Digital twin creation and residual analysis using CLEAN for radar target emulation

Abstract Radar sensors in advanced driver assistance systems sense the traffic scene identifying reflections occurring in the environment. Clutter contained in the data is caused by multipath and scattering effects leading to signal processing challenges to identify weak targets, which is critical for vulnerable road users. To be able to test a radar for its robustness against clutter, suitable clutter models are required. In this paper, an approach for filtering reflections following the CLEAN algorithm is proposed. By iterative decomposition of processed radar data to a superposition of point sources, reflections are isolated for a replay with a radar target simulator (RTS) to form a digital twin. Field datasets recorded in a parking garage are analyzed for its radar target and clutter statistics. After the analysis, a list of possible point targets for replay with an RTS is available and residual radar cross-section levels are analyzed and compared to Swerling I case.