Boosting the EKF for distributed estimation in binary WSN through unscented transformation and iterative processing

This paper deals with the problem of Distributed Estimation (DE) in binary Wireless Sensor Networks (WSN). Specifically, we propose to enhance the well-known Sign Of Innovation (SOI) based Extended Kalman Filter (SOI-EKF) [1] using the Unscented Transformation (UT) and iterative processing. The unscented transformation is especially used here to boost the quality of the predicted observation, already used in the SOI, and thus improve its pertinency; and most of all, improve the Kalman gain computation in the correction step. The so-developed filter, referenced here by the SOI Unscented Kalman Like Filter (SOI-UKLF) exhibit stable convergence behavior, compared to the SOI-EKF, when used in a target tracking application. Besides, an Iterated version of the SOI-EKF (I-SOI-EKF) is also proposed to enhance the tracking performance by stabilizing the filter output.