Variational Bayesian Cubature Kalman Filter for Bearing-Only Tracking in Glint Noise Environment

The problem of state estimation of nonlinear systems in glint noise environment is considered that will deteriorate the tracking performance. In this paper, a cubature kalman filter algorithm based on variational Bayesian learning is proposed. The t-logical-scale distribution is used to model the glint noise and the Gamma distribution is as the prior distribution of the three parameters of the t-logical-scale distribution. The cubature kalman filter and the variational Bayesian learning are combined to estimate the state of system and parameters in the recursive procedure. The simulation results demonstrate that the proposed algorithm has a stronger robustness and a lower computation.