Deep Unfolded Gridless Sparse Recovery without Convex Relaxation for Bistatic Airborne Radar Clutter Suppression
Abstract
This paper proposes a clutter suppression method for bistatic airborne radar based on deep unfolding of a gridless sparse recovery algorithm. Conventional sparse recovery approaches typically rely on relaxing the rank norm to the nuclear norm to obtain tractable solutions, which inevitably introduces approximation errors and degrades reconstruction accuracy. In contrast, the proposed method avoids the performance loss caused by such relaxation, thereby achieving more accurate signal reconstruction. Furthermore, the iterative optimization procedure is unfolded into a deep neural network, where each iteration is mapped to a network layer. This deep unfolding strategy enables automatic learning of key parameters, eliminating the need for manual tuning of regularization and Lagrange parameters required in traditional optimization methods. As a result, the proposed approach combines the interpretability of model-based methods with the adaptability of data-driven learning.