Objective: The current trend in the automotive industry mainly focuses on the autonomous vehicles and driver assistance systems which can provide comfort and safety to the passengers. The scenario-based testing approach is the state of the art to reduce the testing effort by validating such complex functions. However, the representativeness of the test scenarios used is decisive for reliable effectiveness assessment of the system. Reconstructed traffic accidents must be converted into a suitable scenario format to enable virtual testing. This paper deals with the conversion of the GIDAS PCM to other scenario formats and subsequent effectiveness assessment of a generic AEB system. Methodology: A retrospective effectiveness assessment approach is applied. In the first step a sample of 249 real accident scenarios from the GIDAS PCM database are converted into the IPG CarMaker environment. The quality of the conversion is evaluated with several quality criteria. The pre-crash phase is evaluated over the entire time series, taking into account such parameters as velocity, acceleration, yaw angle and position of the both vehicles. A novel X30 criteria is introduced as a robust value to evaluate the specific collision constellation. Real accident scenarios are simulated with a generic Autonomous Emergency Braking (AEB) system and the performance is quantified on the basis of accident avoidance and mitigation. To evaluate the injury severity reduction of not avoided but mitigated cases, MAIS2+ injury risk functions are created in the area of action of the function under test. Results: From 50 % to 80 % of the accidents in the defined area of action of the generic AEB system could be avoided in the simulation. The number of mitigated cases is in the range from 10 % to 35 %. The different safety potential is explained based on the different parameterization of the AEB system. Furthermore, the scenarios simulated with AEB system are used to create a new prospective test case catalog of longitudinal test scenarios. The avoided accidents represent expected near-by collision scenarios. Not avoided but changed accidents represent the new longitudinal accident constellations to be expected. Limitations: The tested AEB system is a generic system, which approximates the behavior of a real system, but cannot identically reproduce it. Furthermore, an ideal quasi-optical sensor model was used to represent a middle range radar sensor. The use of sensor models with phenomenological effects would lead to a more accurate assessment of the tested systems. In the future, the authors will concentrate on the implementation and assessment of further assistance functions, especially focusing on lateral intervention assistance systems. New in the field: Novel is the presented method of the conversion and especially the conversion quality criterias from the original PCM format to the CarMaker road5 and testrun formats. In particular, the introduced X30 criteria to compare the quality of the collision constellation. The injury risk functions are created with a novel approach, by using the k-NN algorithm. Conclusion: The article describes an approach to utilize GIDAS PCM scenarios for driver assistance systems testing, with the software IPG CarMaker. It presents a novel conversion quality criteria and the methodology for effectiveness assessment of AEB, including the evaluation of the injury severity reduction.
Mr. Roman Putter, PhD student & Engineer, Volkswagen AG