Experiencing data-driven simulation

The development of automated vehicles demands extensive testing before releasing this technology to market. Virtual tests, like simulations, are expected to contribute significantly to lower the overall testing effort. One challenge is to create realistic virtual test cases and to transfer critical situations from test drives to simulation. The proposed methodology of data-driven simulation can reproduce real-world drives within a simulation framework. Driving simulators that emulate the physical car do not focus on large amounts of test kilometers but rather on capturing subjective driver experiences. Thus, simulation and driving simulators contribute to reducing time and costs by replacing physical prototypes. In this paper, we propose to combine data-driven simulations with driving simulators. Typically, objective algorithms evaluate the outcome of virtualized tests of simulations. By expanding the data-driven evaluation process with a test driver within a driving simulator, subjective impressions enable the validation of the objective data-driven simulation evaluation. Additionally, if a real-world test drive is evaluated the same way as the virtualized version within the driving simulator, cross-verification is possible. Hence, driving simulators can be validated, and the development process of automated driving and automotive functions is further enhanced.