A Variational Measurement Update for Extended Target Tracking With Random Matrix Under Nonstationary Heavy-Tailed Noise

Existing random matrix-based models rely on the assumption of a Gaussian distribution. However, in real-world, the unreliability of sensors and environmental complexities induce non-stationary heavy-tailed noise, rendering the Gaussian distribution assumption inherent in aforementioned random matrix-based models no longer applicable to these scenarios. To address the challenge of such noise, this paper introduces a new hierarchical linear Gaussian state-space model. Building on this, a Variational Bayesian (VB) method is derived for iteratively estimating the kinematic state and the extension of the extended target. Considering that target maneuvering constitutes one of the contributing factors to non-stationary heavy-tailed noise, an algorithm is proposed by integrating the Information Theoretic Interacting Multiple Model (IT-IMM) framework with the VB method. The effectiveness and robustness of the proposed algorithms were assessed through both simulations and the real experimental dataset. These results demonstrate that the proposed algorithm outperforms existing solutions relying on random matrix based-models in handling Non-Stationary Heavy-Tailed Noise (NSHTN).