Performance Evaluation of Active Safety Sensor Models for Virtual Verification Tool Chain

Abstract

Today’s active safety systems within automotive industry are developing and becoming more complex. Along with this development, much more testing and validation of these systems are required. Due to their high time and cost consumption, Computer Aided Engineering (CAE) simulations are playing a more vital role in the validation stage. Since the performance of active safety systems strongly depends on the performance of sensors, it is important to validate the sensor models for the development of related active safety functions. Traditionally, CAE simulation uses ideal sensor models that reflect the ground truth of measured data. However, active safety systems need sensor models to capture the same inaccuracies as the physical sensors to achieve better agreement with tests in real traffic. As a result, sensor error models that capture the inaccuracies are developed. The combination of ideal sensor model and sensor error model is used to imitate the output of an integrated radar and camera system. In this thesis work, a method for validating sensor error models is presented. The whole validation process is divided into two parts: sensor level validation and function level validation. In sensor level validation, we verify the standalone sensor error model by comparing the simulated sensor error with the measured error using several comparison metrics. In function level validation, the sensor error model is validated under different CAE environment including Model-in-loop (MIL) and Hardware-inloop (HIL) frameworks. Two system responses that can indicate the performance of active safety systems are proposed. Then the experimental results of these responses are compared with measured data as well. In this evaluation level, we also propose several validation metrics to show the discrepancy between simulated and measured data.