Target localization with bistatic MIMO radar based on vandermonde constrained canonical polyadic decomposition

Abstract

We consider target localization based on a bistatic MIMO radar, in which the transmit array and/or the receive array is a uniformly spaced array, i.e., the uniform rectangular array or uniform linear array. By exploiting both the multilinear structure of the dataset and the uniform structure of the array, we formulate the output data into a tensor that admits a Vandermonde constrained canonical polyadic decomposition (VC-CPD) model. We propose a target localization method under the above VC-CPD modeling, consisting of a novel joint eigenvalue decomposition (J-EVD) based algebraic algorithm, a lift-and-project based iterative algorithm to refine the results of J-EVD based VC-CPD, and a post-processing technique for extracting direction of arrival (DOA) and direction of departure (DOD) from the VC-CPD results. Specifically, we exploit the rotational invariance of the Vandermonde factor matrix, link it to multidimensional harmonic retrieval and multiple invariance ESPRIT problems, and solve it using the J-EVD algorithm. Experiments are conducted to demonstrate the effectiveness of the proposed method.