A Methodology for Verifying and Validating the Functional Performance Evaluation of High-Fidelity RADAR Sensor Models at Raw Data and Detection Levels

Abstract

This paper outlines a comprehensive methodology for verifying and validating the functional performance evaluation of high-fidelity radio detection and ranging (RADAR) sensor models at both the raw data level, specifically the range map (RM) and range-Doppler map (RDM), and the detection level. Three key performance parameters (KPPs) are defined at the raw data level: the target's received power, signal-to-noise ratio (SNR), and twoway beam pattern. Four KPPs are identified at the detection level: distance, relative radial velocity, azimuth angle, and elevation angle. This study also introduces three dynamic test scenarios directly applicable to the functional evaluation assessment of RADAR sensor models. Furthermore, a toolchain is presented to integrate real-world test scenarios into the virtual environment, allowing for frame-by-frame validation of the RADAR sensor models' functional evaluation. Quantitative analysis successfully assesses differences between simulations and real measurements, employing the mean absolute percentage error (MAPE) metric. This methodology has been successfully applied to a high-fidelity RADAR sensor model featuring a multiple input and multiple output (MIMO) 2D linear spacing virtual antenna and a complete signal processing toolchain that mimics real RADAR sensors. Results show a strong correlation between the simulation and real measurements, with MAPE values below 7 % for all defined KPPs at raw data and detection levels. Using this methodology, developers can assess whether virtual RADAR sensors are ready for environmental perception during scenario-based automated driving system (ADS) testing.