Maximum Iterations Criterion for OMP in FMCW RADARs

RADAR sensors are a fundamental component of automotive systems for remote sensing and for mapping the surrounding environment. Compared to LIDAR and camera, radar systems operate at a much larger wavelength, and given the limited aperture size, angular resolution is much lower. An often-made compromise between system cost and resolution is obtained by using sparse antenna arrays, such that the aperture is relatively large, but the number of receivers is kept small. The problem of using sparse arrays is the substantial number of spurious lobes that appear due to non-uniform spatial sampling. Sparse sensing algorithms such as Orthogonal Matching Pursuit (OMP) or Matching Pursuit (MP) may be used to iteratively estimate and eliminate the target signatures from the received signal, along with the spurious lobes associated to that target, hence obtaining a clean radar image. A challenge of using greedy algorithms is stopping after the correct number of detections, especially when the dynamic range among targets is large. In this paper we propose a robust stop criterion for greedy/ compressive sensing algorithms.