Diffusion Kalman filter by using maximum correntropy criterion

This paper is concerned with the distributed estimation problem by using diffusion Kalman filters. Instead of the traditional minimum mean square error (MMSE) criterion, the maximum correntropy criterion (MCC) is used to define a new cost function in the diffusion update step and a fixed-point iteration algorithm is adopted to fuse the intermediate estimates of neighboring nodes. It is shown that the proposed filter is a modified version of the traditional diffusion Kalman filter with adaptive weights and the weights are tuned by using fixed-point iteration. An iterated diffusion Kalman filter is further developed to improve estimation performance. Simulation results show that the proposed approach outperforms existing techniques and the proposed filters can efficiently handle non-Gaussian noise and large variations of measurement noise covariances.