Modern vehicles contain nearly 100 embedded control units to realize various comfort and safety functions. These vehicle functions consist of a sensor, a data processing, and an actor layer to react intelligently to stimuli from their context. Recently, these sensors do not only perceive data from the own vehicle but more often also data from the vehicle's surroundings to understand the current traffic situation. Thus, traditional development and testing processes need to be rethought to ensure the required quality especially for safety-critical systems like a collision prevention system. On the example of 1:10 scale model cars, we outline our model-based and composable simulation approach that enabled the virtualized development of autonomous driving capabilities for model cars to compete in an international competition.