A bank of sequential unscented Kalman Filters for target tracking in range-only WSNs

Abstract

The paper is concerned with the target tracking in range-only wireless sensor networks (WSNs). To integrate the separated measurements from the WSN, a sequential fusion estimation method is presented in the sense of linear minimum mean squared error (LMMSE). Moreover, the un-scented transformation is used to implement the recursion of means and covariances, and this kind estimator is termed as sequential unscented Kalman filter (SUKF). A bank of SUKFs are employed to improve the estimation accuracy and stability as a result of that the orientation of the target is not observable. Accordingly, a set of estimates are generated by the filter bank and the estimates are pruned and updated at each estimation instant. Finally, by simulations of a target tracking example, it demonstrated that in contrast to the single SUKF a better estimation accuracy and convergence speed can be obtained by the SUKF bank.