1: \begin{abstract}

2: %Compressive sensing (CS) is known to greatly improve the performance of multiple-input-multiple-output (MIMO) radar for sparse target scenes. In colocated MIMO radar we can resolve targets by direction-of-arrival (DoA) estimation. For a given number of antennas the resolution can be distinctively increased compared to Nyquist-sampling by employing CS. While the antenna spacing is trivial for sampling at Nyquist-rate, it becomes a hard problem for CS. In this work our goal is to place antennas such that the \textit{coherence} of the resulting measurement matrix is minimized. We look at existing work that aims to select optimal antenna positions from a predefined grid, with respect to \textit{coherence} minimization. While this problem is combinatorial in nature, the authors relax it to be convex. This is achieved by optimizing probability distributions of antenna placements with respect to the expected \textit{coherence}, instead of looking for optimal placements directly. Further an iterative approach is used to optimize transmit and receive array alternately until convergence. We adopt this idea and modify the approach in order to improve the performance. We aim to simultaneously allow for convex relaxation of the problem while still optimizing for a concrete antenna placement rather than a probability distribution. We do so by gradually removing antenna slots from the predefined grid as the algorithm iterates until only grid slots for the desired numbers of transmit and receive antennas are left. Finally we conduct a numerical performance analysis to demonstrate that our modification significantly improves the performance of the algorithm while preserving the low complexity.

3: %\end{abstract}

4: