Aim and Scope
Today, automated vehicles already rely heavily on data-driven methods and the relevance of these methods will continue to grow. This leads to an ever-increasing demand for appropriate data. In terms of training data for environment perception algorithms, there already exist many public datasets, such as e.g. KITTI, Cityscapes, Waymo Open, Lyft Level 5 Dataset, etc. The industry has already established processes to generate such data in large amounts. In contrast, the data demand is still not sufficiently met for many other automated driving tasks such as e.g. Safety Validation, Prediction Models, Driver Models, Simulation Agents and Cooperative Behavior Planning, where there is a need for naturalistic road user trajectory dataset.
Therefore, ika and fka have released the highway drone dataset (highD) and intersection drone dataset (inD), which are large scale naturalistic road user trajectory datasets created using camera-equipped drones. Although these datasets are already fostering research a lot, it is expected that many more road user trajectory datasets from other locations and traffic scenarios are needed in future. For this purpose, a panel discussion will be held in the workshop to investigate the current and further need for road user trajectory.
Furthermore, possible applications of road user trajectory datasets shall be presented by invited speakers from industry and the authors of the accepted workshop papers. Those applications of the datasets do not only allow insight into the current state-of-the-art but also the identification of current limitations.