Sparse MIMO radar via structured matrix completion

We explore sub-Nyquist sampling strategies in a bistatic MIMO radar with M transmit and N receive antennas to reconstruct the sparse scatter scene with K ≪ MN targets. We develop a front-end with a matched filter bank at each receive antenna and sample the branch output at random with a total of L samples per pulse. Sparse recovery is then obtained via enhanced matrix completion techniques that make no grid assumptions over the target scene. We demonstrate that as long as L is on the order of O(K log²(MN)), it is possible to recover the target scene under a mild condition with high probability, thus greatly reducing the sampling complexity from the Nyquist rate MN samples per pulse. The performance is numerically examined with comparison against compressive sensing approaches. The framework can also be explored to reduce the size of filter banks at the front-end.