Analysis of a Machine Learning Based Virtual Array Augmentation Technique for Automotive Radar
Maximilian Eschbaumer, Simon Achatz, Gábor Balázs · 2022 19th European Radar Conference (EuRAD) · 2022
The goal of array augmentation in automotive radar is enhancing the direction of arrival estimation performance. The rising demand of low-cost high-resolution radar systems calls for a lightweight extra- and interpolation method. We propose a lightweight neural network that is trained to augment the antenna array of the radar. That is, either extrapolate uniform or both, inter- and extrapolate non-uniform linear arrays. We evaluate the neural network on simulated, and measured data and achieved higher accuracy, and target separability compared to conventional augmentation techniques from literature. With this method, the automotive radar can achieve either higher performance or lower costs.