Learning-based 4D Millimeter Wave Automotive Radar Sensor Model Simulation for Autonomous Driving Scenarios

Abstract

With the development of autonomous driving technology, scenario simulation is considered an important part of autonomous driving development and testing. Due to its ability to work in complex weather, millimeter wave radar is one of the critical sensors in autonomous driving. Therefore, the simulation of millimeter wave radar sensors is also an essential part of the autonomous driving scenario simulation. Unlike conventional millimeter wave radar, 4D automotive millimeter-wave radar improves the angular resolution by cascading antennas and is able to measure the height of the target. In this paper, we present a method for modeling 4D millimeter-wave radar. First, the spatial distribution models of different types of targets are modeled by different Gaussian mixture models. Then, the spatial distribution model of the point cloud is then used to generate the spatial distribution of scattered points for the radar target model. Next, the millimeter wave radar waveform generation model, target echo model, and intermediate frequency (IF) signal generation model is established to simulate the radar radio frequency. Furthermore, the radar signal processing model is established to process the IF signal and obtain the point cloud of the target. Finally, the simulation experiments show the point cloud effects generated by the millimeter wave radar simulation model under different traffic scenarios.