An improved multi-state constraint kalman filter based on maximum correntropy criterion

In recent years, the multi-state constraint Kalman filter has been widely used in the visual-inertial navigation of unmanned systems. However, in most previous studies, the measurement noise of the navigation system was assumed to be Gaussian noise, but this is not the case in practice. In this paper, the maximum correntropy criterion is introduced into the multi-state constraint Kalman filter to improve the robustness of the visual-inertial system. First, the new maximum correntropy criterion-based Kalman filter is introduced, it uses the maximum correntropy criterion to replace the minimum mean square error criterion to suppress the interference of measurement outliers on the filtering results, and it has no numerical problem in the presence of large measurements outliers. Then, an improved multi-state constraint Kalman filter is designed by applying the new maximum correntropy criterion-based Kalman filter to the multi-state constraint Kalman filter, which improved the robustness of the multi-state constraint Kalman filter. The results of numerical simulation and dataset experiments show that the proposed filter improves the accuracy and robustness of the visual-inertial system.