Multiple similarity measure-based maximum correntropy criterion Kalman filter with adaptive kernel width for GPS/INS integration navigation

Abstract

The filter based on the maximum correntropy criterion (MCC) has low sensitivity to non-Gaussian noise, which can effectively improve the robustness of the filter. However, the uncertainty of kernel width restricted the application of the maximum correntropy criterion Kalman filter (MCCKF) in the integrated navigation system. In order to overcome the above difficulty, this paper proposes a multiple similarity measure-based maximum correntropy criterion Kalman filter with adaptive kernel width (ABMCCKF). ABMCCKF uses the MORKF framework and distinguishes the noise polluted by outliers through a designed protecting window, which also ensures that the kernel width is compatible with the noise pollution level. Through the window, ABMCCKF utilizes the optimality of both MCC and the minimum mean square error criterion (MMSE). Simulations and vehicular experiments are used to verify the effectiveness of the method. The results show that the ABMCCKF have a good performance when the optimal kernel width is uncertain.