In technical development, the product, including its robustness, is determined through the requirements and tests. In many cases, the robustness is validated by several sample tests. This paper focuses on the development process in the field of systems for the passive safety of vehicle occupants. The state-of-the-art in the development and assessment of such systems is based on both physical and virtual testing. The physical tests are mainly used for validation and final approval purposes. The prominent role of virtual methods is played with a digital twin of the physical configuration. For the assessment of the interaction of the vehicle with its occupants, few anthropometric configurations are usually tested. Assuming a Gaussian distribution of body shapes in a population, the 5th, 50th, and 95th percentiles crash test dummies are widely applied. Nevertheless, the human body does not necessarily vary in a whole, but in all its measures. The reason for using only those few percentiles lies in the limitation of the resources, especially for physical testing, and the advantage of synchronizing physical and virtual testing. Due to the limitations, the potential of virtual methods cannot be used fully. In this state-of-the-art overview paper, the approaches to speed up and broaden the testing during the system's development to increase the solutions' robustness are presented. The state-of-the-art occupant safety development methods are summarized, including the human representation, tested scenarios, as well as evaluated criteria. The second part of this paper reviews available databases for human body measurements. The transition between the anthropometric data to virtual models plays an essential role in the virtual assessment. This part also includes metamodeling approaches to reduce the parameter space and methods for the automated creation of virtual models. Recent studies using multibody systems or finite element analysis are reviewed. Finally, relevant studies on accelerating the virtual process are presented together with approaches from machine learning, which has rapidly developed in recent years. The comparison of the approaches in the passive safety domain is presented. The supervised learning algorithms are taken as a basis for estimating simulation results for different anthropometric configurations. On the one hand, such algorithms can provide a significant reduction of the development time. However, on the other hand, they need a vast amount of data, while generating simulation results can be a time-consuming process. Therefore, also approaches to overcome this obstacle are also mentioned.

Mr. Franz Plaschkies, Technische Hochschule Ingolstadt / CARISSMA, GERMANY Prof. Dr. Ondřej Vaculín, Technische Hochschule Ingolstadt, GERMANY Prof. Dr.-Ing. Axel Schumacher, Bergische Universität Wuppertal, GERMANY