Validation of safety-critical driver assistance systems is usually performed on test tracks in scenario-based test procedures like the Euro NCAP safety assessment program. With higher automation, the test effort increases exponentially due to the system complexity, which requires new and more efficient test methods that also consider the complexity of every-day road traffic situations. So far, mixed reality environments have been proposed to provide an efficient way of virtual testing and also allow for more realistic interactions between vulnerable road users and automated vehicles. In this work, we investigate the use of different pedestrian representations with varying levels of fidelity (articulated pedestrian target, animated pedestrian, full-body motion capture) regarding their impact on the performance of an exemplary AEB driving function. The results show that the detection of the underlying algorithm and the function's response differ only slightly for the three pedestrian representations. We conclude that all the proposed methods of representing human movement are applicable for testing automated driving functions in mixed reality and can be deployed, depending on the specific test scenario and available resources.
Comparing Different Pedestrian Representations for Testing Automated Driving Functions in Mixed Reality Environments
2022-11-14
2165825 byte
Conference paper
Electronic Resource
English
Pedestrian Environment Model for Automated Driving
IEEE | 2023
|British Library Conference Proceedings | 2017
|