Maximum correntropy unscented Kalman and information filters for non-Gaussian measurement noise

Abstract In this paper, we investigate the state estimation problem of nonlinear systems with non-Gaussian measurement noise. Based on a newly defined cost function which is obtained by a combination of weighted least square (WLS) and maximum correntropy criterion (MCC), we derive our maximum correntropy unscented Kalman filter (MCUKF) and the corresponding maximum correntropy unscented information filter (MCUIF). Comparing with existing MCUKF, our MCUKF avoids the numerical problem occurred when the measurements contain large outliers, and can obtain similar or even better estimation results. When the kernel bandwidth goes infinity, we prove that our MCUKF and MCUIF will converge to UKF and UIF, respectively, while existing MCUIF will not in this case and it generally has poor estimation accuracy as well. Two typical nonlinear models are used to illustrate the advantages of our proposed algorithms.