Compilation of driving simulation languages via retargetable and semantics-based translation

The development of automated driving functions requires the usage of driving simulation as integral part of the development process. Moreover, safety standards such as SOTIF (ISO 21448) and legal regulations give driving simulation a key role for the safety validation of automated driving functions by OEMs and Tier-1s, as well as independent or governmental institutions. The new family of ASAM standards including ASAM OpenDRIVE comes into use and evolves further in frequent version releases, to reflect the dynamics of the field. All this gives rise to the need for tools to translate between different driving simulator languages and versions. The type approval of automated vehicles will strongly rely on simulation, especially to validate the behavior of underlying machine learning algorithms. Up to now, it is not possible to formally verify these algorithms and minor changes in their input might lead to completely different results. From systems engineering perspective it therefore becomes even more important to have a solid foundation without further uncertainties, e.g. regarding the correctness of translations between driving simulation languages. Altogether this raises to two major challenges in the context of translation tools for driving simulation languages that are hitherto not well addressed. These are (1) adaptability to new language versions, and (2) correctness of translations. We present a solution approach for these issues that is inspired by methods established in the domain of compiler design. As cornerstones of this approach we developed at Trier University of Applied Sciences the first translation semantics of driving simulation languages and a suiting research prototype of a retargetable translator supporting ASAM OpenDRIVE and other description formats.