A Sequential Partial Relaxation-Based Technique for Automotive MIMO Radar Imaging

We consider the joint Direction-of-Arrival (DOA), Time-of-Arrival, and Doppler-frequency estimation problem in Multiple-Input-Multiple Output automotive radar systems. To enhance the angular resolution capability, the virtual array concept is employed. For solving the estimation problem we present a multidimensional extension of the recently proposed Partially Relaxed Orthogonal Least Squares Weighted Subspace Fitting (PR-OLS-WSF) algorithm. The PR-OLS-WSF algorithm belongs to the class of computationally efficient greedy algorithms where each source is resolved sequentially. Unlike other greedy algorithms such as the popular Matching Pursuit (MP) and Orthogonal Matching Pursuit (OMP) whose performances are known to severely degrade in both threshold and asymptotic domains, the proposed PR-OLS- WSF algorithm inherits excellent resolution performance from the partial relaxation step involving in the optimization procedure. We show that, based on real measurement data, the proposed algorithm can resolve more targets than the Matching Pursuit algorithm.