Research of MIMU/GPS integrated navigation based on improved Kalman filter

Kalman filter algorithm is basically applied to the MIMU/GPS integrated navigation, but there are always accidents such as abnormal observation noise or noise mutation in reality, where may make the state covariance matrix lose symmetric even the emergence of the divergence of the filter with the consequences of leading to lower the accuracy of integrated navigation system. This paper aims to solve this problem using the principle of singular value decomposition. The main concept behind the algorithm is to achieve updating process via singular value decomposition factors of all the matrices consisting of error covariance, process noise covariance and measurement covariance, instead of Cholesky decomposition on the basis of the conventional Kalman filtering algorithm. The simulation experiments show that the robustness of the proposed algorithm based on singular value decomposition is superior to the Kalman filtering and the square-root filtering algorithm, and improves the navigation accuracy effectively.