Distributed Target Tracking Method Based on Variational Bayesian with Systematic Biases

Abstract

The unknown and time-varying systematic biases and measurement noise and abnormal non-Gaussian process noise will result in low target tracking accuracy. Additionally, traditional Kalman filtering based on Gaussian noise is difficult to meet the requirements of the estimation accuracy. In this paper, a variational Bayesian-based adaptive Kalman filter algorithm under distributed fusion framework (DF-VBAKF) is proposed. It first a priori model of probability density is constructed based on the properties of multiple unknown parameters. And then, through variational Bayesian and adaptive Kalman filter, the posterior distribution probabilities of target state, system biases, and noise covariance are jointly updated by using standard variational method. Finally, collecting local state estimation and its covariance from different platform, the control center fuses them according to the covariance intersection fusion strategy, and feeds results back to platform. The simulation results show that the proposed algorithm has higher estimation accuracy for systematic biases, noise and target state, leading to improvement of target tracking accuracy, too.