Distributed Maximum Correntropy Estimation Under Energy Harvesting Constraints Over Sensor Networks

This paper studies the distributed maximum correntropy estimation issue for nonlinear time-varying systems over energy-harvesting-constrained sensor networks in non-Gaussian noises. The modeled communication scenario is that sensors equipped with energy harvesters select some neighbors for data transmission based on their energy level and the neighbors' priorities. The expectations of a sensor transmitting data to its neighbors can be obtained by recursively computing its energy probability distribution. A cost function based on the maximum correntropy criterion (MCC) rather than the conventional minimum covariance is the optimization index to improve the estimation effect in non-Gaussian environments. The optimal estimator gain and an upper bound of the estimation error covariance are calculated using the MCC and a fixed-point iteration scheme. A sufficient condition is derived to guarantee the convergence of the fixed-point algorithm. The proposed new energy-based maximum correntropy estimator utilizes only local information and information from neighbors, thereby enabling a distributed framework. Finally, a numerical example demonstrates the effectiveness of the estimation design.

I. Introduction

State estimation has always attracted considerable interest due to its applications in smart grids, target tracking, health detection, and other fields. In recent decades, various results have emerged to achieve estimation purposes [1], [2], [3], [4]. The well-known Kalman filter applies the minimum mean square error (MMSE) as the optimization criterion, which is used to process linear Gaussian systems. So far, distributed estimation approaches over sensor networks have become hot topics with the rapid rise of wireless communication technology, see [5], [6], [7], [8], [9], [10], [11]. Sensors over a sensor network perform local estimation by sharing information with their neighbors.