Journal of Software Engineering for Autonomous Systems

Modern advanced driver assistance systems (ADAS) and automated driving (AD) require robust verification and validation (V&V) methods due to their increasing complexity. Simulation-based testing is adopted industry-wide as it provides a controlled, replicable and safe environment with the potential to scale. Yet, fidelity concerns are often faced that still require in situ testing in the real world, which is typically considered time-consuming and being dependent on environmental factors which are hardly controllable. This qualitative study contrasts the practitioners’ view on the fidelity of state-of-the-art virtual test environments in the context of V&V for ADAS and AD in comparison with recent findings from scientific literature. We conducted semi-structured, in-depth interviews with 11 domain experts from the automotive industry to systematically explore their perception of how simulation technologies can reduce the dependency on real-world testing. A thematic analysis was conducted to deduce codes and to develop key themes from the interview data for comparison with the scientific literature. To enhance the validity of our findings, additional focus group interviews with seven experts were performed to complement the initial round of interviews. Our findings reveal necessary steps in bridging the fidelity gap, including systematically measuring simulation model quality, deciding the necessary amount of realism, and the importance of collaboration and transparency among stakeholders.