Target Localization Based on Multi-site Sparse Arrays: A Coupled Canonical Polyadic Decomposition Approach

Sparse arrays have attracted increased attention in the past decades, due to its improved performance over regular arrays in terms of accuracy and identifiability. On the other hand, multi-site arrays have been considered in target localization applications, such as multistatic MIMO radars. In this paper, by combining the above two concepts, we consider the problem of target localization via multi-site sparse arrays and propose a coupled tensor based framework for the solution of this problem. More precisely, a coupled canonical polyadic decomposition (C-CPD) model is established to formulate the array signal and a joint eigenvalue decomposition (J-EVD) based algorithm is developed to compute the C-CPD. Then, a polynomial rooting based technique is used to find the DOA estimates of each target observed by different sites, the fusion of which finally yields the localization of the targets. We discuss the uniqueness conditions and provide numerical experiments to illustrate the improved performance of the proposed approach.