levelXdata – Highly Accurate Real World Scenario Data for Development and Validation of Automated Driving

Authors

Julian Bock, M.Sc., Tobias Moers, M.Sc., Dr.-Ing. Adrian Zlocki, fka GmbH, Aachen, Germany;
Robert Krajewski, M.Sc., Lennart Vater, M.Sc., Univ.-Prof. Dr.-Ing. Lutz Eckstein, Institute for Automotive Engineering (ika), RWTH Aachen University, Aachen, Germany

Summary

It is currently apparent that there is an immensely greater need for data for highly automated driving (HAD) than for advanced driver assistance systems (ADAS). This can be seen for example in the development of HAD systems, where huge data sets are needed as training data, and also in the safety validation of the systems, where all scenarios relevant to the driving function must be identified and described. However, not only the required quantity has increased, but also the quality requirements for the data to be collected. This results in the need for new, supplementary measuring methods for the collection of highly accurate data in large quantities. Especially in the last two years through the publication of the intersection drone dataset (inD) and the highway drone dataset (highD), it has been shown that the method of measuring traffic participant trajectories with the help of drones can generate data sets, which are highly relevant for automated driving. Within this paper, the current status of drone data sets and their applications is analyzed.

Members Login