Improved maximum correntropy cubature Kalman and information filters with application to target tracking under non-Gaussian noise

The cubature Kalman filter (CKF) based on the maximum correntropy criterion (MCC) has been widely used in the target tracking. However, numerical problems usually occur when there are outliers in the measurement noise. In order to solve the problems of state estimation under the non-Gaussian measurement noise, a new combined cost function is defined based on the weighted least squares (WLS) method and MCC. In addition, a new method is also used to adaptively adjust the kernel size, then the improved maximum correntropy CKF (PMCCKF) and its corresponding improved maximum correntropy cubature information filter (PMCCIF) are proposed. Compared with the existing algorithms, the new method can not only obtain similar or even better estimation performance, but also avoid numerical problems. Moreover, when the kernel size is infinite, the performance of proposed algorithms will reduce to the standard CKF and corresponding cubature information filter (CIF) respectively, but the classical maximum correntropy CIF (MCCIF) will not, and even the performance is poor in this case. The advantages of the proposed algorithms are verified by four classical nonlinear models.