Variational Bayesian Bearing-Only Passive Target Localization

This paper proposes a Variational Bayesian-based Extended Kalman Filter (VB-EKF) to address divergence in traditional nonlinear filters caused by unknown measurement noise statistics during bearing-only passive target localization. The algorithm constructs a joint probabilistic model of the state and measurement noise varian-ce. It employs Variational Bayesian inference to estimate the posterior distribution of the variance, embedding it into the EKF framework for adaptive filtering. Simulated datas based on a vector array verify the feasibility of the proposed method. VB-EKF achieves lower localization errors across various noise levels, demonstrates enhanced convergence speed and stability, exhibits robustness to initial deviations, and enables reliable localization of targetwith diverse motion patterns. Results confirm VB-EKF effectively resolves the estimation challenge of unknown noise statistics, maintaining high accuracy and stability even under significant noise.