Mixed-Reality Testing of Multi-Vehicle Coordination in an Automated Valet Parking Environment

The development of highly automated driving functions requires rigorous testing to demonstrate the safety and functionality of the automated vehicle. One open question is how to perform such tests to sufficiently prove the vehicle's capabilities. Designing a proper testing platform is particularly important for multi-vehicle scenarios, because test setups with actual real vehicles are not scalable. This paper proposes a mixed-reality testing framework which seamlessly combines virtual and real testing. The authors conducted a mixed-reality experiment of an automated valet parking (AVP) scenario, where virtual vehicles interact with a real vehicle-in-the-loop system in a simulated world. The experiment was developed to evaluate the algorithmic design of a distributed control scheme for an AVP system. Comparing the mixed-reality results to those of a pure virtual simulation allowed the authors to demonstrate the robustness of the algorithmic design against certain disturbances, while also revealing which parts of the system were sufficiently or inadequately modeled during the development process.