Collaborative design for moving target detection in distributed radar networks

Distributed radar networks (DRNs) can improve target detection by fusing the spatially diverse echoes from multiple receiving stations. However, this potential is limited by the communication burden and the spatial correlation of echoes. To address this challenge, this paper proposes a flexible hierarchical collaborative scheme for moving target detection in DRNs. The proposed scheme supports operating in coherent, noncoherent, and hybrid modes, providing a configurable trade-off between detection performance and communication demand. In particular, the hybrid mode provides an intermediate collaboration strategy between coherent and noncoherent modes. The collaboration design problem is generally non-convex due to the orthogonality constraint. Therefore, a stacked-product Riemannian conjugate gradient algorithm is developed to configure the collaboration matrix, allowing efficient fusion of radar echoes. Additionally, a multi-constraint k-medoids clustering-based topology design is proposed to organize radar nodes into effective cooperative groups based on echo correlation and communication cost. Analytical detection thresholds are derived for each detection mode to ensure a prescribed false-alarm rate under correlated observations. Simulation results demonstrate the effectiveness of all three detection modes, showing up to 55% lower communication overhead compared to the non-collaborative fusion.