscented Kalman filter is used to track the target in order to avoid the bias brought by the linear approximation. Compared to the traditional tracking method and Huber robust method (HFUKF), this method has a more accurate performance and can track the target efficiently while the outliers exist. Last, simulation examples in three different conditions are given and the simulation results show the advantages of the proposed method over the fusion unscented Kalman filter (FUKF) and the Huber robust method (HFUKF).
Highlights
Multisensor networks have received increasing attentions in recent years, due to the huge potential in applications, such as communication, signal process, routing and tracking, and many other areas
We propose a robust method for tracking nonlinear target with the fusion unscented Kalman filter (FUKF)
It is well known that the standard Kalman filter (KF) is an available method to estimate the state parameters of the linear system given by the equation set composed of the dynamic model and the measurement model
Summary
Multisensor networks have received increasing attentions in recent years, due to the huge potential in applications, such as communication, signal process, routing and tracking, and many other areas. Many methods have been presented to extend the KF to the nonlinear dynamic and measurement models by forming a Gaussian approximation to the posterior state distribution such as the extended Kalman filter (EKF) [1] and the unscented Kalman filter (UKF) [2]. The Huber technique has long been used in dynamic filtering problem, including underwater vehicle tracking It is a combined minimum l1 and l2-norm estimation technique and it exhibits robustness with respect to deviations from the assumed Gaussian distribution. The document [7] has proposed an algorithm called Huber-based unscented filtering (HUF), which applies Huber technique into unscented Kalman filter This algorithm approximates the nonlinear measurement equation as linear equation, which will result in a loss of accuracy. The proposed method can perform better than both standard unscented Kalman filter and the linearized Huber-based filter.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
