DOA Estimation for Transmit Beamspace MIMO Radar via Tensor Decomposition With Vandermonde Factor Matrix

We address the problem of tensor decomposition in application to\ndirection-of-arrival (DOA) estimation for transmit beamspace (TB)\nmultiple-input multiple-output (MIMO) radar. A general 4-order tensor model\nthat enables computationally efficient DOA estimation is designed. Whereas\nother tensor decomposition-based methods treat all factor matrices as\narbitrary, the essence of the proposed DOA estimation method is to fully\nexploit the Vandermonde structure of the factor matrices to take advantage of\nthe shift-invariance between and within different subarrays. Specifically, the\nreceived signal of TB MIMO radar is expressed as a 4-order tensor. Depending on\nthe target Doppler shifts, the constructed tensor is reshaped into two distinct\n3-order tensors. A computationally efficient tensor decomposition method is\nproposed to decompose the Vandermonde factor matrices. The generators of the\nVandermonde factor matrices are computed to estimate the phase rotations\nbetween subarrays, which can be utilized as a look-up table for finding target\nDOA. It is further shown that our proposed method can be used in a more general\nscenario where the subarray structures can be arbitrary but identical. The\nproposed DOA estimation method requires no prior information about the tensor\nrank and is guaranteed to achieve precise decomposition result. Simulation\nresults illustrate the performance improvement of the proposed DOA estimation\nmethod as compared to conventional DOA estimation techniques for TB MIMO Radar.\n