Experimental Evaluation of Compressive Sensing for DoA Estimation in Automotive Radar
Abstract
Radar sensors are one of the key elements in highly automated driving systems. Their performance is robust independently of weather and visibility conditions, being able to detect targets up to hundreds of meters distance. In vehicles, radar sensors are required to estimate the range, doppler shift and direction of arrival of the reflected wave. In this contribution we will focus on the direction of arrival (DoA) estimation, which, while being crucial for environmental perception, still has considerable room for improvement. In particular, we aim to measure the performance of Compressive Sensing (CS) based algorithms applied for reconstruction of the DoA. In contrast to the traditional FFT approach, these algorithms are able to exploit sparse antenna configurations with a reduced number of Tx and Rx channels and a large effective aperture. For this purpose, a series of relevant metrics are defined and applied to measurements in representative open-air driving scenarios acquired with an automotive 4x8 MIMO radar operating at 77GHz. The results show that, when compared with the FFT, these algorithms display an overall enhanced angular estimation accuracy, resolution and false alarm ratio.
