Phenomenological modelling of radar sensor and realization of noise effects

The rapid advancement in the autonomous vehicle industry has underscored the critical role of sensors in identifying and tracking traffic participants. Among these sensors, radar plays a pivotal role due to its ability to function reliably in various weather and lighting conditions. This paper presents a phenomenological radar sensor model designed to simulate the behavior of real radar systems under diverse scenarios, including noisy environments and accidental situations. As the complexity of autonomous systems increases, relying solely on on-road and bench testing becomes insufficient for meeting stringent safety and performance standards. These traditional testing methods may not encompass the wide range of potential scenarios that autonomous vehicles might encounter. Consequently, virtual environment modeling has emerged as a crucial tool for validating driving functions, assistance systems, and the strategic placement of multiple sensors. In contrast to high-fidelity radar models, which require detailed environmental inputs and extensive computational resources, phenomenological sensor models offer a balance between performance and accuracy. These medium-fidelity models reduce the need for exhaustive environmental details and complex computations, such as Fourier tracing, thereby enhancing simulation efficiency. Despite the trade-off in accuracy, phenomenological models provide a more realistic and reliable alternative to basic ideal sensor models. This paper details the process of developing the radar model, which incorporates fundamental radar theory and geometric principles. The model simulates radar detection images by integrating synthetic ground truth data with an open simulation interface standard. It also accounts for various degrading effects and environmental noise that impact radar performance. The validity of the proposed radar model is assessed through extensive comparisons with actual radar recordings. Various output parameters of the detection data are analyzed to evaluate the model's realism and reliability. The results demonstrate that the phenomenological radar sensor model closely mirrors the performance of real radar systems, thereby offering a robust tool for advancing the development and testing of autonomous vehicle technologies.