Distributed Tracking of Extended Target With Orientation Using Variational Bayesian

ABSTRACT In this work, we propose an alternative distributed tracking approach for extended target with time‐varying orientation in a sensor network. Within the random matrix framework, we employ a Gaussian prior for the orientation, the inverse Gamma priors for the diagonal elements of the extent matrix, and a Gamma prior for the measurement rate. Using the Gamma Gaussian Inverse Gamma Gaussian (GGIGG) state model, we derive a centralised tracking approach based on the variational Bayesian technique. Subsequently, we introduce a distributed variational measurement update that leverages convex combination fusion. Closed‐form expressions for the unknown variables are derived under a consensus scheme. The resulting algorithm efficiently computes approximate posterior densities for the kinematic state, extent, orientation, and measurement rate in a distributed manner. The effectiveness of the proposed distributed tracking method is validated through numerical experiments, with results showing that the proposed algorithm outperforms existing method based on the multiplicative error model.