Adaptive Front-end for MIMO Radar with Dynamic Matrix Completion

Abstract

This work proposes a dynamic matrix completion (DMC)-based approach for use in the front-end of MIMO radar. The proposed approach is different and complementary to the conventional target tracking algorithms that are widely deployed in the back-end of radar systems. The received signals are modelled as time-varying low-rank matrices and passed through an adaptive singular value thresholding (ASVT) block, resulting in the elimination of noise returns early in the processing chain. When all the antenna elements are not being used and the received signal is only partially observed, the ASVT block imputes the missing entries. Front-end processing results in cleaner signals for the back-end, culminating in fewer range and Doppler bins, increased probability of detection, reduced false alarm rate, and ultimately, improved target tracking performance. Detailed simulation of the radar chain reveal the significant improvements afforded by the proposed algorithm.