Message Passing-Based Distributed Track-Level Fusion Tracking for Space-Based Radar Networks With Time-Varying Topology
Zengfu Wang, Dengliang Qi, Haotian Meng, Hua Lan · IEEE Transactions on Aerospace and Electronic Systems · 2025
Distributed space-based radar (SBR) networking can significantly enhance target tracking performance, where four interdependent subproblems - detection of unknown number of targets, local estimation of target kinematic states, multi-target data association (DA), and distributed state fusion, are required to be addressed, and finding a joint solution is challenging. In response to these challenges, this article proposes a distributed track-level fusion multi-target tracking (MTT) method leveraging the combined belief propagation (BP) and mean-field (MF) approximation in a unified message-passing (MP) framework. This approach formulates the distributed fusion MTT problem as a probabilistic inference problem and employs a factor graph with a BP part and an MF part to characterize the correlations between high-dimensional hidden variables. The posterior probability distribution function (PDF) of the local target kinematic state is approximated using MF due to its conjugate exponential properties, while the posterior PDFs of the target visibility state, consensus-based distributed state fusion and DA are approximated using BP. To address coupling issues among these hidden variables, the marginal posterior PDFs of the hidden variables are iteratively refined in a feedback loop using MP. Considering the variations in the communication relationships between SBRs, the proposed distributed track-level fusion tracking method is adapted to the SBR networks with time-varying topology. Extensive simulations in a multiple high-speed maneuvering target tracking scenario demonstrate the superior performance of the proposed method.