Unbiased minimum-variance simultaneous input and state estimation based on maximum correntropy criterion

The existing unbiased minimum-variance (UMV) filter with simultaneous input and state estimation (SISE) is based on the classical minimum mean square error (MMSE) criterion of the optimal Gaussian system. Nevertheless, when the system is in non-Gaussian noise measurement system, the performance of existing SISE algorithms will deteriorate significantly, and the maximum correntropy criterion (MCC) can solve this problem commendably. In this paper, maximum correntropy filter based unbiased minimum-variance simultaneous input and state estimation method (MCF-UMVSISE) is proposed, SISE filter is transformed into standard form firstly, in this case the input estimation is not required when deriving this estimator. Then the maximum correntropy criterion is used for derivation, which will make the performance better under non-Gaussian noise. In the end, experiments and simulations are carried out under Gaussian and non-Gaussian observation noises respectively to prove the effectiveness of the algorithm.