A Robust Variational Bayesian Student-T CKF Algorithm for Hypersonic Vehicle Tracking

A robust variational Bayesian Student-t Cubature Kalman filter (RVBSTCKF) algorithm is proposed in this paper to address the difficulties arising from the tracking for the complex maneuvering forms of hypersonic vehicle under the non-stationary heavy-tailed measurement noise with the unknown statistics. Firstly, the Student-t (ST) distribution is introduced to approximate the measurement noise of hypersonic vehicle which is caused by non-stationary outliers in complex flight environment. In addition, the inverse Wishart distribution and the Gamma distribution are introduced to improve the ST distribution, which are implemented to approximate the uncertain covariance and non-Gaussian degree, respectively. Moreover, on the basis of the improved ST distribution, the explicit expressions of posterior update can be derived analytically in nonlinear system by the introduction of Cubature Kalman filters (CKF) and variational Bayesian (VB) algorithm. The simulation results demonstrate that the proposed algorithm has the better estimation accuracy on the tracking of hypersonic vehicle.