Advantages of Using Runtime Procedural Generation of Virtual Environments Based on Real World Data for Conducting Empirical Automotive Research

Driving simulations have become important research tools for improving the in-car user experience. Simulations are especially indispensable when conducting empirical studies in a human-centered development process, where iterative testing in real-world environments is very time-consuming, expensive and potentially dangerous. Constantly advancing possibilities offered by new generations of driving simulations increase the demands and expectations of simulation systems. While the implementation of many components such as simulation hardware or driving physics is very mature by now, the realization of virtual driving environments and scenarios is still a challenging endeavor. A partially automated, algorithmic generation of virtual environments based on real world data provides a possible solution to simplify set-up and execution of research scenarios. The data formats used for this purpose are digital elevation models and map data that can be iteratively processed, improved and expanded for the use in a concrete study. This approach usually requires additional manual steps after the procedural generation is completed. This paper outlines an automated, algorithmic generation of virtual environments based on real-world data, realized at runtime and without subsequent manual post-processing. Significant benefits in automotive research enabled by this approach such as rapid prototyping of study scenarios, support for large-scale virtual environments and the analysis of driving behavior in familiar environments are described.