The development of highly automated driving (HAD) is in rapid progress. One of the main remaining challenges, however, is to validate highly automated driving in interaction with humans in the various roles in which they will participate in general traffic situations. This includes not only the role from the potential safety driver's, passenger's or pedestrian's perspective, but also - related to the step from SAE autonomy level 4 to level 5 - the consideration of the safety driver located in a remote control room. In this work, we present a classification of the different human roles in automated driving and outline their requirements. Afterwards, we address the conceptualization of a training room and present our realization of the overall system as well as the underlying software and hardware architecture and finally demonstrate the applicability in different use cases. The suitability of the described training center was exemplary evaluated on one use case with questionnaires. The presented system will further serve for data generation, for data analysis and as a training environment for improving interactions between AI-based highly automated driving systems and humans.