Robust and Efficient Algorithms for Millimeter-Wave Radar Localization and Imaging

Abstract

Millimeter wave radars have been widely used in military reconnaissance and remote sensing, since they can acquire data in all-weather and all-day conditions. However, they still face several challenges, such as array shape limitations, high computation complexity and array manifold errors. In this thesis, we propose three new algorithms to address these challenges and achieve better performance.

First, we consider the problem of localizing multiple targets with a trapezoid virtual antenna array. The goal is to estimate both the number and the 3-D locations of the targets. The proposed algorithm consists of two steps: 1) estimating the number of targets and their ranges by extending Barone’s method to handle data from multiple antennas and 2) estimating the angle of arrival of each target by a Least-Square algorithm.

Second, we propose an efficient imaging method based on robust sparse array synthesis. It first performs range-dimension matched filtering, followed by azimuth-dimension matched filtering using a selected sparse aperture and filtering weights. The aperture and weights are computed offline in advance to ensure robustness to array manifold errors caused by the imperfect radar rotation. We introduce robust constraints on the mainlobe and sidelobe levels of the filter design. The resulting robust SAS problem is a nonconvex optimization problem with quadratic constraints. We devise an algorithm based on feasible point pursuit and successive convex approximation to solve the optimization problem.

Third, due to the unforeseen disturbance and imperfect measurement, radars may not be at their ideal locations for Synthetic Aperture Radar (SAR) imaging. We consider two error models of radar movement in Rotating SAR (ROSAR) systems. To overcome the blurring induced by location deviations of virtual phase centers, we employ an autofocusing algorithm, named Minimum Entropy Algorithm, to improve the image sharpness. The corresponding optimization problem is solved by gradient descent and interior-point methods.

To validate the effectiveness of the proposed algorithms in practice, we built a real-world radar localization system and a robotic ROSAR system. Experimental results show that the systems can localize the targets with higher accuracy and generate sharper SAR images compared with a 2D-FFT based algorithm and the Back Projection Algorithm, respectively.