Drones as a Tool for the Development and Safety Validation of Highly Automated Driving

Authors

Robert Krajewski, Julian Bock, Univ.-Prof. Dr.-Ing. Lutz Eckstein, Institute for Automotive Engineering (ika), RWTH Aachen University

Summary

While partially automated vehicles become commercially available on the market, research and development in the automotive industry is focusing on highly automated driving. The current experiences show that along with the complexity of the driving function, the effort required to develop and validate these systems increases enormously. Thus, new methods are required to facilitate an efficient development and safety validation. We show that modern drones combined with state-of-the-art algorithms from the field of machine learning are an important tool to solve some of the occurring challenges in a better or a new way. On the one hand, drones can be used
to efficiently record large-scale datasets of uninfluenced road user trajectories. These are not only needed for the development, parameterization and evaluation of e.g. motion models, but also for a scenario-based safety validation of the driving function. For this purpose, we present the highD and the inD datasets, which contain road user trajectories on motorways and urban intersections. On the other hand, the advantageous bird's-eye perspective allows the acquisition of reference data for the automated vehicle's environmental perception of individual or fused sensors by actively tracking a test vehicle by the drone. For this purpose, we present a system with which we can obtain reference data with the accuracy of an RTK-GPS.

Members Login