A UKF algorithm based on the singular value decomposition of state covariance

One of the existing problems of General UKF is the stability problem due to the strong nonlinearity and the complexity of the system. When the algorithm can´t ensure the state covariance to be positive semidefinite, the precision will be decreased or even general UKF will be halted. Based on the singular value decomposition of state covariance, this paper proposes a modified UKF algorithm which can enhance the state covariance to be positive semidefinite. Comparing to the general UKF, the SVDUKF algorithm releases the condition of the positive semi-definite of state covariance. Numeric simulation experiments demonstrate the effectiveness of the algorithm.