Classification and Synthetic Data Generation for Automotive Radar

Automotive radar sensors are essential for Advanced Driver Assistance Systems (ADAS) since they enable the perception and thus the understanding of the vehicle’s surrounding. As an automotive radar supplier, it is crucial to ensure the required performance in every signal processing step of the radar system in every imaginably driving scenario. Therefore, a reference platform was designed to infer ground truth information from real world driving scenarios in which the radar system is being tested. The benefit of this reference system is twofold. One, ground truth is generated to label radar targets in range-Doppler map. Second, the ground truth can be used to simulate artificial radar data matching the exact driving scenario under test. We first introduce the basic signal processing chain for automotive radar. Secondly, the reference platform is explained. The above-mentioned benefits will be demonstrated in detail on real data.