Distributed multi-radar fusion of PHD filters with limited radar fields of view
Yue Li, Xinliang Chen, Quanhua Liu · IET conference proceedings. · 2026
In distributed multi-target tracking systems, the sensing fields-of-view (FoVs) of individual nodes are often limited, unknown or difficult to model due to occlusions, platform dynamics, or limited communication. To address this challenge, we propose a sequential partition and fusion algorithm based on the product partitioning criterion, enabling robust fusion of Probability Hypothesis Density Filters without requiring prior FoV knowledge. The algorithm sequentially partitions the Gaussian mixture components of each radar’s intensity function into a set of non-overlapping subsets based on the extent of the spatial density overlap. Each subset is considered to represent a potential target. A Weighted Arithmetic Average (WAA) rule is then applied within each subset, followed by reconstruction of the global intensity estimate. Simulation experiments under scenarios with occlusions and partially overlapping FoVs show that the proposed algorithm significantly outperforms traditional WAA and state-dependent WAA in both cardinality estimation and tracking accuracy, demonstrating strong adaptability and robustness in limited and unknown sensing environments.