Target Tracking Algorithm Based on Variational Bayes

To address the problem of low tracking accuracy caused by unknown excitation and measurement noise in radar systems, a tracking algorithm based on variational Bayes is proposed. This algorithm leverages variational Bayes theory to treat the unknown excitation and measurement noise as hidden variables, enabling the computation of the likelihood probability of the target motion state. By decoupling these hidden variables under the expected maximum criterion and field theory, an approximate estimation of the hidden variables is obtained. These estimates are then combined with the unscented Kalman filter to achieve target state estimation. Through iterative updates of the target state, excitation, and measurement noise parameters, high-precision target tracking is achieved. Additionally, an evaluation method to terminate the iterative process is proposed. Simulation results demonstrate that the variational Bayes-based target tracking algorithm can effectively suppress tracking divergence issues caused by unknown parameters, leading to stable tracking performance and significant improvements in target state estimation accuracy. The algorithm consistently achieves stable target tracking across various target motion models, measurement errors, noise errors, and iteration termination values. These results indicate that the proposed method holds promising potential for practical applications.